THE METAPHYSICS OF LOGIC

Featuring fourteen new essays from an international team of

renowned contributors, this volume explores the key issues, debates,
and questions in the metaphysics of logic. The book is structured in
three parts, looking first at the main positions in the nature of logic,
such as realism, pluralism, relativism, objectivity, nihilism, conceptu-
alism, and conventionalism, then focusing on historical topics such as
the medieval Aristotelian view of logic, the problem of universals, and
Bolzano’s logical realism. The final section tackles specific issues such
as glutty theories, contradiction, the metaphysical conception of
logical truth, and the possible revision of logic. The volume will
provide readers with a rich and wide-ranging survey, a valuable digest
of the many views in this area, and a long overdue investigation of
logic’s relationship to us and the world. It will be of interest to a wide
range of scholars and students of philosophy, logic, and mathematics.

p e n e l o p e r u s h is Honorary Associate with the School of

Philosophy and Online Lecturer for Student Learning at the Univer-
sity of Tasmania. She has published articles in journals including
Logic and Logical Philosophy, Review of Symbolic Logic, South African
Journal of Philosophy, Studia Philosophica Estonica, and Logique et
Analyse. She is also the author of The Paradoxes of Mathematical,
Logical, and Scientific Realism (forthcoming).
THE METAPHYSICS OF LOGIC

edi t ed by
PENELOPE RUSH
University of Tasmania
 University Printing House, Cambridge cb2 8bs, United Kingdom

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781107039643
© Cambridge University Press 2014
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2014
Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library
Library of Congress Cataloging in Publication data
The metaphysics of logic / edited by Penelope Rush, University of Tasmania.
pages cm
Includes bibliographical references and index.
isbn 978-1-107-03964-3 (Hardback)
1. Logic. 2. Metaphysics. I. Rush, Penelope , 1972– editor.
bc50.m44 2014
160–dc23 2014021604
isbn 978-1-107-03964-3 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
 With thanks to Graham Priest for
unstinting encouragement,
and to Annwen and Callum – never give up.
 Contents

List of contributors page ix

Introduction 1
Penelope Rush

part i the main positions 11

1 Logical realism 13
Penelope Rush
2 A defense of logical conventionalism 32
Jody Azzouni
3 Pluralism, relativism, and objectivity 49
Stewart Shapiro
4 Logic, mathematics, and conceptual structuralism 72
Solomon Feferman
5 A Second Philosophy of logic 93
Penelope Maddy
6 Logical nihilism 109
Curtis Franks
7 Wittgenstein and the covert Platonism of mathematical logic 128
Mark Steiner

part ii history and authors 145

8 Logic and its objects: a medieval Aristotelian view 147
Paul Thom

vii
viii Contents

9 The problem of universals and the subject matter of logic 160

Gyula Klima
10 Logics and worlds 178
Ermanno Bencivenga
11 Bolzano’s logical realism 189
Sandra Lapointe

part iii specific issues 209

12 Revising logic 211
Graham Priest
13 Glutty theories and the logic of antinomies 224
Jc Beall, Michael Hughes, and Ross Vandegrift
14 The metaphysical interpretation of logical truth 233
Tuomas E. Tahko

References 249
Index 264
 Contributors

jody azzouni, Professor, Department of Philosophy, Tufts University.

jc beall, Professor of Philosophy and Director of the UCONN Logic
Group, University of Connecticut, and Professorial Fellow at the
Northern Institute of Philosophy at the University of Aberdeen.
ermanno bencivenga, Professor of Philosophy and the Humanities,
University of California, Irvine.
solomon feferman, Professor of Mathematics and Philosophy,
Emeritus, and Patrick Suppes Professor of Humanities and Sciences,
Emeritus, Stanford University.
curtis franks, Associate Professor, Department of Philosophy,
University of Notre Dame.
michael hughes, Department of Philosophy and UCONN Logic
Group, University of Connecticut.
gyula klima, Professor, Department of Philosophy, Fordham University,
New York.
sandra lapointe, Associate Professor, Department of Philosophy,
McMaster University.
penelope maddy, Distinguished Professor, Department of Logic and
Philosophy of Science, University of California, Irvine.
graham priest, Distinguished Professor, Graduate Center, CUNY,
and Boyce Gibson Professor Emeritus, University of Melbourne.
penelope rush, Honorary Associate, School of Philosophy, University
of Tasmania.

ix
x List of contributors
stewart shapiro, Professor, Department of Philosophy, The Ohio
State University.
mark steiner, Professor Emeritus of Philosophy, the Hebrew University
of Jerusalem.
tuomas e. tahko, Finnish Academy Research Fellow, Department of
Philosophy, History, Culture and Art Studies, University of Helsinki.
paul thom, Honorary Visiting Professor, Department of Philosophy,
The University of Sydney.
ross vandegrift, Department of Philosophy and UCONN Logic
Group, University of Connecticut.
 Introduction
Penelope Rush

This book is a collection of new essays around the broad central theme of
the nature of logic, or the question: ‘what is logic?’ It is a book about logic
and philosophy equally. What makes it unusual as a book about logic is
that its central focus is on metaphysical rather than epistemological or
methodological concerns.
By comparison, the question of the metaphysical status of mathematics
and mathematical objects has a long history. The foci of discussions in the
philosophy of mathematics vary greatly but one typical theme is that of
situating the question in the context of wider metaphysical questions:
comparing the metaphysics of mathematical reality with the metaphysics
of physical reality, for example. This theme includes investigations into: on
exactly which particulars the two compare; how (if ) they relate to one
another; and whether and how we can know anything about either of
them. Other typical discussions in the field focus on what mathematical
formalisms mean; what they are about; where and why they apply; and
whether or not there is an independent mathematical realm. A variety of
possible positions regarding all of these sorts of questions (and many more)
are available for consideration in the literature on the philosophy of
mathematics, along with examinations of the specific problems and attrac-
tions of each possibility.
But there is as yet little comparable literature on the metaphysics of
logic. Thus the aim of this book is to address questions about the
metaphysical status of logic and logical objects analogous to those that
have been asked about the metaphysical status of mathematical objects
(or reality). Logic, as a formal endeavour has recently extended far
beyond Frege’s initial vision, describing an apparently ever more com-
plex realm of interconnected formal structures. In this sense, it may
seem that logic is becoming more and more like mathematics. On the
other hand, there are (also apparently ever more) sophisticated logics

1
2 Penelope Rush
describing empirical human structures: everything from natural lan-
guage and reason, to knowledge and belief.
That there are metaphysical problems (and what they might be) for the
former structures analogous to those in the philosophy of mathematics is
relatively easily grasped. But there are also a multitude of metaphysical
questions we can ask regarding the status of logics of natural language and
thought. And, at the intersection of these (where one and the same logical
structure is apparently both formal and mathematical as well as applicable
to natural language and human reason), the number and complexity of
metaphysical problems expands far beyond the thus far relatively small set
of issues already broached in the philosophy of logic.
As just one example of the sorts of problems deserving a great deal more
attention, consider the relationship between mathematics and logic.
Questions we might ask here include: whether mathematics and logic
describe the same or similar in-kind realities and relatedly, whether there
is a line one can definitively draw between where mathematics stops and
logic starts. Then we could also ask exactly what sort of relationship this is:
is it one of application (of the latter to the former) or is it more complex
than this?
Another central problem for the metaphysics of logic is that of pinning
down exactly what it is that logic is supposed to range over. Logic has been
conceived of in a wide variety of ways: e.g. as an abstraction of natural
language; as the laws of thought; and as normative for human reason. But,
what is the ‘thought’ whose structure logic describes; how natural is the
natural language from which logic is abstracted?; and to what extent does
the formal system actually capture the way humans ought to reason?
As touched on above, a key metaphysical issue is how to account for the
apparent ‘double role’ – applying to both formal mathematical and natural
reasoning structures – that (at least the main) formal logical systems play.
This apparent duality lines up along the two central, indeed canonical
applications of logic: to mathematics and to human reason, (and/or human
thought, and/or human language). In many ways, the first application
suggests that logic may be objective – or at least as objective as mathemat-
ics, in the sense that, as Stewart Shapiro puts it (in this volume) we might
say something “is objective if it is part of the fabric of reality”. This in turn
might suggest an apparent human-independence of logic. The second
application, though, might suggest a certain subjectivity or inter-
subjectivity; and so in turn an apparent human-dependence of logic,
insofar as a logic of reason may appear dependent on actual human
thought or concepts in some essential way.
 Introduction 3
Both the apparent objectivity and the apparent subjectivity of logic need
to be accounted for, but there are numerous stances one might take within
this dichotomy, including a conception of objectivity that is nonetheless
human-dependent. In Chapter 4, Solomon Feferman reviews one such
example in his non-realist philosophy of mathematics, wherein “the
objects of mathematics exist only as mental conceptions [and] . . .
the objectivity of mathematics lies in its stability and coherence under
repeated communication”. Others of the various positions one might take
up within this broad-brush conceptual field are admirably explored in both
Stewart Shapiro’s and Graham Priest’s chapters, though from quite differ-
ent stand points: Shapiro explores the nuances and possibilities in concep-
tions of objectivity, relativity, and pluralism for logic, whereas Priest looks
at these issues through the specific lens afforded by the question whether or
not logic can be revised.
There are, then, a variety of possible metaphysical perspectives we can
take on logic that, particularly now, deserve articulation and exploration.
These include nominalism; naturalism; structuralism; conceptual
structuralism; nihilism; realism; and anti-(or non-)realism, as well as
positions attempting to steer a path between the latter two. The following
essays cover all these positions and more, as defended by some of the
foremost thinkers in the field.
The first part of the book covers some of the main philosophical
positions one might adopt when considering the metaphysical nature of
logic. This section covers everything from an extreme realism wherein logic
may be supposed to be completely independent of humanity, to various
accounts and various degrees in which logic is supposed to be in some way
human-dependent (e.g. conceptualism and conventionalism).
In the first chapter I explore the feasibility of the notion that logic is
about a structure or structures existing independently of humans and
human activity. The (typically realist) notion of independence itself
is scrutinised and the chapter gives some reasons to believe that there is
nothing in principle standing in the way of attributing such independence
to logic. So any benefits of such a realism are as much within the reach of
the philosopher of logic as the philosopher of mathematics.
In the second chapter, Jody Azzouni explores whether logic can be
conceived of in accordance with nominalism: a philosophy which might
be taken to represent the extreme opposite of realism. Azzouni argues the
case for logical conventionalism, the view that logical truths are true by
convention. For Azzouni, logic is a tool which we both impose by conven-
tion on our own reasoning practices, and occasionally also to evaluate
4 Penelope Rush
them. But Azzouni shows that although there seems to be a close relation-
ship between conventionality and subjectivity, logic’s being conventional
does not rule out its also applying to the world.
Stewart Shapiro, in the third chapter, argues the case for logical relativ-
ism or pluralism: the view that there is “nothing illegitimate” in structures
invoking logics other than classical logic. Shapiro defends a particular sort
of relativism whereby different mathematical structures “have different
logics”, giving rise to logical pluralism – conceived of as “[the] view that
different accounts of the subject are equally correct, or equally good, or
equally legitimate, or perhaps even (equally) true”.
Shapiro’s chapter looks in some depth at the relationship between
mathematics and logic, identified above as a central problem for our
theme. But in particular, it investigates the extent to which logic can
be thought of as objective, given the foregoing philosophy. He offers a
thorough, precise, and immensely valuable analysis of the central concepts,
and clarifies exactly what is and is not at stake in this particular debate.
In the fourth chapter, Solomon Feferman examines a variety of logical
non-realism called conceptual structuralism. Feferman shares with Shapiro
a focus on the relationship between mathematics and logic, extending the
case for conceptual structuralism in the philosophy of mathematics to logic
via a deliberation on the nature and role of logic in mathematics. He draws
a careful picture of logic as an intermediary between philosophy and
mathematics, and gives a compelling argument for the notion that logic,
as (he argues) does mathematics, deals with truth in a given conception.
According to Feferman’s account, truth in full is applicable only to
definite conceptions. On this picture, when we speak of truth in a
conception, that truth may be partial. Thus classical logic can be concep-
tualised as the “logic of definite concepts and totalities”, but may itself be
justified on the basis of a semi-intuitionist logic “that is sensitive to
distinctions that one might adopt between what is definite and what is
not”. Feferman shows how allowing that “different judgements may be
made as to what are clear/definite concepts”, affords the conceptual
structuralist a straightforward, sensible and clear understanding of the role
and nature of logic.
Penelope Maddy, in the fifth chapter, offers a determinedly second-
philosophical account of the nature of logic, presenting another admirably
clear and sensible account, focusing in this case on the question why logic
is true and its inferences reliable. ‘Second Philosophy’ is a close cousin of
naturalism as well as a form of logical realism and involves persistently
bringing our philosophical theorising back down to earth.
 Introduction 5
In Maddy’s words: “The Second Philosopher’s ‘metaphysics naturalized’
simply pursues ordinary science”. Thus Maddy investigates the question
from this ‘ordinary’ perspective, beginning with a consideration of rudi-
mentary logic, and gradually building up (via idealisations) to classical
logic. On this account, logic turns out to be true and reliable in our actual
(ordinary, middle-sized) world partly because that actual world shares the
formal structure of logic (or at least rudimentary logic). Maddy gives an
extensive account of some of the ways we might come to know of this
structure, presenting recent research in cognitive science that supports
the notion that we are wired to detect just such a structure. She then
offers the (tentative) conclusion that classical logic (as opposed to any non-
classical logic) is best suited to describe the physical world we live in,
despite the fact that classical logic’s idealisations of rudimentary logic are
best described as ‘useful falsifications’.
In the final two chapters of the first part, Curtis Franks questions the
assumption underpinning any metaphysics of logic at all: namely that
there is “a logical subject matter unaffected by shifts in human interest
and knowledge”; and Mark Steiner unpicks Wittgenstein’s idea that “The
rules of logical inference are rules of the language game”.
Steiner points out that for Wittgenstein “There is nothing akin to
‘intuition’, ‘Seeing’ and the like in following or producing a logical
argument. Instead we [only] have regularities induced by linguistic
training”. So, Steiner argues, supposing that logic is grounded by anything
other than the regularities that ground rule following (say by some object-
ive ‘fact’ according to which its rules are determined), is engaging in a kind
of ‘covert Platonism’.
Steiner identifies the key difference (for Wittgenstein) between math-
ematics and logic as the areas their respective rules govern: whereas both
mathematical and logical rules govern linguistic practices, (only) math-
ematical rules also govern non-linguistic practices. Interestingly, while
Steiner argues that the line between mathematics and logic is thus more
substantial than many may think, Franks argues that the line between
maths and logic is illusory, based on a need to differentiate the patterns of
reasoning we have come to associate with logic from other patterns of
reasoning, which itself is grounded on nothing more than a baseless
psychological or metaphysical preconception.
Franks argues that logicians deal not with truth but with the “relation-
ships among phenomena and ideas” – and agrees with Steiner that looking
for any further ‘ontological ground’ is misconceived (note, though, that
Steiner himself does not commit himself to the views he attributes to
6 Penelope Rush
Wittgenstein. Rather he gives what he takes to be the best arguments in
Wittgenstein’s favour). As something of a side note, it is interesting to
compare Sandra Lapointe’s discussion of Bolzano’s notion of definition
(in Part II) to that which Franks presents on behalf of Socrates. Lapointe
argues that, for Bolzano, there is more to a definition than merely fixing its
extension, whereas Franks argues that Socrates was right to prioritise the
fixing of an extension first before enquiring after the nature or essence of a
thing. Steiner’s discussion of the Wittgensteinian distinction between
explanation and description is also relevant here. This debate touches on
another important subtheme running throughout the book: the nature and
role of intentional and extensional motivations of logical systems; and the
related tension (admirably illustrated by Franks’ discussion of the develop-
ment of set theory) between appeals to form/formal considerations and
appeals to our intuitions.
Both Steiner’s Wittgenstein and Franks agree that the image of logic as a
kind of ‘super-physics’ needs to be challenged, even eliminated; but each
takes a different approach to just how this might be achieved, with Franks
arguing for logical nihilism, and Steiner going to pains to show how, for
Wittgenstein, the rules of logic ought to be conceived as akin to those of
grammar and as nothing more than this.
The next part of the book gives an historical overview of past investi-
gations into the nature of logic as well as giving insights into specific
authors of historical import for our particular theme.
In the first chapter of this section Paul Thom discusses the thoughts of
Aristotle and the tradition following him on logic. Thom focuses particu-
larly on what sort of thing, metaphysically speaking, the objects of logic
might be. He traces a gradual shift (in Kilwardby’s work) from a concep-
tion of logic as about only linguistic phenomena, through a conception
wherein logic is also understood as also being about reason, to the
inclusion of ‘the natures of things’ as a possible foundation of logic.
Kilwardby considers a view whereby the principal objects of logic: ‘state-
ables’, are not some thing at all (at least not in themselves), insofar as they
do not belong to any of Aristotle’s categories. Kilwardby opposes this view
on the basis of a sophisticated and complex argument to the effect that
there may be objects of logic that are human dependent but also external
to ourselves, and can be considered both things of and things about nature
itself. These insights are clearly relevant to the modern questions we ask
about the metaphysics of logic and resonate strongly with the themes
explored in the first part. The range of possibilities considered offer a
fascinating and fruitful look into the historical precedents of the questions
 Introduction 7
about logic still open today: e.g. Thom notes that for Aristotle, the types of
things that can belong to the categories are ‘outside the mind or soul’,
and so Kilwardby’s analysis clearly relates to our modern question as to the
possible independence and objectivity of logic. The complexity of that
question is brought to the fore in Kilwardby’s detailed consideration of
the various ‘aspects’ under which stateables can be considered, and
according to which they may be assigned to different categories.
Thom’s chapter goes on to offer a framework for understanding later
thinkers and traditions in logic, some of which (e.g. Bolzano in Lapointe’s
chapter) are also discussed in this part. His concluding section ably
demonstrates that understanding the history of our questions casts useful
light on the modern debate.
Gyula Klima also discusses strategies for dealing with the two way pull
on logic – from its apparent abstraction from human reason and from its
apparent groundedness in the physical world. Klima focuses on the scho-
lastics, comparing the semantic strategies of realists and nominalists
around Ockham’s time. One of these was to characterise logic as the study
of ‘second intentions’ – concepts of concepts. Klima points out that when
logic is conceived of in this way, the core-ontology of real mind-
independent entities could in principle have been exactly the same for
“realists” as for Ockhamist “nominalists”; therefore, what makes the
difference between them is not so much their ontologies as their different
conceptions of concepts, grounding their different semantics.
Klima argues that extreme degrees of ontological and semantic diversity
and uniformity mark out either end of a “range of possible positions
concerning the relationship between semantics and metaphysics, [from]
extreme realism to thoroughgoing nominalism” and points out how the
conceptualisation of the sorts of things semantic values might be varies
according to where a given position sits within this framework. His chapter
illuminates the metaphysical requirements of different historical
approaches to semantics and the way in which the various possible meta-
physical commitments we make come about via competing intuitions
regarding diversity: whether we locate diversity in the way things are or
in the way we speak of or conceptualise them.
In the next chapter, Ermanno Bencivenga picks up a thought Thom
touches on in his closing paragraph – namely that our modern conception
of logic appears to have lost touch with the relevant ways in which actual
human reason can go wrong other than by not being valid. Offering a
Kantian view, Bencivenga suggests we adjust our conception of logic to
that of almost any structure we impose on language and experience, just so
8 Penelope Rush
long as it is a holistic endeavour to uncover how our language acquires
meaning. In this way almost all of philosophy is logic, but not all of what
we commonly call logic makes the grade. For Bencivenga, logic should
focus on meaning: on the way language constructs our world. From this
perspective, the relationship of logic to reason is just one of many connec-
tions between the world we create and the internal structure of any given
logic. For example, while appeals to reason may motivate logic’s claims, so
too do appeals to ethos and pathos.
Sandra Lapointe looks at the sorts of motivations and reasons we might
have for adopting a realist philosophy of logic, pointing out that these
reasons may not themselves be logical and developing a framework within
which different instances of logical realism can be compared. Lapointe
examines Bolzano’s philosophy in particular and shows how his realism
may best be thought of as instrumental rather than inherent: adopted in
order to make sense of certain aspects of logic rather than as a result of any
deep metaphysical conviction.
Lapointe’s chapter shows how Bolzano’s works cast light on a wide array
of issues falling under our theme, from his evocative analogy between the
truths of logic and the spaces of geometry to his critique of Aristotle’s
criteria for validity. Lapointe’s discussion of the latter is worth drawing
attention to as it deals with the topic mentioned earlier – of the tension
between external and intensional; and formal and non-formal motivations
for logical systems. Lapointe compares the results of Bolzano’s motivations
with those of Aristotle for the definition of logical consequence and in so
doing, identifies some central considerations to help further our under-
standing of this topic.
The final part of the book deals with the specific issues of the possible
revision of logic, the presence of contradiction, and the metaphysical
conception of logical truth.
Graham Priest’s chapter deals with the question of the revisability of
logic and in so doing also offers a useful overview of much of what is
discussed in earlier sections and indeed throughout this book. Priest
outlines three senses of the term ‘logic’ and asks of each whether it can
be revised, revised rationally, and (if so) how.
In some ways, Priest’s paper dovetails with Shapiro’s discussion of the
possible criteria used to judge the acceptability of a theory, and draws a
conclusion similar to that of Shapiro’s ‘liberal Hilbertian’: i.e. “[that]
There is no metaphysical, formal, or mathematical hoop that a proposed
theory must jump through. There are only pragmatic criteria of interest
and usefulness” – which, for Priest, are judged against the requirements of
 Introduction 9
its application(s) and by “the standard criteria of rational theory choice”.
And like Shapiro’s, Priest’s chapter is an immensely valuable overview of
the key concepts informing any metaphysics of logic.
In the next chapter, Jc Beall, Michael Hughes, and Ross Vandegrift look
at different repercussions of different attitudes toward “glutty predicates” –
predicates which “in virtue of their meaning or the properties they
express . . . [are] both true and false”. Their chapter shows how our various
theories and attitudes about such predicates may motivate different formal
systems. The formal systems in question here are Priest’s well-known
LP and the lesser-known LA advanced by Asenjo and Tamburino. The
upshot of the discussion is that the latter will suit someone metaphysically
“commited to all predicates being essentially classical or glutty” and
the former someone for whom “all predicates [are] potentially classical
or glutty”.
Thus, Beall et al. draw out some interesting consequences of the
relationships between our intuitions and theories regarding the metaphys-
ical, the material, and the formal aspects of logic. They highlight both the
potential ramifications of the role we afford our metaphysical commit-
ments and the ramifications of the particular type of commitments they
might be. So while Beall et al. look in particular at a variety of metaphysical
theories about contradiction, and the impact of these on two formal
systems, their discussion also gives some general pointers to the way in
which our metaphysical beliefs impact on other central factors in logic:
crucially including the creation of the formal systems themselves and the
evaluation of their differences.
Tuomas Tahko finishes the book by examining a specific realist meta-
physical perspective and suggesting it as another approach we might take
to understand logic, especially to interpret logical truth. His case study
offers an interpretation of paraconsistency which contrasts nicely with that
offered in the penultimate chapter. Tahko’s approach is to judge logical
laws according to whether or not they count as genuine ways the actual
world is or could be. From this perspective, he argues, exceptions to the
law of non-contradiction now appear more as descriptions of features of
our language than of reality. Thus he argues that the realist intuition
grounding logic in how the world is (or could be) gives us good reason
to preserve the LNC. Tahko’s metaphysical interpretation of logical truth
also offers an interesting perspective on logical pluralism. From Tahko’s
metaphysical perspective, pluralism may be understood as about subsets of
possible worlds representing genuine possible configurations of the actual
world. Tahko’s chapter is a meticulous investigation into the links, both
10 Penelope Rush
already in place and that (from this perspective) ought to be, between an
interesting set of metaphysical intuitions and those laws of logic we take to
be true.
In all, this book ranges over a vast terrain covering much of the ways in
which our beliefs about the role and nature of logic and of the structures it
describes both impact and depend on a wide array of metaphysical pos-
itions. The work touches on and freshly illuminates almost every corner of
the modern debate about logic; from pluralism and paraconsistency to
reason and realism.
 part i
The Main Positions
 chapter 1

Logical realism
Penelope Rush

1. The problem
Logic might chart the rules of the world itself; the rules of rational human
thought; or both. The first of these possible roles suggests strong similarities
between logic and mathematics: in accordance with this possibility, both logic
and mathematics might be understood as applicable to a world (either the
physical world or an abstract world) independent of our human thought processes.
Such a conception is often associated with mathematical and logical realism.
This realist conception of logic raises many questions, among which
I want to pinpoint only one: how logic can at once be independent of
human cognition in the way that mathematics might be; and relevant to
that cognition. The relevance of logic to cognition – or, at the very least,
the human ability to think logically – seems indubitable. So any under-
standing of the metaphysical nature of logic will need also to allow for a
clear relationship between logic and thought.1
The broad aim of this chapter is to show that we can take logical
structures to be akin to independent, real, mathematical structures; and
that doing so does not rule out their relevance and accessibility to human
cognition, even to the possibility of cognition itself.
Suppose that logical realism involves the belief that logical facts are
independent of anything human:2 that the facts would have been as they

1
Two things: note I do not claim we can or ought to show that logic underpins, describes, or arises
from cognition. In fact I think the relationship between thought and logic is almost exactly analogous
to that between thought and mathematics (see Rush (2012)), and I disagree with the idea that there is
any especially significant connection between logic and thought beyond this. Two: while this chapter
deals with the notion of ‘independence’ per se, it investigates this from the perspective of applying
that notion especially to logic. That is, my main aim here is to indicate one way in which the realist
conception of an independent logical realm might be considered a viable philosophical position but
one primary way I hope to do this is by showing how attributing independence to logic need be no
more problematic than attributing independence to anything else (e.g. by arguing that the realist
problem applies across any ‘type’ of reality which is supposed to be independent).
2
See Lapointe’s characterisation as IND in this volume.

13
14 Penelope Rush
are regardless of whether or not humans comprehended them, or even had
existed at all. A sturdy sort of objectivity seems guaranteed by this stance.
Janet Folina captures this neatly:
[If logical facts exist independently of the knowers of logic], there is a clear
difference, or gap, between what the facts are and what we take them to be.
(Folina 1994: 204)3

This sturdy objectivity is just one reason we might find logical realism
appealing.4
There is, though, a well-known objection to the idea that we can
coherently posit the independence of facts (including logical facts) from
their human knowers (and human knowledge).
Wilfrid Sellars formulated a version of this objection in 1956. Sellars
argued that in order to preserve both the idea that there is something
independent of ourselves and epistemological processes, and the idea that
we can access this something (e.g. know truths about it), we seem to have
either to undermine the independent status of that thing (by attributing to
it apparently human-dependent features) or to render utterly mysterious
the way in which any knowledge-conferring relationship might arise from
that access.
Sellars’ idea is that we cannot suppose that we encounter reality as it is
independently of us, unless we suppose something like a moment of
unmediated access. But, there can be no relevant relationship between
independent reality and us (e.g. we can make no justificatory or founda-
tional use of such a moment) unless that unmediated encounter can be
taken up within our own knowledge.
The obvious move is simply to say that this initial encounter is available
to knowledge. But this move undermines itself by casting what was
independent as part of what is known: i.e. it attributes an already
in-principle knowability to a supposed fully independent reality (for more
on Sellars’ argument, see Fumerton (2010), and Sellars (1962)).
The broadly applicable Sellarisan objection bears comparison to Bena-
cerraf ’s (1973) objection to mathematical realism, which extends, at least to
a degree, to logical realism.5 Benacerraf argued that even our best theory of

3
Folina was talking about mathematical realism, but the sort of logical realism I want to examine here
is directly analogous to mathematical realism in this respect.
4
Lapointe (this volume) explores a variety of reasons that may play a role in holding some version of
logical realism, so I won’t go into these in depth here.
5
For more on the possible entities a logical realist might posit (e.g. meanings/propositions), see
Lapointe (this volume). Regardless of which entities are selected and where these are situated on the
 Logical realism 15
knowledge could not account for knowledge of mathematical reality just so
long as that reality was conceived of in the usual mathematical realist way:
as abstract, acausal, and atemporal. Part of the problem, as Benacerraf saw
it, was that the stuff being posited as independently real is not sufficiently
like any stuff that we can know, and if it were, it would not be the sort of
thing intended by the mathematical realist in the first place.
Sellars’ objection can be understood as a generalisation of Benacerraf ’s:
common to both is the idea that the fully independent reality posited by
the realist is not the type of thing we can know, or if it is, then it is not the
type of thing the realist says it is.
Thus, even were the mathematical or logical realist to adjust his con-
ception of mathematical or logical reality by ruling out one or all of its
abstractness, atemporality, or acausality, the problem induced by its com-
plete independence of humans and human consciousness would remain.
Recall, the realist idea of independence I am interested in here is one
which posits an in-principle or always possible separation between what
independently is and what we as humans grasp. The basic idea is that were
there no humans to experience or be conscious of it, logic would still be as
it is. So it seems that being the type of thing which is experienced or
known can be no part of what it (essentially) is.6
The problem can be expressed this way: how can independent reality be
part of human consciousness and experience if our human consciousness
and experience of it can be no part of independent reality? A putative
solution, then, might show how independent reality could play a role in
human consciousness, but such a solution would need also to affirm the
necessary condition that being the object of our consciousness is no
(essential) part of independent reality itself.
This notion of independence, then, is not only the most problematic
feature of any logical realism, it may be outright contradictory:
A realist . . . is basically someone who claims to think that which is where
there is no thought. . . . he speaks of thinking a world in itself and
independent of thought. But in saying this, does he not precisely speak of
a world to which thought is given, and thus of a world dependent on our
relation-to-the-world? (Meillassoux 2011: 1)

abstract–physical scale of possible entities, just so long as the realist also posits IND (Lapointe, this
volume), they’ll encounter some version of Benacerraf ’s or Sellars’ problem.
6
For more on the nuances of ‘independence’ available to the realist, see Jenkins (2005) – I take
essential independence to follow from modal independence, and I take modal independence as
characteristic of the sort of realism I want to explore.
16 Penelope Rush
Husserl characterised the realist problem of independence (which he also
called ‘transcendence’) in various ways, one of which is as follows:
[the problem is] how cognition can reach that which is transcendent . . .
[i.e.] the correlation between cognition as mental process, its referent and
what objectively is . . . [is] the source of the deepest and most difficult
problems. Taken collectively, they are the problem of the possibility of
cognition. (Husserl 1964: 10–15)
Each of the above characterisations of the realist’s situation turns on the
central theme of how we can sensibly (and relevantly) conceptualise
the role that a reality independent of human consciousness could play
in the realm of that consciousness.
Husserl’s characterisation of the problem already gives a clue as to his
overall approach: rather than view the problem as bridging a gap of the sort
Folina describes, Husserl suggests we view it as “the possibility of
cognition”.

2. The potential of phenomenology

I hope to show how Husserl’s approach potentially enables us to take
independent reality in both of the ways sitting either side of the gap: i.e.
both as what is and as what is not the end point of a reasonable epistemol-
ogy. That is, I hope to use his approach to see how we might accommodate
the idea that what is cognised, and what must (on a realist account) remain
irreducibly external (or, in principle, separable) to what is cognised – can
be one and the same thing, or (perhaps) more accurately, a dual thing.7
At first glance, this might seem simply to concede the contradiction
Meillassoux graphically outlines. I want to take a second glance – illustrating
how such a concession need be neither simple nor impotent but rather
offer a way to conceptualise the elements underpinning the realist notion
of independent reality and so begin, if not to resolve, then to make some
sense of its intractability. That is, there are ways in which the Husserlian
perspective can motivate us to find reasons and avenues by which we might
begin to accommodate the independent reality the realist posits, even as
potentially contradictory – rather than to take its inherent instability as
reason enough to brush it off as impossible and therefore irrelevant. These
ways all intersect at the possibility opened in the phenomenological
7
As will become clear, I have a very particular notion of duality in mind here – i.e. a (contradictory)
duality of object: ‘one that is also two’ – rather than a duality of an object’s role, or aspect, or
components, etc: ‘one that has two aspects/dimensions/components, etc’.
 Logical realism 17
perspective (admittedly most probably neither envisaged nor anywhere
claimed by Husserl himself ); namely that the realist predicament is itself
an essential ingredient for the possibility of cognition.8
All of the above ways of rendering the realist problem of independence
(barring his own) – i.e. as an intractable and apparently unbridgeable
dichotomy between reality and our knowledge of it – Husserl characterised
as a product of the ‘natural’, ‘scientific’ attitude, which he saw as pervasive
all of philosophy (again, barring his own, e.g. Husserl 1964: 18–19).
By contrast, phenomenology offers a picture of entangled cognition
wherein independent reality is inextricable from cognition itself. This sort
of picture takes the first step toward accommodating both sides of the
divide insofar as it introduces the idea that our internal perspective itself
irreducibly incorporates the possibility, even the necessity, of there being
something outside that perspective.
To be clear, I reiterate that this is my own interpretation of Husserl and
my own exploration of the possibilities his work suggests to me. I do not
attribute these possibilities to Husserl. As I understand him, for Husserl,
experience is always experience of – and so cannot begin to be defined
without allowing (at least) a place or a role for something external toward
which it is directed at the outset. For me, the promising bit is this: that this
something is both somehow outside or external to (‘constituting’) experi-
ence and within it (‘being constituted’) at the same time.
It is by examining and enlarging on this promising bit that I hope to
explore one way in which phenomenology (potentially) offers a role for the
realist predicament itself as the (contradictory) structure of our relationship
to independent reality. I hope to sketch how accepting the predicament in
this way might enable us to make sense of reality, cognition, and experi-
ence within a realist framework – to see the realist’s ‘predicament’ as a
complex and interesting structure that these elements share, as opposed to
an impossible riddle or a problem in need of a solution.
In what follows, I’ll briefly unpack just a couple of aspects of Husserl’s
account in order to show how we might use them to begin to open and
explore this possibility, specifically regarding the idea of a realistically
imagined independent logical structure.9

8
Caveat: I’d like to argue that the predicament can play this role just so far as the basic idea of an
independent reality existing at all can. It is the latter that I see the framework in Husserl’s ideas as
able to directly establish.
9
Or, again, to illustrate how conceiving logic as an independent objective structure akin to
mathematics need not be considered an especially problematic instance of the general idea of
independent reality itself, once that idea is effectively defended.
18 Penelope Rush

3. Key aspects of phenomenology

3.1 The Platonic nature of logic

Husserl had a very broad concept of logic that embraces our usual modern
idea of logic as well as something he called ‘pure logic’, which we can
loosely characterise as something like ‘the fundamental forms of experi-
ence’. For Husserl, logic as formal systems (and so too ‘modern logic’;
incorporating classical, modal, and all the usual non-classical structures), is
to be accounted for in much the same way as is mathematics: by its
relationship to these fundamental forms. This relationship is roughly that
which holds between practice and theory – pure logic is the purely
theoretical structure (or, perhaps, structures – I don’t think it matters
much here) that accounts for logic as practised.
For Husserl, the fundamental forms of pure logic are an in-eliminable
part of experience: i.e. ‘experience’ encompasses direct apprehension of
these inferential relationships. The apprehended structures are abstract
and platonic; discovered, rather than constructed. Theory, empirical
observation, and experience are in this sense fallible: they may or may
not ‘get it right’ and reveal the actual independent structure of logic.
In Husserl’s words:
As numbers . . . do not arise and pass away with acts of counting, and as,
therefore, the infinite number-series presents an objectively fixed totality of
general objects . . . so the matter also stands with the ideal, pure-logical
units, the concepts, propositions, and truths – in short, the significations
dealt with in logic . . . form an ideally closed totality of general objects to
which being thought and expressed is accidental. (1981: 149)
Thus both logic and mathematics, for Husserl, have a ‘pure’, ‘abstract’,
‘theoretical’, ‘definite’, and ‘axiomatic’ foundation. Further, Husserl
believed that:
one cannot describe the given phenomena like the natural number series or
the species of the tone series if one regards them as objectivities in any other
words than with which Plato described his ideas: as eternal, self-identical,
untemporal, unspatial, unchanging, immutable. (Hartimo 2010a: 115–118,
italics mine)
So, according to the prevailing view, both logic and mathematics as
they are characterised by Husserl, should encounter the realist problem
of independence – neither are the sort of thing we can simply take
as part of human cognition; i.e. not without also accommodating
 Logical realism 19
the idea that what cognition accesses is in principle no part of what
either mathematics or logic independently is.

3.2 Inextricability
As touched on above, one of Husserl’s most suggestive and promising ideas
is that consciousness is not separable from consciousness of an object –
intentionality is built into the structure of consciousness and experience
itself.
The leading idea is consciousness as consciousness of: the very definition
of experience and consciousness as involving already what it is directed
toward, or what it is conscious of. Of course, this idea is also what a great
deal of the controversy in Husserlian scholarship centres on. One reason
for the controversy, I think, is the ambiguity in the prima facie simple idea
of an object (or realm, or reality) as an object of anything (including, for
example, consciousness, intention, act, or perception). Even on the most
subjectivist reading, the notion is ambiguous between the idea of objects in
experience, and as experienceable. This ambiguity interplays in obvious
ways with the tension underpinning the realist’s problem: that between the
object as given to an epistemological human-dependent process, and the
object as independent. In turn (as we’ve seen) this ambiguity itself centres
on a distinction between ‘internal’ (what we take the facts to be), and
‘external’ (what the facts are).
I suggest that the urge to disambiguate Husserl on this point should be
resisted,10 since to disambiguate here would be to miss a large part of the
potential of phenomenology. Indeed, Husserl himself seems at times to
deliberately preserve ambiguity here (though whether he meant to or not is
tangential to the point). For example:
First fundamental statements: the cogito as consciousness of something . . .
each object meant indicates presumptively its system. The essential related-
ness of the ego to a manifold of meant objects thus designates an essential
structure of its entire and possible intentionality. (Husserl 1981: 79–80)

On the one hand it has to do with cognitions as appearances, presentations,

acts of consciousness in which this or that object is presented, is an act of
consciousness, passively or actively. On the other hand, the phenomen-
ology of cognition has to do with these objects as presenting themselves in
this manner. (Husserl 1964: 10–12)

10
Thanks to Curtis Franks for help with the expression of this point.
20 Penelope Rush
In the above quotes, both the ‘presenting objects’ and ‘the manner in
which they present’ give cognition its essential structure. It seems that
Husserl resists resolving the ambiguity in these phrases one way or
the other.
Husserl’s “phenomenology of cognition” is accomplished through a
prior conceptual step called the ‘phenomenological reduction’. This
‘reduction’ is related to Descartes’ method of doubt (e.g. in Husserl
1964: 23. A useful elaboration can be found in Teiszen 2010: 80). Teiszen
argues that for Husserl the crucial thing about the phenomenological
reduction was what remains even after we attempt, in Cartesian fashion,
to doubt everything. Teiszen makes the point that if we take a (certain,
phenomenologically mediated) transcendental perspective, we can uncover
in what remains (after Cartesian doubt) a lot more than an ‘I’ who is
thinking. In particular, we can uncover direct apprehension of “the ideal
objects of logic and mathematics” (Teiszen 2010: 9) whose pure forms
extend far further than what Descartes ended up allowing as directly
knowable, and further than the knowable allowed for in Kant’s
philosophy.
Just as there is with what to make of the ‘consciousness as consciousness
of’ idea, so too there is much controversy surrounding exactly what the
phenomenological reduction is and involves. To say that there is disagree-
ment here among Husserl scholars is something of an understatement.
Indeed: “there seem[s] to be as many phenomenologies as phenomenolo-
gists” (Hintikka 2010: 91).
But the clarification of exactly what Husserl may have meant is not
relevant to my purpose here, which is to see if there are ideas we can draw
from Husserl that might help a realist philosopher of logic.
I pause to note, though, that Teiszen’s interpretation of the reduction as
a “‘suspension’ or ‘bracketing’ of the (natural) world and everything in it”
(Teiszen 2010: 9) is standard; and the ‘ideal objects’ recovered in Teiszen’s
consequent ‘transcendental idealism’ (including their ‘constituted mind-
independence’) are also standard for an established tradition of Husserl
scholarship (adhered to by Føllesdal, among others). But these ‘ideal’
objects are very far from the realist mind-independent realm that I want
to imagine has a place here (to hammer this point home, see Teiszen
2010: 18).
Again, it is the (possibly resolute) ambiguity in Husserl’s account
that allows for my alternate reading of phenomenology. Another case
in point: “the description on essential lines of the nature of
consciousness . . . leads us back to the corresponding description
 Logical realism 21
of the object consciously known” (Husserl 1983: 359). The phrase: “the
object consciously known” is ambiguous. It can be read differently
depending on each term’s specific interpretation and on which terms
are emphasized: e.g. the ‘consciously known’ can be read as ‘the object as
we know it’ (i.e. a strictly constituted – internal – object); or as ‘the
object that is known’. It is the latter interpretation that opens the
possibility of an ‘external element’ in the basic ingredients of the nature
of consciousness.
To reiterate: the interesting thing about Husserl for my purpose is
that in his ideas we can discern a (at least potential) role for an
independent objective other, while nonetheless focusing on experience
and consciousness: my thought is that if we can argue that intending
reality as it appears (i.e. in the case of the realist conception of logic: as
objective and independent) is itself constitutive of cognition and even
of the possibility of cognition itself; then we can see a way in which
objective independent reality is (complete with its attendant predica-
ment) already there, structuring the essential nature of consciousness
and experience.
For me, the phenomenological reduction, or ‘ruling out’ of all that can
be doubted, and the subsequent re-discovery of the world (ultimately)
demonstrates an important way that reality, in all of the ways it seems to
us to be (including being independent of us), in fact cannot be ruled
out. Thus, we can see in the basic elements of the phenomenological
analysis how objective, independent reality enters the picture as objective,
and independent – not only as an object of consciousness, but as consti-
tuting consciousness itself. This is the case even if (or, as Husserl would
have it, especially if ) we try to focus only on ‘pure experience’ or ‘pure
consciousness’.
I’ll mention a couple of other perspectives that gesture in a similar
direction to my own before moving on.
From Levinas we get:
the fact that the in itself of the object can be represented and, in knowledge,
seized, that is, in the end become subjective, would strictly speaking be
problematic . . . This problem is resolved before hand with the idea of
the intentionality of consciousness, since the presence of the subject
to transcendent things is the very definition of consciousness. (1998: 114,
italics mine) [and]
the world is not only constituted but also constituting. The subject is no
longer pure subject; the object no longer pure object. The phenomenon
is at once revealed and what reveals, being and access to being. (1998: 118)
22 Penelope Rush
Once we get our heads around the idea that the presence of the subject to
transcendent things defines consciousness,11 it is not a huge leap to see how
this initial subjective/transcendent relationship (even if it’s just one of
mutual ‘presence’) can incorporate the entire problematic outlined above:
i.e. that the Sellars–Meillassoux contradiction is ‘built in’ just so far as it
describes that relationship. Recall that Husserl equates that problematic
with the problem of the possibility of cognition (p. 16 above): it should
now be apparent how his equation can be understood as a means by
which to understand (rather than resolve or dissolve) the ‘natural’,
‘scientific’ perspective, complete with its consequent dilemma. That is,
Husserl’s point:
‘The problem of the possibility of cognition is the traditional realist dilemma’
need not be interpreted thus: ‘the problem of the possibility of cognition
supplants the traditional dilemma’. Rather, it may be interpreted thus: ‘the
traditional dilemma defines (in some way or other) the problem of the
possibility of cognition’.
Hintikka is another who seems to suggest that the contradictory rela-
tionship between the subject and external reality is a part of Husserl’s
(along with Aristotle’s) philosophy. He asks:
Is . . . the object that we intend by means of a noema12 out there in the real
“objective” world? Or must we . . . say that the object “inexists” in the act?
He then points out:
Aristotle [and Husserl] would not have entertained such questions. For him
[/them] in thinking (intending?) X, the form of X is fully actualised both in
the external object and in the soul. If we express ourselves in the phenom-
enological jargon, this shows the sense in which the (formal) object of an act
exists both in the reality and in the act. (2010: 96)
My own point is that this characterisation of the relationship (one I agree
Husserl himself advocates) does not automatically eliminate or supplant
the traditional, ‘natural’ characterisation of the relationship, and so nor
does it eliminate the problem as it arises for that ‘natural’ characterisation.
I suggest that the phenomenological perspective is best understood as a
re-conceptualisation of the same relationship that is characterised and
11
Note that this need not go the other way: we can retain the phenomenological insight without the
inverse claim that the object itself depends on, or even is (either necessarily or always) present to,
consciousness.
12
Husserl’s name for something akin to Fregean ‘sense’, but also apparently akin to (though more fine-
grained than) Fregean ‘reference’ (for some interesting details on these subtleties, see Haddock 2010).
 Logical realism 23
problematised in the natural attitude; and so as capable of engaging
directly with its key concepts (rather than as wholly re-interpreting,
removing, or supplanting those concepts).

4. Overflow
I want now to discuss the idea of the “pregnant concept of evidence”
(Husserl 1964: 46). Husserl says:
If we say: this phenomenon of judgement underlies this or that phenom-
enon of imagination. This perceptual phenomenon contains this or that
aspect, colour, content, etc., and even if, just for the sake of argument, we
make these assertions in the most exact conformity with the givenness of
the cogitation, then the logical forms which we employ, and which are
reflected in the linguistic expressions themselves, already go beyond the
mere cogitations. A “something more” is involved which does not at all
consist of a mere agglomeration of new cogitationes. (1964: 40–1)
Elsewhere, he notes:
The epistemological pregnant sense of self-evidence . . . gives to an inten-
tion, e.g., the intention of judgement, the absolute fullness of content, the
fullness of the object itself. The object is not merely meant, but in the
strictest sense given. (Husserl 1970: 765)
The point I want to draw attention to is that Husserl takes both logical and
physical/perceptual ‘objects’ as the sort of thing that in one sense or
another ‘overflow’, or ‘go beyond’ what is given to cogitation.
The word ‘object’ must . . . be taken in a very broad sense. It denotes not
only physical things, but also, as we have seen, animals, and likewise
persons, events, actions, processes and changes, and sides, aspects and
appearances of such entities. There are also abstract objects . . . (Føllesdal,
in Føllesdal and Bell 1994: 135)
Bearing in mind that in the phenomenological reduction, access to abstract
logical forms is not treated in any especially problematic way, all of what is
given to experience can be explained in much the same fashion: “sensuous
intuition means givenness of simple objects. Categorical intuition . . . means
givenness of categorical formations, such as states of affairs, logical connectives,
and essences” (Hartimo 2010b: 117). The structure underpinning logic – the
form and structure of experience – is constituted and ‘given’ in experience. It is
‘seen’13 analogously to the way physical objects are seen by perception.
13
Or rather, ‘intuited’, where ‘intuition’ is used in the sense of “immediate or non-discursive
knowledge” (Hintikka 2010: 94).
24 Penelope Rush
So, the object of genuine perception and, by the extension I want to
make here, genuine categorical intuition, overflows what is given to the act
of perception or comprehension itself. For this reason it is capable of being
veridical, and is opposed to Hyletic data, which is not.14
This is because genuine perception and intuition involve noema that are
both conceptual and objectual.15 It is because each noema is objectual that
our conceptual grasp can never fully contain the whole noema: i.e. that this
grasp is always ‘pregnant’. Note that Husserl does not commit to there
being two noemata for each act of perception or comprehension, but
neither does he commit to the idea that the conceptual and the objectual
are simply two aspects of the one noema.16 Rather, his claims regarding
objectual (or, to anticipate what’s to come: ‘non-conceptual’) phenomena
and conceptual phenomena are in tension with one another.
In every noema, Husserl says:
A fully dependable object is marked off . . . we acquire a definite system of
predicates either form or material, determined in the positive form or left
“indeterminate” – and these predicates in their modified conceptual sense
determine the “content” of a core identity. (Husserl 1983: 364, italics mine)

It is within this ‘core identity’ we find that which gives the noema its
‘pregnant sense of self evidence’; that which makes what is ‘given to
cognition’ overflow cognition and any (e.g. formal) ‘agglomeration of
new cognitiones’. Other terms Husserl uses for this ‘core identity’ include:
“the object”; “the objective unity”; “the self-same”; “the determinable

14
Shim (2005) nicely characterizes hyletic data as the ‘sensual stuff ’ of experience. He gives the
following helpful example of the process of ‘precisification’ to contrast memory or fantasy with
genuine perception: “In remembering the house I used to live in, I can precisify an image of a red
house in my head. The shape, the color and other physical details of that house must be ‘filled in’ by
hyletic data. Now let’s say I used to live in a blue house and not a red house. There is, however, no
veridical import to the precisifications of my memory until confronted by the corrective
perception . . . there is no sense in talking about the veridical import in the precisifications of [the
memory or] fantasy” (pp. 219–220). In the latter cases, we may mistake merely hyletic data for non-
conceptual (or objectual) phenomena (p. 220). An analogous situation might be said (by a logical
realist) to occur for logical intuition when we encounter counter examples or engage directly with
the meaning of logical operators – in these situations we can see a genuine role for veridical input
capable of correcting or ‘precisifying’ our intuition. On the other hand, perhaps analogously to what
occurs in a fantasy or hallucination, we may mistake the mere manipulation of symbols for genuine
(veridical) comprehension.
15
Shim gives a sophisticated argument for the idea that what provides perceptual noemata with
‘overflow’ is that they have both conceptual and non-conceptual content. My idea is similar, but,
as will be elaborated shortly, the duality I want to consider should not be rendered as (non-
contradictory) aspects of one and the same object, but rather as a contradictory object; whereas
I think that Shim means the duality he proposes to be interpreted in the former sense.
16
Thanks to Graham Priest for pressing this point.
 Logical realism 25
subject of its possible predicates”; “the pure X in abstraction from all
predicates”; “the determinable which lies concealed in every nucleus and
is consciously grasped as self-identical”; “the object pole of intention”; and,
best of all: “that which the predicates are inconceivable without and yet
distinguishable from”. This is conceptually located in a similar variety of
ways, including as: “set alongside [the noema]”; “not separable from it”;
“belonging to it”; “disconnected from it”; and “detached but not separable
[from it]” (all quotations, 1983: 365–367). I simply note here that some of
these characterisations are contradictory. What I hope to indicate, in what
follows, is that this is as it should be.
To review and sum up:
The main points I get from Husserl are these: that independent abstract
‘reality’ is no more difficult to accommodate than is independent physical
reality; that conceptualising logical structures as similar to platonic math-
ematical structures does not preclude conceptualising either as immedi-
ately apprehendable objects of cognition; and thus that the idea of
independent reality as (genuinely, problematically) independent finds a
place in phenomenology.

5. McDowell
It is useful to compare what has so far been drawn from Husserl to a
specific interpretation of McDowell.
Neta and Pritchard in their (2007) article make a point that helps situate
Husserl’s programme: they argue that one way to understand attempts
(specifically McDowell’s, but their ideas extend to Husserl’s) to reach
beyond our ‘inner’ world to an external realm is precisely by close examin-
ation of the assumptions we bring to the Cartesian evil genius thought
experiment. The argument they present demonstrates links between a
particular (perhaps ‘natural’) way of conceiving the distinction between
‘inner’ and ‘outer’, and the commonly held assumption that:
(R): The only facts that S can know by reflection alone are facts that would
also obtain in S’s recently envatted duplicate. (p. 383)
Neta and Pritchard argue that McDowell rejects R on the basis that there is
something about our actual, embodied experience of the world that cannot
be replicated by stimulus, no matter how sophisticated, experienced by a
brain in a vat (compare this with Husserl’s differentiation between genuine
‘pregnant’ perception and hyletic/sensuous data). The clue as to how
McDowell rejects (R) and to uncovering the similarities between his and
26 Penelope Rush
Husserl’s approaches is in the concept ‘experience of the world’. For
McDowell, experience of (the world) is experience as (humans in the
world). The idea is that if indeed that is what we are talking about, then
when we talk of ‘experience in the world’, we cannot, as it were, ‘slice off ’
the part that is us experiencing from the part that is being experienced.
Neta and Pritchard outline McDowell’s position as follows:

McDowell (1998a) allows . . . that one’s empirical reason for believing a

certain external world proposition, p, might be that one sees that p is the
case. Seeing that is factive, however, in that seeing that p entails p. However,
McDowell also holds that such factive reasons can be nevertheless reflect-
ively accessible to the agent – indeed, he demands . . . that they be accessible
for they must be able to serve as the agent’s reasons. (p. 384, italics original)

Thus, for McDowell, ‘it is true that p’; or ‘it being so that p’, are internal to
the knower’s ‘space of reasons’. But her ‘satisfactory standing’ in the space
of reasons in which p is so, involves ‘seeing that p’, which entails p itself.
McDowell’s ‘factive reasons’ are subtle things with clear similarities to
Hintikka’s characterisation of the Aristotelian/Husserlian ‘object of an act’:
they are knowable by reflection alone, but also entail objective ‘external’
states. I remember my then seven-year-old son once saying ‘I think the
trees have faces’, and thinking that this is a nice way of explaining some of
the ideas in McDowell’s Mind and World (1994), which I take as an
attempt to argue that what is external and objectively so is nonetheless
also accessible – available to us as conceptual content.
But I think that the McDowellian/Husserlian sort of manoeuvre can
only work if ‘what is experienced’ genuinely is the realist’s independent
reality (at least as much as it is accessible content). To the extent that
any account re-casts or re-defines that independence, it is hard to see
how the specifically realist problem (which both McDowell and Husserl
identify in the ‘natural attitude’) is the problem their accounts actually
address.
Put another way, if an account implicates the external in our human
(reflective) experience simply by fiat (or by initial (re)design), then it
becomes difficult to see how such an account can help us understand the
problem that inspired it in the first place: i.e. the problem of the realist’s
conception of independence as independence from human experience.
McDowell’s and Husserl’s solution are of a kind, both answer the sceptic
along the following general lines: you can’t take away reference to external
reality (as in the sceptical scenario) just because what we experience has
external reality somehow written into it. But if a position’s ‘inwritten’
 Logical realism 27
externality collapses into (even an interesting) aspect of what remains,
strictly, internal, then that position offers no essential insight into the
dichotomy and the problem with which we began.

6. Effectively defending ~R
The important word in the preceding paragraph is “somehow”. Expanding
on the ‘somehow’, we can find a sense in which neither McDowell nor
Husserl escapes or resolves the traditional, ‘natural’ dilemma. Or rather, to
the extent that they can be said to, their solutions do not address this
original dilemma. Conversely, I want to suggest it is just to the extent that
they don’t escape the dilemma that they may (via expansion on the
‘somehow’) be taken as having offered a sort of solution wherein what
was unintelligible from the traditional/natural perspective, is made at least
a little intelligible. That is, their sort of insight might be taken as offering a
perspective from which the contradiction inherent in speaking of a reality
independent of humans altogether need not automatically undermine the
possibility of a relationship between the two.
To see this, we need to start by outlining the ways in which both
positions “clearly [challenge] the traditional epistemological picture that
has (R) at its core”.
Neta and Pritchard outline McDowell’s challenge to R this way:
McDowell’s acceptance of reflectively accessible factive reasons . . . entails
that the facts that one can know by reflection are not restricted to the
“inner” in this way, and can instead, as it were, reach right out to the
external world, to the “outer”. One has reflective access to facts that would
not obtain of one’s recently envatted duplicate, on McDowell’s picture. If
this is correct, it suggests that the popular epistemological distinction
between “inner” and “outer” which derives from (R) should be rejected,
or at least our understanding of it should be radically revised. (p. 386)

Not believing R is tantamount to taking a more sophisticated or more

complex view of the original Cartesian experiment. To accept ~R, we need
reasons to suppose that the thought experiment of ‘doubting everything’ is
not simply or not only constructible along lines drawn from our ‘natural’
understanding of the ‘outer/inner’ distinction. Husserl offers the broad
reason that consciousness per se is not possible – if we try to imagine such
a thing, we find a sense in which independent reality got there before us:
consciousness itself incorporates ‘potentialities’ that, in turn, cannot be
reduced to wholly ‘subjective’ or ‘internal’ phenomena.
28 Penelope Rush
Neta and Pritchard argue that the temptation to interpret McDowell
either wholly internally or wholly externally rests on believing R. Believing
R then, is very like holding fast to the possibility that in principle what is
given to cognition and what cognition intends, can always be untangled.
For Husserl, only a radically impoverished view of envattedness can deliver
the sceptical conclusion: a closer, careful look at cognition “in general,
apart from any existential assumptions either of the empirical ego or of a
real world” (Husserl 1981: 60) returns the world in all of its “modes of
givenness” (Husserl 1981: 59), as constituted and constituting that cogni-
tion. So I think it is reasonable to take Husserl similarly to McDowell on
the question of envattedness: i.e. to take Husserl as committed to there
being a difference between envatted and non-envatted states.
But I want to take issue with Neta and Pritchard’s claim that: “Once (R)
is rejected . . . these two aspects [internal and external] of the view are no
longer in conflict” (Neta and Pritchard 2007: 38b). And, for the same
reasons, I take issue with similar claims Husserl makes regarding phenom-
enology e.g.: “In . . . phenomenology . . . the old traditional ambiguous
antitheses of the philosophical standpoint are resolved” (Husserl 1981: 34).
A genuine resolution of the ‘traditional antithesis’ could come about only
via an explicit defence of ~R in the original (‘traditional’) terms in which
R itself was conceptualised. In short, a ‘resolution’ of the problem gener-
ated by the original dichotomy must directly address that dichotomy as a
genuine dichotomy.
There are various ways ~R and an alternative conceptualisation of the
internal/external dichotomy might be defended, but only some of these
ways can be said to address and so potentially resolve, the original realist
dilemma. For example, ~R itself, or a set of key reasons offered to
believe ~R, might be used as a sort of first principle, or established by fiat;
then again, an approach might give a bunch of positive reasons or argu-
ments for ~R (independent of the original reasons for R) in order to
convince us that ~R (along with any attendant, independent positive
reasons offered for ~R) ought to replace or provide an alternative perspec-
tive to the ‘traditional’ perspective. But neither of these cases can be said to
resolve the original problem. They might be said to replace that problem,
perhaps; or to render it irrelevant in the face of a potentially more
compelling scenario, but not to resolve that problem.
Any potential resolution would need to directly challenge the original
‘traditional antithesis’ itself, which cannot be done except by explicitly
engaging with that antithesis on its own terms (for a more detailed defence
of these ideas, see Rush 2005). That is, an explicit argument against
 Logical realism 29
R (accommodating the terms and spirit in which it was intended) has to
defend one of the following claims: an internal phenomenon is also not-
internal (i.e. that a phenomenon able to act as internal in the R thought
experiment is also one able to act as not-internal in the thought experi-
ment); an external phenomenon is also not-external (taking ‘internal’
phenomena as ‘not-external’); or, for each case, there is no straightforward
either/or dichotomy (i.e. it is not the case that such phenomena are ‘either
external or not-external’, or ‘internal or not-internal’). That is, in R, the
concepts ‘external’ and ‘internal’ are explicitly (intended as) subject to both
the law of excluded middle (LEM): ~A v A; and the law of non-contradic-
tion (LNC): ~(~A & A).17
So accounts that rest on or incorporate ~R in some way must also
directly challenge the applicability of these classical laws to the internal/
external dichotomy. One such challenge might argue that the point of ~R
is that it gives us reason to doubt that the LEM should hold here. The
relationship between the phenomenological and ‘natural’ perspectives
might then be seen as analogous to the relationship between the intuition-
ist rendering of the continuum as viscous and the classical rendering of the
continuum as discrete. From the intuitionist’s perspective, the continuum
has characteristics it does not have from the classical perspective. To see the
former, we need to allow the LEM to fail, in particular, for 8x8y((x<y) v
(x=y) v (x>y)). In much the same way, we could argue that to see the more
complex characteristics of our human experience in the world, we need to
allow the LEM to fail for 8x(Ix v ~Ix) (where I is ‘internal’) and/or for 8x
(Ex v ~Ex) (where E is ‘external’). (For more on the intuitionists’
continuum, see Posy 2005, especially pp. 345–348.)
Note that this means that the most effective defence of ~R challenges
the universal applicability of the laws of (classical) logic. So knowledge of
(external, independent) logical truths is guaranteed only by an explicit,
rather drastic instance of the corrigibility of that knowledge. Thus, the
knowledge of logic that survives the phenomenological reduction is corri-
gible knowledge – but this is perhaps what we should expect, given the
independence of logical truth: its fundamental role in cognition does not
and cannot guarantee the infallibility of our own intuition.

17
That is, I think arguments for the claim that Husserl’s and McDowell’s accounts do not
‘hypostatise’ ultimately fail (for examples of such arguments, see Hartimo 2010b, and Putnam
2003, particularly p. 178). Or, to the extent that they succeed, the accounts themselves are rendered
largely irrelevant to the philosophical problem I am addressing here.
30 Penelope Rush
Another such challenge could argue that in the case of independent,
external and dependent, internal phenomena, we have an explicit excep-
tion to the LNC: there are occasions where each type of phenomenon
both is and is not that type (for more on this idea see Rush 2005 and
Priest 2009).
Either way, these challenges undermine the notion that the LEM and/or
the LNC apply to internal and external phenomena. My own opinion is
that it makes more sense, for an account wishing to engage with the
philosophical problem, to mount the latter challenge – i.e. to argue that
the LNC does not apply here, (given that it could be argued that LEM
defines the terms of the original thought experiment, R) – but the main
point is that only an explicit argument against (or recognising an implicit
rejection of ) either or both of these classical rules can make such accounts
as Husserl’s and McDowell’s relevant to the original ‘natural’ problem.
And I do think that Husserl was interested in addressing the original
‘natural’ problem,18 but in a particular way:
one’s first awakening to the relatedness of the world to consciousness [i.e.
the philosophical problem] gives no understanding of how the varied life of
consciousness, . . . manages in its immanence that something which mani-
fests itself can present itself as something existing in itself, and not only as
something meant but as something authenticated in concordant experience.
(1981: 28) [and]
We will begin with a clarification of the true transcendental problem, which
in the initial obscure unsteadiness of its sense makes one so very prone . . .
to shunt it off to a side track. (1981: 27)
In Husserl’s account then, there is a duality (akin to McDowell’s)
‘within’ the constituted object itself, insofar as it is also ‘given’ as
independent. That this duality is a genuine counterexample either to the
18
Shim, Teiszen, and others see the duality (which Shim renders as conceptual/non-conceptual) as
residing strictly in the phenomenological attitude, and so Shim (2005) argues that the
phenomenological ‘solution’ cannot neatly slot into a ‘natural’ answer to scepticism. But I think
phenomenology is relevant to the natural answer to scepticism exactly insofar as it provides this
explicit way of differentiating ‘being in the (real) world’ from ‘envattedness’. This differentiation
disrupts a neat holistic story, and so its lesson, carried through to science and the natural attitude, is
perhaps not a ‘categorical mistake’ (Shim 2005: 225), but an alert as to the deficiencies of a
philosophy that disallows any perspective other than its own. What we know from the
phenomenological attitude might resist reduction to naturalist/scientific knowledge, but it
nonetheless can offer an insight into the items with which the scientific/philosophical attitude is
concerned: e.g. reality, experience, and knowledge. It is exactly what makes the phenomenological
perspective “both tempt and frustrate . . . the very philosophical desire it should have satisfied”
(Shim 2005: 225), that can make it relevant to that ‘desire’, and can potentially stop a too quick,
neat, sealed holist answer from gaining complete purchase.
 Logical realism 31
LEM or LNC (or both) is something Husserl seems at times to appreciate –
recall the contradictions in his various accounts of the ‘location of pure x’
listed earlier.
And it is just where it seems able to incorporate the rejection of the
LNC for internal and external reality that phenomenology holds the most
promise. On the other hand the preservation of the LNC in this case calls
for resolution one way or the other and so renders an account open to
being interpreted as wholly internal or wholly external, which I contend,
would drastically impoverish it as an account of human experience. As it
stands though, its own internal inconsistencies bear witness to the richness
of the very idea of phenomenology: of the inescapable, paradoxical, yet
entirely natural thought that our human experience is irreducibly consti-
tuted by the notion (itself inherently either incomplete or inconsistent)
that we might know reality and logic as it independently is.19
19
Thanks to Graham Priest, Curtis Franks, Tuomas Tahko, Sandra Lapointe, and Jody Azzouni for
helpful feedback on earlier drafts.
 chapter 2

A defense of logical conventionalism

Jody Azzouni

1. Introduction
Our logical practices, it seems, already exhibit “truth by convention.”
A visible part of contemporary research in logic is the exploration of
nonclassical logical systems. Such systems have stipulated mathematical
properties, and many are studied deeply enough to see how mathematics –
analysis in particular – and even (some) empirical science, is reconfigured
within their nonclassical confines.1 What also contributes to the appear-
ance of truth by convention with respect to logic is that it seems possible –
although unlikely – that at some time in the future our current logic of
choice will be replaced by one of these alternatives. If this happens, why
shouldn’t the result be the dethroning of one set of logical conventions for
another? One set of logical principles, it seems, is currently conventionally
true; another set could be adopted later.
Quine, nevertheless, is widely regarded as having refuted the possibility
of logic being true by convention. Some see this refutation as the basis for
his later widely publicized views about the empirical nature of logic.
Logical principles being empirical, in turn, invites a further claim that
logical principles are empirically true (or false) because they reflect well (or
badly) aspects of the metaphysical structure of the world. Just as the truth
or falsity of the ordinary empirical statement “There is a table in Miner
Hall 221B at Tufts University on July 3, 2012,” reflects well or badly how a
part of the world is, so too, the Principle of Bivalence is true or false because
it reflects correctly (or badly) the world’s structure. I’ll describe this
additional metaphysical claim – one that I’m not attributing to Quine
(by the way) – as taking logical principles to have representational content.
Most philosophers think logical principles being conventional is

1
The families of intuitionistic and paraconsistent logics are the most extensively studied in this respect.
There is a massive literature in both these specialities.

32
 A defense of logical conventionalism 33
incompatible with those principles having representational content.2
I undermine the supposed opposition of these doctrines in what follows.
That still leaves open the question whether logical principles do have
representational content; but I also undermine this suggestion. That may
seem a lot to do in under eight thousand words. Luckily for me (and for
you too), most of the important work is already done, and I can cite it
rather than have to build my entire case from scratch.

2. Quine’s dilemma
It’s really really sad that almost no one notices that Quine’s refutation of
the conventionality of logic is a dilemma. The famous Lewis Carroll
infinite regress assails only one horn of this dilemma, the horn that
presupposes that the infinitely many needed conventions are all explicit.
Quine (1936b: 105) writes, indicating the other horn:
It may still be held that the conventions [of logic] are observed from the
start, and that logic and mathematics thereby become conventional. It may
be held that we can adopt conventions through behavior, without first
announcing them in words; and that we can return and formulate our
conventions verbally afterwards, if we choose, when a full language is at our
disposal. It may be held that the verbal formulation of conventions is no
more a prerequisite of the adoption of conventions than the writing of a
grammar is a prerequisite of speech; that explicit exposition of conventions
is merely one of many important uses of a completed language. So con-
ceived, the conventions no longer involve us in vicious regress. Inference
from general conventions is no longer demanded initially, but remains to
the subsequent sophisticated stage where we frame general statements of the
conventions and show how various specific conventional truths, used all
along, fit into the general conventions as thus formulated.
Quine agrees that this seems to describe our actual practices with many
conventions, but he complains that (Quine 1936b: 105–106):
it is not clear wherein an adoption of the conventions, antecedently to their
formulation, consists; such behavior is difficult to distinguish from that in
which conventions are disregarded . . . In dropping the attributes of delib-
erateness and explicitness from the notion of linguistic conventions we risk
depriving the latter of any explanatory force and reducing it to an idle label.

2
Ted Sider, a contemporary proponent of the claim that logical idioms have representational content,
represents the positions as opposed in just this way; he (Sider 2011: 97) diagnoses “the doctrine of
logical conventionalism” as supporting the view that logical expressions “do not describe features of
the world, but rather are mere conventional devices.”
34 Jody Azzouni
We may wonder what one adds to the bare statement that the truths of
logic and mathematics are a priori, or to the still barer behavoristic state-
ment that they are firmly accepted, when he characterizes them as true by
convention in such a sense.

These challenges aren’t specifically directed against the conventionality

of logic but against tacit “conventions” of any sort. One challenge is
concerned with making sense of when specific behaviors are in accord
with the proposed tacit conventions and when they’re not. One prob-
lem, that is, is this: if the conventions are explicit, we know what the
conventions are – because they’ve been stated explicitly – and the
behavior can be directly measured against them to determine deviations.
But tacit conventions must be inferred from that very behavior, so the
challenge goes, and therefore a lot of unprincipled play becomes possible
because various conventions may be posited, these conventions differing
in how far the practitioners’s behavior is taken to deviate from them.
A second issue Quine raises is with the label “convention”; he wants to
know what’s distinctive about tacit conventions that makes them stand
apart from the simple “behavioristic” attribution that the population
“firmly accepts them.”
So Quine’s two objections come apart neatly. There is, first, a challenge
to the idea that a set of rules can be attributed to a population in the
absence of explicit indications like a set of official conventions. Even if this
first challenge can be circumvented, the second worry is why the set of
rules so attributed to a population should be called “conventions.”3
If we concede the requirement of explicitness to Quine, we’re forced to
something like the Lewis account of convention:4
A regularity R, in action or in action and belief, is a convention in a
population P if and only if, within P, the following six conditions hold:
(1) Almost everyone conforms to R.
(2) Almost everyone believes that the others conform to R.
(3) This belief that the others conform to R gives almost everyone a good
and decisive reason to conform to R himself.
(4) There is a general preference for general conformity to R rather than
slightly-less-than-general conformity – in particular, rather than con-
formity by all but anyone.

3
See (Quine 1970b) for a reiteration of the first challenge with respect to linguistic rules.
4
See (Lewis 1969: 78) – but I draw this characterization from (Burge 1975: 32–33).
 A defense of logical conventionalism 35
(5) There is at least one alternative R 0 to R such that the belief that the
others conformed to R 0 would give almost everyone a good and
decisive practical or epistemic reason to conform to R 0 likewise; such
that there is a general preference for general conformity to R 0 rather
than slightly-less-than-general conformity to R 0 ; and such that there
is normally no way of conforming to R and R 0 both.
(6) (1)–(5) are matters of common knowledge.
There are many problems with this approach – indeed, it’s no exaggeration
to describe condition (6) as yielding the result that there are almost no
conventions in any human population anywhere. But can Quine’s chal-
lenges be met? Are tacit conventions cogent?

3. Tacit conventions: Burge and Millikan. Suboptimality

Since Quine’s challenges are directed towards tacit conventions of any sort,
let’s look at what appears to be a less-complicated case: purportedly tacit
linguistic conventions. Linguistic conventionality seem less complicated
than logical conventionality if only because the intuitions that seem to
accompany logical principles (ones about necessity, ones about aprioricity)
aren’t present in the linguisitic case. As Burge (1975: 32) writes, “Language,
we all agree, is conventional. By this we mean partly that some linguistic
practices are arbitrary: except for historical accident, they could have been
otherwise to roughly the same purpose.” He adds, “which linguistic and
other social practices are arbitrary in this sense is a matter of dispute.” I’ll
shortly show that this matters to the empirical question whether language is
conventional (and in what ways) – the thing Burge tells us we all agree
about.
But first, notice something important that Burge is explicit about
(although he doesn’t dwell on it): there are psychological mechanisms that
enable these regularities. Burge (1975: 35) writes, “the stability of conven-
tions is safeguarded not only by enlightened self-interest, but by inertia,
superstition, and ignorance.” He makes this point rapidly, and in passing,
because he’s instead intent on undercutting the explicitness assumption for
conventions: “Insofar as these latter play a role, they prevent the arbitrari-
ness of conventional practices from being represented in the beliefs and
preferences of the participants.”
Let’s focus on the important word “inertia.” This is an allusion to an –
ultimately neurophysiological – mechanism of imitation. The point is
made quite explicit some years later by Millikan when she characterizes
36 Jody Azzouni
“natural conventionality” in terms of patterns that are “reproduced.”
Crucial to the idea (Millikan 1998: 2) is that “these [conventional] patterns
proliferate . . . due partly to weight of precedent, rather than due, for
example, to their intrinsically superior capacity to perform certain func-
tions.” That is (Millikan 1998: 3), “had the model(s) been different . . . the
copy would have differed accordingly.”
Some may be worried about this characterization of conventional pat-
terns.5 As I understand the characterization, for it to work we need to
sharply distinguish between the patterns being conventional because they
are proliferating partly due to the weight of precedent, and the patterns
instead only being thought to be conventional because they’re thought to
involve arbitrariness in our choice of a course of action. On the one hand,
we can simply be wrong – thinking that arbitrariness is involved when it
isn’t. On the other hand, there can be “arbitrariness” without our realizing
it: there are other model-options we don’t know about, which, were they
in place, would have been imitated instead.
Consider the venerable practice of rubbing two sticks together to start a
fire.6 A tribal population may simply fail to realize that banging rocks
together will work instead. Their practice of rubbing sticks to start a fire is
conventional despite their failing to realize this. Imagine, however, that
they live where there are no such rocks, and where, presumably, there are
available no other ways to start a fire. Then the practice isn’t conventional.
Suppose (after many moons) the tribe migrates to an area where suitable
rocks are located. Because of a change of location, a practice that wasn’t
conventional has become conventional. (More generally, technological
development can induce conventionality because it creates practical alter-
natives that weren’t there before.) There is a lot of work to do here (much
of it empirical) detailing more fully the notion of “genuine practical
alternatives” – what sort of background factors should be seen as relevant
and which not – but the need for hard empirical work isn’t problematical
for this characterization of tacit convention.
Another worry. Many people believe (and some believe correctly) that
some of their practices P are optimal. They engage (imitate) those practices
(so they believe) precisely because they think these are optimal practices
and not because of the weight of precedent. Conventional or not
conventional? Well, beliefs about optimality aren’t relevant; only the

5
Epstein (2006), for example, is worried. My thanks to him for conversations (and email exchanges)
about this topic that have influenced the rest of this section.
6
I draw this example from (Epstein 2006: 4).
 A defense of logical conventionalism 37
efficacy of P ’s optimality to the spread of the practice through the
population is relevant. Suppose alternative suboptimal practices would
not have spread through the population, if instead they were the models,
precisely because their suboptimality would have extinguished the prac-
tices (or the population engaging in them). Then P isn’t conventional.
Otherwise it is. Superior optimality, of course, can be why a practice
triumphs over alternatives. It’s an empirical question in what ways the
optimality of a practice relates to its popularity, but I’m betting that
superior optimality rarely counts for why a practice P spreads through a
population.7 If a practice has enough optimality over other options to
make its superior optimality efficacious in its spread, then it isn’t conven-
tional. On the other hand, some superior optimality clearly isn’t enough to
erase conventionality. Therefore: How much superior optimality is
required to erase conventionality is an empirical question, turning in part
on how much damage a suboptimal practice will inflict on its population,
how fast this will happen, how fast this will be noticed, and so on. These
empirical complications, although of interest, don’t make the notion of
tacit convention problematical.
One point in the previous paragraph must be stressed further because
I seem to be definitively breaking with earlier philosophers on convention-
ality on just this point. This is that roughly equivalent optimality is invari-
ably built into the characterization of conventionality: the alternative
practices that render a practice conventional are ones that are reasonably
equivalent in their optimality – this is built into Lewis’ approach by
condition (3), that others conforming to such alternatives would give
people “good and decisive” reasons for engaging in them as well – this is
false if the alternative practices are suboptimal enough. It seems built into
Millikan’s approach – at least when conventional patterns serve functions –
because alternatives should serve functions “about as well” (Millikan
2005: 56).
Unfortunately, as Keynes is rumored to have pointed out in a related
context, in the long run we’re all dead. Anthropology reveals that seriously
suboptimal practices are quite stable in human populations (and, to be

7
Is it conventional that we cook some of our food and don’t eat everything raw? I think it is. Is the
alternative suboptimal? There is controversy about this, but I think it is: I think this is why the
alternative eventually died out among our progenitors (after thousands of years, that is). On the other
hand, some of the reasons for why the alternative died out (the greater likelihood of food poisoning,
the inadvertent thriving of parasites in one’s meal, etc.) have been – presumably – eliminated by
technical developments in food processing. So the practice of eating all food raw needn’t be as
suboptimal as it once was.
38 Jody Azzouni
honest, a cold hard look at our own practices reveals exactly the same
thing). Evolution takes a really long view of things – even the extinction of
a population because it engages in a suboptimal practice may occur so
slowly that the conventional fixation of that practice can occur for many
generations, at least.8 How suboptimal a practice can be (in relation to
alternatives) is completely empirical, of course, and turns very much on the
details of the practices involved (and the background context they occur in);
but optimality comparisons should play only a moderate role in an evalu-
ation of what alternative candidates there are to a practice, and therefore in
an evaluation of whether that practice is conventional and in what ways.
(This will matter to the eventual discussion of the conventionality of logic:
that alternative logics are suboptimal in various ways won’t bar them from
playing a role in making conventional the logic we’ve adopted.)
One last additional point about conventionality that I’ve just touched
on in the last sentence. This is that it isn’t – so much – entire practices that
are conventional, but aspects of them that are. “Minor” variations in a
practice are always possible, minor variations that we don’t normally treat
as rendering the practice conventional because we don’t normally treat
those variations as rendering the practice a different one. There are many
variations in how sticks can be rubbed together, for example. How we
describe a practice or label it (how we individuate it) will invite our
recognition of these variations as inducing conventionality or not. It’s
conventional to rub two sticks together in such and such a way, but not
conventional (say) to rub two sticks together instead of doing something
else that doesn’t involve sticks at all (in a context, say, where there are no
rocks). How we individuate “practices” correspondingly infects how and in
what ways we recognize a practice to be conventional; but this is hardly an
issue restricted to the notion of tacit convention, or a reason to think the
notion has problems.

4. Empirical evidence for tacit conventions

More than a serviceable notion of tacit convention is needed to respond to
Quine. Recall his worry about evidence, that “in dropping the attributes of
8
A nice example, probably, is the arrangement of the lettered keys on computer keyboards. No doubt
the contemporary distribution of letters is suboptimal compared to alternatives; it’s clearly an inertial
result of the earlier arrangement of the keys on typewriters – which was probably also suboptimal in
its time and relative to its context at that time. I’m not suggesting, of course, that keyboard
conventions are contributing to a future extinction event – although I have no doubt that a
number of conventions that we currently use are doing precisely that.
 A defense of logical conventionalism 39
deliberateness and explicitness from the notion of linguistic convention we
risk depriving the latter of any explanatory force and reducing it to an idle
label.” As it turns out, and this is an empirical discovery, for conventional
patterns to even be possible for a human population requires neurophysio-
logical capacities and tendencies in those humans. These are currently
being intensively studied, and preliminary results reveal how human
children have a capacity to imitate that’s largely not shared with other
animals.9 The recently discovered mirror-neuron system is crucial to this
capacity (but is hardly the whole story). My point in alluding to this
empirical literature is to indicate how a systematic response to Quine’s
challenges has emerged: Not only is a decent characterization of tacit
conventionality – as noted above – now in place, but an explanation of
the capacity for imitation that underwrites tacit conventions in this sense
(and one that goes far beyond sheer behavioral facts about “firm accept-
ance”) is also emerging due to intensive scientific study.10
Of course, Millikan (and Burge) seem to largely assume that language is
“conventional” in the appropriately tacit sense. But this (on their own
views) should be an entirely empirical question – patently so now that the
neurophysiological mechanisms of imitation are being discovered. It’s an
empirical question, for one thing, whether these mechanisms (mirror
neurons, etc.) are involved in language acquisition – more specifically,
it’s an empirical question how they’re involved in language acquisition.
Imagine (instead) that something like Chomsky’s principles and param-
eters model is at work in language acquisition.11 Then the picture is this:
the child starts language-acquisition with a massive prefixed cognitive
language-structure which is multiply triggered to a final state by specific
things the child hears. Imagine (what’s surely false, but will make the
principle of the point clear) that there are (say) only three thousand and
seventeen human languages that are possible, so that the child has only to
hear a relatively small number of specific utterances for that child’s

9
See the introduction to (Hurley and Chater 2005a&b) for an overview of work as of that date. See
the various articles in the volumes for details. The first sentence of the introduction (Hurley and
Chater 2005a: 1) begins, dramatically enough, with this sentence: “Imitation is often thought of as a
low-level, cognitively undemanding, even childish form of behavior, but recent work across a variety
of sciences argues that imitation is a rare ability that is fundamentally linked to characteristically
human forms of intelligence, in particular to language, culture, and the ability to understand other
minds.” It’s important to stress how recent these discoveries are – only within the last couple of
decades.
10
One almost shocking development is that the study of these mechanisms is successfully taking place
at the neurophysiological level, and not at some more idealized (abstract) level – as is the case with
most language studies to date, specifically those of syntax.
11
See, e.g., (Chomsky and Lasnik 1995).
40 Jody Azzouni
“language-organ” to fixate on a final language. A mechanism like this, even
if it helps itself to the neurophysiological imitation mechanisms to enable
the child to imitate the initial triggers, may leave very little of actual
language as conventional simply because the child’s final-state competence
would leave practically nothing for the child to subsequently learn.12
To respond to Quine, notice, what’s needed are both subpersonal
mechanisms that allow alternative imitations (on the part of a population)
as well as feasible alternative practices made available by the contextual
background a population of humans is in. Without appropriate subperso-
nal imitation mechanisms (as opposed to say, subpersonal mechanisms of
the parameter/principles type), the apparent alternatives don’t render the
current practice conventional – because members of the population are
actually incapable of imitating those alternatives. But if the feasible alter-
native practices are absent from the contextual background then the
practice is rendered nonconventional because of this alone.

5. Three theories of logical capacity

I’ve just finished suggesting that the notion of tacit convention may
(empirically) find almost no foothold in language, despite the appearance
of massive contingency, because the mechanisms of imitation – crucial
to tacit convention – may play only a minimal role in language acquisi-
tion.13 This is an empirical question, unresolved at the moment. But
what about logic?
Despite the subject matter of logic (in some sense) being so ancient, the
actual principles of logic don’t become explicit until the very end of
the nineteenth century. I now attempt to show that – possibly unlike

12
See (Chomsky 2003), specifically page 313. See (Millikan 2003), specifically pages 37–38. This
empirical question is the nub of their disagreement, as Millikan realizes (Millikan 2003: 37): “If
[the child’s language faculty] reaches a steady state, that will be only if it runs out of local
conventions to learn.” I don’t find convincing Millikan’s arguments against the empirical
possibility of a (virtually final) steady-state for the language faculty: They seem to turn only on
the sheer impression that there’s always more language conventions for adults to acquire. But given
that the empirical question is about what actual subpersonal mechanisms are involved in language
acquisition and also in the use of the language by adults who have acquired a language, it’s hard to
see why sheer impressions of conventionality deserve any weight at all.
13
One can always introduce the appearance of massive official conventionality by individuating the
language practices finely enough – e.g., minor sound-variations in the statistical norms of utterances
determining the individuation of utterance practices (recall the last paragraph of Section 3); but I’m
assuming this trivial vindication of the “conventionality” of language isn’t what either Burge or
Millikan have in mind when they presume it as evident that natural language is full of
conventionality.
 A defense of logical conventionalism 41
the case of language, and rather surprisingly – tacit convention has a
genuine place in the characterization of logic.
There are at least three (at times competing) historical characterizations
of logic. The absence, until relatively recently, of explicit logical principles
enables the insight that these models of logic are, strictly speaking, general
theories of the basis of human logical capacities, and not a priori character-
izations of what logic must be. The earliest model, arguably, is the
substitution one. Syllogistic reasoning especially, but also contemporary
reconstruals of logic in terms of schemata, invites the thought that logical
principles require an antecedent segregation of logical idioms. Logical
truths are then characterized as all the sentences generated by the system-
atic substitution of nonlogical vocabulary for nonlogical vocabulary within
what can be characterized as a recursive set of logical schemata or argument
forms. Such a characterization also allows the view that logical principles
can be recognized by their general applicability to any subject area: logical
principles are “formal,” as it’s sometimes put, or “topic neutral.”14
A second model is the content-containment one. Here a notion of
“content” is hypothesized, and the central notion of logic – consequence
(or implication) – is characterized in terms of content-containment: the
content of an implication Im is contained in the content of the statements
Im is an implication of. An intensional version of this model is clearly at
work in Kant’s notion of analytic truth, and in notions of a number
of earlier thinkers as well. An extremely popular contemporary version of
the content-containment approach externalizes the notion of content of a
statement – taking it to be the possible situations, models, or worlds in
which a statement is true. A deductive (intensional) construal of “content”
understands the content of a statement to be all its deductive
consequences.
Yet a third model emerged only in the middle of the last century: what
I’ll call the rule-governed model of logical inference. This is that logical
deduction is to be characterized in terms of a set of rules according to
which logical proofs must be constructed. Part of the reason this model
emerged so late for logic is that it required the extension of mathematical
axiomatic methods to logic, something achieved definitively only by
Frege.15

14
See (Sher 2001) for discussion and for citations of earlier proponents of this approach to logic.
15
Although the axiomatic model anciently arose via Euclidean geometry, it’s striking that it wasn’t
generally recognized – when Euclidean geometry was translated entirely to a language-based
format – how gappy those rules were. An early view was that a nonethymematic mathematical
proof was one without “missing steps” or gaps. But this view, based as it was on a picture of a
42 Jody Azzouni
A requirement, it might have been thought, is that any model of logic
must be adequate to mathematical proof. For mathematical proof – right
from its beginning – exhibited puzzling epistemic properties: We seemed
to know that the conclusion of a mathematical proof had to be true if the
premises were: this was one important ground of the impression of the
“necessity” of mathematical results. This phenomenon seemed to demand
a logical construal, at least in terms of one of the underlying models of logic
I’ve just given: content-preservation. Substitution criteria seem irrelevant
to mathematical proof, and so did explicit rules, since the practice of
mathematical proof – apart from isolated occurrences until the twentieth
century – occurred largely in the absence of explicit rules but instead in
terms of the perceived semantic connections between specialized (explicitly
designed) mathematical concepts.16

6. A case for the conventionality of logic

Let’s grant the suggestion that what logic is has finally stabilized (as of the
middle of the last century). The standard view is that an advantage of first-
order logic over alternative logics is that all three models of logic can be
applied to it – and arguably, all three models converge as equivalent in the
first-order context. The equivalence of the rule-governed model and
the substitution model is established by the existence of equivalent char-
acterizations of first-order logical truths in terms either of sentence-axioms
or in terms of axiom-schemata. The equivalence of these characterizations
in turn with the content-containment model is enabled by Gödel’s com-
pleteness theorem, subject to the model-theoretic characterization of the
content-containment model via models (in a background set theory).
This sophisticated theoretical package of first-order classical logic isn’t
reflected in the psychological capacities of the humans who adhere (collect-
ively) to this model of reasoning. In saying this, I’m not alluding to the rich
and developing literature on human irrationality17; I’m pointing out, rather,
that as we become more sophisticated in our study of the neurophysiological

conceptual relationship between the steps in a mathematical proof, remained purely metaphorical
(or, at best, promissory) until the notion of algorithm in the context of artificial languages emerged
at the hands of Turing, Church, and others in the twentieth century.
16
See (Azzouni 2005: 18–19). It should be noted that this dramatic aspect of informal-rigorous
mathematical proof is still with us despite the presence of formal systems that are apparently fully
adequate to contemporary mathematics. That is, informal-rigorous mathematical proof continues to
operate largely by conceptual implication – supplemented, of course, with substantial
computational bits.
17
Nicely popularized by one of the major researchers in the area: See (Kahneman 2011).
 A defense of logical conventionalism 43
basis for our capacities for mathematics, and for reasoning generally, there is
no echo in our neuropsychological capacities to reason, and to prove, of the
semantic/syntactic apparatus the contemporary view of logic (and even its
competitors) provides.18 That apparatus is an all-purpose topic-neutral piece
of algorithmic machinery; how we actually reason, by contrast, involves
quite topic-specific, narrowly applied, highly componentalized, mental tools.
This means that the role of formal logic can only be a normative one; it has
emerged as a reasoning tool that we officially impose on our ordinary
reasoning practices and that we (at times) can use to evaluate that reasoning.19
The foregoing, if right, makes the conventionality of logic quite plaus-
ible even if it’s an optimal logic, compared to competitors.20 The fore-
going, if right, also makes plausible the emergence of classical logic as
explicitly conventional in the twentieth century; and it makes plausible its
role as tacitly conventional (at least in mathematical reasoning) for earlier
centuries – before sets of rules for logic became explicit. I turn now to
discussing some of the reasons philosophers have for denying logic such a
conventional status. The first kind of objection I’ll consider turns on how
the notion of truth is used in the characterization of validity; next I’ll
evaluate certain arguments that have been offered for why logical principles
have a (metaphysical) representational role.

7. Criticisms of the truth-preservation characterization of logic

We philosophers are all pretty familiar with the apparent truism – the
apparent explanatory cliché, the apparently essential characterization – of
18
See (Carey 2009), especially chapter 4 – also see (Dehaene 1997) – for good introductions to this
remarkable and important empirical literature.
19
I’ve argued that this role of formal logic has emerged in the course of the twentieth century; it first
occurred in mathematics but has spread throughout the sciences in large part because of the
mathematization of those sciences. See (Azzouni 2013), chapter 9, as well as (Azzouni 2005) and
(Azzouni 2008a) for discussion. I should stress that there are several psychological and historical
contingencies that seem involved in why the tacit employment of logical consequence in
mathematical practice turned out to be in the neighborhood of a first-order and classical one: one
of those, I suggest (Azzouni 2005), is the psychological impression (on a case-by-case basis) that the
introduction and elimination rules for the logical idioms (“and,” “or,” “not,” etc.) are content-
preserving, an impression that isn’t sustained for even quite short inference patterns, such as modus
ponens or syllogism.
20
Some philosophers argue that classical first-order logic isn’t optimal because of its representational
drawbacks: proponents of higher-order logics (e.g., Shapiro), on the one hand, think that it can’t
represent mathematical concepts such as “finite,” proponents of one or another paraconsistent
approach (e.g., Priest) think it can’t represent certain global concepts, e.g., “all sentences.” Although
I’ve weighed in on these debates, they don’t matter for the issue of whether logic is conventional
precisely because it’s been established in Section 3 of this chapter that suboptimality in relation to
competitors doesn’t bar a practice from nevertheless being an alternative candidate.
44 Jody Azzouni
deductive reasoning preserving truth: If the premises are true then the
conclusion is true (must be true) as well. Many philosophers have taken
truth-preservation to be a characterization of classical logical principles. If
the notion of truth, in turn, is a correspondence notion, then it would
seem to follow that classical logic is semantically rooted in metaphysics, in
what’s true about the world. And, it might be thought that what follows
from this is that logic cannot be conventional. This argument-strategy fails
for a large number of reasons; for current purposes, I’ll focus on only three
of its failures. The first is that a characterization of deduction as “truth-
preserving” fails to single out any particular set of logical rules – it fails to
even require that a set of logical principles be consistent! The second is
that, in any case, even if a characterization of logic in terms of truth-
preservation singled out only classical logic (and not its alternatives), that
wouldn’t rule out the conventionality of classical logic: suboptimality of
alternatives is no bar to their rendering a practice conventional. The last
reason is that truth, in any case, is too frail an idiom to root logic
semantically in the world. This is because it functions perfectly adequately
in discourses that bear no relationship to what exists.
The first claim is easy to prove. Relevant is that the truth idiom is
governed by Tarski biconditionals: given a sentence S and a name of that
sentence “S,” “S ” is true iff S. Also relevant is that this condition can’t be
supplemented by adding conditions to either wing of the Tarski bicondi-
tionals that aren’t equivalent to the wings themselves.21 But these points
are sufficient to make the truth idiom logically promiscuous: it’s compatible
with any logical principles whatsoever. Let R be any set of logical principles.
And supplement R with the following inference schema T: S ‘ T“S ”,
and T“S ” ‘ S. If the original set of rules is syntactically consistent (as, e.g.,
Prior’s “tonk” isn’t), then so is the supplemented version. That R is “truth-
preserving” follows trivially, regardless of whether R is consistent or
not: If U ‘ V according to R, then, using T, we can show: U ‘ V iff
T“U ” ‘ T“V ” holds in [R, T].
Notice that a characterization of a choice of logic being “legislated-
true” is licensed by the foregoing: Start by choosing one’s logic, and then
supplement that choice with the T-schema. The resulting logic has been
“legislated-true.” It might be thought that more substantial uses of the
truth idiom, in semantics and in model theory, can’t be executed in the
context of a nonclassical logic. But this isn’t true either. In particular,

21
See (Azzouni 2010), 4.8 and 4.9.
 A defense of logical conventionalism 45
a model theory – characterized metalogically using intuitionistic con-
nectives in the metalanguage – is homophonic to classical model
theory.22
The second point has already been established: Imagine (contrary to
what has just been shown) that a population adopts a suboptimal set of
logical principles – ones strictly weaker than ours. (Intuitionist principles,
for example.) Then one possible result would be a failure to know all sorts
of things, both empirically and in pure mathematics, that we proponents
of classical logic know. Let’s say that this is suboptimal;23 but this is hardly
fatal. And so the conventionality of logic isn’t threatened by the presumed
suboptimality of other candidates.
Lastly, a number of philosophers have thought that the Tarski bicondi-
tionals all by themselves characterize “truth” as a correspondence notion.
There are many reasons to think they are wrong about this. Among them is
the fact that if a consistent practice of using nonreferring terms, such as
“Hercules” or “Mickey Mouse” is established, such a practice remains
consistent if it’s augmented with the T schema. Regardless of whether
the truth idiom functions as a correspondence notion for certain dis-
courses, it won’t function that way in this discourse. That shows that talk
of truth has to be supplemented somehow to give it metaphysical traction.
All by itself, it doesn’t do that job.
The point generalizes, of course. In trying to determine whether logic is
conventional, some philosophers focus on specific statements like “Either it
is raining or it is not raining,” and worry about whether this statement is
about the world or not; more dramatically, some philosophers worry about
whether the supposed conventionality of logic yields the result that we
“legislate” the truth of a statement like this.24
But this misses the point. The claim that logic “is conventional” is
orthogonal to the question of whether “logical truths” have content
(worldly or otherwise), or (equivalently?) whether they are or aren’t “about
the world.” No doubt some philosophers have thought these claims
linked – especially philosophers (like the paradigmatic “positivists” influ-
enced by Wittgensteinian Tractarian views) who are driven by epistemic

22
See (Azzouni 2008b), especially sections V and VI.
23
Two issues drive my choice of the qualification-phrase: “let’s say.” First, mathematical possibilities
are richer in the intuitionist context than they are in the classical context – that could easily count
against the supposed suboptimality of intuitionistic mathematics. Second, there are a lot of results
that show that the apparent restrictions of intuitionist mathematics – and constructivist
mathematics, more generally – in applied mathematics can be circumvented.
24
See (Sider 2011: 203–204).
46 Jody Azzouni
motives to deprive logical principles of “content.” The issue, to repeat, isn’t
whether particular logical truths are or aren’t about the world, but instead
whether our current set of logical principles lives in a space of viable
candidate alternatives. In addition, the claim that logical principles (or
truth) are about the world isn’t to be established by ruling out such
worldly content on the part of statements like “Either it is raining or it
is not raining,” but by ruling in such worldly content on the part of
statements like “Either unicorns have one horn or unicorns don’t have
one horn.”
I’d like to close out this section with a couple of remarks about the
curious project of trying to find individual representational contents for
logical idioms, such as disjunction, conjunction, and so on.25 One
extremely natural way to try going about this is to give such notions
content on an individual basis via introduction and elimination rules.
We then understand the content of “and” (“&”) to be characterized
by the rules, for all sentences U and V: U & V ‘ U, U & V ‘
V, U, V ‘ U & V, and so on (familiarily) for the other idioms. An
evident danger with this approach is that the holistic nature of “logical
content” emerges clearly when it’s recognized, for example, that
intuitionistic logic can be characterized by exactly the same introduc-
tion and elimination rules, with the one exception of negation. That
logical truths not involving negation are nevertheless affected is an easy
theorem.26
We can instead attempt to capture the individualized contents of the
connectives “semantically,” via truth tables for example. The problem
here is that truth tables are simply descriptions of truth conditions in
neatly tabular form: e.g., A or B is true iff (A is true and B is not true) or
(A is false and B is true) or (A is true and B is true). As noted earlier in
this section, such an approach simply amounts to a characterization of
logical principles (in a metalanguage) using those very same logical
principles plus the T-schema. The holism problem therefore is still with
us. The appearance that we are semantically characterizing logical idioms
on an individual basis, that is, is still the same illusion that we experience
when we approach the project directly by attempting to characterize the
content of logical idioms individually, using natural deduction principles
(for example).

25
Although the discussion is murky (or perhaps just metaphorical), this seems to be part of the project
undertaken by (Sider 2011), when he speaks of “joint-carving logical notions,” e.g., on page 97.
26
See (Kleene 1971), for lots of explicitly indicated examples.
 A defense of logical conventionalism 47

8. How Walker and Sider beg the question

against logical conventionalism
Much of the argument I’ve offered here has involved technical details that
have been deliberately kept off-stage. That was a necessity because technical
details in a paper restricted to eight thousand words must largely be kept in
the background for purely spatial reasons: to describe these technical details
in even terse self-contained detail would expand the paper greatly – e.g.,
details about the role of the truth idiom in metalanguages when character-
izing a set of logical principles, or details about how the consequence
relation is holistically affected by how individual logical idioms synergistic-
ally interact. But an important warning is in order. Discussion of these
issues – specifically, the issue of the conventionality of logic – often takes
place at an informal level that masks the fact that relevant technical points
are being overlooked. I’ll close with an illustration.
Sider (2011: 104) argues against the idea that logical principles can be
legislated-true, that in particular, the statement “Either it is raining or it is
not raining,” can be legislated-true. Here is the argument:
(i) I cannot legislate-true ‘It is raining’
(ii) I cannot legislate-true ‘It is not raining’
(iii) If I cannot legislate-true j, nor can I legislate-true ψ, then I cannot
legislate-true the disjunction ┌ j or ψ ┐.
(iii) is obviously the key premise. Sider writes (2011: 104),
In defense of iii): a disjunction states simply that one or the other of its
disjuncts holds; to legislate-true a disjunction one would need to legislate-
true one of its disjuncts. . . . It is open, of course, for the defender of truth
by convention to supply a notion of legislating-true on which the argu-
ment’s premises are false. The challenge, though, is that the premises seem
correct given an intuitive understanding of “legislate-true.”
One of the oldest (but still quite popular) ways of begging the question
against proponents of alternative logics (as well as a popular way of begging
the question against logical conventionalism) is to implicitly adopt a lofty
metalanguage stance, and then use the very words that are under conten-
tion against the opponent. That doing this is so “intuitive” evidently
contributes to the continued popularity of the fallacy.
Some readers may be tempted to deny that this is a fallacy. They may
want to speak as Walker (1999: 20) does:
Anyone who refuses to rely on modus ponens, or on the law of non-
contradiction, cannot be argued with. If they insist on their refusal there
48 Jody Azzouni
is therefore nothing to be done about it, but for the same reason there is no
need to take them seriously.27
But this argument is just awful. Even if an opponent refuses to rely on
modus ponens as a law of logic, this doesn’t mean that opponent won’t be
able engage in a debate using specific inferences that fall under classical modus
ponens. This is because all it means to deny modus ponens as a logical
principle is to claim that it has exceptions. That can nevertheless leave
enormous common ground for debate – that is, for arguments that both
debaters take to be sound.28
Even if the reader who has gotten this far in the paper isn’t (or isn’t
fully) convinced by the details of the intricate philosophical argument on
offer (both onstage and off ), I can at least hope the following take-away
message is convincing: This is that the issue of whether or not logic is
“conventional” is a subtle and intricate (and interesting) philosophical
question that can’t be successfully adjudicated by merely superficially
rehearsing Quine’s old arguments against “truth by convention,” and
supplementing that rehearsal with a semantic argument for the representa-
tional content of logic that blatantly presupposes the very logical idioms
under dispute. Also pertinent (or so I would have thought) is a discussion
of the philosophical literature on tacit conventionality that has emerged
subsequent to Quine, including the relevant empirical results. I also think
(and have tried to illustrate) that needed as well is a moderately deep
discussion of whether and in what ways the attribution of “representa-
tional” contents to logical idioms does or doesn’t contradict the supposed
conventionality of logic.

27
See (Lewis 2005), where a similar refusal on similar grounds to debate the law of non-contradiction
is expressed; see (van Inwagen 1981) for the same maneuver directed towards substitutional
quantification. As I said: it’s a popular maneuver with many illustrious practitioners.
28
Metalogical debates, in particular, are ones where proponents can easily debate one another on
common ground, as many clearly do in the philosophical literature. See (Azzouni and Armour-Garb
2005) for details.
 chapter 3

Pluralism, relativism, and objectivity

Stewart Shapiro

I have been arguing of late for a kind of relativism or pluralism concerning

logic (e.g., Shapiro 2014). The main thesis is that there are different logics
for different mathematical structures or, to put it otherwise, there is
nothing illegitimate about structures that invoke non-classical logics, and
are rendered inconsistent if excluded middle is imposed. The purpose of
this chapter is to explore the consequences of this view concerning a core
metaphysical issue concerning logic, the extent to which logic is objective.
In the philosophical literature, terms like “relativism” and “pluralism”
are used in a variety of ways, and at least some of the discussion and debate
on the issues appears to be bogged down because the participants do not
use the terms the same way. One group of philosophers uses the word
“relativism” for what another group calls “contextualism”. So, in order to
avoid getting lost in cross-purposes, we need a brief preliminary concern-
ing terminology.
The central sense of “relativism” about a given subject matter Φ is given
by what Crispin Wright (2008) calls folk-relativism. The slogan is: “There is
no such thing as simply being Φ”. If Φ is relative, in this sense, then in order
to get a truth-value for a statement in the form “a is Φ”, one must implicitly
or explicitly indicate something else. A major discovery of the early twentieth
century is that simultaneity and length are relative, in this sense. To get a
truth-value for “a is simultaneous with b”, one needs to indicate a frame of
reference. Arguably, so-called predicates of personal taste, such as “tasty” and
“fun” are also folk-relative, at least in some uses. To get a truth-value for “p is
tasty”, one must indicate a judge, a taster, a standard, or something like that.
This folk notion of “relativism” seems to be the one treated in Chris
Swoyer’s 2003 article in the Stanford Internet Encyclopedia of Philosophy.
Swoyer suggests that discussions of relativism, and relativistic proposals,
focus on instances of a “general relativistic schema”:
ðGRSÞ Y is relative to X :

49
50 Stewart Shapiro
In other words, in order to formulate a relativistic proposal, one first
specifies what one is talking about, the “dependent variable” Y, and then
what that is alleged to be relative to, the “independent variable” X. So,
according to special relativity, the dependent variable is for simultaneity
and other temporal or geometric notions like “occurs before”, and phrases
like “has the same length as”. The independent variable is for a reference
frame. For predicates of personal taste, the independent variable is for a
given taste notion and the dependent variable is for a judge or a standard
(depending on the details of the proposal).
The main thesis of Beall and Restall (2006) is an instance of folk-
relativism concerning logical validity. They begin with what they call the
“Generalised Tarski Thesis” (p. 29):
An argument is validx if and only if, in every casex in which the premises are
true, so is the conclusion.
For Beall and Restall, the variable x ranges over types of “cases”. Classical logic
results from the Generalized Tarski Thesis if “cases” are Tarskian models;
intuitionistic logic results if “cases” are constructions, or stages in construc-
tions (i.e., nodes in Kripke structures); and various relevant and paraconsis-
tent logics result if “cases” are situations. So Beall and Restall take logical
consequence to be relative to a kind of case, and the General Relativistic
Schema is apt. For them, the law of excluded middle is valid relative to
Tarskian models, invalid relative to construction stages (Kripke models).
Beall and Restall call their view “pluralism”, eschewing the term
“relativism”:
we are not relativists about logical consequence, or about logic as such. We
do not take logical consequence to be relative to languages, communities of
inquiry, contexts, or anything else. (p. 88, emphasis in original)
It seems that Beall and Restall take “relativism” about a given subject matter
to be a restriction of what we here call “folk relativism” to those cases in
which the “independent variable” ranges over languages, communities of
inquiry, or contexts (or something like one of those). Of course, those are
the sorts of things that debates concerning, say, morality, knowledge, and
modality typically turn on. Here, we do not put any restrictions on the sort
of variable that the “independent variable” can range over. However, there
is no need to dispute terminology. To keep things as clear as possible, I will
usually refer to “folk-relativism” in the present, quasi-technical sense.1
1
John A. Burgess (2010) also attributes a kind of (folk) relativism to Beall and Restall: “For pluralism
to be true, one logic must be determinately preferable to another for one clear purpose while
determinately inferior to it for another. If so, why then isn’t the notion of consequence simply
 Pluralism, relativism, and objectivity 51
I propose below, and elsewhere, a particular kind of folk-relativism for
logic. The dependent variable Y is for validity or logical consequence, and
the independent variable X ranges over mathematical theories or, equiva-
lently, structures or types of structures. The claim is that different theories/
structures have different logics.
Once it is agreed that a given word or phrase is relative, in the foregoing,
folk sense, then one might want a detailed semantic account that explains
this. Are we going to be contextualists, saying that the content of the term
shifts in different contexts? Or some sort of full-blown assessment-sensitive
relativist (aka MacFarlane (2005), (2009), (2014))? Questions of meaning,
our present focus, thus come to the fore, and will be broached below. But,
as construed here, folk-relativism, by itself, has no ramifications concern-
ing semantics.
Briefly, pluralism about a given subject, such as truth, logic, ethics, or
etiquette, is the view that different accounts of the subject are equally
correct, or equally good, or equally legitimate, or perhaps even (equally)
true (if that makes sense). Arguably, folk-relativism, as the term is used
here, usually gives rise to a variety of pluralism, as that term is used here.
All we need is that some instances of the “independent variable” in the
(GRS) correspond to correct, or good, versions of the dependent variable.
Define monism or logical monism to be the opposite of logical
relativism/pluralism. The monist holds that there is such a thing as
simply being valid – full stop. The slogan of the monist is that there is
One True Logic.

1. Relativity to structure
Since the end of the nineteenth century, there has been a trend in
mathematics that any consistent axiomatization characterizes a struc-
ture, one at least potentially worthy of mathematical study. A key
element in the development of that trend was the publication of David
Hilbert’s Grundlagen der Geometrie (1899). In that book, Hilbert pro-
vided (relative) consistency proofs for his axiomatization, as well as a
number of independence proofs, showing that various combinations of
axioms are consistent. In a brief, but much-studied correspondence,
Gottlob Frege claimed that there is no need to worry about the

purpose relative” (p. 521). Burgess adds, “[p]erhaps pluralism is relativism but relativism of such a
harmless kind that to use that word to promote it would dramatise the position too much.” The
present label “folk-relativism” is similarly meant to cut down on dramatic effect.
52 Stewart Shapiro
consistency of the axioms of geometry, since the axioms are all true
(presumably of space).2 Hilbert replied:
As long as I have been thinking, writing and lecturing on these things,
I have been saying the exact reverse: if the arbitrarily given axioms do not
contradict each other with all their consequences, then they are true and the
things defined by them exist. This is for me the criterion of truth and
existence.

The slogan, then, is that consistency implies existence.

It seems clear, at least by now, that this Hilbertian approach applies, at
least approximately, to much of mathematics, if not all of it. Consistency,
or some mathematical explication thereof, like satisfiability in set theory, is
the only formal criterion for legitimacy – for existence if you will. Of
course, one can legitimately dismiss a proposed area of mathematical study
as uninteresting, or unfruitful, or inelegant, but if it is consistent, or
satisfiable, then there is no further metaphysical, formal, or mathematical
hoop the proposed theory must jump through before being legitimate
mathematics.
But what of consistency? The crucial observation is that consistency is a
matter of logic. In a sense, consistency is (folk) relative to logic: a given
theory may be consistent with respect to one logic, and inconsistent with
respect to another.
There are a number of interesting and, I think, fruitful theories that
invoke intuitionistic logic, and are rendered inconsistent if excluded
middle is added. I’ll briefly present one such here, smooth infinitesimal
analysis, a sub-theory of its richer cousin, Kock–Lawvere’s synthetic differ-
ential geometry (see, for example, John Bell 1998). This is a fascinating
theory of infinitesimals, but very different from the standard Robinson-
style non-standard analysis (which makes heavy use of classical logic).
Smooth infinitesimal analysis is also very different from intuitionistic
analysis, both in the mathematics and in the philosophical underpinnings.
In the spirit of the Hilbertian perspective, Bell presents the theory
axiomatically, albeit informally. Begin with the axioms for a field, and
consider the collection of “nilsquares”, numbers n such that n2 = 0.
Of course, in both classical and intuitionistic analysis, it is easy to
show that 0 is the only nilsquare: if n2 = 0, then n = 0. But not here.
Among the new axioms to be added, the most interesting is the principle

2
The correspondence is published in Frege (1976) and translated in Frege (1980). The passage here is
in a letter from Hilbert to Frege, dated December 29, 1899.
 Pluralism, relativism, and objectivity 53
of micro-affineness, that every function is linear on the nilsquares.
Its interesting consequence is this:
Let f be a function and x a number. Then there is a unique number d such
that for any nilsquare α, f (x þ α) = f x þ d α.

This number d is the derivative of f at x. As Bell (1998) puts it, the

nilsquares constitute an infinitesimal region that can have an orientation,
but is too short to be bent.3
It follows from the principle of micro-affineness that 0 is not the only
nilsquare:
:ð8αÞðα2 ¼ 0 ! α ¼ 0Þ:
Otherwise, the value d would not be unique, for any function. Recall,
however, that in any field, every element distinct from zero has a multi-
plicative inverse. It is easy to see that a nilsquare cannot have a multiplica-
tive inverse, and so no nilsquare is distinct from zero. In other words, there
are no nilsquares other than 0:


ð8αÞ α2 ¼ 0 ! ::ðα ¼ 0Þ , which is just


ð8αÞ α2 ¼ 0 ! :ðα 6¼ 0Þ :
So, to repeat, zero is not the only nilsquare and no nilsquare is distinct
from zero. Of course, all of this would lead to a contradiction if we also
had (8x)(x = 0_x 6¼ 0), and so smooth infinitesimal analysis is incon-
sistent with classical logic. Indeed, :(8x)(x = 0_x 6¼ 0) is a theorem of
the theory (but, since the logic is intuitionist, it does not follow that
(9x):(x = 0_x 6¼ 0)).
Smooth infinitesimal analysis is an elegant theory of infinitesimals,
showing that at least some of the prejudice against them can be traced to
the use of classical logic – Robinson’s non-standard analysis notwithstand-
ing. Bell shows how smooth infinitesimal analysis captures a number of
intuitions about continuity, many of which are violated in the classical
theory of the reals (and also in non-standard analysis). Some of these
intuitions have been articulated, and maintained throughout the history
of philosophy and science, but have been dropped in the main contem-
porary account of continuity, due to Cantor and Dedekind. To take one

3
It follows from the principle of micro-affineness that every function is differentiable everywhere on its
domain, and that the derivative is itself differentiable, etc. The slogan is that all functions are smooth.
It is perhaps misleading to call the nilsquares a region or an interval, as they have no length.
54 Stewart Shapiro
simple example, a number of historical mathematicians and philosophers
followed Aristotle in holding that a continuous substance, such as a line
segment, cannot be divided cleanly into two parts, with nothing created or
left over. Continua have a sort of unity, or stickiness, or viscosity. This
intuition is maintained in smooth infinitesimal analysis (and also in
intuitionistic analysis), but not, of course, in classical analysis, which views
a continuous substance as a set of points, which can be divided, cleanly,
anywhere.
Smooth infinitesimal analysis is an interesting field with the look and
feel of mathematics. It has attracted the attention of mainstream mathem-
aticians, people whose credentials cannot be questioned. One would think
that those folks would recognize their subject when they see it. The theory
also seems to be useful in articulating and developing at least some
conceptions of the continuum. So one would think smooth infinitesimal
analysis should count as mathematics, despite its reliance on intuitionistic
logic (see also Hellman 2006).
One reaction to this is to maintain monism, but to insist that
intuitionistic logic, or something even weaker, is the One True Logic.
Classical theories can be accommodated by adding excluded middle as a
(non-logical axiom) when it is needed or wanted. The viability of this
would depend on there being no theories that invoke a logic different
from those two. Admittedly, I know of no examples that are as compel-
ling (at least to me) as the ones that invoke intuitionistic logic. For
example, I do not know of any interesting mathematical theories that
are consistent with a quantum logic, but become inconsistent if the
distributive principle is added. Nevertheless, it does not seem wise to
legislate for future generations, telling them what logic they must use, at
least not without a compelling argument that only such and such a logic
gives rise to legitimate structures. One hard lesson we have learned from
history is that it is dangerous to try to provide a priori, armchair
arguments concerning what the future of science and mathematics
must be.
If a set Γ of sentences entails a contradiction in classical, or intuitionistic,
logic, then for every sentence Ψ, Γ entails Ψ. In other words, in classical
and intuitionistic logic, any inconsistent theory is trivial. A logic is called
paraconsistent if it does not sanction the ill-named inference of ex falso
quodlibet. Typical relevance logics are paraconsistent, but there are para-
consistent logics that fail the strictures of relevance. The main observation
here is that with paraconsistent logics, there are inconsistent, but non-
trivial theories.
 Pluralism, relativism, and objectivity 55
If we are to countenance paraconsistent logics, then perhaps we should
change the Hilbertian slogan from “consistency implies existence” to
something like “non-triviality implies existence”. To transpose the themes,
on this view, non-triviality is the only formal criterion for mathematical
legitimacy. One might dismiss a proposed area of mathematical study as
uninteresting, or unfruitful, or inelegant, but if it is non-trivial, then there
is no further metaphysical, formal, or mathematical hoop the proposed
theory must jump through.
To carry this a small step further, a trivial theory can be dismissed on the
pragmatic ground that it is uninteresting and unfruitful (and, indeed, trivial).
So the liberal Hilbertian, who countenances paraconsistent logics, might
hold that there are no criteria for mathematical legitimacy. There is no
metaphysical, formal, or mathematical hoop that a proposed theory must
jump through. There are only pragmatic criteria of interest and usefulness.
So are there any interesting and/or fruitful inconsistent mathematical
theories, invoking paraconsistent logics of course? There is indeed an indus-
try of developing and studying such theories.4 It is claimed that such theories
may even have applications, perhaps in computer science and psychology.
I will not comment here on the viability of this project, nor on how
interesting and fruitful the systems may be, nor on their supposed applica-
tions. I do wonder, however, what sort of argument one might give to dismiss
them out of hand, in advance of seeing what sort of fruit they may bear.
The issues are complex (see Shapiro 2014). For the purposes of this
chapter, I propose to simply adopt a Hilbertian perspective – either the
original version where consistency is the only formal, mathematical
requirement on legitimate theories, or the liberal orientation where there
are no formal requirements on legitimacy at all. And let us assume that at
least some non-classical theories are legitimate, without specifying which
ones those are. I propose to explore the ramifications for what I take to be a
longstanding intuition that logic is objective. One would think logic has to
be objective, if anything is, since just about any attempt to get at the world,
as it is, will depend on, and invoke, logic.

2. What is objectivity?
Intuitively, a stretch of discourse is objective if the propositions (or
sentences) in it are true or false independent of human judgment,
4
See, for example, da Costa (1974), Mortensen (1995), (2010), Priest (2006), Brady (2006), Berto
(2007), and the papers in Batens et al. (2000). Weber (2009) is an overview of the enterprise.
56 Stewart Shapiro
preferences, and the like. Many of the folk-relative predicates are charac-
teristic of paradigm cases of non-objective discourses. Whether something
is tasty, it seems, depends on the judge or standard in play at the time. So
taste is not objective (or so it seems). Whether something is rude depends
on the ambient location, culture, or the like. So etiquette is folk-relative
and, it seems, not objective. Etiquette may not be subjective, in the sense
that it is not a matter of what an individual thinks, feels, or judges, but,
presumably, it is not objective either. It is not independent of human
judgment, preferences, and the like.
One would be inclined to think that simultaneity and length are
objective, even though both are folk-relative, given relativity. As is the
case with much in philosophy (and everywhere else), it depends on what
one means by “objective”. We are told that whether two events are
simultaneous, and whether two rods are of the same length, depends on
the perspective of the observer. Does that undermine at least some of the
objectivity? But, vagueness and such aside, time and length do not seem to
depend on anyone’s judgment or feelings, or preferences. A given observer
can be wrong about whether events are simultaneous, even for events
relative to her own reference frame.
One might say that a folk-relative predicate P is objective if, for each
value n of the independent variable, the predicate P-relative-to-n does not
depend on anyone’s judgment or feelings. For example, if a given subject
can be wrong about P-relative-to-n, then the relevant predicate is objective.
However, even an established member of a given community can be wrong
about what is rude in that community. But one would not think that
etiquette is objective, even when restricted to a given community.
Clearly, to get any further on our issue, we do have to better articulate
what objectivity is, at least for present purposes. Again, objectivity is tied to
independence from human judgment, preferences, and the like. There is a
trend to think of objectivity in straightforward metaphysical terms. It must
be admitted that this has something going for it. The idea is that some-
thing, say a concept, is objective if it is part of the fabric of reality. The
metaphor is that the concept cuts nature at its joints, it is fundamental.
Theodore Sider (2011) provides a detailed articulation of a view like this,
but the details do not matter much here.
Presumably, taste and etiquette are not fundamental; tastiness and
rudeness do not cut nature at its joints (whatever that means). Does logic,
or, in particular, logical validity cut nature at its joints? It is hard to say,
without getting beyond the metaphor. What are the “joints” of reality?
Does it have such “joints”? How does logic track them?
 Pluralism, relativism, and objectivity 57
One might argue that there can be at most one logic that is objective, in
this metaphysical sense. Sider does argue that at least parts of logic are
fundamental. As it happens, the logic he discusses is classical, but, so far as
I know, there is no argument supporting that choice of logic. It might be
compatible with his overall program that, say, parts of intuitionistic logic
or a relevant logic are fundamental instead. But perhaps two distinct logics
cannot both be fundamental. Contraposing, if the present folk-relativism
about logic is correct, then logic is not objective, in the foregoing
metaphysical sense.
For what it is worth, I would not like to tie objectivity to such deep
metaphysical matters as Sider-style fundamentality. First, things that are
not so fundamental can still be objective. Intuitively, the fact that the
Miami Heat won the NBA title in 2012 is objective (like it or not), but (I
presume) it is hardly fundamental. One can call a proposition objective if it
somehow supervenes on fundamental matters, but that requires one to
accept a contentious metaphysical framework, and to articulate the rele-
vant notion of supervenience.
More important, perhaps, several competing philosophical traditions
have it that there simply is no way to sharply separate the “human” and
the “world” contributions to our theorizing. Protagoras supposedly said
that man is the measure of all things. On some versions of idealism, not to
mention some postmodern views, the world itself has a human character.
The world itself is shaped by our judgments, observations, etc. Perhaps
such views are too extreme to take seriously. A less extreme position is
Kant’s doctrine that the ding an sich is inaccessible to human inquiry. We
approach the world through our own categories, concepts, and intuitions.
We cannot get beyond those, to the world as it is, independently of said
categories, concepts, and intuitions.
On the contemporary scene, a widely held view, championed by
W. V. O. Quine, Hilary Putnam, Donald Davidson, and John Burgess,
has it that, to use a crude phrase, there simply is no God’s eye view to be
had, no perspective from which we can compare our theories of the world
to the world itself, to figure out which are the “human” parts of our
successful theories and which are the “world” parts (see, for example,
Burgess and Rosen 1997). On such views, the world, of course, is not of
our making, but any way we have of describing the world is in human
terms. As Friedrich Waismann once put it:

What rebels in us . . . is the feeling that the fact is there objectively no

matter in which way we render it. I perceive something that exists and put
58 Stewart Shapiro
it into words. From this, it seems to follow that something exists inde-
pendent of, and prior to language; language merely serves the end of
communication. What we are liable to overlook here is the way we see a
fact – i.e., what we emphasize and what we disregard – is our work.
(Waismann 1945: 146)
This Kant–Quine orientation may suggest that there simply is no object-
ivity to be had, or at least no objectivity that we can detect. Perhaps
objectivity is a flawed property, going the way of phlogiston and caloric, or
witchcraft. If this is right, then there simply is no answering the question of
this paper – folk-relativism or no folk-relativism. Logic is not objective,
since nothing is. Despite having sympathy with the Kant – Quine orien-
tation, I would resist this rather pessimistic conclusion. There may not be
such a thing as complete objectivity – whatever that would be – but it still
seems that there is an interesting and important notion of objectivity to
be clarified and deployed. There seems to be an important difference – a
difference in kind – between statements like “the atmosphere contains
nitrogen” and statements like “the Yankees are disgusting”. The distinction
may be vague and even context dependent, but it is still a distinction, and,
I think, an important one. Our question concerns whether the present
folk-relative logic falls on one side or the other of this divide (or perhaps on
or near its borderline).
Crispin Wright’s Truth and objectivity (1992) contains an account of
objectivity that is more comprehensive than any other that I know of,
providing a wealth of detailed insight into the underlying concepts. Wright
does not approach the matter through metaphysical inquiry into the fabric
of reality, wondering whether the world contains things like moral prop-
erties, funniness, or numbers. He focuses instead on the nature of various
discourses, and the role that these play in our overall intellectual and
social lives.
As Wright sees things, objectivity is not a univocal notion. There are
different notions or axes of objectivity, and a given chunk of discourse can
exhibit some of these and not others. The axes are labeled “epistemic
constraint”, “cognitive command”, “the Euthyphro contrast”, and “the
width of cosmological role”. In a previous paper, (Shapiro 2000), I argue
that logic easily passes all of the tests. The conclusion is (or was) that, on
each of the axes, either logic is objective (if anything is) or matters of logic,
such as validity and consistency, lie outside the very framework of object-
ivity and non-objectivity, since most of the tests presuppose logic. That is,
to figure out whether a given stretch of discourse is objective, on this or
that axis, one must do some logical reasoning or figure out what is
 Pluralism, relativism, and objectivity 59
consistent, or the like. So it is hard to even apply the framework to matters
of logic. The main target of Shapiro (2000) was Michael Resnik’s (1996),
(1997) non-cognitivist account of logical consequence, a sort of Blackburn
(1984)-style projectivism, which would make logic non-objective at least
on the intuitive conception of objectivity. According to Resnik, to call an
argument valid, or to call a theory consistent, is to manifest a certain
attitude toward the theory.5
The present relativism/pluralism was not on the agenda then. The plan
here is to return to the matter of objectivity with the present folk-
relativism concerning logic in focus. Sometimes we will concentrate on
general logical matters, such as validity and consistency, as such, and
sometimes we will deal with particular instances of the folk-relativism,
such as classical validity, intuitionistic consistency, and the like. We will
limit the discussion to Wright’s axes of epistemic constraint and cognitive
command.

3. Epistemic constraint
Epistemic constraint is an articulation of Michael Dummett’s (1991a)
notion of anti-realism. According to one of Wright’s formulations, a
discourse is epistemically constrained if, for each sentence P in the
discourse,
P $ P may be known: ðp: 75Þ
In other words, a discourse exhibits epistemic constraint if it contains no
unknowable truths.6
It seems to follow from the very meaning of the word “objective” that if
epistemic constraint fails for a given area of discourse – if there are
propositions in that area whose truth cannot become known – then
that discourse can only have a realist, objective interpretation. As
Wright puts it:
To conceive that our understanding of statements in a certain discourse is
fixed . . . by assigning them conditions of potentially evidence-transcendent

5
It is perhaps ironic (or at least interesting) that Resnik argues against pluralism and relativism about
logic. He claims that there “ought to be” but one logic; the logic he favors is classical.
6
Actually, if the background logic is intuitionistic, there is a difference between the absence of
unknowable truths and the truth of the biconditional: P $ P may be known. That difference
does seem to bear on Wright’s argument that if epistemic constraint fails – in the sense that there are,
or could be, unknowable propositions in that area – then the discourse is objective, but we will not
pursue this further here.
60 Stewart Shapiro
truth is to grant that, if the world co-operates, the truth or falsity of any
such statement may be settled beyond our ken. So . . . we are forced to
recognise a distinction between the kind of state of affairs which makes such
a statement acceptable, in light of whatever standards inform our practice of
the discourse to which it belongs, and what makes it actually true. The
truth of such a statement is bestowed on it independently of any standard
we do or can apply . . . Realism in Dummett’s sense is thus one way of
laying the essential groundwork for the idea that our thought aspires to
reflect a reality whose character is entirely independent of us and our
cognitive operations. (p. 4)

In other words, if epistemic constraint fails for a given discourse, then it is

objective, and that is the end of the story. The other axes of objectivity –
cognitive command, cosmological role, and the Euthyphro contrast – are
then irrelevant; they do not track a sense of objectivity (if the axis can be
applied at all). Or so Wright argues.
So what of logic? Are there, or could there be, unknowable truths
concerning logical consequence, consistency, and the like? The present
folk-relativism concerning logic pushes that question in a certain direction.
Consider a given argument, or argument form Δ, and let P be a statement
that Δ is valid. Could something like P be an unknowable truth?
Not as it stands, but that is because, absent context, P is not a truth
(or a falsehood) at all. According to the present folk-relativism, in order
to get a truth-value for P, one must specify something else, such as a
particular mathematical theory, a structure, or perhaps just a logic. We
have to ask separately whether Δ is valid in classical logic, in intuitionistic
logic, in various relevant logics, etc. So it seems to me that in order to
ask whether logic is epistemically constrained, we have to consider
statements of validity and the like with the logic made explicit. We must
consider statements in the form, Δ is valid in logic L, where L is one of
the logics that can go in for the dependent variable in the general
relativistic scheme.
To push the analogy, consider, again, relativity. Let p and q be two
events. Say that p is a runner in baseball leaving third base, and q is an
outfielder catching a fly ball. Consider the statement S that p occurred
before q (which an umpire sometimes has to adjudicate). According to
relativity, we cannot get a truth-value for S without specifying a frame of
reference. So, a fortiori, we cannot even ask if there is an unknowable truth
for a statement about what happened before what without indicating a
reference frame. If a reference frame is specified (implicitly or explicitly),
then, it seems, there can be unknowable truths in this area. For example, it
 Pluralism, relativism, and objectivity 61
may be unknowable whether the runner left base before the ball was
caught, from the perspective of the home plate umpire. For example, it
may be too dark or no human can see distinctions that fine, or whatever.
So matters of temporal order, from a given reference frame, are not
epistemically constrained. And, intuitively, matters of temporal order are
objective, vagueness aside.
The point here is that with folk-relative discourses, we can only ask
about epistemic constraint for statements that have the relevant parameters
fully specified, at least implicitly. So the central question is whether there
can be unknowable truths concerning whether the argument (form) Δ is
valid in a given logic L?
In effect, this matter was dealt with in my earlier paper (Shapiro 2000),
and also in Shapiro (2007), which concerns mathematics. Classical logic
was in focus then, but to some extent, the argument generalizes. Whether
there are unknowable truths in this area depends on what one means by
“unknowable”. If we do not idealize on the knowers, then of course there
can be unknowable truths. Suppose that our argument Δ is an instance of
&-elimination in which the premise and conclusion each have, say, 10100
characters. Then Δ is valid in, say, classical logic, but no one can
know that, since no one can live long enough to check that Δ is an instance
of &-elimination.
So, to give epistemic constraint a chance of being fulfilled, we have to
idealize on the knowers. One sort of idealization is familiar. We assume
that our knowing subjects have unlimited (but still finite) time, attention
span, and materials at their disposal, and that they do not make any simple
computation errors. These idealizations are common throughout math-
ematics, and we take them to be conceptually unproblematic (and thus we
set aside issues concerning rule-following, as in, say Kripke (1982)). Then,
if L is classical first-order logic, or intuitionistic logic, or most of the
relevant logics, and Δ is an arbitrary argument form (with finitely many
premises and conclusions), then a statement that Δ is valid in logic L is true
if and only if that fact is knowable (by one of our ideal agents). That is
because each of those logics has an effective and complete deductive
system.
Things are not so clear if the logic in question is classical second-order
logic, since its consequence relation is not effectively enumerable. Nor are
things so clear for statements that a given argument Δ is not valid in one of
the aforementioned logics. Invalidity is not recursively enumerable, and so
checking invalidity is not a matter of running an algorithm. So if we are to
insist that all matters of logic are epistemically constrained, once the logic
62 Stewart Shapiro
is fixed, we have to attribute to our knowers the abilities to decide
membership in non-recursive sets.
Things get vexed here. It is not at all clear what the relevant modality is
for the key phrase “knowable”. Moreover, as noted, the issues are essen-
tially the same as with monism concerning logic. So I propose to just take
it as given, for the sake of argument, that the relevant discourse is
epistemically constrained, in at least some relevant sense, so that we can
move on to another of Wright’s axes of objectivity.

4. Cognitive command
Assume that a given area of discourse serves to describe mind-independent
features of a mind-independent world. In other words, assume that the
discourse in question is objective, in an intuitive, or pre-theoretic sense.
Suppose now that two people disagree about something in that area. It
follows that at least one of them has misrepresented reality, and so some-
thing went wrong in his or her appraisal of the matter. Suppose, for
example, that two people are arguing whether there are seven, as opposed
to eight, spruce trees in a given yard. Assuming that there is no vagueness
concerning what counts as a spruce tree and no indeterminacy concerning
the boundaries of the yard, or whether each tree is in the yard or not, it
follows that at least one of the disputants has made a mistake: either she
did not look carefully enough, her eyesight was faulty, she did not know
what a spruce tree is, she misidentified a tree, she counted wrong, or
something else along those lines. The very fact that there is a disagreement
suggests that one of the disputants has what may be called a cognitive
shortcoming (even if it is not always easy to figure out which one of them it
is that has the cognitive shortcoming).
In contrast, two people can disagree over the cuteness of a given baby or
the humor in a given story without either of them having a cognitive
shortcoming. One of them may have a warped or otherwise faulty sense of
taste or humor, or perhaps no sense of taste or humor, but there need be
nothing wrong with his cognitive faculties. He can perceive, reason, and
count as well as anybody.
The present axis of objectivity turns on this distinction, on whether
there can be blameless disagreement:
A discourse exhibits Cognitive Command if and only if it is a priori that
differences of opinion arising within it can be satisfactorily explained only
in terms of “divergent input”, that is, the disputants working on the basis of
 Pluralism, relativism, and objectivity 63
different information (and hence guilty of ignorance or error . . .), or
“unsuitable conditions” (resulting in inattention or distraction and so in
inferential error, or oversight of data, and so on), or “malfunction” (for
example, prejudicial assessment of data . . . or dogma, or failings in other
categories . . . (Wright 1992: 92)

Intuitively, cognitive command holds for discourse about spruce trees

(vagueness and indeterminacy aside) and it fails for discourse about the
cuteness of babies and the humor of stories.
Later in the book, Wright (1992: 144) adds some qualifications to the
formulation of cognitive command, meant to deal with matters like
vagueness. A discourse exerts cognitive command if and only if
It is a priori that differences of opinion formulated within the discourse,
unless excusable as a result of vagueness in a disputed statement, or in the
standards of acceptability, or variation in personal evidence thresholds, so to
speak, will involve something which may properly be described as a cogni-
tive shortcoming.

So what of logic, again assuming the correctness of the foregoing folk-

relativism? Let Δ be a given argument form, and consider two folks who
disagree – or seem to disagree – whether Δ is valid. One says it is and the
other says it is not. Our question breaks into two, depending on whether
we fix the logic. For our first type of case, let Δ be an instance of excluded
middle or double-negation elimination, and consider the “dispute”
between advocates of classical logic and advocates of intuitionistic logic.
The inference is valid in classical logic, invalid in intuitionistic logic. For
the other sort of case, we fix the logic and ponder disputes concerning that
logic. We imagine two folks who disagree – or seem to disagree – whether
Δ is valid in L, where L is, say, a particular relevant logic.
We start with the second sort of case, disagreements that concern a fixed
logic. I would think that there is room for blameless disagreement con-
cerning how a given argument, formulated in natural language, should be
rendered in a formal language. However, such issues would take us too far
afield, broaching matters of the determinacy of meaning, the slippage
between logical terms and their natural language counterparts, and the
intentions of the arguer. It is not so clear whether a disagreement in how
to render a natural language argument is “excusable as a result of
vagueness . . . or in the standards of acceptability, or variation in personal
evidence thresholds” or the like.
So let us set such matters aside, and just assume that our target
argument Δ is fully formalized. One of our characters says that Δ is valid
64 Stewart Shapiro
in the given logic L and the other says that Δ is invalid in that logic L.
Do we know (a priori) that at least one of them has a cognitive
shortcoming?
Suppose that the logic L is either defined in terms of a deductive system
or that there is a completeness theorem for it. So L can be classical first-
order logic, intuitionistic logic, or one of the various relevant and para-
consistent logics that are given axiomatically. So our disputants differ on
whether there is a deduction whose undischarged premises are among the
premises of Δ and whose last line is the conclusion of Δ. So, up to
Church’s thesis, our disputants differ over a Σ1-sentence in arithmetic,
one in the form (9n)Φ, where Φ is a recursive predicate. So our question
concerning cognitive command for this logic L reduces to whether cogni-
tive command holds for these simple arithmetic sentences. I would think
that cognitive command does hold here. One of our disputants has made
(what amounts to) a simple arithmetic error, and that surely counts as a
cognitive shortcoming. But I will rest content with the reduction. Cogni-
tive command holds in this case if and only if it holds for 91-sentences (or,
equivalently, Π1-sentences).
Now suppose that our fixed logic L is not complete. Say it is second-
order logic, with standard, model-theoretic semantics. In that case, the
question at hand reduces to set theory. Suppose, for example, that our
target argument Δ has no premises and that its conclusion is, in effect, the
continuum hypothesis (see Shapiro 1991: 105). So Δ is valid if and only if
the continuum hypothesis is true. So, in effect, our disputants differ over
the truth of the continuum hypothesis. Is that dispute cognitively blame-
worthy? Surely, that would take us too far afield (but see Shapiro 2000,
2007, 2011, 2012), and we will leave this case with the reduction.
Let us briefly consider the analogues of our question concerning
cognitive command with our other examples of folk-relative predicates.
Suppose that two judges differ on whether a certain event a occurred
before another event b from the same frame of reference (putting aside the
fact that this discourse is not epistemically constrained). Assume, for
example, that the two judges are in the same reference frame. Then,
unless the disagreement is “excusable as a result of vagueness . . . or in the
standards of acceptability, or variation in personal evidence thresholds”,
at least one of them exhibits a cognitive shortcoming. She did not look
carefully enough, or did not time the events properly, or forgot some-
thing. So cognitive command holds, and, of course, matters of temporal
order from a fixed frame of reference are intuitively objective. The same
goes for matters of length.
 Pluralism, relativism, and objectivity 65
Now consider two folks who disagree over whether a given food is tasty
for one and the same subject. Suppose, for example, that Tom and Dick
differ over whether licorice is tasty to Harry. To keep things simple,
assume that neither Tom nor Dick is Harry. Tom’s and Dick’s judgments
would presumably be based on what Harry has told them and their
observations of his reactions when eating licorice. We should assume that
Tom and Dick have exactly the same body of such evidence (since
otherwise one of them has the cognitive shortcoming of lacking relevant
evidence). And we should set aside matters of “vagueness . . . standards of
acceptability, [and] variation in personal evidence thresholds”. Tom and
Dick may have come to opposite conclusions because they weighed certain
pronouncements or reactions differently. In this case, perhaps, neither of
them has a cognitive shortcoming – each is cognitively blameless. If so,
cognitive command fails.7 I take it that talk about taste in general –
concerning what is tasty (simpliciter) – is a paradigm of a non-objective
discourse, but I am not sure whether discourse about Harry’s taste is
objective, intuitively speaking. Maybe we have a borderline case.
Returning to matters logical, I’ve saved the hardest sort of situation for
last. That is on prima facie disagreements when the logic is not held fixed.
To focus on a specific example, let Δ be an instance of the law of
excluded middle (with no premises). Let h be a classicist who says that
Δ is valid and let b be an intuitionist who insists that Δ is not valid. Is
this a disagreement that is (cognitively) blameless? If so, then cognitive
command fails here, and this aspect of logic falls on the non-objective
side of this particular axis (assuming that cognitive command tracks a sort
of objectivity).
According to the foregoing folk-relativism, h and b are both right. Each
has spoken a truth, and so presumably there is nothing to fault either of
them. So each is (cognitively) blameless, at least concerning this particular
matter. The only question remaining is whether they disagree. Here we
encounter a matter that is treated extensively in the philosophical litera-
ture, and I must report that the issues are particularly vexed. There does
not seem to be much in the way of consensus as to what makes for a
disagreement. John MacFarlane (2014), for example, articulates several

7
A referee for Shapiro (2012) suggested that the failure of cognitive command does not distinguish
cases which are not objective from those in which evidence is scant. The situation sketched above,
with Tom and Dick, is not that different (in the relevant respect) from cases in science where
available evidence must be evaluated holistically – say in cosmology. Two scientists might both be in
reflective equilibrium, having assigned slightly different weights to various pieces of evidence.
Cognitive command might fail there, too, despite science being a paradigm case of objectivity.
66 Stewart Shapiro
different, and competing senses of “disagreement”. We will keep things at
a more intuitive level, as far as possible.
One thesis, perhaps, is that a necessary condition on disagreement is
that the parties in question cannot both be correct. If so, and continuing to
assume our folk-relativism concerning logic, we have that h and b do not
disagree. A fortiori, we do not have a case of blameless disagreement. We
can still maintain that cognitive command holds when the logic is held
fixed, as above, and so logic passes this test for objectivity.
The thesis that in a disagreement both parties cannot be correct is
controversial. It is sometimes taken as a criterion of being non-objective
that parties can disagree and both be correct (see, for example, Barker
2013). Suppose that Harry announces that licorice is tasty, and Jill
responds, “no it is not; licorice is disgusting”. That looks like a disagree-
ment; Jill uses the language of disagreement, apparently denying Harry’s
assertion. And yet, one might say, both are correct. Or at least one might
say that both are correct.
To make any progress here, we have to get beyond the loose character-
ization of folk-relativism and address matters of semantics. I’ll briefly
sketch the framework proposed by John MacFarlane (2005), (2009),
(2014) for interpreting expressions in a folk-relative discourse. The terms
used by other philosophers and linguists can usually be translated into this
framework, though sometimes with a bit of loss.
Indexical contextualism about a given term is the view that the content
expressed by the term is different in different contexts of use. The clearest
instances are the so-called “pure indexicals”, words like “I” and “now”. The
content expressed by the sentence “I am hungry”, when uttered by me on a
given day, is different from the content expressed by the same sentence,
uttered by my wife at the same time. Intuitively, the first one says that I am
hungry (then) and the second says that she is hungry (then). Clearly, these
are different propositions; they don’t say the same thing about the world –
not to mention that one might be true and the other false.
Although very little is without controversy in this branch of philosophy
of language, words like “enemy”, “left”, “right”, “ready”, and “local” seem
apt for indexical contextualist treatments.8 Suppose, for example, that Jill,
sitting at a table says that the salt is on the left while, at the same time,
Jack, who is sitting opposite her, says that the salt is not on the left (since it
is on the right). Intuitively, Jack and Jill do not disagree with each other,
and the propositions they express are not contradictories. The reason is
8
Of course, this is not to say that these terms are like the standard indexicals in every manner.
 Pluralism, relativism, and objectivity 67
that the content of the word “left” is different in the two contexts. In the
first, it means something like “to the left from Jill’s perspective” and in
the second it means “to the left from Jack’s perspective”. And they can
both be correct – intuitively one of them is correct just in case the
other is.
Non-indexical contextualism, about a given term, is the view that its
content does not vary from one context of use to another, but the extension
can so vary according to a parameter determined by the context of
utterance.9 Suppose, for example, that a graduate student sincerely says
that a local roller coaster is fun, and her Professor replies “No, that roller
coaster is not fun, it is lame”. According to a non-indexical contextualism
about “fun” (and “lame”), each of them utters a proposition that is the
contradictory of that uttered by the other – so they genuinely disagree. Yet,
assuming both are accurately reporting their own tastes, each has uttered a
truth, in his or her own context. For the graduate student, at the time,
the roller coaster is fun, since it is fun-for-the-graduate-student. For the
professor, the roller coaster is not fun, since it is not fun-for-the-professor.
Indeed, it is lame-for-the-professor.
Finally, assessment-sensitive relativism, sometimes called “relativism
proper”, about a term agrees with the non-indexical contextualist that
the content of the term does not vary from one context of use to another,
and so, in the above scenario, the relativist holds that the graduate student
and the professor each express a proposition contradictory to one expressed
by the other. However, for the assessment-sensitive relativist, the term gets
its extension from a context of assessment. Suppose, for example, that a third
person, a Dean, overhears the exchange between the graduate student and
professor and, assume that the roller coaster is not fun-for-the-Dean.
Then, from the context of the Dean’s assessment, the student uttered a
false proposition and the professor uttered a true one. And, from the
graduate student’s context of assessment, the Professor uttered a false
proposition, and from the Professor’s context of assessment, the student
uttered a false proposition.
According to MacFarlane, the difference between non-indexical
contextualism and assessment-sensitive relativism is made manifest by
the phenomenon of retraction. That difference does not matter here, and
we can lump non-indexical contextualism and assessment-sensitive

9
Nearly all terms have different extensions in different possible worlds. That is not the sort of
contextual variation envisioned here. For terms subject to non-indexical contextualism, the
relevant contextual parameter is for a judge, a time, a place, etc.
68 Stewart Shapiro
relativism together. If we go with a contextualist treatment of the dispute
between our logicians h and b, then they do not disagree, and so cognitive
command is saved. If we opt for a non-indexical contextualist or an
assessment-sensitive interpretation, we do have a disagreement – in the
sense that each of them accepts a content that is the contradictory of that
accepted by the other. As above, the disagreement is blameless (since both
are correct), and so cognitive command fails.
Recall that h says that our (fully formalized) argument Δ is valid and b
says that Δ is not valid. Recall that Δ is an instance of excluded middle
Φ_:Φ, with no premises. There are two places to look here, but both
deliver the same range of verdicts.
We can ask first about the content of the argument Δ. Do h and b mean
the same thing by the disjunction “_” and by negation “:”? We thus
broach the longstanding question of whether the classicist and the intu-
itionist (or, indeed, advocates of any rival logics) are talking past
each other.
Michael Dummett (1991a: 17) argues that the “disagreement” is merely
verbal:
The intuitionists held, and continue to hold, that certain methods of
reasoning actually employed by classical mathematicians in proving the-
orems are invalid: the premisses do not justify the conclusion. The imme-
diate effect of a challenge to fundamental accustomed modes of reasoning is
perplexity: on what basis can we argue the matter, if we are not in
agreement about what constitutes a valid argument? In any case how can
such a basic principle of rational thought be rationally put in doubt?
The affront to which the challenge gives rise is quickly allayed by a resolve
to take no notice. The challenger must mean something different by the
logical constants; so he is not really challenging the laws that we have always
accepted and may therefore continue to accept.

Dummett goes on to argue that the classicist has no coherent meaning he

can assign to the connectives, but we can set that aside here (as inconsistent
with the foregoing folk-relativism).
From a very different perspective, W. V. O. Quine (1986: 81) also holds
that the various connectives change their content in the different logical
theories. Concerning the debate over paraconsistent logics, he wrote:
My view of this dialogue is that neither party knows what he is talking
about. They think they are talking about negation, “”, “not”; but surely
the notation ceased to be recognizable as negation when they took to
regarding some conjunctions in the form “p.p” as true, and stopped
 Pluralism, relativism, and objectivity 69
regarding such sentences as implying all others. Here, evidently, is the
deviant logician’s predicament: when he tries to deny the doctrine he only
changes the subject.

And Rudolf Carnap (1934: §17):

In logic, there are no morals. Everyone is at liberty to build his own logic, i.e.
his own form of language, as he wishes. All that is required of him is that, if
he wishes to discuss it, he must state his methods clearly, and give syntac-
tical rules instead of philosophical arguments.

Again, the key idea is that each logic is tied to a specific language.
Presumably, the meaning of the logical terms differs in the different
languages.
So the Dummett–Quine–Carnap perspective has it that we have a kind
of indexical contextualism here. The logical terms themselves have differ-
ent contents for our characters h and b. Using a subscript-C to indicate a
classical connective and a subscript-I for the corresponding intuitionistic
connective, we have that h holds that Φ_C:CΦ is valid, while b holds that
Φ_I:IΦ is invalid. This is the same sort of situation as with Jack and Jill
and the salt. There is no disagreement between h and b unless it be over
whether the other has a coherent meaning at all. If they are sufficiently
open-minded, h and b might agree that Φ_C:CΦ is valid and that Φ_I:IΦ
is invalid. So we do not have a failure of cognitive command.
The Dummett–Quine–Carnap perspective is not shared by all. Beall
and Restall (2006), for example, insist that their “pluralism” concerns
the notion of validity for a single language, with a single batch of
logical terms. So there is not, for example, a separate “_C” and “_I”.
There is just “_”. Restall (2002: 432) puts the difference with Dummett–
Quine–Carnap well:
If accepting different logics commits one to accepting different languages
for those logics, then my pluralism is primarily one of languages (which
come with their logics in tow) instead of logics. To put it graphically, as a
pluralist, I wish to say that
A, :A ‘C B, but A, :A⊬R B
A and :A together, classically entail B, but A and :A together do not
relevantly entail B. On the other hand, Carnap wishes to say that
A, :C A ‘ B, but A, :R A⊬B
A together with its classical negation entails B, but A together with its
relevant negation need not entail B.
70 Stewart Shapiro
So (Beall and) Restall reject an indexical contextualism concerning the
connectives (and quantifiers). Either there is no folk-relativism at all for the
connectives – each has a single, uniform content – or we have a non-
indexical contextualism or an assessment-sensitive view.
Recall that h says that Δ is valid and b says that Δ is invalid. On the
option considered now, championed by Beall and Restall, we have that h
and b mean the same thing by Δ. What about “valid”? Does that have the
same content in the two pronouncements?
Recall Beall and Restall’s (2006: 29) “Generalised Tarski Thesis”:
An argument is validx if and only if, in every casex in which the premises are
true, so is the conclusion.

I presume that Beall and Restall did not intend to make a claim about the
semantics of an established term of philosophical English. However, the
presence of the subscript x in the statement of the thesis might indicate that
the word “valid” has a sort of elided constituent, a slot where a logic can be
filled in. This suggests a sort of indexical contextualism about the word “valid”.
The same idea is suggested by the use of subscripts in the above passage from
Restall [2002], when he is using his own voice. He says that, for him:
A, :A ‘ C B, but A, :A ⊬R B:
So the technical term “‘” seems to have an elided constituent, and that
suggests a kind of contextualism.
So, on the Beall and Restall view – as on the opposing Dummett–
Quine–Carnap view – our logicians h and b do not have a genuine
disagreement. They are in the analogous situation as Jack and Jill with
the salt. Beall and Restall insist that h and b give the same content to the
argument Δ, but not to “valid”. For h, it is “classically valid”, “‘C”, and for
b it is “intuitionistically valid”, “‘I”. So, once again, we do not have a
failure of cognitive command.
To get cognitive command to fail, we have to assume that our logicians
h and b assign the same content to the terms in the argument Δ and we
have to assume that they assign the same content to the word “valid”.
Given that Δ has the same content, “valid” must be folk-relative (since
both h and b are correct). The options for that term are thus non-indexical
contextualism and assessment-sensitive relativism. I do not know of
anyone who explicitly defends that combination of views, and I won’t
consider how plausible it is (but see Shapiro 2014).
To summarize and conclude, Wright’s criterion of epistemic constraint
concerns the possibility of unknowable truths. Given the present
 Pluralism, relativism, and objectivity 71
folk-relativism, statements of validity do not get truth-values unless one
somehow indicates a particular logic. If a particular logic is so indicated,
then it depends on how much idealization goes into the notion of
“knowable”.
If we fix a particular logic, then either cognitive command holds
trivially, or, at worst, the question is reduced to one concerning math-
ematics which is, I would think, almost a paradigm case of objectivity. If
we do not fix a particular logic, and consider statements of validity
simpliciter, then the question of cognitive command depends on some
delicate, and controversial semantic theses concerning both the logical
terminology and the word “valid”.
Prima facie, it might seem strange that matters of cognitive command,
and indirectly, matters of objectivity, should turn on semantics. After all,
we are concerned with validity and not with the meanings of words, like
“or”, “not”, and, indeed “valid”. However, the notion of cognitive com-
mand depends on the notion of disagreement and, as we saw, that does
turn on notions of meaning.
Recall the Kant–Quine thesis articulated above, that there is no way to
sharply separate the “human” and the “world” contributions to our theor-
izing (perhaps with some emphasis on “sharply”). So we might expect
some tough, borderline cases of objectivity. Add to the mix some widely
held, but controversial views that meaning is not always determinate,
involving open-texture, and the like (e.g., Waismann 1945, Quine 1960,
Wilson 2006). Then perhaps the connection between objectivity and
semantics is not so surprising.
 chapter 4

Logic, mathematics, and conceptual structuralism

Solomon Feferman

1. The nature and role of logic in mathematics:

three perspectives
Logic is integral to mathematics and, to the extent that that is the case, a
philosophy of logic should be integral to a philosophy of mathematics. In
this, as you shall see, I am guided throughout by the simple view that what
logic is to provide is all those forms of reasoning that lead invariably from
truths to truths. The problematic part of this is how we take the notion of
truth to be given. My concerns here are almost entirely with the nature and
role of logic in mathematics. In order to examine that we need to consider
three perspectives: that of the working mathematician, that of the math-
ematical logician, and that of the philosopher of mathematics.
The aim of the mathematician working in the mainstream is to establish
truths about mathematical concepts by means of proofs as the principal
instrument. We have to look to practice to see what is accepted as a
mathematical concept and what is accepted as a proof; neither is deter-
mined formally. As to concepts, among specific ones the integer and real
number systems are taken for granted, and among general ones, notions of
finite and infinite sequence, set and function are ubiquitous; all else is
successively explained in terms of basic ones such as these. As to proofs,
even though current standards of rigor require closely reasoned arguments,
most mathematicians make no explicit reference to the role of logic in
them, and few of them have studied logic in any systematic way. When
mathematicians consider axioms, instead it is for specific kinds of struc-
tures: groups, rings, fields, linear spaces, topological spaces, metric spaces,
Hilbert spaces, categories, etc., etc. Principles of a foundational character
are rarely mentioned, if at all, except on occasion for proof by
contradiction and proof by induction. The least upper bound principle
on bounded sequences or sets of real numbers is routinely applied without
mention. Some notice is paid to applications of the Axiom of Choice. To a

72
 Logic, mathematics, and conceptual structuralism 73
side of the mainstream are those mathematicians such as constructivists or
semi-constructivists who reject one or another of commonly accepted
principles, but even for them the developments are largely informal with
little explicit attention to logic. And, except for some far outliers, what
they do is still recognizable as mathematics to the mathematician in the
mainstream.
Turning now to the logicians’ perspective, one major aim is to model
mathematical practice – ranging from the local to the global – in order to
draw conclusions about its potentialities and limits. In this respect, then,
mathematical logicians have their own practice; here I shall sketch it and
only later take up the question how well it meets that aim. In brief:
Concepts are tied down within formal languages and proofs within formal
systems, while truth, be it for the mainstream or for the outliers, is
explained in semantic terms. Some familiar formal systems for the main-
stream are Peano Arithmetic (PA), Second-Order Arithmetic (PA2), and
Zermelo–Fraenkel set theory (ZF); Heyting Arithmetic (HA) is an
example of a formal system for the margin. In their intended or “standard”
interpretations, PA and HA deal specifically with the natural numbers, PA2
deals with the natural numbers and arbitrary sets of natural numbers, while
ZF deals with the sets in the cumulative hierarchy. Considering syntax
only, in each case the well-formed formulas of each of these systems are
generated from its atomic formulas (corresponding to the basic concepts
involved) by closing under some or all of the “logical” operations of
negation, conjunction, disjunction, implication, universal and existential
quantification.
The case of PA2 requires an aside; in that system the quantifiers are
applied to both the first-order and second-order variables. But we must be
careful to distinguish the logic of quantification over the second-order
variables as it is applied formally within PA2 from its role in second-order
logic under the so-called standard interpretation. In order to distinguish
systematically between the two, I shall refer to the former as syntactic or
formal second-order logic and the latter as semantic or interpreted second-order
logic. In its pure form over any domain for the first-order variables,
semantic second-order logic takes the domain of the second-order variables
to be the supposed totality of arbitrary subsets of that domain; in its
applied form, the domain of first-order variables has some specified inter-
pretation. As an applied second-order formal system, PA2 may equally well
be considered to be a two-sorted first-order theory; the only thing that
acknowledges its intended second-order interpretation is the inclusion of
the so-called Comprehension Axiom Scheme: that consists of all formulas
74 Solomon Feferman
of the form 9X8x[x 2 X $ A(x,. . .)] where A is an arbitrary formula of the
language of PA2 in which ‘X ’ does not occur as a free variable. Construing
things in that way, the formal logic of all of the above-mentioned systems
may be taken to be first-order.
Now, it is a remarkable fact that all the formal systems that have been set
up to model mathematical practice are in effect based on first-order logic,
more specifically its classical system for mainstream mathematics and its
intuitionistic system for constructive mathematics. (While there are formal
systems that have been proposed involving extensions of first-order logic
by, for example, modal operators, the purpose of such has been philosoph-
ical. These operators are not used by mathematicians as basic or defined
mathematical concepts or to reason about them.) One can say more about
why this is so than that it happens to be so; that is addressed below.
The third perspective to consider on the nature and role of logic in
mathematics is that of the philosopher of mathematics. Here there are
a multitude of positions to consider; the principal ones are logicism (and
neo-logicism), “platonic” realism, constructivism, formalism, finitism, pre-
dicativism, naturalism, and structuralism.1 Roughly speaking, in all of these
except for constructivism, finitism, and formalism, classical first-order logic
is either implicitly taken for granted or explicitly accepted. In constructivism
(of the three exceptions) the logic is intuitionistic, i.e. it differs from the
classical one by the exclusion of the Law of Excluded Middle (LEM).
According to formalism, any logic may be chosen for a formal system. In
finitism, the logic is restricted to quantifier-free formulas for decidable
predicates; hence it is a fragment of both classical and intuitionistic logic.
At the other extreme, classical second-order logic is accepted in set-theoretic
realism, and that underlies both scientific and mathematical naturalism; it is
also embraced in in re structuralism. Modal structuralism, on the other
hand, expands that via modal logic. The accord with mathematical practice
is perhaps greatest with mathematical naturalism, which simply takes prac-
tice to be the given to which philosophical methodology must respond. But
the structuralist philosophies take the most prominent conceptual feature of
modern mathematics as their point of departure.

2. Conceptual structuralism
This is an ontologically non-realist philosophy of mathematics that I have
long advanced; my main concern here is to elaborate the nature and role of
1
Most of these are surveyed in the excellent collection Shapiro (2005).
 Logic, mathematics, and conceptual structuralism 75
logic within it. I have summarized this philosophy in Feferman (2009) via
the following ten theses.2
1. The basic objects of mathematical thought exist only as mental
conceptions, though the source of these conceptions lies in everyday
experience in manifold ways, in the processes of counting, ordering,
matching, combining, separating, and locating in space and time.
2. Theoretical mathematics has its source in the recognition that these
processes are independent of the materials or objects to which they
are applied and that they are potentially endlessly repeatable.
3. The basic conceptions of mathematics are of certain kinds of rela-
tively simple ideal-world pictures that are not of objects in isolation
but of structures, i.e. coherently conceived groups of objects inter-
connected by a few simple relations and operations. They are com-
municated and understood prior to any axiomatics, indeed prior to
any systematic logical development.
4. Some significant features of these structures are elicited directly from
the world-pictures that describe them, while other features may be
less certain. Mathematics needs little to get started and, once started,
a little bit goes a long way.
5. Basic conceptions differ in their degree of clarity or definiteness. One
may speak of what is true in a given conception, but that notion of
truth may be partial. Truth in full is applicable only to completely
definite conceptions.
6. What is clear in a given conception is time dependent, both for the
individual and historically.
7. Pure (theoretical) mathematics is a body of thought developed
systematically by successive refinement and reflective expansion of
basic structural conceptions.
8. The general ideas of order, succession, collection, relation, rule, and
operation are pre-mathematical; some implicit understanding of
them is necessary to the understanding of mathematics.
9. The general idea of property is pre-logical; some implicit understand-
ing of that and of the logical particles is also a prerequisite to the
understanding of mathematics. The reasoning of mathematics is in
principle logical, but in practice relies to a considerable extent on
various forms of intuition in order to arrive at understanding and
conviction.

2
This section is largely taken from Feferman (2009), with a slight rewording of theses 5 and 10.
76 Solomon Feferman
10. The objectivity of mathematics lies in its stability and coherence
under repeated communication, critical scrutiny, and expansion
by many individuals often working independently of each other.
Incoherent concepts, or ones that fail to withstand critical examin-
ation or lead to conflicting conclusions are eventually filtered out
from mathematics. The objectivity of mathematics is a special case of
intersubjective objectivity that is ubiquitous in social reality.

3. Two basic structural conceptions

These theses are illustrated in Feferman (2009) by the conception of the
structure of the positive integers on the one hand and by several concep-
tions of the continuum on the other. Since our main purpose here is to
elaborate the nature and role of logic in such structural conceptions, it is
easiest to review here what I wrote there, except that I shall limit myself to
the set-theoretical conception of the continuum in the latter case.
The most primitive mathematical conception is that of the positive
integer sequence as represented by the tallies: |, ||, |||, . . . From the
structural point of view, our conception is that of a structure (Nþ, 1,
Sc, <), where Nþ is generated from the initial unit 1 by closure under the
successor operation Sc, and m < n if m precedes n in the generation
procedure. Certain facts about this structure (if one formulates them
explicitly at all), are evident: that < is a total ordering of Nþ for which
1 is the least element, and that m < n implies Sc(m) < Sc(n). Reflecting on
a given structure may lead us to elaborate it by adjoining further relations
and operations and to expand basic principles accordingly. For example, in
the case of Nþ, thinking of concatenation of tallies immediately leads us to
the operation of addition, m þ n, and that leads us to m  n as “m added
to itself n times.” The basic properties of the þ and  operations such as
commutativity, associativity, distributivity, and cancellation are initially
recognized only implicitly. We may then go on to introduce more dis-
tinctively mathematical notions such as the relations of divisibility and
congruence and the property of being a prime number. In this language, a
wealth of interesting mathematical statements can already be formulated
and investigated as to their truth or falsity, for example, that there are
infinitely many twin prime numbers, that there are no odd perfect
numbers, Goldbach’s conjecture, and so on.
The conception of the structure (Nþ, 1, Sc, <, þ, ) is so intuitively
clear that (again implicitly, at least) there is no question in the minds of
mathematicians as to the definite meaning of such statements and the
 Logic, mathematics, and conceptual structuralism 77
assertion that they are true or false, independently of whether we can
establish them in one way or the other. (For example, it is an open
problem whether Goldbach’s conjecture is true.) In other words, realism
in truth values is accepted for statements about this structure, and the
application of classical logic in reasoning about such statements is auto-
matically legitimized. Despite the “subjective” source of the positive inte-
ger structure in the collective human understanding, it lies in the domain
of objective concepts and there is no reason to restrict oneself to intuitio-
nistic logic on subjectivist grounds. Further reflection on the structure of
positive integers with the aim to simplify calculations and algebraic oper-
ations and laws leads directly to its extension to the structure of natural
numbers (N, 0, Sc, <, þ, ), and then the usual structures for the integers
Z and the rational numbers Q. The latter are relatively refined conceptions,
not basic ones, but we are no less clear in our dealings with them than for
the basic conceptions of Nþ.
At a further stage of reflection we may recognize the least number
principle for the natural numbers, namely if P(n) is any well-defined
property of members of N and there is some n such that P(n) holds then
there is a least such n. More advanced reflection leads to general principles
of proof by induction and definition by recursion on N. Furthermore, the
general scheme of induction,
Pð0Þ ^ 8n½PðnÞ ! PðScðnÞÞ ! 8nPðnÞ,
is taken to be open-ended in the sense that it is accepted for any definite
property P of natural numbers that one meets in the process of doing
mathematics, no matter what the subject matter and what the notions used
in the formulation of P. The question – What is a definite property? –
requires in each instance the mathematician’s judgment. For example, the
property, “n is an odd perfect number,” is definite, while “n is a feasibly
computable number” is not, nor is “n is the number of grains of sand
in a heap.”
Turning now to the continuum, in Feferman (2009) I isolated several
conceptions of it ranging from the straight line in Euclidean geometry
through the system of real numbers to the set of all subsets of the natural
numbers. The reason that these are all commonly referred to as the
continuum is that they have the same cardinal number; however, that
ignores essential conceptual differences. For our purposes here, it is suffi-
cient to concentrate on the last of these concepts. The general idea of set or
collection of objects is of course ancient, but it only emerged as an object
of mathematical study at the hands of Georg Cantor in the 1870s. Given
78 Solomon Feferman
the idea of an arbitrary set X of elements of any given set D, considered
independently of how membership in X may be defined, we write S(D) for
the conception of the totality of all subsets X of D. Then the continuum in
the set-theoretical sense is simply that of the set S(N) of all subsets of N.
This may be regarded as a two-sorted structure, (N, S(N), 2), where 2 is
the relation of membership of natural numbers to sets of natural numbers.
Two principles are evident for this conception, using letters ‘X ’, ‘Y ’ to
range over S(N) and ‘n’ to range over N.
I. Extensionality 8X 8 Y [8n(n 2 X $ n 2Y ) ! X = Y ]
II. Comprehension For any definite property P(n) of members of N,

9 X 8n½n 2 X $ PðnÞ:
What is problematic here for conceptual structuralism is the meaning of
“all” in the description of S(N) as comprising all subsets of N. According to
the usual set-theoretical view, S(N) is a definite totality, so that quantifica-
tion over it is well-determined and may be used to express definite
properties P. But again that requires on the face of it a realist ontology
and in that respect goes beyond conceptual structuralism. So if we do not
subscribe to that, we may want to treat S(N) as indefinite in the sense that
it is open-ended. Of course this is not to deny that we recognize many
properties P as definite such as – to begin with – all those given by first-
order formulas in the language of the structure (N, 0, Sc, <, þ, ) (i.e.
those that are ordinarily referred to as the arithmetical properties); thence
any sets defined by such properties are recognized to belong to S(N).
Incidentally, even from this perspective one can establish categoricity of
the Extensionality and Comprehension principles for the structure
(N, S(N), 2) relative to N in a straightforward way as follows. Suppose
given another structure (N, S 0 (N), 20 ), satisfying the principles I and II,
using set variables ‘X 0 ’ and ‘Y 0 ’ ranging over S 0 (N). Given an X in S(N), let
P(n) be the definite property, n 2 X. Using Comprehension for the
structure (N, S 0 (N), 20 ), one obtains existence of an X 0 such that for all
n in N, n 2 X iff n 2 X 0 ; then X 0 is unique by Extensionality. This gives a
1–1 map of S(N) into S 0 (N) preserving N and the membership relation; it is
seen to be an onto map by reversing the argument. This is to be compared
with the standard set-theoretical view of categoricity results as exemplified,
for example, in Shapiro (1997) and Isaacson (2011). According to that view,
the subject matter of mathematics is structures, and the mère structures of
mathematics such as the natural numbers, the continuum (in one of its
various guises), and suitable initial segments of the cumulative hierarchy of
 Logic, mathematics, and conceptual structuralism 79
sets are characterized by axioms in full second-order logic; that is, any two
structures satisfying the same such axioms are isomorphic.3 On that
account, the proofs of categoricity in one way or another then
appeal prima facie to the presumed totality of arbitrary subsets of any
given set.4
Even if the definiteness of S(N) is open to question as above, we can
certainly conceive of a world in which S(N) is a definite totality and
quantification over it is well-determined; in that ideal world, one may
take for the property P in the above Comprehension Principle any formula
of full second-order logic over the language of arithmetic. Then a number
of theorems can be drawn as consequences in the corresponding system
PA2, including purely arithmetical theorems. Since the truth definition for
arithmetic can be expressed within PA2 and transfinite induction can be
proved in it for very large recursive well-orderings, PA2 goes in strength far
beyond PA even when that is enlarged by the successive adjunction of
consistency statements transfinitely iterated over such well-orderings.
What confidence are we to have in the resulting purely arithmetical
theorems? There is hardly any reason to doubt the consistency of PA2
itself, even though by Gödel’s second incompleteness theorem, we cannot
prove it by means that can be reduced to PA2. Indeed, the ideal world
picture of (N, S(N), 2) that we have been countenancing would surely lead
us to say more, since in it the natural numbers are taken in their standard
conception. On this account, any arithmetical statement that we can prove
in PA2 ought simply to be accepted as true. But given that the assumption
of S(N) as a definite totality is a purely hypothetical and philosophically
problematic one, the best we can rightly say is that in that picture,
everything proved of the natural numbers is true.

3
Those who subscribe to this set-theoretical view of the categoricity results may differ on whether
the existence of the structures in question follows from their uniqueness up to isomorphism.
Shapiro (1997), for example, is careful to note repeatedly that it does not, while Isaacson (2011)
apparently asserts that it does (cf., e.g., Isaacson 2011, p. 3). In any case, it is of course not a logical
consequence.
4
In general, proofs of categoricity within formal systems of second-order logic can be analyzed to see
just what parts of the usual impredicative comprehension axiom scheme are needed for them. In the
case of the natural number structure, however, it may be shown that there is no essential dependence
at all, in contrast to standard proofs. Namely, Simpson and Yokoyama (2012) demonstrate the
categoricity of the natural numbers (as axiomatized with the induction axiom in second-order form)
within the very weak subsystem WKL0 of PA2 that is known to be conservative over PRA (Primitive
Recursive Arithmetic). By comparison, it is sketched in Feferman (2013) how to establish categoricity
of the natural numbers in its open-ended schematic formulation in a simpler way that is also
conservative over PRA. For an informal discussion of the categoricity of initial segments of the
cumulative hierarchy of sets in the spirit of open-ended axiom systems, see D. Martin (2001, sec. 3).
80 Solomon Feferman
Incidentally, all of this and more comes into question when we move
one type level up to the structure (N, S(N), S(S(N)), 21, 22) in which
Cantor’s continuum hypothesis may be formulated. A more extensive
discussion of the conception of that structure and the question of its
definiteness in connection with the continuum problem is given in
Feferman (2011). We shall also see below how taking N and S(N) to be
definite but S(S(N)) to be open-ended can be treated in suitable formal
systems.

4. Where and why classical first-order logic?

Logic, as I affirmed at the outset, is supposed to provide us with all those
forms of reasoning that lead invariably from truths to truths, i.e. it is given
by an essential combination of inferential and semantical notions. But
from the point of view of conceptual structuralism, the classical notion of
truth in a structure need not be applicable unless we are dealing with a
conception (such as that of the structure of natural numbers) for which the
basic domains are definite totalities and the basic notions are definite
operations, predicates, and relations. It is clear that at least the classical
first-order predicate calculus should be admitted both on semantical and
inferential grounds, since we have Gödel’s completeness theorem to pro-
vide us with a complete inferential system. But why not more? For example,
model-theorists have introduced generalized quantifiers such as the cardin-
ality quantifiers (Qκx)P(x) expressing that there are at least κ individuals x
satisfying the property P, where κ is any infinite cardinal; one could
certainly consider adjoining those to the first-order formalism. A much
more general class of quantifiers defined by set-theoretical means was
introduced by Lindström (1966); each of those can be used to extend
first-order logic with a model-theoretic semantics for arbitrary first-order
domains. But for which such extensions do we have a completeness
theorem like that of Gödel’s for first-order logic? It is well known that
no such theorem is possible for the quantifier (Qωx)P(x) which expresses
that there are infinitely many x such that P(x). For, using that quantifier
and thence its dual (“there are just finitely many x such that P(x)”) we can
characterize the structure of natural numbers up to isomorphism, so all the
truths of that structure are valid sentences in the logic. But the set of such
truths is not effectively enumerable, indeed far from it, so it is not given by
an effectively specified formal system of reasoning.
Surprisingly, Keisler (1970) obtained a completeness theorem for the
quantifier (Qκx)P(x) when κ is any uncountable cardinal; as it happens,
 Logic, mathematics, and conceptual structuralism 81
that has the same set of valid formulas as for the case that κ is the first
uncountable cardinal. In view of the leap over the case κ = ω, one may
suspect that the requirement that the set of valid formulas be given by
some effective set of axioms and rules of inference is not sufficient to
express completeness in the usual intended sense. We need to say some-
thing more about how such axioms and rules of inference ought specific-
ally to be complete for a given quantifier. The key is given by Gentzen’s
(1935) system of natural deduction NK (or sequent calculus LK) where
each connective and quantifier in the classical first-order predicate calculus
is specified by Introduction and Elimination rules for that operation only.
Moreover, for each pair of such rules, any two connectives or quantifiers
satisfying them are equivalent, i.e. they implicitly determine the operator
in question. So a strengthened condition on a proposed addition by a
generalized quantifier Q to our first-order language is that it be given by
axioms and rules of inference for which there is at most one operator
satisfying them. That was the proposal of Zucker (1978) in which he gave a
theorem to the effect that any such quantifier is definable in the first-order
predicate calculus. In particular, that would apply to the Lindström
quantifiers. However, there were some defects in Zucker’s statement of
his theorem and its proof; I have given a corrected version of both in
Feferman (to appear). To summarize: we have fully satisfactory semantic
and inferential criteria for a logic to deal with structures whose domains are
first-order and that are completely definite in the sense described above,
and these limit us to the standard first-order classical logic.
Let us turn now to conceptions of structures with second-order or
higher-order domains, such as (N, S(N), 2, . . .) where the ellipsis indicates
that this augments an arithmetical structure on N such as (N, 0, Sc, <, þ, ).
Again, if S(N) is considered as a definite totality, the classical notion of
truth is applicable and the semantics of second-order logic must be
accepted. But as is well known there is no complete inferential system
that accompanies that, since again the arithmetical structure is categorically
axiomatized in this semantics and in consequence the set of its truths is not
effectively enumerable. In any case, as I have argued above, S(N) ought not
to be considered as a definite totality; to claim otherwise, is to accept the
problematic realist ontology of set theory. As Quine famously put it,
second-order logic is “set theory in sheep’s clothing.” Boolos (1975, 1984)
tried to get around this via a reduction of second-order logic to a “nomin-
alistic” system of plural quantification. This was incisively critiqued by
Resnik in his article “Second-order logic still wild”: “Boolos is involved in a
circle: he uses second-order quantification to explain English plural
82 Solomon Feferman
quantification and uses this, in turn, to explain second-order quantifica-
tion” (Resnik 1988, p. 83).
Though the Lindström quantifiers are restricted to apply to first-order
structures and thus bind only individual variables they may well be defined
using higher-order notions in an essential way, in particular those needed
for the cardinality quantifiers. Another example where the syntax is first-
order on the face of it but the semantics is decidedly second-order is IF
(“Independence Friendly”) logic, due to Hintikka (1996). This uses for-
mulas in whose prenex form the existentially quantified individual vari-
ables are declared to depend on a subset of the universally quantified
individual variables that precede it in the prefix list. Explanation of the
semantics of this requires the use of quantified function variables; over any
given first-order structure (D, . . .) those variables are interpreted to range
over functions of various arities with arguments and values in D. Indeed,
Väänänen (2001, p. 519) has proved that the general question of validity of
IF sentences is recursively isomorphic to that for validity in full second-
order logic. Thus, as with the Lindström quantifiers, the formal syntax can
be deceptive. See Feferman (2006) for an extended critique of IF logic.

5. Where and why intuitionistic first-order logic?

Now let us turn to the question which logic is appropriate to structural
conceptions that are taken to lack some aspect of definiteness. Offhand,
one might expect the answer in that case to be intuitionistic logic, but the
matter is more delicate. The problem is that there is not one clear-cut
semantics for it; among others that have been considered, one has the
so-called BHK interpretation, Kripke semantics, topological semantics,
sheaf models, etc., etc. Of these, the first is the most principled one with
respect to the basic ideas of constructivity; it is that that leads one directly to
intuitionistic logic but it does not determine it via a precise completeness
result. By contrast, as we shall see, not only does Kripke semantics take care
of the latter but it relates more closely to the question of dealing with
conceptions of structures involving possibly indefinite notions and
domains. For the details concerning both of these I refer to Troelstra and
van Dalen (1988), a comprehensive exposition of constructivism in math-
ematics that includes treatments of the great variety of semantics and proof
theory that have been developed for intuitionistic systems.
The BHK (Brouwer–Heyting–Kolmogoroff ) constructive explanation
of the connectives and quantifiers is described in Troelstra and van Dalen
(1988, p. 9). It uses the informal notions of construction and constructive
 Logic, mathematics, and conceptual structuralism 83
proof; for each form of compound statement C necessary and sufficient
conditions are provided on what it is for a construction to be a proof of
C, in terms of proofs of its immediate sub-statements. Namely, a proof
of A ^ B is a proof of A and a proof of B; a proof of A _ B is a proof of A or
a proof of B; a proof of A ! B is a construction that transforms any proof
of A into a proof of B; and a proof of :A is a construction that transforms
any proof of A into a proof of a contradiction ⊥, i.e. is a proof of A ! ⊥.
In the case of the quantifiers, where the variables range over a given
domain D, a proof of (8x)A(x) is a construction that transforms any d in
D into a proof of A(d); finally, a proof of (9x)A(x) is given by a d in D and a
proof of A(d). (D must be a constructively meaningful domain, so that it
makes sense to exhibit each individual element of D and for constructions
to be applicable to elements of D.)
A statement A of the first-order predicate calculus is constructively valid
according to the BHK interpretation if there is a proof of A, independently
of the interpretation of the domain D and the interpretation of the
predicate symbols of A in D. The axioms of intuitionistic logic in any of
its usual formulations are readily recognized to be constructively valid and
the rules of inference preserve constructive validity. But since there are no
precise notions of proof and construction at work here, we cannot state a
completeness result for the BHK interpretation. Instead, the literature uses
“weak counterexamples” to show why it is plausible on that account that a
given classically valid form of statement is not constructively valid. Thus,
for example, to show that A _ :A is not constructively valid as a general
principle one argues that otherwise one would have a general method for
obtaining for any given statement A, either a proof of A or a proof that
turns any hypothetical proof of A into a contradiction. But if we had such
a universal method, we could apply it to any particular statement A that
has not yet been settled, such as the twin prime conjecture, to determine
its truth or falsity. Similarly, the method of weak counterexamples is used
informally to argue against the constructive validity of many other such
schemes, for example ::A ! A, though the converse is recognized to be
valid.5
Let us turn now to Kripke semantics for the language of first-order
predicate logic (Troelstra and van Dalen 1988, Ch. 2.5–2.6). A Kripke
model is a quadruple (K, , D, v), where (i) (K, ) is a non-empty

5
Various methods of realizability, initially introduced by Kleene in 1945, can be used to give precise
independence results for such schemes, but are still not complete for intuitionistic logic. Cf. Troelstra
and van Dalen (1988, Ch. 4.4).
84 Solomon Feferman
partially ordered set, (ii) D is a function that assigns to each k in K a non-
empty set D(k) such that if k  k 0 then D(k)  D(k 0 ), and (iii) v is a
function into f0, 1g at each k in K, each n-ary relation symbol R in the
language and n-ary sequence of elements of D(k), such that if k  k 0 and
d1,. . .,dn 2 D(k) and v(k, R(d1,. . .,dn)) = 1 then v(k 0 , R(d1,. . .,dn)) = 1. One
motivating idea for this is that the elements of K represent stages of
knowledge, and that k  k 0 holds if everything known in stage k is known
in stage k 0 . Also, v(k, R(d1,. . .,dn)) = 1 means that R(d1,. . .,dn) has been
recognized to be true at stage k; once recognized, it stays true. The domain
D(k) is the part of a potential domain that has been surveyed by stage k;
the domains may increase indefinitely as k increases or may well bifurcate
in a branching investigation so that one cannot speak of a “final” domain
in that case.
The valuation function v is extended to a function v(k, A(d1,. . .,dn)) into
f0, 1g for each formula A(x1,. . .,xn) with n free variables and assignment
(d1,. . .,dn) to its variables in D(k); this is done in such a way that if k  k 0
and d1,. . .,dn 2 D(k) and v(k, A(d1,. . .,dn)) = 1 then v(k 0 , A(d1,. . .,dn)) = 1.
The clauses for conjunction, disjunction, and existential quantification are
just like those for ordinary satisfaction at k in D(k). The other clauses are
(ignoring parameters): v(k, A ! B) = 1 iff for all k 0  k, v(k 0 , A) = 1 implies
v(k 0 , B) = 1; v(k, ⊥) = 0; and v(k, 8x A(x)) = 1 iff for all k 0  k and d in
D(k), v(k 0 , A(d)) = 1. As above, we identify :A with A ! ⊥; thus v(k, :A) = 1
iff for all k 0  k, v(k 0 , A) = 0. We say that k forces A if v(k, A) = 1; i.e. A is
recognized to be true at stage k no matter what may turn out to be known
at later stages. A formula A(x1,. . .,xn) is said to be valid in a model (K, ,
D, v) if for every k in K and assignment (d1,. . .,dn) to its free variables in
D(k), v(k, A(d1,. . .,dn)) = 1. Then the completeness theorem for this
semantics is that a formula A is valid in all Kripke models iff it is provable
in the first-order intuitionistic predicate calculus. We shall see in the next
section how Kripke models can be generalized to take into account
differences as to definiteness of basic relations and domains.
Satisfying as this completeness theorem may be, there remains the
question whether one might not add connectives or quantifiers to those
of intuitionistic logic while retaining some form of its semantics. Though
intuitionistic logic is part of classical logic, the semantical and inferential
criterion above for classical logic doesn’t apply because of the differences in
the semantical notions. But just as for the classical case, on the inferential
side each of the connectives and quantifiers of the intuitionistic first-order
predicate calculus is uniquely identified via Introduction and Elimination
rules in Gentzen’s natural deduction system NJ. Even more, Gentzen first
 Logic, mathematics, and conceptual structuralism 85
formulated the idea that the meaning of each of the above operations is
given by its characteristic inferences. Actually, Gentzen claimed more: he
wrote that “the [Introduction rules] represent, as it were, the ‘definitions’
of the symbols concerned” (Gentzen 1969, p. 80). Prawitz supported this
by means of his Inversion Principle (Prawitz 1965, p. 33): namely, it follows
from the normalization theorem for NJ that each Elimination rule for a
given operation can be recovered from the appropriate one of its Introduc-
tion rules when that is the last step in a normal derivation. Without
subscribing at all to this proposed reduction of semantics to inferential
roles, we may ask whether any further operators may be added via suitable
Introduction rules. The answer to that in the negative was provided by the
work of Zucker and Tragesser (1978) in terms of the adequacy of what they
call inferential logic, i.e. of the logic of operators that can simply be marked
out by Introduction rules. As they show, every such operator is defined in
terms of the connectives and quantifiers of the intuitionistic first-order
predicate calculus. To be more precise, this is shown for Introduction rules
in the usual sense in the case of possible propositional operators, while in
the general case of possible operators on propositions and predicates – now
in accord with the BHK interpretation – “proof ” parameters and con-
structions on them are incorporated in the Introduction rules, but those
are eventually suppressed.6

6. Semi-intuitionism: the logic of partially

open-ended structures
An immediate generalization of Kripke structures is to allow many-sorted
domains, possibly infinite in number. Let I be a collection of sorts. Then
the definition of Kripke structure is modified to have each of K, , and
D indexed by I, and the valuation function modified to accord with the
different sorts. Thus we deal with n-tuples k = (k1,. . .,kn) where km is of
specified sort im; the  relation then holds between such n-tuples if it
holds term-wise. Of course the basic predicates come with specified arities
to show what sorts of objects they relate, and the variables in the first-order
language over these predicates are always of a specified sort. Then the
definition of the valuation function on arbitrary formulas for a many-
sorted structure (K, I, , D, v) proceeds in the same way as above. Now an

6
Incidentally, as Zucker and Tragesser show (p. 506), not every propositional operator given by
simple Introduction rules has an associated Elimination rule; a counterexample is provided by
(A ! B) _ C.
86 Solomon Feferman
n-ary relation R may be considered to be definite if v(k, R(d1,. . .,dn)) = v(k0 ,
R(d1,. . .,dn)) whenever k  k 0 . A domain Di is definite if Di(k) = Di(k 0 )
for all k and k 0 in Ki, otherwise indefinite or open-ended. While the
formulas valid in the structure obey intuitionistic logic in general, one
may apply classical logic systematically to formulas involving definite
relations as long as the quantified variables involved range only over
definite domains.
This is illustrated by reasoning about the ordinary two-sorted struc-
ture (N, S(N), 2, . . .) where (N, . . .) is conceived of as definite with
definite relations, while S(N) is conceived of as open-ended. To treat
this as a two-sorted Kripke structure, take I = f0, 1g where N is of sort
0 and S(N) is of sort 1. We may as well take K0 to consist of a single
element, while K1 could be indexed by all collections k of subsets of
N, ordered by inclusion. Now the membership relation is definite
because sets are taken to be definite objects, i.e. if X is in both the
collections k and k 0 then n 2 X holds in the same way whether
evaluated in k or in k 0 . So classical logic applies to all formulas A that
contain no bound set variables, though they may contain free set
variables, i.e. A is what is usually called a predicative formula. But
when dealing with formulas in general, only intuitionistic logic is
justified on this picture. This leads us to the consideration of semi-
intuitionistic (or semi-constructive) theories in general, i.e. theories in
which the basic underlying logic is intuitionistic, but classical logic is
taken to apply to a class of formulas distinguished by containing
definite predicates and quantified variables ranging over definite
domains. A number of such theories have been treated in the paper
Feferman (2010), corresponding to different structural notions in which
certain domains are taken to be definite and others indefinite. They fall
into three basic groups: (i) predicative theories, (ii) theories of countable
(tree) ordinals, and (iii) theories of sets. The general pattern is that in
each case one has a semi-intuitionistic version of a corresponding
classical system, and they are shown to be proof-theoretically equivalent
and to coincide on the classical part. Moreover, the same holds when
the semi-intuitionistic system is augmented by various principles such
as the Axiom of Choice (AC) that would make the corresponding
classical system much stronger. It is not possible here to explain the
results in adequate detail, so only some of the ideas behind the formu-
lations of the systems involved are sketched. The reader who prefers to
avoid even the technicalities that remain can easily skim (or even skip)
the rest of this section.
 Logic, mathematics, and conceptual structuralism 87

6.1 Semi-intuitionistic predicative theories

Here the language of arithmetic is extended by variables for function(al)s
in all finite types; following Gödel (1958, 1972) in his so-called Dialectica
interpretation, we also add primitive recursive functionals in all finite
types. In many-sorted intuitionistic logic, the system obtained is denoted
HAω. In the process of obtaining reduction to a quantifier-free system,
Gödel showed that this system is of the same strength as Peano Arithmetic,
PA; in fact the same holds for HAω þ AC. Now the latter is turned into a
semi-intuitionistic system by adding the Law of Excluded Middle for all
arithmetical formulas. For the proof-theoretical work on that, it proves to
be more convenient to add the least-number operator μ and an axiom (μ)
that says that when the operator is applied to a function f : N ! N for
which there exists an n with f(n) = 0, it yields the least such n. Under this
axiom, all arithmetical formulas become equivalent to quantifier-free (QF)
formulas, for which the LEM then holds. Thus one is led to consider
HAω þ AC þ (μ), which turns out to be proof-theoretically equivalent to
PAω þ QF-AC þ (μ), and both are equivalent to ramified analysis through
all ordinals less than Cantor’s ordinal ε0. If one adds the Bar Rule for
arithmetical orderings in both the semi-intuitionistic and the classical
systems, we obtain systems of proof-theoretical strength full predicative
analysis, i.e. ramified analysis up to the least impredicative ordinal Γ0. (The
Bar Rule on an ordering allows us to infer transfinite induction w.r.t.
arbitrary formulas from well-foundedness of the ordering.) On the other
hand, if in the basic system we restrict the primitive recursive functionals
to those with values in N and restrict induction to QF formulas, we obtain
a semi-intuitionistic system Res-HAω þ AC þ (μ) that turns out to be of
exactly PA in strength.

6.2 Semi-intuitionistic theories of countable tree ordinals

By countable tree ordinals one means the members of the open-ended
collection O of countably branching well-founded trees. Add a sort for
the members of O to the preceding systems; extend the higher type
variables accordingly; add the operator of supremum that joins a sequence
of trees f: N ! O into a single tree sup( f ) in O; add the inverse operator
that takes each sup( f ) in O and n in N and produces f (n); and, finally, add
operators for transfinite recursion on O. The resulting system is denoted
SOO in intuitionistic logic and COO in classical logic; then SOO þ (μ) is a
semi-intuitionistic system intermediate between these two. The main
88 Solomon Feferman
result in this case is that the following are of the same proof-theoretical
strength: SOO þ AC þ (μ), COO þ QF-AC þ (μ), and ID1, the theory of
arbitrary arithmetical inductive definitions. It is known that the latter
has the same proof-theoretical strength in intuitionistic logic as in
classical logic.

6.3 Semi-intuitionistic theories of sets

We turn finally to the picture of the cumulative hierarchy structure, the
standard classical view of which leads us to the system ZFC, i.e. ZF þ AC.
However, if we identify definite totalities with sets then by Russell’s
paradox, the “universe V of all sets” must be considered to be an open-
ended indefinite totality if we are to avoid contradiction. But in the
Separation Axiom scheme for ZF, 8a9b8x[x 2 b $ x 2 a ^ A(x)], one
allows the formula A to contain bound variables that range without
restriction over V, and hence in general do not represent definite proper-
ties; the same criticism applies to the formulas A(x, y) in the Replacement
Axiom scheme. By a Δ0 formula is meant one in which all quantified
variables are restricted, i.e. take the form 8y(y 2 x ! . . .) or 9y(y 2 x
^ . . .), written respectively (8y 2 x)(. . .) and (9y 2 x)(. . .). The system KP
of Kripke–Platek set theory in classical logic has, like ZF, the axioms of
extensionality, ordered pair, union, infinity, and the scheme of transfinite
induction on the membership relation. In place of the Separation Axiom
scheme it takes Δ0-Separation, i.e. the Separation Axiom scheme restricted
to Δ0 formulas. And in place of the Replacement Axiom scheme, it takes
what is called Δ0-Collection, i.e. the scheme that for each Δ0 formula
A, (8x 2 a)9yA(x, y) ! 9b(8x 2 a)(9y 2 b)A(x, y). This implies the
Replacement Axiom scheme for Δ0 formulas. It is known that the system
KP is of the same strength as ID1.
The system IKP is taken to be the same as KP but restricted to
intuitionistic logic. It turns out that we can strengthen it considerably by
adding a bounded form ACS of the Axiom of Choice, namely (8x 2 a)9yA
(x, y) ! 9f [Fun( f ) ^ (8x 2 a)A(x, f (x))], where Fun( f ) expresses that
the set f is a function in the set-theoretical sense, and where now A is an
arbitrary formula of the language of set theory. Under the assumption ACS
we can infer Collection for arbitrary formulas and hence Replacement for
arbitrary formulas. Finally, since sets are considered to be definite totalities,
we obtain a semi-intuitionistic system from IKP by adjoining the law of
excluded middle for Δ0 formulas. The main result of Feferman (2010) is
that the semi-intuitionistic system IKP þ ACS þ Δ0-LEM is of the same
 Logic, mathematics, and conceptual structuralism 89
proof-theoretical strength as KP and hence of ID1 in its classical and
intuitionistic forms. Moreover, if we add the Power Set Axiom (Pow) we
obtain a system that is of strength between that of KP þ Pow and that of
KP þ Pow þ (V = L).7,8
It is natural in the context of semi-intuitionistic theories T to say that a
sentence A in the language of T is definite (relative to T) if T proves LEM
for A, i.e. A _ :A. A question in set theory that has caused considerable
discussion in recent years is whether Cantor’s continuum hypothesis CH is
a definite mathematical problem. One formulation of it is that every subset
of S(N) is either countable or in 1-1 correspondence with S(N). Of course,
that is definite in the theory IKP þ Pow þ Δ0-LEM, because quantifica-
tion over subsets of S(N) is bounded once we have existence of S(S(N))
[i.e., S(S(ω))] by the Power Set Axiom. That suggests – as I did in
Feferman (2011) – considering the weaker system T = IKP þ Pow(N)
þACS þ Δ0-LEM, where Pow(N) simply asserts the existence of S(N) as a
set. I conjectured there that CH is not definite relative to that system.9 Of
course, that would not show that CH is not a definite mathematical
problem, but it might be considered as an interesting bit of evidence in
support of that.

7. Conceptual structuralism and mathematical practice

One criterion for a philosophy of mathematics that is often heard is that it
should accord with mathematical practice. It’s very hard to know just what
that means since there are so many dimensions along which practice can be
viewed. One particular interpretation of the criterion is that philosophers
have no business telling mathematicians what does or doesn’t exist. Fam-
ously, David Lewis wrote:
I’m moved to laughter at the thought of how presumptuous it would be to
reject mathematics for philosophical reasons. How would you like the job of
telling the mathematicians that they must change their ways, and abjure
countless errors, now that philosophy has discovered that there are no classes?
(Lewis 1991, p. 59)10

7
There is a considerable literature on semi-intuitionistic theories of sets including the power set
axiom going back to the early 1970s. See Feferman (2010, sec. 7.2) for references to the relevant
work of Poszgay, Tharp, Friedman, and Wolf.
8
Mathias (2001) proved that KP þ Pow þ (V = L) proves the consistency of KP þ Pow, so the usual
argument for the relative consistency of (V = L) doesn’t work.
9
Michael Rathjen (2014) has recently verified this conjecture.
10
Curiously, this quote is from Lewis’ book, Parts of Classes, which offers a revisionary theory of classes
that differs from the usual mathematical conception of such.
90 Solomon Feferman
But this is a caricature of what philosophy is after; philosophers take for
granted that mathematicians have settled problematic individual questions
of existence like zero, negative numbers, imaginary numbers, infinitesi-
mals, points at infinity, probability of subsets of [0, 1], etc., etc., using
purely mathematical criteria in the course of the development of their
subject. The existence of some of these has been established by reduction
to objects whose existence is unquestioned, some by qualified acceptance,
and some not at all. But what the philosopher is concerned with is, rather,
to explain in what metaphysical sense, if any, mathematical objects exist, in
a way that cannot even be discussed within ordinary mathematical par-
lance. Lewis could equally well have laughed at the idea that some general
principles accepted in the mathematical mainstream such as the Law of
Excluded Middle or the Axiom of Choice would be dismissed as false (or
unjustified) for philosophical reasons. But again, the use of truth in
ordinary mathematical parlance is deflationary and the reasons for
accepting such and such principles as true has either been made without
question or for mathematical reasons in the course of the development of
the subject. The philosopher, by contrast, is concerned to explain in what
sense the notion of truth is applicable to mathematical statements, in a way
that cannot be considered in ordinary mathematical parlance. Whether the
mathematician should pay attention to either of these aims of the philoso-
pher is another matter.
Conceptual structuralism addresses the question of existence and truth
in mathematics in a way that accords with both the historical development
of the subject and each individual’s intellectual development. It crucially
identifies mathematical concepts as being embedded in a social matrix that
has given rise, among other things, to social institutions and games; like
them, mathematics allows substantial intersubjective agreement, and like
them, its concepts are understood without assuming reification.11 What
makes mathematics unique compared to institutions and games is its
endless fecundity and remarkable elaboration of some basic numerical
and geometrical structural conceptions. To begin with, mathematical
objects exist only as conceived to be elements of such basic structures.
The direct apprehension of these leads one to speak of truth in a structure
in a way that may be accepted uncritically when the structure is such as the
integers but may be put into question when the conception of the structure
is less definite as in the case of the geometrical plane or the continuum, and

11
For an interesting social institutional account of mathematics see Cole (2013); this differs from
conceptual structuralism in some essential respects while agreeing with it in others.
 Logic, mathematics, and conceptual structuralism 91
should be put into question when it comes to the universe of sets. One
criticism of conceptual structuralism that has been made is that it’s not
clear/definite what mathematical concepts are clear/definite, and making
that a feature of the philosophy brings essentially subjective elements into
play.12 Actually, conceptual structuralism by itself, as presented in the
theses 1–10, takes no specific position in that respect and recognizes that
different judgments (such as mine) may be made. Once such are con-
sidered, however, logic has much to tell us in its role as an intermediary
between philosophy and mathematics. As shown in the preceding section,
one can obtain definitive results about formal models of different stand-
points as to what is definite and what is not. Moreover, the results can be
summarized as telling us that to a significant extent, the unlimited (de
facto) application of classical logic in mainstream mathematics – i.e., the
logic of definite concepts and totalities – may be justified on the basis of
a more refined mixed logic that is sensitive to distinctions that one might
adopt between what is definite and what is not.13 In other words, once
more they show that, at least to that extent, you can have your cake and
eat it too.
There are other dimensions of mathematical practice that reward meta-
mathematical study motivated by the philosophy of conceptual
structuralism. One, in particular, that I have emphasized over the years
is the open-ended nature of certain principles such as that of induction for
the integers and comprehension for sets. This accords with the fact that in
the development of mathematics what concepts are recognized to be
definite evolve with time. Thus one cannot fix in advance all applications
of these open-ended schematic principles by restriction to those instances
definable in one or another formal language, as is currently done in the
study of formal systems. This leads instead to the consideration of logical
models of practice from a novel point of view that yet is susceptible to
metamathematical study. One such is via the notion of the unfolding of
open-ended schematic axiom systems, that is used to tell us everything that
ought to be accepted if one has accepted given notions and principles.
Thus far, definitive results about the unfolding notion have been obtained
by Feferman and Strahm (2000, 2010) for schematic systems of non-finitist
and finitist arithmetic, resp., and by Buchholtz (2013) for arithmetical

12
In particular, this criticism has been voiced by Peter Koellner in his comments on Feferman (2011);
cf. http://logic.harvard.edu/EFI_Feferman_comments.pdf.
13
These kinds of logical results can also be used to throw substantive light on philosophical discussions
as to the problem of quantification over everything (or over all ordinals, or all sets) such as are found
in Rayo and Uzquiano (2006).
92 Solomon Feferman
inductive definitions. As initiated in Feferman (1996), I am optimistic that
it can be used to elaborate Gödel’s program for new axioms in set theory
and in particular to draw a sharper line between which such axioms ought
to be accepted on intrinsic grounds and those to be argued for on extrinsic
grounds.

AC K N O W L E D G E M E N T S
I would like to thank Gianluigi Bellin, Dagfinn Føllesdal, Peter Koellner,
Grigori Mints, Penelope Rush, Stewart Shapiro, and Johan van Benthem
for their helpful comments on a draft of this essay.
 chapter 5

A Second Philosophy of logic

Penelope Maddy

What’s hidden in my hand is either an ordinary dime or a foreign coin of a

type I’ve never seen. (I drew it blindfolded from a bin filled with just these
two types of objects.) It’s not a dime. (I can tell by the feel of it.) Then,
obviously, it must be a foreign coin! But what makes this so?
It’s common to take this query as standing in for more general questions
about logic – what makes logical inference reliable? what is the ground of
logical truth? – and common, also, to regard these questions as properly
philosophical, to be answered by appeal to distinctively philosophical
theories of abstracta, possible worlds, concepts, meanings, and the like.
What I’d like to do here is step back from this hard-won wisdom and try to
address the simple question afresh, without presumptions about what
constitutes ‘logic’ or even ‘philosophy’. The thought is to treat inquiries
about reliability of the coin inference and others like it as perfectly ordinary
questions, in search of perfectly ordinary answers, and to see where this
innocent approach may lead.
To clarify what I have in mind here, let me introduce an unassuming
inquirer called the Second Philosopher, interested in all aspects of the
world and our place in it.1 She begins her investigations with everyday
perceptions, gradually develops more sophisticated approaches to observa-
tion and experimentation that expand her understanding and sometimes
serve to correct her initial beliefs; eventually she begins to form and test
hypotheses, and to engage in mature theory-formation and confirmation;
along the way, she finds the need for, and pursues, first arithmetic and
geometry, then analysis and even pure mathematics;2 and in all this, she
often pauses to reflect on the methods she’s using, to assess their

1
The Second Philosopher is introduced in (Maddy 2007), and her views on logic detailed in Part III of
that book. The discussion here reworks and condenses the presentation there (see also (Maddy to
appear)).
2
For more on the Second Philosopher’s approach to mathematics, see (Maddy 2011).

93
94 Penelope Maddy
effectiveness and improve them as she goes. When I propose to treat the
question of the reliability of the coin inference as an ordinary question,
I have in mind to examine it from the Second Philosopher’s point of view.
She holds no prior convictions about the nature of the question; she sees it
simply as another of her straightforward questions about the world and her
investigations of it.
The first thing she’s likely to notice is that neither the reliability of the
coin inference nor the truth of the corresponding if–then statement3
depends on any details of the physical composition of the item in her
hand or the particular properties that characterize dimes as opposed to
other coins. She quickly discerns that what’s relevant is entirely independ-
ent of all but the most general structural features of the situation: an object
with one or the other of two properties that lacks one must have the other.
In her characteristic way, she goes on to systematize this observation – for
any object a and any properties, P and Q, if Qa-or-Pa and not-Qa, then
Pa – and from there to develop a broader theory of forms that yield such
highly general forms of truth and reliable inference. In this way, she’s led
to consider any situation that consists of objects that enjoy or fail to enjoy
various properties, that stand and don’t stand in various relations; she
explores conjunctions and disjunctions of these, and their failures as well;
she appreciates that one situation involving these objects and their inter-
relations can depend on another; and eventually, following Frege, she
happens on the notion that a property or relation can hold for at least
one object, or even universally – suppose she dubs this sort of thing a
‘formal structure’.4
Given her understanding of the real-world situations she’s out to
describe in these very general, formal terms, she sees no reason to suppose
that every object has precise boundaries – is this particular loose hair part
of the cat or not? – or that every property (or relation) must determinately
hold or fail to hold of each object (or objects) – is this growing tadpole now
a frog or not? She appreciates that borderline cases are common and fully
determinate properties (or relations) rare. Thinking along these lines, she’s
led to something like a Kleene or Lukakasiewicz three-valued system: for a
given object (or objects), a property (or relation) might hold, fail, or be
indeterminate; not-(. . .) obtains if (. . .) fails and is otherwise indetermin-
ate; (. . .)-and-(__) obtains if both (. . .) and (__) obtain, fails if one of
them fails, and is otherwise indeterminate; and so on through the obvious

3
I won’t distinguish between these, except in the vicinity of footnote 5.
4
In (Maddy 2007) and (Maddy to appear), this is called KF-structure, named for Kant and Frege.
 A Second Philosophy of logic 95
clauses for (. . .)-or-(__), (there is an x, . . .x. . .) and (for all x, . . .x. . .).
A formal structure of this sort validates many of the familiar inference
patterns – for example, the introduction and elimination rules for ‘not’,
‘and’, ‘or’, ‘for all’, and ‘there exists’; the DeMorgan equivalences; and the
distributive laws – but the gaps produce failures of the laws of excluded
middle and non-contradiction (if p is indeterminate, so are p-or-not-p and
not-( p-and-not-p).5 The subtleties of the Second Philosopher’s depend-
ency relation undercut many of the familiar equivalences: not-(the rose is
red)-or-2 þ 2 = 4, but 2 þ 2 doesn’t equal 4 because the rose is red.
Fortunately, modus ponens survives: when both (q depends on p) and
p obtain, q can’t fail or be indeterminate. Suppose the Second Philosopher
now codifies these features of her formal structures into a collection of
inference patterns; coining a new term, she calls this ‘rudimentary logic’
(though without any preconceptions about the term ‘logic’). She takes
herself to have shown that this rudimentary logic is satisfied in any
situation with formal structure.
This is a considerable advance, but it remains abstract: what’s been
shown is that rudimentary logic is reliable, assuming the presence of formal
structure. Common sense clearly suggests that our actual world does
contain objects with properties, standing in relations, with dependencies,
but the Second Philosopher has learned from experience that common
sense is fallible and she routinely subjects its deliverances to careful
scrutiny. What she finds in this case is, for example, that the region of
space occupied by what we take to be an ordinary physical object like the
coin does differ markedly from its surroundings: it contains a more dense
and tightly organized collection of molecules; the atoms in those molecules
are of different elements; the contents of that collection are bound together
by various forces that tend to keep it moving as a group; other forces make
the region relatively impenetrable; and so on. Similarly, she confirms that
objects have properties, stand in relations, and that situations involving
them exhibit dependencies.
Now it must be admitted that there are those who would disagree, who
would question the existence of ordinary objects, beginning with
Eddington and his famous two tables:
One of them is familiar to me from my earliest years. . . . It has extention; it
is comparatively permanent; it is coloured; above all it is substantial. By

5
Here, briefly, the distinction between logical truths and valid inferences matters, because the gaps
undermine all of the former. Inferences often survive because gaps are ruled out when the premises
are taken to obtain.
96 Penelope Maddy
substantial I do not mean merely that it does not collapse when I lean upon
it; I mean that it is constituted of ‘substance’. (Eddington 1928: ix)
Table No. 2 is my scientific table. . . . It . . . is mostly emptiness. Sparsely
scattered in that emptiness are numerous electric charges rushing about
with great speed; but their combined bulk amounts to less than a billionth
of the bulk of the table itself. Notwithstanding its strange construction it
turns out to be an entirely efficient table. It supports my writing paper as
satisfactorily as table No. 1; for when I lay the paper on it the little electric
particles with their headlong speed keep on hitting the underside, so that
the paper is maintained in shuttlecock fashion at a nearly steady level. If
I lean upon this table I shall not go through; or, to be strictly accurate, the
chance of my scientific elbow going through my scientific table is so
excessively small that it can be neglected in practical life. (Eddington
1928: x)
So far, the Second Philosopher need have no quarrel; Eddington can be
understood as putting poetically what she would put more prosaically:
science has taught us some surprising things about the table, its properties
and behaviors.
But this isn’t what Eddington believes:
Modern physics has by delicate test and remorseless logic assured me that
my second scientific table is the only one which is really there. (Eddington
1928: xii)
The Second Philosopher naturally wonders why this should be so, why the
so-called ‘scientific table’ isn’t just a more accurate and complete descrip-
tion of the ordinary table.6 In fact, it turns out that ‘substance’ in
Eddington’s description of table No. 1 is a loaded term:
It [is] the intrinsic nature of substance to occupy space to the exclusion of
other substance. (Eddington 1928: xii)
There is a vast difference between my scientific table with its substance (if
any) thinly scattered in specks in a region mostly empty and the table of
everyday conception which we regard as the type of solid reality . . . It
makes all the difference in the world whether the paper before me is poised
as it were on a swarm of flies . . . or whether it is supported because there is
substance below it. (Eddington 1928: xi–xii)
Here Eddington appears to think that being composed of something like
continuous matter is essential to table No. 1, that one couldn’t come to
6
Some writers reject the ordinary table on the grounds that its boundaries would be inexact. As we’ve
seen, the Second Philosopher is happy to accept this sort of ‘worldly vagueness’.
 A Second Philosophy of logic 97
realize that its supporting the paper or resisting my elbow arise very
differently than I might have at first imagined – that one couldn’t come
to realize this, that is, without also coming to realize that there is no such
thing as table No. 1. But why should this be so? Why should our initial
conceptualization be binding in this way? For that matter, is it even clear
that our initial conceptualization includes any account at all of how and
why the table supports paper or resists elbows? The Second Philosopher
sees no reason to retract her belief in ordinary macro-objects.7
So let’s grant the Second Philosopher her claim that formal structure
as she understands it does turn up in our actual world. This means not
only that rudimentary logic applies in such cases, but that it does so
regardless of the physical details of the objects’ composition, the precise
nature of the properties and relations, any particular facts of spatiotem-
poral location, and so on. This observation might serve as the first step
on a path toward the familiar idea, noted earlier, that questions like
these are peculiarly philosophical: the thought would be that if the
correctness of rudimentary logic doesn’t depend on any of the physical
details of the situation, if it holds for any objects, any properties and
relations, etc., then it must be quite different in character from our
ordinary information about the world; indeed, if none of the physical
details matter, if these truths hold no matter what the particular contin-
gencies happen to be, then perhaps they’re true necessarily, in any
possible world at all – and if that’s right, then nothing particular to
our ordinary, contingent world can be what’s making them true.
By a series of steps like these, one might make one’s way to the idea
that logical truths reflect the facts, not about our world, but about a
platonic world of propositions, or a crystalline structure that our
world enjoys necessarily, or an abstract realm of meanings or concepts,
or some such distinctively philosophical subject matter. Many such

7
Eddington’s two tables may call to mind Sellars’ challenge to reconcile ‘the scientific image’ with ‘the
manifest image’. In fact, the manifest image includes much more than Eddington’s table No. 1 – ‘it is
the framework in terms of which, to use an existentialist turn of phrase, man first encountered
himself ’ (Sellars 1962: 6) – but Sellars does come close to our concerns when he denies that ‘manifest
objects are identical with systems of imperceptible particles’ (Sellars 1962: 26). He illustrates with the
case of the pink ice cube: ‘the manifest ice cube presents itself to us as something which is pink
through and through, as a pink continuum, all the regions of which, however small, are pink’ (Sellars
1962: 26), and of course the scientific ice cube isn’t at all like this. Here Sellars seems to think, with
Eddington, that science isn’t in a position to tell us surprising things about what it is for the ice cube
to be (look) pink; he seems to agree with Eddington that some apparent features of the manifest ice
cube can’t be sacrificed without losing the manifest ice cube itself. Indeed the essential features they
cling to are similar: a kind of substantial continuity or homogeneity. The Second Philosopher
remains unmoved.
98 Penelope Maddy
options spring up in the wake of this line of thought, but ordinary facts,
ordinary information about our ordinary world has been left behind,
and ordinary inquiry along with it – we’ve entered the realm of philoso-
phy proper.
But suppose our Second Philosopher doesn’t set foot on this path.
Suppose she simply notices that nothing about the chemical makeup of
the coin is relevant, that nothing about where the coin is located is
relevant, – that only the formal structure matters to the reliability of the
rudimentary logic she’s isolated. From here she simply continues her
inquiries, turning to other pursuits in geology, astronomy, linguistics,
and so on. At some point in all this, she encounters cathode rays and
black body radiation, begins to theorize about discrete packets of energy,
uses the quantum hypothesis to explain the photo-electric effect, and
eventually goes on to the full development of quantum mechanics. And
now she’s in for some surprises: the objects of the micro-world seem to
move from one place to another without following continuous trajectories;
a situation with two similar particles A and B apparently isn’t different
from a situation with A and B switched; an object has some position and
some momentum, but it can’t have a particular position and a particular
momentum at the same time; there are dependencies between situations
that violate all ordinary thinking about dependencies.8 Do the ‘objects’,
‘properties’, ‘relations’, and ‘dependencies’ of the quantum-mechanical
micro-world enjoy the formal structure that underlies rudimentary logic?
The Second Philosopher might well wonder, and sure enough, her doubts
are soon realized. In a case analogous to, but simpler than position and
momentum, she finds an electron a with vertical spin up or vertical spin
down, and horizontal spin right or horizontal spin left – (Ua or Da) and
(Ra or La) – but for which the four obvious conjunctions – (Ua and Ra) or
(Ua and La) or (Da and Ra) or (Da and La) – all fail. This distributive law
of rudimentary logic doesn’t obtain!
We’re now forced to recognize that those very general features the
Second Philosopher isolated in her formal structures actually have some
bite. Though it wasn’t made explicit, an object in a formal structure was
assumed to be an individual, fundamentally distinct from all others; having
a property – like location, for example – was assumed to involve having a
particular (though perhaps imprecise) property – a particular location, not
just some location or other. These features were so obvious as to go
unremarked until the anomalies of quantum mechanics came along to
8
For more on these quantum anomalies, with references, see (Maddy 2007, §III.4).
 A Second Philosophy of logic 99
demonstrate so vividly that they can in fact fail.9 Those of us who ventured
down that path the Second Philosopher didn’t take were tempted to think
that her formal structure is to be found in every possible world, but it turns
out it isn’t present even in every quarter of our own contingent world!10
Rudimentary logic isn’t necessary after all; its correctness is contingent on
the very general, but still not universal, features isolated in the Second
Philosopher’s formal structure.
We’ve focused so far on the metaphysics – what makes these inferences
reliable, these truths true? – but there’s also the epistemology – how do we
come to know these things? If we followed the philosopher’s path and
succeeded in dismissing the vicissitudes of contingent world as irrelevant
well before the subsequent shocks dealt the Second Philosopher by quan-
tum mechanics, then we might continue our reasoning along these lines: if
logic is necessary, true in all possible worlds, if the details of our contingent
world are beside the point, then how could coming to know its truths
require us to attend to our experience of this world?11 Again a range of
options flourish here, from straightforward theories of a priori knowledge

9
In yet another twist in the tradition of Eddington and Sellars, Ladyman and Ross (2007) begin from
this observation – that the micro-world doesn’t seem to consist of individual objects – then go on to
classify the ordinary table, along with the botanist’s giant redwoods and the physical chemist’s
molecules, as human constructs imposed for ‘epistemological book-keeping’ (p. 240) on an entirely
objectless world. I suspect that this disagreement with the Second Philosopher traces at least in part
to differing pictures of how ‘naturalistic’ metaphysics is to be done. The Second Philosopher’s
‘metaphysics naturalized’ simply pursues ordinary science and ends up agreeing with the folk, the
botanist and the chemist that there are tables, trees and atoms, that trees are roughly constituted by
biological items like cells, cells by chemical items like molecules, molecules by atoms, and so on. She
doesn’t yet know, and may never know, how to extend this program into the objectless micro-
world, but she has good reason to continue trying, and even if she fails, she doesn’t see that this
alone should undermine our belief in the objects of our ordinary world. In contrast, the ‘naturalized
metaphysics’ of Ladyman and Ross is the work of ‘naturalistic philosophical under-labourers’
(p. 242), designed to show ‘how two or more specific scientific hypotheses, at least one of which
is drawn from fundamental physics, jointly explain more than the sum of what is explained by the
two hypotheses taken separately’ (Ladyman and Ross 2007: 37) – and it’s this project that delivers
the surprising result that ordinary objects are constructed by us. From their perspective, the Second
Philosopher ‘metaphysics naturalized’ is just more science: the botanist and the physical chemist
make no contribution to ontology; metaphysics only begins when their hypotheses are unified with
fundamental physics. From the Second Philosopher’s perspective, there’s no reason to suppose that
ordinary objects are human projections or to insist that assessments of what there is must involve
unification with fundamental physics. Indeed, from her perspective, given our current state of
understanding (see below), quantum mechanics is perhaps the last place we should look for
ontological guidance!
10
This incidentally removes another sort of skeptical challenge to the Second Philosopher’s belief in
ordinary macro-objects, namely, the charge that an inquiry starting with objects with properties,
etc., will inevitably uncover objects with properties.
11
An inference from necessary to a priori is less automatic in our post-Kripkean age, when many
philosophers recognize a posteriori necessities, but logical truth seems a poor candidate for this sort
of thing. In any case, what I’m tracing here are tempting paths, not conclusive arguments.
100 Penelope Maddy
to complex accounts of how logic serves to constitute inquiry and thus
can’t itself be confirmed. But let’s return to the Second Philosopher’s more
naïve inquiries, still well clear of the philosopher’s path, and ask how she
answers the simple question: how do we come to know that rudimentary
logical inference is reliable?
In general, the Second Philosopher’s epistemological investigations take
the form of asking how human beings – as described in biology, physi-
ology, psychology, linguistics, and so on – come to have reliable beliefs
about the world – as described in physics, chemistry, botany, astronomy,
and so on.12 Work in psychology, cognitive science, and the like is primary
here, but the Second Philosopher’s focus is somewhat broader; not only
does she study how people come to form beliefs about the world, she also
takes it upon herself to match these beliefs up with what her other
inquiries have told her about how the world actually is, and to assess
which types of belief-forming processes, in which circumstances, are reli-
able. Though her epistemology is naturalized – that is, it takes place
roughly within science – it’s also normative.
In the case of rudimentary logic, the Second Philosopher’s focus is on
formal structure: her other studies of the world have revealed the existence
of many objects, with properties, standing in relations, with dependencies,
and she now asks how we come to be aware of these worldly features. Here
she recapitulates the work of an impressive research community in con-
temporary cognitive science.13 The modern study of our perception of
individual objects reaches back at least to the 1930s, when Piaget used
experiments based on manual search behavior14 to argue that a child
reaches the adult conception of a permanent, external object by a series
of stages ending at about age 2. Conflicting but inconclusive indications
from visual tracking suggested that even younger children might have the
object concept, but it wasn’t until the 1980s that a new experimental
paradigm emerged for testing this possibility: habituation and preferential
looking. In such an experiment, the infants are shown the same event over
and over until they lose interest, as indicated by their decreased looking
time (habituation); they’re then shown one or another of two test displays,
one that makes sense on the adult understanding of an object, the other

12
This is reminiscent of many of Quine’s descriptions of his ‘epistemology naturalized’, but Quine
also tends to fall back on more traditional philosophical formulations, asking how we manage to
infer our theory of the external world from sensory data (see Maddy 2007, §I.6; for more).
13
I can only give the smallest sampling of this work here. For more, with references, see (Maddy 2007,
§III.5).
14
E.g., does the child lift a cloth to find a desirable object she’s seen hidden there?
 A Second Philosophy of logic 101
inconsistent with the adult understanding; if an infant is thinking like
the adult, the inconsistent display should draw a longer gaze (preferential
looking).
So, for example, suppose a metal screen is attached to a long hinge that
extends from left to right on a stage; the screen can lie flat toward the
viewer on the stage surface, and it can pivot through 180˚ arc to lie flat
away from the viewer. The infant is habituated to seeing the screen move
through this range of motion. Then the screen is positioned toward
the infant, a box is placed behind it, and the screen is rotated backwards.
The consistent display shows the screen stopping when it comes to rest
on the now-hidden box; the inconsistent display shows it moving as before
and coming to rest on the stage surface away from the viewer. If the infant
thinks the box continues to exist even when it’s hidden by the screen, and
that the space it occupies can’t be penetrated by the screen, then the
inconsistent display should draw the longest gaze. (Notice that the incon-
sistent display is exactly the one the infant has been habituated to, so its
very inconsistency would be sufficiently novel to overcome the habitu-
ation.) In this early use of the new paradigm, this is exactly what was
observed in infants around five months of age.
Obviously this is only the beginning of the story. For example, does the
infant understand the box as an individual object, as a unit, or just as an
obstacle to the screen? Experiments of similar design soon indicated that
infants as young as four months perceive a unit when presented with a
bounded and connected batch of stuff that moves together. Now imagine
a display with two panels separated by a small space. An object appears
from stage left, travels behind one screen, after which an object emerges
from behind the second screen, and vanishes stage right. One group of
four-month-olds is habituated to seeing an object appear in the gap
between the screens, as if it moved continuously throughout; another
group is habituated to seeing an object disappear behind the first screen
and an object emerge from behind the second screen without anything
appearing in the gap. The test displays are then without panels, showing
either one or two objects. The result was that the infants habituated with
the apparently continuous motion looked longer at the two-object test
display than the infants habituated with the scene where no object was
seen in the gap. It seems an object is regarded as an individual if its motion
is continuous.
Of course there’s much more to this work than can be summarized here,
but the current leading hypothesis is that these very young infants concep-
tualize individual units in these terms: they don’t think that such a unit
102 Penelope Maddy
can be in two places at once or that separate units can occupy the same
space, and they expect them to travel in continuous trajectories. In the
words of Elizabeth Spelke, a pioneer in this field, the infant’s objects are
‘complete, connected, solid bodies that persist over occlusion and maintain
their identity through time’ (Spelke 2000: 1233):
Putting together the findings from studies of perception of object boundar-
ies and studies of perception of object identity, young infants appear to
organize visual arrays into bodies that move cohesively (preserving their
internal connectedness and their external boundaries), that move together
with other objects if and only if the objects come into contact, and that
move on paths that are connected over space and time. Cohesion, contact,
and continuity are highly reliable properties of inanimate, material objects:
objects are more likely to move on paths that are connected than they are to
move at constant speeds, for example; and they are more likely to maintain
their connectedness over motion than they are to maintain a rigid shape.
Infants’ perception appears to accord with the most reliable constraints on
objects. (Spelke et al. 1995: 319–320)

Partly because so much of this research depends on experiments conducted

with habituation/preferential-looking and closely related designs, partly for
other reasons,15 these conclusions can’t be taken as irrevocably established,
but then the fallibility of ongoing science is an occupational hazard for
the Second Philosopher. Let’s take the early emergence of a modest
human ability to detect (some of ) the world’s individual objects as a
tentative datum.
As for properties and relations, the infants’ sensitivity to these plays a
role in the habituation/preferential-looking studies mentioned earlier:
habituating to green objects then preferentially looking at red ones must
involve noticing those colors, likewise the spatial relations of objects and
the screens. What’s surprising is that object properties aren’t initially used
to individuate them. Ten-month-old infants watched as a toy duck
emerged from the left side of a single screen, followed by a ball emerging
from the right side of the screen; one of the two test displays then showed
the duck and the ball, the other just the duck – and no significant
difference between their reactions was found! The same experiment run

15
E.g., Hatfield (2003) argues that the findings of Spelke and her collaborators only establish that
young infants perceive ‘bounded trackable volumes’ not ‘individual material objects’. Of course,
Spelke (e.g., in Spelke et al. (1995), cited by Hatfield) does allow that the infant’s object concept
continues to develop in early childhood, so there is room here for clarification of levels or degrees of
object perception. A question like this would prompt the Second Philosopher to get down to sorting
things out, but I’m not so idealized an inquirer and leave these further investigations to others.
 A Second Philosophy of logic 103
on slightly older one-year-olds delivered what an adult would have
expected: the test display with the duck alone drew greater attention.
On reflection these results aren’t so bewildering. While infants begin
with simple but highly reliable spatiotemporal constraints on object iden-
tity (as Spelke notes), property distinctions require more judicious appli-
cation: a red ball can turn blue and still be the same object; a human can
change clothes and still be the same person. Some experience and learning
must be needed for the child to realize that ducks don’t generally turn into
balls, and considerably more to reach the full adult concept:

We are inclined to judge that a car persists when its transmission is replaced,
but would be less inclined to judge that a dog persists if its central nervous
system were replaced. . . . Because we know that dogs but not cars have
behavioral and mental capacities supported by certain internal structures, we
consider certain transformations of dogs to be more radical than other,
superficially similar transformations of cars. (Spelke et al. 1995: 302–303)

With this in mind, it’s less surprising that the beginnings of the child’s
identification of objects by their properties comes a couple of months later
than their identification by the more straightforward spatiotemporal
means, and perhaps even that this new development apparently coincides
with the acquisition of their first words – property nouns like ‘ball’ and
‘duck’!
So as not to belabor this fascinating developmental work, let me just
note that similar studies have shown that young infants detect conjunc-
tions and disjunctions of object properties, the failure of properties or
relations, simple billiard-ball style causal dependencies, and so on. It’s also
notable that many of these abilities found in young infants are also present,
for example, in primates and birds. This suggests an evolutionary origin,
and clearly the advantages conferred by the ability to track objects spatio-
temporally, to perceive their properties and relations, to notice dependen-
cies, would have been as useful on the savanna as they are in modern life.
All this leaves the Second Philosopher with two well-supported hypoth-
eses: the ability to detect (at least some of ) the formal structure present in
the world comes to humans at a very early age, perhaps largely due to our
evolutionary inheritance; whether by genetic endowment, normal matur-
ation, or early experience, the primitive cognitive mechanisms underlying
this ability are as they are primarily because humans (and their ancestors)
interact almost exclusively with aspects of the world that display this
formal structure. From here it’s a short step to the suggestion that the
presence of these primitive cognitive mechanisms, all tuned to formal
104 Penelope Maddy
structure, is what makes the simpler inferences of rudimentary logic strike
us as so obvious. Assuming that the Second Philosopher has this right –
that the formal structure is often present, that we are configured to detect
it, and that this accounts for our rudimentary logical beliefs – then a
sufficiently externalist epistemologist might count this as a case of a priori
knowledge. An epistemologist of more internalist leanings might hold that
the sort of a posteriori inquiry undertaken here would be required to
support actual knowledge of rudimentary logic. The Second Philosopher
isn’t confident that this disagreement has a determinate solution, isn’t
confident that the debate is backed by anything more substantial than
the various handy uses of ‘know’, so she’s content to offer a fuller version
of the story sketched here, and to leave the decision about ‘knowledge’
to others.
Notice, incidentally, that if this is right, if the Second Philosopher’s
formal structure is so deeply involved in our most fundamental cognitive
mechanisms, this explains why it’s so difficult for us to come up with a
viable interpretation of quantum mechanics, where formal structure goes
awry. But this observation raises another question: if formal structure and
hence rudimentary logic are missing in the micro-world, and if these are
so fundamental to our thought and reasoning, how do we manage to
carry out our study of quantum mechanics? Some suggest that we should
adopt a special logic for quantum mechanics,16 but the question posed
here is how we manage to do quantum mechanics now, apparently using
our ordinary logic. I think the answer is fairly simple: what we actually
have in quantum mechanics isn’t a theory of particles with properties, in
relations, with dependencies, but a mathematical model, an abstract
Hilbert space with state vectors.17 This bit of mathematics displays all
the necessary formal structure – it consists of objects with properties, in
relations, with (logical) dependencies – so our familiar logic is entirely
reliable there.18 The deep problem for the interpretation of quantum

16
See the discussion of deviant logics below.
17
As noted above (footnote 9), Ladyman and Ross (2007) argue on grounds similar to the Second
Philosopher’s that the micro-world doesn’t consist of objects. Given her account of how our
cognition and our logic work, the Second Philosopher would predict that these authors should
encounter some difficulty when it comes to describing the subject matter of quantum mechanics,
and in fact, what they say on that score is consistent with the line taken here: ‘it is possible that
dividing a domain up into objects is the only way we can think about it’ (Ladyman and Ross (2007:
155); ‘we can only represent [the non-objectual structures of the micro-world] in terms of
mathematical relationships’ (Ladyman and Ross 2007: 299).
18
In fact, the mathematical world is in some ways more amenable to our logical ways (see footnote
19 below).
 A Second Philosophy of logic 105
mechanics is to explain how and why the mathematical model works so
well, to figure out what worldly features it’s tracking, but in the math-
ematics itself, our natural ways of thinking and reasoning are on impec-
cable footing.
Now for all its advertised virtues – reliability in a wide variety of worldly
settings, harmony with our most fundamental cognitive mechanisms –
rudimentary logic is in fact a rather unwieldy instrument in actual use.
We’ve seen, for example, that the presence of indeterminacies eliminates
the law of excluded middle, the principle of non-contradiction, and indeed
all logical truths. An inference rule as central as reductio ad absurdum can
be seen to fail: that (q-and-not-q) follows from p only tells us that p is
either false or indeterminate. And the substantive requirements on
dependency relations undercut most of our usual manipulations with the
conditional. Though he’s speaking of a full Kleene system, with a truth-
functional conditional, I think Feferman’s assessment applies to rudimen-
tary logic as well: ‘nothing like sustained ordinary reasoning can be carried
on’ (Feferman 1984: 95).
Under the circumstances, a stronger, more flexible logic is obviously to
be desired. The Second Philosopher has seen this sort of thing many times:
she has a theoretical description of a given range of situations, but that
description is awkward or unworkable in various ways. To take one
example, she can give a complete molecular description of water flowing
in a pipe, but alas all practical calculation is impossible. In hope of making
progress, she introduces a deliberate falsification – treating the water as a
continuous substance – that allows her to use the stronger and more
flexible mathematics of continuum mechanics. She has reason to think
this might work, because there should be a size-scale with volumes large
enough to include enough molecules to have relatively stable temperature,
energy, density, etc., but not so large as to include wide local variations in
properties like these. This line of thought suggests that her deliberate
falsification might be both powerful enough to deliver concrete solutions
and benign enough to do so without introducing distortions that would
undercut its effectiveness for real engineering decisions. She tests it out,
and happily it does work! This is what we call an ‘idealization’, indeed a
successful idealization for many purposes. (It would obviously be unaccept-
ably distorting if we were interested in explaining the water’s behavior
under electrolysis.) In similar ways, we ignore friction when its effects are
small enough to be swamped by the phenomenon we’re out to describe; we
treat slightly irregular objects as perfectly geometrical when this does no
harm; and so on.
106 Penelope Maddy
With the technique of idealization in mind, the Second Philosopher
looks for ways to simplify and streamline her theoretical account of
formal structure, that is, her rudimentary logic, in ways that make it
more flexible, more workable, and to do so without seriously undermin-
ing its reliability. To this end, she makes two key idealizations, introduces
two falsifying assumptions – that there is no indeterminacy, that any
particular combination of objects and properties or relations either holds
or fails; and that dependencies behave as material conditionals19 – and at
a stroke, she transforms her crude rudimentary theory into our modern
classical logic. There can be no doubt that full classical logic is an
extraordinarily sophisticated and powerful instrument; the only open
question is whether or not the required idealizations are benign. And
as in the other examples, this judgment can be expected to vary from
case to case.
This is where some of the so-called ‘deviant logics’ come in. Proponents
of one or another of the various logics of vagueness, for example, may insist
that indeterminacy is a real phenomenon,20 may condemn ‘the lamentable
tendency . . . to pretend that language is precise’ (J. A. Burgess 1990: 434).
On the first point, the Second Philosopher agrees – indeterminacy is real –
but she views the classical logician’s pretending otherwise as no different in
principle than the engineer’s pretending that water is a continuous fluid;
what determines the acceptability of either pretense isn’t the obvious fact
that it is a pretense, but whether or not it is beneficial and benign in the
situation at hand. Most logics of vagueness begin from a picture not unlike
the Second Philosopher’s, in which, for example a property can hold of an
object, fail to hold, or be indeterminate for that object; there’s also the
problem of higher-order indeterminacy, that is, of borderline cases
between holding and being indeterminate, between being indeterminate
and failing. So far, I think it’s fair to say that there is no smooth and
perspicuous logic of vagueness, no such logic that escapes Feferman’s
critique. It is, of course, true that classical logic can lead us astray in
contexts with indeterminacy – this is the point of the sorites paradox –
but at least for now the Second Philosopher’s advice is simply to apply

19
Though these idealizations involve falsification in her description of the physical world, they are
satisfied in the world of classical mathematics: excluded middle holds and the dependencies are
logical. For more on the ontology of mathematics, see (Maddy 2011).
20
There is serious disagreement between various writers over the source of the indeterminacy: is it
purely linguistic or does the world itself include borderline cases and fuzzy objects? Here the Second
Philosopher sides with the latter, but this shouldn’t affect the brief discussion here, despite the
formulation in the quotation in the next clause above.
 A Second Philosophy of logic 107
classical logic with care,21 as one should any idealization, rather than switch
to a less viable logic.22
Advocates of various conditional logics protest the Second Philosopher’s
other bold idealization: replacement of real dependencies with the simple
material conditional. There are many proposals for a more substantial
conditional, far too many to consider here (even if my slender expertise
allowed it), but perhaps the conditional of relevance logic can be used as one
representative example. The motivation here speaks directly to the falsifica-
tion in question: the antecedent of a conditional should be ‘relevant’ to the
consequent.23 To return to our earlier example, the redness of this rose isn’t
relevant to the fact that 2 þ 2 = 4, despite the truth of the corresponding
material conditional (if the rose is red, then 2 þ 2 = 4). Of course, as before,
the Second Philosopher fully appreciates that the material conditional is a
falsification, that the rose inference is an anomaly, but the pertinent
questions are whether or not the falsification is beneficial and benign, and
whether or not the relevance logician has something better to offer. Again
I think that for now, we do best to employ our classical logic with care.
So we see that some deviant logics depart from the Second Philosopher’s
classical logic by rejecting her idealizations,24 and that our assessment then
depends on the extent to which the falsifications introduced are beneficial and
benign, and on the systematic merits of the proposed alternative. But not all
deviant logics fit this profile; some concern not just the idealizations of classical
logic, but the fundamentals of rudimentary logic itself. Examples include
intuitionistic logic – which rejects double negation elimination – quantum

21
Sorensen (2012) credits this approach to H. G. Wells: ‘Every species is vague, every term goes cloudy
at its edges, and so in my way of thinking, relentless logic is only another name for a stupidity – for a
sort of intellectual pigheadedness. If you push a philosophical or metaphysical enquiry through a
series of valid syllogisms – never committing any generally recognized fallacy – you nevertheless
leave behind you at each step a certain rubbing and marginal loss of objective truth and you get
deflections that are difficult to trace, at each phase in the process’ (Wells 1908: 11).
22
Williamson (1994) also advocates retaining classical logic, but his reason is quite different: because
there is no real vagueness, because apparent borderline cases really just illustrate our ignorance of
where the true borderline lies. This strikes many, including me, as obviously false.
23
Relevance logicians are particularly unhappy with what they call ‘explosion’, the classical oddity that
anything follows from a contradiction. For related reasons, full relevance logic rejects even some
rudimentary logical inferences not involving the conditional, like disjunctive syllogism, but I leave
this aside here. (For a bit more, see Maddy 2007: 292, footnote 24.)
24
Some other deviant logics respond to idealizations of language rather than the worldly features of
rudimentary logic: e.g., free logicians counsel us to reject the falsifying assumption that all naming
expressions refer. Here, too, our assessment depends on the effectiveness of the idealization and the
viability of the alternative. In practical terms, leaving aside the various technical studies in the theory
of free logics, I’m not sure using a free logic is readily distinguishable from being careful about the
use of existential quantifier introduction in the context of classical logic. In any case, our concern
here is with worldly idealizations, not linguistic ones.
108 Penelope Maddy
logic – which rejects the distributive laws – and dialetheism – which holds
that there are true contradictions. Given the connection of rudimentary logic
with the Second Philosopher’s formal structure, the challenge for each of
these is to understand what the world is like without this formal structure,
what the world is like that this alternative would be its logic.25 Of the three,
intuitionistic logic comes equipped with the most developed metaphysical
picture, but it’s suited to describing the world of constructive mathematics,
not the physical world.26 Quantum logic at first set out to characterize the
non-formal-structure of the micro-world, but in practice it has not succeeded
in doing so;27 the problem of interpreting quantum mechanics remains
open. And dialetheism faces perhaps the highest odds: as far as I know, its
defenders have focused for the most part on the narrower goal of locating a
compelling example of a true contradiction in the world, perhaps so far
without conspicuous success.28 The Second Philosopher tentatively con-
cludes that rudimentary logic currently has no viable rivals as the logic of
the world, and that classical logic likewise stands above its rivals as an
appropriate idealization of rudimentary logic for everyday use.
In sum, then, the Second Philosopher’s answer, an ordinary answer to
the question of why that coin must be foreign, is that the coin and its
properties display formal structure and the inference in question is reliable
in all such situations. This answer doesn’t deliver on the usual philosoph-
ical expectations: the reliability of the inference is contingent, our know-
ledge of it is only minimally a priori at best. The account itself results from
plain empirical inquiry, which may lead some to insist that it isn’t
philosophy at all. Perhaps not. Then again, if the original question –
why is this inference reliable? – counts as philosophical – and it’s not clear
how else to classify it – then the answer, too, would seem to have some
claim to that honorific. But the Second Philosopher doesn’t care much
about labels. After all, even ‘Second Philosophy’ and ‘Second Philosopher’
aren’t her terms but mine, used to describe her and her behavior. In any
case, philosophy or not, I hope the Second Philosopher’s investigations do
tell us something about the nature of that inference about the coin.29

25
Our interest here is in the logic of the world, not the logic that best models something else, as, e.g.,
paraconsistent logic (a variety of relevance logic) might serve to model belief systems (see Maddy
2007: 293–296).
26
See the discussion of Creator Worlds in (Maddy 2007: 231–233, 296) and (Maddy to appear, §II).
27 28
See (Maddy 2007: 276–279, 296). See (Maddy 2007: 296–297).
29
My thanks to Patricia Marino for helpful comments on an earlier draft and to Penelope Rush for
editorial improvements.
 chapter 6

Logical nihilism
Curtis Franks

1. Introduction
The idea that there may be more than one correct logic has recently
attracted considerable interest. This cannot be explained by the mere fact
that several distinct logical systems have their scientific uses, for no one
denies that the “logic” of classical mathematics differs from the “logics” of
rational decision, of resource conscious database theory, and of effective
problem solving. Those known as “logical monists” maintain that the
panoply of logical systems applicable in their various domains says nothing
against their basic tenet that a single relation of logical consequence is
either violated by or manifest in each such system. “Logical pluralists” do
not counter this by pointing again at the numerous logical systems, for
they agree that for all their interest many of these indeed fail to trace any
relation of logical consequence. They claim, instead, that no one logical
consequence relation is privileged over all others, that several such relations
abound.
Interesting as this debate may be, I intend to draw into question the
point on which monists and pluralists appear to agree and on which their
entire discussion pivots: the idea that one thing a logical investigation
might do is adhere to a relation of consequence that is “out there in the
world,” legislating norms of rational inference, or persisting some other
wise independently of our logical investigations themselves. My opinion is
that fixing our sights on such a relation saddles logic with a burden that it
cannot comfortably bear, and that logic, in the vigor and profundity that
it displays nowadays, does and ought to command our interest precisely
because of its disregard for norms of correctness.
I shall not argue for the thesis that there are no correct logics. Although
I do find attempts from our history to paint a convincing picture of a
relation of logical consequence that attains among propositions (or sen-
tences, or whatever) dubious, I should not know how to cast general doubt

109
110 Curtis Franks
on the very idea of such a relation. By “drawing this point into question”
I mean only to invite reflection about what work the notion of a correct
logic is supposed to be doing, why the debate about the number of logical
consequence relations is supposed to matter to a logician, and whether the
actual details of logic as it has developed might be difficult to appreciate if
our attention is overburdened by questions about the correctness of logical
principles. Rather than issue any argumentative blows, I propose merely to
lead the reader around a bit until his or her taste for a correct logic sours.

2. The law of excluded middle

Surely the most notorious bone of contention in the discussion of logical
correctness is the law of excluded middle, f _ :f. Is this law logically
valid, so that we know that its instances, “Shakespeare either wrote all
those plays or he didn’t,” “Either the continuum hypothesis is true or it is
not true,” “He’s either bald or he isn’t,” etc., each are true in advance of
any further information about the world?
One hardly needs to mention that hundreds of spirited disavowals and
defenses of lem have been issued in the last century. Many of these have
even been authored by expert logicians. But let us turn our backs to these
ideological matters and consider briefly some of what we have learned
about lem quite independently of any question about its correctness.
The simplest setting for this is the propositional calculus.1 The intui-
tionistic propositional calculus (IPC) differs from the classical propos-
itional calculus (CPC) precisely in its rejection of lem. For a concrete
and standard formalization of IPC one may take a typical Hilbert-style
axiomatization of CPC and erase the single axiom for double negation
elimination, ‘ ::f ⊃f, leaving the rules of inference as before.
In fact, one of the first things observed about lem is its equivalence with
dne, which is most easily seen in the setting of natural deduction. Standard
natural deduction presentations of CPC have a rule allowing one to infer
any formula f from the single premise ::f. It is easy to derive in such a
system the formula f _ :f. It is similarly easy to show that if we modify

1
I emphasize that this really is a matter of perspicuity. One should not think that the phenomena
described below are artifacts of peculiar features of propositional logic. They are nearly all
consequences of decisions about lem that are invariant across a wide spectrum of logics. Consider:
the structural subsumption of lem applies also to the predicate calculus; the admissible propositional
rules of Heyting Arithmetic (and of Heyting Arithmetic with Markov’s principle) are exactly those of
IPC (Visser 1999); the disjunction property holds for Heyting Arithmetic (Kleene 1945) and for
intuitionistic Zermelo–Fraenkel set theory (Myhill 1973).
 Logical nihilism 111
this calculus by disallowing dne but now introduce a new rule that allows
one to write down any instance of lem in any context, f will be derivable
in all contexts in which ::f is derivable.
In (1934–35), Gentzen observed to his own surprise2 that the sequent
calculus presentation of CPC admits an even more elegant modification
into a presentation of IPC than the one just described. One simply
disallows multiple-clause succedents and leaves the calculus otherwise
unchanged. Thus lem and with it the entire distinction between
intuitionistic and classical logic is subsumed into the background structure
of the logical calculus. All the inference rules governing the logical particles
(^, _, ⊃, :) and all the explicit rules of structural reasoning (identity, cut,
weakening, exchange, and contraction) are invariant under this transform-
ation. Thus it appears that a duly chosen logical calculus allows a precise
analysis of what had been thought of as a radical disagreement about the
nature of logic. When classicists and intuitionists are seen to admit
precisely the same inference rules, their disagreement appears in some
ways quite minor, if more global than first suspected.
Exactly how minor, on closer inspection, is the difference between these
superficially similar calculi CPC and IPC? Not very. Even before
Gentzen’s profound analysis, Gödel (1932) observed that IPC satisfied
the “disjunction property”: formulas such as f _ ψ are provable only if
either f or ψ is as well. At first sight this might appear to be no more than
a restatement of the intuitionist’s rejection of lem. After all, that rejection
was motivated by the idea that instances of lem are un-warranted when
neither of their disjuncts can be independently established. But, one might
think, if any disjunction is warranted in the absence of independent
verification of one of its disjuncts, those like f _ :f are, so rejecting
lem should lead to something like the disjunction property. This reasoning
strikes me as worthy of further elaboration and attention, but it should be
unconvincing as it stands. For one thing, the formal rejection of lem only
bars one from helping oneself to its instances whenever one wishes. The
gap between this modest restriction and the inability ever to infer any
disjunction of the form f _ :f at all, unless from a record of that
inference one could effectively construct a proof either of f or of :f, is
a broad one. More, there are infinitely many formulas f unprovable in

2
Gentzen described the fact that lem prescribes uses of logical particles other than those given by their
introduction and elimination rules as “troublesome.” The way that in the sequent calculus the logical
rules are quarantined from the distinction between classical and intuitionistic logic he called
“seemingly magical.” He wrote, “I myself was completely surprised by this property . . . when first
formulating that calculus” (Gentzen 1938: 259).
112 Curtis Franks
IPC, such that IPCþf is consistent but insufficient to prove lem.
Examples like :p _ ::p may well fuel suspicions that even in the shadow
of global distrust of lem, some disjunctions f _ ψ are more plausible
than any instance of lem even in the absence of resources sufficient to
derive f or ψ.
Thus the disjunction property is a non-trivial consequence of the
invalidity of lem. In fact, this situation exemplifies a recurring phenom-
enon in logic, wherein from the assumption of a special case of some
general hypothesis, that hypothesis follows in its full generality. This often
happens even when, as in the present case, the general phenomenon does
not appear, even in hindsight, to be a “logical consequence” of its instance
in any absolute sense. The disjunction property has further consequences
of its own, however, to which we can profitably turn.
In the approach to semantics known as inferentialism, the meaning of a
logical particle should be identified with the conditions under which one is
justified in reasoning one’s way to a statement governed by that particle.
From this point of view, which is given expression already in some of
Gentzen’s remarks, and owing to the separation in sequent calculus of lem
from the logical rules, the meanings of the familiar logical particles might
be said not to differ in intuitionistic and classical logic.
However, the disjunction property gives rise to an alternative interpret-
ation of the logical particles of IPC in which each theorem refers back to
the notion of provability in IPC itself. For if f _ ψ is provable only if one
of f and ψ is as well, then a candidate and interesting reading of the
sentence ‘ IPC f _ ψ is “Either ‘ IPC f or ‘ IPC ψ.” Expanding on this
idea, one might suggest that the provability of a conjunction “means”
that each of its conjunctions is provable, that the provability of a condi-
tional, f ⊃ ψ, “means” that given a proof of f one can construct a proof
of ψ, and that the provability of :f “means” that a contradiction can be
proved in the event that a proof of f is produced.
As we shall see shortly, this so far informal interpretation of
intuitionistic logic is riddled with ambiguities. All the same, some
reflection should bring home the idea that some disambiguation of this
reading is a possible way to understand the theorems of IPC. When one
compares the situation with CPC, where conditional truth comes so
cheap and the disjunction property fails badly, one can only conclude
that the difference between these calculi is in some sense great after all,
greater even than the debate over the validity of lem alone first suggests.
When one then recalls the earlier observation that these calculi can be
presented so that their formal differences are slight and their rules
 Logical nihilism 113
identical, the observation that radical differences in meaning result from
so subtle a change in form is striking.
If we are to take seriously the idea that theorems of IPC refer back to
IPC provability, then some care must be taken in making this interpret-
ation precise. For if the provability of a conditional, f ⊃ ψ, “means” that
‘ IPC ψ in the event that ‘ IPC f, then one should expect ‘ IPC f ⊃ ψ in
every situation in which the set of theorems of IPC is closed under the rule
“from f, infer ψ.” However, the disjunction property implies that these
expectations will not be met.
To see this, consider the Kreisel–Putnam rule, “From :f ⊃ (ψ _ χ),
infer (:f ⊃ ψ) _ (:f ⊃ χ).” The only derivations of :f ⊃ (ψ _ χ) in
natural deduction are proofs whose last inferences are instances of ⊃-elim
(modus ponens), ^-elim, dne, or ⊃-intro. It is easy to see that a proof
ending in ⊃-elim or ^-elim cannot be the only way to prove this formula,
that in fact any such proof can be normalized into a proof of the same
formula whose last inference is an instance of one of the other two rules. If,
further, we consider the prospects of this formula being a theorem of IPC,
then dne is no longer a rule, and we may conclude that any proof
necessarily contains a subproof of ψ _ χ from the assumption :f (to
allow for ⊃-intro). What might this subproof look like? Once more, dne is
not an option, so again by insisting that the proof is normalized (so that it
doesn’t end needlessly and awkwardly with ⊃-elim or ^-elim) we ensure
that its last step is an instance of _-elim. But this means that an initial
segment of this subproof is a derivation in IPC either of ψ or of χ from the
assumption :f (it is here that the disjunction property rears its head), and
in each case it is clear how to build a proof of (:f ⊃ ψ) _ (:f ⊃ χ).
Putting this all together, we see that whenever :f ⊃ (ψ _ χ) is a theorem
of IPC, so too is (:f ⊃ ψ) _ (:f ⊃ χ).
When a logical system’s theorems are closed under a rule of inference,
we say that the rule is “admissible” for that logic. The above argument
established that the Kreisel–Putnam rule is admissible for IPC. One
might expect that the rule is also “derivable,” that ‘IPC (:f ⊃ (ψ _ χ))
⊃ ((:f ⊃ ψ) _ (:f ⊃ χ)). However, it is not (Harrop 1960). This
situation is dis-analogous to that of classical logic, where all admissible
rules are derivable so that the distinction between admissibility and
derivability vanishes. In the parlance, we say that CPC is “structurally
complete” but that IPC is not.
In fact much of the logical complexity of IPC can be understood as a
residue of its structural incompleteness. For the space of intuitionistically
valid formulas is far more easily navigated than the space of its admissible
114 Curtis Franks
rules: although it is decidable whether or not a rule is admissible in IPC,
there is no finite basis of rules that generates them all (Rybakov 1997).
What ought one make of the structural incompleteness of IPC? One
thing that can definitely be said is that reading the expression f ⊃ ψ as
“There is a procedure for transforming a proof of f into a proof of ψ” is
problematic for both the classicist and the intuitionist, but for different
reasons. This reading is wrong for the classicist, because the idea of
procedurality simply does not enter into the conditions of classical
validity. By contrast, procedures of proof transformation are central for
the intuitionist. However, we now know that there are procedures for
transforming a IPC-proof of f into a IPC-proof of ψ in cases where f ⊃ ψ
is not a theorem of IPC. So at best one could say that IPC is incomplete
with respect to this semantics, and more plausibly one should say that this
reading of ‘IPC f ⊃ ψ is erroneous.
Thus we see a sense in which the phenomenon of structural complete-
ness is related to a sort of semantic completeness: a structurally incomplete
logic will be incomplete with respect to the most naive procedural reading
of its connectives. It also happens that structural completeness bears a
precise relation to the phenomenon of Post-completeness, the situation in
which any addition made to the set of theorems of some logic will trivialize
the logic by making all formulas in its signature provable. To state this
relationship, we refer to a notion of “saturation.” For a logical calculus
L whose formulas form the set S, let Sb(X ) be the set of substitution
instances of formulas in X  S and let CnL(X ) be the set of formulas f
such that X ‘L f. L is saturated if for every X  S CnL(X ) = CnL(Sb(X ))
for every X  S. By a (1973) theorem of Tokarz, a Post-complete calculus is
structurally complete if, and only if, it is saturated.
For these and perhaps other reasons many authors have felt that the
presence of non-derivable, admissible rules is a deficiency of systems like
IPC. The very term “structural incompleteness” suggests that something is
missing from IPC because correct inferences about provability in this logic
are not represented as theorems in IPC. Rybakov (1997), for example,
suggests that “there is a sense in which a derivation inside a [structurally
incomplete] logical system corresponds to conscious reasoning [and] a
derivation using [its] admissible rules corresponds to subconscious
reasoning.” He faults such systems for having rules that are “valid in
reality” yet “invalid from the viewpoint of the deductive system itself ”
(10–11). Structurally complete systems, by contrast, are “self contained” in
the sense that they have the “very desirable property” of being conscious of
all the rules that are reliable tools for discovering their own theorems (476).
 Logical nihilism 115
It seems to me that this attitude derives from wanting to preserve the naive
procedural understanding of the logical connectives. The situation ought
rather, I counter, lead one to appreciate the subtlety of procedurality
exhibited in intuitionism. For the logical lesson to be learned is that
in the absence of lem the context of inference takes on a new role. Thus
f ⊃ ψ means that given any background of assumptions from which f is
provable, a proof of f can be transformed into a proof of ψ under those
same assumptions, and this understanding does not reduce, as it does with
logics insensitive to context, to the idea that any proof of f can be
transformed into a proof of ψ. This irreducibility strikes me as a “very
desirable property” for many purposes. I should like to know more about
the conditions that lead to it.
From this point of view, it is natural to ask whether there are logics that,
unlike classical logic, admit a constructive interpretation but, like classical
logic, are not sensitive in this way to context. Perhaps the constructive
nature of IPC derives from its context-sensitivity. Surprisingly, Jankov’s
logic, IPCþ:f _ : :f, appears to undermine any hope of establishing a
connection between these phenomena. Consider the Medvedev lattice of
degrees of solvability. The setting is Baire space ωω (the set of functions
from ω to ω) and the problem of producing an element of a given subset of
this space. By convention, such subsets are called mass problems, and their
elements are called solutions. One says that one mass problem reduces to
another if there is an effective procedure for transforming solutions of the
second into solutions of the first. If one defines the lattice of degrees of
reducibility of mass problems, it happens that under a very natural
valuation, the set of identities of corresponds to the set of theorems
of Jankov’s logic, so that the theory of mass problems provides a
constructive interpretation of this logic.3 However, Jankov’s logic is struc-
turally complete (Prucnal 1976). Thus one sees that the so-called weak law
of excluded middle preserves the context insensitivity of CPC despite,
standing in the place of full lem, allowing for a procedural semantics.
In my graduate student years, several of my friends and I were thinking
about bounded arithmetic because of its connections with complexity
theory and because the special difficulty of representing within these
theories their own consistency statements shed much light on the fine
details of arithmetization. We had a running gag, which is that formal
theories like PA are awfully weak, because with them one can’t draw very
many distinctions. The implicit punchline, of course, is that the bulk of
3
For the details of this interpretation, see (Terwijn 2004).
116 Curtis Franks
the distinctions one can draw in theories of bounded arithmetic are among
statements that are in fact equivalent. Lacking the resources to spot these
equivalences is no strength!
Something perfectly analogous happens in the case of substructural
logics. There are theorems of CPC that are unprovable in IPC, but not
vice versa, so the latter logic is strictly weaker. Moreover, CPC proves all
sorts of implications and equivalences that IPC misses. But if we stop
believing for a moment, as the discipline of logic demands we do, and ask
about the fine structure of inter-dependencies among the formulas of
propositional logic, IPC delivers vastly more information. Consider just
propositional functions of a single variable p. In CPC there are exactly
four equivalence classes of such formulas: those inter-derivable with p, :p,
p _ :p, and p ^ :p. In IPC the equivalence classes of these same formulas
exhibit a complicated pattern of implications, forming the infinite Rieger–
Nishimura lattice.
One thought one may have is that IPC should be considered an expan-
sion of CPC: every classical tautology can be discovered with IPC via the
negative translation of Gödel (1933) and Gentzen, so with IPC one gets all
the classical tautologies and a whole lot more. (Gödel at times suggested
something like this attitude.) But, of course, neither the negative transla-
tion nor the very idea of a classical tautology arises within intuitionistic
logic. The thought that I encourage instead is this: The logician is loath to
choose between classical and intuitionistic logic because the phenomena of
greatest interest are the relationships between these logical systems. Who
would have guessed that the rejection of a single logical principle would
generate so much complexity – an r.e. set of admissible rules with no finite
basis, an infinite lattice of inter-derivability classes?
The intuitionist and the classicist have very fine systems. Perhaps with
them one gains some purchase on the norms of right reasoning or the
modal structure of reality. The logician claims no such insight but
observes that one can hold fixed the rules of the logical particles and, by
merely tweaking the calculus between single conclusion and multi con-
clusion, watch structural completeness come in and out of view. The same
switch, he or she knows, dresses the logical connectives up in a construct-
ive, context-sensitive interpretation in one position and divests them of
this interpretation in the other. These connections between sequent
calculus, constructive proof transformation, structural completeness, and
lem are fixtures from our logical knowledge store, but they cannot
seriously be thought of as a network of consequences in some allegedly
correct logic.
 Logical nihilism 117

3. Logic imposed and logic discovered

If I have conveyed my attitude successfully, then I will have inspired the
following objection:

You speak about an unwillingness to embrace any one or select few logical
systems because of an interest in understanding all such systems and how
their various properties relate to one another. But by making logical systems
into objects of investigation, you inhabit an ambient space in which you
conduct this investigation. It is legitimate to ask which logic is appropriate
in this space. What is your “metalogic”?

I shall explain that this objection rests on various preconceptions that I do

not share. I hope the explanation functions to aid the reader in seeing these
as misconceptions. If it does, then logical nihilism will be understood.
If we agree that as logicians we are interested, not in factual truth, but in
the relationships among phenomena and ideas, then the point of view we
must hasten to adopt should be the one that assists us in detecting and
understanding these relationships. Which relationships? Presumably, it has
been suggested, those that accurately pick out grounds and consequences,
those that answer the question “What rests on what?” But why stop here?
What sort of purpose is served by simultaneously disregarding factual states
of affairs and pledging allegiance to factual relations of ground and conse-
quence? I have an intuitive sense of what a fact is; I have no such sense of
ontological grounding. Nor have I seen any reason to expect that the study
of logic can foster such a sense in me.
More appropriate seems to be a disregard for privileged relationships
similar to our disregard for truth. Suppose that we are interested in
detecting and understanding whatever relationships we can find. Then
we might wish not to be wedded to any point of view. We might, instead,
try on a few hats until some interesting patterns appear where before there
seemed to have been only disorder. We might find that one hat helps time
and again, but we will be well-advised not to forget that we are wearing it.
For if we never take it off, then we risk forever overlooking logical
relationships of considerable interest. Worse, we risk coming to think
of the relationships we can detect as “in the world,” “preconditions of
thought,” or some such thing.
Allow me to illustrate this point with an example. At least since Aristotle
it was expected that great complexity could be uncovered by a proper
analysis of the quantifiers that occur in natural language. Notably, quanti-
fiers allow us to reason in succinct strokes about infinite collections. About
118 Curtis Franks
a century ago, David Hilbert had the idea of analyzing quantifiers
with consistency proofs, and he devised an intricate calculus with “trans-
finite axioms” which allows proof figures involving quantifiers to be
transformed (in principle) into figures without them. Hilbert’s idea was
that reasoning about this transformation (which by its nature requires
being attentive to constructibility) would expose the quantifiers as innocu-
ous parts of our mathematical language and also make perspicuous their
complexity.4
Hilbert conducted this investigation in the shadow of a great ideological
quarrel about the validity of various logical and mathematical principles. In
one quarter were dour skeptics who distrusted not only lem but other
forms of infinitary reasoning. Chief among them was Kronecker, against
whom Hilbert (1922: 199–201) railed because he “despised . . . everything
that did not seem to him to be an integer.” Less famously, but perhaps
more importantly, he faulted Kronecker also because “it was far from his
practice to think further” about what he did accept, “about the integer
itself.” In a similar vein, Hilbert observed that Poincaré “was from the start
convinced of the impossibility of a proof of the axioms of arithmetic”
because of his belief that mathematical “induction is a property of the
mind.” Thus Hilbert viewed these figures as short-sighted, not only in
their rejection of mathematical techniques that he wished to defend, but
also in their belief that the manifest validity of a principle precludes any
hope of our analyzing it. Both attitudes, he cautioned, “block the path to
analysis.”5
After only a decade of partial successes, it was discovered that the sort of
consistency proofs Hilbert envisioned are not available. Specifically, Gödel
(1931) demonstrated that no proof of the consistency of a reasonably strong
and consistent mathematical system could be carried out within that same
system. Typically it is the recursion needed to verify that the proof
transformation algorithm halts that cannot be so represented. This situ-
ation raises the question whether proving with such principles that a
system is consistent is not obscurum per obscurius.
In (1936) Gentzen made these circumstances much more precise by
providing a perspicuous proof of the consistency of PA. Gentzen’s proof is
carried out in the relatively weak theory PRA together with the relatively
strong principle of transfinite induction up to the ordinal

4
For details, see section 2 of (Franks 2014).
5
For more on Hilbert’s view, see Professor Shapiro’s contribution to this volume.
 Logical nihilism 119
...ωn
 0 ¼ sup ω1 ω2
n<∞

Together with Gödel’s result, this proof demonstrates that transfinite

induction through  0 is unprovable in PA. All of this is well known.
Most people familiar with the history of logic are aware that Gentzen
proved also that transfinite induction to any ordinal beneath  0, any
...ωn
ordinal ω1 ω2 for n  N, is in fact provable in PA. But Gentzen pressed
even further. One can consider fragments of PA defined by restricting the
induction scheme to formulas with a maximum quantifier complexity (call
these the theory’s class of inductive formulas). Gentzen showed in (1943)
that the height of the least ordinal sufficient for a proof of the consistency
of such a fragment corresponds with the quantifier complexity of that
theory’s class of inductive formulas, and that transfinite induction to any
smaller ordinal is provable in the fragment. So the number of quantifiers
over which mathematical induction is permitted equals the number of
exponentials needed to express the ordinal that measures the theory’s
consistency strength. One quantifier equals one exponential. Thus the
theory of constructive proof transformations has turned up a precise
mathematical analysis of the complexity of natural language quantifiers, a
remarkable realization of Hilbert’s original ambition.
Why has so little attention been given to this result? The discussion of
Gentzen’s work has been dominated by debate about whether or not the
proof of PA’s consistency can really shore up our confidence in this theory.
To anyone who has witnessed a talk about ordinal analysis sabotaged by
this debate, the scene will be familiar: someone reminds us that the proof
uses a principle that extends the resources of PA. Someone else defends the
principle despite this fact and points out that in every other way the proof
is extremely elementary compared to the full strength of PA. Because the
two theories PA and PRAþti06 are in this basic way incomparable, the
jury is out as to the gains made by reducing the consistency of one to that
of the other. From the point of view of logic, however, this is all a
distraction from what Gentzen actually achieved: he showed that the
question of the consistency even of elementary theories can be formulated
as a precise problem, and he showed that the solution to this problem
requires new perspectives and techniques and carries with it unexpected
insights about logical complexity. If philosophers did not harbor skepti-
cism about PA, then they would likely not be interested in Gentzen’s result

6
The theory of primitive recursive arithmetic extended with a principle of transfinite induction to  0.
120 Curtis Franks
one way or the other. Their disinterest in the analogous results about
fragments of PA is just evidence that they harbor no skepticism about these
theories. We recognize Gentzen’s analysis of first-order quantifiers as one
of the deepest results in the history of logic as soon as, and no sooner than,
we stop believing.
I now wish to respond more directly to the objection that opened this
section. Of course it is true that in the study of logical systems one must
engage in reasoning of some sort or another. This reasoning can possibly
be described by one or a few select logical systems. But why should anyone
assume that this amount of reasoning is anything more than ways of
thinking that have become habitual for us because of their proven utility?
Further, why should anyone assume that there is any commonality among
the principles of inference we deploy at this level over and above the fact
that we do so deploy them?
To expand on the first of these points, it may be helpful to draw an
analogy between rudimentary logic and set theory. Often it is thought that
decisions about which principles should govern the mathematical theory of
sets should be made by appealing to our intuitions about the set concept
and even about the cumulative hierarchy of sets. Doubtless such appeals
have figured centrally in the development of set theory. But the history of
the subject suggests that a complete inversion of this dynamic has also been
at play. Kanamori (2012: 1) explains:
[L]ike other fields of mathematics, [set theory’s] vitality and progress have
depended on a steadily growing core of mathematical proofs and methods,
problems and results . . . from the beginning set theory actually developed
through a progression of mathematical moves, whatever and sometimes in
spite of what has been claimed on its behalf.

Whereas one can today find endless phenomenological and metaphysical

justifications of, for example, the replacement axiom, Kanamori contends
that set theory in fact evolved primarily by absorption of successful
techniques, like transfinite induction, devised to answer mathematical
questions. “It was von Neumann’s formal incorporation of this method
into set theory, as necessitated by his proofs, that brought in Replacement”
(33). In similar fashion, the power set existence assumption, which origin-
ally had many detractors, was not finally embraced in the wake of any
argument or philosophical insight. It merely happened that “iterated
cardinal exponentiation figured prominently” in Kurepa’s proofs in infini-
tary combinatorics, “so that shedding deeper concerns the power set
operation became further domesticated” (46). The upshot? “Set theory is
 Logical nihilism 121
a particular case of a field of mathematics in which seminal proofs and
pivotal problems actually shaped the basic concepts and forged axiomatiza-
tions, these transmuting the very notion of set” (1).
The point is not that we have no reliable intuitions about the set
concept, nor even that they should play no role in the development of
formal set theory. The point is that those intuitions have evolved partly in
response to our need to make sense of routinely counter-intuitive scientific
discoveries. If our intuitions have been at least partially, perhaps largely,
shaped by developments in logic, then the fact that we appeal to them on
occasion in order to refine our definitions and techniques or in order to
choose new axioms seems to lose its significance. Rather than develop a
theory of sets that unpacks the fundamental truths we intuit, we have
developed an intuition about sets that makes sense of mathematically
interesting relations we have discovered.
I need not argue that rudimentary logic has taken shape in a similar
fashion first because Professor Maddy has been persuasive on this same
point in her contribution to this volume and second because it does not
really matter for my purposes whether, in the end, this is true. I am
content simply to reveal the picture that holds us captive when we begin
to think about logic. For the objection we solicited was that if our modern
science of logic thrives on an unprejudiced consideration of the full gamut
of properties exhibited by logical systems and the relations among them, so
that issues of correctness do not arise, then it has only smuggled those
issues in through the back door, in the metalogic that makes the science
possible. But it is a preconception that science is made possible by
ahistorical norms of right reasoning. Once one considers the possibility
that logic may be studied with patterns of thought adapted to what we
learn along the way, it becomes hard to understand what special status the
rudimentary principles we find ourselves reflexively appealing to are sup-
posed to have.
This brings us to the second of the points above, the idea that the
principles that have found their way into our basic toolkit must presum-
ably have some features in common that led them there. Even if these
principles have been adopted over the course of time, the idea goes, there
must be some reason for their being adopted instead of other principles.
Perhaps this reason can be repackaged as an explanation of their being the
true logical principles.
In response to this suggestion, I wish only to expose the presupposition
driving it. Whoever said that there must be some property of logical
validity that some principles of inference enjoy and others do not? If we
122 Curtis Franks
knew in advance of all evidence that some such property attains, then it
might be reasonable to look for it in whatever classes of inference rules we
happen to find collected together. But we have no such foreknowledge
and, in fact, the evidence suggests that the arrangement of our toolkit is a
highly contingent matter. Are we not better off shedding this vestigial
belief that among all the intricate and interesting consequence relations out
there, some have a special normative status? Can we not get by with the
understanding that principles of inference with a rather wide range of
applicability differ from those suited only to specific inference tasks only in
having a wider range of applicability? Had this been the understanding
that our culture inherited, would anything we have learned from studying
logic lead us to question it?
At the end of Plato’s Phaedrus, Socrates explains that prior to investi-
gating the essence of a thing, it is important to devise an extensionally
adequate definition of that thing so that we will be in agreement about
what we are investigating. This attitude seems right to me, and it seems to
me that the familiar debates about, for example, where logic leaves off and
mathematics begins violate this principle unabashedly. Suppose it were
clear to everyone that some but not all patterns of reasoning are inescap-
able and furthermore that it were easy to tell which these are. Then we
would have good reason to label this “logic” in distinction to patterns of
reasoning that we all recognize as the province of some special science or
particular application. It would be reasonable to wonder what accounts for
the privileged role that these principles play in our lives. As things stand,
however, many of us seem instead to assume that there simply must be
patterns of reasoning that differ in kind from others. Typically, our minds
are already made up about the psychological or metaphysical circumstances
that underwrite this difference. This is what Wittgenstein stressed with his
observation that “the crystalline purity of logic was not the result of an
investigation,” that instead “it was a requirement” (Wittgenstein 1953:
§107). Driven by this assumption, we thrash about looking for some
extensional definition that we can hang our ready-made distinction on.
These definitions are simply unconvincing on their own. They can satisfy
only people who cannot tolerate the thought that there is no line to
be drawn.
When Gentzen began his study of logic, he parted ways with his
predecessors7 by not first defining logical validity and then seeking out
logical principles that accord with that definition. He simply observed that
7
There are historical precedents for Gentzen’s attitude: Aristotle, Condillac, Mill.
 Logical nihilism 123
some patterns of reasoning abound. Consider that to define “predicate
logic” he said only that it “comprises the types of inference that are
continually used in all parts of mathematics” (1934–35: 68). This homely
definition set Gentzen on a task of empirically tallying the techniques used
in mathematical proofs, ignoring those that are unique to geometry,
arithmetic, and other specific branches of mathematics.8 Of course such
a survey is by no means guaranteed to be exhaustively executable, and it
was Gentzen’s good fortune that his subject matter happens to exhibit few
instances.
One should not, however, write off his success as purely a matter of
luck. Gentzen devised an ingenious argument to the effect that his tally
was in fact exhaustive. This involved constructing an innovative type of
logical calculus that is at once formal and patterned on the informal
reasoning recorded in mathematical proofs. This scheme enabled him to
do more than construct a serial tally of proof techniques, because the
inference types identifiable with it are extremely few and systematically
arranged so that one can be sure that none have been overlooked. After all,
if there are any inference types that went unnoticed, then for that very
reason they fail to meet the criterion of ubiquity in mathematical practice
that Gentzen imposed.
This empirical “completeness proof ” bears little semblance to familiar
conceptions of logical completeness and is interesting for this reason.
I mention it now only to draw attention to the fact that while Gentzen’s
definition of predicate logic does pick out a well-defined body of infer-
ences,9 he did not concoct the definition in the service of a preconceived
notion of logical validity. He did not, for example, first stipulate a semantic
notion of logical consequence based on his own intuitions and then ask
whether his calculi adequately capture this notion. Gentzen simply pro-
posed that the intuitions guiding mathematicians in their research would
be worth isolating and studying, and he therefore modeled a logical
calculus on the inferences mathematicians actually make.10
Of course mathematicians also deploy proof techniques that are less
universal, and the only observable difference between these and the ones
that meet Gentzen’s criterion of ubiquity is their relative infrequency.
Mathematicians do not report a feeling that arithmetical reasoning is less

8
This is explicit in (Gentzen 1934–35) and even more vividly depicted in section 4 of (Gentzen 1936).
9
Actually Gentzen vacillated over the inclusion of the principle of mathematical induction,
ultimately deciding against it.
10
For details of this conception of completeness, see section 3 of (Franks 2010).
124 Curtis Franks
valid or valid in some other way than general reasoning, and even if they
did, we should be inclined to ignore these reports if they did not reflect in
mathematical practice. For these reasons Gentzen could never bring him-
self to describe the distinction between inference rules that appear in the
predicate calculus and those that belong specifically to arithmetic as a
distinction between the logical and the non-logical. He only thought that
he had designated a logical system, one that by design encodes some of the
inferences he was bound to make when reasoning about it but whose
logical interest derives solely from what that reasoning brings to light.
Contrast this with one of the more famous attempts to demarcate the
logical: Quine’s defense of first-order quantification theory. Second-order
quantification, branching quantifiers, higher set theory, and such can each
be dissociated from logic for failing to have sound and complete proof
systems, for violating the compactness and basic cardinality theorems, and
other niceties. There is even a (1969) theorem, due to Lindström, to the
effect that any logic stronger than first-order quantification theory will fail
to exhibit either compactness or the downward Löwenheim–Skolem the-
orem. Second-order logic, Quine concluded, is just mathematics “in
sheep’s clothing” because by using second-order quantifiers one is already
committed to non-trivial cardinality claims (Quine 1986: 66).
How true will these remarks ring to someone who doesn’t know in
advance that they are expected to distinguish logical and mathematical
reasoning? Quine’s consolation is telling: “We can still condone the more
extravagant reaches of set theory,” he writes, “as a study merely of logical
relations among hypotheses” (Quine 1991: 243). I should have thought that
this accolade, especially in light of the intricate sorts of logical relations that
set-theoretical principles bear to one another and that set theory bears to
other systems of hypotheses, would be used rather to enshrine a discipline
squarely within the province of logic. For if we never suspected that among
the plenitude of logical relations are a privileged few that capture the true
inter-dependencies of propositions, what else would we mean by “logic”
than just the sort of study Quine described?
As to the properties that characterize first-order quantification theory, it
should now go without saying that from our perspective Lindström’s
theorem, far from declaring certain formal investigations extra-logical,
exemplifies logic. So too do results of Henkin (1949) and others to the
effect that second-order quantification theory and first-order axiomatic set
theory each are complete with respect to validity over non-standard
models. For a final example, I can think of none better than the recent
result of Fan Yang that Väänänen’s system of dependence logic (with
 Logical nihilism 125
branching quantifiers), extended with intuitionistic implication (the con-
text sensitive and constructive operator that results from the denial of lem),
is equivalent to full second-order quantification theory (Yang 2013).
Wittgenstein (§108) expected his readers to recoil from the suggestion
that we shed our preconceived ideas by “turning our whole examination
around.” Rather than impose our intuitions about logic on our investi-
gations by asking which principles are truly logical, let us first ask if a close
look at the various inference principles we are familiar with suggests that
some stand apart from others. If some do, then let us determine what it is
that sets them apart. But the question he puts in our mouths – “But in that
case doesn’t logic altogether disappear?” – suggests that we know deep
down that our empirical investigation is bound to come up empty. Various
criteria will allow us to demarcate different systems of inference rules to
study, but when none of these indicate more than a formal or happen-
stance association we will find ourselves hard pressed to explain why any
one of them demarcates “the logical.”
I am more optimistic than Wittgenstein. The conclusion that I expect
my reader to draw from the absence of any clear demarcation of the logical
is not that there is no such thing as logic. Let us agree instead that no one
part of what logicians study, contingent and evolving as this subject matter
is, should be idolized at the expense of everything else. Logic outstrips our
preconceptions both in its range and in its depth.

4. Conclusion
Traditional debates about the scope and nature of logic do not do justice to
the details of its maturation. In asking whether certain inferential practices
are properly logical or more aptly viewed as part of the special sciences, for
example, we ignore how modern logic has been shaped by developments in
extra-logical culture. Similarly, questions about whether logic principally
traces the structure of discursive thought or the structure of an impersonal
world presuppose a logical subject matter unaffected by shifts in human
interest and knowledge.
I mean, by saying this, not just to suggest that the principles of
rudimentary logic are contingent, not different in kind from principles
that we use only some of the time or very rarely and only for specific tasks.
I do urge this attitude. But the caution against mistaking our default,
multi-purpose habits of reasoning for something monumental is only
preparation for a second, more valuable reaction. One should warm up
to the trend of identifying logic with the specialized scientific study of the
126 Curtis Franks
relationships among various systems and their properties. This is, after all,
how logicians use the word. Our preference to ignore questions about a
logic’s correctness stems not only from an interest in exploring the proper-
ties of possible logical systems in full generality but also from an appreci-
ation, fostered by the study of logic, that no one such system can have all
the properties that might be useful and interesting.
In closing, let me re-emphasize that the idea of a true logic, one that
traces the actual inter-dependencies among propositions, is unscathed by
all I have said. Part of the difficulty in questioning that idea is that it is a
moving target: argue against it, you feel it again in that very argument;
close the door, it will try the window. But this very circumstance only
underlines the fact that the idea is a presupposition, nothing that
emerges from any discovery made in the study of logic. For the same
reason that we can marshal no evidence against it, we see that if we can
manage to forget it our future discoveries will not reveal to us that we
have erred.
This realization, coupled with the observation that a fixation on the true
logical relationships “out there” hinders the advancement of logic, certainly
recommends nihilism on practical grounds. The question that remains is
whether we are capable of sustaining a point of view with no direct
argumentative support.
The proper antidote to our reflexive tendencies will surely extend an
analysis of modern logic and include a rehearsal of the subject’s history.
I cannot offer that here.11 I can only mention that logic as a discipline has
evolved often in defiance of preconceived notions of what the true logical
relations are. Logic has been repeatedly reconceived, not as a fallout from
our better acquaintance with its allegedly eternal nature, but in response to
the changing social space in which we reason. There is reason neither to
expect nor to hope that logic will not be continually reconceived. Such
reconceptions have been and likely will again be fundamental, so that what
makes the moniker “logic” apt across these diverse conceptions is not an
invariable essence.
In these pages I have indicated instead logic’s modern contours, high-
lighting the fact that the deepest observations logic has to offer come with
no ties to preconceptions about its essence. The richness of logic comes
into view only when we stop looking for such an essence and focus instead
on the accumulation of applications and conceptual changes that have

11
For the details of the evolution of one central concept – that of logical completeness – in the past
two centuries, see (Franks 2013).
 Logical nihilism 127
made current logical investigations possible. The study of logic might be
the best practical antidote to the view of it that we have inherited.
In his Logic of 1780, Condillac wrote: “People would like to have had
philosophers presiding over the formation of languages, believing that
languages would have been better made. It would, then, have required
other philosophers than the ones we know” (237–8). Our interest in a
better made language is an interest in a language that traces a pre-existing
logical structure. Like Condillac, Wittgenstein warned that presupposing
such a structure fosters dismissive attitudes about the languages we have:
“When we believe that we must find that order, must find the ideal, in our
actual language, we become dissatisfied with what are ordinarily called
‘propositions,’ ‘words,’ ‘signs’” (§105). When we stop believing for a
moment, as the discipline of logic demands we do, the structures we find
immanent in our several, actual languages command our interest more
than anything we could have devised in the service of our ideal.
 chapter 7

Wittgenstein and the covert Platonism

of mathematical logic
Mark Steiner

(I use the following abbreviations: PI = Philosophical investigations

(Wittgenstein 2009), RFM = Remarks on the foundations of mathematics
(Wittgenstein 1978), LFM = Wittgenstein’s lectures on the foundations of
mathematics, Cambridge, 1939 (Wittgenstein 1976), PG = Philosophical
grammar (Wittgenstein and Rhees 1974).)
By the end of the 1930s, Wittgenstein’s thought on mathematics had
undergone a major, if often undetected, change.1 The change had to do
with the relationship between arithmetic, including elementary number
theory and geometry,2 and empirical regularities, including behavioral
regularities that are induced by training. During the first part of the
decade, Wittgenstein continued to regard mathematical theorems as akin
to grammatical rules. As such, there was no need to seek a general theory of
mathematical applicability, as Frege did.3 “The applications,” he repeated,
“take care of themselves.” (E.g. PG, III, 15: 308.) After all, grammatical
rules have no applications outside grammar itself, being norms, not
descriptions of nonlinguistic objects or processes. This does not imply,
to be sure, that the environment in which language operates has no effect
on which rules we use in language to describe the environment. In a
1939 lecture at Cambridge (LFM XX: 194) Wittgenstein remarked “there
is, in all the languages we know, a word for ‘all’ but not for ‘all but one’.”

1
See (Steiner 2009).
2
In this chapter, for the most part, I will not address the complicated question of to what extent
Wittgenstein’s ideas were intended to describe advanced mathematics, and to what extent he actually
succeeded in describing advanced mathematics. Hence, we will focus upon arithmetic and
elementary number theory, and Euclidean geometry.
3
As Dummett points out (Dummett 1991b), for arithmetic, Frege’s theory of application involved (a)
rendering all arithmetic statements in second-order logic, universally quantified, where the predicate
letters range over “concepts.” Then to apply an arithmetic proposition, all one needs to do is to
perform universal instantiation, replacing each predicate variable with a constant predicate that
expresses a particular concept. To apply arithmetic to empirical situations, for example, all we
need to do is instantiate empirical predicates for the universally quantified second-order variables.
Note that mathematical applicability is in this account the same thing as logical applicability.

128
 Wittgenstein and the covert Platonism of mathematical logic 129
He certainly meant to say that this is true because in our world it is most
convenient to have a universal quantifier, not because logic is itself empirical.
I believe that this part of what he says there, though not all of it, can be
attributed to him well before 1939.4 Even in 1939, Wittgenstein told the class:
To say “A reality corresponds to ‘2 þ 2 = 4’” is like saying “A reality
corresponds to ‘two’.” It is like saying a reality corresponds to a rule, which
would come to saying: “It is a useful rule, most useful – we couldn’t do
without it for a thousand reasons, not just one.”5 (LFM: 249)

Such a view of mathematics places it on a par with the rules of logic. Both are
grammatical rules, the difference being which vocabulary the rules govern.
This is not to say, with Frege and Russell, that mathematics is logic. The
rules of logic are used to prove mathematical theorems, to be sure, but this
does not make mathematics into logic: logic is used in every discourse.
During the period 1936–7, Wittgenstein began to study in earnest the
concept of rule-following which was to loom so large in his Philosophical
investigations. The connection between rules and regularities (Regelmäßigkeit)
becomes manifest to those who study his notebooks.
Rules are norms which evaluate what happens or what is done by people;
regularities are what happen most of the time, or what people do most of
the time – when they are trained the same way. Rules label the deviations
from these regularities “mistakes,” “abnormalities,” “perturbations”6 (in

4
Wittgenstein goes on to say:
This is enormously important: this is the sort of fact which characterizes our logic. “All but one”
seems to us a complex idea – “all”, that’s a simple idea. But we can imagine a tribe where “all but
one” is the primitive idea. And this sort of thing would entirely change their outlook on logic.
This further idea, expressed, as I say, in 1939, I would not want to attribute to Wittgenstein in the
early 1930s (which is what I am discussing here) – and I will discuss it below, in the context of (what
I will call) his revolution of 1937. It is reminiscent of Nelson Goodman’s relativism concerning
natural kinds (“grue”).
5
The idea that the relationship between the empirical world and mathematical propositions is that the
former makes the latter useful is not replaced in 1939, but augmented by a much deeper connection
between mathematical propositions and empirical reality, which we will discuss later.
6
I don’t mean to say that the mathematical technique of perturbation theory is “normative.” I bring
the subject of perturbations in because Wittgenstein himself does:
Suppose we observed that all stars move in circles. Then “All stars move in circles” is an
experiential proposition, a proposition of physics – Suppose we later find out they are not
quite circles. We might say then, “All stars move in circles with deviations” or “All stars move
in circles with small deviations.” (LFM IV: 43)
If it is a calculation we adopt it as a calculation – that is, we make a rule out of it. We make
the description of it the description of a norm – we say, “This is what we are going to
compare things with.” It gives us a method of describing experiments, by saying that they
deviate from this by so much. (LFM X: 99)
130 Mark Steiner
physics) and the like. There are a number of possible explanations of the
utility of stigmatizing deviations in this way, at least in the areas of language,
logic, and mathematics.7 Society has an interest in rendering certain prac-
tices as uniform as possible, and adding negative and positive incentives
may do the trick.8 This account is plausible in the areas of language and
mathematics, which is our topic here.
The so-called “rule-following paradox,” as Wittgenstein himself labels it
in PI, is (and is intended to be) a paradox only for academic philosophers.
I9 use this term to refer to those who take the goal of philosophy to
explain10 human practices like rule-following – i.e., almost all philosophers
besides Wittgenstein. The explanations emerge from diverse philosophies,
from mentalism to physicalism, but all agree that there must be some fact
about a person, beyond the regularities of his behavior, in virtue of which
we can say he is following a specific rule at a specific time and not another.
The explanation that follows here follows that of Saul Kripke, though,
as will become clear, it differs from his in some crucial details.11 Kripke
attributes a “skeptical argument” concerning rule-following to Wittgen-
stein, and has drawn much criticism on this account. I agree with Kripke
that Wittgenstein did construct a skeptical argument, but I hold that the
argument is supposed to be valid only for academic philosophy, as distinct
from Wittgenstein’s own philosophy.

7
Wittgenstein never attempted to found an account of ethics on this basis – there are regularities in
the way people treat one another, and moral norms arise from stigmatizing deviational behavior –
and it is an interesting question why.
8
“When someone whom I am afraid of orders me to continue the series, I act quickly, with perfect
certainty, and the lack of reasons does not trouble me.” (PI: 212) I believe that this passage reflects
actual occurrences in Wittgenstein’s life. After the publication of the Tractatus, Wittgenstein left
academics and went into school teaching. Ray Monk (Monk 1991, pages 195–196, 232–233) reports
that Wittgenstein used to inflict corporal punishment on his pupils if he thought they were not
applying themselves to the arithmetic lessons he was giving them. Not enough has been said about
the connection between the rule-following arguments in Philosophical investigations and
Wittgenstein’s short-lived experience as a schoolteacher, which came to an end, when one of his
pupils lost consciousness as a result of being struck by Wittgenstein.
9
Wittgenstein himself referred to “academic philosophy” in a letter, but not in his published or (so
far as I know) unpublished works. Felix Mühlhölzer draws my attention to the following passage
from Zettel, 299:
We say: “If you really follow the rule in multiplying, it MUST come out the same.” Now,
when this is merely the slightly hysterical style of university talk, we have no need to be
particularly interested. . .
10
Wittgenstein condemns this kind of philosophy in one of the most famous passages in Philosophical
investigations: “We must do away with explanation and description must take its place.”
(Philosophical investigations, 109; in the 4th edition this is translated: “All explanation must
disappear. . .”)
11
Cf. (Kripke 1982).
 Wittgenstein and the covert Platonism of mathematical logic 131
Since it turns out that there is no such “fact” which can serve as an
explanatory criterion for rule-following, academic philosophers are faced
with a paradox since it now follows that there is no such thing as rule-
following at all.12 This could be considered a skeptical argument, though an
ad hominem one.13
Let us now examine the arguments Kripke brings to support the
contention (on behalf of Wittgenstein) that there is no fact in virtue of
which somebody is following a rule. Kripke adduces two arguments
for this, of which only one is actually in Wittgenstein. This is the
“normative” argument: to ascribe rule-following to someone is to assert
that someone is acting according to a norm, i.e. following the rule
“correctly” – as we say. The gap between “is” and “ought” then implies
that no fact or state of the person at time t could be identified with
following the rule at t. The situation is different in other cases of
explanation and reduction in science.14 Disposition terms (e.g. “solubil-
ity”) can in principle be reduced to “state descriptions” of a substance,
which actually replace the disposition term. This is the so-called “place
holder” theory of dispositional terms. Given the normative nature of
rule-following, i.e. its social nature, it applies to what interests society:
behavior. Thus it cannot be reduced to, or identified with, but only
correlated with, an underlying state description either of the mind or of
the brain; that is Wittgenstein’s argument.
Kripke has another argument, the so-called “infinity” argument,
according to which any state of the brain,15 for example, is necessarily
finite; while rule-following commits the trainee to infinitely many

12
The case is formally similar to Frege’s foundation of arithmetic upon “logic.” When Russell’s
paradox showed that Frege’s “logic” is inconsistent, Frege overreacted by saying “arithmetic
totters.”
13
Kripke compares Wittgenstein’s “skeptical argument” to that of David Hume, and the comparison
is just, but not in the way that Kripke imagines: both Wittgenstein and Hume use skeptical
arguments to dispose of various kinds of academic philosophy, without themselves being skeptics.
As I have argued in the text above, Hume’s skeptical argument disposes of necessary connections
between events, which are used in rationalist explanations of causal reasoning. It is a skeptical
argument only for them, because they hold that, without the necessary connections there is no
causal reasoning at all. See here (Steiner 2009: 26ff ).
14
I am here offering my own opinions, not those of Wittgenstein. In fact, I am not at all sure that
Wittgenstein distinguished clearly between dispositions in science and abilities in humans, since in
Philosophical investigations, 193–194, he claims that academic philosophers make the same kind of
mistakes in discussing the abilities of humans with dispositions of machines. See also Philosophical
investigations, 182, where he compares “to fit” (said of bodies in holes) and “to be able,” “to
understand,” said of humans.
15
As my colleague Oron Shagrir has cogently argued, Kripke seems to be thinking of a brain state as
the physical realization of a finite digital computer.
132 Mark Steiner
applications of the rule – as in the rule “add two.” Not only is this
argument not in Wittgenstein, it couldn’t be, as I will argue below.16
The “rule-following paradox” is, then, a paradox only for academic
philosophy. For Wittgenstein himself there is no paradox to begin with.
For paradox to loom, our ordinary discourse about some topic must be
seen to lead to catastrophe. For example, Zeno’s paradoxes began with
ordinary conceptions of motion and showed that they lead to inconsist-
ency, or to the conclusion that no motion is possible. Wittgenstein’s
account of rule-following involves the claim that all rules are supervenient
upon regularities. In the case of the rule “add 2” which surfaces in
Philosophical investigations – how do we know that our trainee is following
this rule if he manages to go 2, 4, 6, 8, . . ., 1000 – or another rule which
says that after 1000 one starts “adding 4”? We don’t know, but our
experience, both as students AND as teachers, is that almost all who
produce this series go on the same17 way: to 1,002 and not to 1,004.
(The claim is not that we reflect upon these regularities – that would be
a misinterpretation – but that the regularities make our practice in this
regard possible and coherent.) This regularity allows us to attribute the rule
“add two” (the rule which is “hardened” from just this regularity) with
great confidence to our trainee, and to call his response “erroneous” or
perhaps “provocative” if he says next 1,004. (We may filter out frivolous
responses by warning the trainee that he will be severely sanctioned if he
doesn’t give the right answer.) Wittgenstein expressed this idea quite
clearly in 1939:
Because in innumerable cases it is enough to give a picture or a section of
the use, we are justified in using this as a criterion of understanding, not
making further tests, etc. (LFM I: 21)

The position I am attributing to Wittgenstein is not that of Kripke. Kripke

seems to regard as a criterion that our trainee is following the rule þ2, if he
16
In fairness, however, I must add that my own presentation of Wittgenstein’s “normative” argument
that there is no “fact” in virtue of which somebody is following a rule, also “improves” the argument
somewhat. The distinction I draw between disposition terms – which are in principle reducible to
state descriptions – and rule-following ascriptions – which are not – is not in Wittgenstein. On the
contrary, Wittgenstein tends to see rule-following as precisely a disposition, but denies that
disposition terms are reducible to underlying state descriptions. Since I think Wittgenstein is
mistaken here, not only about the issue itself, but also about how to make his own argument
(a malady which many philosophers are prone to), I have made the necessary adjustments. Kripke’s
“infinity” argument, on the other hand, is one which in fact contradicts basic Wittgensteinian
insights and has no place even in an “improved” Wittgensteinian corpus.
17
I will deal below, page 133, with the objection that there is no objective meaning to “the same” and
hence that the claim is circular.
 Wittgenstein and the covert Platonism of mathematical logic 133
goes on to give the same responses that “we” would give, i.e. agreement
with society. On the contrary, society is itself predicated on empirical
regularities of its members. The criterion, then, is simply that our trainee
has successfully followed the rule þ2 up to now. It is true that without
behavioral regularity upon training, this criterion would have no point.
But to apply the everyday criterion, one does not have to reflect on this
regularity, or even know about it.
Wittgenstein argues in Philosophical investigations that the notion of a
rule and that of regularity are picked up simultaneously as a result of our
common training, so that there is no circularity in saying that a rule is
founded on regularity even though detecting a regularity requires an ability
of ours to follow rules. The same training teaches the concept of “the
same.” For this reason our previous statement that “People trained the
same act the same” is not susceptible to skeptical doubt.18
In the passages of Philosophical investigations we are discussing, though
not necessarily in RFM, Wittgenstein is employing a very simple concept
of “applying” a rule. We may understand Wittgenstein as saying that
applying a rule is simply following it (correctly). Since rules are grounded
in regularities, it is the ability to continue the series by doing what almost
everybody does when placed in the same situation, which grounds the
ability to apply the rule. In principle, there is no difference between rules
of logic and grammatical rules. The “hardness of the logical must” is a kind
of projective superstition, much as the superstition that Hume thought he
had exposed in the idea there is “necessary connection” between causes and
their effects. It is similar to the superstition of thinking that when one is
reading he is having a characteristic experience of being “influenced” or
being “guided” by the text. (PI, §§170ff.)
Since rules are norms, there is no equivalence between saying that
somebody is following a rule and saying that his behavior falls under the
underlying regularity. Saying that somebody is following a rule is simply
evaluating his behavior, not describing it – even though the evaluation
results from observing his previous behavior and responding to it in light
of our own training in following rules, and the regularities that are
instilled by that training. In other words, we can describe the criteria
that a teacher is using to evaluate the student’s behavior as successfully
following a rule, even though the teacher may not be aware of these
criteria. In fact one of the purposes of Wittgenstein’s analysis in

18
One could say it is Wittgenstein’s version of a synthetic a priori proposition.
134 Mark Steiner
Philosophical investigations is to unearth these criteria in support of his
view of philosophy as being purely descriptive.
To the question “How does the rule follower himself know that he is
following a rule” Wittgenstein answers that the exclamation “Now I can go
on” is often not a description at all, certainly not of his own mental state,19
though Wittgenstein does not deny that there are such states. It is certainly
no form of self-knowledge, and given what Wittgenstein says in On
certainty (1974), the conviction on the part of the rule follower who
exclaims “Now I can go on” actually rules out knowledge.20
In summation, logic consists of rules governing the use of logical
expressions like “and,” “or,” “if. . .then,” “everything,” etc. As Wittgen-
stein himself put it, even in the 1940s, “The rules of logical inference are
rules of the language-game.” (RFM, VII: 401) There is nothing akin to
“intuition,” “seeing,” and the like in following or producing a logical
argument. Instead we have regularities induced by linguistic training,
which in hindsight are interpreted, or misinterpreted, by us as some kind
of determination. Deviation from this regularity is labeled by society as
“incorrect” reasoning. Wittgenstein’s aim is to demystify logic and logical
necessity, just as Hume’s aim was to demystify causation by eliminating
the alleged “necessary connection” between events. The image of logic as a
kind of “super-physics” is what needs to be debunked. Philosophical investi-
gations contains a number of references to this mystification of logic:
89. With these considerations we find ourselves facing the problem: In what
way is logic something sublime?
For logic seemed to have a peculiar depth, a universal significance.
Logic lay, it seemed, at the foundation of all the sciences.
97. Thinking is surrounded by a nimbus. –Its essence, logic, presents an
order: namely, the a priori order of the world; that is, the order of possibil-
ities, which the world and thinking must have in common.
108. . . .The preconception of crystalline purity [in logic] can only be
removed by turning our whole inquiry around. (One might say: the inquiry
must be turned around, but on the pivot of our real need.)
Wittgenstein devoted a great amount of thought to this topic in LFM.
Concerning the “law of contradiction” (actually the law of “noncontradic-
tion”) he stated:
19
PI, §§151ff. I understand the Private Language Argument of Wittgenstein as saying that what is
called “referring” to our mental states is more like expressing them than naming them.
20
Can one say “Where there is no doubt there is no knowledge either”?
 Wittgenstein and the covert Platonism of mathematical logic 135
Let us go back to the law of contradiction. We saw last time that there is a
great temptation to regard the truth of the law of contradiction as some-
thing which follows from the meaning of negation and of logical product
and so on. Here the same point arises again. (LFM: 211).
The expression “follows from” is circular here, Wittgenstein is pointing out,
since logic itself is the criterion of “what follows.”21 The term “follows from
the meaning,” is incoherent, since meaning is tied to use, and it does not
make sense to speak of “following from” use. The “temptation” of which
Wittgenstein speaks here, is the attraction of the academic philosopher (and
the early Wittgenstein, of course) to a covert Platonism. Wittgenstein’s own
“demystified” view is that logical laws are a special case of rules that are based
on regularities of speakers of language – i.e. rules of “grammar.” As in the
general case of rule following, in which the rules are grounded in regularities,
and nothing more, so are logical laws the “application” of training in rules to
new cases.
In another lecture, Wittgenstein said:
If we give a word one particular partial use, then we are inclined to go on
using it in one particular way and not in another. “Not” could be explained
by saying such things as “There’s not a penny here” or saying to a child
“Must not have sugar” (preventing him). We haven’t said everything but we
have laid down part of the practice. Once this is done, we are inclined
when we go on to adopt one step and not another – for example, double
negation being equivalent to affirmation. (LFM: 242–243)
We can say, then, that logical laws arise in a two step process. First, the
child is trained in the use of words like “not.” The training induces a
regularity in this use, a regularity which society reinforces as “correct”
usage. Within this regularity, however, there arises a subregularity, when
the rules for using “not” are to be applied to special cases like double
negation. Most trainees find themselves using double negation as they
would affirmation. This regularity is then “put in the archives” as a law
of logic.
Something similar happens in arithmetic, according to Wittgenstein. In
applying the rules for division to 1/7, most proficient students find them-
selves repeating the sequence 0.142857142857. . .22 In fact, most proficient
students in dividing m by n always get a finite decimal or a repeating
decimal. This subregularity is then converted into a rule in itself, a law or
21
This is a point that Quine also made during the very same period. See (Quine 1936a), reprinted in
(Quine 1976).
22
See LFM, p. 123.
136 Mark Steiner
proposition of arithmetic, as a result of a proof. This view is in accordance
with Wittgenstein’s arguments in PI (186ff.) that rule-following is
grounded, and grounded only, in the actual behavioral regularities of
individuals and group. The covert Platonist wishes to say more: that, for
example, there is an objective fact of the matter by which the theorem
about repeating decimals is “determined,” in some further – perhaps
metaphysical – sense by the rules for division, once they are accepted.
The search for a fact like that, however, Wittgenstein has argued, collapses
into paradox.
It would be an error, however, to conclude that the only difference
between arithmetic and logic is that they control different vocabularies:
that arithmetic controls numerical terms and logic controls sentential
connectives and quantifiers. The year 1937 saw a revolution in Wittgen-
stein’s view of arithmetic, and mathematics in general: arithmetic propos-
itions remained rules as always – it was the nature of the rules that
changed. Mathematical rules were to govern nonlinguistic practices as well
as linguistic ones.
Arithmetic propositions, in Wittgenstein’s post-1937 thought, are rules
that govern our practice of counting. Geometrical propositions are
rules that govern our practice of measuring.23 Not only does the applica-
tion no longer “take care of itself,” it is the very heart of the mathematical
proposition. The canonical application is now precisely the regularity of
counting or measuring which is “hardened” into a rule. The applications of
arithmetic and geometry are outside mathematics; they are empirical
applications. The applications of logic remain, as before, within logic: an
application of modus ponens, for example, is simply an inference of the
form “If A, then B; A; therefore B.” To see the difference between these
two kinds of applications, consider an example Wittgenstein loves to use:
the game of chess. When we apply the rules of chess, we are only playing
chess. The rules can apply to infinitely many “chess sets,” which are
unlimited in their physical composition and also shapes. However,
although chess is essentially a war game between two “kingdoms,” there
are no applications of chess outside the game itself, even to war itself. The
more abstract theory of games, which is real mathematics, does have such
“external” applications.
Wittgenstein even ventured the idea in 1939 that set theory is not
mathematics at all, because it has only imaginary applications. In the

23
This is not Wittgenstein’s only interpretation of geometry: see (Mühlhölzer 2001) for another one.
 Wittgenstein and the covert Platonism of mathematical logic 137
1940s he stated that the meaning of a mathematical proposition, as well as
of a mathematical concept, is determined by the application.24
In this scheme, proofs bring the mathematician to convert regularities to
rules. Wittgenstein came up with the idea that they do so by being
schematic “pictures” of these nonmathematical regularities, something like
flowcharts in computer science or schematic diagrams in electronics:
You might say that the relation between a proof and an experiment is that
the proof is a picture of the experiment, and is as good as the experiment.
(LFM, VII: 73)
What is interesting is that some of these new rules function as rules
that determine whether previous rules have been followed. (Since, by
Wittgenstein’s rule-following considerations, there is no fact by which
the previous rule has been followed, the idea as such does not harbor
any contradiction. For Wittgenstein, the “rule-following paradox” is
not only not a paradox, but it bolsters his account of mathematics.)
Consider again the theorem that:
1/7 = 0.142857142857 and so on ad infinitum (a repeating decimal)
One might think that the infinite expansion of 1/7 is determined25 from
the beginning by the rules for division that are learned in school (or were
once learned in school). But the rules cannot outstrip the regularities that
are their basis, and the regularities, being regularities of human beings
cannot go on forever, and in fact, at some finite point, the regularities will
peter out: the deviation will increase to the extent that no rule could be
founded on human practice.
Mathematics to the rescue of mathematics: the theorem gives a sche-
matic picture of doing the division. Using a pigeonhole principle it is
“clear” that the algorithm will run out of remainders, and thus that the
24
Wittgenstein asserts:
It is the use outside mathematics, and so the meaning of the signs, that makes the sign-game
into mathematics. (RFM, V: 2)
Here we have the extreme anti-formalist statement that the applications of mathematics give
meaning to its language.
In case the message has been missed, Wittgenstein relays it again at once:
What does it mean to obtain a new concept of the surface of a sphere? How is it then a concept
of the surface of a sphere? Only insofar as it can be applied to real spheres. (RFM, V: 4)
25
Wittgenstein recognizes a number of meanings for the concept of “determination” in mathematics,
and some of them he might regard here as innocuous. See PI, 189. I thank Felix Mühlhölzer for this
reference. Compare also LFM, p. 28.
138 Mark Steiner
first remainder, 3, will recur, and thus that the whole cycle will start again.
This induces the mathematician and the rest of us to label as “wrong” any
calculation which does not lead to a repeating decimal; it overrules the
naïve use of the school rules. This kind of proof is characteristic of
mathematics:
It is just the same with 1:7 = 0.142857142. . . You say, “This must give
so-and-so.”
Suppose it doesn’t.
Suppose what doesn’t?
Here I am adopting a new criterion for seeing whether I divide this
properly – and that is what is marked by the word “must”. But it is a
criterion which I need not have adopted. For just as bricks measured with
all exactness might give a curve (‘space is curved’), so 1 : 7 = 0. . . . looked
through with all exactness might give something else. But it hardly ever does
[my italics – i.e., we have noticed an empirical regularity]. And now I’ve
made up a new criterion for the correctness of the division. And I have
made it up because it has always worked. If different people got different
things, I’d have adopted something different. (LFM XIII: 129)

In fact, Wittgenstein remarked in RFM that one should not regard

calculations with very large numbers as simple applications of the rules
for the operations which we learned on small numbers:
We extend our ideas from calculations with small numbers to ones with
large numbers in the same kind of way as we imagine that, if the distance
from here to the sun could be measured with a footrule, then we should get
the very result that, as it is, we get in a quite different way. That is to say, we
are inclined to take the measurement of length with a footrule as a model
even for the measurement of the distance between two stars. (RFM,
Part III: 147)

The reader should not be surprised to find Wittgenstein in a somewhat

ambivalent attitude towards “finitism”:26
If one were to justify a finitist position in mathematics, one should say just
that in mathematics “infinite” does not mean anything huge. To say
“There’s nothing infinite” is in a sense nonsensical and ridiculous. But it
does make sense to say we are not talking of anything huge here. (LFM: 255)

26
By “finitism” Wittgenstein always means what is now called “strict finitism,” according to which it
is incorrect or false to assert “There are infinitely many natural numbers.”
 Wittgenstein and the covert Platonism of mathematical logic 139
We now can see where Kripke went wrong in attributing the “infinity”
argument to Wittgenstein. The argument was supposed to defeat the idea
that “following a rule” is identifiable with some (perhaps dispositional)
state of the brain. When we say that somebody is following the rule “þ2”
or even “plus,” we are saying that he is committed to infinitely many
(correct) responses to the question, “What is . . . þ2”? But the brain, being
finite, cannot produce infinitely many answers to questions of this kind.
Kripke discusses a number of possible responses to this argument and finds
fault with them all. He does not realize, however, that the major premise of
his argument is in direct conflict with a basic feature of Wittgenstein’s
account of arithmetic: the idea that adopting an algorithm like “plus”
determines in some physical, mental, or metaphysical way one’s response
to infinitely many exercises is nothing but covert Platonism, in many ways
worse than the Platonism of objects.
These reflections reflect on the application of logic to arithmetic. By the
application of logic to arithmetic I mean simply the substitution of
arithmetic propositions in the variables (or schematic letters, if you prefer)
of logical rules or “truths.” Consider the law of the excluded middle, a law
of the “Propositional Calculus,” p_  p. An application of this would
be: Either the Goldbach conjecture is true or its negation is true. The
Goldbach conjecture states that every even number greater than 2 is the
sum of two primes (e.g. 8 = 5 þ 3). The conjecture has been shown to hold
for very large numbers, and there are corollaries of the conjecture which
have been proved. But no proof of the full conjecture has been given,
though most mathematicians are persuaded that it is “true.” (There are
pseudo-probabilistic arguments for this, based on the fact that as the
numbers get larger, the “probability” that a given number can be parti-
tioned into two primes rises monotonically, since the number of the
partitions themselves rises.)
The intuitionists hold that it is a form of metaphysics to assert the law
of excluded middle for such a case. To assert it here is to presuppose that
the natural numbers form a closed totality, or what Aristotle called an
“actual infinite,” so that we can say that either there is, or is not, a
counterexample to the Goldbach conjecture in this closed totality. If we
think of the natural numbers through the metaphor of “becoming,”
rather than “being,” then the present absence of a proof or of a refuta-
tion of the Goldbach conjecture means only that the truth of the
conjecture is not determined, and the law of the excluded middle cannot
be asserted. As an invalid rule of inference, it is thus banished from
classical mathematics.
140 Mark Steiner
Let us now apply Wittgenstein’s ideas to the Intuitionist program.
Wittgenstein agrees entirely with the Intuitionist critique of the law of
excluded middle. For the Goldbach conjecture to be “true” in the sense of
classical mathematics, we have to say that the operations of arithmetic
determine in advance that every even number, no matter how large, can be
partitioned into two primes. Wittgenstein agrees that this is not math-
ematics, but metaphysics: a statement like this cannot be grounded on the
behavioral regularities inculcated in grade school. A statement true of all
the natural numbers can be based only upon a theorem – which lays down
a new norm (on the basis of a proof ) which labels any deviation from
the Goldbach conjecture a “mistake.” Hence, Wittgenstein agrees with the
Intuitionists that one cannot regard the law of excluded middle for the
Goldbach conjecture as a theorem of mathematics. It cannot be regarded
as the “hardening” of a regularity.
How, then, are we to square this with Wittgenstein’s explicit disavowal
of Intuitionism (“Intuitionism is all bosh,” he said, “entirely” (LFM
XXIV: 237))?
There are two explanations available. The first has to do with the
connection of Intuitionism with . . . intuition. Brouwer writes as if the
numbers themselves are mental constructions, and the law of excluded
middle does not apply to mental constructions, which can never produce a
closed totality. Wittgenstein is an implacable opponent of the concept of
mathematical intuition – he believes, among other things, that it has no
explanatory value, and hence its only rationale fails. From this point of
view, Intuitionism is a form of mentalism, the other side of the coin from
Platonism. Both are unacceptable foundations of mathematics.
It should be noted, however, that Michael Dummett (Dummett 1975)
championed a non-metaphysical version of Intuitionism, one which has
little or nothing to do with mathematical intuition. According to this point
of view, which is presumably heavily influenced by Wittgenstein’s
thought, truth in general is associated with “assertibility.” And since
mathematical propositions are asssertible only when provable, Dummett
thinks,27 one cannot assert an instance of the law of excluded middle at
time t unless we can show at t that one of the two alternatives can be
proved. Thus a proof of the following form is invalid at t, despite the
acceptance of it by almost all mathematicians:

27
I actually deny this, and have given examples of mathematical propositions that were assertible even
when there was no proof of them in (Steiner 1975). But I will take for granted that Wittgenstein
agrees with Dummett on this point, an agreement that has a solid basis in the corpus.
 Wittgenstein and the covert Platonism of mathematical logic 141
If the Goldbach conjecture is true, then T
If the Goldbach conjecture is false, then T
Therefore, T.
For Wittgenstein, to blacklist mathematical theorems, on the basis of the
Intuitionist attack on the law of excluded middle means: to revise math-
ematics on the basis of a philosophical argument. Wittgenstein is quite
explicit on this point in one of the most famous passages of Philosophical
investigations: no mathematical discovery is relevant to philosophy, and no
philosophical argument can revise accepted mathematical practice. Phil-
osophy describes practice; and the only reason we need philosophy is that
we have a strong tendency to misdescribe it (i.e., practice).
We now seem to have reached an impasse: Wittgenstein upholds the
behavioral/empirical basis of the mathematical propositions, or rules. At
the same time he refuses to revise mathematical practice on the basis of
Dummett’s arguments, themselves based on Wittgensteinian ideas!
The resolution of this “paradox” is based on another Wittgensteinian
idea: that mathematics is a “motley”28 of proofs. The idea that mathemat-
ical theorems are “hardenings” of regularities was never meant to be a
characterization of the “essence” of mathematics. The philosopher who
emphasized so strongly the idea that the referents of certain terms (and
really all terms) are related only by a “family resemblance” did not become
an “essentialist” suddenly when he studied mathematics.
And in fact, Wittgenstein told the students in his 1939 Lectures at
Cambridge that the law of excluded middle in the infinite case (i.e. either
all natural numbers have property P or not all natural numbers have
property P) should be regarded as a “postulate” and was used as such in
mathematics. Presumably the postulate should be judged by its usefulness
in mathematics, though Wittgenstein, ironically, rejected the most cele-
brated attempt (Hilbert 1983) to “justify” the law of excluded middle –
namely, by showing – without using the law of excluded middle – that the
law of excluded middle does not lead to contradiction, when applied to
“infinitary” statements: “Either all numbers have property P, or there is a
number that does not have property P.”
A consistency proof can be compared to theorems to the effect that, in
chess, a forced checkmate is not possible from a certain position. And the
attempt to find one is associated with David Hilbert’s programmatic “On
28
Mühlhölzer protests this translation, which has become entrenched in the philosophical
Wittgenstein discourse, and insists that the right phrase is “multi-colored.”
142 Mark Steiner
the Infinite” – the program which is almost universally thought to have
been refuted by Gödel’s second theorem – which states that arithmetic
cannot prove its own consistency even if the law of excluded middle is
used, to say nothing of the kind of combinatorial, “metamathematical,”
proof Hilbert had in mind. At the risk of digressing, I would now like to
discuss in a little more detail why a consistency proof for Wittgenstein is
not what we are seeking in showing the usefulness of the law of excluded
middle.
Wittgenstein’s rejection of Hilbert’s program had nothing to do with
Gödel’s theorem, which he in any case regarded with suspicion. On the
contrary, he regarded Gödel’s theorems as part and parcel of what was
wrong with the program to begin with – the concept of “metamathe-
matics.” Nor did he regard the search for consistency proofs for math-
ematics as having anything to do with showing the usefulness of the
“postulate” of the law of excluded middle as he saw it.
Wittgenstein’s discussion of contradictions and consistency is of a piece
with his theory of rule-following in general.
How do we get convinced of the law of contradiction? – In this way: We
learn a certain practice, a technique of language; and then we are all inclined
to do away with this form – on which we do not naturally act in any way,
unless this particular form is explained afresh to us. (LFM: 206)

He later explained this point this way:

This simply means that given a certain training, if I give you a contradiction
(which I need not notice myself ) you don’t know what to do. This means
that if I give you orders I must do my best to avoid contradictions; though it
may be that what I wanted was to puzzle you or to make you lose time or
something of that sort.

The Law of Contradiction is thus nothing but the hardening of a linguistic

practice into a rule.
From this it follows that the concept of a “hidden contradiction” does not
have a clear meaning.
There is always time to deal with a contradiction when we get to it. When
we get to it, shouldn’t we simply say, “This is no use – and we won’t draw
any conclusions from it”? (LFM: 209)

At this point, Alan Turing (who attended Wittgenstein’s lectures at

Cambridge in 1939) remarked that the problem could arise in the applica-
tions of logic and mathematics.
 Wittgenstein and the covert Platonism of mathematical logic 143
The real harm will not come in unless there is an application, in which case
a bridge may fall down or something of that sort. (LFM: 211)

Wittgenstein responded to Turing’s attack in the following way:

The question is: Why should they be afraid of contradictions inside
mathematics? Turing says, “Because something may go wrong with the
application.” But nothing need go wrong. And if something does go
wrong – if the bridge breaks down – then your mistake was of the kind
of using a wrong natural law. (LFM: 217)

For as long as an actual inconsistency does not turn up, Wittgenstein held,
we need not worry that the “bridges will fall down.” Like any other
mathematical proposition, inconsistency is either a rule, or nothing. As
long as it is not a rule, i.e. a proven theorem, physical applications go on as
before.
But let’s look at this a little closer. Wittgenstein discusses whether a
bridge could fall down because somebody divided by zero. This is certainly
possible; consider the equation x2 = x. 93 percent of precalculus students at
City College of the City University of New York, in a recent test, divided
by x and got the (only) answer x = 1.29 Not knowing about the solution
x = 0 could, in some scenarios, indeed cause a bridge to fall down. Much
more sophisticated cases could be constructed in which somebody does
not know he is dividing by zero.
Is this a case, however, of an inconsistency of a formal system, or is it
just a simple mistake in informal mathematics? One could imagine a case of
teaching students an axiomatic number theory in which cancellation of
zero is possible, in other words an inconsistent system. The students might
not even notice that ac = bc ! a = b yields 1 = 2 if we allow c to be zero,
because they have little cause to divide by zero. But it is hard to think of an
actual case in which a hidden contradiction in a formal axiomatic system
caused “bridges to fall down.”
A good example of this quandary is the theory of quantum electrodynam-
ics (QED), pioneered by, among others, Schwinger and Feynman.30 The
calculations afforded by this theory are remarkably accurate, but nobody
knows how to base the calculations in a consistent axiomatic mathematical
system. In fact, there are mathematical physicists who think it cannot be
done. One reason is as follows. In calculating the probability of events in

29
See for example www.nydailynews.com/new-york/education/cuny-math-problem-report-shows-
freshmen-city-hs-fail-basic-algebra-article-1.418801.
30
I am grateful to Barry Simon and Shmuel Elitzur who helped me with the details.
144 Mark Steiner
quantum mechanics, according to the usual rules, we come upon infinite
integrals. To “renormalize” these integrals, which are intended to be (Feyn-
man 1985) functions of the basic constants of physics, we change the rules:
instead of theoretical values like e, the charge of the electron, we substitute
the observed value of the electronic charge, a value derived from experiment,
not theory. But now we have a new problem. While the new rules work very
well for observed values at the low scales of energy with which we are
familiar, as the energies get higher the observed electronic charge becomes
larger and larger, so that even the new rules are not valid. In fact, mathemat-
ical physicists think that this charge may become infinite at some finite high
energy, so that QED is not defined at all as a universal “theory of light and
matter.”31
For Wittgenstein, this would just show what he was claiming all the
time: that the ideal of a formal system does not fit the reality of mathemat-
ical physics.32 This would be a perfect example of “The disastrous invasion
of mathematics by logic.” (RFM, V: 281) It is also plausible that the
inconsistency which appears in the infinite integrals has its source in the
physics, not the mathematics, exactly as Wittgenstein says. But, further,
the physicists managed to eliminate the troublesome integrals, albeit by
tweaking the rules for calculations in QED by using, as we said, not the
“naked” magnitudes that appear in the “Hamiltonian” of the system, such
as e, but the “dressed” magnitudes as measured in the laboratory. This
artifice works, and no physicist worries that a possible inconsistency (which
is suspected though not proved) could somehow spoil the calculations we
make at familiar energies.
Coming back to the law of excluded middle, we see that the problem is
not that it has no formalist justification in terms of a combinatorial
consistency proof. It is rather that the law of excluded middle cannot be
regarded as a hardened regularity in cases in which we are applying it to a
putative infinite totality. But precisely because of this, there is no direct
comparison possible between empirical observations and mathematical
theorems in this type of proof. That is what Wittgenstein means by a
“postulate.” The justification of such a postulate would be, in Quine’s
pithy words, “where rational, pragmatic.”33 It would seem that Wittgen-
stein, in accommodating classical mathematics and rejecting the intuition-
ist revisionism, ends up where Quine began: in holism.34

31
I am of course referring to the title of (Feynman 1985).
32 33
Tim Chow made this point on the FOM list, on August 15, 2013. (Quine 1953: 46).
34
I am grateful to Felix Mühlhölzer and Penny Rush for their helpful comments.
 part ii
History and Authors
 chapter 8

Logic and its objects: a medieval Aristotelian view

Paul Thom

What is logic about? According to one familiar account logic tells us

when arguments are valid; logic is thus about arguments. On another
account logic tells us which propositions are (unconditionally) neces-
sary; logic is thus about propositions (Smith 2012). Less familiar than
either of these accounts is the Aristotelian tradition of thinking about
logic. Aristotelians have standardly thought of logic as being about
terms, as well as propositions and arguments. Let us call propositions
and arguments, and whatever else logic has been supposed to be
about, the objects of logic. The general question that interests me is:
What are the metaphysical types to which the objects of logic belong?
More specifically, I will look at the way this question has been
addressed in the Aristotelian tradition. I will not be dealing with
answers to our question proposed by Platonists or with the Stoic
concept of lekta.
I use the expression ‘the Aristotelian tradition’ to cover the writings of
Aristotle himself as well as those over time who have broadly sympathised
with his views. The latter include the ancient Greek commentators, a
multitude of medieval logicians writing in Arabic or Latin, and a smaller
number of later thinkers (notably Bernard Bolzano). But my main focus
will be on just one of these, the thirteenth-century philosophical logician
Robert Kilwardby (d. 1279). Kilwardby dealt with our question at some
length, and his discussion is also useful in that it considers several views
other than his own. Let us begin with Aristotle’s own ideas on our
question.

1. Aristotle
There is not much in Aristotle’s own writings that bears directly on our
question. Four passages are noteworthy.

147
148 Paul Thom
First, the Categories makes a remark about statements:
Statements and beliefs, on the other hand, themselves remain completely
unchangeable in every way; it is because the actual thing changes that the
contrary comes to belong to them. For the statement that somebody is
sitting remains the same; it is because of a change in the actual thing that it
comes to be true at one time and false at another. Similarly with beliefs.
Hence at least the way in which it is able to receive contraries – through a
change in itself – would be distinctive of substance, even if we were to grant
that beliefs and statements are able to receive contraries. However, this is
not true. For it is not because they themselves receive anything that
statements and beliefs are said to be able to receive contraries, but because
of what has happened to something else. For it is because the actual thing
exists or does not exist that the statement is said to be true or false, not
because it is able itself to receive contraries. No statement, in fact, or belief
is changed at all by anything. So, since nothing happens in them, they are
not able to receive contraries. (Aristotle 1963: 4a5)

Here Aristotle leaves two positions open: either statements do not change
truth-value at all, or else any change in their truth-value is due to a change
in something external to them, namely the things which the statements
are about.
Second, in the De Interpretatione we find Aristotle apparently proposing
a general semantic theory according to which the meaning of spoken and
written utterances is to be found in the existence of mental items that
somehow correspond to them:
Now spoken sounds are symbols of affections in the soul, and written marks
symbols of spoken sounds. . . . Just as some thoughts in the soul are neither
true nor false while some are necessarily one or the other, so also with
spoken sounds. For falsity and truth have to do with combination and
separation. (Aristotle 1963: 16a2)

Here, the meaning of spoken and written language is derived from ‘affec-
tions in the soul’, and truth and falsity are seen as residing primarily in the
combination or separation of mental items.
Third, there is a remark in the Posterior Analytics which, again, seems to
point to the soul as the locus of truth and demonstration.
By contrast, it is always possible to find fault with ‘external’ arguments (i.e.
spoken or written ones): For demonstration is not addressed to external
argument – but to argument in the soul – since deduction is not either. For
one can always object to external argument, but not always to internal
argument. (Aristotle 1994: 76b23)
 Logic and its objects: a medieval Aristotelian view 149
Finally, there is a remark in Metaphysics Book 6 which again locates truth
and falsity in the soul rather than in external reality:
But since that which is in the sense of being true, or is not in the sense of
being false, depends on combination and separation, and truth and false-
hood together are concerned with the apportionment of a contradiction (for
truth has the affirmation in the case of what is compounded and the
negation in the case of what is divided, while falsity has the contradictory
of this apportionment – it is another question, how it happens that we
think things together or apart; by ‘together’ and ‘apart’ I mean thinking
them so that there is no succession in the thoughts but they become a
unity – ; for falsity and truth are not in things – it is not as if the good were
true, and the bad were in itself false – but in thought; while with regard to
simple things and essences falsity and truth do not exist even in thought): –
we must consider later what has to be discussed with regard to that which is
or is not in this sense; but since the combination and the separation are in
thought and not in the things. (Aristotle 1993: 1027b30)
In sum, Aristotle thinks that
1. statements, as the bearers of truth and falsity, are in the soul and are
either unchanging or any change in them is due to a change in
something else;
2. the meaning of written and spoken language is to be explained by
reference to what goes on in our minds;
3. truth and falsity belong in the first instance to combinations and
separations that occur in our minds.
These are scattered remarks. Aristotle doesn’t show how they could be
combined in a coherent theory of terms, propositions and arguments. We
do not find such a theory in Aristotle; we find only some materials that
seem to have the potential for theoretical development.
An interpreter of Aristotle, faced with this situation, might try to
develop a theory in one of two ways. One option would be to enlist
elements drawn from Aristotle’s metaphysics or his account of scientific
knowledge. Another would be to import non-Aristotelian ideas. We will
see that both approaches were used by later Aristotelians in their efforts to
flesh out Aristotle’s sketchy remarks.
One obvious place to look for theoretical help in this enterprise is the
Philosopher’s division of all beings into the ten categories (substances,
quantities, relatives, qualities etc). From the standpoint of the theory of the
categories, our question becomes: Do the objects of logic belong to any of
the Aristotelian categories, and if they do, to which category or categories
150 Paul Thom
do they belong? This question was explicitly posed by a number of
thinkers in the twelfth and thirteenth centuries.

2. Robert Kilwardby
The thirteenth-century English philosopher and churchman Robert
Kilwardby commented extensively on Aristotle’s logic, as well as compos-
ing a treatise On the origin of the sciences and a set of questions on the four
books of Peter Lombard’s Sentences. Over the course of his career he
showed a continuing interest in the nature of the objects of logic, and
indeed the nature of logic itself.
In his early question-commentary on the Prior Analytics Kilwardby takes
the view that logic is one of the language-related sciences along with
grammar and rhetoric. In the work’s first sentence he adopts Boethius’s
characterisation of logic as an art of discoursing (Kilwardby 1516: 2ra).1 He
goes on to consider the meaning of the words ‘proposition’ [propositio,
Aristotle’s protasis] and ‘syllogism’ [syllogismus] as they occur in Boethius’s
translation of Aristotle’s text, distinguishing propositions from statements
[enuntiationes]. A statement is put forward on its own account, a propos-
ition on account of the conclusion it is intended to support. A statement
expresses what is in the speaker’s soul, and accordingly is defined as that
which is either true or false since truth and falsity reside in the soul
(Kilwardby 1516: 4rb).2 In his other writings Kilwardby will generally
preserve this distinction, reserving the term ‘proposition’ for the premise
of an argument.
He asks whether a syllogism should be defined as a kind of process,
rather than a kind of discourse (following Aristotle’s definition). He agrees
that there is a sense in which a syllogism is a mental process, but says that
this is a metaphorical sense (Kilwardby 1516: 4vb).3 And it must indeed be
regarded as a transferred usage for someone whose starting-points are
Aristotle’s usage of ‘syllogism’ to mean a kind of discourse and Boethius’s
characterisation of logic as a science of language.
In his later work On the rise of the sciences logic is no longer characterised
purely as a linguistic science, and the syllogism is no longer a purely
linguistic phenomenon. Logic is there presented under two guises. It is a
science of reason as well as being a language-related science:

1
Robert Kilwardby, Notule libri Priorum, Prologue.
2
Kilwardby, Notule libri Priorum Lectio 2 dubium 5.
3
Robert Kilwardby, Notule libri Priorum Lectio 4 dubium 1.
 Logic and its objects: a medieval Aristotelian view 151
It is called a science of reason not because it considers things belonging to
reason as they occur in reason alone, since in that case it would not properly
be called a science of discourse, but because it teaches the method of
reasoning that applies not only within the mind but also in discourse,
and because it considers the things belonging to reason as the reasons
why things set forth in discourse can be reasoned about by the mind. . . ..
It is, therefore, a ratiocinative science, or science of reason, because it
teaches one how to use the process of reasoning systematically, and a science
of discourse because it teaches one how to put it into discourse systematic-
ally. (Kilwardby 1988: 265)

On this view, which places mental reasoning on a par with reasoning in

words, it cannot be right to dismiss as merely metaphorical a conception of
the syllogism as a mental process.
He raises the issue of the basis or foundation of logic, declaring that
there are three different kinds of basis on which a body of scientific
knowledge can be founded. The science might be based on things that
actually exist. Or it might be based on potentialities, even when they are
unactualised. Finally, a science might be based not on potentialities but on
aptitudes of things. These are incomplete potentialities, such as the apti-
tude for sight which exists even in a blind eye. Now, even though speech
passes away as soon as it is uttered, something remains, namely certain
natural principles wherein potentialities or aptitudes reside. Because of
these, speech contains enough on which to found a science, even when
no-one is speaking (Kilwardby 1976: 429).
Later in On the rise of the sciences he adds that an art of reasoning has a
sufficient foundation in the natures of things through which they are
susceptible to a rational account. Among these natures he mentions
antecedents, consequents, incompatibles, universality, particularity,
middles, extremes, figure and mood (Kilwardby 1976: 463).
This interest in the foundations of the art of logic is even more
evident in a late theological work, Kilwardby’s questions on Peter
Lombard’s four books of the Sentences. Question 90 on Book One of
the Sentences contains a detailed exposition of the metaphysical status of
the objects of logic. It seems that Kilwardby himself attached some
importance to this exposition, for in the alphabetic index which he
compiled of the matters covered in his questions, he refers on four
occasions to question 90 on Book One.4 It is therefore worth examining
his exposition in detail.

4
Kilwardby 1995: Entries under ‘Ens’, ‘Predicamentum’, ‘Ratio’ and ‘Res’.
152 Paul Thom
Here is his question:
The next question is about divine knowledge in respect of things of reason,
which namely are in the human reason and are brought about by reason –
things such as propositions, syllogisms and the like, and all manner of
complex and incomplex things insofar as they concern reason. And the first
question about these is whether they are something, in such a way that they
are things in one or more of the categories. (Kilwardby 1986: 1, q.90: 1)
Although he refers here to ‘propositions’, he proceeds to discuss instead
what he calls ‘stateables’ (enuntiabilia). This is no doubt partly because of
the distinction he had made earlier between propositions and statements;
but this doesn’t explain why he talks about stateables rather than state-
ments. Christopher Martin takes the expression enuntiabile in earlier
authors to refer to a statement’s content rather than to the statement itself
(C. Martin 2001: 79). But I will argue later that there is reason for
doubting that this is Kilwardby’s meaning.
Concerning the nature of the objects of logic, Kilwardby mentions a
view according to which stateables cannot be assigned to any of the
Aristotelian categories. Among the arguments he mentions in favour of
this view, two rest on Aristotelian texts. First, there is the chapter of the
Categories where the ten categories are presented as a classification of
‘things said without any complexity’; stateables on the other hand, if they
are things at all, are things possessing complexity. The second Aristotelian
text is the one we noted above from Metaphysics book 6. Here, says
Kilwardby, the ten categories are presented as being truly outside the mind
or soul, whereas composition and division are said to belong to cognition,
not to external things (Kilwardby 1986: 1, q.90: 57).
Views denying categorial status to stateables or similar quasi-entities were
not uncommon in the twelfth and early thirteenth centuries. Peter Abelard,
for one, in using the word dictum to refer to a that-clause, or an accusative and
infinite construction in Latin, thought that the question of what sort of things
these dicta are simply does not arise: they are not things at all (King 2010).5
5
King 2010: ‘Abelard describes this as signifying what the sentence says, calling what is said by the
sentence its dictum (plural dicta). To the modern philosophical ear, Abelard’s dicta might sound like
propositions, abstract entities that are the timeless bearers of truth and falsity. But Abelard will have
nothing to do with any such entities. He declares repeatedly and emphatically that despite being
more than and different from the sentences that express them, dicta have no ontological standing
whatsoever. In the short space of a single paragraph he says that they are “no real things at all” and
twice calls them “absolutely nothing.” They underwrite sentences, but they aren’t real things. For
although a sentence says something, there is not some thing that it says. The semantic job of
sentences is to say something, which is not to be confused with naming or denoting some thing. It is
instead a matter of proposing how things are, provided this is not given a realist reading.’
 Logic and its objects: a medieval Aristotelian view 153
Again, the twelfth-century Ars Burana denies that enuntiabilia belong to any
of the Aristotelian categories.They exist, but belong to a category of their own
(Ars Burana, 208).6
But Kilwardby doesn’t have these versions in mind when he refers to the
view that the stateables are not to be found in any Aristotelian category.
Rather, he is thinking of the version of the view advanced by the English
theologian Alexander of Hales (Hales 1951–1957: 1 d.39 n.1). Alexander held
that the ontological type to which a statement belongs depends on whether
the statement expresses an essential or an accidental predication. In the
former case the statement is nothing other than its subject, and thus
belongs to the same Aristotelian category as its subject. Thus the statement
‘Fido is a dog’ is a substance, and is the very same substance as Fido. In the
case of accidental predications, the statement can be reduced to the
Aristotelian categories in one of two ways: either it reduces to the category
in which its accidental predicate is located, or partly to that category and
partly to the category of the subject. Thus ‘Fido is white’ turns out either
to be a quality (and then it is the quality of whiteness) or partly a quality
and partly a substance (and then it is partly Fido and partly whiteness)
(Kilwardby 1986: 1 q.90: 70).
In opposition to this view, Kilwardby holds that compositions, state-
ables and the other objects of logic can be assigned to the Aristotelian
categories in their own right without having to be reduced to the categories
to which their subjects and predicates belong. His view involves a complex
reduction to the Aristotelian categories.
Every thing, he declares, is either divine or human. The products of
nature he includes among the divine, along with things that issue from
God by himself. Human things, in his parlance, do not include what issues
from humans solely in virtue of their existence as natural beings, but only
what comes about through human activity in the form of industry or skill.
He classes the objects of logic, not among divine things, but among
human things in this narrow sense (Kilwardby 1986: 1 q.90: 102).
Among such things he distinguishes those that are internal to a human
and those that are external. The former include actions of combining,
dividing or reasoning, as well as the corresponding acts which he calls

6
Anon 1967: ‘If you ask what kind of thing it is, whether it is a substance or an accident, it must be
said that the sayable [enuntiabile], like the predicable, is neither substance nor accident nor any kind
of other category. For it has its own mode of existence [suum enim habet modum per se existendi]. And
it is said to be extracategorial, not, of course, in that it is not of any category, but in that it is not of
any of the ten categories identified by Aristotle. Such is the case with this category, which can be
called the category of the sayable [praedicamentum enuntiabile].’
154 Paul Thom
combinations, divisions, reasonings etc. The human things that are exter-
nal include utterances, the making of works and the works made (e.g. the
making of a house, and the house that is made). This distinction between
what is internal to the human and what is external appears to rest on a
distinction between doing and making. While making can be considered
as a kind of doing, it can also be distinguished from other kinds of doing
insofar as it involves the production of something, or at least a process
aimed at the production of something. Thus when we mentally combine
or separate concepts, or when we reason ‘in our heads’, we do not thereby
produce anything external to ourselves: we have done something but we
haven’t made anything. But when we utter something, or build a house,
we do produce something external, we make something. If this is what
Kilwardby means, then the acts which he distinguishes from actions, and
which he also considers to be internal, cannot be products of those actions.
Being purely internal, they have no product. It is clear that the relation
between acts and actions should be similar to the relation between works
and the making of works. But works stand to the making of works in more
than one relation. The relation of product to process is one such relation,
but it is of no use to us here because doings which are not makings have no
products. There is, however, another relation connecting works to their
making: the relation of completion. All actions, in principle, have comple-
tions; and it is these completions, I believe, that Kilwardby refers to as acts
or things-done. Thus the human things that are the objects of logic include
completed acts of stating and reasoning, as well as the actions that have
those acts as their completions. According to Kilwardby, all of these are
things of reason. They are secondarily in a category, because they are
founded on things of nature in one of two ways. In the case of makings
and actions of reason, they are founded on things of nature in the sense
that the latter constitute their subject matter. In the case of things-done or
made by reason and art, they are founded on things of nature in the sense
that they are certain relations or accidental conditions of things of nature.
Kilwardby takes both of these senses to indicate that the things of reason
and art have things of nature as their subjects; and he means here the
metaphysical subject that underlies these things of reason and of art. Thus,
while it is the things of nature that are primarily and of themselves in the
categories, the things of reason and art can be assigned to the categories in
a secondary sense, via the things of nature that are their metaphysical
subjects (Kilwardby 1986: 1 q.90: 111).
Stateables and arguments, whether completed or incomplete, may exist
in writing, in speech or merely in thought; and Kilwardby applies the
 Logic and its objects: a medieval Aristotelian view 155
above analysis to all three types of case. A written statement or argument,
he says, is a string of characters whose order is in accordance with the rules
of some art (he has in mind the arts of grammar and logic), and which has
the purpose of communicating knowledge of something through visual
perception. The characters are, let us say, written in ink; and their
metaphysical subject is then the primary substances which these blobs of
ink constitute. The written argument or the stateable is not these blobs of
ink; it is constituted by certain relations and accidental features of the ink.
Entities satisfying this complex description can be considered under more
than one aspect; and accordingly they will belong to different categories,
depending on the aspect under which they are considered. Considered as
signs they belong to the category of relatives. Considered as an ordering of
characters they could be assigned to the category of location or the category
‘Where’. Considered as exhibiting a certain syntactic form they can be
assigned to the fourth species of quality. In all these ways the relations or
properties which constitute the rational entity in question are accidental
features of the underlying subject: it is not essential to the ink that it be a
sign, nor is it essential to the characters that they be so ordered as to make
propositions.
Similar treatments can be given of spoken and mental statements and
arguments. Whether spoken or merely thought, these are signs and thus
belong to the category of relatives. As spoken they are qualities. As thought
they are dispositions of the mind – either states or passions and qualities.
Equally, the basic mental components which are combined or separated –
Aristotle’s ‘passions of the soul’, and Kilwardby’s intentiones or concepts –
can be considered either as qualities residing in the soul, or as relatives
insofar as they are signs of external things (Kilwardby 1986: 1 q.90: 163).
In sum, Kilwardby holds that stateables and the other objects of logic
have the following features:
1. They are human things.
2. Some of them are spoken, some written, some mental.
3. They are things of reason.
4. They are grounded in things of nature.
5. Considered as signs, they fit into the category of relatives.
Let us return to the meaning of ‘stateable’ in the light of this overall
picture. Whatever Kilwardby means by this word, it is evident that state-
ables must satisfy the above five conditions. They must also satisfy the
terms in which question 90 was framed: they have to be ‘in the human
reason and are brought about by reason’. Given these things, it is
156 Paul Thom
impossible to suppose that Kilwardby meant stateables to be propositional
contents, in the modern sense of eternal abstract objects. If he meant
stateables to be the contents of statements, he would have to have meant it
in a sense that complies with the above constraints. Now, it might be
proposed that a suitable notion of content can be devised, according to
which contents exist only when the things of which they are the contents
also exist. Such a notion, it might be argued, complies with the above
constraints. Alternatively, using the Aristotelian notion of potentiality, we
could say that a stateable is just a potential statement. The second of these
approaches, unlike the first, allows for the possibility that some stateables
are not (yet) actually stated.
How well does Kilwardby’s account of logic and its objects fit the sketch
given by Aristotle? Aristotle envisaged two possible answers to the question
whether statements are immutable. His first suggestion (that they are entirely
immutable) does not figure in Kilwardby’s account. The objects of logic, on
his account, are human things and thus subject to change. And if stateables are
potential statements then they change when their potentiality is actualised.
However, Aristotle’s alternative suggestion, that statements might be
such that any change in them is really a change in other things, is
consistent with Kilwardby’s account. Mental compositions, considered as
signs, are relative to that of which they are signs. Moreover, theirs is a
special kind of relativity – a kind that gives rise to Cambridge change. I can
change from being on your left to being on your right simply because you
walk around to my other side while I remain stationary; and similarly the
stateable that Socrates is sitting can change from being false to being true
simply because Socrates sits down.
The mentalistic semantics sketched in the De interpretatione is also
consistent with Kilwardby’s account of the objects of logic, as is his
account of composition and separation as located in the mind.
But only the second of the five points listed above is found explicitly in
Aristotle. Kilwardby’s specific conception of logic as an art – an art that
deals with human things which are grounded in things of nature – is not to
be found in Aristotle. It is Kilwardby’s way of turning Aristotle’s sketchy
account into a theory of the objects of logic.
Notwithstanding its departures from Aristotle’s own remarks on the
nature of the objects of logic, Kilwardby’s account is wholly Aristotelian in
its motivation. But the Aristotelian ideas on which he draws do not belong
in logic itself; they belong in natural philosophy and metaphysics. His
account is thus, to use the terminology of Sandra Lapointe (this volume),
an external one.
 Logic and its objects: a medieval Aristotelian view 157

3. Later thinkers
Jakob Schmutz argues that scholastic ideas were transmitted to the early
modern period along two paths. The first of these paths, which he calls ‘the
idealistic main road’, took the subject-matter of logic to be the activity of
the mind. The second path, ‘the realistic by-pass’, took logic to deal with
independent objects and structures (Schmutz 2012: 249). We have seen a
version of the first path in the writings of Kilwardby. Kilwardby was a
moderate realist. But other versions of this path can be found in nominal-
ists like William Ockham, for whom the objects of logic are individual
written, spoken, or mental tokens.7 Walter Burley, who opposed Ock-
ham’s views in most matters, appears to be working within the second
path: for him, propositions are either complexes depending on mental acts
of composition and separation, or intentional complexes existing in the
mind, or complexes existing outside the mind, which are signified by those
mental complexes. These extra-mental propositions [propositiones in re] are
the causes of truth of mental propositions (Cesalli 2007: 234).
The second path is taken up in the nineteenth century by Bernard
Bolzano then by Frege. Bolzano believed in propositions in themselves
(Sätze an sich), and held that it is the job of logicians to describe these
entities and their properties (Lapointe, this volume). He outlines his
notion of a proposition as follows:
One will gather what I mean by proposition as soon as I remark that I do
not call a proposition in itself or an objective proposition that which the
grammarians call a proposition, namely, the linguistic expression, but rather
simply the meaning of this expression, which must be exactly one of the
two, true or false; and that accordingly I attribute existence to the grasping
of a proposition, to thought propositions as well as to the judgments made
in the mind of a thinking being (existence, namely, in the mind of the one
who thinks this proposition and who makes the judgment); but the mere
proposition in itself (or the objective proposition) I count among the kinds
of things that do not have any existence whatsoever, and never can attain
existence. (Bolzano 2004: 40)
The objects of logic, on Bolzano’s view, are not human things and are not
grounded in the things of nature. As Rusnock and George say, ‘It should
be possible, [Bolzano] thought, to characterize propositions, ideas, infer-
ences, and the axiomatic organization of sciences without reference to a
thinking subject’ (Rusnock and George 2004: 177).
7
See (Panaccio 2004: 55).
158 Paul Thom
For Bolzano, propositions are not human things, they do not exist in the
mind or in language or in any way at all, and they are objective not relative.
His view is designed to pare down our conception of the objects of logic to
a bare minimum so that propositions are understood simply as that which
is true or false, and arguments are understood as configurations of
propositions.

4. Concluding remarks
In his essay in the present volume Graham Priest asks whether logic can be
revised, whether this can be done rationally, and if so how. And he
distinguishes logic as something that is taught, logic as something that is
used, and logic as the correct norms of reasoning (Priest, this volume).
I would like to add a few comments on Priest’s questions.
The history of logic contains plenty of examples of logicians proposing
to revise what hitherto had been accepted as the correct norms of
reasoning. Some of the great logicians – Abelard and Ockham along with
the well-known greats of the nineteenth century – saw themselves as not
just revising but reforming logic. Sometimes these reforms are motivated
by a sense that accepted logics are erroneous or in other ways inadequate to
accepted ideals of what logic should be. And sometimes what motivates a
reforming logician is a new vision of what logic should be. I think that the
major reformers of the nineteenth century had this sort of motivation.
Looking at the ‘traditional’ logic of their day, which was a watered-down
version of medieval logic, usually along the lines of Schmutz’s ‘idealist
road’, they worked with a vision of logic as an objective science. We benefit
today from the fruits of that vision. But it can be salutary occasionally at
least to look back to the different aims of the ‘idealist’ logicians of the high
Middle Ages.
The reason why Kilwardby and other ‘idealist’ medieval logicians con-
ceived of the objects of logic as human things is to be found in the aims
which they thought logic should have. In treating logic as an art, they were
committed to thinking that it should teach us how to construct good
definitions, divisions and arguments. So the objects of logic had to include
human activities of defining, dividing and arguing.
Everyone agrees that an argument is faulty if it allows the conclusion to
be false while the premises are true; and accordingly any good logical
theory has to include among its norms that one should not argue from
truths to a falsehood. Faults and norms go together.
 Logic and its objects: a medieval Aristotelian view 159
But in the opinion of most medieval logicians arguments were subject to
a variety of faults besides invalidity. Kilwardby is expressing a commonly
held view when he says that Aristotle identifies two of these faults in his
definition of the syllogism: first when he specifies that the conclusion must
be other than the premises, second when he requires that the conclusion
follow from the premises that are explicitly stated. According to Kilwardby,
the first of these specifications rules out begging the question, and the
second excludes the fallacy of stating as a reason what is not a reason [non
causa ut causa] (Kilwardby 1516: 4vb).8
In order to see why begging the question, and failing to state explicitly
what premises on which the conclusion relies, are faults in reasoning, one
has to look at what the point of reasoning is. Many of the medievals
believed that it is the function [opus] of the activity of reasoning to make
something known by proving it:
the function of the syllogism is to prove and make known. (Kilwardby
1516, 12rb)9
If a particular argument is not suitable for making its conclusion known by
proving it, then it is faulty in the way that a functional object is faulty
when it is incapable of performing its function. And if a form of reasoning
is not suitable for making anything known, then it is faulty. Kilwardby’s
idea here is that forms of reasoning which are intrinsically question-
begging, or which include redundant material, cannot perform the func-
tion that belongs to reasoning:
And it is to be said that there isn’t always a demonstration when the
conclusion follows of necessity, but there has to be proof of the conclusion
and it has to be made known, and further it is required that the premises are
apt to prove the conclusion and to make it known. But this is lacking when
the question is begged. (Kilwardby 1516: 72vb)10
In turning an art of human reasoning into an objective science, modern
logic has made enormous gains in comprehension and rigour. But it has
lost its connection with a conception of reasoning as an activity whose
point in human affairs makes it subject to other faults than invalidity.

8
Robert Kilwardby, Notule libri Priorum Lectio 4 dubium 1.
9
Kilwardby, Notule libri Priorum Lectio 11 dubium 3.
10
Kilwardby, Notile libri Priorum Lectio 67 dubium 3.
 chapter 9

The problem of universals and

the subject matter of logic
Gyula Klima

1. Introduction: the subject matter of logic

and the problem of universals
It might seem that the problem of universals should have little to do with
the issue of the subject matter of logic. After all, in (formal) logic we deal
with the deductive validity of arguments based on their formal structure,
whereas the problem of universals, at least in one of its possible formula-
tions, is the question of what corresponds to the universal terms of our
language, which constitute precisely the “material” part of arguments, the
part we disregard or abstract from in formal logic. However, upon a closer
look, there is a certain connection. On the semantic conception of validity
(which is also the intuitive motivation for syntactic rules of inference in
deductive systems), a formally valid argument has to be truth-preserving,
i.e., the truth of the premises has to guarantee the truth of the conclusion.
In a formal semantic system, this notion of “truth-preservation” is spelled
out in terms of the idea of compositionality, namely, the idea that the
semantic values of complex expressions are a function of the semantic
values of their components. Given this idea of compositionality and the
range of all possible evaluations of the components of the propositions
constituting an argument, the semantic notion of validity can be spelled
out by saying that an argument is valid just in case there is no possible
evaluation of the primitive components of its propositions that would,
based on the composition of these components, render the premises true
and the conclusion false. Obviously, this notion of validity presupposes
that we have a pretty clear idea of what the range of all possible semantic
values of the primitive components in question are and how those deter-
mine the truth and falsity of propositions based on their compositional
structure. But then, when we deal with predicate logic, some of those
possible semantic values are precisely the correlates of our universal terms,
the bone of contention in the problem of universals.

160
 The problem of universals and the subject matter of logic 161
So, in the end, the semantic notions of truth and logical validity in
predicate logic, being dependent on what the correlates of our universal
terms are, demand at least a certain semantic clarification of the issue of
universals. Contemporary conventional wisdom that we can glean from
ordinary logic textbooks would tell us that those correlates are sets, the
“extensions” or “denotations” of common terms. (See, e.g., Hurley 2008:
82–84) And if we press the issue of what sets are, then we are told that they
are possibly completely arbitrary collections of just any sorts of things, yet
somehow they are “abstract entities”. Clearly, ordinary logic text books can
just stop there. After all, they are not supposed to go into the metaphysical
problems of “abstract entities”: qua logic texts, they are just supposed to
provide some validity-checking machinery, and need not worry about the
possible ontological qualms of metaphysicians these machineries involve,
just like elementary math texts, as such, need not worry about the
ontological status of “mathematical entities” when they concern them-
selves only with providing reliable methods of calculation or construction.
This sort of attitude of the logician toward the metaphysical issues raised
by his subject is almost as old as the subject itself, as is testified by
Porphyry’s famously raising the fundamental questions concerning univer-
sals just in order to set them aside as pertaining to “deeper enquiries”, but
not to logic. (Spade 1994: 1) And of course it is one of the famous ironies of
the history of ideas that it was precisely on account of these questions that
medieval logicians got so much involved in these “deeper enquiries” that
John of Salisbury in his Metalogicon (John of Salisbury 2009: 111–116) had
to complain about how his contemporaries’ endless debates over these
issues confuse, rather than instruct, their students of introductory logic.
But despite the pedagogical validity of John’s objection to this practice,
one cannot really blame those logicians who get involved in these issues;
after all, as we shall see, the answers to Porphyry’s questions determine to a
large extent the construction of logical semantics in general, and thus the
understanding of the relationship between the subject matters of logic and
metaphysics in particular.

2. Realism, nominalism, conceptualism

Apparently, the primary issue concerning universals is ontological: are
there universal entities? After all, nobody in their right mind would doubt
whether we have universal words, i.e., words that on account of their
meaning apply to a multitude, indeed, to a potential infinity of entities.
However, the question then is: how come we can have such universal
162 Gyula Klima
terms at all? Plato’s “realist” answer, namely, that the difference between
universal and singular terms hinges on the ontological difference between
the kinds of entities these terms primarily name, rests on a relatively
simplistic understanding of the semantic relations of these terms: i.e. the
notion that their meaning consists in naming these different kinds of
entities in the same way. In fact, generalizing on this idea we might say
that on a realist conception semantic differences are accounted for in terms
of the ontological differences of the semantic values of syntactical items of
different categories, and not in terms of the differences in the semantic
functions of these items themselves: on this approach, in realism we can
have semantic uniformity at the expense of ontological diversity.
By contrast, those medieval thinkers who were convinced by Aristotle’s
and Boethius’s arguments against platonic universals (by John of Salis-
bury’s time practically everybody (Klima 2013a: n. 27)) would account for
the semantic diversity of singular and common terms not on the basis of
the ontological differences of the kinds of entities these terms denote, but
rather in terms of how they denote the same kind of entities, namely,
individuals, the only kind of real entities there are. Thus, on this under-
standing of the Aristotelian view, we can have ontological uniformity on the
basis of semantic diversity. As we shall see, the two formulae just italicized
can be regarded as the two extremes of a whole range of possible positions
concerning the relationship between semantics and metaphysics, ranging
from extreme realism to thoroughgoing nominalism. Indeed, let me call
the theoretical extreme of extreme realism the position that holds that all
semantic differences are ontological differences: different items in seman-
tically different syntactical categories differ in what kinds of entities their
semantic values are and not in what kinds of semantic functions relate
them to their semantic values. By contrast, on the other theoretical
extreme we have the position of extreme nominalism, which would hold
that all different items in semantically different syntactical categories differ
only in the kinds of semantic functions that relate them to their semantic
values, but all those semantic values are ontologically of the same kind, the
same, single kind of entities (or just the one single entity) there is. But in
order to see how actual historical positions can be arranged on this
theoretical scale, we should get into some further details concerning each
extreme.
On the platonic view, as we could see, the semantic relation between
common and singular terms and their semantic values would be of the
same kind: namely, denoting a single entity. What would make the
difference would be just the further ontological relation of the entity
 The problem of universals and the subject matter of logic 163
denoted by the common term, the universal, to its singulars as their
exemplar. It is only on account of this ontological relation that we can
use these terms to denote secondarily the singulars imitating or participat-
ing in their exemplar, but what the terms truly and primarily denote is the
exemplar itself. So, on this platonic understanding, the semantic function
of universal terms would be the same as that of singular terms, namely,
denoting a single entity, just like the representative function of a portrait is
to represent a single individual. However, just as the portrait of a monarch
can stand for a whole nation and thus can identify someone as a member
of that nation (say, in a passport), so the name of the universal can stand
for a whole kind and thus identify any individual participating in it as a
member of that kind.
On the Aristotelian view, on the other hand, universal terms are
universal precisely because they apply to a multitude of singular entities,
the same singular entities we can denote by their proper names, but
differently, namely, in a universal fashion, in abstraction from their individ-
ual differences. So, on this conception, what accounts for universality is
abstraction, a mental activity, the activity of the Aristotelian agent intellect
(nous poietikos, intellectus agens), which by this activity produces the first
universal representations, the so-called intelligible species out of the singular
representations of sensible singulars stored in sensory memory, the so-
called phantasms. The intelligible species, however, although they are
universally representing mental acts, generally were not regarded as the
universals Porphyry meant to consider in his work. An intelligible species
on this conception is rather an acquired disposition enabling the receptive
intellect (nous pathetikos, intellectus possibilis) to form a universal concept in
actual use. For example, once I acquire the intelligible species of circles,
that enables me to form actual thoughts about circles in general, but that
does not mean that I am thinking of circles all the time. Thus, in
possession of the intelligible species my mind still needs to form time
and again another mental act, the so-called formal concept, to form an
actual thought, as when I actually think that all circles touch a straight line
in one point. However, this mental act is still not the universal. It is a
universally representing singular act of a singular human mind; so, my
universal concept of circles is not the same item as your universal concept
of circles, even if those concepts are exactly alike in their representational
content, just like my dance moves I perform with my body are not the
same items you perform with yours, even if we are making exactly the same
kinds of moves, say, in a chorus line. What is the universal in the intended
sense is the common representational content of both your concept and
164 Gyula Klima
mine, on account of which we can be said to have the same concept,
despite the individual differences of the mental acts whereby we have it,
just like we can be said to make the same dance moves, despite the
individual differences of our bodies whereby we make them. Therefore,
this commonly intended object, the universal representational content of
both of our individual mental acts, was rightly called by later scholastic
thinkers the objective concept or intention, both because it is the universal
representation of the ultimately intended objects, namely, all singulars of
the same kind from some of which the intelligible species giving rise to this
concept was abstracted in the first place, and because it is the common
objective content of the formal concepts of all those individual human
minds that are capable of thinking this objective concept at all.
Now, even if this notion of a universal (as the objective representational
content of individual mental acts representing a natural kind of singulars in
an abstract fashion) may seem to be rather contrived from a contemporary
perspective, it should be clear that the conception that treats universals as
objective concepts, the universality of which is the result of the intellectual
activity of abstraction, does not allow in its “core ontology” the sort of
“abstract objects” Plato entertained. On this view, the intellect can form
universal objects of thought, but those objects of thought are not objects or
things absolutely speaking. Since they are the results of a mental activity,
they are ontologically posterior to that activity. (Although Scotus and his
followers would insist that among individuals of a certain kind there is a
certain less-than-numerical unity that is ontologically prior even to this
activity, and even Aquinas would admit a certain formal unity among
individuals of the same kind prior to any activity of the intellect (Klima
2013a: n. 39)). As Averroes was often quoted by medieval authors: intellec-
tus facit universalitatem in rebus – it is the understanding that generates
universality among things.

3. Scholastic “conceptualisms”
To see this issue in a little more detail, we should see exactly how the pieces
of the theory presented so far fit together in this tradition of medieval logic,
which I like to call “via antiqua semantics”, in contrast to a radically
different medieval logical tradition that emerged from the works of Wil-
liam Ockham, John Buridan, and their fellow nominalists, which I refer to
as “via moderna logic” (Klima 2011a, 2013a). As we shall see, both of these
approaches to logical semantics are basically variations on what may still be
called conceptualism; however, they are based on radically different
 The problem of universals and the subject matter of logic 165
conceptions of what concepts are and how they are related to their
objects, and accordingly give rise to very different constructions of logical
semantics.
The easiest way to make this contrast is through the analysis of an
example. Take one of the staples of scholastic lore: “Every man is an
animal”. This is an affirmative, universal categorical proposition (in the
medieval sense of ‘proposition’, meaning sentence-token), both terms of
which are common or universal terms, joined by a copula and determined
by a universal sign of quantity (a universal quantifier, as we would say). On
the common via antiqua analysis, the subject and predicate terms of this
proposition, its categorematic terms, have their semantic property of
signifying human and animal natures, respectively, on account of being
subordinated to the respective concepts our minds abstracted from their
individuating conditions in the humans and animals we have been exposed
to. Thus, although whatever it is on account of which I am a man (i.e., a
human being, regardless of gender) is a numerically distinct item from
whatever it is on account of which you are a man, the concept we
abstracted from humans we have been exposed to in forming our concept
of man abstracts from any individual differences (“individuating condi-
tions”). This is precisely the reason why this concept will represent not
only the humans we have been exposed to, but any past, present, future
and merely possible humans, that is to say, whatever it is that did, does,
will or can satisfy the condition of being human, whatever this condition
is, and whatever means we have (or don’t have) for verifying the satisfac-
tion of this condition (which would be a question of epistemology and not
of semantics). Accordingly, the corresponding term (‘man’ in English or
‘homo’ in Latin) can stand for any of these individuals in a proposition.
Indeed, this is what it does in this proposition: it stands or (to use the
Anglicized form of the scholastic technical term commonly used in the
secondary literature) supposits for all human beings that presently exist.
(For an overview of scholastic theories of “properties of terms”, including
supposition, see Read 2011) The reason why this term supposits only for
presently existing humans is the present tense of the copula, which restricts
the supposition (reference) of the term to present individuals that actually
satisfy the condition of its signification, namely, those individuals that
actually have human nature signified in general by this term. By contrast,
with different tenses or modalities, or when construed with verbs and their
derivatives that signify acts of the cognitive soul (i.e., sensitive or intellect-
ive, as opposed to the purely vegetative, soul) that are capable of targeting
objects beyond the presently existing ones (such as memory, imagination,
166 Gyula Klima
anticipation, abstract thought, etc.), the supposition of this term would be
extended, or ampliated, to use the Anglicized form of the scholastic term,
to past, future, or merely possible humans. (Klima 2001a, 2014) Since
medieval philosophers did not equate ontological commitment with quan-
tification à la Quine, they did not find any special ontological difficulty in
talking about “non-existent objects”, that is, objects of our cognitive
faculties beyond the objects directly perceived in our present environment.
In fact, even the ontologically most squeamish nominalists would not
hesitate to quantify over mere possibilia, simply because the flexibility of
their theory of quantification and reference, namely, the theory of suppos-
ition coupled with the theory of ampliation, allowed them to contend that
these mere objects of thought (and of other cognitive acts) are simply
nothing, and so to inquire into their nature and ontology would be just a
wild goose chase, amounting to nothing. (Cf. Klima 2014, 2009: c. 10.)
The nominalists, however, did have a bone (or two) to pick with via
antiqua semanticists on other aspects of their theory. In the first place,
and perhaps most fundamentally, the medieval “realists” (practically any-
body before Ockham), even if they did not buy into Plato’s “stratified
ontology” of universals vs. singulars, and had a much more sophisticated
semantic theory than the uniform naming relation between different
kinds of words and correspondingly different kinds of things, they did
preserve some sort of semantic uniformity at the expense of some sort of
ontological diversity.
As we have seen, the signification of common terms, based on the idea
of words being subordinated to concepts to inherit their natural semantic
features, coupled with the Aristotelian theory of abstraction, led to a
peculiar theory of predication within this framework, often referred to in
the literature as the inherence theory of predication. The theory is simple
enough: the predication ‘x is F’ is true, just in case the F-ness of x actually
exists, or equivalently, just in case F-ness, the form or property signified by
the predicate F in the individual x actually inheres in x. The problems start
when we consider all sorts of substitution instances of F. For then we start
realizing that, apparently, by the lights of via antiqua semantics, as
Ockham put it “a column is to the right by to-the-rightness, God is
creating by creation, is good by goodness, just by justice, mighty by might,
an accident inheres by inherence, a subject is subjected by subjection, the
apt is apt by aptitude, a chimera is nothing by nothingness, someone blind
is blind by blindness, a body is mobile by mobility, and so on for other,
innumerable cases” (Ockham 1974: I, 51). In short, to the nominalists,
starting with Ockham, it appeared that their realist opponents (in the case
 The problem of universals and the subject matter of logic 167
of Ockham, especially John Duns Scotus and Walter Burleigh) generate
metaphysical problems where there shouldn’t be any, simply on account of
a misconception of semantics, because their conception would “multiply
beings according to the multiplicity of terms” (Ockham 1974: I, 51).
To be sure, the “realists” did make a number of metaphysical distinc-
tions between the types and modes of being depending on the substitution
instances of F to avoid apparent metaphysical absurdities (such as a thing
undergoing change without losing or acquiring a property, action at a
distance, non-beings undergoing change, etc. – cf. Klima 1999), but for
Ockham and his ilk, that is precisely the problem: to maintain a certain
type of semantic uniformity, the realists introduce ontological diversity
where there shouldn’t be any, since the difference is not in the things
signified by our different terms, but in the different concepts signifying the
same things in different ways (Klima 2011a).
To illustrate the sort of semantic uniformity and the requisite onto-
logical diversity in the via antiqua approach to semantics, let us briefly
return to the via antiqua analysis of the meaning and conditions of truth of
‘Every man is an animal’. The two categorematic terms both have the same
type of significative function, namely, signifying the individualized natures
of individuals represented by their corresponding concepts in an abstract,
universal fashion. The subject term, in turn, has the function of standing
for those individuals that actually have this nature at the time connoted by
the tense of the copula, whereas the predicate has the function of attribut-
ing the nature it signifies to the individuals thus picked out.
And since the universal sign in front of the subject indicates that the
truth of the entire proposition requires that all these individuals have the
nature signified by the predicate in actuality at the time connoted by
the copula, the propositional complex, variously called dictum, enuntiabile,
or complexe significabile, resulting from the combination of subject and
predicate by the copula as further determined by the universal sign, will be
actual just in case all individuals supposited for by the subject actually have
the nature signified by the predicate. As can be seen, the via antiqua
analysis of this single proposition apparently requires an extremely com-
plex, multilayered ontology; however, the payoff in the end is the simple,
uniform semantic criterion of truth originally proposed by Aristotle: a
proposition is true just in case what it signifies exists. But it is instructive
to take a closer look at the ontological status of the items required by this
analysis.
In the first place, the analysis requires the existence of some ordinary
primary substances, namely, humans. However, for something to count as
168 Gyula Klima
a human, it has to have humanity. And since humans are essentially
rational animals, their existence also requires the existence of their
rationality and animality. Furthermore, on this conception, if something
exists in actuality, its existence also has to be in actuality, and it is this
actual existence that is supposed to be signified by the copula. But the
copula also co-signifies the union of what is signified by the predicate and
by the subject, thereby indicating that the existence of the thing signified
by the predicate is also the existence (whether substantial or accidental
existence, but in the case of the proposition at hand, it is the substantial
existence) of the thing supposited for by the subject, which is actual at the
time connoted by the tense of the copula.
Now, these are just the real, mind-independently existing items required
for the truth of this proposition. However, as we could see, these items can
be picked out by the relevant syntactical items from reality only on account
of these syntactical items being subordinated to their respective concepts
that renders them meaningful in the first place. So, on this analysis,
all propositions require a further “ontological layer”, as it were, the layer
of concepts.
But, as we could see, concepts come in two necessarily connected sorts,
namely, formal and objective concepts. The formal concepts are real,
inherent, individualized qualities of the individual minds that form them.
The objective concepts, on the other hand, are the direct objects of these
individual mental acts, some of which represent extra-mental individuals in
a universal manner, but without representing the sorts of universal things
imagined by Plato. Thus, these objective concepts form another onto-
logical layer, the layer of beings of reason, which in the strict sense are mere
objects of thought (the representational contents of formal concepts), but
with a more or less remote “foundation in reality” (as opposed to mere
figments). In the case of universal concepts, this “foundation in reality”
consists in the individualized natures of the things from which these
objective concepts derive in the process of abstraction and concept forma-
tion (through the generation of intelligible species).
But the objective concepts do not occur to the mind in isolation. They
enter into the composition of complex thoughts, which are formed by
means of syncategorematic concepts, such as the copula, which, as we
could see, besides its syncategorematic function of joining the concepts of
subject and predicate also has the categorematic function of signifying the
existence of what is signified by the predicate in the relevant supposita
(referents) of the subject, the relevant supposita being determined by the
syncategorematic concept of the sign of quantity, in the present case the
 The problem of universals and the subject matter of logic 169
universal sign (quantifier). The propositional complex thus formed is
another being of reason with a foundation in reality. (Klima 2011b, 2012)
Indeed, the obtaining of this complex is conditioned both on the side of
the mind forming it and on the side of the things serving as its foundation.
For the state of affairs (dictum, enuntiabile, etc.) that every man is an
animal actually obtains just in case there are humans and each of them
actually has its animality. But then, providing the rules of composition for
all types of thought, based on the syntactical structure of the proposition
expressing it, one can provide the uniform Aristotelian criterion of truth
for all types of propositions, and based on that, the uniform criterion for
the formal validity of an inference or consequence. In fact, since in this
framework truth is defined in terms of the content of propositions, a
stronger entailment relation is also definable, in terms of the more fine-
grained notion of content-containment, as was proposed by some authors
in this tradition (Martin 2010, 2012; Read 2010; for comparison, an
interesting contemporary development of the idea of entailment based
on content- or meaning-containment can be found, for instance, in Brady
and Rush 2009). However, as we could see, this could be obtained only in
terms of the multi-layered ontology that provoked Ockham’s and his
fellow-nominalists’ charges.
Nevertheless, we should also emphasize that Ockham’s and his follow-
ers’ charges were not entirely justified, and, accordingly, the ultimate
difference between late-medieval realists and nominalists did not lie simply
in their different ontologies or simply in their different semantics that
allowed them to handle their ontological problems in rather different ways,
but rather in their different conceptions of concepts underlying even their
semantic differences.
The Ockhamist charge of multiplying entities with the multiplicity of
terms was unjustified for several reasons. In the first place, even “realists”
had at their disposal at least two different kinds of strategies to reduce the
ontological commitment of their semantics: (1) the identification of the
semantic values of terms belonging to different categories, and (2) attrib-
uting a “reduced” form of existence to some of the semantic values of some
terms in some categories.
The first strategy could rely already on the authority of Aristotle, who in
his Physics identified action and passion with the same motion, but several
original considerations allowed scholastic thinkers to identify relations
with their foundations (i.e., with entities in the “absolute” categories of
substance, quantity and quality), and in general entities in the remaining
six categories with those in the first three.
170 Gyula Klima
The second strategy, as we could already see, was based on the idea that
the mind’s different ways of conceiving of mind-independent entities of
external reality produces certain mind-dependent, intentional objects, the
objective concepts, the information contents of our mental acts, by means of
which we variously conceive ultimately those mind-independent objects
that satisfy the criteria of applicability set by these objective concepts, or
intentions. This is precisely why in this tradition the subject matter of logic
was generally characterized as the study of second intentions, that is, of
concepts of concepts (such as the concepts of subject, predicate, propos-
ition, negation, or the ultimately targeted notion of valid consequence).
So, the core-ontology of real mind-independent entities could in principle
have been exactly the same for these “realists” as for Ockhamist
“nominalists”.
In fact, both late-medieval “realists” and “nominalists” were conceptual-
ists, but based on a rather different conception of concepts and their role in
logic, semantics, and epistemology. In this connection, it is informative to
compare Ockham’s earlier, fictum-theory of universals with that of the via
antiqua conception discussed so far. For the important difference between
the two is that even if Ockham’s ficta are ontologically on the same footing
as the objective concepts of the realists (they are beings of reason), and they
would be best characterized in the same way, namely, as the objective
information content of individual mental acts, they do not have the same
role in Ockham’s theory.
In fact, as prompted by the arguments of his confrere, Walter Chatton,
Ockham came to realize that ficta did not play any significant role in his
logic at all, and so, grabbing his famous razor, he painlessly cut them out
from his ontology. The reason why Ockham could do so is that for him
the universality of universal representations (whether ficta or universally
representing mental acts) consists merely in their indifferent representation
of a number of individuals (in the case of a natural kind, all past, present,
future, and merely possible individuals of the same kind). However, this
indifferent representation is due not to some abstracted condition of having
a certain nature that individuals of a given kind satisfy, but, as a matter of
brute fact, to the indifference of the causal impact of one individual or
another of the same natural kind on the human mind.
Accordingly, for Ockham, there is no question whether there is a real
distinction between the nature of an individual represented by a universal
concept and the individual itself (as this emerged as a metaphysical
question in the via antiqua), because what these concepts indifferently
represent are just the individuals themselves. Therefore, for him, the
 The problem of universals and the subject matter of logic 171
supposita of the terms subordinated to these concepts are not the individ-
uals that actually have these natures relative to the time connoted by the
copula (as was conceived in the via antiqua), but simply the individuals
represented by the concept that are actual at that time. As a result, terms in
the predicate position do not signify inherent natures either, so Ockham
and his followers endorse the identity-theory of predication, as opposed to
the inherence-theory. According to the identity-theory, an affirmative predi-
cation is true, just in case the terms of the proposition supposit for the
same thing or things. But this is obviously not a general “definition
of truth”. In order to achieve a truth-definition on this approach, one
should provide similar “satisfaction clauses” for all logically different
proposition-types, such as negatives, universals, particulars, not to mention
the propositional complexes, such as conjunctions, disjunctions, etc.
(Klima 2009: c. 10).
As can be seen, on this nominalist approach, just as terms do not figure
into the calculation of truth-values with their intensions, but their exten-
sions, so too, the truth of propositions themselves is not determined in
terms of their intension or signification, but solely by the extensions (sets of
supposita) of their categorematic terms. Accordingly, nominalist semantics
as such has no use for enuntiabilia or complexe significabilia, as is brilliantly
illustrated by the logic of John Buridan.
On Buridan’s theory, propositional signification is simply the set of all
significata (and connotata) of a proposition’s categorematic terms, which of
course yields a very coarse-grained conception of propositional significa-
tion. In fact, on this conception, contradictory propositions must signify
the same, although differently, on account of the concept of negation
included in the one, but not in the other of the contradictory pair of
propositions (Klima 2009: c. 9). However, Buridan does not have to care
much. On his account, truth is not a function of signification, so, two
propositions of the same signification can have opposite truth-values.
Thus, when he needs a more fine-grained semantics of propositional
signification (as in intentional contexts) he can always refer to the diversity
of the corresponding propositions on the mental level, where, of course, in
line with his nominalist ontology, the mental propositions in question are
just inherent qualities, individual acts of individual human minds, just as
are the concepts entering into their semantic make-up (Klima 2009: c. 8).
So, nominalist semantics can afford to be based on an entirely homoge-
neous, parsimonious ontology (containing only two or three distinct
categories of entities, namely, substances, quantities – sometimes identified
with substances or qualities, as by Ockham – and qualities). However, this
172 Gyula Klima
parsimonious homogeneity is achieved at the “expense” of massive seman-
tic diversity, assigning some of these entities various, distinctive semantic
functions, especially on the mental level. However these semantic func-
tions are always defined in terms of the extensions of these mental items,
the formal concepts inherent in individual human minds, to the exclusion
of items in the “ontological limbo” of the objective concepts of the
older model.
Still, even the nominalist version of scholastic conceptualism could
maintain that logic is the study of second intentions without lapsing into
subjectivism, conventionalism, or skepticism, let alone psychologism – features
that in a modern context are so often associated with conceptualism. Well,
how come? Actually, answering this question will allow us to draw some
general conclusions concerning both major versions of scholastic concep-
tualism sketched out here, and some general lessons we can learn from
these scholastic theories concerning the subject matter of logic and
metaphysics.

4. Conclusion: the lessons we can learn from the scholastics

In the first place, it should be quite obvious that the objective concepts of
the via antiqua conception are objective not only in the medieval sense, i.e.,
in the sense that they are the objects of individual mental acts inherent in
individual human minds as their individualized forms (the formal con-
cepts), but also in the modern sense of being intersubjectively accessible
and the same for all. For an objective concept is the common, abstract
information content of any formal concept that carries this information,
and any formal concept that does not carry the same information is just
not a formal concept of the same objective concept. Thus, in this frame-
work it simply cannot happen that you and I have different concepts of the
same kind of entities as such, or the same concept of entities of different
kinds. If I have the concept of H2O and you have the concept of XYZ,
then we are just not talking about the same thing, no matter that we use
the same word in our miraculously matching English idioms of Putnam’s
Twin Earth scenario. (Putnam 2000: 422) Since I use the term ‘water’ as
subordinated to my concept and you use it as subordinated to yours, we
use our words equivocally, no matter how phenomenally similar the two
kinds of things are, and how similarly we would describe their phenomenal
properties. Therefore, if I say ‘This is water’ pointing to a glass of H2O and
you say, pointing to the same, ‘No, that is not water’, we actually do not
contradict each other, although, of course, it can take a while until we
 The problem of universals and the subject matter of logic 173
figure out just why. But that is an issue of epistemology, not semantics. As
far as meaning is concerned, on this conception you can have the same
concept as I do only if our individual mental acts latch onto the same kind
of objects in the same way, carrying exactly the same information. To be
sure, one of us may have a better understanding of the nature of the thing
or things thus conceived, on account of being more aware of the relation-
ships among this concept and others, picking out the same nature differ-
ently, on account of other, more specific or more generic information, as
when one of us knows the genuine quidditative definition of the kind of
thing in question. But regardless of whether either of us has this definition
in mind or knows what it would be, we can be said to have a concept of
this kind of thing as such, only if we managed to form the objective
concept of its essence, which must be the same for both of us, or we just
do not have this concept at all (Cf. Aquinas 2000: Sententia Metaphysicae,
lib. 9. l. 11. n. 13.).
From this it should also be clear that these objective concepts are non-
conventionally objective. For what determines the information content of
our abstracted, simple concepts is what kinds of things they are abstracted
from, that is to say, the nature or essence of those things themselves. To be
sure, we can construct complex concepts out of these simple ones as we
wish and agree to express them by words we wish (ad placitum, as the
scholastics said), but whether the concept we both abstracted from samples
of H2O will apply to all and only samples of that kind of thing (even if we
cannot infallibly identify all such samples in all possible scenarios) is clearly
not a matter of our wishes.
Finally, even if our psychological mechanisms require that when we
form these simple concepts and their combinations our minds work with
their own individual, subjective mental acts, their formal concepts; the
logical relations among these mental acts are not a matter of the causal or
other psychological relations among them, but a matter of the relations of
their objective semantic contents, the relations among their objective
concepts. So, no wonder scholastic thinkers working in this tradition
would identify the subject matter of logic as those second intentions or
objective concepts of our objective concepts that express precisely these
objective conceptual relations. (Schmidt 1966; Natalis 2008)
Therefore, it should also be clear that the laws of logic in this framework
are supposed to be fundamentally different from the laws of psychology.
For while the former are the laws of the logical relations among objective
concepts, the latter are the laws of the causal relations among formal
concepts. Thus, whereas logic can be normative, prescribing the laws of
174 Gyula Klima
valid inference, cognitive psychology can only be descriptive, describing
and perhaps explaining those psychological mechanisms that can, for
instance, make us prone to certain types of logical errors.
But similar observations can be made about the nominalist approach,
although with some interesting, and from a modern perspective, especially
instructive modifications. As for the issue of objectivity, the nominalist
authors, after Ockham had dropped his ficta as being ontologically both-
ersome and theoretically unnecessary, still insisted on our simple mental
concepts being externally determined and not just subjectively made up by
us, grounding this belief in the generally reliable natural mechanisms of
sense perception and universal concept formation (Panaccio 2004; Klima
2009: c. 4).
So, for the nominalists there are no longer quasi-entities in mere
objective being: our concepts are anchored in extramental reality through
the objective (in the modern sense, meaning mind-independent) laws of
natural causality. However, there is a slight, but very significant shift in the
way nominalists conceived of this “anchoring”, as opposed to their realist
counterparts. For realists, what did the “anchoring” was a certain “formal
unity”, the sameness of the information content that was encoded in the
mental acts carrying this information and that was realized in the very
nature of the things these concepts represented to the subjects having
them. This conception of formal unity provided a much stronger
“anchoring” for the via antiqua conception than what is available in the
nominalist via moderna.
The via antiqua conception, as we could see, builds the identity of the
nature of the represented objects into the identity-conditions of a concept
itself, hence tying the identity of its objects by logical necessity to the
identity of the concept in question. By contrast, the via moderna concep-
tion ties the identity of the concept by mere natural, causal necessity to the
identity of its objects. In a medieval theological context, however, this
difference amounts to the difference of what could and could not be done
by divine absolute power. Therefore, it should come as no surprise that
anticipations of Descartes’ famous “Demon argument” crop up precisely in
this context, once the nominalist conception opened up at least the logical
possibility of a cognitive subjects’ having exactly the same concepts planted
in his mind by a deceptive God without the mediation of these concepts’
adequate objects; i.e., the subject having exactly the same phenomenal
consciousness, regardless of whether any items of it are veridical, faithful
representations of reality or not (Klima and Hall 2011; Karger
forthcoming).
 The problem of universals and the subject matter of logic 175
As for conventionalism, the nominalists usually laid even more emphasis
on the conventionality of spoken and written languages than their realist
counterparts; however, they equally strongly emphasized the non-
conventional, natural character of the language of thought or mental
language, based on the fixed laws of nature. So, even for the nominalists,
our simple concepts, anchored in natural kinds by causality, are not made
up by us at will in the way the words we express them by are.
Nevertheless, as we could see, the nominalist conception still allows for
the possibility of “supernatural skepticism”, providing a whole range of
different reactions to this possibility, but perhaps most typically offering a
“reliabilist” solution, dismissing the overblown certainty-criteria of the
skeptic in terms of different sorts of reliability-criteria for our various
cognitive powers and mechanisms utilized differently in different cognitive
scenarios (Aristotle providing again a good authority by his remark that
one should not expect mathematical certainty in all fields of inquiry)
(Klima 2009: c. 12). In any case, in deductive logic, nominalists still
required the same, highest form of certainty as in mathematics.
For despite their conception of concepts as being simply individualized
mental acts tied to their objects by mere natural necessity, the nominalists
did not take logic to collapse into psychology. Perhaps, the best illustration
of this fact comes from Adam Wodeham’s “thought experiment” concern-
ing the presumed “perfect telepathy” of human minds uncorrupted by
original sin (of which now we have no actual example) and of good angels
(the ones that did not fall with Lucifer). These minds, according to
Wodeham, are perfectly capable of intuiting each other’s thoughts, how-
ever, this would still not amount to communication, because they would
not be able, simply on account of this intuition, to decode the contents of
those thoughts (pretty much like brain scans can give us some information
about some sort of brain activity, but not about what that activity is
about). Accordingly, based on these observations, these minds could come
up with a natural science describing the regularities of these mental
activities, but that science would tell us nothing about the content,
let alone the validity of the thought processes couched by these activities,
which would be the concern of a different science, namely, logic. (Karger
2001: 295–6)
So, what conclusions can we draw for ourselves from this however
sketchy, general comparison of the two main scholastic approaches to
the problem of universals and the subject matter of logic?
The traditionally recognized alternative answers to the problem of
universals come in many shades and colors. But especially in their
176 Gyula Klima
sophisticated medieval versions, they are primarily differentiated not by
their different ontologies, but rather by their different conceptions of
concepts, determining different kinds of constructions of logical semantic
theories.
These different theories can be arranged on a “theoretical scale”, ranging
from extreme realism to extreme nominalism, meaning maximal semantic
uniformity along with maximal ontological diversity on the realist end
(every linguistic item has the same type of semantic function, say, naming
some entity, while these items differ semantically on the basis of what type
of entity they name), and maximal ontological uniformity with maximal
semantic diversity on the nominalist end (having just one ontological type
of entities, while all semantic differences consist in the different semantic
functions of some of these entities of the same ontological type). Even if,
perhaps, no actual historical theory can be placed on either extreme
(although, if Parmenides had had one, it would have probably been close
to the extreme nominalist end, whereas if Plato had had an articulated
semantic theory, then it might have been close to the extreme realist end,
as probably so would be Wittgenstein’s caricature of Augustine’s theory),
the actual, well-articulated theories can better be understood as variously
removed from either extreme on account of various elements of variety
introduced either on the side of ontology, by multiplying the semantically
relevant distinct categories of entities, or on the side of semantics, by
multiplying the different types of semantic relations that map the syntac-
tical categories of language onto the ontological categories distinguished by
the theory.
In the scholastic theories discussed here, these different types of
semantic relations were understood in terms of how our different kinds
of concepts relate our words to things in our ontology. Here, in the via
antiqua framework “things” can be understood rather loosely for any
object or quasi-object of our thought, whereas in the via moderna
framework, they would be restricted to really existing entities in the
category of substance and quality (or for Buridan and his followers also
in the category of quantity). It is very telling, however, that the “core
ontology” (i.e., the categories of real entities to the exclusion of beings of
reason) of the via antiqua framework could be just as parsimonious as
the nominalist “core ontology” was (as is illustrated, for instance, by the
ontology of the late-scholastic Domingo Soto). Furthermore, even the via
moderna framework could have in principle reduced its ontology to one
homogeneous category, had it not been for certain theological worries
 The problem of universals and the subject matter of logic 177
concerning the Eucharist, and metaphysical worries concerning atomism,
widely taken to have been refuted by Aristotle.
Finally, in view of the foregoing comparative analysis of how both
medieval viae would avoid, in their own ways, contemporary worries about
conceptualism in general, we can conclude that such comparisons can be
especially useful for refining our understanding of the implications of the
various theories that can be arranged on the theoretical scale set up in this
chapter. Such refinements in the end will allow us to overcome certain
modern theoretical reflexes (nominalism entails skepticism, conceptualism
leads to psychologism, etc.) by shining a new light on the historical origins
of these reflexes themselves.
 chapter 10

Logics and worlds

Ermanno Bencivenga

There are no specifically logical objects. A logic is a theory of the logos, of

meaningful discourse: a theory of how discourse acquires the meaning it
does. Traditionally, logics have investigated the behavior of syncategore-
matic words like “and,” which contribute to the meanings of their contexts
while having no meaning in isolation, and hence have studied the contrast,
say, between “and” as it occurs in sentences like
(1) They are married and have a child
(2) They got married and had a child.
But there is no reason why they should not also study the relation, say,
between (1) and
(3) They are married and have a pet;
in particular, why they should not inquire into whether (1) and (3) are
interchangeable. A logic that studied the contrast between (1) and (2) would
be (among other things) a logic of “and” (a theory of the meaningful use of
“and”); a logic that studied the relation between (1) and (3) would be
(among other things) a logic of “child” (and of “pet”). Nothing other than
greater generality, and attendant lesser detail, is gained by concentrating on
logics of conjunctions, adverbs, and pronouns; and at no level of generality
does it make any sense to grace a word (or maybe a diacritical sign, like a
parenthesis) with the label “logical constant.” If you ever get sidetracked
into a Quinean fatuous search for the nonexistent Eldorado of “pure
logic,” I recommend a refreshing immersion into Buridan’s subtle, percep-
tive study of the infinite nuances of signification and supposition.1
But, if there is no specific ontological realm for logic, there definitely are
ontological questions pertaining to it. Two questions, primarily: Do logical
laws (the laws bringing out the meaningful behavior of various words) have
1
See Klima 2001b.

178
 Logics and worlds 179
an objective status? And, if so, how do they acquire it? No answers to such
questions can be attempted without a substantive view of what objectivity,
and an ontology, are. Since this is not the place to defend my Kantian,
transcendental-idealist position on the matter, I will simply state it before
moving on.2 (Though I must note that, here, the matter dealt with is not
innocent: an ontology is a logic of being, hence what ontological status a
logic has is not independent of what logic it is.)
A transcendental philosophy as described and practiced by Kant is itself
a logic. It is not intended to decide such factual questions as whether there
is a God or humans are free, but to address semantical issues like what the
meaning of “God” or “freedom” is. The reason why the formidable epithet
“transcendental” is attached to it is precisely the misunderstanding
I alluded to above: if you think that logic only deals with (some) conjunc-
tions, adverbs, and pronouns, then you are forced to qualify this narrow
concern as “general logic” and to conjure up some other name for the full
line of business.
Within the semantical space where the (transcendental) logical enter-
prise is located, one can take different words as primitives and establish a
network of semantical relations and dependencies based on those primi-
tives and involving other words, each time resulting in (the beginning of ) a
different transcendental philosophy/logic; as more such structure is
exposed, the meanings of the words involved will become correspondingly
better established and clearer. If we want, we can even talk about “con-
cepts”: clusters of largely interchangeable words resonating with a common
theme, not necessarily spoken but suggestively intimated by the resonance.
A transcendental realism (TR) is a transcendental philosophy/logic
that takes a cluster of largely interchangeable words including “object,”
“substance,” “thing,” and “existence” as primitives, and then turns to
the (hopeless) task of defining words like “experience” or “knowledge”
on that basis. A transcendental idealism (TI) – my chosen course – is a
transcendental philosophy/logic that takes its cue from a different
cluster including “experience,” “representation,” and “consciousness,”
and then moves to defining “object” and “existence.” Not surprisingly,
a TI has a lot more to say about objectivity – what makes an object an
object – than a TR: of primitives we will forever be dumb and, though
occasionally that incapacity is depicted as mystical depth, the bottom
line is that no interesting account of what primitives mean is forthcom-
ing. In a TI, however, objectivity belongs to a derived cluster; hence its
2
For additional details, see my 1987 and 2007.
180 Ermanno Bencivenga
derivation provides for a rich and complex contribution to the logos –
one we need to work out now.
Representations (or experiences, or consciousness) are always of some-
thing: their so-called intentional objects, which despite their name are not
objects yet, indeed never will be. And neither representations nor their
intentional objects can be objective in isolation: they can only be object-
ive to the extent that they are categorially connected – that they are
mutually consistent; that there are relations of mutual determination
among them; that there is a definite fact of the matter of how many of
them there are, hence how they are identical with, or distinct from, one
another. The graduality signaled by the locution “to the extent that”
would only be redeemed at the limit: by a system of representations to
which nothing further could be added and where each member were fully
determined to be what it is by its relations with all others. Within that
system (suddenly, as soon as completeness were reached), all representa-
tions would be objective and all their intentional objects would be
objects, period: existent objects. The limit cannot be experienced, in
the strongest sense of “cannot”: it would be contradictory (antinomical)
to suppose otherwise, hence all intentional objects will forever stay that
way – remain appearances. But this conclusion is only going to worry
those who reduce a TI to a series of empirical claims about what takes
place (or can take place) in a mind. As none of that is implied here, for
what is in question is rather the semantics of “objectivity,” we have all we
could ask for: a regulative idea that orients our everyday, always fallible
vicissitudes, signaling the direction in which we are likely to find more
objectivity and the standards we must enforce to maximize it (coherence,
agreement, inclusiveness, mathematical structure), inevitably staying clear
of such a complete realization of the idea as would make ( per impossibile)
the whole project fall apart.
What the system of representations envisioned at the limit would
represent is (as one might expect) a system of objects to which nothing
further could be added and where each member were fully determined to
be what it is by its relations with all others. I call (and Kant calls) this
system a world. No one ever experiences a world, though most everyone
(everyone but severely disturbed people) ordinarily presumes herself to be
experiencing part of one, and sometimes goes through the catastrophe of
seeing what she took to be part of a world explode into incoherence and
disconnectedness. When such unfortunate events take place, we try hard
to blame them on contingent occurrences (on misreadings of data) while
keeping faith with the semantical laws that organize our logic. What I saw
 Logics and worlds 181
in the corner was not an elephant; it was an armchair; but the meanings of
“elephant” and “armchair” are not disrupted by this mishap. And yet, it is
not always that easy; for, what about the semantical necessity, until circa
1905, that a wave is not a particle? And about the logical clash that ensued
when we were forced to deny that necessity: a clash whose logical character
would be missed by more parochial characterizations of logic? If, on the
other hand, you want to insist on a parochial characterization and attribute
that clash to the empirical realm, I urge you to consider what happened a
few years earlier, when the very logic of sets blew up in people’s face. As
very unfortunate happenings of this kind can never finally be ruled out,
logics align themselves with worlds, in the following way:
A logic cannot be a theory of meaningless discourse (of alogos). But any
word we use can only be meaningful if our whole discourse is meaningful:
if all words we use belong in an ideal complete dictionary that sets
consistent, connected relations among them – once again, it is an all-or-
nothing affair. Only a logic associated with this kind of dictionary would
be objective in the sense of possibly describing a world of objects (would be
a real, not an apparent, logic), independently of the data that gave empir-
ical content to its entries. As no such dictionary can ever be at hand, we are
never in possession of a logic but only of something we presume to be a
fragment of one, and which is always at risk of dissolving into the stuff
dreams are made of. A logic (like a world) is worse than a territory
constantly under threat of being conquered by enemies: it is constantly
under threat of vanishing into thin air.
That being the case, a major consequence follows, of a sign opposite to
the Quinean puritanism mentioned earlier. Just as, in the absence of a
complete system of representations or a complete world, we are to maxi-
mize the consistency, connectedness, and inclusiveness of what systems of
representations or of intentional objects we do have, in the absence of a
logic we are to maximize our closeness to one, walking away from the
depopulated citadels of the propositional and the predicate calculi toward a
finer and finer appreciation of the logical distinctions between “crowd”
and “mob,” or “magenta” and “scarlet.” In a true Kantian vein,
completeness will be not actual but set as a task, so logic will graduate
from a tenseless doctrine into a concrete practice ready to uncover seman-
tical treasures under any rock, and carry semantical threads around any
corners. The routine of jotting down a few axioms, “formally interpreting”
them by translating them into the stock language of set theory, and
“vindicating” them by proving a completeness theorem will be shunned
in favor of the completeness that really matters: the one that is never
182 Ermanno Bencivenga
achieved but demands that we trace more and more connections,
across a larger and larger field.
The two most obvious examples of this search for an objective logic, in
our tradition, come from Aristotle and Hegel3 – or, I should say, from
Hegel and from Aristotle as interpreted by Hegel. For the official story
about Aristotle is that logic for him is an organon, a neutral tool to be
prefixed to research proper, indeed to be done with exhaustively before
embarking in any research. If that were true, Aristotle’s would no more be
a logic than quantification theory or S4 are: it would be an abstract,
uninformative, and ultimately irrelevant repertory of (logical) platitudes.
But, fortunately, such is not the case (as Hegel points out): every segment
of Aristotle’s philosophy (and science) deepens and widens his logical
analysis4 – his logic as analysis, his analytic logic. Whether he is talking
about the challenge sea-anemones bring to the logical distinction between
animals and plants, or he is illuminating through a careful examination of
courage or friendship the relation between focal and extended/analogical
meanings, Aristotle is reshaping his dictionary (including what it is to be a
dictionary) every step of the way.
So this is Hegel’s Aristotle I am talking about. It is also Kant’s.
Aristotle’s text could be used as a prime example of a commitment to a
TR. But one can also see it as a major avenue that is open to us when we,
within a TI, get to the point of spelling out categorial connectedness; more
precisely, when we spell out the part that has to do with counting objects,
hence identifying and distinguishing them. If we go with Aristotle there
(with the Aristotle that maintains a not-entirely-comfortable presence
inside Kant) then the issue is simple: as soon as we face a contradiction
between two representations, or their intentional objects, a distinction
must be made – there must be at least two things. The Aristotelian world
is structured by contraries: by what cannot be true together and invokes a
splitting. If waves can cause interference phenomena and particles cannot,
then waves are not particles, and for a definition of light we have only two
choices: we can either have particles or waves but not both, or give up on

3
For a systematic account of the contrast between Aristotelian and Hegelian logic, see my 2000.
4
See the following passages from Hegel’s 1995: “in his metaphysics, physics, psychology, etc., Aristotle
has not formed conclusions, but thought the concept in and for itself ” (p. 217; translation modified);
“it must not be thought that it is in accordance with . . . syllogisms that Aristotle has thought. If
Aristotle did so, he would not be the speculative philosopher that we have recognized him to be”
(p. 223); “Like the whole of Aristotle’s philosophy, his logic really requires recasting, so that all his
determinations should be brought into a necessary systematic whole” (p. 223). While he thus
acknowledged the comprehensive character of Aristotle’s logic, however, Hegel did not see it as an
alternative to his own, as I do, but rather as a step toward the latter.
 Logics and worlds 183
both and think of something else entirely. That Kant’s criteria of objective
identity be spatiotemporal shows him committed to this Aristotelian route:
one and the same thing cannot be at two different locations at the same
time. But, when it comes to the semantics of regulative ideas, including the
ones that determine the criteria of objectivity, his inclination seems to be
proto-Hegelian, witness his “derivation” of positive from negative free-
dom, or of reciprocal action from simultaneity.5
With Hegel, on the other hand, contradiction is not a threat: it is an
opportunity. When the semantics of a word faces a bifurcation between
contradictory options, its fate is to take both, and its job is to evolve in
such a way that both options be present in a dialectical overcoming of their
contrast. Light is both particles and waves: the two are complementary
descriptions of one and the same complex reality, indeed belong to the
very substance of that reality, which is nourished (adds to its “concrete-
ness,” Hegel would say) by their antagonism. Therefore the world that no
one will ever experience but of which everyone takes herself to be experi-
encing a portion is a monistic one: as not even contradictions can divide,
no two things are radically divided; all divisions are but chapters of one
story. And the very unfortunate events that might bring this logic to a crisis
will not be the surfacing of contradictions, as is the case with its analytic
counterpart. It will rather be the confronting of occurrences (the Holo-
caust, say) that simply cannot be integrated within one and the same
comprehending, rationalizing, “spiritual” narrative.
I said that these are the two most obvious examples of the search for a
logic. There are countless, less obvious, others; except that they are not to
be found where one would be most likely to look for them. As I pointed
out already, individual calculi cannot be regarded as logics, unless they are
part of an ambitious program that extends over a substantial area of
experience, indeed potentially all of it. But, whereas most of what falls
under the academic discipline of logic does not qualify as logic for me, a lot
of traditional philosophy does. Transcendental philosophy is not a new
way of doing philosophy initiated by Kant: it is a new way of looking at
what philosophy has always done, without much awareness and hence
with considerable self-deception. Of course pre-Kantian, and many post-
Kantian, philosophers typically took themselves to be establishing factual
claims like the existence of God or human freedom, but the way they did

5
For the former, see Groundwork of the Metaphysics of Morals, in Kant 1996, p. 94: “The preceding
definition of freedom is negative and therefore unfruitful for insight into its essence; but there flows
from it a positive concept of freedom. . .”. For the latter, see my 1987, p. 149.
184 Ermanno Bencivenga
this was by launching apodeictic demonstrations, that is: by trying to
prove that the existence of God or human freedom were more than facts –
that they were necessities. Therefore, it was not facts about God or
freedom they were directly addressing: it was the meanings of “God” and
of “freedom,” and facts only insofar as they were inescapable consequences
of what those meanings were. Virtually every one of them was doing,
largely unbeknowst to himself, transcendental philosophy, which is to say:
(transcendental) logic. The logic of the State and of justice, if they were
Plato or Hobbes; the logic of art, if they were Plotinus or Schiller; the logic
of economic exchanges, if they were Ricardo or Marx. Far from being just
an organon of philosophy, logic constitutes the very body of it: all philo-
sophical theses, arguments, and theories are but logical matters – pages of
an ideal dictionary by which we try to make sense of experience. And, as
I showed earlier with Aristotle, most of these theses, arguments, and
theories, though often grown on TR soil, can be put to profitable use in
developing a TI, specifically in defining objectivity for various TI philoso-
phies/logics.
If each logic (within a TI framework) develops its own definition of
objectivity, one important feature of logic as ordinarily understood turns
out to be mistaken – and one more way emerges in which traditional
philosophers were deluded about their own work. The feature I refer to is
that logic, one often believes, allows us to conclusively refute an opponent
by proving him conceptually confused, or to conclusively establish our
position by proving it sustained by necessary argument; which is just what
philosophers traditionally took themselves to be doing when they offered
the apodeictic demonstrations I mentioned. Here too it might help to get
to our point by a digression through Kant, this time through Kant’s ethics.
(In what follows, for the sake of simplicity, I will adopt an analytic,
Aristotelian perspective, hence speak of inferences, that only have currency
within an analytic framework. Hegelian, dialectical logic has other conver-
sational and confrontational modes – which explains the repeated failure of
attempts at coming up with a dialectical theory of inference. But mutatis
mutandis what I say could be extended to the Hegelian camp.)
According to Kant, ethics is rationality,6 and never mind at the moment
that rationality, like objectivity, cannot be definitively established. Suppose
we pronounce an ethical judgment that, on as solid grounds as we can
manage, stigmatizes a certain behavior as immoral. Assuming the grounds
to be as solid as they appear, this is a case of reason itself speaking, and one
6
For a detailed discussion of this thesis, see my 2007.
 Logics and worlds 185
would imagine that, when reason speaks, everyone will stop and listen.
Not so: behavior occurs in empirical reality, Kant thinks, subject to
empirical laws, so only empirical factors like temperament, education, or
emotions can have motivating force there. Reason has none. One can only
hope (if one takes reason’s side) that those empirical factors will promote
what reason would want to see done; that moral feeling, say, will ally itself
with rational judgment and make the agent move the “right” way – in this
case, take his distance from the stigmatized behavior. If the agent decides
otherwise, there is nothing reason can do. It can call the agent irrational,
even deny him the status of an agent; but the (non?)agent need not be
impressed by any of this. In fact, he can appropriate words like “reason”
and “rational” and provide them with his own semantics; and there will be
no forcing him to recognize that as an error. Reason (whatever that is) is
playing its own game and, however consistent and connected the game
might be, one can always, simply, opt out of it.
Same thing here. Every transcendental philosophy/logic sets out its own
game, to be played by its own set of rules. Now suppose that, by the rules
current in a particular game, I prove that, than which nothing greater can
be thought, necessarily to exist. If I am a believer, I rejoice in thus seeing
my faith confirmed and I generously broadcast my proof to all others, so
that they can see the light also. And I am puzzled when many of those
others, instead of coming to a harmonious, reasonable agreement with me,
use their disbelief as the premise of a modus tollens and start looking for
what is wrong with my proof. Eventually they might focus on something
I took to be included in the semantics of “greater”: that existing, say, is
greater than not existing. And they might deny it: adopt an alternative
account of “greater.” What can I do then? Clearly, they are playing a
different game; and, no matter how loud I protest, there is no convincing
them that my game is the one they should be playing. This last judgment is
internal to my game, and of course from that internal perspective it looks
irrefutable. From the outside, it just looks like something else one
could say.
In and by itself (more about this qualification shortly), logic has no
persuasive force. Despite the metaphors of constraint that are invariably
brought up in its wake, it can constrain no one. If anything, the practice of
logic (as opposed to the often deceptive theory of it) has a liberating effect.
You felt constrained to making a certain inferential step (say, from some-
thing being necessary to its being necessarily so); but, when you bring
logical acuity and attention to bear upon it, you realize that it was a matter
of habit, that you can dislodge yourself from that straightjacket and make
186 Ermanno Bencivenga
the step not a forced but an optional one – that you are free to go either
way, you have a choice in the matter that you had missed at first. Think of
the story long told about Girolamo Saccheri:7 of how he wanted to firm
up, once and for all, the necessity of Euclid’s fifth postulate (to withdraw
the option of having it, or not having it, as an independent assumption)
and ended up unwittingly freeing thought from Euclidean fetters.
What, then, is the use of a logic? How is the search for its objectivity
ever going to pay off? Its value judgments, I said, are internal to the game
the logic is playing; it is only from within that game that certain principles
appear secure, certain inferential steps apodeictical, certain objections
untenable. So it is only internally that a logic, in and by itself, has a use.
The development of my transcendental philosophy/logic will be like the
development of an organism: a realization of its own potential and a
functional interaction of all its components. Repeatedly, I will come upon
theoretical options, and the game I am playing (I decided to play,
I committed myself to playing) will sometimes determine my choice of
one of them, in which case I will “naturally” accept that choice, and
sometimes not, in which case I will reflect on what else I want to add to
the rules of the game in order to have it cover more ground, to make it
more delicately responsive to the rugged terrain on which I must travel.
And the places I get to by traveling on that terrain will retroact on my
initial commitments: I will regard those commitments as confirmed to the
extent that I approve of my destination; I will correct them to the extent
that I find it unwelcome. There will even be surprises along the way:
locations I never thought I would reach but my rules irresistibly take me
to, either to be more powerfully reassured that I am on the right track, or
more anxiously aware that I must be doing (assuming) something wrong.
A logic is a self-organizing structure, self-enclosed and self-referential,
that provides the bare scaffolding of a world and, if given enough data,
even a large part of its actual construction. (So, as anticipated earlier, a
logic includes its own ontology.) Luigi Pirandello called it a “corrosive”,
infernal “little machine”8 because from any imagined variation in the
existing circumstances it could engineer, one step after the other, the
most horrid outcomes; and he considered it something to be afraid of.
For me, the fear at issue here is the one that always accompanies freedom.

7
A Jesuit priest and professor of mathematics at Pavia, who published in 1733 Euclides ab omni naevo
vindicatus, a presumed reductio proof of Euclid’s fifth postulate from his other assumptions, long
regarded as the first (unintended) development of a non-Euclidean geometry.
8
Pirandello 1990, pp. 1108–1109.
 Logics and worlds 187
Whatever the outcome, whether horrid or even benevolent, logic is
revealing of our powers: of the creative process by which we shape our
world and hence of the responsibility that follows from it. Decades of
existentialist thought (starting with Kant!) have made it clear that none of
that is taken lightly, or should be.
In the Rhetoric, Aristotle discusses three means of persuasion: ethos (an
appeal to the speaker’s character, intended to suggest authority and cause
respect), pathos (an appeal to the public’s emotions), and logos (an appeal to
reasoning and argument). On the face of it, this taxonomy would seem to
contradict my (and Kant’s) claim that reason has no motivational force,
indeed it might look like the premise of an edifying call to exercising
personal and emotional restraint and letting the austere business of logos
take control of public exchanges and safely guide the community to
perfectly reasonable outcomes. In light of what I said so far, this attitude
would be delusional: if in fact a speaker were to convince an interlocutor to
switch to her side by the use of logical argument, it would only be by
cleverly hiding the optional character of her principles and the controver-
sial nature of her inferential steps, and that itself would happen, most
likely, because of the competence, hence the authority, the interlocutor
attributes to her, hence ultimately because of her implicit use of ethos, her
implicit appeal to her superior ability (and honesty) in dealing with these
matters.9 Rhetoric aficionados are fond of making some such point, and of
collapsing logos into a fraudulent mannerism, which will succeed (when it
does) by couching in impressive, authoritarian pseudorational garb sub-
jective (and often repressive) opinions and policies. This extreme, unwar-
ranted stance issues from a gut reaction to the equally extreme, and equally
unwarranted, claim that logic, in and by itself, can persuade anyone; and
we have now prepared the ground for a more plausible and balanced
posture – and for finally explaining what I have meant by the qualification
“in and by itself.”
When creationists say that evolution is only a theory, they are saying
something clearly, even trivially, true; and their opponents’ angry retorts
that evolution is a fact are only signs of bad faith. But that evolution (and
creation as well) be a theory is the beginning of a story, not the end of one,
as some theories are better than others along a significant list of parameters:
they are more detailed, more discriminating, more resourceful, more
ingenious. And, when compared with creation, evolution is all that. One
can imagine that, at the limit of becoming more and more detailed,
9
Also, crucially, because of her interlocutor’s deference to her authority. See the following footnote.
188 Ermanno Bencivenga
discriminating, etc., a theory could be judged to be the factual description
of a world; as that is one more limit that cannot be reached, we will make
do with what approximations we can reach, and develop theories that
weave a finer and finer texture of a presumed reality. Then we will throw
our theories onto the marketplace of ideas and defend them as best we can,
hoping that others will buy them. Some theories (creationism, for
example) will have a clear advantage in terms of pathos: they will line up
the support of strong feelings – the fear of death, the already mentioned
fear of freedom. Others will have more modest, though no less genuine
emotions on their side: intellectual curiosity, the fascination of complex
solutions, the aesthetic satisfaction of seeing things fall into place. And they
will stir such emotions more, the more detailed, discriminating, resource-
ful, and ingenious they are, that is: the more reason structures them
internally – the reason that is forever asking pointed questions and
expecting relevant answers, the reason Socrates taught us how to use.10
A logic is a highly ambitious theory: one that attempts to construct a
universal language. In and by itself, this theory will be found persuasive only
by those who are already committed to the particular view it expresses and
articulates. But, the more the view is articulated, the more material it includes
and makes fit in a well-organized, thoroughly sensible structure, the more it
will look to others like the groundwork for a majestic cathedral, and the more
they might find it attractive. Despite the attraction, they might never leave
the hovels they are used to, since those give them more comfort and
reassurance; still, however slim a chance logos has of winning over fear, by
eliciting the waner passions germane to itself, this is a chance, that in happy
(safe, relaxed, sociable) circumstances might well come to fruition. Forcefully
asserting our axioms and proudly marching to the tune of our proofs will
never get us but a reputation for arrogance; patiently working out a thing of
beauty and making it a paradigm (an example, that is) of internal richness
and consistency might make a few others want to play with us. Not because
they have to; but because that internal richness and consistency – the logos
that internally paces it – might make them feel that it would be fun to do.
And so they might, if perhaps only for a while, come to inhabit our world.

10
Similar points could be made about ethos. The character and competence of a speaker, in and by
themselves, will have no power to persuade an audience unless the latter feels respect for them. So, as
appropriate to a discussion of persuasion – that is, of how an audience can be manipulated – , it is
always pathos (the emotions the speaker is able to instigate in the audience) that works if anything
does; and the real distinction is among the emotions that are in play. Both logos and ethos will only
be successful if the speaker can raise the emotions akin to them, and if these emotions, under the
circumstances, prevail over conflicting ones.
 chapter 11

Bolzano’s logical realism

Sandra Lapointe

1. Framework
The term ‘logical realism’, as it is commonly understood, picks out a
family of views that are committed to at least two theses.1 The first, let
us call it ‘(LF)’, is that there are “logical facts”. Here (LF) is construed in
the widest possible sense to include any theory that assumes that there is a
fact of the matter when it comes to the truth-value of claims about logic. (LF)
can thus be cashed out in more or less robust terms. Take for instance the
putatively true claim that modus ponens is a valid principle of inference.
The realist may be committed to there being “something” – whatever this
turns out to imply – that makes the claim that modus ponens is valid true.
Or she may understand the idea that the validity of modus ponens “is a fact”
to mean merely that the corresponding claim is true. Both interpretations
of (LF), and every other one in between, raise a number of questions that
go beyond the scope of the present chapter. (For instance: what is truth?
What is it for a fact to “make true” a truth?) What is relevant here is the
following: whatever she understands “logical facts” to be, what makes the
adherent to (LF) a realist about logic is a further assumption (IND), that
“logical facts” are independent of our cognitive and linguistic make-up and
practices; they are independent of our minds and languages. In this sense,
for the logical realist the truth or falsity of logical claims is “objective”.
History offers a number of theoretical alternatives to logical realism.
What’s common to nihilism, pragmatism and pluralism,2 for instance, is
the fact that they deny (LF). By contrast, the proponents of naturalism
(of which there are many variants, including logical psychologism) and

My thanks to Matt Carlson, Nicholas F. Stang, Penny Rush, David Sanson, Ben Caplan and Peter
Hanks, Julie Brumberg and Teresa Kouric for their input on previous versions of this chapter.
1
This characterization of logical realism draws on Resnik’s 2000: 181.
2
I revert to the definition of “pluralism” given by Stewart Shapiro in the chapter included in this
collection.

189
190 Sandra Lapointe
conventionalism need not reject (LF). A psychologistic logician – think of
John Stuart Mill – need not disagree with the idea that there is a fact of the
matter as to whether or not ‘Modus ponens is valid’ is true. Rather, he
might be denying that the truth of this claim can be established independ-
ently of psychological knowledge and therefore independently of certain
facts concerning our mind. Likewise, the conventionalist may assume that
there are determinate facts concerning our (linguistic) practices that deter-
mine whether, for instance, the claim that modus ponens is valid ought to
count as true or false. What logical psychologism and conventionalism
share is the fact that they reject (IND).
As I’ve characterized it so far, logical realism is compatible with certain
kinds of relativism. In the chapter included in this volume, Shapiro
describes the view he calls “logical folk-relativism”. While one who holds
this view assumes (i) that the truth-value of ‘y is a consequence of x’, for
instance, varies from one logical framework to another; she also admits (ii)
that there is a fact of the matter as to whether y is a consequence of x in a
given framework (i.e., LF); and (iii) that this fact is objective (i.e., IND).
For the purpose of this chapter, I will use ‘logical realism’ in a narrower
sense that does not include relativism of this sort; the type of logical
realism I will be discussing below is “monistic”.
The ontological questions that underlie logical realism – e.g. what kinds
of “facts”, if any, ground the truth or falsity of logical claims? – are to be
strictly separated from the types of concerns that arise when explaining
how we come to know the truth of a claim about logic. The distinction
between questions about the epistemology of logic and questions about its
metaphysics is important, among other reasons, for assessing the consist-
ency of some theories. Take Edmund Husserl, for instance. At least in the
first edition of the Logical Investigations (1900–1901), he adopts a form of
logical realism of the more robust kind. What makes claims about, say,
validity, true according to the Logical Investigations are certain features of
abstract entities that exist independently of us: “Bedeutungen”. Nonethe-
less, Husserl believed that the only way to know the truth-value of logical
claims is to engage in certain (admittedly rather esoteric types of ) psycho-
logical analyses. Whatever its other merits, Husserl’s theory is not
inconsistent. The ontological position according to which there are
mind-independent “logical facts” need not be at odds with the epistemo-
logical position according to which we can only discover the truths of logic
through an investigation of the mind. More generally, logical realists, while
they hold that a claim about logic, if it is true, is true independently of
what we believe or do, may also believe that the recognition of the
 Bolzano’s logical realism 191
truth-value of such claims require us to investigate the way our brain or
mind works and/or reflect upon our cognitive abilities, psychological
dispositions, linguistic conventions or other uses and practices. The alter-
native would presumably be to assume that we come to recognize the
truths of logic through some sort of immediate logical “grasp”. And while
this cannot be excluded a priori, it is an assumption that might seem
dubious to anyone who has ever taught introductory logic to college
students.3
Logical realism raises a number of interesting metaontological questions.
Consider two simple toy semantic theories. Let us assume that the content
of both theories is defined as the set of true instances of:
(1) s means p
where ‘s’ is taken to stand for a sentence of natural language and ‘p’ for the
meaning of this sentence, say the “proposition” that p. To the extent that
one holds that at least some instances of (1) are true, both theories commit
one to there being sentences and, more controversially, to there being
propositions. What makes the two theories different theories may be a
variety of things: they may diverge on which instances of (1) are true, they
may rest on different accounts of what a sentence is, or have different views
on what propositions consist of (e.g. structured entities, sets of possible
worlds). Or they may agree on all this and still not be identical.
When it comes to comparing types of logical realism, differences that
reside in the metatheory, and in particular in the kinds of “grounds” that
underlie commitment to the existence of proposition-like entities, can be
especially enlightening. I want to take the notion of ground in a broad,
intuitive sense: Agent A’s belief that x is a ground for her belief that y if her
holding x to be true has explanatory value when it comes to accounting for
A’s belief that y. The notion of explanation used here is to include the case
in which y follows from x (in a sense of ‘follows’ to be specified) as well as a
range of other cases I will discuss below. What’s peculiar about all these
cases is the fact that the relation between y and x is to be construed in
epistemic terms. As I use the terms ‘ground’ and ‘explain’ here, whether y
“objectively” follows from x is not ultimately what matters when it comes
3
The theory of “eidetic variation” and “Wesensschau” Husserl eventually committed to is an instance
of this kind of epistemology, and this explains in good part why, in many circles, his theories
eventually fell into disrepute. Gödel adopted a similar view, and one directly inspired by Husserl.
(Cf. Kennedy 2012.) The idea that the realist might be bound to adopt an epistemology seems to be a
common objection to the doctrine as a whole. In her chapter in this volume Penny Rush argues for
the potential of phenomenology as regards this problem.
192 Sandra Lapointe
to determining whether A’s belief that y can be explained by A’s belief
that x. It is sufficient in order for A’s belief that x to have explanatory value
(to be a ground) in the relevant sense when it comes to accounting for A’s
belief that y that A effectively believes that y is a consequence of x. This
qualification is important if we are to account for the fact that grounds that
are unclear, implausible or otherwise mistaken nonetheless have explana-
tory value when it comes to understanding an agent’s motivation for
certain claims. If it makes sense to say that A holds the belief that y because
A holds the belief that x then A’s belief that x – and the corresponding
claim – is a ground for A’s belief that y.
There are at least two kinds of ground to adhere to (LF) and (IND) and,
accordingly, two main types of realism in logic. The proponent of logical
realism may have “external” grounds to assume that there are putative
logical facts, even if these grounds are implicit, unconvincing or otherwise
flawed. In the context of logical realism, what I mean by “external
grounds” are grounds that arise out of a concern that is not itself for
logic. While it might be difficult to define precisely what counts as a logical
concern, the idea that some concerns pertain to logic while others don’t is
uncontroversial enough. On the contemporary understanding the defin-
ition of validity and logical consequence belongs to logic – construed
widely enough to include semantics. The investigation of what is involved
in perception and cognition, what moral principle(s) we should abide by
and what there is in the world, by contrast, do not.
The (more or less well defined) boundaries between the various philo-
sophical subdisciplines are not hermetic and indeed are often such that the
grounds we have to hold a belief in one, are effectively driven by another.
For instance, there’s nothing that forbids that a logical realist’s grounds to
commit to (LF) and (IND) be external to the extent that they are driven by
metaphysical concerns i.e. concern for what there is in the world in addition
to rocks and chairs (assuming that there are such things). But this sort of
“metaphysical” realism in logic is uncommon – if it exists at all – and the
kinds of grounds that underlie realist commitments, when they are external,
are typically not metaphysical. The realist’s grounds for positing mind and
language-independent logical facts, when they are external, are typically
driven by other aspects of her philosophical theory altogether.
One may be a logical realist on epistemological grounds, for instance.
Take Leibniz’s arguments that truths exist eternally in the mind of God,4

4
“But it will be further asked what the ground is for this connection, since there is a reality in it which
does not mislead. The reply is that it is grounded in the linking together of ideas. In response to this
 Bolzano’s logical realism 193
and that God “displays” (some of ) these truths to us.5 The former commits
Leibniz to truths that are independent of human minds (and language).
And taken together, these two assumptions explain how human knowledge
is possible on Leibniz’s view. Leibniz’s primary concern in introducing
propositio is not for what there is, but for how we acquire knowledge.
Leibniz’s grounds to commit to the existence of proposition-like entities
are thus (in part) that the supposition of such entities – and the further
assumption that a benevolent God exists! – is required to provide a
coherent theory of knowledge. Similarly, Popper’s grounds for thinking
that there is “objective knowledge” – in (Popper 1968) for instance –
whatever their merit, is that this allegedly explains certain features of the
sciences, such as the relatively autonomous character of scientific theories
and problems. Whether they are epistemological or otherwise, as long as
the logical realist’s grounds for believing in the existence of logical facts are
not themselves logical, I will call the kind of realism she adopts ‘external’ or
‘extra-semantic’.
One’s grounds to subscribe to (LF) and (IND) and to the idea that there
are proposition-like entities, in particular, need not be external. What
often underlies one’s commitment to logical facts may correspond to
(implicit) theoretical desiderata or aims. Desiderata and aims are types of
grounds in the relevant sense: they have explanatory value when it comes
to accounting for the ontological commitments that come with a logical
theory. Let us call the kind of logical realism that would underlie such a
theory ‘internal’. Historically, many instances of realism in logic have been
internal. The exact nature of the grounds that underlie the internal realist’s
commitment to logical facts vary. It may be that the logician desires to see
certain “intuitions” satisfied or certain epistemic “purposes” fulfilled by the
logical theory. Why precisely these intuitions and purposes ought to be
satisfied by the theory is bound to be a matter of contention, but there’s a
case to be made to the effect that they pervade logic and its philosophy.6
What I call “intuitions” here correspond to certain claims that seem
more certain, more epistemically salient or otherwise accessible to the

it will be asked where these ideas would be if there were no mind, and what would then become of
the real foundation of this certainty of eternal truths. This question brings us at last to the ultimate
foundation of truth, namely to that Supreme and Universal Mind who cannot fail to exist and whose
understanding is indeed the domain of eternal truths. . .That is where I find the pattern for the ideas
and truths which are engraved in our souls.” IV.xi.447 (my emphasis added).
5
IV.v.397. I wish to thank Chloe Armstrong for an informative discussion concerning this point.
6
The exact nature of the distinction between intuitions and purposes would benefit from a closer
investigation, but this would go beyond the scope of the present chapter.
194 Sandra Lapointe
agent (though there might not be a fact of the matter as to whether they
really are). The logical realist may be convinced, for instance, that truth,
whatever it is, is “immutable”, in the sense that it cannot be changed or
destroyed and she won’t regard the theory as adequate unless the immut-
ability of truth is a consequence of it. Since she is also likely to hold the
belief that individual sentence- and thought-tokens do not persist indefin-
itely (for they don’t) and thus cannot be the fundamental bearers of truth
(and falsity), she might deem it necessary to introduce ontologically robust
abstract entities, precisely in order to satisfy this intuition. If that is the
case, then A’s (desire to satisfy this) intuition has explanatory value when it
comes to accounting for her commitment to proposition-like entities:
there is a definite sense in which A believes that there are propositions
because she believes that truth is immutable.
Quantifying over meanings may also serve certain more or less clandes-
tine “purposes” within the theory. The logical realist may, for instance, be
guided by the fact that systematically including instances of (1), above, in a
semantic theory (surreptitiously) introduces a paraphrastic procedure
that can be used to “clarify” natural language sentences or make them
more “exact”.7 If one’s motive, be it explicitly or not, in introducing the
semantic operator ‘means’ and in quantifying over propositions are the
(stealthy) epistemic gains that come from translations of this type, one’s
grounds to commit to propositions are subservient to the semantic theory
and the type of realism they embrace is internal.
Admittedly, in certain cases, it could be unclear whether one’s grounds
are internal or external. Take the case in which A’s belief that there are
logical facts is the consequence of certain assumptions concerning
the relation between language and the world. A may believe that there
are objective logical facts because A believes (TM):
(TM) The truth of a claim implies its correspondence to something that
makes it true (or the existence of a “truth-making” relation), whatever this
turns out to be.

Assuming that some claims about logic are true, (TM) implies the exist-
ence of entities that fulfil this truth-making role in logic. Commitment to
(TM) explains the commitment to logical facts. Indeed, (TM) epitomizes
7
See (Lepore and Ludwig 2006). They write: “The assignment of entities to expressions, which was to
be the key to a theory of meaning, turns out to have been merely a way of matching object-language
expressions with metalanguage expressions thought of as used (in referring to their own meaning), so
that we are given an object-language expression and a matched metalanguage expression we
understand, in a context which ensures that they are synonymous” (Lepore and Ludwig 2006: 31).
 Bolzano’s logical realism 195
precisely the type of “full-bodied” correspondence Wittgenstein’s Tractatus
was meant to put into question.8 But is (TM) an “external” ground to
subscribe to logical facts? After all (TM) is a claim that belongs to the
metaphysics of logic and one could argue that logic does include its own
metaphysics. This raises a question – what is the scope of logic/semantics –
which I am inclined to answer liberally but which I leave open for now.
It is sufficient for our purposes that this question has an answer in principle
even if it is a difficult one.

2. Bolzano’s internal realism in logic

Bolzano’s Theory of Science (1837) presents the first explicit and methodical
espousal of internal logical realism. It also contains a formidable number of
theoretical innovations. They include (i) the first account of the distinction
between “sense” (Sinn, Bedeutung) and “reference” (or “objectuality”:
Gegenständlichkeit), (ii) definitions of analyticity and consequence, i.e.
“deducibility” (Ableitbarkeit) based on a new substitutional procedure that
anticipates Quine’s and Tarski’s, respectively, and (iii) an account of math-
ematical knowledge that excludes, contra Kant, recourse to extraconceptual
inferential steps and that is rooted in one of the earliest systematic reflections
on the nature of deductive knowledge. (i)–(iii) all assume the existence of
mind- and language-independent entities Bolzano calls “propositions and
ideas in themselves” (Sätze an sich). Take (i) for instance. Appeal to propos-
itions in themselves in this context serves Bolzano’s antipsychologism in
logic: according to Bolzano, the sense (Sinn) of a sentence – the proposition
it expresses – is to be distinguished from the mental act in which it is
grasped. Just like what is the case in Frege, a sentence has the semantic
properties it has (e.g. truth) on Bolzano’s account derivatively, by virtue of
its relation to mind-independent entities: the primary bearers of semantic
properties are the propositions that constitute their Sinne.
Bolzano’s version of logical realism is among the more robust. It yields a
unique form of “semantic descriptivism”: there are objective,9 immutable10
entities, “propositions” (for short), that bear certain properties and rela-
tions, which it is the task of logicians to describe. The first two books of the
Theory of Science (together they make up the Theory of Propositions and

8 9
On this, see (Mulligan, Simons and Smith 1984: 289). Cf. (Bolzano 1837, §21: 84).
10
Cf. (Bolzano 1837, §125: 7). That truth is immutable – that ‘is true’ is not a relativized predicate – is
thus an intuition Bolzano’s theory seeks to satisfy and one of the grounds that motivates his
commitment to logical realism.
196 Sandra Lapointe
Representations in Themselves) are divided into chapters whose headers
include reference to propositions’ and representations’ “general character-
istics”, “properties”, “relations” (among themselves, to objects) and
“internal constitution”. Bolzano even devotes entire sections of the book
to the analysis of claims in which such properties and relations are ascribed
to representations and propositions.11 As Bolzano sees it, truths of logic –
including definitions of such notions as meaning, analyticity and apriority –
amount to descriptions in which (often multifaceted) properties or relations
are ascribed to propositions in themselves, their parts or classes thereof.
Bolzano’s descriptive approach to logic is both original and noteworthy.
Nonetheless it comes with an explicit commitment to the existence of
certain kinds of non-natural entities which, because it is explicit and
indeed unequivocal, is perhaps somewhat perplexing: one could be left
with the impression that Bolzano’s ontology of logic is a more direct target
for standard naturalistic objections than some other varieties of logical
realism.12 There are at least two grounds why this impression is misleading.
First, to the extent that ontological commitments cannot be measured on a
scale and that all logical realists subscribe to (LF) and (IND), all variants of
realism are equally ontologically “candid” from a naturalistic standpoint.
There is in principle nothing more ontologically damning about Bolzano’s
semantic descriptivism than about the kind of realism Frege will eventually
put forward in (1918).
Second, while ontological commitments do not come in degrees,
metatheoretical considerations are not irrelevant and some kinds of
“grounds” for positing non-natural entities may be more palatable to the
naturalist than others. The naturalistic criticism of logical realism is typically
motivated by a concern for metaphysical economy (Is it consistent to
postulate entities that do not exist in the causal realm?) or by related
epistemological reservations (What does it mean to “grasp” or cognize or
be epistemically related to something that does not exist causally?). For this
reason, one who has independent (external) metaphysical or epistemological
grounds to subscribe to (LF) and (IND) is a more direct target for naturalistic
criticism. But Bolzano’s commitment to the existence of non-natural
entities, while it is uncompromising, is also clearly motivated by the kind
of internal grounds that makes him least susceptible to the naturalistic
concern.

11
See for instance (Bolzano 1837, §§164–168).
12
There are other types of objection to realism. Rush (in this volume) for instance discusses Sellars’
objection. I won’t be discussing this point.
 Bolzano’s logical realism 197
As Bolzano sees it, the main reason for positing propositions is their
“usefulness” for certain theoretical purposes, in particular for the purpose
of reaching satisfactory definition of logical notions (understood broadly):
The usefulness of the distinction [between propositions in themselves and
thought propositions] manifests itself in tens of places and in the most
surprising way in that it allows the author to determine objectively a
number of concepts that had not been explained before or that were
explained incorrectly. For instance, the concept of experience, a priori,
possibility, necessity, contingency, probability, etc. (1839: 128)

Bolzano is clear that the positing of propositions should not be considered

to have bearing outside of logic. As Bolzano sees it, logicians should be
allowed to appeal to entities that may reveal themselves to be inconsistent
with paradigmatic metaphysical and/or epistemological theories:
Thus, to give another example, the logician must have the same right to
speak of truths in themselves as the geometrician who speaks of spaces in
themselves (i.e., of mere possibilities of certain locations) without thinking
of them as filled with matter, although it is perhaps possible to give
metaphysical reasons why there is no, and cannot be any, empty space.
(1837, §25: 113–14)

What’s perhaps most remarkable about Bolzano’s internal logical realism is

the fact that, while he argues that he needs to posit what he calls ‘propos-
itions’ to arrive at satisfactory definitions, and while he assumes that the
bearers of the logical properties and relations he defines in fact bear this
name: Bolzanian definitions of logical notions are in principle compatible
with a number of different ontologies. This idea – it amounts to claiming
that logic is “topic neutral” – emerges from a series of remarks Bolzano
makes in a text he published some years after the Theory of Science, the
Wissenschaftslehre (Logik) und Religionswissenschaft in einer beurtheilenden
Uebersicht (1841) whose (failed) purpose was to arouse the public’s interest
for Bolzano’s theories. Bolzano writes:
Everything the author asserts of propositions in themselves in the first
section – with the exception of what he says at §122, namely that they
don’t exist – holds of thought propositions; likewise, in the second section,
the “Differences amongst propositions as regards their internal properties”
are all such that whoever admits of thought propositions can also admit
of them. (Bolzano 1841: 50)13

13
See also (Bolzano 1841: 34–35).
198 Sandra Lapointe
But now the question arises whether someone who rejects the concept of
propositions in themselves and accepts only that [for instance] of thought
propositions could nonetheless admit of a connection amongst the latter
more or less like the one Bolzano describes as objective. And this, we think,
should be answered in the affirmative. (Bolzano 1841: 68)
On the face of it, the kind of semantic descriptivism Bolzano adopts seems
incompatible with the claim that someone who rejects the notion of
proposition could still admit his definitions of logical notions. Indeed such
statement contradicts what Příhonský, Bolzano’s close collaborator, seems
to have assumed in the New Anti-Kant, namely that:
All will be lost if they cannot grant us this concept [of a proposition], if they
keep representing truths in terms of certain thoughts, appearances in the
mind of a thinking being . . . (Příhonský 1850: 5)
If Příhonský is right, Bolzano’s move – the claim that definitions of logical
notions are topic neutral – is at best a rhetorical concession made in order
to win a reluctant public. The problem with this exegetical line is not that
it is implausible. The problem is that it does not do justice to the
coherence of Bolzano’s views. Notwithstanding what Příhonský assumes
(more on this below), and even granting that Bolzano as an internal logical
realist has less of an axe to grind when it comes to defending the existence
of propositions, it remains that the claim that definitions of logical notions
are topic neutral is not a mere rhetorical ploy. For Bolzano has the
theoretical resources to make sense of this idea systematically. This is what
I argue in Section 3. Nonetheless, if part of Bolzano’s point is that the value
of a logical theory does not reside in the nature of the entities that bear the
properties it defines, but in the properties and relations they are meant to
epitomize and that he would be willing to revise some of his ontological
commitment as long as some other aspects of his theory are preserved, the
onus is on him to show that his theory does present an advantage over that
of his predecessors and contemporaries. Section 4 is dedicated to arguing
that it does.

3. Topic neutrality and implicit definition

Bolzano claims that everything he asserts of propositions in themselves,
with the exception of their being non-actual, in the first and second
section of the second volume of the Theory of Science – what Bolzano calls
“general characteristics” and “differences that arise from their internal
 Bolzano’s logical realism 199
constitution” – holds of thought propositions, the doxastic states in which
such propositions are grasped.14 These characteristics and differences
include the following:

(i)
For all x, if x is a proposition, then x contains several ideas (§123)
(ii)
For all x, if x is a proposition, then x can be viewed as part of
another proposition, even a mere idea (§124)
(iii) For all x, if x is a proposition, then x is either true or false (for always
and everywhere) (§125)
(iv) For all x, if x is a proposition, then x is of the form ‘A has b’
(§§127–128)
(v) For all x, if x is a proposition, then the extension of x is identical
with the extension of the subject-representation of x (§130)
(vi) For all x, if x is a proposition, then x is either simple or complex
(§132)
(vii) For all x, if x is a proposition, then x is either conceptual or
intuitional (§133)
and so on. What is relevant here is the following observation: while
(i)–(vii) take the form of descriptive statements, it is more accurate to
think of Bolzano as resorting to what he calls “definition on the basis of use
or context” (1837, §668: 547), that is, implicit definitions. The idea is that
in (i)–(vii), ‘proposition’ designates a primitive (simple) concept (i)–(vii)
define implicitly.15 Bolzano was aware from very early on of the benefit of
this procedure when it comes to defining primitive notions. Though
Bolzano’s paradigmatic examples come from mathematics, the procedure
applies across the board, including in logic. It consists in:

stating many propositions in which the concept that needs to be under-

stood occurs in different combinations and which is designated by the word
that is associated to it. By comparing these propositions, the reader himself
will abstract exactly the concept designated by the unknown word. Thus for

14
The same holds for what he describes as the “objective connections” between propositions,
including “formal properties”. More on this in the next section.
15
That ‘proposition’ is a simple concept is something Bolzano suggests at (1837, §128: 8) when he
writes: “From the mere fact that representations are the components of propositions we cannot infer
that the concept of a representation must be simpler than that of a proposition. On the contrary,
there is a lot to say for the idea that this mark which I use in §48 merely as an explanation of the
concept of a representation is the actual definition of the latter.” At (1837, §48: 216) Bolzano had
written: “Anything that can be part of a proposition in itself, without being itself a proposition,
I wish to call a [representation] in itself. This will be the quickest and easiest way of conveying my
meaning to those who have understood what I mean by a proposition in itself ”.
200 Sandra Lapointe
example, every one can grasp which concept is designated by the word point
on the basis of the following propositions: the point is the simple object in
space, it is the limit of the line without being part of the line, it is extended
neither lengthwise nor according to width, nor according to depth, etc. As
is well known, this is the means through which we all learn the first
significations of our mother tongue. (1810, II, §8: 54–55)

What defines propositions, then, ultimately, is the system of all relevant

implicit definitions. Implicit definitions (including (i)–(vii)) define
propositions as much as necessary for the purpose of Bolzano’s logic. In
(i)–(vii), ‘proposition’ occurs only as the name of what is in effect being
defined and, in this light, part of Bolzano’s point is that substituting
‘thought (proposition)’ for ‘proposition (in itself )’ in (i)–(vii) has no
bearing on the nature and structure of the properties and relations
involved. Indeed, Bolzano has nothing to object to someone who would
claim that the bearer of the properties involved in (i)–(vii) have further
properties, e.g. the property of being types of mental processes, as long as
she admits that mental processes have the properties involved in (i)–(vii).
There is at least one other reason to take Bolzano’s suggestion that his
definitions are topic neutral seriously. Bolzano (1841, 68) claims that
“objective connections” need not be predicated of propositions, that
properties such as validity (§147), analyticity (§148), compatibility (§154)
and deducibility (§155) – “formal” properties – could equally well be
recognized by one who admits only thoughts. By a formal property,
Bolzano means a property that is defined for “entire genera of propos-
itions”, on the basis of the substitutional procedure.16 Beyond what I’ve
argued above, I want to show that Bolzano’s definitions of what counts as
formal properties are topic neutral in the relevent sense.
Formal properties are not properties of individual propositions but prop-
erties of what Bolzano calls ‘forms’, i.e. schematic expressions.17 Bolzano
makes copious use of schematic expressions – or their equivalent18 – when it

16
Cf. Bolzano (1837, §12: 51) where he explains (my emphasis): “The clearest definitions say hardly
more than that we consider the form of propositions and ideas when we keep an eye only on what
they have in common with many others, that is, when we speak of entire species or genera of the
latter. . . . one calls a species or genus of proposition formal if in order to determine it one only needs to
specify certain parts that appear in these ideas or propositions while the rest of the parts which one calls the
stuff or matter remain arbitrary.”
17
Cf. (Bolzano 1837, §9: 42f ).
18
Bolzano often speaks of propositions containing “variable representations” and he does not always
revert to schemata to indicate variability. If [Caius] is taken to be variable in [Caius who has
mortality, has humanity], the latter can in principle be designated by the schematic expression
‘X who has humanity, has mortality’; and [Caius is Caius] by ‘A is A’.
 Bolzano’s logical realism 201
comes defining formal notions. Take for instance Bolzano’s claims that there
are logically analytic or tautological propositions. Bolzano writes:
The following are some very general examples of analytic propositions
which are also true: “A is A”, “An A which is a B is an A”, “An A which
is a B is a B”, “Every object is either B or non-B”, etc. Propositions of the
first kind, i.e., propositions cast in the form “A is A” or “A has (the
attribute) a” are commonly called identical or tautological propositions.
(Bolzano 1837, §148: 84)
In this passage, Bolzano ascribes the property of being logically analytic to
individual propositions, yet his examples – “A is A”, “An A which is a B is
an A”, “An A which is a B is a B”, “Every object is either B or non-B”,
etc. – are not examples of individual propositions at all. If we follow what
Bolzano says in the Theory of Science, ‘A is A’ does not stand for any
“proposition” in particular. On Bolzano’s account schematic expressions of
the kind ‘A is A’ represent classes of propositions that are defined through a
substitutional procedure. To say that a proposition “falls under a certain
form” is to say that it belongs to a certain substitution class designated by
this schematic expression.19 ‘A is A’ represents the class of all propositions
in themselves that correspond to the substitution instances of ‘A is A’. If
we use ‘[‘and’]’ to form designations for individual propositions (and their
parts), we find among the propositions designated by the substitution
instances of ‘A is A’ the following:
[Caius is Caius]
[Redness is Redness]
[1 is 1]
and so on.
It is certainly not incongruous for Bolzano to claim that ‘A is A’ – the
schematic expression – is logically analytic. Indeed, it would seem that one
need understand what it means for a proposition to belong to such a class –
or to fall under such a “form” – in order to understand how it itself can be
said to be analytic. The proposition: [Caius is Caius], for instance, falls
under the form ‘A is A’: ‘A is A’ is a “determinate connection of words or
signs” through which the class to which [Caius is Caius] belongs can be
“represented”.20 To say that the individual proposition [Caius is Caius] is
19
In one of his numerous historical digressions, Bolzano notes that “the Latin word forma . . . was in
fact used as equivalent to the word species, i.e. the word class” (Bolzano 1837, §81: 391).
20
See (Bolzano 1837, §81: 393).
202 Sandra Lapointe
logically analytic is to say that it is a member of a class of the latter kind: a
class that can be represented by a determinate type of schematic expres-
sions, namely one all of whose substitution instances designate propos-
itions that have the same truth value.
What’s interesting here is the fact that, on this interpretation, Bolzano is
ultimately committed to the following view of logical analyticity:
(LA) x is logically analytic if x belongs to a class that can be represented by a
schematic expression in which only logical terms occur essentially.

But (LA) is topic neutral. Nothing compels us to think of ‘x’ in (LA) in

terms of a proposition in itself. And as we have seen, once the logical work
is done, Bolzano would not be disconcerted by such a move since “some-
one who rejects the concept of propositions in themselves and accepts only
that [for instance] of thought propositions could nonetheless admit of a
connection amongst the latter more or less like the one Bolzano describes
as objective” (Bolzano 1841: 68).

4. Bolzano’s logic
Internal grounds to adhere to logical facts – or in Bolzano’s case to fully
fledged semantic entities – are typically certain desiderata or aims the
theory is meant to fulfil. In Bolzano’s case, one of the main purposes in
introducing propositions in themselves is to achieve precise and satis-
factory definitions. By way of consequence, on Bolzano’s own account
the success of the endeavour depends on whether his commitment to
propositions allows him to deliver a “good” theory of logic, or at least
one that is preferable to its rivals. To a large extent, Bolzano succeeds. It
is not only that the Theory of Science is furnished with rich and
remarkably well-articulated distinctions and theoretical innovations
but also that he set out to redefine the very nature of a logical investi-
gation in a way that is largely consistent with well-established contem-
porary endeavours.
As Bolzano sees it, at its core, the purpose of logic is to tell us what it
means for something to follow from something else, i.e. what it means for
an inference to be valid or for a claim to be the consequence of some other
claim(s). As an explanation of what it means for a truth to follow from
others, Bolzano’s views on “deducibility” (Ableitbarkeit) are comparatively
close to the ones that have become standard following Tarski in the
twentieth century. Bolzano defines deducibility in the following terms:
 Bolzano’s logical realism 203
Hence I say that propositions M, N, O, . . . are deducible from propositions
A, B, C, D, . . . with respect to variable parts i, j, . . . if every collection of
representations whose substitution for i, j, . . . makes all of A, B, C, D, . . .
true, also makes all of M, N, O, . . . true. Occasionally, since it is custom-
ary, I shall say that propositions M, N, O, . . . follow, or can be inferred or
derived, from A, B, C, D, . . . Propositions A, B, C, D, . . . I shall call the
premises, M, N, O, . . . the conclusions. (1837, §155: 114)
The modern character of Bolzano’s definition, in itself, and especially the
semantic machinery on which it rests is noteworthy enough. On Bolzano’s
account:21
The propositions T, T0 , T00 . . . are ableitbar from S, S0 , S00 with respect to
representations i, j, . . . if and only if:
(i) i, j, . . . can be varied so as to yield at least one true variant of S, S0 ,
S00 , . . . and T, T0 , T00 , . . .
(ii) whenever i, j, . . . are varied so as to yield true variants of S, S0 , S00 . . .,
the corresponding variants of T, T0 , T00 , . . . are also true.
To logicians and philosophers of logic today, the idea that the aim of logic
is to define validity via the elaboration of a theory of logical consequence is
unremarkable. Pointing to the similarities (and dissimilarities) between
Bolzano’s definition of deducibility and Tarski’s definition of logical
consequence has become commonplace in the literature.22 This goes to
show that at least some of the desiderata and aims that underlie Bolzano’s
logic rest on the kind of intuitions that have proven to be enduring. This
should be emphasized for at least two reasons. First, when he published the
Theory of Science in 1837, Bolzano’s views on deducibility were perfectly
anachronistic. For one thing, by the end of the eighteenth century it had
become usual for philosophers to think of logic as invested in the study of
“reason” through an investigation of “thought” and to conceive of such an
investigation to involve the study of mind-dependent operations and
products. Though the methodologies underlying these investigations
varied widely – contrast Locke’s empirical approach in the Essay on Human
Understanding with Kant’s transcendental philosophy – they largely con-
tributed to either discredit formal logic as a discipline23 or, at best, to
convey the opinion that it could not be improved on.24 In this light,
Bolzano’s efforts toward a new logic based on an objective doctrine of
21
For a more detailed discussion of Bolzano’s theory of deducibility, see (Lapointe 2011: 72–90).
22
See, for instance, (van Benthem 1985; George 1986; Siebel 2002; Lapointe 2011).
23
See (George 2003: 99s).
24
See Kant’s famous claim that logic is closed and complete (1781: Bviii).
204 Sandra Lapointe
inferences “in themselves” constitutes an important break from his imme-
diate modern predecessors.
Second, while Bolzano reaches back to Aristotle, his approach to the
definition of validity also marks an important departure from Aristotle and
most of his (early) traditional scholastic commentators.25 Aristotle intro-
duces the notion of a good “deduction” (i.e. syllogism) in the Prior
Analytics. He writes:
A deduction (syllogismos) is speech (logos) in which, certain things having
been supposed, something different from those supposed results of necessity
because of their being so. (Prior Analytics I.2, 24b18–20)
Let us call this the “intuitive Aristotelian notion of validity”. Contempor-
ary attempts at a definition of logical consequence – one may think of
Tarski-type model-theoretic definitions in particular – are generally under-
stood to account for the intuitive Aristotelian notion of validity. The same
holds for Bolzano’s. What makes Bolzano’s account historically distinctive
is the assumption that a good definition of the intuitive Aristotelian notion
of validity needs the support of a semantic theory. In this, his definitional
strategy ought to be contrasted with that of much of the Aristotelian
tradition itself. Aristotle and his early medieval successors are mostly
known for their understanding of validity as epitomized in traditional
syllogistic theories. But traditional syllogistic definitions of validity are
not concerned with providing a semantic account of validity.26 The
standard and paradigmatic methodology behind traditional syllogistic
theories of valid inference, and the one that is best known, is two-pronged.
It first consists in making a list of all possible forms of arguments (syllo-
gisms) and then in identifying those forms whose instances effectively fulfil
the intuitive Aristotelian definition of validity. In order to determine
whether a particular inference is valid, one is thus required to determine
whether it instantiates one of the forms identified as valid.
There are at least three problems, from Bolzano’s perspective, with this
approach. First, traditional syllogistic definitions of validity suppose that
there is a finite (and implausibly small) number of possible forms of
inference. Bolzano is right. If we follow the teachings of the Schoolmen,
25
Here I am not concerned with comparing the Bolzanian and Aristotelian conception of the object of
logic (see Thom, this volume, for such a discussion) but their views on validity.
26
Here, I exclude from what I call “syllogistic tradition” the theories of consequentia that emerged in
the fourteenth century – those we find in Occam and Buridan, for instance. The latter were
attempts to generalize syllogistic and aimed at providing a new insight into the intuitive Aristotelian
notion based on semantic considerations. On this topic, see (Novaes 2012). Something similar holds
for Abelard. I am grateful to Julie Brumberg-Chaumont for this precision.
 Bolzano’s logical realism 205
there are exactly 256. This number comes up as a result of various
assumptions concerning the number of sentences involved in an argu-
ment – three! – and the form of such sentences. In particular, classical
syllogistic theory assumes (i) that only categorical sentences (i.e. sentences
of the form subject-copula-predicate) are involved in arguments, (ii) that
there are four variants of such forms (a, e, i, o) and (iii) that any given
inference contains at most three different terms – subject, middle term and
predicate – which yields four possible syllogistic “figures”. Second, (i)–(iii)
mark out a syntax whose expressive resources are too limited to account for
the richness of actual inferential practices. Hence, it cannot adequately
model (even some of the most basic forms of ) inference. For instance, it
cannot model disjunctive and hypothetical syllogisms that require separate
theories (at least if understood in its original sense, i.e. as a propositional
logic). This is tributary to a third more general problem, namely the fact
that traditional syllogistic definitions of validity are bound to a given
syntax (namely the one defined by (i)–(iii) above). But as is obvious from
the relevant passage in Aristotle the intuitive notion of validity is not
bound to any particular syntax – it is a “semantic” definition.27
Bolzano was aware of these three related problems. He writes:
Aristotle began with such a broad definition of the word syllogism that one
is astonished that he could have subsequently restricted the concept of this
kind of inference so severely. He writes (in Anal Pr. I, 1) “syllogism is a
discourse in which, certain things being stated, something other than what is
stated follows of necessity from their being so”. This definition obviously fits
every inference, not only with two, but also with three and more premises,
and not only simple inferences but complex ones as well. (1837, §262: 535)
As Bolzano sees it, one need not suppose that the number of (valid) forms
of inferences is finite or that it is linked to a determinate syntax, for
instance that it can only be defined for inferences that have only two
categorical premises.28
Moreover, the three above problems concerning traditional syllogistic
treatments of validity are linked to a fourth more general one. There are
various ways of fixing the extension of a concept, not all of which amount
to definition. The mere fact of knowing which inferential forms satisfy the
27
This is even more obvious when one reads the beginning of the second book of the Prior
Analytics, which was devoted to the relationship between premises and conclusion as regards
their truth-value.
28
See (Příhonský 1850: 115f ). Some passages of the Prior Analytics suggest that Aristotle was aware of
the problem. See for instance (Prior Analyics I, 32). But Aristotle himself did not provide a
systematic account of what it is for an inference that is not a syllogism to result of necessity.
206 Sandra Lapointe
intuitive Aristotelian definition of validity does not, on Bolzano’s account,
amount to having a definition of this notion. On Bolzano’s account, my
merely knowing what falls in the extension of a concept – say the class of
all putatively valid syllogistic inferential forms – does not amount to my
having a definition (Erklärung) of that concept. Definition is a conceptual
exercise: one that requires us to identify the components of a concept as
well as the way in which they are connected. As Bolzano sees it, the theory
of deducibility and the proposed definition above is what allows us to grasp
the concept of validity.
More importantly perhaps – though this might go beyond Bolzano’s
criticism – it seems that a good definition of validity is one that is epistemic-
ally fruitful in the following sense: a good definition of validity is one on the
basis of which one can ascertain systematically for any newly encountered
inference, whether or not it is valid. But the traditional syllogistic definition
of validity is not epistemically fruitful. There is no obvious reason to think
that one could decide whether an as yet unknown argument form is valid
when presented with it in any other way than by reverting to the intuitive
notion. By contrast, Bolzano’s definition is epistemically fruitful: equipped
with Bolzano’s definition, one can in principle determine for any new
argument whether or not it instantiates the property in question.

5. Conclusion
In light of what precedes, Bolzano’s internal realism is vindicated: Bolzano’s
positing of propositions in themselves allows him to articulate a theory of
deducibility that could do what the syllogistic theories of his predecessors
could not: provide us with a general semantic theory of validity. Nonetheless,
as those acquainted with recent scholarship know, there are problems with
Bolzanian deducibility. (See, e.g. Siebel 2002.) For one, despite Bolzano’s
claim to the contrary, his definition of deducibility fails to capture what is
usually taken to be the modal insight that underlies the intuitive Aristotelian
notion of validity, namely the idea that the conclusion of a good argument
“results of necessity”. Consequently, it overgenerates. Bolzanian deducibility
systematically includes inferences that are merely materially valid. I say
“systematically” because if it is the case that all As are Bs, then ‘X is B’ is
invariably deducible from ‘X is A’. For instance, on Bolzano’s account:
X is no taller than three metres
is deducible from:
X is a man
 Bolzano’s logical realism 207
with respect to X. This failure may strike as the result of a misunderstand-
ing (coupled with contentious exegetical choices). Bolzano interprets the
relevant passage of the Prior Analytics in the following terms:
Since there can be no doubt that Aristotle assumed that the relation of
deducibility can also hold between false propositions, the results of necessity
can hardly be interpreted in any other way than this: that the conclusion
becomes true whenever the premises are true. Now it is obvious that we
cannot say of one and the same collection of propositions that one of them
becomes true whenever the others are true, unless we envisage some of their
parts as variable. For propositions none of whose parts change are not
sometimes true and sometimes false; they are always one or the other.
Hence when it was said of certain propositions that one of them becomes
true as soon as the others do, the actual reference was not to these propos-
itions themselves, but to a relation which holds between the infinitely many
propositions which can be generated from them, if certain of their repre-
sentations are replaced by arbitrarily chosen other representations. The
desired formulation was this: as soon as the exchange of certain representa-
tions makes the premises true, the conclusion must also become true. (1837,
§155: 129)
The main problem with Bolzano’s interpretation is that he assumes that
“results of necessity”, in this context, means the same as “preserves truth from
premises to conclusion”. Whatever the explanation for this confusion is –
Bolzano does have a systematic account of necessity and one may wonder why
he did not revert to it to interpret Aristotle on this occasion – it is unfortunate.
Nonetheless one should not conclude from the fact that Bolzano’s definition
of deducibility fails to grasp the modal insight that underlies the intuitive
notion of validity that he achieved little toward a theory of logical conse-
quence or that he missed the point entirely. This would not do justice to
Bolzano’s accomplishment, both historical and philosophical. For one thing,
while Bolzano’s own use of the substitutional method fails to do so, other
philosophers have put a wager on a substitutional procedure of the type
Bolzano was first to introduce for the purpose of providing a satisfactory
account of logical consequence. Tarskian-type model-theoretic approaches
for instance can be seen as an extension of Bolzano’s theory.
Few would deny that Bolzano’s views on deductive knowledge were
overall largely preferable to those of his predecessors and contemporaries.
In particular, it is important to stress the fact that Bolzano did have views
on epistemic modality – though unfortunately, there is no place for a
discussion of the latter here.29 At the very least, it ought to be mentioned
29
See (Lapointe 2014).
208 Sandra Lapointe
that as an alternative to Kant’s theory of pure intuition in arithmetic and
geometry, Bolzano was first to propose an account of epistemic necessity
that rests on (i) the idea that truth by virtue of meaning can be defined
systematically (in a deductive system) and that (ii) a priori knowledge is
accordingly always deductive. Regardless of the execution, (i) and (ii) are
both manifestly valuable philosophical insights that deserve the attention
of historians and philosophers alike. For one thing, one committed to (i)
and (ii) cannot appeal to subjective justificatory devices such as “certitude”
or “evidence” to warrant the truth of a priori claims. And, again, many
even today would consider this to be an important lesson. What’s relevant
here is the fact that to the extent that Bolzano’s views on a priori
knowledge and deductive systems are parts and pieces of his theory of
propositions in themselves, they are inseparable from his commitment to
mind-independent logical facts. What this means is that logical realism also
informs his views on a priori knowledge and nourishes insights that many
of his successors, realist or not, will share.
 part iii
Specific Issues
 chapter 12

Revising logic
Graham Priest

1. What’s at issue
Much ink has been spilled over the last few decades in disputes between
advocates of “classical logic” – that is, the logic invented by Frege and
Russell, and polished by Hilbert and others – and advocates of non-
classical logics – such as intuitionist and paraconsistent logics. One move
that is commonly made in such debates is that logic cannot be revised.
When the move is made, it is typically by defenders of classical logic.
Possession, for them, is ten tenths of the law.
The point of this chapter is not to enter into substantive debates about
which logic is correct – though relevant methodological issues will tran-
spire in due course. The point is to examine the question of whether logic
can be revised.1 (And let me make it clear at the start that I am talking
about deductive logic. I think that matters concerning non-deductive logic
are much the same, but that is an issue for another occasion.) Three
questions, then, will concern us:

• Can logic be revised?

• If so, can this be done rationally?
• If so, how is this done?
Unfortunately, debates about the answers to these questions are often
vitiated by a failure to observe that the word ‘logic’ is ambiguous. Only
confusion results from running the senses of the word together. Once the
appropriate disambiguations are made, some of the answers to our ques-
tions are obvious; some are not. It pays, for a start, to be clear about which
are which.

1
Thanks go to Hartry Field for many enjoyable and illuminating discussions on the matter. We taught
a course on the topic together in New York in the Fall of 2012. Many of my views were clarified in the
process.

211
212 Graham Priest
We may distinguish between at least three senses of the word, which
I will call:
• Logica docens
• Logica utens
• Logica ens
What each of these is will require further discussion and clarification. But
as a first cut, we may characterise them as follows.
Logica docens (the logic that is taught) is what logicians claim about
logic. It is what one finds in logic texts used for teaching. Logica utens (the
logic which is used) is how people actually reason. The first two phrases are
familiar from medieval logic. The third, logica ens (logic itself ) is not. (I
have had to make the phrase up.) This is what is actually valid: what really
follows from what.
Of course, there are important connections between these senses of
‘logic’, as we will see in due course. But the three are distinct, both
intensionally and extensionally, as again we will see.
I will proceed by discussing each of these senses of ‘logic’, and asking
each of our three target questions about them. We have, then a nine-part
investigation.

2. Logica docens

2.1 Can it be revised?

Let us start with logica docens. The discussion of this will form the longest
part of the essay, since it informs the discussion with respect to the other
two parts. The question of whether the logica docens can be revised is,
however, the easiest to deal with. It can be revised because it has been
revised.
The history of logic in the West has three great periods.2 The first was in
Ancient Greece, when logic was founded by Aristotle, the Megarians, and
the Stoics. The second was in the new universities of Medieval Europe,
such as Oxford and Paris, where Ockham, Scotus, and Buridan flourished.
The third starts in the late nineteenth century, with the rise of mathemat-
ical logic, and shows no signs yet of ending. Between these three periods
were periods of, at best, mainly maintaining what was known, and at worst
forgetting it. Much of Greek logic was forgotten in Europe, but fortunately
2
The history of logic in the East has its own story to tell, but that will not be our concern here.
 Revising logic 213
preserved by the great Arabic scholars such as Al Farabi and Ibn Rushd.
Most of medieval logic was simply wiped out by the rise of the Enlighten-
ment, and the consequent obliteration of Scholasticism. It is only in
the twentieth century that we have started to rediscover what was lost
in this period.
At any rate, one needs only a passing acquaintance with logic texts in the
history of Western logic to see that the logica docens was quite different in
the various periods. The differences between the contents of Aristotle’s
Analytics, Paul of Venice’s Logica Magna, the Port Royale Logic, or the
Art of Thinking, Kant’s Jäsche Logik, and Hilbert and Ackermann’s
Principle’s of Mathematial Logic would strike even the most casual observer.
It is sometimes suggested that, periods of oblivion aside, the develop-
ment of logic was cumulative. That is: something once accepted, was never
rejected. Like the corresponding view in science, this is just plain false. Let
me give a couple of examples.
One of the syllogisms that was, according to Aristotle, valid, was given
the name Darapti by the Medievals, and is as follows:

All As are Bs
All As are C s
Some Bs are C s
As anyone who has taken a first course on modern first-order logic will
know, this inference is now taken to be invalid.3
For another example: Classical logic is not paraconsistent; that is, the
following inference (Explosion) is valid for all A and B: A, :A ├ B. It is
frequently assumed that this has always been taken to be valid. It has not.
Aristotle was quite clear that, in syllogisms, contradictions may or may not
entail a conclusion. Thus, consider the syllogism:
No As are Bs
Some Bs are As
All As are As
This is not a valid syllogism, though the premises are contradictories.
There are usually three distinct terms in a syllogism. The above has only
two. But Aristotle is also quite explicit that two terms of a syllogism may be
the same.
So when did Explosion enter the history of Western logic? Matters are
conjectural, but the best bet is that it entered with the ideas of the twelfth-

3
For further discussion of the matter, see (Priest 2006a: 10.8).
214 Graham Priest
century Paris logicians called the Parvipontinians, whose members
included Adam of Balsaha and William of Soissans, who may well have
developed the argument to Explosion using extensional connectives and
the Disjunctive Syllogism. After that, the validity of Explosion was
debated. But it certainly did not become entrenched in Western logic till
the rise of classical logic.4

2.2 Can it be revised rationally?

Logica docens, then, has been revised, and not in a cumulative fashion. The
next question is whether revision can be rational.
Arguably, not all the changes in the history of logic were rational (or
perhaps better: occurred for reasons that were internal to the subject).
Thus, logic fell into oblivion in the early Middle Ages in Western Chris-
tendom because the institutions for the transmission of philosophical texts
collapsed. And later Medieval logic was written off on the coat-tails of the
rejection of Scholasticism during the Enlightenment.5
However, many changes that did arise were the result of novel ideas,
reason, argument, debate. These are the things of which rational change
are made. This should be pretty obvious with respect to the only change
that most logicians are now familiar with: the rise of mathematical logic.
In the mid nineteenth century, text book logic (“traditional logic”) was
a highly degenerate form of medieval logic: essentially, Aristotelian
syllogistic with a few medieval accretions, such as “immediate infer-
ences” like modus ponens. But this was a period in which high standards
of rigour in mathematics were developing. Mathematicians such as
Weierstrass and Dedekind were setting the theory of numbers on a
firm footing. And when it came to examining the reasoning required in
the process, notably by Frege, it became clear that traditional logic did
not seem to be up to the job. Hence Frege invented a logic that did
much better: classical logic. The extra power of this logic made it much
preferable rationally; and within 50 years it had replaced traditional
logic as the received logica docens.
I will come back to this in the next section. For the present, let us move
on to our third question.

4
For references and further discussion on all these matters, see (Priest 2007: sec. 2).
5
Actually, my knowledge of the history of these periods is pretty sketchy; but I think that these claims
are essentially correct.
 Revising logic 215

2.3 Logic as theory

So, what, exactly, is it in virtue of which one logica docens is rationally
preferable to another, and so may replace it? To answer this question, we
need to draw some new distinctions.
Let us start with geometry. There are many pure geometries: Euclidean
geometry, elliptical geometry, hyperbolic geometry, and so on. And as
pieces of pure mathematics, all are equally good. They all have axiom
systems, model theories; each specifies a perfectly fine class of mathemat-
ical structures. Rivalry between them can arise only when they are applied
in some way. Then we may dispute which is the correct geometry for a
particular application, such as mensurating the surface of the earth. Each
applied geometry becomes, in effect, a theory of the way in which the
subject of the application behaves.
Geometry had what one might call a canonical application: the
spatiotemporal structure of the physical cosmos. Indeed this application
was coeval with the rise of Euclidean geometry. It was only the rise of non-
Euclidean geometries which brought home the conceptual distinction
between a pure and an applied geometry. And nowadays the standard
scientific view is that Euclidean geometry is not the correct geometry for
the canonical application.
So much, I think, is relatively uncontestable. But exactly the same
picture holds with respect to logic. There are many pure logics: classical
logic, intuitionist logic, various paraconsistent logics, and so on. And as
pieces of pure mathematics, all are equally good. They all have systems of
proof, model theories, algebraicisations. Each is a perfectly good math-
ematical structure. But pure logics are applied for many purposes: to
simplify electrical circuits (classical propositional logic), to parse grammat-
ical structures (the Lambeck calculus), and it is only when different logics
are taken to be applied for a particular domain that the question of which
is right arises. Just as with geometries, each applied logic provides, in effect,
a theory about how the domain of application behaves.
And just as with geometries, pure logics have a canonical application:
(deductive) reasoning. A logic with its canonical application delivers an
account of ordinary reasoning. One should note that ordinary reasoning,
even in science and mathematics, is not carried out in a formal language,
but in the vernacular; no doubt the vernacular augmented by many
technical terms, but the vernacular none the less. (No one reasons à la
Principia Mathematica.) And so applied, different pure logics may give
different verdicts concerning an inference. If it is not the case that it is not
216 Graham Priest
the case that there is an infinitude of numbers, does it follow that there
is an infinitude of numbers? Classical logic says yes; intuitionist logic
says no.6
In other words, a pure logic with its canonical application is a theory of
the validity of ordinary arguments: what follows (deductively) from what.
How to frame such a theory is not at all obvious. Many approaches have
been proposed and explored. One approach is to take validity to be
constituted modally, by necessary truth-preservation (suitably under-
stood). Another is to define validity in terms of probabilistic constraints
on rational belief. Perhaps the most common approach at present is to take
a valid inference to be one which obtains in virtue of the meanings of (at
least some of ) the words employed in it. This strategy has itself two ways
in which it can be implemented. One takes these meanings to be spelled
out in terms of truth conditions, giving us a model-theoretic account of
validity; the other takes these meanings to be spelled out in inferential
terms, giving us a proof-theoretic account of validity.
It is clear that a theory of validity is no small undertaking. It requires an
account of many other notions, such as negation and quantification.
Moreover, depending on the theory in question, it will require an articula-
tion of other important notions, such as truth, meaning, probability. No
wonder it is hard to come up with plausible such theories!
At any rate, it is crucial to distinguish between logic as a theory (logic
docens, with its canonical application), and what it is a theory of (logica
ens). In the same way we must clearly distinguish between dynamics as a
theory (e.g., Newtonian dynamics) and dynamics as what this is a theory of
(e.g., the dynamics of the Earth). This is enough to dispose of the Quinean
charge (still all too frequently heard): change of logic means change of
subject.7 If one changes one’s theory of dynamics, one can still be
reasoning about the same thing: the way the Earth moves.

2.4 What is the mechanism of rational revision?

With this substantial prolegomenon over, we can now address the question
of the mechanism of rational change of logica docens. As we have seen, a
pure logic with its canonical appication is essentially a theory of validity
and its multitude of cognate notions. How do we determine which theory
is better? By the standard criteria of rational theory choice.

6 7
Further on the above, see (Priest 2006a: chs. 10, 12). (Quine 1970a: p. 81).
 Revising logic 217
Given any theory, in science, metaphysics, ethics, logic, or anything
else, we choose the theory which best meets those criteria which determine
a good theory. Principal amongst these is adequacy to the data for which
the theory is meant to account. In the present case, these are those
particular inferences that strike us as correct or incorrect. This does not
mean that a theory which is good in other respects cannot overturn
aberrant data. As is well recognised in the philosophy of science, all things
are fallible: both theory and data.
Adequacy to the data is only one criterion, however. Others that are
frequently invoked are: simplicity, non-(ad hocness), unifying power,
fruitfulness. What exactly these criteria are, and why they should be
respected, are important questions, which we do not need to go into here.
One should note, however, that whatever they are, they are not all
guaranteed to come down on the same side of the issue. Thus (the
standard story goes), Copernican and Ptolemaic astronomy were about
equal in terms of adequacy to the data; the Copernican system was simpler
(since it eschewed the equant); but the Ptolemaic system cohered with the
accepted (Aristotelian) dynamics. (The Copernican system could handle
the motion of the Earth only in an ad hoc fashion.) In the end, the theory
most rational to accept, if there is one, is the one that comes out best on
balance. How to understand this is not, of course, obvious. But we do not
need to pursue details here.8
I observe that this procedure does not prejudice the question of logical
monism vs logical pluralism. If there is “one true logic” one’s best appraisal
of what this is is determined in the way I have indicated. If there are
different logics for different topics, each of these is determined in the same
way. Whether one single logic is better than many, is a “meta-issue”, and is
itself to be determined by similar considerations of rational theory-choice.
Let me finish this discussion by returning, by way of illustration, to the
replacement of traditional logic by mathematical logic in the early years of
the twentieth century. In the nineteenth century, much new data had
turned up: specifically, the microscope had been turned on mathematical
reasoning, showing all sorts of inferences that did not fit into traditional
logic. Mathematical logic was much more adequate to this data. This is not
to say that enterprising logicians could not try to stretch traditional logic to
account for these inferences. But mathematical logic scored high on many
of the other theoretical criteria: simplicity, unifying power, and so on. It
was clearly the much better theory.
8
Matters are spelled out in detail on (Priest 2006a: ch. 8), and especially, (Priest to appear).
218 Graham Priest
A word of warning: it would be wrong to infer that classical logic did not
have its problems. It had its own ad hoc hypotheses (to deal with the
material conditional, for example). It had areas where it seemed to perform
badly (for example, in dealing with vague language). And why should one
expect a logic that arose from the analysis of mathematical reasoning to be
applicable to all areas of reasoning? It was just these things which left the
door open for the development of non-classical logics. That, however, is
also a topic for another occasion.9 We have seen, at least in outline, what
the mechanism of rational change for a logica docens is.

3. Logica utens

3.1 What is this?

So much for the discussion of logica docens. Let us now turn to the next
disambiguation. Before we address our three questions, however, there is
an important preliminary issue to be addressed. What exactly is logica
utens?
I said that it is the way that people actually reason. This may make it
sound like a matter of descriptive cognitive psychology; but it is not this,
for the simple reason that we know that people often reason invalidly. Set
aside slips due to tiredness, inebriation, or whatever. We know that people
actually reason wrongly in systematic ways.10
To take just one very well established example: the Wason Card Test.
There is a pack of cards. Each card has a letter on one side and a positive
integer on the other. Four cards are laid out on the table so that a subject
can see the following:
A K 4 3
The subject is then given the following conditional concerning the dis-
played situation: If there is an A on one side of the card, there is an even
number on the other. They are then asked which cards should be turned
over (and only those) to check this hypothesis. The correct answer is:
A and 3. But a majority of people (even those who have done a first course
in logic!) tend to give one of the wrong answers: A, or A and 4.
Exactly what is going on here has occasioned an enormous literature,
which we do not need to go into. The experiment, and ones like it, show
that people can reason wrongly systematically. Of course, people are able
9 10
Some discussion can be found in (Priest 1989). See (Wason and Johnson-Laird 1972).
 Revising logic 219
to appreciate the error of their reasoning when it is pointed out to them.
But how to draw a principled distinction between correcting a standard
performance error, and revising an actual practice is not at all obvious.
Fortunately, we do not need to go into this here. I point these facts out
only to bring home the point that logica utens is not a descriptive notion; it
is a normative one. A logica utens is constituted by the norms of an
inferential practice. Subjects in the Wason Card Test can see, when it is
pointed out to them, that they have violated appropriate norms. How to
understand the normativity involved here is a particularly hard question,
which, fortunately, we also do not need to pursue. We have sufficient
understanding to turn to the first of our three questions. Can a logica utens
be revised?

3.2 Can it be revised?

Clearly, different reasoning practices come with different sets of norms.
Thus, the norms that govern reasoning in classical mathematics are differ-
ent from those that govern reasoning in intuitionist mathematics. I was
trained as a classical mathematician, and have no difficulty in reasoning in
this way. But I have also studied intuitionist logic, and can reason (more
falteringly) in this way too. Clearly, then, it is possible to move from one
logica utens to another. I can reason like a classical mathematician on
Mondays, Wednesdays, Fridays, and like a intuitionist on Tuesdays,
Thursdays, and Saturdays. (And on Sundays flip a coin.) So practices
can be changed.
At this point one might wonder about the nature of inference sketched
in Wittgenstein’s Philosophical Investigations. According to this, correct
reasoning is simply how we feel compelled to go on after suitable training.
If such is the case, then how can one change? The answer is that we must
take the suitable training seriously. I can follow my training as a classical
logician some days, and my training as an intuitionist on others – just as
I can follow my training in cricket on some days, and my training in
baseball on others.

3.3 Can it be revised rationally?

So logica utens can change. Can it be changed rationally? Unless one is a
complete relativist about inferential practices, the answer must be yes:
some practices are better than others. And to move from one that is less
good to one that is more good for principled reasons is clearly rational.
220 Graham Priest
Moreover, being a relativist about such practices is a hard pill to
swallow. For we use reasoning to establish what is true, and what is not,
about many things. A relativism about these practices therefore entails a
relativism about truth. And such a relativism is problematic. To take an
extreme example: suppose that reasoning in one way, we establish that the
theory of evolution is correct, but that reasoning in another way, we
establish that creationism is true and the theory of evolution false. Some-
thing, surely, must be wrong with one of these forms of reasoning.11

3.4 How is it revised rationally?

Assuming, then, that rational change is possible, how is this to be done?
The answer to that is easy. We determine what the best theory of reasoning
is (the best docens), and simply bring our practice (utens) into line with
that. How else could one be rational about the matter?

4. Logica ens

4.1 Can it be revised?

We now turn to what I think is the hardest of the three disambiguations:
logica ens. These are the facts of what follows from what – or better, to
avoid any problems with talk of facts: the truths of the form ‘that so and so
follows from that such and such’. Can these be revised? The matter is
sensitive for a number of reasons.
As we have seen, our logica docens, with its canonical application, is a
theory about what claims of this form are true. Now, if one changes one’s
theory of dynamics, the dynamics of the Earth do not themselves change.
Such realism about the physical world is simply common sense. But logic
is not a natural science. It is a social science, and concerns human practices
and cognition. When a theory changes in the social sciences, the object of
the science may change as well. One has to look only at economics to see
this. When free-market economics became dominant in the capitalist
world in the 1980s, so did the way that the then deregulated economy
functioned. So, in the social sciences one is not automatically entitled to
the view that a change of theory does not entail a change of object.

11
It is quite compatible with this point that sometimes truth may be internal to a practice – for
example, within classical and intuitionist pure mathematics. See (Priest 2013).
 Revising logic 221
But the object of a social scientific theory may not change when the
theory does, for all that. (Many basic laws of psychology are, presumably,
hard-wired in us by evolution.) Whether the truth of validity-claims can
change will depend on what, exactly, constitutes validity. Let me illustrate.
Suppose that one held a “divine command” theory of validity: something is
valid just if God says so. Then, God being constant and immutable, what
is valid could not change. On the other hand, suppose that one were to
subscribe to the “dentist endorsement” view of validity: what is valid is
what 90 per cent of dentists endorse. Clearly, that can change.
These theories are, of course, rather silly. But they make the point: the
truth of validity-claims may or may not change, depending on what
validity actually is. An adequate answer to our question would therefore
require us to settle the issue of what validity is, that is, to determine the
best theory of validity. That is far too big an issue to take on here.12
I shall restrict myself in what follows to some remarks concerning the
model-theoretic and proof-theoretic accounts of validity. According to
the first, an inference is valid iff every model of the premises is a model
of the conclusion. But a model is a structured set, that is, an abstract
object, the premises form a set, another abstract object, and the premises
and conclusions themselves are normally taken to be sentence types, also
abstract objects. According to the second, an inference is valid if there is a
proof structure (sequence or tree), at every point of which there is a
sentence related to the others in certain ways. But a proof structure is an
abstract object, as, again, are the sentences.
In other words, validity, on these accounts, is a realtionship between
abstract objects. As usual, we may take these all to be sets. If this is so,
then, at least if one is a standard platonist about these things, the truth of
claims about validity cannot change.13 Claims about mathematical objects
are not significantly tensed: if ever true true, always true.

4.2 Can meanings change?

That is not an end of the matter, though. The propositions about validity
may not change their truth values. But we express these in language. It
might be held that the words involved may change their meanings – and,
moreover, do this in such a way that the truth values of the sentences

12
I have said what I think about the matter in (Priest 2006a: ch. 11).
13
Certain kinds of constructivists may, of course, hold that the truth about numbers and other
mathematical entities may change – for example, as the result of our acquiring new proofs.
222 Graham Priest
involved may change. If this is the case, then the sentences expressing
validity claims can change their truth values.
Can meanings change in such a way as to affect truth value? Of course
they can. When Nietzsche wrote The Gay Science, it was a reference to the
art of being a troubadour. Nowadays, one could hear it only as concerning
a study of a certain sexual preference. In modern parlance, Nietzsche did
not write a book about (the) gay science.
Now, could there be such change of meaning in the case we are
concerned with? Arguably, yes. In both a proof-theoretic and a model-
theoretic account of validity, part of the machinery is taken as giving an
account of meanings – notably, of the logical connectives (introduction or
elimination rules, truth conditions). If we change our theory, then our
understanding of these meanings will change. This does not mean that the
meanings of the vernacular words corresponding to their formal counter-
parts changes. You can change your view about the meaning of a word,
without the word changing its meaning. However, if one revises one’s
theory, and then brings one’s practice into line with it, in the way which
we noted may happen, then the usage of the relevant words is liable to
change. So, then, will their meanings – assuming that meaning supervenes
on use (and some version of this view must surely be right). So the
sentences used to express the validity claims, and maybe even which
propositions the language is able to express, can change.14
It might be thought that this makes such a change a somewhat trivial
matter. Suppose we have some logical constant, c, which has different truth
or proof conditions according to two different theories. Can we not just
use two words, c1 and c2, which correspond to these two different senses?
Perhaps we can sometimes; but certainly not always: for meanings can
interact. Let me illustrate. Suppose that our logic is intuitionist. Then
“Peirce’s law”, ((A ! B) ! A) ! A, is not logically valid. But suppose that
we now decide to add a new negation sign to the language, which behaves
as does classical negation. Then Peirce’s law becomes provable. The
extension is not conservative. Another case: given many relevant logics,
the rules for classical negation can be added conservatively, as can the
natural introduction and elimination rules for a truth predicate. But the

14
A pertinent question at this point is whether the meaning of ‘follows deductively from’ – or however
this is expressed – can itself change. Perhaps it can; and if it does, this adds a whole new dimension
of complexity to our investigation. However, I see no evidence that the meaning of the phrase (as
opposed to our theories of what follows from what) has changed over the course of Western
philosophy. So I ignore this extra complexity here.
 Revising logic 223
addition of both (when appropriate self-reference is available) produces
triviality. Meanings, then, are not always “separable”.15

4.3 Can meanings change rationally?

So meanings can change, and not necessarily in a straightforward way. Can
this happen rationally, and if so, how? The answers to these questions are
implicit in the preceding discussion. Suppose we change our logica docens
to a rationally preferable one. Suppose that we then change our logica utens
rationally to bring it in line with this. Then the meanings of our logical
constants, and so the language used to express the facts of validity, may
also change. And the whole process is rational.

5. Conclusion
Let me end by summarising the main conclusions we have reached, and
making a final observation.
A logica docens may be revised rationally, and this happens by the
standard mechanism of rational theory choice. A logica utens may be
changed by bringing it into line with a logica docens; and if the docens is
chosen rationally, so is the utens. The answer to the question of whether or
not the logica ens may change depends on one’s best answer to the question
of what validity is. However, under the model- or proof-theoretic accounts
of validity, the answer appears to be: no. This does not mean, however,
that the sentences used to express these facts may not change. And a
rational change of logica utens may occasion such a change.
Now the observation. The rational logica utens depends on the rational
logica docens. The true logica docens depends on the facts of validity. And
assuming a model- or proof-theoretic account of meaning, the language
available to express these may depend on the logica utens. It is clear that we
have a circle. If one were a foundationalist of some kind, one might see this
circle as vicious: there is no privileged point where one can ground the
entire enterprise, and from which one can build up everything else.
However, I take it that all knowledge, about logic, as much as anything
else, is situated.16 We are not, and could never be, tabulae rasae. We can
start only from where we are. Rational revision of all kinds then has to
proceed by an incremental and possibly (Hegel notwithstanding) never-
ending process.
15 16
On these matters, see (Priest 2006a: ch. 5). See (Priest to appear).
 chapter 13

Glutty theories and the logic of antinomies

Jc Beall, Michael Hughes, and Ross Vandegrift

1. Introduction
There are a variety of reasons why we would want a paraconsistent
account of logic, that is, an account of logic where an inconsistent
theory does not have every sentence as a consequence. One relatively
standard motivation is epistemic in nature.1 There is a high probability
that we will come to hold inconsistent beliefs or inconsistent theories
and we would like some account of how to reason from an inconsistent
theory without everything crashing. Another motivation, rooted in the
philosophy of logic or language, is that we want a proper account of
entailment or relevant implication, where there is a natural sense in
which inconsistent claims do not (relevantly) entail arbitrary propos-
itions – where not every claim follows from arbitrary inconsistency.2
A third motivation, the one which will occupy our attention here, is
metaphysical or semantic. One might, for various reasons, endorse that
there are ‘true contradictions’, or as they are sometimes called, truth-
value gluts – true sentences of the form φ ^ : φ, claims which are both
true and false. We shall say that a glut theorist is one who endorses
glutty theories – theories that are negation-inconsistent – with the full
knowledge that they are glutty.
There are different kinds of metaphysical commitments that can lead
one to be a glut theorist. One route towards glut theory arises from
views about particular predicates of a language or the properties that
those predicates express. Along these lines, a familiar route towards glut
theory holds that certain predicates like ‘is true’, ‘is a member of ’, or
‘exemplifies’ are essentially inconsistent: they cannot be (properly)

1
For work in this tradition, see Rescher and Manor (1970); Schotch et al. (2009); Schotch and
Jennings (1980).
2
For work in this tradition, see Anderson and Belnap (1975); Anderson et al. (1992); Dunn and Restall
(2002); Mares (2004); Slaney (2004).

224
 Glutty theories and the logic of antinomies 225
interpreted in a way that avoids there being objects of which these
predicates are both true and false. Such essentially glutty predicates –
everywhere glutty with respect to something if glutty anywhere with
respect to anything – are antinomic, as we shall say. Of course, one need
not hold that a predicate is essentially inconsistent to think that it can
give rise to gluts: there may be only contingently glutty predicates. For
some predicates, whether they are properly interpreted consistently
or inconsistently may depend on facts about the world. Priest, for
example, has suggested that predicates like ‘is legal’ and ‘has the right
to vote’ are of this sort; see Priest (2006b). Acceptance of either sort of
(essentially or contingently) inconsistent predicates is sufficient for
being a glut theorist – though not necessary.
Another path one might take towards being a glut theorist is inevitable
ignorance about the exact source of gluttiness. One might think that our
best – and true – theory of the world will inevitably be inconsistent, even
though we might, for all we know, remain ignorant of the source of the
inevitable inconsistency. Indeed, one might have reason to be agnostic
about the source of gluttiness: one is convinced that our best theory of the
world (including truth, exemplification, sets, computability, modality,
whatever) will be inconsistent, though also convinced that we will never
be in good position to pinpoint the exact source of the inconsistency.
Agnosticism about the particular predicates responsible for gluttiness
remains an option for the glut theorist.
The question that arises is: how do our metaphysical commitments
inform our choice of logic? We cannot ask this question without attending
to the difference between formal and material consequence. Briefly, a logic
takes a material approach to consequence when it builds in facts about the
meaning of predicates, the properties they express, or the objects those
predicates are about. A logic takes a formal approach to consequence when
it abstracts away from all of these concerns. There are various ways a logic
could be said to ‘build in’ such facts, and one of our aims below is to
explore these in the context of metaphysical commitments to gluts. We
carry out our discussion via a comparison of two paraconsistent logics,
namely, the logic of paradox (LP) and the logic of antinomies (LA). The
former is well-known in philosophy, discussed explicitly and widely by
Priest (1979, 2006b);3 the latter is a closely related but far less familiar and

3
LP is the gap-free extension of FDE, the logic of tautological entailments; it is the dual of the familiar
glut-free extension of FDE called ‘strong Kleene’ or ‘K3’. See Dunn (1966, 1976), Anderson and
Belnap (1975), and Anderson et al. (1992).
226 Jc Beall, Michael Hughes, and Ross Vandegrift
equally less explored approach in philosophy, an approach advanced by
Asenjo and Tamburino (1975).4
Below, we consider various philosophical motivations that could explain
the logical differences between LA and LP. We shall argue that LA reflects
a fairly distinctive set of metaphysical and philosophical commitments,
whereas LP, like any formal logic, is compatible with a broad set of
metaphysical and philosophical commitments. We illustrate these
points below.
The discussion is structured as follows. §§2–3 present the target logics in
terms of familiar model theory. §4 discusses the main logical differences in
terms of differences in philosophical focus and metaphysical commitment.
§5 closes by discussing the issue of detachment.

2. The logic of antinomies

The logic of antinomies (LA) begins with a standard first-order syntax. The
logical vocabulary is _, :, 8. Constants c0, c1, . . . and variables x0, x1, . . .
are the only terms. The set ℙ of predicate symbols is the union of
two disjoint sets of standard predicate symbols:  = fA0, A1, . . .g and
 = fB0, B1, . . .g. (Intuitively,  contains the essentially classical predi-
cates and  the essentially non-classical, essentially glutty predicates.) The
standard recursive treatment defines the set of sentences.5
An LA interpretation I consists of a non-empty domain D, a denotation
function d, and a variable assignment v, such that:

• for any constant c, d(c) 2 D,

• for any variable x, v(x) 2 D,
• for any predicate P, d(P) = 〈Pþ, P〉, where Pþ [ P = D.
The only difference from the standard LP treatment appears here, in the
form of a restriction that captures the distinction between the antinomic
(i.e., essentially glutty) and essentially classical predicates:

4
For purposes of accommodating glutty theories, the propositional logic LP was first advanced in
Asenjo (1966) under the name calculus of antinomies; it was later advanced, for the same purpose,
under the name ‘logic of paradox’ by Priest (1979), who also gave the first-order logic under the same
name (viz., LP). What we are calling ‘LA’ is the first-order (conditional-free) logic advanced by
Asenjo and Tamburino (1975), which was intended by them to be a first-order extension of Asenjo’s
basic propositional logic. Due to what we call the LA Predicate Restriction (see page 227) LA isn’t a
simple first-order extension of Asenjo’s propositional LP – as will be apparent below (see §4).
5
For simplicity, we focus entirely on unary predicates. Both LA and LP cover predicates of any arity,
but focusing only on the unary case suffices for our purposes.
 Glutty theories and the logic of antinomies 227
LA Predicate Restriction. For any predicate P:
þ 
• if P is in , then the intersection Pþ \ P must be empty;
• if P is in , then the intersection P \ P must be non-empty.
As above, the Ais are the essentially classical predicates, while the Bis are
those which are antinomic.6
|φ|v is the semantic value of a sentence φ with respect to a variable
assignment v, which is defined in the standard recursive fashion. (We leave
the relevant interpretation implicit, as it will always be obvious.) For atomics:
8
> = P þ and I ðtÞ 2 P 
0 if I ðtÞ 2
< þ
jPtjv ¼ 1 if I ðtÞ 2 P and I ðtÞ 2 = P
>
: 1
otherwise:
2
The inductive clauses are as follows:
1. |φ _ ψ|v = maxf|φ|v |ψ|vg.
2. |:φ|v = 1  |φ|v.
3. |∀xφ|v = minf|φ|v0 : v0 is an x-variant of vg.
Conjunction and existential quantification can be defined from these in
the normal way.
LA consequence ‘LA is defined as preservation of designated value,
where the designated values are 1 and 21 . Thus, Γ ‘LA φ holds (i.e., Γ
implies/entails φ according to LA) if and only if no LA interpretation
designates everything in Γ and fails to designate φ.

3. The logic of paradox

We obtain the logic LP simply by dropping the LA predicate restriction,
but leaving all else the same. Thus, for purposes of ‘semantics’ or model
theory of LP, there’s no difference between -predicates and -predicates:
they’re all treated the same.

4. Contrast: LA and LP
We begin with formal contrast. While both logics are paraconsistent (just
let jφjv ¼ 21 and jψjv = 0, for at least some formulae φ and ψ), there are
some obvious but noteworthy formal differences between the logics LA
6
The presentation in Asenjo and Tamburino (1975) is rather different; but we present their account in
a way that affords clear comparison with LP.
228 Jc Beall, Michael Hughes, and Ross Vandegrift
and LP. LP permits the existence of a maximally paradoxical object – an
object of which every predicate is both true and false – whereas LA does
not. Indeed, LA – but not LP – validates ‘explosion’ for certain
contradictions; for example, for any Ai in  and any φ,
Ai t ^ :Ai t ‘LA φ:
Similarly, LA validates the parallel instances of detachment (modus
ponens):
Ai t, Ai t ⊃ φ ‘LA φ
where φ ⊃ ψ is defined as usual as :φ _ ψ. But LP is different: not even a
restricted version of detachment is available; see Beall et al. (2013a).
As a final and nicely illustrative example, LA validates some existential
claims that go beyond those involved in classical logic (e.g., 9x(φ _ :φ),
etc.), whereas LP does not. To see this, note that, for any Bi in , the
following is a theorem of LA:
9xB i x:
Since any predicate Bi in  must have at least some object in the intersec-
tion of its extension and anti-extension, it follows that something is in its
extension.

4.1 Metaphysics, formal and material consequence

How are we to understand the logical differences between LA and LP? For
a first pass, they might be naturally understood as arising from different
notions of consequence: namely, material and formal consequence. The
distinction may not be perfectly precise, but it is familiar enough.7
Material consequence relies on the ‘matter’ or ‘content’ of claims, while
formal consequence abstracts away from such content. Example: there is
no possibility in which ‘Max is a cat’ is true but ‘Max is an animal’ is not
true; the former entails the latter if we hold the meaning – the matter, the
content – of the actual claims fixed. But the given entailment fails if we
abstract away from matter (content), and concentrate just on the standard
first-order form: Cm does not entail Am.
The notion of formal consequence delivers conclusions based on logical
form alone. Material consequence essentially requires use of the content of
the claims or the meaning of things like predicates that appear in them.
7
See Read (1994, Ch. 2) wherein Read provides a defense of material consequence as logical
consequence, and also for further references.
 Glutty theories and the logic of antinomies 229
One way to understand Asenjo and Tamburino’s proposal is that it gives
a material consequence relation of a language arising from certain meta-
physical commitments. It is clear that the logic reflects an assumption that
certain predicates are essentially classical, and other predicates are essen-
tially glutty – antinomic, as we have said. On this interpretation, the
incorporation of essentially classical predicates reflects a metaphysical
commitment that gluts cannot arise absolutely anywhere. Similarly, the
semantic restriction on the Bi predicates reflects a metaphysical commit-
ment that certain predicates, in virtue of their meaning, or the properties
they express, must give rise to gluts: there is bound to be at least some
object of which Bi is both true and false. Beall (2009) gives such a view:
inconsistency unavoidably arises in the presence of semantic predicates like
‘is true’. The typical semantic paradoxes like the liar require an inconsistent
interpretation of the truth predicate, but this is compatible with the
commitment to the essential classicality of all predicates in the truth-free
fragment of the language.
But what if you wanted to give the formal consequence relation of a
language that is motivated by Asenjo and Tamburino’s metaphysical
commitments? LP, we suggest, provides the formal consequence relation
of such a language – abstracting from the matter or content to mere form.
LA’s predicate restriction is not a purely formal matter: that a predicate is
either antinomic or essentially classical depends on its meaning. If we
ignore content, and focus just on purely formal features of sentences,
LA’s predicate restriction falls away as unmotivated. And that’s precisely
what happens in LP: if we abstract away to ‘pure form’ then the content of
predicates doesn’t matter.
We observe that one might be concerned with material consequence,
and yet still be motivated to adopt LP rather than LA. Suppose that all
predicates are on par with respect to (in-) consistency: each might be glutty
with respect to something – or not glutty at all. If one held this commit-
ment, then LA’s predicate restriction is inappropriate, or at least unmoti-
vated. Indeed, even one predicate which is either contingently consistent
or contingently inconsistent arrests the motivation for LA’s predicate
restriction. And one might think that ordinary cases of such predicates
are not hard to find. Priest (2006b, Ch. 13), for example, discusses such
cases arising from considerations of the law. Suppose that we had laws that
all citizens have a right to vote and no felons have a right to vote. It is then a
contingent matter whether or not there are any gluts about rights to vote;
it depends on whether anyone commits any felonies, and whether or not
anything is classified as a felony. And you might hold a view wherein all
230 Jc Beall, Michael Hughes, and Ross Vandegrift
predicates are like that: potentially glutty, one and all, but none anti-
nomic – none essentially glutty.
There is another metaphysical route to LP. We might not start with any
commitments about the nature of any predicates, their meaning, the
properties they express, and whether or not they are essentially inconsist-
ent. One might start with the commitment that one’s theory is both true
and inconsistent, while remaining agnostic about where to locate the
origins of the inconsistency. There is no reason to think that this position
excludes a material approach to consequence. It’s just that such a view
lacks any particular metaphysical commitments that would motivate a
restriction on predicates like the LA predicate restriction.
Of course, from the material point of view, LA and LP far from exhaust
the possibilities. So far we’ve mentioned fairly strong, all-or-nothing
approaches. On a material approach to consequence, the proponent of
LA is committed to all predicates being essentially classical or glutty, while
the proponent of LP is committed to all predicates being potentially
classical or glutty. Mixed approaches are available. These are achieved by
adding obvious combinations to the LA predicate restriction – for
example, some antinomic, some essentially classical, some neither, etc.
We leave these to the reader for exploration.
We turn (briefly) to an issue peculiar to the logics under discussion:
detachment or modus ponens.

5. Detachment
A salient problem for LP is that there is no detachable (no modus-ponens-
satisfying) conditional definable in the logic (Beall et al. (2013)); and thus,
historically, LP has been viewed as unacceptably weak for just that reason.
A lesson one might try to draw from the above observations is that LP can
be improved by shifting focus to the material notion of consequence. But
this is not quite right. Though one fragment of LA differs from LP in that
it satisfies detachment, LA is like LP in that detachment doesn’t hold
generally: arguments from φ and φ ⊃ ψ to ψ have counterexamples.
On this score, Asenjo and Tamburino (1975), along with Priest (1979,
2006b), have a solution in mind. The remedy is to add logical resources to
the base framework to overcome such non-detachment.8 But the remedy

8
Until very recently, Beall (2013), all LP-based glut theorists focused their efforts on the given task:
adding logical resources to the base LP framework to overcome its non-detachment. Whether this is
the appropriate response to the non-detachment of LP is something we leave open here.
 Glutty theories and the logic of antinomies 231
offered by Asenjo and Tamburino doesn’t work, as we now briefly
indicate.
Asenjo and Tamburino define a conditional ! that detaches (i.e., φ and
φ ! ψ jointly imply ψ). The conditional is intended to serve the ultimate
purpose of the logic, namely, to accommodate paradoxes in non-trivial
theories (e.g., theories of naïve sets), and is defined thus:
8 n1 o
>
> if jψjv ¼ 0 and jφjv 2 ,1
>
<
0
n2 o
jφ ! ψjv ¼ 1 1
if jψjv ¼ and jφjv 2
1
,1
>
>
>
:2 2 2
1 otherwise
The resulting logic, which we call LA!, enjoys a detachable conditional.
In particular, defining ‘LA! as above (no interpretation designates the
premise set without designating the conclusion), we have:
φ, φ ! ψ ‘LA! ψ:
The trouble, however, comes from Curry’s paradox. Focusing on the set-
theoretic version (though the truth-theoretic version is the same), Meyer
et al. (1979) showed that, assuming standard structural rules (which are in
place in LP and LA! and many other logics under discussion), if a
conditional detaches and also satisfies ‘absorption’ in the form
φ ! ðφ ! ψÞ ‘ φ ! ψ
then the given conditional is not suitable for underwriting naïve founda-
tional principles. In particular, in the set-theory case, consider the set
c ¼ fx : x 2 x ! ⊥g
which is supposed to be allowed in the Asenjo and Tamburino (and
virtually all other) paraconsistent set theories.9 By unrestricted comprehen-
sion (using the new conditional, which is brought in for just that job),
where $ is defined from ! and ^ as per usual, we have
c 2 c $ ðc 2 c ! ⊥Þ:
But, now, since the Asenjo–Tamburino arrow satisfies the given absorp-
tion rule, we quickly get
c2c!⊥

9
Throughout, ⊥ is ‘explosive’ (i.e., implies all sentences).
232 Jc Beall, Michael Hughes, and Ross Vandegrift
which, by unrestricted comprehension, is sufficient for c’s being in c,
and so
c 2 c:
But the Asenjo–Tamburino arrow detaches: we get ⊥, utter absurdity.
The upshot is that while LA may well be sufficient for standard first-
order connectives, the ‘remedy’ for non-detachment (viz., moving to LA!)
is not viable: it leads to absurdity.10 Other LP-based theorists, notably
Priest (1980) and subsequently Beall (2009), have responded to the non-
detachability of LP by invoking ‘intensional’ or ‘worlds’ or otherwise ‘non-
value-functional’ approaches to suitable (detachable) conditionals. We
leave the fate of these approaches for future debate.11

6. Closing remarks
Philosophy, over the last decade, has seen increasing interest in paracon-
sistent approaches to familiar paradox. One of the most popular
approaches is also one of the best known: namely, the LP-based approach
championed by Priest. Our aim in this chapter has been twofold: namely,
to highlight an important predecessor of LP, namely, the LA-based
approach championed first by Asenjo and Tamburino, and to highlight
the salient differences in the logics. We’ve argued that the differences in
logic reflect a difference in both background philosophy of logic and
background metaphysics. LA is motivated by a material approach to logical
consequence combined with a metaphysical position involving antinomic
predicates, while LP is compatible with both a formal and material
approach to consequence and can be combined with a large host of
metaphysical commitments (including few such commitments at all).12

10
We note that Asenjo himself noticed this, though he left the above details implicit. We have not
belabored the details here, but it is important to have the problem explicitly sketched.
11
We note, however, that Beall has recently rejected the program of finding detachable conditionals
for LP, and instead defends the viability of a fully non-detachable approach (Beall (2013)), but we
leave this for other discussion.
12
We note that Priest’s ultimate rejection of LP in favor of his non-monotonic LPm (elsewhere called
‘MiLP’) reflects a move ‘back’ in the direction of the original Asenjo–Tamburino approach, where
one has ‘restricted detachment’ and the like, though the latter logic (viz., LA) is monotonic. We
leave further comparison for future debate. For some background discussion, see Priest (2006,
Ch. 16) and Beall (2012) for discussion.
 chapter 14

The metaphysical interpretation of logical truth

Tuomas E. Tahko

1. Two senses of logical truth

The notion of logical truth has a wide variety of different uses, hence it is not
surprising that it can be interpreted in different ways. In this chapter I will
focus on one of them – what I call the metaphysical interpretation. A more
precise formulation of this interpretation will be put forward in what follows,
but I wish to say something about my motivation first. Part of my interest
concerns the origin or ground of logic and logical truth, i.e., whether logic is
grounded in how the world is or how we (or our minds) see the world.1
However, this is not my topic here. Rather, I will assume that logic is grounded
in how the world is – a type of realism about logic – and examine the status of
logical truth from the point of view of logical realism. The upshot is an
interpretation of logical truth that is of special interest to metaphysicians.2
My starting point is the apparent difference between what we might call
absolute truth and truth in a model, following Davidson (1973). The notion
of absolute truth is familiar from Tarski’s T-schema: ‘Snow is white’ is true
if and only if snow is white – in the world and absolutely. Instead of being
a property of sentences as absolute truth appears to be, truth in a model,
that is relative truth, is evaluated in terms of the relation between sentences
and models.3 Davidson suggested that philosophy of language should
be interested in absolute truth exactly because relative truth does not yield
T-schemas, but I am not concerned with this proposal here.4
1
For a recent discussion on this topic, see Sher (2011), who examines the idea that logic is grounded
either in the mind or in the world, and defends that it is grounded in both – hence logic has a dual
nature. See also the opening chapter of this volume.
2
See Chateaubriand Filho (2001, 2005) for a version of the metaphysical interpretation of logical truth
partly similar to mine.
3
‘Models’ are to be interpreted in a wide sense: they may for instance be interpretations, possible
worlds, or valuations. We will return to this ambiguity concerning ‘model’ below.
4
I should mention that I will omit discussion of Carnap and Quine on logical truth, as their debate is
not directly relevant for my purposes. However, see Shapiro (2000) for an interesting discussion of
Quine on logical truth.

233
234 Tuomas E. Tahko
To clarify, relative truth is an understanding of logical truth in terms of
truth in all models. One can be a realist or an anti-realist about the models,
hence about logical truth. But there are choices to be made even if one is
realist about the models, as the models can be understood interpretationally
or representationally, along the lines suggested by John Etchemendy (1990).
We will discuss the difference between these views in the next section, but
ultimately none of these alternatives are expressive of the metaphysical
interpretation of logical truth. Instead, we need a way to express absolute
truth, which is not possible without spelling out the correspondence intu-
ition, to be discussed in a moment.
Given the topic of this chapter, one might expect that Michael
Dummett’s view would be discussed, or at least used as a foil, but
I prefer not to dwell on Dummett. The primary reason for this is that
Dummett’s methodology is entirely opposite to the one that I use. Here is
a summary of Dummett’s method:

My contention is that all these metaphysical issues [questions about truth,

time etc.] turn on questions about the correct meaning-theory for our
language. We must not try to resolve the metaphysical questions first, and
then construct a meaning-theory in the light of the answers. We should
investigate how our language actually functions, and how we can construct
a workable systematic description of how it functions; the answers to those
questions will then determine the answers to the metaphysical ones.
(Dummett 1991a: 338)

Since I am analyzing logical truth from a realist, metaphysical point of

view, Dummett’s methodology is obviously not going to do the trick. In
my view, there is a bona fide discipline of metaphysics and I am interested
in finding a use for logical truth within that discipline. I doubt there is
enough initial common ground to fruitfully engage with Dummett.
Let me briefly return to Davidson and Tarski before proceeding. When
considering the distinction between absolute and relative truth, an initial
point of interest is absolute truth’s characterization by the T-schema. One
question that emerges is the connection between the T-schema and
metaphysics. A likely approach is to explicate this connection in terms of
correspondence. However, at least according to one reading, Tarski (1944)
considered truth understood as a semantic concept to be independent of
any considerations regarding what sentences actually describe, that is,
independent of issues concerning correspondence with the world. Indeed,
the T-schema is now rarely considered to play a crucial role in correspond-
ence theories of truth, despite the appearance of a correspondence relation
 The metaphysical interpretation of logical truth 235
between sentences and the world.5 Yet, Tarski’s (1944: 342–343) initial
considerations on the meaning of the term ‘true’ explicitly take into
account an ‘Aristotelian’ conception of truth, where correspondence with
the world is central. Davidson (1973: 70) as well seems to have some
sympathy for the idea that an absolute theory of truth is, in some sense,
a ‘correspondence theory’ of truth, although he insists that the entities that
would act as truthmakers here are ‘nothing like facts or states of affairs’, but
sequences (which make true open sentences).
I will not aim to settle the status of the correspondence theory here, but
it will be necessary to discuss it in some more detail. I suggest adopting an
understanding of the correspondence relation which is neutral in terms of
our theory of truth. It is this type of weak correspondence intuition that
I believe central to the metaphysical interpretation of logical truth. But it
should be stressed that the correspondence intuition itself is not necessarily
expressive of realism (Daly 2005: 96–97). For instance, Chris Daly’s
suggested definition of the intuition is simply that a proposition is true if
and only if things are as the proposition says they are. Daly explains the
neutrality of (his version of ) the correspondence intuition as follows:6
Consider the coherence theorist. He may consistently say ‘If <p> is true, it
has a truthmaker. <p> corresponds to a state of affairs, namely the state of
affairs which consists of a relation of coherence holding between <p> and
the other members of a maximal set of propositions’. Consider the pragma-
tist. He may consistently say, ‘If <p> is true, it has a truthmaker. <p>
corresponds to a state of affairs, namely the state of affairs of <p>’s having
the property of being useful to believe’. It is controversial whether there
exist states of affairs. Let that pass. My point here is that the coherence
theory and the pragmatic theory are each compatible with the admission of
states of affairs. Furthermore, each of these theories is compatible with the
admission of states of affairs standing in a correspondence relation to truths.
(Daly 2005: 97)

A neutral version of the correspondence intuition is desirable because I do

not want to rule out the possibility of different approaches to truth, despite
assuming realism in the present context. A central appeal of the corres-
pondence intuition is, I suggest, its wide applicability. However, a slightly

5
Furthermore, the idea that the T-schema or the correspondence theory are somehow expressive of
realism has been forcefully disputed. See for instance Morris (2005) for a case against the connection
between realism and correspondence; in fact Morris argues that correspondence theorists should be
idealists. See also Gómez-Torrente (2009) for a discussion about Tarski’s ideas on logical
consequence as well as on Etchemendy’s critique of Tarski’s model-theoretic account.
6
The angled brackets describe a proposition, following Horwich (1998).
236 Tuomas E. Tahko
better formulation than Daly’s can be found by following Paolo Crivelli
(2004), who interprets Aristotle as an early proponent of the correspond-
ence theory. Crivelli defines correspondence-as-isomorphism as follows: ‘a
theory of truth is a correspondence theory of truth just in case it takes the
truth of a belief, or assertion, to consist in its being isomorphic with reality’
(Crivelli 2004: 23).7 This type of view, which Crivelli ascribes to Aristotle,
is expressive of the correspondence intuition, but avoids mention of
propositions, or indeed states of affairs.8 Hence, we may define the
correspondence intuition as follows:
(CI) A belief, or an assertion, is true if and only if its content is isomorphic
with reality.
This formulation preserves Daly’s idea. ‘Reality’ in CI may consist, say, of
what it is useful to believe, as the pragmatist would have it, so neutrality is
preserved. If we accept that CI is neutral in terms of different theories of
truth, then we can characterize the issue at hand as follows. There is an
apparent and important difference between truth understood along the
lines of CI, and truth understood as a relation between sentences and
models. I take this to be at the core of Davidson’s original puzzle concern-
ing absolute and relative truth. We ought to inquire into these two senses
of truth before we give a full account of logical truth. This is exactly what
I propose to do, arguing that the metaphysical interpretation of logical
truth must respect CI.
Tarski and the model-theoretic approach may have made it possible
to talk about logical truth in a manner seemingly independent of
metaphysical considerations, but important questions about the meta-
physical status of logical truth and the interpretation of models remain.
One thing that makes this problem topical is the recent interest in
logical pluralism, or pluralism about logical truth (e.g., Beall and Restall
2006). In the second section I will assess the metaphysical status of the
notion of logical truth with regard to the two senses of truth familiar
from Davidson. The third section takes up the issue of interpreting
logical truth in terms of possible worlds and contains a case study of the
7
Crivelli also defines a stricter sense of correspondence, which can be found in Aristotle. But
sometimes Aristotle’s view on truth is also considered as a precursor to deflationism about truth,
so we shouldn’t put too much weight on the historical case. For a more historically inclined
discussion, see Paul Thom’s chapter in this volume.
8
Admittedly, once we explicate isomorphism, reference to propositions, states of affairs or something
of the sort could easily re-emerge. This shouldn’t worry us too much, because it is likely that we want
a structured mapping from something to reality. The reason to opt for isomorphism here is merely to
keep the door open for one’s preferred (structured) ontology.
 The metaphysical interpretation of logical truth 237
law of non-contradiction. A brief discussion of logical pluralism will
take place in the fourth section, before the concluding remarks.

2. Reconciling the two senses of truth

Can we reconcile the two senses of truth familiar from Davidson, the
absolute and the relative? As Etchemendy (1990: 13) notes, the obvious way
to attempt this would be in terms of generalization: if absolute truth is a
monadic predicate of the form ‘x is true’, then it may be helpful to analyze
it in terms of a relational predicate of the form ‘x is true in y’, for instance
‘x is a brother’ could be analyzed by first analyzing ‘x is a brother of y’, thus
using the generalized concept of brotherhood. However, this does not
apply to truth: ‘[C]learly the monadic concept of truth, the concept we
ordinarily employ, is no generalization of any of the various relational
concepts. A sentence can be true in some model, yet not be true; a
sentence can be true, yet not be true in all models’ (1990: 14). Accordingly,
generalization will not help in reconciling the two senses of truth.
Another alternative that Etchemendy considers is to interpret absolute
truth as a specification of truth in a model, namely, absolute truth could be
considered equivalent to truth in the right model, the model that corres-
ponds with the world. This maintains the correspondence intuition
expressed by CI above, but note that ‘correspondence with the world’
already suggests a realist theory of truth, so the neutrality of the formula-
tion is in question.9
However, there are good reasons to think that the notion of ‘model’ is
not entirely appropriate when discussing absolute truth, as it is closely
associated with relative truth. Hence, interesting as Etchemendy’s charac-
terization may be, it is unlikely to result in a metaphysical account of logical
truth. Still, Etchemendy’s account may help pinpoint the issue; consider
the following passage:
Once we have specified the class of models, our definition of truth in a
model is guided by straightforward semantic intuitions, intuitions about the
influence of the world on truth values of sentences in our language. Our
criterion here is simple: a sentence is to be true in a model if and only if it
would have been true had the model been accurate – that is, had the world
actually been as depicted by that model. (Etchemendy 1990: 24)

9
Note that the question concerning which model is ‘right’ is not, strictly speaking, a question for the
logician. For instance, as Burgess (1990: 82) notes, it is the metaphysician’s task to determine the
correct modal logic, as this depends on our understanding of (metaphysical) modality. In contrast,
the question about the ‘right’ sense of logical validity remains in the realm of logic.
238 Tuomas E. Tahko
There is an important requirement in the passage above, namely, it must
be the case that the model could have been true. How do we interpret the
modality in effect here? If we understand it as saying that it must be the
case that the world could have turned out to be like the model depicts,
then this supports the case for a metaphysical interpretation of logical
truth, for it introduces as a requirement for the notion of ‘model’ that it is
a possible representation of the world. This representational approach, or
‘representational semantics’ can be contrasted with ‘interpretational
semantics’, which Etchemendy discusses later on:
[I]n an interpretational semantics, our class of models is determined by the
chosen satisfaction domains; our definition of truth in a model is a simple
variant of satisfaction. (Etchemendy 1990: 50)

Etchemendy claims that the Tarskian conception of model-theoretic

semantics is of the ‘interpretational’ kind, although his interpretation of
Tarski can certainly be questioned (e.g. Gómez-Torrente 1999). But I do
not wish to enter the debate about Tarski or interpretational semantics.
According to Etchemendy, in the representational approach models must
represent ‘genuinely possible configurations’ of the world, and I am
interested in the correct understanding of these possible configurations
(cf. Etchemendy 1990: 60). However, instead of developing Etchemendy’s
representational account, I will propose a pre-theoretic account of absolute
truth, which aligns nicely with Etchemendy’s analysis. The biggest compli-
cation is the interpretation of the modal content in Etchemendy’s picture;
we will need to return to this issue later (in the next section).
What I propose to draw from Etchemendy is that once we have specified
the class of ‘genuinely possible configurations’, we can define relative truth
according to Etchemendy’s suggestion. In this regard, my analysis will not
follow that of Etchemendy’s, as the case for absolute truth will come before
Etchemendy’s account. Etchemendy’s representational approach notwith-
standing, the notion of ‘model’ is not ideal for this task, as it is strongly
reminiscent of relative truth.10
Instead of ‘models’, I propose to resort to talk of ‘possible worlds’. What
I have in mind is interpreting possible worlds as metaphysical possibilities.

10
It has been suggested to me (by Penny Rush) that relative truth may be problematic because of its
underlying metaphysical commitment to relativism, rather than not being up to the job of giving a
metaphysical interpretation of logical truth at all. This may indeed be the case. I have attempted to
preserve ontological neutrality while at the same time making it clear that I am presently only
interested in putting forward a realist interpretation of logical truth. But I will set this issue aside for
now, whether or not it is possible to combine relative truth and realism.
 The metaphysical interpretation of logical truth 239
This is, of course, somewhat controversial, but as we will see, there are
reasons to think that only metaphysical modality is fitting for the task. In
any case, more needs to be said about how the space of metaphysical
possibilities is restricted. We will return to this in the next section.
We are now in the position to define a provisional sense of logical truth
which I propose to call metaphysical:
(ML) A sentence is logically true if and only if it is true in every genuinely
possible configuration of the world.
ML leaves open the criteria for a ‘genuinely possible configuration of the
world’. But it does preserve CI and it provides us – via the possible worlds
jargon – a ‘metaphysician friendly’ interface to the notion of logical truth.
It is time to see if we can actually work with that interface.

3. Genuinely possible configurations and the case

of the law of non-contradiction
The puzzle can now be expressed in the following form: What sort of
criteria can be established to evaluate whether a given possible world is a
genuinely possible configuration of the world, i.e., could have turned out
to correspond with the actual world? Let me approach the problem with a
case study. Take, arguably, one of the most fundamental laws of logic, the
law of non-contradiction (LNC). When I say that the law of non-
contradiction is true in the ‘metaphysical sense’, I mean that LNC is true
in the sense of absolute truth, i.e., it is a genuine constraint on the structure
of reality. The metaphysical formulation of LNC takes a form familiar
from Aristotle (Metaphysics 1005b19–20), although my proposed formula-
tion is somewhat weaker, defined as follows:
(LNC) The same attribute cannot at the same time belong and not belong
to the same subject in the same respect and in the same domain.
The above formulation differs from Aristotle’s only with regard to the
qualification regarding ‘the same domain’ – here the domain is the set of
genuinely possible configurations of the world. How do we know whether
LNC is true in this sense? I have previously argued (Tahko 2009) that we
do have a good case for the truth of LNC in the metaphysical sense – the
primary opponent here is Graham Priest (e.g., 2005, 2006b).11 I will not
11
See also Berto (2008) for an attempt to formulate a (metaphysical) version of LNC which even the
dialetheist must accept. Berto’s idea, to which I am sympathetic, is that LNC may be understood as
240 Tuomas E. Tahko
repeat my arguments here, but it may be noted that this is not strictly a
question for logic. For instance, Priest’s most celebrated arguments in favor
of true contradictions (in the metaphysical sense) concern the nature of
change and specifically motion, the paradoxical nature of which is sup-
posedly demonstrated by Zeno’s well-known paradoxes. Although these
paradoxes can quite easily be tackled by mathematical means, the relevant
question is whether change indeed is paraconsistent.12 The answer to this
question requires both metaphysical and empirical inquiry. I will return to
this point briefly below, but first I wish to say something about the
methodology of logical-cum-metaphysical inquiry.
In terms of ML, demonstrating the falsity of LNC would first require a
genuinely possible configuration of the world where LNC fails. That is, it
is not enough that we have a model where LNC is not true, such as
paraconsistent logic, but we would also need to have some good reasons to
think that the world could have been arranged in such a way that the
implications of the metaphysical interpretation of LNC do not follow. This
point deserves to be emphasized, for it would be much easier to show that
a paraconsistent model can be useful in modelling certain phenomena, or
interpreted in such a way that it is compatible with all the empirical data.
But what is required here is that LNC, fully interpreted in the metaphysical
sense, can be shown to fail.
Note that we may also ask whether LNC is necessary, i.e., are there any
possible worlds in which LNC does not hold – even if we did have a good
case for its truth in the actual world? In fact, this is the question we should
begin with, since if LNC is necessary, then it could not fail in the actual
world either. However, it is not clear how we could settle this question
conclusively, given that we are dealing with the metaphysical interpret-
ation of LNC. Moreover, I do think that there could (in an epistemic
sense) be possible worlds in which LNC fails, and hence I take the debate
about LNC seriously. Yet, I am uncertain about whether such a paracon-
sistent possible world is in fact a genuinely possible configuration, as I will
go on to explain.13 In any case, if a possible world in which LNC is not true

a principle regarding structured exclusion relations (between properties, states of affairs, etc.), and
the world is determinate insofar as it conforms to this principle.
12
For discussion regarding Zeno’s paradoxes, see for instance Sainsbury (2009: Ch. 1).
13
It is worth pointing out here that in my proposed construal, the distinction between absolute truth
and truth in a model is not quite so striking for dialetheists. The idea, which I owe to Francesco
Berto, is that the world cannot be a model, because it contains everything, and there’s no domain of
everything, on pain of Cantor’s paradox. The result is that something can be a logical truth in the
sense of being true in all models, without being true in the absolute sense, for the world is not a
model. My proposed treatment of this issue proceeds by understanding absolute truth in terms of
 The metaphysical interpretation of logical truth 241
were genuinely possible, then LNC would obviously not be necessary.
This should be relatively uncontroversial, but I should finally say some-
thing more about ‘genuine possibility’.
As was mentioned in the previous section, there are reasons to under-
stand genuine possibility in terms of metaphysical possibility, as only
metaphysical modality could secure the correspondence between a possible
world and the structure of reality – this is also what CI requires. The
relevant modal space must consist of all possible configurations of the
world and only them. Logical modality cannot do the job because it is not
sufficiently restrictive. This can be demonstrated with any traditional
example of a metaphysical, a posteriori necessity, such as gold being the
element with atomic number 79. Assuming that it is indeed metaphysically
necessary that gold is the element with atomic number 79, we must be able
to accommodate the fact that gold failing to be the element with atomic
number 79 is nevertheless logically possible. But since we are interested in
genuinely possible configurations of the world, we ought to rule out
metaphysically impossible worlds, such as the world in which gold fails
to be the element with atomic number 79. The upshot is that if we accept
the familiar story about metaphysical a posteriori necessities of this type,
then there are necessary constraints for the structure of reality which logical
necessity does not capture.14
The only other viable alternative in addition to metaphysical and logical
modality is conceptual modality, i.e., necessity in virtue of the definitions of
concepts. Nomological modality is already too restrictive, as we sometimes
need to consider configurations of the world that are nomologically impos-
sible but at least may be genuinely possible (e.g., superluminal travel).
However, conceptual modality is too liberal, quite like logical modality, as
it also accommodates configurations of the world which are not genuinely
possible, such as violations of the familiar examples of metaphysical a poster-
iori necessities. If we accept these examples, then neither definitions of
concepts nor laws of logic rule out things like gold failing to be the element
with atomic number 79. Accordingly, if one accepts that there are metaphys-
ical necessities that are not also conceptually and logically necessary – some-
thing that most metaphysicians would accept – the only available
interpretation of genuine possibility is in terms of metaphysical possibility.

metaphysical modality, but the dialetheist could, in principle, endorse paraconsistent set theory and
posit that absolute truth is just truth in the world-model – the model whose domain is the world.
14
I should add that cashing out these constraints is, I think, a much more complicated affair than the
traditional Kripke–Putnam approach to metaphysical a posteriori necessities suggests.
242 Tuomas E. Tahko
There is, however, a way to understand logical modality which may
do a better job in capturing the relevant sense of logical truth. This type
of understanding has been proposed by Scott Shalkowski, who suggests
that ‘logical necessities might be explained as those propositions true in
virtue of the natures of every situation or every object and property, thus
preserving the idea that logic is the most general science’ (Shalkowski
2004: 79). On the face of it, this suggestion respects the criteria for
genuine possibility. According to this approach, logical modality con-
cerns the most general (metaphysical) truths, such as the law of non-
contradiction when it is considered as a metaphysical principle (as in
Tahko 2009). In this view, logical relations reflect the relations of
individuals, properties, and states of affairs rather than mere logical
concepts. Indeed, this understanding effectively equates metaphysical
and logical modality. The idea is that the purpose of logic is to describe
the structure of reality and so it is ‘the most general science’. As
Shalkowski (2004: 81) notes, denying the truth of LNC would, in terms
of this understanding, amount to a genuine metaphysical attitude
instead of, say, the fairly trivial point that a model in which the law
does not hold can be constructed.
Do we have any means to settle the status of LNC in the suggested
sense? A simple appeal to its universal applicability may not do the trick,
but the burden of proof is arguably on those who would deny LNC.
One might even attempt to distill a more general formula from this:
logical principles – which are presumably reached by a priori means –
are prima facie metaphysically necessary principles. They may be chal-
lenged and sometimes falsified even by empirical means, but merely the
fact that we can formulate models in which they do not hold is not
enough to challenge their truth; it will also have to be demonstrated that
there are possible worlds which constitute genuinely possible configur-
ations of the world. However, this approach seems biased towards
historically prior logical principles, the ones that were formulated first.
It is not implausible that the reason why they were formulated first is
because they are indeed the best candidates for metaphysically necessary
principles: for Aristotle, the law of non-contradiction is ‘the most certain
of all principles’ (Metaphysics 1005b22). But this is admittedly quite
speculative – we ought to be allowed to question even the ‘first’
principles.
It would certainly be enough to challenge the metaphysical necessity of
LNC, or other logical principles, if empirical evidence to the effect that the
principle is not true of every situation or every object and property would
 The metaphysical interpretation of logical truth 243
be found.15 This is what Priest has attempted to show with the case of
change and Zeno’s paradoxes, but I remain unconvinced. As I have argued
(Tahko 2009), Priest’s examples can all be accounted for in terms of
semantic rather than metaphysical dialetheism – a distinction developed
by Edwin Mares (2004). The idea is that there may be indeterminacy in
semantics, but this does not imply that there is indeterminacy in the world.
Only the latter type of indeterminacy would corroborate the existence of a
genuinely possible paraconsistent configuration of the world. Since I have
not seen a convincing case to the effect that such a configuration is
genuinely possible, I take it that LNC is a good candidate for a metaphys-
ically necessary principle. If I am right, this means that a paraconsistent
possible world could not have turned out to accurately represent the actual
world. The fact that there are paraconsistent models has no direct bearing
on this question. I do not claim to have settled the status of LNC once and
for all, but I think that a strong empirical case for the truth of LNC can be
made, on the basis of the necessary constraints for the forming of a stable
macrophysical world, i.e., the emergence of stable macrophysical objects.
I have developed the preceding line of thought before with regard to the
Pauli Exclusion Principle (PEP) (Tahko 2012), and electric charge (Tahko
2009). For instance, as PEP states, it is impossible for two electrons (or
other fermions) in a closed system to occupy the same quantum state at the
same time. This is an important constraint, as it is responsible for keeping
atoms from collapsing. It is sometimes said that PEP is responsible for the
space-occupying behavior of matter – electrons must occupy successively
higher orbitals to prevent a shared quantum state, hence not all electrons
can collapse to the lowest orbital. Here we have a principle which captures
a crucial constraint for any genuinely possible configuration of the world
that contains macroscopic objects. Whether or not there are genuinely
possible configurations that do not conform to PEP is an open question,
but it seems unlikely that such a configuration could include stable
macroscopic objects.
Consider the form of PEP: it states that two objects of a certain kind
cannot have the same property (quantum state) in the same respect (in a
closed system) at the same time. Compare this with Aristotle’s formulation
of LNC: ‘the same attribute cannot at the same time belong and not
belong to the same subject in the same respect’ (Metaphysics 1005b19–20).
LNC is of course a much more general criterion than PEP – it concerns

15
I have in mind concrete objects in the first place; see Estrada-González (2013) for a case to the effect
that there are abstracta which violate LNC in this sense.
244 Tuomas E. Tahko
one thing rather than things of a certain kind – but its underlying role is
evident: if any fermion were able to both be and not be in a certain
quantum state at the same time, then PEP would be violated and macro-
scopic objects would collapse. If LNC is needed to undergird PEP, then we
have a strong case in favor of the metaphysical interpretation of LNC in
worlds that contain macrophysical objects, given the necessity of PEP for
the forming of macrophysical objects. This is of course not sufficient to
establish the metaphysical necessity of either principle, but it is an interest-
ing result in its own regard.

4. Pluralism about logical truth

Now that we have a rough idea about the metaphysical interpretation of
logical truth, we can consider the implications of this interpretation in
a wider context. Here I would like to focus on the topic of logical
pluralism, which has lately received an increasing amount of attention.
Perhaps the most influential form of logical pluralism derives from
pluralism about logical consequence, i.e., the view that there are models
in which the logical consequence relation is different, and irreconcilably
so. Beall and Restall have formulated and defended this type of
pluralism:
Given the logical consequence relation defined on the class of casesx, the
logicalx truths are those that are true in all casesx. If you like, they are the
sentences that are x-consequences of the empty set of premises. The logicalx
truths are those whose truth is yielded by the class of casesx alone. Since we
are pluralists about classes of cases, we are pluralists about logical truth.
(Beall and Restall 2006: 100)
If this is indeed what pluralism about logical truth amounts to, then it
appears that anyone who accepts multiple classes of cases is a pluralist
about logical truth. But what does ‘being true in a case’ mean? On the face
of it, one might think that it means exactly the same as ‘being true in a
model’, that is, we are talking about a type of relative truth familiar from
Davidson. This would imply that anyone who accepts multiple classes of
models will also be a pluralist about logical truth. Pluralism about logical
truth would then mean only that there are multiple models, and we can
talk about logical truth separately in each one of these models. But this
would be a rather uninteresting sense of logical pluralism, at least from the
point of view of the metaphysical interpretation of logical truth. However,
as Hartry Field has recently pointed out, this cannot be what Beall and
 The metaphysical interpretation of logical truth 245
Restall have in mind. Moreover, Field suggests two reasons why model-
theoretic accounts are irrelevant to logical pluralism:
One of these reasons is that by varying the definition of ‘model’, this
approach defines a large family of notions, ‘classically valid’, ‘intuitionisti-
cally valid’, and so on; one needn’t accept the logic to accept the notion of
validity. A classical logician and an intuitionist can agree on the model-
theoretic definitions of classical validity and of intuitionist validity; what
they disagree on is the question of which one coincides with genuine
validity. For this question to be intelligible, they must have a handle on
the idea of genuine validity independent of the model-theoretic definition.
Of course, a pluralist will contest the idea of a single notion of genuine
validity, and perhaps contend that the classical logician and the intuitionist
shouldn’t be arguing. But logical pluralism is certainly not an entirely trivial
thesis, whereas it would be trivial to point out that by varying the definition
of model one can get classical validity, intuitionist validity, and a whole
variety of other such notions. (Field 2009: 348)
And the second reason:
[I]f we were to understand ‘cases’ as models, then there would be no case
corresponding to the actual world. There is no obvious reason why a
sentence couldn’t be true in all models and yet not true in the real world.
This connects up with the previous point: the intuitionist regards instances
of excluded middle as true in all classical models, while doubting that they
are true in the real world. (Field 2009: 348; italics original)
Field goes on to suggest that Beall and Restall must have meant that there
is an implicit requirement for interpreting ‘truth in a case’, namely, that
truth in all cases implies truth. Field then argues that this will not produce
an interesting sort of logical pluralism as the pluralist notion of logical
consequence suggested by Beall and Restall does not capture the normal
meaning of ‘logical consequence’. But it should be noted that Beall and
Restall (2006: 36 ff.) do say something about the matter. Specifically, they
suggest that on one reading of ‘case’ (the TM account), Tarskian models
are to be understood as cases. Another reading (the NTP or necessary truth-
preservation account) takes possible worlds to be cases. Beall and Restall
(2006: 40) add that the existence of a possible world that invalidates an
argument entails the existence of an actual (abstract) model that invalidates
the argument.
So, it is not clear that Field’s critique is accurate, as Beall and Restall do
suggest that there is a case that corresponds with the actual world – on the
TM account it is a Tarskian model and on the NTP account it is a possible
world. The latter is of immediate interest to us, given that the metaphysical
246 Tuomas E. Tahko
interpretation of logical truth also makes use of the possible worlds jargon.
Yet, Beall and Restall do not provide an interpretation of possible worlds,
so it is not quite clear what the connection, if any, between the NTP
account and the metaphysical interpretation of logical truth is.
Connecting all this with the analysis provided in the previous section,
one might suggest that classes of cases are sets of metaphysically possible
worlds, distinguished in terms of logical truths that are true in each set of
possible worlds. Only one possible world is actual, but the logical truths
that are true in the actual world will also be true in all worlds which are in
the same set of possible worlds, i.e., these worlds may differ in other
regards, but they are close to the actual world in the sense that all the
logical truths are shared.
Accordingly, pluralists about logical truth, in the metaphysical sense,
hold that there are distinct sets of possible worlds in which different logical
truths hold. The metaphysical interpretation of logical truth can accom-
modate this sense of logical pluralism, provided that possible worlds are
interpreted appropriately – this also enables us to preserve CI.16 However,
accommodating pluralism in the metaphysical interpretation of logical
truth does require a revision in our original definition (ML), which defined
a sentence as logically true if and only if it is true in every genuinely
possible configuration of the world. Since in this view of logical pluralism
there can be proper subsets of genuinely possible configurations with
different laws of logic, we must revise ML as follows:
(ML-P) A sentence is logically true if and only if it is true in every possible
world of a given subset of possible worlds representing genuinely possible
configurations of the world.

ML-P can of course also accommodate the situation where the laws of
logic are the same across all subsets of genuinely possible configurations,
i.e., logical monism – in that case the relevant subset of possible worlds
would not be a proper subset of the genuinely possible configurations.
An alternative formulation of ML-P is possible, dismissing subsets
altogether. We could understand logical pluralism by giving different
interpretations to ‘genuinely possible configurations’.17 This formulation

16
Why is interpreting logical truth on the basis of metaphysical possibility the only way to preserve
CI? Because we’ve seen that only by restricting our attention to metaphysically possible worlds can
we preserve a sense of correspondence between logical truth and genuinely possible configurations of
the world. Only metaphysically possible worlds are sufficiently constrained to take into account all
the governing principles such as metaphysical a posteriori necessities.
17
Thanks to Jesse Mulder for suggesting this type of formulation.
 The metaphysical interpretation of logical truth 247
could be developed by adopting a line of thought from Gillian Russell
(2008). Russell suggests that we can distill a sense of pluralism by under-
standing logical validity as the idea that in every possible situation in which
all the premises are true, the conclusion is true (2008: 594), where possibility is
ambiguous between logical, conceptual, nomological, metaphysical, or
other senses of modality, hence producing a similar ambiguity concerning
validity. A friend of the metaphysical interpretation of logical truth could
accept this idea, but only provided that we prioritize the reading where
possible situations reflect metaphysical possibility, as CI is preserved only
in this reading. Nevertheless, there may still be room for a type of
pluralism concerning metaphysical possibility and hence genuinely pos-
sible configurations. Unfortunately I have no space to develop this
approach further.
It may be noted that since I have been discussing logical pluralism only
with regard to the law of non-contradiction, the resulting sense of
pluralism is limited. Given that I consider there to be strong reasons to
think that LNC holds in the actual world, we can define a set of possible
worlds in which the law of non-contraction holds, call it WLNC. The
assumption is that WLNC includes the actual world. But since I have made
no mention of any other laws of logic that hold (in the metaphysical sense)
in WLNC, the sense in which we can talk of a logic may be questioned. In
other words, it may be wondered if the resulting sense of logical pluralism
is able to support a rich enough set of logical laws to constitute a logic.
However, I suspect that the case can be extended beyond LNC. That is, we
can extend the metaphysical interpretation to other laws of logic as well in
such a way that a subset of WLNC may be defined. This is not quite as
straightforward in other cases though.
Very briefly, consider modus ponens (A ^ (A ! B)) ! B. If thought of
as a rule, it is not obvious that modus ponens can be applied to the world
in the sense that I have suggested with regard to LNC. Yet, there are clear
cases of physical phenomena that feature a modus ponens type structure.
As a first pass, causation might be offered as a candidate of ‘real world
modus ponens’, but there are obvious complications with this suggestion,
as it depends on one’s theory of causation. However, there are better
candidates. Take the simple case of an electron pair in a closed system,
where two electrons occupy the same orbital. As we’ve already observed,
two electrons in a closed system are governed by the Pauli Exclusion
Principle. In particular, since the electrons cannot be in the same quantum
state at the same time, we know that the only way for them to occupy the
same orbital (i.e., having the same orbital quantum numbers) is for them
248 Tuomas E. Tahko
to differ in spin (i.e., to have different spin quantum numbers). Accord-
ingly, when we observe electron A having spin-up, we immediately know
that any electron, B, in the same orbital as A must have spin-down.
Moreover, there can be only two electrons in the same orbital and they
must always have opposite spin.
If cases such as the one for a ‘real world modus ponens’ can be found,
then we may indeed have a rich enough set of logical laws to constitute a
logic, enabling the suggested interpretation of logical pluralism. The
resulting subset could be called WLNCþMP.
This hardly exhausts the debate about logical pluralism, but it appears
that there are ways, perhaps several ways, to accommodate pluralism about
logical truth within the metaphysical interpretation.

5. Conclusion
In conclusion, I have demonstrated that there is a coherent metaphysical
interpretation of logical truth, and that this interpretation has some
interesting uses, such as applications regarding logical pluralism. It has
not been my aim to establish that this interpretation of logical truth is the
correct one, but only that it is of special interest to metaphysicians. I have
assumed rather than argued for a type of realism about logic for the
purposes of this investigation, but I contend that for realists about logic,
one interesting interpretation of logical truth is the one sketched here.18

18
Thanks to audiences at the University of Tampere Research Seminar and the First Helsinki-Tartu
Workshop in Theoretical Philosophy, where earlier versions of the paper were presented. In
particular, I’d like to thank Luis Estrada-González for extensive comments. In addition,
I appreciate helpful comments from Franz Berto and Jesse Mulder. Thanks also to Penny Rush
for editorial comments. The research for this chapter was made possible by a grant from the
Academy of Finland.
 References

Anderson, A. R., and Belnap, N. D. 1975. Entailment: The logic of relevance and
necessity, volume i. Princeton University Press
Anderson, A. R., Belnap, N. D., and Dunn, J. M. 1992. Entailment: The logic of
relevance and necessity, volume ii. Princeton University Press
Anon. 1967. Ars Burana, in L.M. de Rijk, Logica modernorum vol. ii: The origin
and early development of the theory of supposition. Assen: Van Gorcum
Aquinas, T. 2000. Opera omnia, edited by E. Alarcón. Pamplona: University of
Navarre, www.corpusthomisticum.org/iopera.html
Aristotle. 1963. Categories and de interpretatione, translated with notes by J. l.
Ackrill. Oxford: Clarendon Press
1984. Metaphysics, translated by W. D. Ross, revised by J. Barnes. Princeton
University Press
1993. Metaphysics books gamma, delta and epsilon, translated with notes by
C. Kirwan, second edition. Oxford: Clarendon Press
1994. Posterior analytics, translated with a commentary by J. Barnes, second
edition. Oxford: Clarendon Press
2009. Prior analytics, translated and with a commentary by G. Striker. Oxford
University Press
Asenjo, F. G. 1966. “A calculus of antinomies”, Notre Dame Journal of Formal
Logic 7(1), 103–105
Asenjo, F. G., and Tamburino, J. 1975. “Logic of antinomies”, Notre Dame
Journal of Formal Logic 16(1), 17–44
Azzouni, J. 2005. “Is there still a sense in which mathematics can have
foundations”, in G. Sica, (ed.) Essays on the foundations of mathematics and
logic. Monza, Italy: Polimetrica International Scientific Publisher
2006. Tracking reason: Proof, consequence, and truth. Oxford University Press
2008a. “Why do informal proofs conform to formal norms?” Foundations of
Science 14, 9–26
2008b. “Alternative logics and the role of truth in the interpretation of
languages”, in D. Patterson, (ed.) New essays on Tarski and philosophy.
Oxford University Press, 390–429
2010. Talking about nothing: numbers, hallucinations and fictions. Oxford
University Press

249
250 References
2013. Semantic perception: How the illusion of a common language arises and
persists. Oxford University Press
Azzouni, J. and Armour-Garb, B. 2005. “Standing on common ground”, Journal
of Philosophy 102(10), 532–544
Barker, C. 2013. “Negotiating taste”, Inquiry 56, 240–257
Batens, D., Mortensen, C., and Van Bendegem, J.P. (eds.) 2000. Frontiers of
paraconsistent logic. Dordrecht: Kluwer Publishing Company
Beall, Jc 2009. Spandrels of truth. Oxford University Press
2012. “Why Priest’s reassurance is not reassuring”, Analysis 72(3), 739–741
2013. “Free of detachment: logic, rationality, and gluts”, Noûs 47(3), 1–14
Beall, Jc and Restall, G. 2006. Logical pluralism. Oxford University Press
Beall, Jc, Forster, T., and Seligman, J. 2013. “A note on freedom from detachment
in the Logic of Paradox”, Notre Dame Journal of Formal Logic 54(1)
Beebee, H. and Dodd, J. (eds.) 2005. Truthmakers: The contemporary debate.
Oxford University Press.
Bell, J. 1998. A primer of infinitesimal analysis. Cambridge University Press
Benacerraf, P. 1973. “Mathematical truth”, Journal of Philosophy 70, 661–79
Bencivenga, E. 1987. Kant’s Copernican revolution. New York: Oxford University
Press
2000. Hegel’s dialectical logic. New York: Oxford University Press
2007. Ethics vindicated: Kant’s transcendental legitimation of moral discourse.
New York: Oxford University Press
Benthem, J. van, 1985. “The variety of consequence, according to Bolzano”,
Studia Logica 44(4), 389–403
Berto, F. 2007. How to sell a contradiction: Studies in Logic 6. London: College
Publications
2008. “Άδύνατον and material exclusion”, Australasian Journal of Philosophy
86(2), 165–90
Blackburn, S. 1984. Spreading the word. Oxford: Clarendon Press
Bolzano, B. 1810. Beyträge zu einer begründetereu Darstelluns der Mathematik.
Prague: Widtmann
1837. Wissenschaftslehre. Sulzbach: Seidel
1839. Dr Bolzano und seine Gegner. Sulzbach: Seidel
1841. Wissenschaftslehre (Logik) und Religionswissenschaft in einer beurtheilenden
Uebersicht. Sulzbach: Seidel
1985–2000. Wissenschaftslehre, critical edition by Jan Berg. Stuttgart: Fromman-
Holzboog
2004. On the mathematical method and correspondence with Exner, translated by
Rusnock, P., and George, R. Amsterdam-New York: Rodopi
Boolos, G. 1975. “On second-order logic”, Journal of Philosophy 72, 509–527.
(Reprinted in Boolos 1998, pp. 37–53)
1984. “To be is to be the value of some variable (or some values of some
variables)”, Journal of Philosophy 81, 430–450. (Reprinted in Boolos 1998,
pp. 54–72)
1998. Logic, logic, and logic. Cambridge, MA: Harvard University Press
 References 251
Brady, R. 2006. Universal logic. Stanford, CA: Centre for the Study of Language
and Information
Brady, R. and Rush, P. 2009. “Four Basic Logical Issues”, The Review of Symbolic
Logic 2(3), 488–508
Buchholtz, U. 2013. The unfolding of inductive definitions, PhD Dissertation,
Dept. of Mathematics, Stanford University
Burge, T. 1975. “On knowledge and convention”, in Foundations of mind (2007),
32–37. Oxford University Press
Burgess, J. A. 1990. “The sorites paradox and higher-order vagueness”, Synthese 85,
417–474
2010. “Review of Beall and Restall 2006,” Philosophy and Phenomenological
Research 81, 519–522
Burgess, J. P., and Rosen, G. 1997. A subject with no object: Strategies for
nominalistic interpretation of mathematics. Oxford University Press
1999. “Which modal logic is the right one?” Notre Dame Journal of Formal Logic
40, 81–93
Carey, S. 2009. The origin of concepts. Oxford University Press
Carnap, R. 1934. Logische Syntax der Sprache, Vienna, Springer, translated as The
logical syntax of language, New York, Harcourt, 1937; another edition edited
by Amethe Smeaton, Patterson, New Jersey, Littleton, 1959
Cesalli, L. 2007. Le réalisme propositionnel: sémantique et ontologie des propositions
chez Jean Duns Scot, Gauthier Burley, Richard Brinkley et Jean Wyclif. Paris:
Vrin
Chateaubriand Filho, O. 2001. Logical forms: Part i – truth and description.
Campinas: UNICAMP
2005. Logical forms: Part ii – logic, language, and knowledge. Campinas:
UNICAMP
Chomsky, N. 2003. “Reply to Millikan”, in L. M. Antony and N. Hornstein
(eds.) Chomsky and his critics. Oxford: Blackwell Publishing, 308–315
Chomsky, N. and Lasnik, H. 1995. “The theory of principles and parameters”, in
Noam Chomsky The minimalist program. Cambridge, MA: The MIT Press,
13–127
Cole, J. 2013. “Towards an institutional account of the objectivity, necessity, and
atemporality of mathematics”, Philosophia Mathematica 21, 9–36
Condillac, E. 1780. La logique (Logic), translated by W. R. Albury. New York:
Abaris
Crivelli, P. 2004. Aristotle on truth. Cambridge University Press.
da Costa, N. 1974. “On the theory of inconsistent formal systems”, Notre Dame
Journal of Formal Logic 15, 497–510
Daly, C. 2005. “So where’s the explanation?” In Beebee and Dodd 2005,
85–103.
Davidson, D. 1973. “In defense of convention T”, in H. Leblanc (ed.) Truth,
syntax and modality. Amsterdam: North-Holland. Reprinted in his Inquiries
into truth and interpretation. Oxford University Press.
Dehaene, S. 1997. The number sense. Oxford University Press
252 References
Dummett, M. 1975. “The philosophical basis of intuitionistic logic”, Logic
Colloquium ’73, 5–40
1991a. The logical basis of metaphysics. Cambridge, MA: Harvard University
Press
1991b. Frege: Philosophy of mathematics. Cambridge, MA: Harvard University
Press
Dunn, J. M. 1966. The algebra of intensional logics. PhD thesis, University of
Pittsburgh
1976. “Intuitive semantics for first-degree entailments and ‘coupled trees’”,
Philosophical Studies 29, 149–168
Dunn, J. M., and Restall, G. 2002. “Relevance logic”, in D. M. Gabbay and
F. Günthner editors Handbook of philosophical logic (second edition),
volume vi. Dordrecht: D. Reidel
Eddington, A. S. 1928. The nature of the physical world, New York: Macmillan
Epstein, B. 2006. “Review of Ruth Millikan, Language: A biological model,”
Notre Dame Philosophical Reviews 5
Estrada-González, L. 2013. “On two arguments against realist dialetheism”,
unpublished manuscript.
Etchemendy, J. 1988. “Models, semantics and logical truth”, Linguistics and
Philosophy 11, 91–106
1990. The concept of logical consequence. Cambridge, MA: Harvard University
Press.
Feferman, S. 1984. “Toward useful type-free theories I”, Journal of Symbolic Logic
49, 75–111
1996. “Gödel’s program for new axioms. Why, where, how and what?”, in P.
Hájek (ed.) Gödel ’96; Lecture Notes in Logic, 6. Cambridge University Press,
3–22
2006. “What kind of logic is ‘Independence Friendly’ logic”, in R. E. Auxier
and L. E. Hahn (eds.) The philosophy of Jaako Hintikka; Library of living
philosophers vol. 30. Chicago: Open Court, 453–469
2009. “Conceptions of the continuum”, Intellectica 51, 169–189
2010. “On the strength of some semi-constructive theories”, in S. Feferman and
W. Sieg (eds.) Proof, categories and computation: Essays in honor of Grigori
Mints. London: College Publications, 109–129; reprinted in U. Berger, et al.
(eds.) Logic, Construction, Computation. Frankfurt: Ontos Verlag, 2012,
201–225
2011. “Is the Continuum Hypothesis a definite mathematical problem?” lecture
text, Harvard University, Dept. of Philosophy, EFI project. http://math.
stanford.edu/~feferman/papers/IsCHdefinite.pdf
2013. “Categoricity of open-ended axiom systems”, lecture slides for the conference,
“Intuition and Reason”, in honor of Charles Parsons, Tel-Aviv, Dec. 2, 2013;
http://math.stanford.edu/~feferman/papers/CategoricityOpenEnded.pdf
(to appear). “Which quantifiers are logical? A combined semantical and inferential
criterion”, http://math.stanford.edu/~feferman/papers/WhichQsLogical(text).
pdf.
 References 253
Feferman, S. and Strahm, T. 2000. “The unfolding of non-finitist arithmetic”,
Annals of Pure and Applied Logic 104, 75–96
2010. “The unfolding of finitist arithmetic”, Review of Symbolic Logic 3, 665–689
Feynman, Richard P. 1985. QED: the strange theory of light and matter. Princeton
University Press
Field, H. 2009. “Pluralism in logic”, The Review of Symbolic Logic 2(2), 342–359
Folina, J. 1994. “Poincaré on objectivity”, Philosophia Mathematica 2(3), 202–227
Føllesdal, D. and Bell, D. 1994, “Objects and concepts”, in Proceedings of the
Aristotelian Society, Supplementary Volumes, 68, 131–166
Franks, C. 2010. “Cut as consequence”, History and Philosophy of Logic 31(4), 349–79
2013. “Logical completeness, form, and content: an archaeology”, in J. Kennedy
(ed.) Interpreting Gödel: Critical essays. New York: Cambridge University Press
2014. “David Hilbert’s contributions to logical theory”, in A. P. Malpass (ed.)
The history of logic. New York: Continuum
Frege, G. 1918, “Der Gedanke: Eine Logische Untersuchung”, in Beiträge zur
Philosophie des Deutschen Idealismus i. 58–77
1976. Wissenschaftlicher Briefwechsel, edited by G. Gabriel, H. Hermes, F.
Kambartel, and C. Thiel. Hamburg: Felix Meiner
1980. Philosophical and mathematical correspondence. Oxford: Basil Blackwell
Fumerton, R. 2010, “Foundationalist theories of epistemic justification”, The
Stanford encyclopedia of philosophy, Edward N. Zalta (ed.), http://plato.
stanford.edu/archives/sum2010/entries/justep-foundational/
Gentzen, G. 1934–35. “Untersuchungen über das logische Schliessen”. Gentzen’s
doctoral thesis at the University of Göttingen, translated as “Investigations
into logical deduction” in (Gentzen 1969): 68–131
1936. “Die Widerspruchsfreiheit der reinen Zahlentheorie”, Mathematische
Annalen 112, translated as “The consistency of elementary number theory” in
(Gentzen 1969): 132–213
1938. “Neue Fassung des Widerspruchefreiheitsbeweises für die reine
Zahlentheorie”, Forschungen zur logik und zur Grundlegung der exackten
Wissenschaften, New Series 4, translated as “New version of the consistency
proof for elementary number theory”, in (Gentzen 1969): 252–86
1943. “Beweisbarkeit und Unbeweisbarkeit von Anfangsfällen der transfiniten
Induktion in der reinen Zahlentheorie”. Gentzen’s Habilitationschrift at the
University of Göttingen, translated as “Provability and nonprovability of
restricted transfinite induction in elementary number theory”, in (Gentzen
1969): 287–308
1969. The collected papers of Gerhard Gentzen, M. E. Szabo (ed.). Amsterdam:
NorthHolland
George, R. 2003. “Bolzano and the problem of psychologism” in D. Fisette (ed.)
Husserl’s logical investigations reconsidered. Dordrecht: Kluwer, 95–108
1986. “Bolzano’s concept of consequence”, Journal of Philosophy 83, 558–563
Gödel, K. 1931. “Über formal unentscheidbare Sätze der Principia Mathematica
und verwandter Systeme, I”, Monatshefte für Mathematik und Physik 38,
173–98
254 References
1932, “Zum intuitionistischen Aussagenkalkül”, Anzeiger der Akademie der
Wissenschaftischen in Wien 69, 65–66
1933, “Zur intuitionistischen Arithmetik und Zahlentheorie”, Ergebnisse eines
mathematischen Kolloquiums 4, 34–38
1958. “Über eine bisher noch nicht benützte Erweiterung des finiten
Standpunktes”, Dialectica 12, 280–287; reprinted with facing English
translation in Gödel (1990), pp. 240–251
1972. “On an extension of finitary mathematics which has not yet been used”,
in Gödel 1990, 271–280
1990. Collected works, vol. ii : Publications 1938–1974 (S. Feferman, et al., eds.),
Oxford University Press
Gómez-Torrente, M. 1999. “Logical truth and Tarskian logical truth”, Synthese
117: 375–408
2009. “Rereading Tarski on logical consequence”, The Review of Symbolic Logic
2(2), 249–297
Haddock, G. 2010. “Platonism, phenomenology, and interderivability”, in Hartimo
2010a, 23–44
Hales, Alexander of, 1951–1957. Glossae in quatuor libros Sententiarum Magistri
Petri Lombardi. Quaracchi: Florence
Harrop, R. 1960. “Concerning formulas of the types A! B _ C, A! (Ex)B(x) in
intuitionistic formal systems”, Journal of Symbolic Logic 25, 27–32
Hartimo, M. 2010a. Phenomenology and mathematics. Dordrecht: Springer
2010b. “Development of mathematics”, in Hartimo 2010a, 107–121
Hatfield, G. 2003. “Getting objects for free (or not)”, reprinted in his Perception
and Cognition. Oxford University Press, 2009 212–255
Hegel, G. 1995. Lectures on the history of philosophy, vol. ii, translated by
E. S. Haldane and F. H. Simson. Lincoln, NE and London: University of
Nebraska Press
Hellman, G. 2006. “Mathematical pluralism: the case of smooth infinitesimal
analysis”, Journal of Philosophical Logic 35, 621–651
Henkin, L. 1949. “The completeness of the first-order functional calculus”,
Journal of Symbolic Logic 14, 159–166
Hilbert, D. 1899. Grundlagen der Geometrie, Leipzig, Teubner; Foundations of
geometry, translated by E. Townsend. La Salle, IL: Open Court, 1959
1922. “Neubergründung der Mathematik. Erste Mitteilung”, translated by
William Ewald as “The new grounding of mathematics: first report”, in
P. Mancosu (ed.) From Brouwer to Hilbert: the debate on the foundations of
mathematics in the 1920’s. New York: Oxford, 1998, 198–214
1983. “On the infinite”, in P. Benacerraf and H. Putnam (eds.) Philosophy of
mathematics: Selected readings, second edition. Cambridge University Press,
183–201
Hintikka, J. 1996. The principles of mathematics revisited (with an appendix by
G. Sandu). Cambridge University Press
2010. “How can a phenomenologist have a philosophy of mathematics?”, in
Hartimo 2010a, 91–104
 References 255
Horwich, P. 1998. Truth, second edition. Oxford University Press
Hurley, P. J. 2008. A concise introduction to logic, eighth edition. Belmont, CA:
Wadsworth/Thomson Learning
Hurley, S. and Chater, N. 2005a. Perspectives on imitation: From neuroscience to
social science, Volume i: Mechanisms of imitation and imitation in animals.
Cambridge, MA: The MIT Press
2005b. Perspectives on imitation: From neuroscience to social science, Volume ii:
Imitation, human development, and culture. Cambridge, MA: The MIT Press
Husserl, E. 1964. The idea of phenomenology, translated by W. P. Alston and
G. Nakhnikian. The Hague: Martinus Nijhoff
1970. Logical investigations VII, translated by J. N Findlay. New York:
Humanities Press
1981. Husserl shorter works, edited by P. McCormick and F. A. Elliston.
University of Notre Dame Press
1983. Edmund Husserl: Ideas pertaining to a pure phenomenology and to a
phenomenological philosophy (first book), translated by F. Kersten. Boston,
MA: Nijhoff
Isaacson, D. 2011. “The reality of mathematics and the case of set theory”, in
Z. Novak and A. Simonyi (eds.) Truth, reference and realism. Budapest:
Central European University Press, 1–76
Jenkins, C. S. 2005. “Realism and independence”, American Philosophical Quarterly,
42(3)
John of Salisbury 2009. The Metalogicon of John of Salisbury: A twelfth-century
defense of the verbal and logical arts of the trivium, translated by D. D.
McGarry, First Paul Dry Books, edition, 2009; originally published
Berkeley, CA: University of California Press, 1955
Kahneman, D. 2011. Thinking, fast and slow. New York: Farrar, Straus and Giroux
Kanamori, A. 2012. “Set theory from Cantor to Cohen”, in D. M. Gabbay,
A. Kanamori, and J. Woods (eds.) Handbook of the history of logic, vol. vi:
sets and extensions in the twentieth century. Amsterdam: Elsevier
Kant, I. 1781. Kritik der reinen Vernunft, second edition 1787(B), in Kants
gesammelte Schriften, vol. iii and iv, Royal Prussian Academy of Science
Berlin: de Gruyter, 1905
1996. Practical philosophy, translated by Mary J. Gregor, Cambridge University
Press
Karger, E. 2001. “Adam Wodeham on the Intentionality of Cognitions”, in
D. Perler (ed.) Ancient and Medieval Theories of Intentionality. Leiden,
Boston, MA, Köln: Brill, 283–300
(forthcoming). “Was Adam Wodeham an internalist or an externalist?” in
Klima (2014)
Keisler, H. J. 1970. “Logic with the quantifier ‘there exist uncountably many’”,
Annals of Mathematical Logic 1, 1–93
Kennedy, J. 2012. “Kurt Gödel”, The Stanford encyclopedia of philosophy (Winter
2012 Edition), Edward N. Zalta (ed.), URL http://plato.stanford.edu/
archives/win2012/entries/goedel/.
256 References
Kilwardby, R. 1516. [Pseudo-] Aegidius Romanus. In libros priorum analeticorum
Aristotelis Expositio. Venice
1976. De ortu scientiarum, edited by A. Judy. Oxford University Press for the
British Academy
1986. Quaestiones in primum librum Sententiarum, herausgegeben von Johannes
Schneider. Munich: Verlag. der Bayerischen Akademie der Wissenschaften
1988. “On the nature of logic”, in N. Kretzmann and E. Stump (eds.)
Cambridge translations of medieval philosophical texts Volume i: logic and the
philosophy of language. Cambridge University Press, 264–276
1995. Quaestiones in quatuor libros Sententiarum, Appendix : Tabula ordine
alphabeti contexta (Cod. Worcester F 43) Herausgegeben von Gerd
Haverling. Munich: Verlag der Bayerischen Akademie der Wissenschaften
King, P. 2010. ‘Peter Abelard’, in The Stanford encyclopedia of philosophy (Winter
2010 Edition), Edward N. Zalta (ed.), URL http://plato.stanford.edu/
archives/win2010/entries/abelard/
Kleene, S. C. 1945. “On the interpretation of intuitionistic number theory”,
Journal of Symbolic Logic 10, 109–124
1971. Introduction to metamathematics. Amsterdam: North-Holland
Klima, G. 1999. “Ockham’s semantics and ontology of the categories”, in P. V. Spade,
(ed.) The Cambridge Companion to Ockham. Cambridge University Press, 118–142
2001a. “Existence and reference in medieval logic”, in A. Hieke and E. Morscher
(eds.) New Essays in Free Logic. Dordrecht: Kluwer Academic Publishers 197–226
2001b. John Buridan’s Summulae de dialectica, translated by Gyula Klima, New
Haven, CT and London, Yale University Press
2009. John Buridan, Great Medieval Thinkers, Oxford University Press
2011a. “Two summulae, two ways of doing logic: Peter of Spain’s ‘realism’ and
John Buridan’s ‘nominalism’”, in M. Cameron and J. Marenbon (eds.):
Methods and methodologies: Aristotelian logic East and West, 500–1500.
Leiden and Boston, MA: Brill Academic Publishers, 109–126
2011b. “Being”, Lagerlund, H. Encyclopedia of Medieval Philosophy, Springer:
Dordrecht, 100–159
2012. “Being”, in Marenbon, J. 2012, 403–420
2013a. “The medieval problem of universals”, The Stanford Encyclopedia of
Philosophy (Fall 2013 Edition), Edward N. Zalta (ed.), URL http://plato.
stanford.edu/archives/fall2013/entries/universals-medieval/
2013b. “Being and cognition”, in D.D. Novotný and L. Novák (eds.) Neo-
Aristotelian perspectives in metaphysics. To Be Published February 1st 2014 by
Routledge: New York, 104–115
2014 (ed.) Intentionality, cognition and mental representation in medieval
philosophy. New York: Fordham University Press
Klima, G. and Hall, A. 2011. (eds.) The demonic temptations of medieval nominalism,
Proceedings of the Society for Medieval Logic and Metaphysics, Vol. 9,
Cambridge Scholars Publishing: Newcastle upon Tyne
Kripke, S. A. 1982. Wittgenstein on rules and private language. Cambridge, MA:
Harvard University Press
 References 257
Ladyman, J. and Ross, D. (with D. Spurrett and J. Collier). 2007. Every thing
must go, Oxford University Press
Lapointe, S. 2011. Bolzano’s theoretical philosophy. Houndsmill: Palgrave
2014. “Bolzano, Quine and logical truth” in G. Harman, and E. Lepore (eds.)
Blackwell companion to Quine. London: Blackwell
Lepore, E. and Ludwig, K. 2006. “Ontology in the theory of meaning”,
International Journal of Philosophical Studies 14, 325–335
Levinas, E. 1998. Discovering existence with Husserl. Illinois: Northwestern
University Press
Lewis, D. 1969. Convention: A philosophical study. Cambridge University Press
1991. Parts of classes. Oxford University Press
2005. Quotation, in G. Priest, Jc Beall, and B. Armour-Garb (eds.) The law of
contradiction: New essays. New York: Oxford University Press, 1
Lindström, P. 1966. “First order predicate logic with generalized quantifiers”,
Theoria 32, 186–195
1969. “On extensions of elementary logic”, Theoria 35, 1–11
MacFarlane, J. 2005. “Making sense of relative truth”, Proceedings of the
Aristotelian Society 105, 321–339
2009. “Nonindexical contextualism”, Synthese 166, 231–250
2014. Assessment sensitivity: Relative truth and its applications. Oxford University
Press
Maddy, P. 2007. Second philosophy. Oxford University Press
2011. Defending the axioms. Oxford University Press
(to appear) The logical must. Oxford University Press
Marenbon, J. 2012. (ed.) The Oxford handbook of medieval philosophy. Oxford
University Press
Mares, E. D. 2004. “Semantic dialetheism”, in G. Priest, Jc Beall, and
B. Armour-Garb (eds.) The law of non-contradiction. Oxford: Clarendon
Press, 264–275
2007. Relevant logic: A philosophical interpretation. Cambridge University Press
Martin, C. J. 2001. “Obligations and liars”, in M. Yrjönsuur (ed.), Medieval
formal logic: Obligations, insolubles and consequences. Springer, 63–94
2010. “The development of logic in the twelfth century”, in Pasnau 2010,
129–145
2012. “Logical consequence”, in Marenbon 2012, 289–311
Martin, D. A. 2001. “Multiple universes of sets and indeterminate truth values”,
Topoi 20, 5–16
Mathias, A. R. D. 2001. “The strength of Mac Lane set theory”, Annals of Pure
and Applied Logic 110, 107–234
McDowell, J. 1994. Mind and world. Cambridge, MA: Harvard University Press
Meillassoux, Q. 2011. “Contingency & the absolutization of the one”, A lecture
delivered at the Sorbonne for a symposium. Source: http://speculativeheresy.
wordpress.com/2011/03/24.
Meyer, Robert K., Routley, R. and Dunn, J. Michael. 1979. “Curry’s paradox”,
Analysis 39, 124–128
258 References
Millikan, R. G. 1984. Language, thought, and other biological categories. Cambridge,
MA: The MIT Press
1998. “Language conventions made simple”, The Journal of Philosophy 95(4),
61–180. Reprinted in Language: A biological model (2005) Oxford University
Press, 1–23
2003. “In defense of public language”, in L. M. Antony and Norbert Hornstein
(eds.) Chomsky and his critics. Oxford: Blackwell Publishing, 215–237.
Reprinted in Language: A biological model (2005). Oxford University Press,
24–52
2005. “Meaning, meaning, and meaning”, in Language: A biological model.
Oxford University Press, 53–76
Monk, R. 1991. Ludwig Wittgenstein: The duty of genius. Penguin
Morris, M. 2005. “Realism beyond correspondence”, in Beebee and Dodd 2005,
49–66
Mortensen, C. 1995. Inconsistent mathematics, Dordrecht: Kluwer Academic
Publishers
2010. Inconsistent geometry, Studies in logic 27. London: College Publications
Mühlhölzer, F. 2001. “Wittgenstein and the regular heptagon”. Grazer
Philosophische Studien: Internationale Zeitschrift für Analytische Philosophie,
62, 215–247
Mulligan, K., Simons, P., and Smith, B. 1984. “Truth-makers”, Philosophy and
Phenomenological Research 44, 287–321
Myhill, J. 1973. “Some properties of Intuitionistic Zermelo-Fraenkel set theory”,
Proceedings of the 1971 Cambridge summer school in mathematical logic (Lecture
Notes in Mathematics 337): 206–231
Natalis, H. 2008. A treatise of Master Hervaeus Natalis: On second intentions. Vol.
i–An English translation, and Vol. ii–A Latin edition, edited and translated,
by John P. Doyle. Milwaukee: Marquette University
Neta, R., and Pritchard, D. 2007. “McDowell and the new evil genius”,
Philosophy and Phenomenological Research 74(2), 381–396
Novaes, C. D. 2012. “Medieval theories of consequence”, The Stanford Encyclopedia
of Philosophy (Summer 2012 Edition), Edward N. Zalta (ed.), URL http://
plato.stanford.edu/archives/sum2012/entries/consequence-medieval/
Ockham, W. 1974. Summa logicae, opera philosophica i, edited by G. Gál and
S. Brown, St. Bonaventure, NY: Franciscan Institute, St. Bonaventure
University
Panaccio, C. 2004. Ockham on concepts. Aldershot: Ashgate
Pasnau, R. 2010. (ed.) C. Van Dyke (assoc. ed.) The Cambridge history of medieval
philosophy (2 vols.), Cambridge University Press
Pirandello, L. 1990. Novelle per un anno, vol. iii, edited by M. Costanzo, Milano,
Mondadori.
Plato. 1997. “Phaedrus”, translated by A. Nehamas and P. Woodruff. In J. M.
Cooper (ed.) Plato: Complete works. Indianapolis, IN: Hackett
Popper, K. 1968. “Epistemology without a knowing subject”, in Logic,
Methodology and Philosophy of Science iii. Elsevier, 333–373
 References 259
Posy, C. 2005. “Intuitionism and philosophy”, in Shapiro, S. (ed) The Oxford
handbook of philosophy of mathematics and logic. Oxford University Press,
pp. 318–356
Prawitz, D. 1965. Natural deduction: A proof-theoretical study; second edition
(2006), Mineola, NY: Dover Publications
Priest, G. 1979. “The logic of paradox”, Journal of Philosophical Logic 8, 219–241
1980. “Sense, entailment and modus ponens”, Journal of Philosophical Logic 9,
415–435
1987. In contradiction. The Hague: Martinus Nijhoff
1989. “Classical logic Aufgehoben”, in G. Priest, R. Routley, and J. Norman
(eds.) Paraconsistent logic: Essays on the inconsistent. Munich: Philosophia
Verlag, ch. 4
2006a. Doubt truth to be a liar. Oxford University Press.
2006b. In contradiction, second edition. Oxford University Press
2007. “Paraconsistency and dialetheism”, in D. Gabbay and J. Woods (eds.)
Handbook of the history of logic. Amsterdam: North-Holland, 194–204
2009. “Beyond the limits of knowledge”, in J. Salerno (ed.) New essays on the
knowability paradox, Oxford University Press, 93–105
2013. “Mathematical Pluralism”, Logic Journal of IGPL 21, 4–20
(to appear). “Logical disputes and the a priori”
Příhonský, F. 1850. Neuer Anti-Kant. Bautzen: Weller
Prucnal, T. 1976. “Structural completeness of Medvedev’s propositional calculus”,
Reports on Mathematical Logic 6, 103–105
Putnam, H. 2000. “Meaning and reference”, in E. D. Klemke and Heimir
Geirsson (eds.) Contemporary analytic and linguistic philosophies, second
edition. Amherst, NY: Prometheus Books
2003. “McDowell’s mind and world”, in Reading McDowell. London:
Routledge, 174–191
Quine, W. V. O. 1936a. “Truth by convention. In O. H. Lee (ed.), Philosophical
essays for A. N. Whitehead. New York: Longmans, 90–124
1936b. “Truth by convention”, in The ways of paradox, revised and enlarged
edition (1976) Harvard University Press, 77–106
1953. Two dogmas of empiricism from a logical point of view. Cambridge, MA:
Harvard University Press
1960. Word and object. Cambridge, MA: The MIT Press
1970a. Philosophy of logic. Prentice Hall
1970b. “Methodological reflections on current linguistic theory”, Synthese 21,
386–398
1976. The ways of paradox, and other essays (revised and enlarged edition).
Cambridge, MA: Harvard University Press
1986. Philosophy of logic, second edition. Cambridge, MA: Harvard University
Press
1991. “Immanence and validity”, Dialectica 45, reprinted with emendations
in W. V. O. Quine, Selected logic papers. Cambridge, MA: Harvard, 1995,
242–250
260 References
Rathjen, M. 2014. “Indefiniteness in semiintuitionistic set theories: On a
conjecture of Feferman”, arXiv: 1405448
Rayo, A. and Uzquiano, G. (eds.) 2006. Absolute generality. Oxford: Clarendon
Press
Read, S. 1994. Thinking about logic: An introduction to the philosophy of logic.
Oxford University Press
2010. “Inferences”, in Pasnau 2010 173–184.
2011. “Medieval theories: Properties of terms”, The Stanford encyclopedia of
philosophy (Spring 2011 Edition), Edward N. Zalta (ed.), URL = http://
plato.stanford.edu/archives/spr2011/entries/medieval-terms/
Rescher, N. and Manor, R. 1970. “On inference from inconsistent premises”,
Theory and Decision, 1(2), 179–217
Resnik, M. 1988. “Second-order logic still wild”, Journal of Philosophy, 85,
75–87
1996. “Ought there to be but one logic?”, in B. J. Copeland (ed.) Logic
and reality: Essays on the legacy of Arthur Prior. Oxford University Press,
489–517
1997. Mathematics as a science of patterns. Oxford University Press
Resnik, Michael, 2000. “Against logical realism”, History and Philosophy of Logic
20, 181–194
Restall, G. 2002. “Carnap’s tolerance, meaning, and logical pluralism”, Journal of
Philosophy 99, 426–443
Rush, P. 2005. PhD thesis: The philosophy of mathematics and the independent
“other”, School of History, Philosophy, Religion and Classics, The
University of Queensland. http://espace.library.uq.edu.au
2012. “Logic or reason?”, Logic and Logical Philosophy 21, 127–163
Rusnock, P. and George, R. 2004. “Bolzano as logician”, in D. M. Gabbay and
J. Woods (eds.) The rise of modern logic: from Leibniz to Frege. Handbook of
the history of logic, vol. iii. Springer, 177–205
Russell, G. 2008. “One true logic?” Journal of Philosophical Logic 37, 593–611
Rybakov, V. 1997. Admissibility of logical inference rules. Amsterdam: Elsevier
Sainsbury, R. M. 2009. Paradoxes, third edition Cambridge University Press
Schmidt, R. W. 1966. The domain of logic according to Saint Thomas Aquinas. The
Hague: Martinus Nijhoff
Schmutz, J. 2012. “Medieval philosophy after the Middle Ages”, in Marenbon
2012, 245–270
Schotch, P., Brown, B. and Jennings, R. (eds.) 2009. On preserving: Essays on
preservationism and paraconsistent Logic. University of Toronto Press
Schotch, P. and Jennings, R. 1980. “Inference and necessity”, Journal of
Philosophical Logic 9, 327–340
Sellars, W. 1956. “Empiricism and the philosophy of mind”, in H. Feigl and
M. Scriven (eds.) Minnesota Studies in the Philosophy of Science, Vol i,
Minneapolis: University of Minnesota Press, 253–329
1962. “Philosophy and the scientific image of man”, reprinted in his Science,
perception and reality. London: Routledge & Kegan Paul, 1–40
 References 261
Shalkowski, S. 2004. “Logic and absolute necessity”, The Journal of Philosophy
101(2), 55–82
Shapiro, S. 1991. Foundations without foundationalism: A case for second-order logic.
Oxford University Press
1997. Philosophy of mathematics. Structure and ontology. Oxford University Press
2000. “The status of logic”, in P. Boghossian and C. Peacocke (eds.) New essays
on the a priori. Oxford University Press, 333–366; reprinted in part as “Quine
on Logic”, in T. Childers (ed.) Logica Yearbook 1999 Prague: Czech Academy
Publishing House, 11–21
(ed.) 2005. The Oxford handbook of the philosophy of mathematics and logic.
Oxford University Press
2007. “The objectivity of mathematics”, Synthese 156, 337–381
2011. “Mathematics and objectivity”, in J. Polkinghorne (ed.) Meaning in
mathematics. Oxford University Press, 97–108
2012. “Objectivity, explanation, and cognitive shortfall”, in A. Coliva (ed.)
Mind, meaning, and knowledge: themes from the philosophy of Crispin
Wright. Oxford University Press, 211–237
2014. Varieties of logic, forthcoming
Sher, G. 2001. “The formal-structural view of logical consequence”, The
Philosophical Review 110(2), 241–261
2011. “Is logic in the mind or in the world?” Synthese 181, 353–365
Shim, M. 2005 “The duality of non-conceptual content in Husserl’s phenomenology
of perception”, Phenomenology and the Cognitive Sciences 4, 209–229
Sider, T. 2011. Writing the book of the world. Oxford University Press
Siebel, M. 2002. “Bolzano’s concept of consequence”, The Monist 85(4), 580–599
Simpson, S. G., and Yokoyama, K. 2012. “Reverse mathematics and Peano
categoricity”, Annals of Pure and Applied Logic 164, 229–243
Slaney, J. 2004. “Relevant logic and paraconsistency”, in L. Bertossi, A. Hunter,
and T. Schaub (eds.) Inconsistency tolerance. Heidelberg: Springer-Verlag,
270–293
Smith, Nicholas J. J. 2012. Logic: The laws of truth. Princeton University Press
Sorensen, R. 2012. ‘Vagueness’, The Stanford encyclopedia of philosophy (Summer
2012 Edition), Edward N. Zalta (ed.), URL http://plato.stanford.edu/
archives/sum2012/entries/vagueness/
Spade, P. V. 1994. (trans.) Five texts on the mediaeval problem of universals:
Porphyry, Boethius, Abelard, Duns Scotus, Ockham. Indianapolis,
Cambridge: Hackett
Spelke, E. 2000. “Core knowledge”, American Psychologist 55, 1233–1243
Spelke, E., Grant, G. and Van de Walle, G. 1995. “The development of object
perception”, in S. Kosslyn and D. Osherson (eds.) Visual Cognition.
Cambridge, MA: MIT Press, 297–330
Steiner, M. 1975. Mathematical knowledge. Ithaca and London: Cornell University
Press
2009. “Empirical regularities in Wittgenstein’s philosophy of mathematics”,
Philosophia Mathematica 17, 1–34
262 References
Swoyer, C. 2003. “Relativism”, Stanford internet encyclopedia of philosophy, http://
plato.stanford.edu/entries/relativism
Tahko, T. E. 2009. “The law of non-contradiction as a metaphysical principle”,
The Australasian Journal of Logic 7, 32–47. URL: www.philosophy.unimelb.
edu.au/ajl/2009/
2012. “Boundaries in reality”, Ratio 25 (4), 405–24
Tarski, A. 1944. “The semantic conception of truth”, Philosophy and
Phenomenological Research 4, 13–47
Teiszen, R. 2010. “Mathematical realism and transcendental phenomenological
idealism”, in Hartimo 2010a, 1–22
Terwijn, S. A. 2004. “Intuitionistic logic and computation”, Vriendenboek ofwel
Liber Amicorum ter gelegenheid van het afscheid van Dick de Jongh, Institute
for Logic, Language, and Computation. Amsterdam
Tokarz, M. 1973. “Connections between some notions of completeness of
structural propositional calculi”, Studia Logica 32(1), 77–91
Troelstra, A. S., and van Dalen, D. 1988. Constructivism in mathematics. An
introduction, Vol. i. Amsterdam: North-Holland
Väänänen, J. 2001. “Second-order logic and foundations of mathematics”, Bulletin
of Symbolic Logic 7, 504–520
van Inwagen, P. 1981. “Why I don’t understand substitutional quantification”, in
Ontology, identity, and modality: Essays in metaphysics (2001). Cambridge
University Press, 32–6
Visser, A. 1999. “Rules and arithmetics”, Notre Dame Journal of Formal Logic 40,
116–140
Waismann, F. 1945. “Verifiability”, Proceedings of the Aristotelian Society,
Supplementary Volume 19, 119–150; reprinted in A. Flew (ed.) Logic and
language. Oxford: Basil Blackwell, 1968, 117–144
Walker, Ralph C.S. 1999. “Induction and transcendental argument”, in R. Stern
(ed.) Transcendental arguments. Oxford University Press, 13–29
Wason, P. C., and Johnson-Laird, P. 1972. Psychology of reasoning: Structure and
content. Cambridge, MA: Harvard University Press
Weber, Z. 2009. “Inconsistent mathematics”, Internet encyclopedia of philosophy,
www.iep.utm.edu/math-inc/
Wells, H. G. 1908. First and last things. Teddington Echo, 2011
Willard, D. 1981. Introduction to “A reply to a critic”, in Husserl 1981, 148–151
Williamson, T. 1994. Vagueness. London: Routledge
Wilson, M. 2006. Wandering significance. Oxford University Press
Wittgenstein, L. 1953. Philosophical investigations, second edition, translated by
G. E. M. Anscombe. Oxford: Blackwell. 2001
1974. On certainty, edited by G.E.M. Anscombe and G.H. von Wright.
Translated by D. Paul, and G.E.M. Anscombe. Oxford: Blackwell
1976. Wittgenstein’s lectures on the foundations of mathematics, Cambridge, 1939.
Ithaca: Cornell University Press
1978. Remarks on the foundations of mathematics (revised edition). Cambridge,
MA: MIT Press
 References 263
2009. Philosophical investigations, translated by P. M. S. Hacker, G. E. M.
Anscombe, and J. Schulte, fourth revised edition, edited by P. M. S. Hacker
and J. Schulte. Wiley-Blackwell
Wittgenstein, L. and Rhees, R. 1974. Philosophical grammar. Oxford: Basil
Blackwell
Wright, C. 1992. Truth and objectivity. Cambridge, MA: Harvard University Press
2008. “Relativism about truth itself: haphazard thoughts about the very idea”,
in M. García-Carpintero and M. Kölbel (eds.) Relative truth. Oxford
University Press
Yang, F. 2013. “Expressing second-order sentences in intuitionistic dependence
logic”, Studia Logica 101(2), 323–342
Zucker, J. I. 1978. “The adequacy problem for classical logic”, Journal of
Philosophical Logic 7, 517–535
Zucker, J. I., and Tragesser, R. S. 1978. “The adequacy problem for inferential
logic”, Journal of Philosophical Logic 7, 501–516
 Index

admissible inference rule, 110, 113–114, 116 contradiction, 8–9, 16, 27, 53–54, 72, 83, 107–108,
analyticity, 195–196, 200, 202 134–135, 137, 141–143, 149, 182–183, 213,
Aristotelian, 147, 204 224, 228, 240
categories, 149–150, 152–153 convention, 3, 33, 39, 115
Aristotle, 6–8, 22, 54, 117, 139, 147–150, 205, 207, Lewis account of, 34–35
212–213, 239, 242–243 tacit, 34, 36–38, 40–41
Crivelli interpretation of, 236 truth by, 32, 34, 47–48
notion of validity, 204, 206–207, See Bolzano conventionalism
Axiom of Choice, 72, 86, 88, 90 linguistic, 35
logical. See logical conventionalism
Beall, Jc, 229–230, 232 conventions, 33–35, 45, 191
Beall and Restall, 245 explicit, 33–35
Beall and Restall’s pluralism, 50, 69–70, 236, logical, 32–33
244–246, See Restall, Greg optimality of, 36–38, 44–45
Beall, Hughes and Vandegrift, 9 tacit, 34–35
Priest and Beall, 232 correspondence, 44–45, 195, 234–237, 241, 246
bounded arithmetic, 116 1-1, 89
Burgess, J.A., 50–51, 57, 237 criteria for
validity, 8
Carnap, R, 69, 233 criterion for
Dummett-Quine-Carnap, 69–70 a philosophy of mathematics, 89
classical validity, 59, 114, 245 legitimacy, 52
cognitive command, 58–60, 62–63, 65–66, mathematical legitimacy, 55
68–71 rule-following, 131
completeness, 80–83, 114, 116, 123, 180–181 validity, 169
theorem, 42, 64, 80, 84, 181, See Gödel
Condillac, 122, 127 Darapti, 213
conditional Davidson, D, 57, 233–234, 236, 244
logic, 107 dialetheism, 108
material, 106–107, 218 metaphysical, 243
the, 105, 107, 231 semantic, 243
consistency, 51–52, 55, 58–60, 79, 115, 118–119, disjunction property, 110–113
141–142, 144, 188 Dummett, M, 59–60, 68–70, 128, 140–141, 234
constructive, 82–83, 115–116, 119, 125
mathematics, 74, 108 Eddington, A.S., 95–97, 99
semi-constructive, 86 epistemic constraint, 58–61, 70
constructivism, 74, 82 Etchemendy, J, 234–235, 237–238
contextualism, 49, 66–67, 69–70, 257 Explosion, 107, 213–214, 228
continuum, 29, 54, 64, 76–80, 90,
97, 105 Feferman, S, 3–4, 75–77, 79–82, 86, 88–89,
Continuum Hypothesis, 89, 110 91–92, 105–106

264
 Index 265
Field, H, 211, 244–245 mind-, 20, 56
first-order logic, 42–43, 61, 64, 74, 80, 82, 213, of facts, 14
226 of logic, 7
Føllesdal, D, 20, 92 of logical truth, 29
formalism, 74 proofs, 51
Frege, G, 1, 41, 94, 128, 131, 157, 195, 214 realist, 3, 15–18
and Russell, 129, 211 results, 83
correspondence with Hilbert, 51–52 intuitionism, 115, 140
Fregean, 22 intuitionist validity, 245
realism, 196 intuitionistic, 69, 74, 112
analysis, 52, 54
Gentzen, G, 111–112, 116, 122–124 consistency, 59
meaning, 85 intuitionistic logic, 116
proof, 118–120 intuitionistically, 45, 70, 113
system of natural deduction, 81, 84 logic, 46, 50, 52, 54, 57, 60–61, 63–64, 74, 77,
geometry, 8, 52, 93, 123, 128, 136, 208 82–83, 108, 111–112
application, 136, 215 predicate calculus, 84–86
axioms, 52 propositional calculus, 110
Euclidean, 41, 77, 128, 215 semi-, 86, 89, See logic, intuitionist
non-Euclidean, 186, 215 semi-intuitionism, 85
Gödel, K, 87, 92, 111, 116, 118–119, 142, 191
completeness theorem, 42, 80 Jankov’s logic, 115
incompleteness theorem, 79
Goldbach conjecture, 76, 139–141 Kant, 20, 41, 57, 94, 180, 183, 187, 195, 203, 208,
213
Hatfield, G, 102 Anti-, 198
Hilbert ethics, 184–185
Hilbertian, 8 Kantian, 7, 179, 181, 183
Hilbert, D, 51–52, 110, 118–119, 211 Kant-Quine, 58, 71
Hilbertian, 52, 55 KF-structure, 94
Hilbert’s program, 142
space, 72, 104 Ladyman, J, 99
Husserl, E, 17, 25–28, 30 Ladyman, J and Ross, D, 99, 104
cognition, 22 language acquisition, 40, 103
concept of evidence, 23–24 law of excluded middle (LEM), 29, 50, 65, 74,
conception of logic, 18–19 87–88, 90, 139–142, 144
Husserlian, 26 weak, 115
logical realism, 190 law of non-contradiction (LNC), 9, 29, 48, 239,
phenomenological reduction, 19–21 242
transcendence, 17 Lewis Carroll regress, 33
logic
idealization, 71, 106–107 applied, 215
classical logic, 106–108, See logic, classical canonical application, 2, 215–216, 220
of rudimentary logic, 5, See logic, rudimentary classical, 4–5, 42–45, 50, 52–53, 77, 81, 84,
technique of, 105–106 86–88, 107, 111–113, 115, 211, 214, 216, 218,
throughout mathematics, 61 228
implication, 41, 73 application to mathematics, 91
intuitionistic, 125 idealization, 106–108
incompleteness, 113–114 rise of, 214
theorem, 79, See Gödel valid in, 60–61, 63
independence, 3, 13 conditional, 107
conceptions of, 26 content-containment model of, 42
essential and modal, 15 deviant, 69, 104, 106–107
human-, 2, 15 intuitionist, 215–216, 219
IF Independence Friendly, 82 mathematical, 72–73, 212, 214, 217
266 Index
logic (cont.) metalogical debates, 48
medieval, 147, 158–159, 161, 164, 212–214 mirror neuron, 39
Megarian, 212 model theory, 44–45, 226–227
non-classical logic, 5, 49, 211, 218 model-theoretic, 42, 64, 80, 204, 207, 235–236,
paraconsistent, 50, 54–55, 64, 68, 108, 211, 215, 238, 245
225, 240, See paraconsistency account of validity, 221–223
Port Royale, 213 modus ponens, 43, 48, 95, 113, 136, 189–190, 214,
pure, 18, 178, 215–216 228, 230, 247–248
relevance, 54, 107 monism, 51, 54, 62, 217, 246
rudimentary, 5, 95, 97–100, 104–108, 120–121, 125
rule-governed model of, 41–42 naturalism, 3–4, 74, 189
semi-intuitionist, 4 necessity, 42, 159, 186, 204–207, 241,
substitution model of, 42 See possibility
traditional, 214, 217 causal, 174
logica docens, 212–216, 218, 220, 223 epistemic, 208
logica ens, 212, 216, 220, 223 follows of, 205
logica utens, 212, 218–219, 223 logical, 35, 134, 174, 241
logical metaphysical, 242, 244
connectives, 23, 115–116, 222 natural, 175
consequence, 8, 51, 59–60, 79, 109–110, 112, semantical, 181
123, 235 non-realist, 3–4, 74
Beall and Restall’s, 50, 244–245 norms of reasoning, 158
Bolzano, 203–204, 207
in mathematical practice, 43 objectivity, 3, 14, 56–60, 65–66, 71, 174, 179–180,
material approach to, 232 184, 186
Read’s defense of material, 228 axes of, 62
traditional definition of, 192 criteria of, 183
conventionalism, 3, 33, 47, 190 of logic, 7
inference, 5, 41, 93, 100, 134 of mathematics, 76
pluralism, 4, 9, 217, 237, 244–248 open-texture, 71
realism, 4, 8, 13–15, 189–192, 195–197, 208, 233
schemata, 41 paraconsistency, 9–10, See logic, paraconsistent
logical validity, 50, 56, 121–123, 161, 237, 247 Peano Arithmetic, 73, 87
logicism, 74 Piaget, 100
Plato, 18, 122, 162, 164, 166, 168, 176, 184
MacFarlane, J, 51, 65–67 platonic, 18, 25, 74, 97, 162
Maddy, P, 5, 121 platonism, 5, 135, 139–140
mathematical platonist, 136, 147, 221
objects, 1, 90, 221 pluralism, 9, 49, 51, 69, 189, 247, See logical:
proof, 42, 120, 123 pluralism
realism, 14 possibility, 70, 197, 228, 247, See necessity
reality, 1, 15 genuine, 241
McDowell, J, 25–30 logical, 174
meaning metaphysical, 241, 247
of a mathematical proposition, 137 of cognition, 13, 16–17, 21–22
of ‘all’, 78 of logic, 32
of logical operations, 24, 85 Priest, G, 3, 9, 158, 225–226, 229–230, 232
of logical particles, 112 arguments against LNC, 240, 243
of logical predicates, 225, 228 principle of bivalence, 32
of logical terms, 69, 135 principles and parameters model, 39
of ‘proposition’, 150
of spoken and written utterances, 148–149 quantum mechanics, 98–99, 104, 108, 144
of ‘stateable’, 155 Quine, W.V.O.
Medvedev lattice, 115 on second-order logic, 124
metalogical, 45 substitutional procedure, 195
 Index 267
Quine, W.V.O., 32, 81, 100, 124, 135, 166 -type, 204
challenge, 32–35, 38–40, 48 Tarskian conception, 238
Dummett-Quine-Carnap, 68–69 Tarskian model, 50, 207, 245
holism, 144 theory choice, 9, 216, 223
Kant-Quine, 58, 71 truth
Putnam and Davidson, 57 absolute, 233–234, 237–239, 241
Quinean, 178, 181, 216 by convention, 32, 47–48
in a model, 233
rationality, 168, 184 logical, 3, 9, 29, 41, 46, 95, 99, 233, 240, 242
relativism, 49, 51, 129, 190, 220, 238 all, 105
folk-, 49–51, 57–60, 63, 65–66, 68, 70–71, 190 first-order, 42
logical, 49, 51 ground of, 93
proper, 67–68, 70 interpretation, 233–234, 248
rule-following, 61, 129–132, 137, 142 metaphysical, 8–9, 233, 235–239, 244–247
pluralism about, 244–245
second-order logic, 61, 64, 73, 79, 81, 128 Quine on, 233
classical, 74 realist, 238
full, 79, 82 reflecting facts, 97
Quine on, 81, 124 preservation, 44, 160, 222, 245
semantics, 81 preserving, 44, 160
Sellars, W, 22, 97, 99, 196 relative, 234, 236–238, 244
objection, 14–15 truth tables, 46
set theory, 6, 64, 81, 88, 92, 124, 136, 181, 231 T-schema, 44, 46, 233–235
background, 42
development of, 120–121 vagueness, 56, 61–65
Kripke-Platek, 88 logics of, 106
paraconsistent, 241 real, 107
satisfiability in, 52 worldly, 96
Zermelo-Fraenkel (ZF), 73
Shapiro, S, 2–4, 9, 43, 61, 78–79, 118, 189, 233 Waismann, F, 57, 71
smooth infinitesimal analysis, 52–54 Wason Card Test, 218–219
Spelke, E, 102–103 Wittgenstein, L, 5, 125, 144, 195, 219
structuralism, 3, 74 and physics, 144
conceptual, 3–4, 78, 80, 90–91 means by ‘postulate’, 144
in-re, 74 on mathematics, 128–129
modal, 74 on rule-following, 129–132, 143
rejection of Hilbert’s program, 142
Tarski, A, 195, 202, 234–236, 238 Steiner on, 6
biconditionals, 44–45 Wittgensteinian, 45, 132, 141
definition of logical consequence, 203 Wright, C, 49, 58–60, 62–63, 70
Generalised Tarski Thesis, 50, 70
T-schema, 233 Zermelo-Fraenkel set theory, 73