Azzouni - Deflating Existential Consequence
Azzouni - Deflating Existential Consequence
Azzouni - Deflating Existential Consequence
Jody Azzouni
3
2004
3
Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
São Paulo Shanghai Taipei Tokyo Toronto
2 4 6 8 9 7 5 3 1
Printed in the United States of America
on acid-free paper
Acknowledgments
Introduction 3
1
As a result, we find recent philosophers, e.g., van Inwagen (2000)—first writing:
—and then, without going on to tell us how we should understand this criterion (except to
say, vaguely, that “it is a strategy or technique, not a thesis”), using it (or something like
it?) to commit us to the existence of fictional entities.
2
And so on? Absolutely: aesthetics, for example.
3
4 Introduction
3
An interesting historical point: Carnap (1939) offers what Psillos (1999, pp. 10–1)
explicitly describes as an indispensability argument for “theoretical entities.”
Introduction 5
Naturalizing Ontology
Some philosophers have been exploring what might be called the program
of naturalized ontology,4 the idea that in science ontological practices are
already in place that dictate scientific commitments, and that to engage
in “naturalized ontology” is simply to adopt those practices. I’ve some
sympathy with this move, and broadly speaking, it’s my escape route from
ontological nihilism5 to the safety of a particular criterion for what exists.
But there are at least two crucial ways I break with this approach. First, I
doubt that one finds in science autonomous practices exemplifying a crite-
rion for what exists (or a criterion for what scientific discourse commits
us to): Science (I claim, and I hope) is not as revolutionary as that. The
criterion for what exists operating in science is drawn from epistemic and
ontological practices at work in the human community at large—it’s a
criterion we’ve adopted in common. (I say a little more about this
shortly.) Second, I continue to hold that the “ontological nihilism” of
my 1998 is right: there are no philosophically conclusive ways to argue
for our criterion for what exists. That is, we can imagine alternative com-
munities with the same science we have but with different beliefs about
4
Esp. Penelope Maddy.
5
I borrow this depressing label for my earlier (1998) position from a (currently) un-
published paper of Penelope Maddy’s.
6 Introduction
what exists because (and solely because) they have a different criterion
for what exists; they’re otherwise unaffected by their choice. In particular,
there is nothing we can point to, either practically speaking or in terms
of some implicit incoherence in their practices or theories, that shows
they’ve got the wrong criterion for what exists (see chapter 4 for a discus-
sion of this).
Notice what’s required when it’s claimed that there is nothing to
adjudicate between our criterion for what exists and the different one of
the imagined alternative community: I’m not asking for an a priori resolu-
tion of the disagreement (I don’t believe in the existence of things like
that). I’m asking for some way of being able to show (empirically) that
either we or the members of the alternative community are worse off be-
cause of our (or their) choice of criterion for what exists—that, for exam-
ple, our (or their) scientific or epistemic practices fall short in some way.
Only then can we say that there is a genuine issue to be judged rationally.
Applied Mathematics
Unlike the Quinean criterion, which commits us to everything the object-
ual existential quantifier in (regimented) scientific doctrine ranges over,
my criterion is more discriminating. Each scientific case must be individu-
ally investigated in fair detail to see what it commits us to and what it
doesn’t. This is because when mathematical discourse is applied empiri-
cally, the terms in it (which indicate abstracta) can play two very different
roles: sometimes they proxy for items that exist, and sometimes they
don’t.
As an illustration of this complex and interesting phenomenon, part
II of this book examines a (small) portion of Newtonian physics. The
point of this application is threefold: first, to see in a real but relatively
straightforward case how my approach sorts out what exists from what
doesn’t; second, to see how ontological evaluation can yield principled
results even in a context that far outstrips the commonsense arena that
our ontological promptings were first honed in; and third, to see how
powerful methodological prejudices about explanation can conspire with
routine generalizing practices in mathematics to produce (when that mathe-
matics is applied) strong (and yet unjustified) ontological intuitions. I single
out the substantivalist/relationist debate about space and space-time as
an excellent illustration of this last phenomenon. I’ll show how the appli-
cation of geometries to motion that yields simple and unified laws leads
to intuitions about the existence of absolute space.
Principled Results
In a way, the hope for principled results, raised as my second aim earlier,
brings us back to Carnap’s original motivation for deflating ontology.
What look like bizarre positions—at least from the commonsense point
Introduction 7
6
I don’t care much for Kant’s ultimate position, I should add, but that’s a relatively
minor matter, all things considered.
7
Quine (1951) writes that one effect of his abandoning what he describes as the two
dogmas of empiricism is “a blurring of the supposed boundary between speculative meta-
physics and natural science” (p. 20).
8 Introduction
without some redistribution of microphysical states” (p. 98). Later on the same page, he
describes the job of physics as “full coverage” in this sense and (still on the same page)
describes his physicalism as “nonreductive” and “nontranslational.” But what, we have a
right to wonder, entitles Quine to such a position?
13
In earlier writing, esp. my 2000b, I unthinkingly adopted Quine’s phrase “web of
beliefs” even as I labored to undercut the theoretical deductivism that made that phrase
attractive to him. Once this terminological inconsistency was pointed out to me by Jonathan
Hodge, I felt I had to find an alternative. Unfortunately, what I came up with is not as
attractive as Quine’s phrase.
My phrase, and Quine’s too, is unfortunate in a different respect: Belief is a psychologi-
cal term, but these beliefs are at best ones held in common. Alas, the naive picture of like-
minded scientists peacefully assembling a system of beliefs held in common is surely wrong.
The situation is complicated by various “deference practices” that detach the beliefs “held
10 Introduction
in common” from the psychological states of the individuals these “common beliefs” belong
to. To sort out the sociological mechanisms involved in the detail they need goes far beyond
what’s possible, given the themes this book is supposed to cover. Nevertheless, the reader
should keep in mind the implicit complexity of what’s actually going on (when I refer to
our “body of beliefs”), especially when psychological issues, about when we’re committed
to the truth of something or when we tentatively believe something, are brought to bear
on “beliefs held in common” that aren’t genuine psychological entities. I apologize for the
brevity and cryptic nature of these remarks: I plan to discuss this issue more thoroughly at
some future time.
I will, however, continue to use the phrase “web of beliefs” when speaking of Quine’s
views.
Introduction 11
Why Empirically
Indispensable
Mathematical Doctrine
and (Some) Scientific
Law Must Be Taken as
True: Preliminary
Considerations
1
Some of the arguments in this chapter and the next originally appeared in my 2000b,
pp. 49–51.
15
16 Truth and Ontology
one by exhibiting the sentence(s) that the term in the blind ascription
picks out.2
It’s widely known that no explicit use of the truth predicate is ever
indispensable. This is because Tarski biconditionals, the equivalences be-
tween sentences and explicit truth ascriptions to those sentences, allow us
to replace explicit truth ascriptions with the sentences themselves that
truth is ascribed to. From this, it follows that only blind truth ascriptions,
if any, involve instances of the truth idiom that are indispensable, and this
is true only of ineliminable blind truth ascriptions. Are there such?
Yes—two kinds, actually. First, consider my brilliant roommate who
regularly proves mathematical truths of all sorts and who never rashly
asserts anything. Suppose he steps into my room at the end of the day
and tells me: “I’ve just proved something new.” Suppose he even says
what it is that he has just proved (but, of course, I don’t quite catch it).
I might say later, to someone else: “The thing Paul proved about Hilbert
spaces today—whatever those are—is true.” As long as Paul has proved
one and only one thing about Hilbert spaces that day, not only does the
definite description in my statement determinately refer, but such a state-
ment is also exactly the sort of straightforward truth-attribution regularly
engaged in when what’s taken to be true (or false) can’t itself be used
explicitly. The point that it’s only possible to construct an explicit truth
ascription to a statement if we have “transparent access” to it (i.e., only if
we’re acquainted with the statement itself, vs. merely having a description
of it) let us call the transparency condition on explicit truth ascriptions.
The other kind of ineliminable blind truth ascription involves assent
to infinitely many sentences. If I want to assert three short sentences, I
can do so by asserting a conjunction of them. If I want to assert my belief
in the implications of a theory T (all infinitely many of them), I then have
to say, “All the implications of T are true.” Explicit truth ascriptions to
infinite sets of sentences are impossible because any explicit truth ascrip-
tion must exhibit the sentences it attributes truth to. Call this the finite
statability condition on explicit truth ascriptions.
Of course, it’s not just infinite sets of sentences that resist explicit
exhibition: Finite (but large) sets of sentences can resist, as well. This is
important to notice because the indispensability of the truth predicate is
an indispensability to the practice of asserting statements. It’s not right,
therefore, to describe the truth predicate as dispensable in a case where
someone can assent to a (long) conjunction of statements but only if
given a certain outrageous amount of time (a thousand years, say). There
is no principled point at which a finite set of sentences is too large to
exhibit explicitly; this turns on whether simple abbreviations of them are
in use, the medium in which they occur (print, orally), and so on. But
2
Definitions for blind and explicit false ascriptions are similar. There is an explicit use
of the truth predicate without quote marks: “It is true that. . . . ” For our purposes, there’s
no difference between it and explicit uses with quote marks.
Why Doctrine and Law Must Be Taken as True 17
it’s clear that, beyond a certain point, a finite set of sentences can’t be
exhibited explicitly, and then the truth idiom is essential to our assertion
practices. Perhaps this should be called the manageable exhibition condi-
tion on explicit truth ascriptions. For ease of exposition in what follows, I
focus only on those cases where the ineliminability of the truth idiom is
due to commitments to infinite sets of sentences, and for the most part I
leave aside concerns with unmanageably large finite sets of sentences. The
reader should not forget them, however.
When it comes to truth ascriptions to theories or statements in sci-
ence, it may seem obvious that it’s never failure to meet the transparency
condition that breeds ineliminable blind truth ascriptions; after all, scien-
tists applying a statement or theory know what statement or theory
they’re applying.3
But then, why should ineliminable blind truth ascriptions ever come
up in scientific practice? They can only arise—according to the foregoing
reasoning—because of blind truth ascriptions to infinite sets of sentences;
but aren’t infinite sets of sentences, after all, rather unwieldy? No, recur-
sive ones need not be: They’re not if there is a perspicuous decision proce-
dure for recognizing any member of such an infinite set. Infinite but
recursive sets of sentences of this sort are (practically) quite as usable as
finite sets of sentences, and in fact they come up in science often: The
identification of a scientific or mathematical theory by a set of axioms,
where that set of axioms is infinite, is common.
For expository purposes, I temporarily adopt an idealized description
of an aspect of scientific practice, that of deducing implications from ma-
thematized (and axiomatized) empirical theories.4 A recursive set of
axioms—that is, a set of axioms for which a decision procedure is avail-
able—can be one that’s as easy to use and learn as a finite set of axioms.5
3
Let’s accept this claim for now: I show later in this chapter that some accepted scien-
tific theories aren’t explicitly exhibitable, and so blind truth ascriptions to such theories will
be ineliminable.
4
I say “mathematized empirical theory” because, so often, the language of an empiri-
cal theory involves ineliminable mathematical idioms. This deductive picture, however, is a
falsification in at least two respects. First, scientific theories are often deductively intractable:
The desired results are ones we cannot (and do not) extract from the theory by deduction.
Second, scientific theories need not be well-defined axiomatic systems. I touch on both these
points in what follows. See my 2000b for further details, especially on the first observation.
5
Notoriously, Davidson 1965 seems to claim otherwise. Pace that view, the axioms for
Peano arithmetic (PA), e.g., include an induction schema: [A(0) & (x)(A(x) → A(x′))] →
(x)A(x), an infinite set of axioms that arise from this schema by the substitution of any
formula (with the variable x free) for A(x). This infinite set of sentences is particularly
perspicuous because it contains only substitution-instances of a schema. A less perspicuous
set of sentences (but equally usable and routinely studied) is the set of tautologies of any
first-order language. Axiomatizations of such languages usually just include tautologies
wholesale (because truth tables provide an easy decision procedure for them). See, e.g.,
Enderton 1972.
Useful terminology: A set of sentences T is finitely axiomatizable if there is a finite set
of sentences from which the sentences of T can be deduced. PA is not finitely axiomatizable.
18 Truth and Ontology
This is because in both the finite and infinitely recursive cases, mechanical
methods are available for verifying that purported proofs are valid and
sound.6
But, and this is the crucial point, although implications can be drawn
from PA (say) just as easily as from a finitely axiomatized theory T, blind
truth ascriptions can be eliminable in the second case, although not in
the first.
For consider: Presume a commitment to the truth of T, and thus to
the truth of some implication of T, which I represent here by the sche-
matic symbol “Im.” Provided T has a manageably exhibitable conjunction
of its axioms, which I represent by the schematic letter “A,” the informal
description of our truth-commitments just given (involving “truth”) can
be eliminated by using the sentences themselves instead: A. (A → Im).
Im.7 In the first case, however (where “Im” represents an implication of
PA), we must say
PA is true. PA implies ‘Im’. ‘Im’ is true. Im.
(The last step requires the Tarski biconditional: ‘Im’ is true ⇔ Im.) Be-
cause PA is not finitely axiomatizable, it can’t be exhibited explicitly; the
use of the truth idiom in this inference is ineliminable.
A study of the axiomatizability of first-order theories reveals how
quickly and easily recursively but not finitely axiomatizable mathematical
theories come up.8 Whenever such theories are used to draw implications,
blind truth ascriptions are ineliminable.
The use of schemas where a finite axiomatization isn’t available is not
restricted to applied mathematical doctrine, however. Consider Newton’s
6
Because not all mechanical methods are equally practical, an infinite set of sentences
can sometimes be more useful than a quite large finite set of sentences. The former may
come in an especially perspicuous form (e.g., iterated conjunctions of the single sentence
“Peter is running”) that makes the decision procedure for them easy, whereas a list for the
latter, even though finite, may be quite difficult to peruse.
7
To do this we identify T with a particular conjunction of its set of axioms. In general
T can be (finitely) axiomatized, and the axioms ordered, in more than one way. Similarly,
a statement can be expressed in more than one way. In both cases, fixation on a particular
incarnation of a statement or theory for purposes of assertion is harmless.
In what’s forthcoming, “‘Im’” is a schematic symbol representing the Tarskian meta-
language name of the implication that the schematic symbol “Im” represents. “PA” will
function as a name for Peano Arithmetic both in the discourse of this book and in the
discourse containing the inferences I give as examples. This will also be true of variables
standing for theories that arise later in this chapter.
I should say that I won’t be particularly fussy about use/mention matters, quotation,
or quasi-quotation except when I fear misunderstanding.
8
See, e.g., Chang and Keisler 1973, pp. 41–3. Although PA is an appealing illustration
because it’s an intuitively simple axiomatization of ordinary arithmetic, and yet is both in-
complete and not finitely axiomatizable, there is, in general, no connection between finite
axiomatizability and completeness. Q (p. 43) is a well-known subtheory of PA that is finitely
axiomatizable and incomplete. Presburger arithmetic, although not finitely axiomatizable,
is complete.
Why Doctrine and Law Must Be Taken as True 19
9
A substantial discussion of schemas in scientific contexts may be found in Koslow
2000, although my understanding of “schema” differs from his. He describes Newton’s
second law as schematic because it leaves open what sorts of forces there can be. I agree
but regard an instantiation of any law as still schematic unless all the parameters in it have
values. (One of Kuhn’s examples, as well, is not an instantiation in my sense; what’s missing
is the substitution of a specific scalar for “l.”) Koslow also takes schemas to have meanings,
whereas I adopt the Quine line that—strictly speaking—schemas don’t say anything them-
selves but stand for a set of sentences. See what’s forthcoming later in this chapter.
Arnold Koslow tells me in an e-mail dated October 2000 that Sidney Morgenbesser
had made the point (to him) about the schematic character of physical laws as early as 1955.
20 Truth and Ontology
Quine has claimed again and again that schemas themselves can’t be
used to express anything.10 His point is that schemas are only (diagram-
matic) tools used to talk about the sentences of a language: they allow us
to codify an infinite set of sentences and to talk about that set in the
context of a metalanguage. If, therefore, we’re interested in characteriz-
ing a set of sentences by schemas and then assenting to them, the truth
idiom is still needed, and what results is a blind ascription (e.g., “All the
instances of the induction schema are true”).
One recalls how notoriously finicky Quine is about use/mention er-
rors; he’s positively frightened that someone will unconsciously treat
schematic letters as variables, bind them with quantifiers, and in this way
inadvertently adopt a higher-order theory of some sort. So it’s true that
in Quine’s hands schemas—with their “dummy letters”—are understood
as diagrams that—horrors!—one had better not quantify into unless one
wants to slip into unintended commitments to attributes (or worse).
Consequently, on this view, schemas are not statements and can’t be used
to assert collections of sentences. But why share Quine’s terminological
puritanism?
Instead, let’s introduce substitutional quantifiers (Σ and Π, say) with
substitution-classes containing the sentences of the language, predicates
holding of classes of those sentences, and special quotation devices which
the substitutional quantifiers can bind variables within. If “I” is im-
pounded as a predicate so that all substitutions of instances of the induc-
tion schema of PA for “x” in “I‘x’” yield truths, we can then commit
ourselves to all the instances of the induction schema by asserting:
(Πp)(I ‘p’→p).11
Unfortunately, if the aim is to circumvent indispensability of the truth
idiom, this fails because once substitutional quantifiers are available, we
can define truth outright.12 Moral: Dressing up the truth idiom in impos-
ing notation won’t dispense with it.
A similar point applies to those tempted to reach out and grab what
they need: The truth idiom is needed to assert a recursively infinite collec-
10
Quine 1941, pp. 18–9; 1986, p. 67; 1955, pp. 153–4; 1972a; and esp. 1972b, pp.
272–3. In Quine (1980) he says of sentential calculus schemas: “No meaning is to be
attached to such expressions: they serve only in the manner of diagrams, in connection with
general discussions of truth-functional structure” (p. 36). Similar passages occur often in
his corpus.
11
Useful terminology: Objectual quantifiers come with a given domain of objects that
they range over. An existential sentence with an objectual quantifier (e.g., (∃x)Sx) is true
(roughly) if there is something in the domain that is in the extension of the formula Sx.
Substitutional quantifiers come with a substitution class of terms that can be substituted
for the substitutional variable. An existential sentence with a substitutional quantifier (e.g.,
(ΣX)SX) is true (roughly) if there is a term O in the substitution class of the quantifier and
if the sentence SO is true. As Quine (1969, p. 106) points out, any set of terms can serve
as a substitution class, even the left parenthesis.
12
Given equality, we can define truth thus: Ta ⇔ (Σp)(p & ‘p’ = a), where a is a
name-schema of sentences in the substitution class of (Σp).
Why Doctrine and Law Must Be Taken as True 21
tion of sentences? Fine, let’s just assert the collection of sentences itself
by means of an infinite conjunction of those sentences. Perhaps it can
even be argued that this is routine in ordinary language: We indicate how
additional cases can be constructed, and then we wave our hands and say,
“and so on.” But the infinitary conjunction also gives us a logical idiom
from which a truth predicate can be directly defined—just as before. In
both cases, we don’t escape the truth predicate as much as express it by
other means.
One last strategy of this type may have occurred to the reader. Finite
axiomatizability seems relative to the ontology (in the Quinean sense) of
the theory in question. Introduce more objects of a certain sort (quantifi-
cation over sets or properties, in the case of PA; quantification over forces,
in the case of (the bit of) Newtonian mechanics given above; quantifica-
tion over propositions or sentences in the case of tautologies of the senten-
tial calculus), and then finite axiomatizability (of the resulting theory)
follows. Logically speaking, that is, indispensable use of the truth predi-
cate isn’t very deep: Essential use of the truth predicate can be eliminated
by introducing more substantial quantifier commitments. A quite similar
strategy is to argue that the finite-axiomatizability problem is a parochial
one due to the choice of first-order logic. Use a higher-order logic in-
stead; second-order PA, for example, is finitely axiomatizable.13
One argument that seems available against this (and which I’m sym-
pathetic to) is that, again, the truth predicate has not really been elimi-
nated; rather, its powers have been tucked into another notation, into the
ontology, as it were. This is clearest and most explicit in the case of the
second-order move: The second-order version of PA is more powerful
than first-order PA; that it’s more powerful is provable because the truth
predicate for first-order PA is definable in second-order PA.14
As I said, I’d like to apply this point to cases where the language
remains first order but the ontology (in Quine’s sense) has been aug-
mented. This may not be convincing, however, in all such cases: The
postulation of propositions or forces can be motivated apart from the
13
The induction schema is replaced by the second-order sentence: ∀X[(X0 & ∀x(Xx →
Xsx)) → ∀xXx].
14
See, e.g., Shapiro 1991, p. 110. The moral here, and with respect to the previous
approaches just considered, is that logical devices that enable us to bypass the truth idiom
are (pretty much) powerful enough to define that idiom in their own terms. This shouldn’t
be surprising: There’s a sense in which the truth predicate is the minimally powerful logical
device for the purposes it serves.
One caveat: Of course second-order PA provides (at least directly) neither a truth
predicate for itself nor a truth predicate for the (empirical) sentences to which number
theory is to be applied. On the first point, this is the same situation that we find with
(predicative) truth idioms generally. The second point can, in general, usually be finessed
by extending the range of the truth predicate (e.g., for first-order PA) to include those
(true) sentences. As discussed in part II, this is, in any case, relatively unimportant because
most applications of mathematics to empirical subject matters involve, in a very real sense,
absorbing those empirical subject matters into that of the mathematics being applied.
22 Truth and Ontology
desire to avoid ineliminable blind truth ascriptions. And so, rather than
attempt the uphill struggle of arguing that quantification over a type of
thing amounts (in certain cases) to an implicit coding of a truth predicate,
I’d prefer to go a different route. This is, first, to show that there are
numerous cases in empirical science where indispensable blind truth as-
criptions come up in a way that can’t be eliminated by the supplementa-
tion of ontology in Quine’s sense. I do this in the very next paragraph.
My second move is to show that if the role of the truth predicate is for
the purposes of blind truth ascription, then even if indispensable use of
the truth predicate (in certain cases) is absent, this doesn’t absolve us of
a commitment to truths. This argument is given in chapter 2 (see esp.
the summary and concluding remarks).
I’ve allowed myself (up until now) a description of scientific theories
where their implications are recognized via deduction. But as footnote 4
indicated, science is more complicated; we must often use tools other
than deduction to find the implications of a theory. I now discuss certain
aspects of this that reveal how the transparency condition contributes to
the indispensability of the truth idiom to scientific practice.
There is a tendency to divide scientific methodology discretely into
induction and deduction. Crudely characterized, the view takes the de-
ductive side of scientific methodology to be one where a theory is in place
and implications are deduced from it (using logic). Of course, back-
ground mathematics is in place, as well, but this is neatly described—on
the deductive picture—as just so many more premises in deductions of
the consequences of a scientific theory.
Unfortunately, deductive consequences about a type of situation are
often out of reach: Using the above talk of instantiations of physical-law-
schemas, either (a) the instantiation (of the law-schemas in question) can
be written down, but the mathematics involved makes the instantiation
too intractable to extract the desired consequences from,15 or (b) an accu-
rate description of the physical situation is too complicated to allow us
to write down the instance to begin with (the required parameters and
constraints are too complicated and numerous), or worse, we’re simply
ignorant of crucial details, and so we’re prevented from giving the in-
stance (the actual shapes of the molecules are too irregular to discover
the details of, the actual intermolecular forces are too complicated to get
a grip on). In case (b), notice, the desired instantiation itself can’t be
15
Sussman and Wisdom (1992):
Our physical model is the same as that of QTD except in our treatment of the effects
of general relativity. General relativistic corrections can be written in Hamiltonian
form, but we have not been able to integrate them analytically. Instead we used the
potential approximation of Nobili and Roxburgh . . . which is easily integrated, but
only approximates the relativistic corrections to the secular evolution of the shape and
orientation of the orbit. (p. 57)
Why Doctrine and Law Must Be Taken as True 23
16
This should be familiar. For all practical purposes—and practical purposes are what
matter when asserting statements—most instances of the induction schema can’t be written
down either. (They’re too long.)
17
These examples are drawn and modified from my 2000b, pt. 1, § 2. It’s worth
mentioning that these tricks—and others—are often applied simultaneously.
18
For details on the derivation, see Purcell (1985, pp. 276–9). Other examples, such
as Ohm’s law and self-inductance, are in Purcell (1985, pp. 128–33, 283). Also see Feyn-
man et al. (1963, 15–10) for approximating formulas in relativistic dynamics, oscillations
in various contexts (23–4, 25–5, 30–11), polarization of light (33–9), etc. All these exam-
24 Truth and Ontology
ples have the general form just described: An aspect of the phenomenon is ignored, and if
its numerical contribution to the resulting effect is sufficiently tiny, then using the derived
formula will result in deviations small enough to live with (given our purposes, and measure-
ment capacities, at that time).
19
Often, not merely is the instantiation simpler, it may be—strictly speaking—an in-
stantiation of a simpler set of law-schemas. For example, we may neglect a certain type of
force altogether (e.g., in calculating the trajectory of certain planets, we not only leave out
a number of other small bodies—which we presume won’t have much impact—but also
ignore the effects of electromagnetic forces, because those are also small). In effect, we
replace a model of a world where certain laws apply with one in which those laws don’t
apply at all.
Why Doctrine and Law Must Be Taken as True 25
20
Sheer geometry can pose problems for the instantiation of law-schemas: E.g., con-
tours are shaped in all sorts of ways—but the number of contours that can be neatly and
precisely characterized by tractable mathematical equations is tiny by comparison. Often,
however, a mathematical description of the shape of a contour must be inserted into a law-
schema in order to construct an instance (e.g., if a current is traveling along a wire bordering
an area). So, in practice, contours that can’t be (tractably) described by mathematical equa-
tions must be replaced by ones that can be.
26 Truth and Ontology
these cases. Such a strategy can’t be carried out unless the theory can at
least be indicated in schematic form.
Something even odder is possible. We may be unable to write down
the physical-law-schemas themselves even though we’re able to recognize
instances of those schemas (in certain idealized situations) or at least im-
plications of instances of those law-schemas. Can we become committed
to a scientific theory in a case where we can’t write down a physical-law-
schema for it? Of course. We may find that all the implications we’re able
to recognize as following from instances of such a theory-schema are ones
that convince us that all the instances of that theory-schema are true. In
such a case, commitment to the truth of the theory can follow despite
the inexpressibility of the theory—even in schematic form.
21
Why the explicit restriction of the examples to our own language? First, because
“true” as used in the vernacular applies to sentences whether they occur in our language or
not; the value of blind truth ascription isn’t exhausted by its application to our own lan-
guage. But, second, what licenses the truth predicate in its application to other languages
can’t be those Tarski biconditionals expressible in our own language. So there is a serious
sense in which the logic of the truth predicate (in the vernacular) is not captured by Tarski
biconditionals. See my 2001 for details about this.
Why Doctrine and Law Must Be Taken as True 27
Bob regards John as a totally sober and trustworthy individual. Then, for
any sentence A of our language, Bob can engage in the following bit of
reasoning:
Everything John says is true. John said A. Therefore A.
22
In my 2000b, I described this requirement as the topic neutrality of truth—but
univocality is perhaps better jargon.
28 Truth and Ontology
Concluding Remarks
What’s been shown about indispensably applied mathematical doctrine
and empirical scientific law? The example of PA is suggestive, and what it
suggests is that we’re simply committed to the truth of every sentence in
a body of statements that—as a whole—is characterized by a description
(e.g., reference to a schema), which requires blind truth ascription for
application, and which is indispensable to empirical science.
This isn’t quite true, however. Imagine that (for whatever reason) we
refuse to commit ourselves to any implication of PA larger in length than
some number n, where n is large enough that no sentence with that
number of symbols (or more) can ever be applied by us in an inference.
Then, to avail ourselves of an implication Im of PA, we needn’t engage
in the inference: PA is true. PA implies ‘Im’. ‘Im’ is true. Im. We can say
instead: Every implication of PA that is smaller than n is true. ‘Im’ is an
implication of PA that is smaller than n. ‘Im’ is true. Im.
What this shows is that if we can descriptively subdivide a set of sen-
tences in an appropriate way, we needn’t be committed to the truth of
all of them, despite having to use blind truth ascriptions. Thus, we
needn’t be committed to all the sentences in a (schematically presented)
theory if we can indicate a subclass of them the truth of which we prefer
to be exclusively committed to. Whether this can be turned to the advan-
tage of that philosopher who wishes not to be committed to the truth of
empirically applied mathematical doctrine or to (some) scientific laws is
something I address in chapter 2 (see esp. the discussion in section (B3)).
One last point that should be stressed: Those who are convinced that
the ascription of truth to a sentence only amounts to the assertion of the
sentence itself may think that escaping the need to use the truth predicate
in the case of a class of sentences won’t suffice for an escape from a
commitment to those sentences (and, really, to their truth). This is cor-
rect, as the discussion of the manifestation condition (also in chapter 2,
(B3) will make clear.
2
Circumventing
Commitment
to Truth
despite Empirical
Indispensability
Here’s where we are: The indispensability of the truth idiom coupled with
the indispensability of (certain) statements of mathematics and empirical
science seems to require us to assent to the truth of such statements in
order to apply them empirically. But a number of philosophers are either
fictionalistic or agnostic about the truth of empirically applied mathemat-
ics, or even about (certain) empirical laws themselves.1 One can easily
imagine strategies to enable such philosophers to retain their respective
positions by circumventing, one way or another, the points about blind
truth ascription raised in chapter 1. This chapter is dedicated to showing
why most of these strategies, which many readers will have in mind, won’t
work, and why the one that does is insufficient for the purposes of philo-
sophical agnostics and fictionalists.
1
Examples of those who deny that indispensable scientific laws are true, or at least are
agnostic about them: Cartwright 1983, Hacking 1983, van Fraassen 1980. Examples of
those who reject the Quine-Putnam indispensability thesis by denying the truth, and thus
our need to assent to the truth, of applicable mathematics: Maddy 1992, Sober 1993.
29
30 Truth and Ontology
(B1) Epistemic
One can argue that the mere indispensability of a statement doesn’t suf-
fice for a commitment to its truth—one needs something stronger, for
example, a certain sort of confirmation or the statement’s passing a certain
sort of empirical test. So Cartwright (1983), for example, argues that
2
See Field 1980 for an attempt at (i) and (ii) for a restricted bit of physics. See Burgess
and Rosen 1997 for a more recent discussion of the technical issues.
3
There are tricks that finesse the need for (iii), perhaps the need for (ii), and that can
weaken (i). See section (B3). Nevertheless, as I show, the execution of even weakened forms
of this program still pose insurmountable obstacles.
4
It’s not easy to attribute strategies outlined in sections (B1) and (B3) cleanly to
distinct specific thinkers, so in what follows I attribute more than one strategy to the same
philosophers.
Circumventing Commitment to Truth 31
(B2) Semantic
Real Truth, one can argue, isn’t an idiom restricted to blind truth ascrip-
tion as its role; that deflationary view of truth is philosophically inade-
quate. Everyone (let’s say) can agree that the Tarski biconditionals are
one necessary condition on a truth idiom, but other conditions are neces-
sary too. Some philosophers may think that one particular additional con-
dition is required; others may claim that there are several competing
richer notions of truth. All may agree, however, that unless one or an-
other additional condition is satisfied by indispensable mathematical doc-
trine or by empirical laws, we shouldn’t consider them (really) TRUE,
although they may be (deflatedly) true.5
A variant of this move is to try to claim that such indispensable uses
of the truth idiom when applying mathematics or empirical laws are nev-
ertheless nonliteral uses. Thus, although one must say that such state-
ments are true, one nevertheless doesn’t literally mean it.6
These won’t work either. Let’s first consider the richer notion of
truth idea and then turn to the nonliteralist ploy. The problem is that
any supplementation of deflationist (Tarski-biconditional governed) truth
faces a dilemma. Either it falsifies a Tarski biconditional (and so proves
unfit for blind ascription), or it fails to be a genuine supplementation of
the deflationist notion of truth. Let me illustrate these claims with some
examples.
5
See Wright 1992 for the view that the Tarski biconditionals provide an unstable
minimal notion of truth that must be supplemented with additional conditions in order to
yield (one or another) richer notion of truth we can use. Candidate additional conditions
may include (among others) (i) some sort of correspondence to facts or the world, (ii)
verification under ideal circumstances of inquiry, (iii) ultimate acceptance by the community
of knowers, and (iv) best coherence with our background beliefs.
6
My thanks to Sarah McGrath for urging me to consider this objection. Attributions
of metaphoricality or pretense to avoid otherwise apparent commitment to the truth of
certain sorts of statements or to the existence of certain sorts of entities, e.g., fictional or
mathematical ones, are becoming popular. I further discuss these views, and my objections
to them, in chapter 3.
32 Truth and Ontology
7
In particular, (ii), (iii), and (iv) of n. 5 are examples. See my 2000b, pt. 2, § 7, for
further details on this argument and for consequences.
8
My thanks to Penelope Maddy for urging me to explicitly discuss the compositional
approach to truth. I should add that my argument against supplementations of the sheer
quantificatory role of truth appears (in its essentials) in my 2000b, p. 51 (see (ii), in partic-
ular).
Circumventing Commitment to Truth 33
(B3) Instrumentalist
This last strategy is one I spend a lot of time on (pretty much the rest of
this chapter) because it’s a strategy that attracted many early practitioners
9
Notice that this is not an objection to compositional analyses of language per se, or
to the intimate connection such analyses, following Tarski, have had to the notion of truth.
Any axiomatization of any notion can be supplemented—we can always add new terminology
to the language and then connect that terminology to the old language in a way that doesn’t
disturb the previous axioms; that’s not the issue. The point is that the Tarski biconditionals
can’t be supplemented in a way that can legitimately be described as enriching the notion
of truth (vs. merely wedding other concepts to it). Compositional analyses of language
don’t themselves illuminate the notion of truth, and, related to this, attempts to wed no-
tions of truth to ontology via compositional analyses of language are misguided. The second
point will be illustrated in what follows. For more details on the first point, see my 2001
and my forthcoming(b).
One caveat: I’ve again suppressed refinements about the truth predicate in the interests
of expository ease. See the general remarks about truth that chapter 1 closes with, esp. n. 21.
34 Truth and Ontology
10
The reader will notice in what follows that I’m not distinguishing between taking
an empirically applied (and, sometimes, indispensable) scientific theory to be truth-valueless
and taking it to be false. For our purposes, this particular distinction doesn’t matter. So
hereon, both applied empirical theories that are taken to be false and those that are taken
to be truth-valueless are described as “instrumentally construed.”
11
What makes Ptolemaic astronomy empirically indispensable (at a particular time) is
that at that time it offers the only way to predict and retrodict the apparent movements of
the planets, the sun, and the moon. Many philosophers, following Duhem, think Ptolemaic
astronomy was understood instrumentally in just this way by its practitioners. Hatfield
(1990) writes sternly: “Much mischief has been caused by the acceptance of Duhem’s char-
acterization of Ptolemaic astronomy as inherently ‘instrumental’ . . . ” (p. 155, n. 37). To
avoid historical impropriety, I subsequently substitute the ahistorical example of P, a theory
used instrumentally by its (imagined) practitioners, for actual Ptolomaic astronomy as his-
torically practiced. It certainly was possible for Ptolomaic astronomy to be treated instrumen-
tally, and besides, there are other examples of theories that are treated instrumentally in
current science, as I indicate later.
Circumventing Commitment to Truth 35
12
I use “quarantine” with some misgivings. I like its sound, and its connotations,
except for the suggestion (for some readers) that all quarantines are temporary. They
needn’t be in this case. Although some theories may shift in their status from an instrumen-
tal tool to something we whole-heartedly believe, in many cases this isn’t possible, and the
implications of the theory placed in quarantine remain there forever. (“Exile,” an earlier
option I considered, is missing the connotation of temporality but sometimes gives rise to
awkward phrasing; “ostracize” is just too ugly. So “quarantine” it is.)
13
I’ve previously (1997a, pp. 196–7; 2000b, p. 54) discussed this under the rubric of
“isolating a theory understood instrumentally from the rest of our web of beliefs.”
14
A special case of a theory with quarantined implications (which may be the most
common case in practice) is where none of the implications of that theory are contained in
our body of beliefs; such cases, where none of the implications of a theory are taken as true
but nevertheless they’re still seen as useful, is when such implications numerically approxi-
mate what’s true (relative, of course, to a background theory that we do take to be true).
36 Truth and Ontology
15
Presumably, this approach is indispensable if we want to make predictions about the
magnetic field at distant points from an arbitrary piece of material of this sort. This is be-
cause of, in most cases, our inability to compute the actual field generated by the slab’s
microstructure. Nevertheless, it’s perfectly clear that it’s entirely legitimate for scientists to
take an instrumental attitude toward this theory.
Circumventing Commitment to Truth 37
16
This is important. The potential evidence for or against a theory is always subject to
change because certain implications of a theory (heretofore insulated against confirmation
or infirmation) can become testable via mathematical breakthroughs or instrumental ad-
vances. I put this point to polemical use a few pages hence.
38 Truth and Ontology
etry. Consider the old Bob Dylan line: “The sun’s not yellow; it’s
chicken.” (Surely we don’t take all, and perhaps any, of the literal implica-
tions of this sentence to be true, and this is how we intuitively recognize
the presence of metaphoricality in this case.) Thus, it’s reasonable to as-
sert that a metaphorically construed sentence is not part of our body of
beliefs; its literal paraphrase (if it has one) is part of our body of beliefs;
and even if it hasn’t a literal paraphrase, still, some of its implications may
(literally) be part of our body of beliefs.17
Three caveats: First, of course, it’s always possible for the status of a
theory to change—an instrumental theory may be one we later take to
be true (or, more commonly, vice versa). And this change in attitude will
be due to our (now) concerning ourselves with implications of the theory
that we once ignored.
Second, there is one way of taking a theory instrumentally that I
should be explicit about. This is when a paraphrase for it is available (and
so it’s only out of sheer laziness, as it were, that we use the unparaphrased
theory). I’m not explicitly concerning myself with this sort of case (and
so I’m not qualifying my claim that placing (some) implications of a the-
ory in quarantine provides both necessary and sufficient conditions for a
theory to be instrumentally construed because I’m implicitly treating a
theory as empirically indispensable when no such paraphrase is available).
It’s worth pointing out, however, that it’s well known to philosophers
that there are scientific theories that scientists seem to treat instrumentally
but for which no paraphrase seems available. Maddy (1997) tells us that
on the paraphrase approach:
[A] claim about a continuous fluid would be replaced by, or translated as, a
claim about what happens to actual fluids as they approximate ever more
closely to ideal fluids, that is . . . as their molecules become ever more tightly
packed together, approximating continuous matter. But this is all wrong. If
the molecules of an actual fluid are packed tightly enough, it stops being a
fluid, and even if it didn’t, the best we could approximate in this way would
be density, not full continuity. (p. 145)18
17
There’s more to say about this interesting topic but I can’t get into it in this book.
One point that needs stressing is that placing some implications of a sentence or theory in
quarantine is hardly a sufficient condition for the presence of metaphor: The quarantined
implications of instrumentally construed scientific theories hardly show those theories to be
metaphorical.
18
These particular considerations are hardly fatal to the paraphrase proponent—for
whoever said that the only way to paraphrase (and thus eliminate) a scientific theory was to
show that it approximates theories that we take as really true, in the sense that a limit
operation on the entities (presupposed in the approximating theory) that takes them to real
entities similarly takes the approximating theory to the true theory. (Well, there may be
places in Quine’s writings where he, at least, seemed to imply something like this. But he
shouldn’t have.) But in any case, Maddy is right about the prospects of the paraphrase
option: scientists are perfectly willing to use any theory at all—whether they believe it or
not—if it can be shown to have (some) empirically valuable implications. They don’t stop
to show that the theory so used (and yet instrumentally construed) is paraphrasable away.
Circumventing Commitment to Truth 39
Notice by contrast how easily this case is dealt with on the quarantine
view of instrumentally construed empirical theories.
Third, suppose we had no truth idiom and no way of expressing blind
ascriptions. Even so, we could only mark a difference between sentences
we were committed to and ones we were merely taking instrumentally by
placing (some) implications of the instrumentally construed sentences in
quarantine. For someone who sees facilitating blind truth ascription as
the sole purpose of the truth idiom, this is the heart of the indispensability
argument. As I’ve shown, if it’s claimed that some sentence in our body
of beliefs—although (1) ineliminable by paraphrase and (2) none of its
implications are in quarantine—is still not to be taken as literally true
(see, e.g., Yablo 2000 for this kind of view), the response is that if no
implication of the sentence—literally construed—is in quarantine, then
the claim that this sentence is not literally taken to be true has no manifes-
tation whatsoever with respect to the role of that sentence in our body
of beliefs. (Offstage whisperings about instrumentality or metaphoricality
don’t count as “manifestations.”)19
Let’s turn now to philosophical attempts to attribute instrumental
status to scientific theories. The threat of an alleged instrumental attitude
toward a theory evaporating into the ineffable seems to sour a (tradi-
tional) instrumental approach to theories—to introduce a division in the
vocabulary of the language (theoretical vs. observational) that is then to
impose a separation of the empirically valuable observational truth-bear-
ing consequences of a theory from the rest of its purely theoretical conse-
quences; the latter can then be treated as truth-valueless. Many philoso-
phers have pointed out that this project is doomed if only because the
required separation of vocabulary into observational and theoretical ter-
minology can’t be carried out;20 this is right, of course, but I’m pressing
a different concern, which is that such a distinction in vocabulary is taken
to support a distinction in attitude toward sentences we take to be true
and those we treat merely as instruments for deducing something else we
take to be true. But such a difference in attitude can’t reside solely in
nomenclature: It must be grounded in a difference in how the termino-
logically distinct sentences are treated or used in our body of beliefs to
avoid the charge that the purported difference in attitude involves (at
most) an ironic stance toward vocabulary—in particular, as I pointed out
19
The considerations just raised reverse a claim made in my 2000b, pp. 50–1. Also
see the second-to-last paragraph of this chapter.
The reader may think there are still other ways to manifest an instrumental attitude
toward theories. I’m skeptical that any of these possibilities will prove valuable in the way
that philosophers who urge an instrumentalist attitude toward scientific law or mathematical
doctrine need; see the discussion later of Cartwright (1983), where I raise another such
possibility.
20
For classical animadversions to the observation/theory distinction as applied to
terms, see, e.g., Putnam 1962, Maxwell 1962, or Achinstein 1965.
40 Truth and Ontology
21
I’m going along (for the sake of argument) with this psychologistic take on com-
monly held beliefs, but see the second paragraph of n. 13 of the introduction.
42 Truth and Ontology
22
Recall that strategy described in section (A) is the attempt to show that applied
mathematics only appears to be indispensable, and that mathematical statements are actually
dispensable instrumental inference tickets between empirical statements by (i) recasting em-
pirical subject matters, which commit us to undesirable mathematical abstracta, so that they
don’t so commit us; (ii) showing that the mathematics so applied is conservative with re-
spect to the mathematics-free empirical subject matter of (i); and (iii) showing that it’s
practically possible to dispense with the inferential role that mathematics has on this view.
Circumventing Commitment to Truth 43
23
This is obvious if the logic in which these theories are couched is first order. But
what if it’s second order? Then it still can be done—but in such a case, because the logic is
incomplete, we have to treat the alternatives not as alternative models of S but as alternative
semantic frameworks for second-order logics (with different principles). This is as far as I
can get into this now.
44 Truth and Ontology
Hacking (1983) and Cartwright (1983) point out that more than one
scientific theory may be applicable to the same phenomenon. In studying
the electron, say, there are any number of competing theories—which
disagree—that scientists may apply to yield results about electrons. This
motivates the claim that, according to scientific practice, although the elec-
tron is taken to be real, theories about electrons aren’t generally taken to
be true.
I’ve two objections to the foregoing. First, I disagree with the de-
scription given of scientific practice: One must distinguish between theo-
ries recognized by scientists to be only approximating and ones not so
understood. Approximating theories (ones idealizing away from one or
another aspect of the phenomena they’re applied to) can approximate in
different ways. Because of this, they can give different answers when ap-
plied to the same phenomenon. But this hardly shows a commitment to
inconsistent theories.
But, second, the considerations about blind truth ascriptions show
that even if scientists were committed to contradictory theories, one could
not blithely gloss this as a matter of the theories not being taken to be
true. That would be to mistake why the truth idiom is involved in scien-
tific practice to begin with. The right move—if a philosopher insists on
the collective inconsistency of scientific doctrine—is to focus on the logic
used in science, to argue that scientists are using one or another paracon-
sistent logic, for example, instead of a classical one.
As I’ve said, I prefer to leave the logic alone and take a closer look
at how the so-called competing theories are justified in applications, in
particular, to locate what more general theory they’re being culled from,
and in which ways they’re differing approximations of that theory. Of
course, I’m making an empirical claim here: It could turn out that scien-
tists really are committed to inconsistent theories in the way Hacking and
Cartwright suggest. But I doubt it. And further, I suspect that if this
could be shown, scientists would be really upset. After all, as I’ve already
mentioned, the apparent incompatibility of general relativity with quan-
tum mechanics already makes (many) physicists uncomfortable. It’s this
perceived incompatibility that is the primary factor driving so much of
the research toward a unified theory of “everything.”
Let me stress explicitly an implication of my last paragraph: Attribu-
tion of an instrumental attitude toward a scientific theory is an empirical
matter on my view, and one that doesn’t turn solely on what a scientist
or group of scientists happens to claim. Further, this means that attribu-
tions of instrumentality toward indispensable scientific theories aren’t
matters for philosophers to stipulate (on the grounds of one or another
philosophical program). Instead, an instrumental attitude toward such a
theory must manifest itself the way I illustrated earlier: (some) implica-
tions of that theory must be placed in quarantine.
Two last points: First, this take on how attributions of instrumentality
are to be recognized allows the empirical possibility of a science where
Circumventing Commitment to Truth 45
her 1997 (pp. 143–4), for example, she notes that many empirical theo-
ries are taken to be (literally) false.
I’ve already given a couple of examples of such, but they’re numer-
ous, and Maddy alludes to many more:
[W]e treat a section of the earth’s surface as flat, rather than curved, when
we compute trajectories; we assume the ocean to be infinitely deep when we
analyse the waves on its surface; we use continuous functions to represent
quantities like energy, charge, and angular momentum, which we know to
be quantized; we take liquids to be continuous substances in fluid dynamics,
despite atomic theory. (p. 143)
we are committed to everything our theory says ‘there is’. Perhaps the notion that all exis-
tence claims are on the same footing—the ‘univocality of “there is”’ as Quine calls it—is
inaccurate as a reflection of the function of scientific language” (p. 143); but her immedi-
ately following suggestion that one should perhaps consider pre-Perrin “there are atoms”
as fictional and post-Perrin “there are atoms” as literal indicates that, in any case, she sees
a strong analogy between how scientists deny (or reserve judgment on) the existence of
empirical entities, and similar instrumental attitudes toward mathematical objects. It’s this
purported analogy that I primarily want to challenge. See my 1997b, or better yet (because
the reader presumably has this book in hand), see chapters 3, 4, and 5. For purposes of the
immediate discussion, I’m focusing only on the suggestion that because scientists cogently
take indispensable empirical doctrine to be false, there is hope that something similar can
be done with indispensable mathematical doctrine.
26
See n. 25.
27
Maddy (1997, p. 153, n. 32) takes note of Resnik’s objection but describes herself
as “unconvinced” without explaining why.
Circumventing Commitment to Truth 47
1
This is nontrivial, and there may not be consensus on which regimentation of a body
of doctrine best reflects it. My objections to there being a definitive criterion for what a
discourse commits us to won’t turn on this issue.
2
Restricted, as we are, to first-order languages, the criterion may be indifferently
phrased deductively or implicationally. Of course, even if we use a model-theoretic notion of
implication, the criterion still doesn’t require an understanding or translation of nonlogical
vocabulary in order to apply it to (foreign) discourse.
That Quine’s criterion should be so glossed has been suggested before. Scheffler and
Chomsky 1958–59 deplore this version of Quine’s criterion on the grounds of its triviality
(they write: “Not a very exciting result” (p. 81))—an objection Quine is unlikely to have
recognized as an objection. They also claim (p. 82) that this version of the criterion serves no
philosophical purposes “that cannot as economically be served by reference to the theories
themselves.” Here I demur. As economically? That’s not obvious.
Cartwright 1954 considers several versions of the criterion but never quite suggests
this one, perhaps because he takes ontological commitment to be a relation between sen-
tences and collections of entities. Admittedly, Quine’s own gloss on his criterion (“to be is
to be the value of a variable”) makes his text seem hospitable to the relation view of his
criterion, but it’s better to see that slogan as indicating the nature of the connection be-
tween objectual quantifiers and the semantics they’re given in a metalanguage, especially
because relational views, as both Cartwright 1954 and Scheffler and Chomsky 1958–59
make clear, don’t even have a chance of working unless murky intensionalist maneuvers are
introduced.
I like Gottlieb’s (1980, chap. 2) defense of the version of the criterion I adopt, apart
from two reservations: It’s clear that Quine means his criterion to apply only to first-order
languages, where the quantifier is standardly objectual, and it’s also clear that he doesn’t
mean it to apply to individual sentences—except parasitically—but to whole discourses,
which presuppose the use of first-order logic.
Criteria for the Ontological Commitments of Discourse 51
3
I should add that my take on regimentation (with its recognition of, and partial
respect for, the presence of ontologizing in ordinary language) does not square well with
everything Quine says on the subject. E.g., he says (1981a): “Ontological concern is not a
correction of a lay thought and practice; it is foreign to the lay culture, though an outgrowth
of it” (p. 9). My exegetical problem here is that I don’t quite see how to get this more
extreme thought—even with careful attention to its context—to fit compatibly with Quine’s
subscription to his triviality thesis regarding ontological commitment. So my regimenter
has a more moderate view, one compatible with much of what Quine has written on regi-
mentation and on explication. See the discussion later of Quine’s triviality thesis.
52 Truth and Ontology
or “what’s in” are most naturally read as ontologically committing), E ranges over a “sub-
class of what’s in the domain.” Should this be described as the claim that “existence is a
special property that not everything shares”? No—this substantial way of putting the posi-
tion is misleading because (in this context) it makes it sound like there are things (which
subsist?) but that don’t exist; further these (subsisting?) things have properties—e.g., they
don’t exist. The best way to put the point (in this context) is from a perspective of semantic
ascent: Our (regimented) terminology allows a statement “(∃x)Px” to be true even if noth-
ing is a P. If, however, “(∃x)(Px & Ex)” is true, then something is a P. (Ps exist.) Does
this mean there are things that subsist but don’t exist in the domain of discourse? No—to
put the point this way is really misleading. I say more about this shortly.
7
Quine 1981c:
9
It’s striking that both Barcan-Marcus (1971) and Parsons (1980) assume that use of
objectual quantification convicts one of ontological commitments (to the domain of dis-
course). Parsons even assumes that his use of objectual quantifiers for nonexistent objects
commits him to nonexistent objects. Also see Smart 1999, pp. 109–10, and Davidson 1999,
p. 123, an exchange in which both sides virtually take it to be a truism that Tarskian satisfac-
tion involves commitment to entities; and Schiffer 1996 (pp. 151–2, 156), where it’s pre-
sumed that objectual quantification must bear ontological commitments (even if such com-
mitments can be to kinds of “deflated” objects such as properties and fictional beings). One
place I know of in the literature where it has been suggested that “objectual quantifiers”
can be taken to be ontologically neutral (besides my 1997a and 1998) is in Landesman
(1975). However, Routley (1980) may also claim this (I’ve not—as of this date—worked
my way through the entire book). See, however, the seven paragraphs preceding the section
“Some Important Illustrations.”
10
Why the scare quotes? I don’t want the reader to presume that a “domain of dis-
course” or a language being “interpreted” amounts to (by definition) its terms picking out
what exists. All that’s meant is that such a language isn’t a “meaningless formalism”: It has
a subject matter.
11
This enables the definition of satisfaction for two-place relations: “Rxy” is satisfied
by 〈o 1,o 2〉 if and only if R*o 1 o 2 (where “R*” is a metalanguage relation coextensive with
the object language relation “R”).
Criteria for the Ontological Commitments of Discourse 55
12
How are such coextensional predicates “provided”? Well, it’s asserted in the metalan-
guage that they have the same extensions as the object-language predicates they’re paired
with. What else is possible?
13
I.e., a theorem of set theory such as (s)(t)(∃y)(x)(x ε y ⇔ x = s ∨* x = t), and then
a definition of ordered pairs using the sets guaranteed by that theorem, e.g., 〈x, y〉 =def {{x},
{x, y}}. See any standard book on basic set theory, e.g., Levy 1979, pp. 20, 24.
“Existence claims . . . are there to provide them”: On the view I’m pushing, these
aren’t genuine existence claims. But mathematicians use such phrases as “existence proof ”
all the time. Are they misusing the word “exists”? Has the term been co-opted by them for
a technical use (e.g., “ring,” “field”)? See chapter 4, where I give an answer to this question.
14
The last six paragraphs explicate an argument from my 1997a, pp. 206–7, n. 19. It
appears again (also in a fairly encapsulated form) in my 1998, p. 3.
56 Truth and Ontology
15
Both Landesman and Routley officially draw their inspiration for this point from
Meinong. Nevertheless, two philosophers from such different (philosophical) backgrounds
claiming the same (weird-sounding) thing? Then it has to be addressed (however bizarre
one thinks it is).
16
Here are some other illustrative quotes from Routley (1980): “Meinong maintained
that although the number two does not exist it has being. On the new theory of number
two neither exists nor is assigned being of any sort; however it does have properties, it has
indeed a nature,” and a few lines later, “the number two has no being of any kind; even so
it is an object and can be talked about” (p. 4). Finally, we find that “‘[i]tem’ is introduced as
an ontologically neutral term: it is intended to carry no ontological, existential, or referential
commitment whatsoever” (p. 5).
17
Maybe, that is, ordinary uses of “exist” (and “subsist,” surely, because the latter is
largely philosophical language untouched by ordinary usage) don’t compel the claim that if
something has properties and is not nothing then that’s tantamount to claiming that it
exists in some sense. I think, ahem, there’s something to this. See chapter 5.
18
Compare my discussion below of Quine’s challenge regarding the individuation con-
ditions for what doesn’t exist with Routley’s to be found in his 1980, pp. 411–26. On my
view, the individuation conditions are simply made up (which is no surprise given that, on
Criteria for the Ontological Commitments of Discourse 57
The reader may fear that if things (e.g., Mickey Mouse) exist in no
sense at all (in my sense of “no sense at all”), then nothing true can be
said when “Mickey Mouse” appears as a noun phrase (as in, e.g., “Mickey
Mouse doesn’t exist,” or “Many people think Mickey Mouse is cute”).
This simply isn’t true, and the insight that it isn’t true is one that can be
gleaned from the substitutionalist view of quantification.
For the insight that substitutionalists have, when they deny that their
form of quantification need incur ontological debts, is that the sentences
“Mickey Mouse is a fictional mouse” and “Mickey Mouse is a fictional
dog,” for example, can be taken to be, respectively, true and false without
their having any truthmakers (to use currently fashionable language),
without there being any objects the properties of which determine the
truth-value status of those sentences. That is, the truth values of sentences
containing terms like “Mickey Mouse” can be determined on the basis of
factors other than the properties of an item, Mickey. And this is a very
good thing (on my view) because things that don’t exist—I have to say
it again—don’t have properties.
It’s also this insight, by the way, that I’m taking advantage of when
I claim that objectual quantification doesn’t necessarily betray ontological
commitments of any sort because objectual quantification is only a body
of semantic constraints of a certain sort couched in metalanguage sentences,
for we have to evaluate how those semantic claims themselves are made
true before we can convict the object-language quantifiers of carrying
ontological commitment.
my view, these things exist in no sense at all). Routley’s view is very different: Because such
objects have “determinate” natures, Routley is required to engage in metaphysical reasoning
to determine what these natures are.
58 Truth and Ontology
ics. On the regimentation view, these entities are required in our best
regimentation of the vernacular; on the scientific study of the logical form
of the vernacular view, these entities are ongoing (inadvertent) commit-
ments of ordinary speakers.
Two Examples: Davidson (1967, 1977) is very explicit that the analysis
of the logical form of adverbial inferences offers metaphysical insights.
Consider the inferences from
(1) Jones nicked his cheek while shaving with a razor in the bathroom
on Saturday,
to
(2) Jones nicked his cheek in the bathroom,
to
(3) Jones nicked his cheek.
Davidson (1977) writes:
It seems that some iterative device is at work. . . . The books on logic . . .
analyse these sentences to require relations with varying numbers of places
depending on the number of adverbial modifications, but this leads to the
unacceptable conclusion that there is an infinite basic vocabulary, and it fails
to explain the obvious inferences. By interpreting these sentences as being
about events, we can solve the problems. Then we can say that “Jones nicked
his cheek in the bathroom on Saturday” is true if and only if there exists an
event that is a nicking of his cheek by Jones, and that event took place in
the bathroom, and it took place on Saturday. (p. 212)
Davidson quickly concludes that an ontology of events is required: He
relies on the presupposition that the objectual quantifiers so introduced
to explain obvious inferences and to avoid an infinite basic vocabulary
bring ontological commitments in their train. I can’t now go into the
question of whether it’s plausible or not to instead treat events as fictional
items introduced for the reasons Davidson explicitly gives; this will turn
on other issues about events. My aim here is only to illustrate how ontol-
ogy is taken to ride on considerations of logical form alone, and to regis-
ter a complaint against the presupposition in this practice that first-order
objectual quantification automatically secures ontological commitments.
As a second example, Jackson (1976) argues that mental objects must
exist because there is no way to preserve the logical properties of state-
ments about mental states while eliminating the apparent relations of per-
sons to the objects in such states, where such relations seem to commit
us to both persons and these mental objects. So, for example, I may have
a painful itch and a (distinct) ticklish itch. Further, my painful itch may
burn, and the burning sensation of it may be intermittent. Further, the
burning sensation of that itch may be the same burning sensation as a
painful scrape on my knee. These ways of distinguishing and comparing
Criteria for the Ontological Commitments of Discourse 59
19
As these two examples already indicate, our ways of carving up subject matters often
require the entrance of objectual quantifiers onto the semantic stage in many areas, with
the resulting yield of items we’d otherwise be loathe to regard ourselves as ontologically
committed to. An especially clear case, which I describe later, is that of discourse about
fiction: a form of discourse that I argue involves talk of items we take to exist in no sense
whatsoever. It’s nice to think that the idioms that come up there, and that are best seen as
nonontologically committing uses of “there are,” in the vernacular can be nicely regimented
by a first-order quantifier that is similarly noncommitting.
20
See, e.g., Quine 1960, pp. 142, 167; 1950, pp. 75–6; 1953b; 1975, p. 102. In
raising these concerns, Quine often acknowledges Frege.
60 Truth and Ontology
21
Why not? Not for any metaphysically deep reason. In classical contexts, inconsistency
yields triviality—everything is provable—and the resulting family of algorithmic systems is
useless.
22
By “fictional entity,” here and hereafter, I mean those items in fiction that are en-
tirely fictional. Neither Bush, should one appear in fiction, is fictional in this sense.
Criteria for the Ontological Commitments of Discourse 61
23
Resnik 1997, pp. 209–12, when addressing the same problems, opts to treat mathe-
matical objects, although “mind independent” as “incomplete objects”: There is no fact of
the matter about certain identity claims that we can formulate about such objects.
24
Note: A standardly interpreted first-order language need not have identity among
its vocabulary items, and its domain can certainly have more than one equivalence relation
defined on it. In a case where the objects in that domain exist in no sense at all, there need
be no residual worries that there is some genuine notion of identity implicitly at work,
viz., that equivalence relation that most finely individuates the “objects” in the domain of
discourse.
62 Truth and Ontology
in this usage even though, as it’s well known, Tarski’s theory of truth
allows us to define a “theory of denotation.”25 For, given what I’ve said
about the ontological neutrality of objectual quantifiers, it follows that
such a theory of denotation can easily be produced for terms that refer
to nothing at all.
Nevertheless, when terms that refer to nothing at all occur in a lan-
guage with identity, and with consistent identity conditions that allow
statements like A = B to be true for distinct terms A and B, I’ll describe
such terms as co-referring* and will also, in general, speak of a term A as
referring* in this sense even should it (like “Mickey Mouse” or “1,” as I
eventually claim) actually refer to nothing at all. I understand referring*
and co-referring* as more general notions that include their narrower
cousins referring and co-referring.
It should now be clear that the mere presence of quantifiers in the
logical analysis of an idiom, even if they’re objectual ones, needn’t bring
ontological commitments in its train. Still untouched, of course, is the
Quinean claim (which is independent of the position that objectual first-
order quantifiers are ontologically committing by virtue of their semantics
alone) that ordinary uses of “there is” in the vernacular are understood
by speakers to carry ontological commitment.26 Notice that if this claim
could be shown to be true, it would yield the ontologically committing
nature of the first-order objectual quantifier even though the claim, that
objectual quantifiers are ontologically committing by virtue of their se-
mantics, fails. One problem, as I’ve mentioned, with this take on the
triviality thesis is that it puts a lot of weight on ontological commitment
as it’s indicated in the vernacular. However, given the failure of the at-
tempt to extract ontological commitments from the semantics of the
quantifiers, the Quinean has no choice.
I turn now to an evaluation of the claim that uses of “there is” in the
vernacular carry ontological commitment. Consider this ordinary sen-
tence:
(M) There are fictional mice that talk.
25
See, e.g., Field 1972.
26
An especially neat use of this latter claim (which is not connected to any claim about
formal quantifiers) may be found in Lewis 1986, pp. 8–17. The argument, briefly, is that
the use of “could” in many contexts in the vernacular isn’t served well by regimentation in
standard modal notation because its role seems to be the broadening of the scope of quanti-
fiers within its range so that possible objects fall under their domains of discourse. The
conclusion that “we’d better have other-worldly things to quantify over—” (p. 17) has the
ontological force Lewis wants (support for his modal realism) because of the already in
place assumption that vernacular use of quantifiers carries ontological commitment. Also
see Colyvan (2001, pp. 117–8), where his presupposition that the vernacular use of “there
is” carries ontological commitment induces him to the conclusion that statements like
“there is an even prime” or “there are differences between Clinton and Bush” can’t be
obviously true because it can’t be obvious that differences or numbers exist. (Notice that
Criteria for the Ontological Commitments of Discourse 63
this is a modus tollens on the Carnapian-flavored view that such (internal) statements are
ontologically insignificant, in part, because they are obvious.) Van Inwagen 2000, p. 239,
straightforwardly embraces the claim that the vernacular “there is” is ontologically commit-
ting.
27
Of course, philosophers are clever people, and no doubt other options will emerge.
(Indeed, I actually go on to consider seven. See the discussion of the fourth option.)
By the way, I’m offering options (except for one) that will support Quine’s triviality
thesis; in doing so, however, I’m (in many cases) not offering options that Quine could
stomach. But that’s OK—I think the triviality thesis false—and this is so even if non-Quin-
ean resources are mounted in its defense.
28
This is why the paraphrase proponent is not committed to any particular way of
producing paraphrases. See chapter 2, n. 18.
64 Truth and Ontology
30
This distinction between properties had and properties depicted may be only termi-
nologically distinct from van Inwagen’s (2000, pp. 245–6) separation of properties that
fictional characters have from those that they hold. Also see Salmon 1998, where a similar
distinction is made.
31
Barcan-Marcus (1971) argues for the usefulness of substitutional quantification to
regiment mythological and fictional discourse without incurring ontological commitments.
66 Truth and Ontology
mitting, and this is the use that supports the triviality thesis. The other
use isn’t. The two uses are distinguished (in the vernacular) because we
recognize that a noncommitting use must be inferentially connected to
an instantiation with respect to a name (the way that (M) and (O) are so
connected), which is referentially empty.
32
The view, if it’s to support Quine’s triviality thesis, requires distinguishability in the
vernacular between metaphorical and literal uses. Yablo (1998; 2000, pp. 304–5) argues
that such isn’t to be had, and he draws the conclusion that Quine’s criterion for what a
discourse commits us to thus fails because it’s understood to apply only to literal uses of
words. Although I’m sympathetic to attacks on Quine’s criterion, I don’t like this particular
way of going about it. To some extent, Yablo’s argument is ad hominem: It attacks that
textually relaxed and easy-going Quine who seems to allow that certain usages are meta-
phorical even without the availability of paraphrase. I interpret Quine more severely on this
point: Without paraphrase, there can be no claim of metaphorical usage that allows us to
repudiate commitments to what’s metaphorically referred to. This, by the way, makes Quine’s
particular brand of physicalism programmatic: Until paraphrases are available, if I must refer
to cities or angry clouds to express what I want, then I’m committed to cities and angry
clouds. Period.
We can weaken Quine’s constraint, as I indicated in chapter 2, but not in a way that
supports Yablo’s attack on Quine’s criterion: Sentences we take literally, and believe, are
ones all of whose implications are part of our body of beliefs. Sentences understood meta-
phorically are ones not all of whose literal implications are part of our body of beliefs. This
provides a way of distinguishing metaphorical from literal usage of sentences, even without
a full explication of what that metaphoricality consists in.
Criteria for the Ontological Commitments of Discourse 67
Any of options (1) − (5) can be used to support the triviality thesis. So
let’s take them in turn.
sider the first paraphrase, “Some mice that talk don’t exist.” The problem
with this item is that the locution “some mice” as used here seems to
come to the same thing as “there are mice,” so “Some mice that talk
don’t exist” seems to come to the same thing as “There are mice that
talk but that don’t exist.” But then, arguably, we’ve still got a use of the
locution “there are,” which isn’t ontologically committing: “There is”
isn’t ontologically committing in contrast to the predicate “exists” that
does carry commitment (at least in this sentence).
Still, the paraphrase proponent can try to avail himself of a legality at
this point: The issue is whether “There are Ms” is ontologically commit-
ting in the vernacular; it’s not whether “Some Ms” are ontologically com-
mitting. So the fact that we can paraphrase away “There are Ms” this way
tells us that “There are Ms” is ambiguous and that the paraphrase propo-
nent need only attach the success of the triviality thesis to one of those
meanings. We have (1) an ontologically committing use and (2) an onto-
logically noncommitting use. The latter is recognized because it comes to
the same thing as “Some Ms.” Unfortunately, this ploy doesn’t succeed in
cleanly separating two uses of “there is” in the vernacular because “Some
Ms” can be substituted for “There are Ms” equally well regardless of
whether “There are Ms” is taken to be ontologically committing or not.
So let’s now consider the paraphrase of (M) that characterizes it as
“There are no mice that talk.” This option works perfectly well when
what’s being paraphrased in the vernacular is a simple denial of the exis-
tence of something. But I should make it perfectly clear that it’s unac-
ceptable if what’s involved is a claim about what goes on in fictional
discourses of various sorts. In that case, there are not, presumably, fic-
tional mice willing and able to do anything at all. There are no fictional
mice, for example, that eat the planet Neptune in one gulp.35 Also, there
are many things fictional mice do that real mice also do (eat cheese, for
example). But this paraphrase makes each statement about what fictional
mice do true if real mice don’t do what the statement has fictional mice
doing, and false otherwise.
It may be thought that paraphrasing via depictions does better be-
cause it makes the implicit reference to a fictional medium explicit:
“There are depictions of mice that talk” is true; “there are depictions of
mice that eat the planet Neptune in one gulp” is false—just as it should
be. But consider this sentence: “There are fictional detectives who are
admired by many real detectives.” Paraphrasing this via depictions can
give sentences with the wrong truth values: “There are depictions of de-
tectives that are admired by many real detectives” may be true, but one
is perfectly capable of admiring the fictional detective depicted without
35
I hope this is right: I hope no novel (play, cartoon, or whatever) exists in which a
mouse eats the planet Neptune in one gulp. But if so, substitute some more outlandish
example, or use a simple contradiction: “There are fictional mice that aren’t fictional mice,”
or perhaps “There are fictional mice that aren’t depicted as mice.”
Criteria for the Ontological Commitments of Discourse 69
admiring his depiction (e.g., one may admire Sherlock Holmes but utterly
deplore Doyle’s prose style, or vice versa). This problem isn’t easily sur-
mountable by paraphrase: We think about, admire, and react to characters
in fiction, mythology, and so on—this is how we talk, anyhow; and we
do so even when entirely aware that such things don’t exist in any sense
at all. Nonetheless, these thoughts, admirations, and reactions aren’t—or
at least aren’t naturally—construed as thoughts, admirations, and reac-
tions to depictions or text or anything like that. This is what makes these
paraphrases poor options.36
Some may want to fault this sort of counterexample by noting that
it’s an opaque context, and thus will counsel banning quantification into
such a context. The idea is that the logic of propositional attitude terms
is tangled enough that paraphrase proponents—even though they’re re-
stricted to the vernacular for their paraphrases—can ignore such pur-
ported counterexamples. This line, I think, is mistaken, not because prop-
ositional attitude contexts don’t pose formidable logical problems, but
because those problems have nothing to do with the issues at hand.
The real problem is that the depiction option seals off the content of
what’s depicted from quantification, and so anaphora becomes impossible
even when the context isn’t a propositional attitude one or, more gener-
ally, an opaque context. Consider this example:
There is a fictional mouse who is (sometimes) depicted as having a
round friendly face (and big eyes), and who is (sometimes) depicted as
looking quite unpleasantly ratlike (with small eyes).
36
Walton (1990, p. 203) denies this, and there has been quite a bit of discussion about
whether he’s right or wrong in so doing. I think he’s wrong, but that needn’t detain us
now because, in any case, the paraphrase proponent won’t find this strategy of much use.
For suppose there is a kind of special “pretending*” that we engage in when we “have”
emotions toward characters we know to be fictional: if we replace “There are fictional detec-
tives who are admired by many real detectives” with “There are fictional detectives who are
pretend*-admired by many real detectives,” we still have the same paraphrase problem.
Walton’s ultimate solution—which I discuss later—is more drastic than anything the para-
phrase proponent will want—it requires extending the pretense to the proposition itself:
“There are fictional detectives who are admired by many real detectives” doesn’t really
express a proposition; we only pretend that it does.
70 Truth and Ontology
37
I.e., when fictional items are being discussed, and two distinct terms A and B are
co-referential*, then a depiction is a depiction of A if and only if it is a depiction of B.
Criteria for the Ontological Commitments of Discourse 71
38
A sensible ordinary person will find it bizarre if you claim that fictional objects exist,
either in the “strict” Meinongian sense or in the sense of being abstracta. Indeed, many
sensible ordinary people will find it bizarre if you claim that abstracta exist—they will, any-
way, once you’ve made it perfectly clear what you really mean by this.
39
Salmon’s motivation is clear: True (and false) sentences with nonreferring expres-
sions pose a serious prima facie problem for his strong Millian view, “according to which
the semantic contents of certain simple singular terms, including at least ordinary proper
names and demonstratives, are simply their referents, so that a sentence containing a non-
vacuous proper name expresses a singular proposition, in which the name’s bearer occurs
directly as a constituent,” and even a (prima facie) problem for the “less committal” “theory
of direct reference, according to which the semantic content of a name or demonstrative is
not given by any definite description” (Salmon 1998, p. 278).
Nevertheless, a general theme of this book is that if a theory needs quantification over
“entities” of a certain sort, it doesn’t follow that such things exist. Although I can’t get
into details about proper names in this book, the moral is the same one for semantic theories
as it is for any other scientific theory. I say a little more about this shortly; see esp. n. 55.
40
I’m surprised not to have found this argument in the literature already, because
many characters in fiction are nameless, e.g., most of the soldiers who fight in the Trojan
War in The Illiad, or most of those who live and die in the background of War and Peace.
Criteria for the Ontological Commitments of Discourse 73
It’s understood that there are many more hobbits, dwarves, and elves in Middle Earth than
those that literally appear on the pages of The Lord of the Rings. By the way, depicting a
character as having a name isn’t the same thing as the discourse actually giving the character
a name. (We can presume that all the characters who are understood to die in War and
Peace are understood to have names, too.) Substitutional semantics for discourse about
fiction can’t manage with mere depictions of characters as having names. There are desper-
ate maneuvers still available, of course. One can understand fictional discourse to be far
more bloated than it appears to be (containing names for every character alluded to—even
though neither author nor readers have any idea what these could be).
41
Recall that this is the suggestion that there are metaphorical or pretend uses of
“there is” in the vernacular that aren’t ontologically committing.
42
Yablo (2000) argues that “there is” and “exists” are metaphorically used in a wide
range of cases. The contrast argument just given generalizes to these examples in a way that
makes this claim extremely dubious, unless what’s meant is that these locutions are always
used metaphorically. I rule this latter suggestion out in what immediately follows.
Jeff McConnell has pointed out to me that a synonymous version of (R) is “There are
fictional mice that talk, but no real ones that do.” Thus, the two appearances of “there are”
in (R) cannot differ in literality.
43
Despite my label “metaphor/pretend option,” I do not mean to suggest that the
claim that “there is” is sometimes metaphorical is identical to the claim that we sometimes
“pretend” to utter propositions with “there is” in them. Walton (2000) carefully distin-
74 Truth and Ontology
“Mickey Mouse is a talking mouse,” it’s not that any particular word in
that sentence is being used metaphorically; rather, it’s that the presence
of a fictional name reveals that the speaker is only pretending to refer (by
means of “Mickey Mouse”) and that (really) no proposition has been
uttered at all, but only fictionally so, or in pretense. Fictional names are
ones that speakers can never use in ordinary propositions but can only
pretend to refer with.44 Not only do we pretend to refer when we use
fictional names, but also (Walton 1990, pp. 428–9) we pretend to attri-
bute properties to things when we use predicates such as “mythical,”
“illusory,” “imaginary,” “fake,” “real,” “genuine,” and so on. Thus, in
uttering (R), we only pretend to mark a contrast between fictional mice
and real ones.
I find this view too drastic in two respects. The first is that it generally
requires discourse about fictional objects to be meaningless. That is, not
only is fictional discourse itself meaningless on this view, but discourse
on the part of literary critics about characters in fiction is meaningless, as
well.45 However, just as it’s a grave drawback for a theory to commit us
to fictional objects (because fictional objects are paradigmatic of what
doesn’t exist in any sense at all), it’s also a grave drawback to require
palpably meaningful discourse to be meaningless. On both these points,
the paraphrase option does much better: It treats fictional discourse as
meaningful and fictional objects as nonexistent (in every sense).
A second drawback of this view is that speakers are taken to be pre-
tending to refer, and as a result, to be pretending to be uttering proposi-
tions, even if speakers (sincerely) disavow themselves as engaged in pre-
tense. The problem isn’t that speakers can’t often sincerely disavow true
claims about themselves: Speakers are notoriously ignorant about their
own motivations, for example. But in this case an ordinary psychological
state, pretending, has been drafted as a device that allows us to avoid the
difficult task of analyzing the logical forms of sentences in such a way that
their logical properties are preserved, in particular, what implications they
seem to have and what truth values speakers attribute to them.46 And this
guishes his pretend-doctrine from the claim that, e.g., “exists” is sometimes used metaphori-
cally. The state of play is this: I regard the metaphoricality claim as refuted by the immedi-
ately preceding remarks on (R), and I’m allowing the proponent of the metaphor/pretend
option to adopt a more extreme position toward (R), viz., some version or other of Wal-
ton’s claim that (R) isn’t (really) a proposition at all.
44
Salmon (1998) attributes the view to Kripke that “a sentence using a proper name
in an ordinary context (not within quotation marks, etc.) expresses a proposition only if the
name refers” (p. 312, n. 19). Donnellan (1974) writes: “If a child says, ‘Santa Claus will
come tonight,’ he cannot have spoken the truth, although, for various reasons, I think it
better to say that he has not even expressed a proposition” (pp. 20–1).
45
No doubt a great deal of literary criticism is meaningless. But surely it doesn’t have
to be.
46
See Walton 1990, pp. 416–9. Walton does offer paraphrases, but such paraphrases
“seek to capture . . . what speakers say in uttering the sentences cited, not what the sen-
tences themselves mean or what propositions they express, if any” (p. 417). They are thus
not required to respect the logical form of the sentences they paraphrase, or even to be
sentences speakers would recognize themselves as having said.
Criteria for the Ontological Commitments of Discourse 75
raises the most important objection to this approach, that it has insulated
itself against the only data available to discriminate among alternative the-
ories of this sort. If the claims of speakers to have asserted meaningful
statements in a sincere way (without intent of pretense or play) can be
denied for purely theoretical reasons, in what sense can the resulting the-
ory be regarded as sensitive to evidence?47
A Digression
Semantic pretense views on large chunks of our discourse—fiction, dis-
course about fiction, mathematics, propositional attitude ascriptions, even
truth attributions—are becoming fashionable, and I’ve detected gestures
toward the previous objections, in one form or another, at conferences
and in the literature. I find them unanswerable (at least insofar as they
provide evaluative comparisons with the other candidates I consider),
however, and so I’ve given them in my own way for the record. I should
add that I like the view that discourse in fiction itself, in novels, short
stories, and so on, involves pretense, provided the view is taken that the
statements of fiction are pretended-true rather than the view being that a
pretense is made of asserting statements themselves.48 But I don’t want to
extend this view to the sorts of claims about fictional characters I’ve con-
sidered above (e.g., (M))—because these are statements not pretended-
true but actually true; and for similar reasons, I’m loathe to extend the
pretend-true view to mathematics or to truth attributions themselves.
This point is worth expanding on.49 The literature on pretense has
not, to my knowledge, engaged itself explicitly with the issue of the
Quine-Putnam indispensability thesis—but it has to if its program is to
work. The problem is that statements we’re pretending to utter, on this
view, must be ones that can be cleanly separated from ones that we’re
not pretending to utter, on pain of global pretense: that our entire body
of utterances collapses into pretend utterances.
Certainly, when it comes to the pretense take on abstracta or on
truth, this problem must be faced, because what fuels the indispensability
argument right from the beginning is that (in the important cases) an
47
A way of seeing the point about evidence is to ask, “Where will this all stop?” Why
not, e.g., claim that we have no ontological commitments at all? No matter what we say,
we’re always pretending, even though we don’t think we are and our statements seem to
have logical forms that we respect when we draw inferences from them. Call this the threat
of global pretense, and see the digression immediately below.
48
This is similar to Urmson’s (1976) view: “[M]aking up fiction is not a case of
stating, or asserting, or propounding a proposition and includes no acts such as referring”
(p. 155). On the view I like, however, “pretending-true” does not exclude statements in
fiction—some of them, anyway—from being actually true, and it does not exclude certain
names, “Nixon,” for example, from referring.
49
The pretend view, apart from its application to sentences in fiction itself, is open to
the same objections I raised against the instrumentalist strategy (B3) in chapter 2, simply
because it is an instrumentalist strategy, as I indicated there. Nevertheless, some additional
remarks are in order.
76 Truth and Ontology
much in the way of real confirmational impact): Early evolutionary theory required the sun
to be much older than preatomic physics seemed to allow the possibility for. As we all know,
evolutionary theory was vindicated, not preatomic physics.
52
This, recall, is the view that terms like “fictional” cancel, either pragmatically or
semantically, the ontological force of “there is.”
53
But does it? Consider the following sentences: “Although unicorns are fictional crea-
tures, this doesn’t mean they don’t exist. Of course they do: They just don’t exist in space
and time—they’re abstracta.” I don’t agree with this view, and I’ve argued against it. But
I don’t think I can argue against it on the grounds that this sentence is contradictory by
virtue of the meaning of the word “fictional.” However, just this claim must be made by
proponents of this version of the cancellation option.
78 Truth and Ontology
Here’s what I’ve tried to do: I’ve not (in every case) attempted to show
that the alternatives to the ontologically neutral anaphora option or even
a combination of such can be definitively refuted. Rather, I’ve tried to
show that the ontologically neutral anaphora option is far more likely
than the other options. That is to say—and let me stress this—the claim
I’ve attempted to establish is that the best option among those I’ve con-
sidered is that “there is,” as used in the vernacular, isn’t ontologically com-
mitting. Some might be inclined to draw a weaker conclusion: “There is”
sometimes isn’t ontologically committing. I draw the stronger conclusion
on the grounds that uses of “there is” are not ambiguous (recall the
discussion of the metaphor/pretend option).
Some further remarks, however, should probably be made in support
of the ontologically neutral anaphora option. Some may be concerned, in
particular, that this option isn’t compatible with views that take the role
of (genuine) singular terms to at least have to include the contribution,
of what’s referred to, to the proposition that such singular terms occur
in. As I said in the second paragraph of footnote 39, I really can’t get into
this now because it will involve a substantial discussion of the semantics of
proper names, and our concern here is only with Quine’s triviality thesis.
However, let me say this much: Just as in the case of objectual quantifiers,
where a separation between the semantics of objectual quantifiers and the
ontological force such quantifiers carry is possible, one can adopt direct
reference* semantics for proper names while stripping “what’s referred
to” of ontological significance. That is, one can take the role of a proper
name to supply to a proposition only what it refers* to, and yet that
reference can be drawn from a domain (given in the metalanguage) that
54
I owe this suggestion to Mike Resnik.
Criteria for the Ontological Commitments of Discourse 79
contains items that exist in no sense at all. Of course, a story must be told
about how the sentences in which such a proper name appears are as-
signed truth values (a causal story, for example—which traces names back
to items the properties of which determine the truth values of such sen-
tences—isn’t available). But this isn’t as hard as it looks.55
Recapitulation
I detect (in Quine’s text) two possible arguments for taking the first-
order objectual existential quantifier to carry ontological commitment
rather than, say, an existence predicate doing so. The first is that the
quantifier does so by virtue of its semantics alone. The second is that this
quantifier is a regimentation of the ordinary language “there is,” and the
latter (in the vernacular) already carries ontological commitment. The
first argument is dispatched fairly quickly. The second argument requires
a lengthy tour of possibilities because there are methods of finessing pur-
ported examples of vernacular uses of “there is” that don’t carry ontologi-
cal commitments. The upshot, I conclude, is that, despite appearances,
Quine has no argument—at least no argument deserving the name “triv-
ial”—for his taking the first-order objectual existential quantifier to be
the sole conveyer of ontological commitment (in regimented languages).
Concluding Remarks
The mere falsity of Quine’s triviality thesis doesn’t refute his criterion;
what’s been shown is only that his criterion must be compared with alter-
native criteria, for example, that what a regimented discourse commits us
to should be evaluated in terms of a special existence predicate minted
precisely to carry commitment. (This last suggestion is the one that I’ll
eventually conclude—in chapter 4—is our best choice.)
This much I can say now: It may be argued against Quine’s criterion
that our discourse includes discourse about fiction, and that the foregoing
55
It’s not as hard as it looks because, in the case of nonreferring names, co-reference*
between names in the object language and names in the metalanguage can be stipulated,
and the truth or falsity of sentences involved can be a matter of the truth values we take
them to have. This means, in particular, that names function (semantically and grammati-
cally) the same way regardless of whether or not there’s an object that they refer to. That
proper names function in no semantically special way in pure mathematics is one of the
major burdens of my 1994, where the view presupposed is that the semantics of mathemati-
cal language is the same as that for the rest of our language, and yet, our understanding of
mathematical practice is distorted if we try to supply any content to the idea that some sort
of mathematical object is anywhere involved in that practice. That the same must be true of
discourse about fictional objects I simply take to be an indisputable datum. So as far as the
semantics of proper names is concerned, the foregoing is compatible with the claim that
names contribute their referents* to the propositions they appear in (and where their refer-
ents* are described in the metalanguage)—and this is so whether “what they refer to” exists
in any sense at all. Recall the slogan about semantics and ontology given earlier: It applies
to the semantics of every locution, not just to the semantics of the quantifiers.
80 Truth and Ontology
has shown that any discourse concerned (in part) with a fictional subject
matter can’t be regimented by a first-order language where objectual
quantifiers bear ontological commitment. This is because such regimenta-
tion requires options that have been shown to be inadequate.
It may also be argued that the indispensability of mathematical doc-
trine to science shows that introducing an existence predicate into regi-
mented discourse to indicate ontological commitment is better than hav-
ing objectual quantifiers carry commitment, because otherwise we find
ourselves committed to the existence of mathematical objects, and we can
exclude these undesirables by crafting an existence predicate that mathe-
matical abstracta don’t fall under.
These last two suggestions, however, which pit Quine’s criterion
against alternatives, raise new considerations, because both moves turn
on an antecedent desire to avoid commitments, respectively, to fictional
objects and to abstracta. It seems, that is, that adjudication among criteria
for what a discourse is committed to requires antecedent criteria for what
exists. This requires an extended discussion of its own, to which I now
turn.
4
1
All parties to the debate recognize this, and the standard example of a case where
linguistic convenience leads to a commitment to numbers is “The number of As is the same
as the number of Bs,” where, say, the number of As and Bs is indeterminately large. See
chapter 3, n. 50.
81
82 Truth and Ontology
2
See my 1998, p. 2.
Criteria for What Exists 83
fered isn’t a criterion for what exists; rather, an epistemic claim is being
made about what sorts of things we can have knowledge of, or about
what sorts of things we’re justified in claiming knowledge of.3
2. Second example: A philosopher may want to argue that the only
objects we should ever be committed to are ones that in some sense we’ve
causal access to. And the argument may be that, in fact, it isn’t that,
metaphysically speaking, anything that exists must have causal powers (of
some sort)—how on earth would we ever show that?—rather, the claim
is an epistemic one: Any thing we’re in a position to know something
about must be something we can causally interact with. Again, what’s
being offered isn’t a criterion for what exists; rather, it’s an epistemic
claim about what sorts of things we can take ourselves as justified in claim-
ing knowledge of.4
3. Third example. A philosopher may want to argue that the only
objects we should ever be committed to are concreta or items in space
and time. And the argument may be that, in fact, it isn’t that, metaphysi-
cally speaking, anything that exists is concrete, that abstracta are, meta-
physically speaking, the sorts of things that don’t exist; rather, the claim
is an epistemic one: Any thing we can maneuver ourselves into a position
to know something about must be something that we’ve managed to
causally interact with, or something to which our knowledge claims are
reliably connected. And the only candidates for which that is possible are
concreta or items in space and time.5
Contrast these sorts of arguments with the one about fiction that I
closed chapter 3 with: It was suggested that fictional objects don’t exist,
not because such objects are ones that it’s hard to find things out about;
rather, it’s all too easy to find things out about fictional objects because
authors simply make the facts up. It seems, therefore, that the denial of the
existence of fictional objects turns on a metaphysical claim, something to
the effect of, What’s totally made up doesn’t exist.6 That is, a necessary con-
dition has been placed on anything that exists: it can’t be totally made up.
3
Some such view as roughly described may be attributed to van Fraassen (1980).
4
Some such position as roughly described may be attributed to Cartwright (1983),
and perhaps also to Hacking.
5
Some such roughly described position may be attributed to certain nominalistically
inclined philosophers of mathematics, e.g., Field (1980).
6
Fictional objects aren’t like artifacts, sculptures, for example. There is something—
call it stuff—that sculptures are made out of, and this is true of artifacts in general. Fictional
objects aren’t made out of anything. (It’s confused to think fictional objects are made up of,
say, words. It’s similarly confused—but perhaps a little harder to see why—to think that
movie characters are made out of, say, celluloid, or that painted characters are made out of
paint. Of course, what a sculpture depicts needn’t exist any more than fictional objects do.)
84 Truth and Ontology
of purported objects. And so, on one construal of examples (1) − (3), the
central targets are truths about such objects: We can’t take ourselves to
think that certain sentences (about nonobservables, about causally inac-
cessible items, about abstracta) are true.7 As I showed in chapters 1 and
2, however, this position is untenable, because regardless of how certain
statements are established, unless they’re shown to be dispensable to sci-
entific practice, or to have implications that are quarantined, we’re com-
mitted to their truth (and no matter how much this grates on the episte-
mic sensitivities of some philosophers).
We can, however, reconstrue examples (1) − (3) as making claims not
about what sentences should be taken as true or not on the basis of our
epistemic practices but, rather, as making claims about what entities we
should take ourselves to be committed to. Examples (1) − (3) may be
described, that is, as trying to mark out the instrumental character of
certain sentences in our conceptual scheme, not by identifying them as
false or as truth-valueless or as statements the truth-values we’re ignorant
of, but as containing terms that refer to objects that don’t exist (refer to
nothing at all).
Here’s how to so reconstrue example (1): Use a predicate, “observ-
able” (“Ob”), and then argue that a sentence (∃x)Sx is purely instrumen-
tal (even if true) if (x)(Sx → ¬Obx) holds. Argue further that we commit
ourselves to a class of items S only if the sentence (∃x)(Sx) & (x)(Sx →
Obx) is an implication of our discourse.8
It may seem that substantial debates in epistemology are in the offing.
For example, one may argue that scientists do take observation of some-
thing to be the only appropriate grounds for regarding that something as
(provisionally) existing: They’re not willing to be committed to a kind of
entity unless they can observe it (in some sense), and thus, even though
scientists take, for example, mathematical claims or claims about theoreti-
7
This is how I understand the positions, respectively, of van Fraassen and of Cart-
wright and Hacking. Field’s program is similarly motivated: Show that applied mathematical
sentences are dispensable instrumental tools, and so, that mathematical predicates can be
taken to hold of nothing, that mathematical constants can be taken to refer to nothing (and
thus, mathematical sentences can be assigned truth values according to this null interpreta-
tion of their predicates and constants).
8
I.e., only in this way are we committed to every S. (∃x)(Sx & Obx) merely commits
us to some Ss existing.
There seem to be two possible construals of the ontological interpretation of Ob: On
the first, it’s not acting—strictly speaking—as an “existence” predicate but rather as an
“epistemically licensed to regard as existing” predicate; on the second, it is an “existence”
predicate, but (of course) like every predicate, (1) we can be wrong about what that predi-
cate (actually) applies to, and further (in the case of predicates we invest with ontological
significance), (2) we can be wrong about which predicate we should treat as ontologically
committing. (Perhaps some of what we observe doesn’t exist, or perhaps many things we
don’t observe do exist.) For a philosopher who thinks that, anyway, we can be wrong—
pretty much—about anything, these two construals seem to come to the same thing. De-
spite this, I give reasons at the end of this chapter for why the “epistemically licensed to
regard as existing” option for any candidate existence predicate is not a good choice.
Criteria for What Exists 85
9
See my 1997b.
86 Truth and Ontology
10
The second claim about how to distinguish epistemically significant differences be-
tween our transactions with abstracta and our transactions with concreta is open to debate.
See chapter 5. Because the move fails in any case, I’m granting (temporarily) that the dis-
tinction can be made out in one or another of these ways.
11
Causal access to entities nicely generalizes from sensory access to objects to include
instrumental manipulation of items out of our sensory range. So, too, observation general-
izes to what I call “thick epistemic access.” See my 1997a, 1997b, and chapter 6.
Criteria for What Exists 87
It’s important not to beg questions. What is it about the fact that
toads are interacted with (and numbers aren’t) that makes the number 7
(say) inessential to the epistemic process that established (A), and toads
essential to it? The thought is this: If, for all we know, there are no num-
bers, it doesn’t change anything about the events in the garden that es-
tablished the truth of sentence (A).
This last point, it must be said, amounts only to the truism that num-
bers neither participate in causal processes nor are the sorts of things that
one can observe (this is why the possible nonexistence of numbers has no
impact on those processes). But it’s hard to see how any of this shows
what’s needed. For why can’t this entire line of thought be challenged
like so: Why is it a requirement, on the sorts of objects we can be justified
in knowing exist on the basis of the truth of a discourse, that only those
items that participate in the evidential process itself (either by being ob-
served or by causal involvement) are candidates? If numbers exist, and
have properties knowable by the methods we use to establish truths about
numbers, why isn’t it just irrelevant that they don’t participate in any way
in the epistemic processes that established those truths?
Well, one might say plaintively, we’re forced to believe in cats and
dogs, and we’re not forced to believe in numbers by the epistemic process
of establishing the truth of (A). But again, so what? And anyway, why
can’t it be said that we’re forced to believe in numbers by our belief in
the truth of sentence (A)?12
There seems no way around it. We’re forced to an excursion into
metaphysics—specifically, into a study of necessary and sufficient condi-
tions for what exists—because only then can we raise (and perhaps an-
swer) the question of whether things that exist can be the sorts of things
that we can learn about in these various (epistemically lightweight) ways.
Let’s, therefore, consider various candidate criteria (necessary and suffi-
cient conditions) for what exists and address the question of how to adju-
dicate among them.
Before doing this, we must guard ourselves against potential fallacies
due to the trickiness of reasoning about nothing at all. Consider, for a
moment, the nature of necessary and sufficient conditions. Given, as I’ve
argued, that things that don’t exist don’t have any properties, it seems to
follow pretty cleanly that if a condition A is necessary for the existence of
something (if, i.e., anything that exists must have the property A), it
12
There are echoes here of an earlier debate in epistemology, for the above is just
another example of the common temptation to restrict knowledge claims to the purported
total evidence for such claims. So, e.g., if we think that all our evidence for anything consists
of observations, we can be tempted to think that we’re only justified in knowledge claims
about the existence of observable entities.
I should add that during the course of this chapter I’ll eventually settle on the position
that numbers exist in no sense at all, a position that someone plying this verificationist-
flavored argument will no doubt find appealing. But, as the above indicates, I can’t accept
a verificationist route to that position and must employ very different arguments.
88 Truth and Ontology
13
Wittgenstein’s (1922) dictum, 5.5301, on this is even more harsh: “Roughly speak-
ing, to say of two things that they are identical is nonsense, and to say of one thing that it
is identical with itself is to say nothing at all.”
Criteria for What Exists 89
true, but not because it describes (correctly) that something has a certain
property.
However, when speaking, metaphysically, of criteria (necessary and
sufficient conditions) for what exists, a (benign) split arises between how
these conditions apply to what exists and how these conditions arise as
applied to our sentences, and this split is similar to what happens with
identity. As far as (real) objects are concerned, what there is is only what
exists, and because there isn’t anything else, and so, nothing else that has
properties, talk of necessary and sufficient conditions on what exists seems
strangely trivial (any property is sufficient for existence if it’s instantiated,
otherwise not; any property that is necessary for what exists is sufficient
for what exists).
But given a language with nondenoting noun phrases—noun phrases
that pick out nothing at all—and given that predicates can apply to such
terms just as they apply to any term, and yield truths, necessary and suffi-
cient conditions may exist for the existence predicate, that is, conditions
of the sort (x)(Ax ⇔ Ex), where “A” picks out some condition or other
(“causally efficacious,” “located in space and time,” etc.) and “E” is the
existence predicate. The language allows the cogency of such a condition
“A” being co-extensive*14 with the existence predicate, even though,
strictly speaking, it’s only terms that denote nothing that the co-exten-
sionality* of “A” with “E” excludes from yielding truths when appended
to either “E” or “A,” and not the exclusion of the application of “A” to
ghostly nonexistent objects; further, this can be a significant claim to
make because “A” may be distinguished from other predicates “B” and
“C,” which apply to nondenoting noun phrases to yield truths, and thus
can nontrivially fail to be necessary for existence, sufficient for existence,
or both.
This take on the cogency of necessary and sufficient conditions for
existence raises, however, what I’ll call Kant’s challenge:15 that any list of
properties can apply to something that doesn’t exist, and so no property
can be co-extensive* with “exists,” when construed ontologically.
I don’t understand “Kant’s challenge” as a metaphysical claim about
what doesn’t exist, and how properties can be attributed to such things;
rather, it’s a claim about how the predicates in our language operate, so
that it amounts to the claim that, in fact, there is no predicate or set of
predicates A for which the assertion (x)(Ax ⇔ Ex) can be made. The
reason for this, if it’s true, lies in how we allow ourselves to assert truths
14
Recall the notions of co-extensionality* and reference* from chapter 3. What’s perti-
nent to the relationship between a predicate A and the existence predicate E, if the former
is to function as a criterion for the latter, is of course not just that they’re both co-exten-
sional (on (all) the same (real) items), but that they both fail to yield truths when applied
to all nondenoting noun phrases, i.e., that they’re also co-extensional*.
15
See chap. III, § 4 of Kant’s (1787) Transcendental Dialectic, pp. 500–7, esp. the
cute bit on p. 505 about the thalers. Of course I’m not claiming that Kant’s challenge, as I
construe it, is what Kant had in mind. Also see Hume 1739, bk. I, pt. II, § VI.
90 Truth and Ontology
involving nondenoting noun phrases: The claim is that for any predicate
“A,” which is a candidate for co-extension* with “E,” there is some non-
denoting noun phrase n, where An is true.
This claim may look plausible: Consider any predicate you like. Surely
a piece of fiction is possible in which that predicate is attributed to some-
thing. Consider the list of potential criteria for what exists that I offered
at the beginning of this chapter: “Observable,” “causally efficacious,” “in
space and time,” all surely apply to Sherlock Holmes. He lived in Lon-
don, didn’t he; and he could be seen on many days by Dr. Watson, isn’t
that true, too? Even “not ontologically dependent on linguistic or psy-
chological processes in certain ways” seems true of him—doesn’t it?—
because he wasn’t a hallucination of the other characters, something not
true of everyone in fiction.16
But this last suggestion gives the game away. I chose my examples of
truths about fictional objects in chapter 3 carefully. I’m willing to accept
as true the claim that “Sherlock Holmes is admired by many real detec-
tives”; I’m not willing to accept as true the claim that “Sherlock Holmes
lived in London in the late nineteenth century.” We must distinguish true
claims about fictional objects from those only pretended true. What takes
place in fiction (usually) is pretended true, no doubt about that. This
doesn’t force every claim about fictional objects to be only pretended true
(as I argued in chapter 3), but this is enough to meet Kant’s challenge.
It’s simply not true that fictional beings or other denizens of the nonexis-
tent, such as hallucination beings, dream beings, and so on, are causally
efficacious or not ontologically independent of us in the appropriate re-
spects.17
That abstracta, the other class of items that I’ll argue exist in no sense
at all, palpably have predicates holding of them that do not hold of con-
creta is, I dare say, obvious.18
16
It’s not true, e.g., of Cale in Jonathan Lethem’s Amnesia Moon (1995).
17
Are dream beings observable? Are hallucination beings in space and time and observ-
able? I’d hate to have to argue that these predicates don’t apply truthfully to noun phrases
referring* to such things, given how we talk. Luckily, it doesn’t matter to me one way or
the other. For Kant’s challenge to succeed, it has to be that no predicate exists that can be
used to distinguish what’s real from what’s not. To deflect Kant’s challenge, therefore, I
need only find at least one predicate (other than “really exists”) that yield truths when
applied (to all) things that exist but does not yield any truths when applied to nonreferring
noun phrases: My candidates are “causally efficacious” and “ontologically independent”—
the latter in the appropriate sense.
18
Being a claim about our language, it’s possible that Kant’s challenge could have
succeeded—after all, couldn’t our language have evolved so that any predicate applicable
to terms that refer could also have been applied to nonreferring terms (of one sort or
another) and yield truths? Well, yes, of course that’s in principle possible. I offer some
speculations later in this chapter for why things didn’t go this way.
I should add that the reader shouldn’t overestimate the significance of my needing a
response to Kant’s challenge: It wouldn’t have changed very much if that challenge had
Criteria for What Exists 91
Two last points before returning to the main theme of this chapter,
criteria for what exists: First, the temptation (stylistically speaking) is
overwhelming to descend semantically when speaking of the nonexistent,
to say, for example, that “it’s not the case that, given any property, there
is some item that exists in no sense at all which has that property,” rather
than “it’s not the case that, given any predicate, there is some nondenot-
ing noun phrase that when appended to that predicate yields a truth.” I
hope a cursory comparison of the two sentences just quoted explains why.
The related considerations raised in chapter 3 explain why the temptation
to speak directly of nonexistent objects and their properties rather than
go systematically metalinguistic is so natural; I hope I have also made the
case there that doing so does not (at least on these grounds alone) involve
ontological commitments of any sort. Lastly, given the delicate nature of
the discussion just engaged in, I’ve tried very hard to couch it almost
entirely in metalinguistic terms to avoid misunderstanding. I hope (hope
is a major theme of this paragraph) that the reader forgives me for not
being rigorously consistent in this regard. I’ll continue to fail to be rigor-
ously consistent in this regard: I’ll sometimes speak of nonexistent objects
and their properties in what follows. No more than before should such
talk be seen as ontologically committing to things (of any sort) with
properties.
Second, although in discussing criteria for what exists we’re implic-
itly engaged (much of the time) in a discussion of the properties of our
language, I’ll nevertheless continue to describe this as an analysis of meta-
physical considerations to contrast it with the previous epistemological
ones.
OK. Let’s return now to the question of what sorts of candidates we
actually have for necessary and sufficient conditions on existence. If we
go back to that list of candidate criteria I opened this chapter with, it
doesn’t seem that any of them are items that one can argue successfully
for on purely metaphysical grounds.19 Take observationality, for example:
As anyone with eyes can tell you, observationality is a deeply parochial
feature of the observing creature. It’s hard to believe that anything that
been sustained. After all, whoever said that a philosopher (me, for instance) couldn’t coin
terms “causally efficacious*,” “in* space and time,” etc., stipulated to apply truth-success-
fully only to terms that refer?
19
There are two general reasons why attempts to provide genuine metaphysical argu-
ments for constraints on what exists prove elusive. Briefly, “exists,” like “true,” “gold,” and
other terms, is criterion-transcendent. It’s a word that’s averse to fixed criteria that constrain
what terms it can apply to in order to yield truths. See my 2000b, pt. 4, § 3. The second
is even more significant: It’s that we can’t—on purely linguistic grounds—claim that the
word “exists” itself always indicates ontological commitment. This should already be clear
from the fact that the use of what should be merely a rhetorical enhancer, “really exists,”
does not sound redundant to the untutored ear. See chapter 5 for further discussion on
this.
92 Truth and Ontology
exists must be something that could be observed; it’s hard, anyway, to see
how something like that could be established without a substantial analy-
sis of observation itself. Certainly, it looks like an entirely empirical ques-
tion whether what could be observed is co-extensional* with what exists;
it doesn’t look like something that can be established by merely treating
observationality as a criterion for what exists.20
Causal efficacy isn’t much better. Why should anything that exists be
causally efficacious? Why couldn’t (some) things be causally inert? Notice
that to argue that we can explain what needs to be explained about the
universe on the basis of what’s causally efficacious, and that therefore by,
say, Occam’s Razor, only such items exist, isn’t to provide an argument
of the desired form. The latter is an epistemic argument based on the
idea that we need only commit ourselves to what we need (ontologically
speaking) for explanation. But what relevance has this epistemic point to
a criterion for what exists? How does it show that what exists must be
causally efficacious (vs. showing the weaker claim that on epistemic grounds
we need never find ourselves committed to anything that isn’t causally
efficacious to explain what we need to explain—an argument, as I’ve al-
ready shown, that won’t work on its own)?
Location in space and time is no better. (How do we establish that
what exists must be in space and time? Why doesn’t this, in the crassest
way imaginable, simply beg the question against the traditional Plato-
nist?)
The fourth candidate criterion, that what exists must not be ontologi-
cally dependent on any linguistic or psychological process, is also trou-
bled. However, it’s a candidate criterion for what exists that, on the face
of it, is more plausible than the others just considered. Furthermore, it’s
a candidate that, under certain descriptions, for example, “mind indepen-
dence,” philosophers have been metaphysically motivated to adopt as a
criterion for what exists (vs. having a motivation really due to epistemic
considerations).21 The reasons, therefore, for why there is no (purely
metaphysical) argument for its being a constraint on what exists deserves
a more extended discussion.
It’s not as easy to characterize what’s meant by claiming that some-
thing is “mind dependent” as it is to give examples.22 The most straight-
20
Also see n. 17 with regard to this and with regard to the criterion: location in space
and time.
21
See Vinueza (2001), and the literature he cites. Also see my 2000c.
22
The problem with producing a crisp definition is that this notion seems to include
too much. Minds, mental states, language, speech acts, works of art, and artifacts are all
items we’re prone to think do exist but that also seem to be mind dependent or ontologically
dependent on us in some sense. See Vinueza 2001.
One suggestion Vinueza doesn’t explore, and that seems not to have been previously
considered in the literature, is to presuppose something like a causal theory of reference as
already in place. We can consider a term to, at best, pick out something ontologically
dependent on us in the desired sense if it fails to denote anything causally speaking. Onto-
logical dependence (in this specific sense) then emerges as an extremely natural necessary
Criteria for What Exists 93
forward ones are these: characters in dreams, imaginary beings that one
thinks about, fictional characters, myths, and so on. And among these,
the most straightforward examples are fictional characters designed from
scratch by authors for the first time. The idea is that in deciding what’s
true about fictional entities, the author isn’t required to square what’s
attributed to such entities with anything: He or she can “make it all up.”
Perhaps it can even be said that what’s true about such fictional entities
is only what the author stipulates as true of them. And something all of
whose properties are ones stipulated as true isn’t something that is “mind
independent” or “ontologically independent of us.”23
Suppose something is dependent on our linguistic practices in exactly
the way that fictional entities have just been described as being. Why must
it be that such objects don’t exist? What sort of argument could be of-
fered for this metaphysical claim? That is, why couldn’t fictional practices
be ones that commit its practitioners to entities—entities, admittedly,
that exist in their own fictional worlds, but that nevertheless exist? Is there
anything philosophically conclusive we can say about sheer existence that
shows it’s inconsistent with the “mind dependence” or the “ontological
dependence” of fictional entities? No, there isn’t, as the following (ex-
tended) example makes clear.
Imagine two communities: The first (us) takes fictional beings to ex-
ist in no sense at all (we’re naive fictionalists); the second community
practices fiction exactly the same way we do but does take fictional beings
to exist—in as robust a sense as anything else exists—although they think
fictional entities (of course) exist in their own (fictional) worlds (they’re
fictional realists).
condition for what exists because we can say, as we’d like to, that such terms fail to
denote anything at all. We can then beef it up to a necessary and sufficient condition by
generalizing to any possible language whatsoever (not just ours). Although I find this
suggestion appealing, an exploration of whether it can be ultimately made to work raises
issues that can’t be pursued now.
23
I should point out that even in this case the author may be constrained in what he
or she is allowed to imagine—by concerns with what the public will like, the author’s own
taste, etc. But no constraints arise from the properties of the object being described. As I
said, this is the easiest case. Dream beings, hallucinations, or cases where authors write
about characters from other books or from mythology, etc., offer something more by way
of complication. In particular, the use of the term “stipulation” is less apt when involuntary
psychological events such as the contents of hallucinations or dreams are involved, or when
the fictional entity is one whose characteristics are publicly known. However, the tidying
up that’s needed to explain the desired sense in which all these items are ontologically
dependent on our linguistic practices, although a little more complicated, isn’t difficult.
Some of these items can be described as “language-created” in Schiffer’s (1996) sense.
He takes himself to be sketching out a middle position according to which such “language-
created” items are ontologically deflated because they’ve no “hidden and substantial natures
for empirical investigation to discover” (p. 159), beyond what our language practices estab-
lish them to have, but nonetheless are items that exist unaccompanied by language-creating
entities in other possible worlds. Later in this chapter and in chapter 5, I give reasons to
reject this compromise ontological position.
94 Truth and Ontology
At least two views of how fictional realists view fictional existence are
possible.24 The first requires “fictional space,” a metaphysical haven of
already existing fictional worlds in which every fictional possibility eter-
nally resides. When someone writes a novel, say, every term in that novel
refers to the beings in at least one fictional world that matches perfectly
the descriptions in that book. Indeed, the terms in the novel actually refer
to the events and beings in more than one fictional world—they refer to
fictional objects in infinitely many such fictional worlds—because a novel
leaves incomplete many details that are completed in various ways in the
various fictional worlds.25
This first view faces serious problems. Consider an alternative: There
are such fictional worlds, but fictional terms refer to nothing in any of
these because fictional terms don’t refer at all. Or (a worse possibility),
fictional terms refer to fictional beings in such a way that nearly every-
thing novelists say about their characters, and what happens to them, is
wrong. On what grounds can we adjudicate between these views? There
is nothing in the second community’s practices that fixes what its fictional
terms refer to in such a way that evidence can be brought to bear for and
against these alternatives to decide among them.
What if the fictional realists in the second community claim that their
fictional terms pick out fictional worlds as described? Isn’t that evidence?
So what if they claim this? Why couldn’t they be wrong? Notice that not
only is there nothing about their fictional practices that can be brought
to bear successfully on the question of what they refer to, but it’s not
even imaginable what facts could ever arise to adjudicate among these
alternatives. (That no data, even in principle, bear on this question one
way or the other suggests not only that the fictional beings are made up,
but also that the referential story has to be made up, too. But why think a
“made up” referential story bears any resemblance to genuine reference?)
Some may argue that because this view relies on a “description the-
ory” of the reference of fictional terms, it doesn’t allow “referential mis-
haps” (terms pick out exactly what the descriptions associated with them
pick out), and so the concern about these alternative possibilities is an
unreal one. But this response overlooks a worse problem, which is that
this approach can’t guarantee, of the fictional objects taken to be inde-
pendent of us, that they have the properties needed by the descriptions
24
Views that separate what’s attributed to fictional objects (while practicing fiction)
from the properties such objects have (see the discussion of option (2) in chapter 3) elude
the specific objection to the “fictional space” view that I raise, but face easily constructed
generalizations of that objection. In any case, my aim here isn’t to evaluate the cogency of
every realist position about fictional nomenclature, but to find one that’s immune to philo-
sophical attack on metaphysical grounds.
25
Alternatively, one might claim that fictional worlds (and fictional objects) are “in-
complete objects” individuated precisely by the predicates attributed to them, and no oth-
ers. In this case, the terms in a novel refer to the items in one and only one fictional world.
This latter view is similar to that taken of nonexistent objects by Parsons (1980).
Criteria for What Exists 95
we forge to pick them out. This problem persists even if, on this view, all
fictional possibilities (in some logically broad sense) are present in fic-
tional space. For that space, and all the possible fictional worlds within it,
are still taken to be ontologically independent of us, and this means we’ve
no guarantees that the wealth of possibilities that we need (given our
capacity to generate fictional descriptions) are out there.26
The second view is that fictional objects are brought spontaneously
into existence by their authors in the process of creating fiction; not only
that, but the traits of such objects change as authors change their views
about them. (Revising a novel has metaphysical consequences!) In this
case, the objects are regarded as stipulated in the sense that the properties
they actually have are ones they’re stipulated to have by their inventors.
This view of fictional realism doesn’t face the referential problems that
the first view faced because there are no antecedently existing entities that
fictional terms are supposed to refer to. Instead, the properties of the
objects are dictated by the very things said about them.27
Given the second view of the metaphysics of the practice of fiction
by the second community, ontological debate between us and them is
26
I raise the same objection against Ballaguer’s “full blooded Platonism.” See my
2000c, pp. 238–42. It’s important to realize that this objection can be pressed against any
view that (1) takes a collection of objects to be correctly characterized by a set of descrip-
tions we take to be true of them, (2) nevertheless takes such objects to be ontologically
independent of us, i.e., to be items we can’t stipulate the properties of, and (3) uses only
logical or conceptual resources of some sort to bridge the (epistemic) gap between (1) and
(2). See the discussion below of nontrivial explanations of the satisfaction of the reliability
requirement.
Deutsch (2000) speaks of a “fictional plenitude [that] is defined by a logical compre-
hension principle that says, in effect, that the domain contains any and every variety of
object and event imaginable—and then some” (p. 155). Further, “a description recorded
in the course of making up a story cannot fail to describe something and it cannot be a
misdescription of anything” (pp. 155–6). We seem, on this view—in fiction, anyway—to
have the remarkable power to stipulate both the contents of the fictional plenitude and to
refer to those contents. Such things, therefore, cannot be ontologically independent of us,
but must be “made up” by us in the strongest sense of that phrase. Deutsch seems to make
it clear that he doesn’t take his descriptions of the fictional plenitude to be describing items
that exist in no sense at all. He describes comprehension principles as “ontological” (p. 150),
and he opposes his view explicitly (p. 155) to one that takes fictional terms to refer to
nothing. On the other hand, he claims that he has “little or no commitment” regarding
what the things in the fictional plenitude are, suggesting that his general view is even com-
patible with “something of a nominalist approach” (p. 157–8). Finally, he describes himself
as seriously committed to the idea that one makes things up “out of whole cloth,” or out
of “thin air” in fiction, although he glosses this as recording descriptions “that are to be
understood (philosophically) as describing elements of the fictional domain” (p. 167). Re-
gardless of the truth or falsity of my views about ontological independence, I can certainly
still deny that Deutsch can have everything he wants.
27
Thus, the previous objection against the first fictional realist view, of how the com-
munity can be sure that what it says about fictional entities is true, is easily answered: What
it says dictates what properties those entities have, and so of course what’s said is true about
them. The properties of these entities, and even the identity conditions among them, are
seriously incomplete, of course, depending, as they do, only on what’s said about them.
96 Truth and Ontology
fruitless, because we agree on all the facts available to adjudicate the de-
bate: We both agree that fictional entities are “made up” by the authors
that invent them (“but how does this show the entities don’t exist?” fic-
tional realists ask us). Nor is it possible to show that the fictional realist
gets into some sort of trouble by claiming that fictional objects exist. It
isn’t, after all, that the fictional realist thinks this implies that an author
could get wrong what he or she writes about fictional characters (fictional
entities aren’t that sort of entity). Nor do they think that their claim that
fictional beings exist has any impact on what they say about any other
kind of entity. To the objection that fictional realists are lacking a robust
sense of reality, they respond that, on the contrary, they have an exquisite
sense of reality, for they never confuse fictional entities with other sorts
of entities or think that the epistemic practices used to discover things
about nonfictional objects are relevant to knowledge claims about fic-
tional objects. Indeed, fictional realists are even willing to grant that the
properties fictional objects have are stipulated by their authors. But what,
they ask disingenuously, does the fact of stipulation have to do with exis-
tence?
Both fictional realists and naive fictionalists agree: If one uses lan-
guage to refer to a made-up character, then just because a real person
shows up with the same properties, it doesn’t mean that one was (against
one’s will) referring to that person rather than to the made-up character.
Indeed, both communities have straightforward methods for preventing
problems should such things happen. Suppose, for example, that B makes
up a fictional character for a screenplay, and then someone from Ohio
shows up with a lawyer because B has inadvertently described his life,
looks, and personality entirely. Luckily, B (in either community) can pro-
tect herself against such eventualities by writing a claim of intent at the
end of screenplay: The characters in this work are fictional, and any resem-
blance to anyone, living or dead, is purely coincidental.28
Of course, this disclaimer isn’t relevant if we think that a work is
fiction and learn instead that the author was actually engaged in nothing
of the sort. And, of course, vice versa is possible too: We can think a
character mythological and learn that he isn’t. This wouldn’t be a matter
of our discovering that there is someone with characteristics similar to the
mythological being; rather, this would be a matter of our learning that a
group of people were referring to him. In all these cases, what has hap-
pened is that we’re mistaken about the linguistic processes at work. In
the first two cases, the individual is pretending to engage in one sort of
activity (with one set of conventions) but in fact is engaged in another.
In the third case, we’re under the impression that the language describing
the purported mythological being arose the way mythological language
usually arises—through religious misinformation, say—whereas instead,
28
Notice that this disclaimer, on the face of it, anyway, takes no position about the
ontological status of fictional entities.
Criteria for What Exists 97
it’s a sober description (say) of an unusual person that all the speakers
actually know. Both communities have access to these ways of adjudicat-
ing among these cases; existence, one way or the other, doesn’t come
into it at all.
If the foregoing is correct, then the last candidate criterion for what
exists, that something exists if and only if it’s independent of linguistic
and psychological processes, isn’t any more of a genuine criterion for
what exists than the first three are, because it fails to even be a necessary
condition. Does this mean that it’s truly indeterminate whether fictional
entities exist or not? In one sense, the philosophical sense, the answer is
“Yes.” Should fictional realists stubbornly insist that fictional entities do
exist, and should they invoke a version of the second fictional realist story,
there is no philosophical argument—no version of Occam’s Razor, for
example—available to nudge them from their position. What is true,
however, is that our community of speakers, for whatever reason, does
not take fictional entities to exist in any sense at all. One can give socio-
logical reasons for why our folk ontology took this route, but such rea-
sons won’t be philosophically binding.
Here’s an example of what such an explanation might look like: The
first step is to note that the word “exists” plays a contrastive role in local
discourses—we use it to contrast actual horses with flying horses, to con-
trast even primes greater than 2 with odd numbers, to contrast the ether
with black holes. That its contrastive role is local explains why the word
“exists” doesn’t have necessary and sufficient conditions attached to it
(we can’t predict ahead of time which contrasts we may need to make in
a discourse, and how). Notice that this view bears a strong resemblance
to views that take ontological positing in mathematics, for example, to
have originally arisen by a metaphorical extension of the ontological pos-
iting of ordinary life. I agree for the sake of argument that an extension
in usage has occurred; I disagree that there is a respectable sense in which
it can still be labeled “metaphorical.”
The second step goes like this: We’re willing to extend the word
“exist” to a collection of noun phrases in a discourse, provided the exten-
sion doesn’t breed confusion in already established discourses. Talk of
distinct fictional worlds is a rather sophisticated, and recent, construal of
the ontology of storytelling; one not available to the general fiction-
listening public several centuries ago. But without distinct worlds to seg-
regate fictional entities within, one must distinguish them from the flora
and fauna one is likely to meet in one’s travels. And this motivates the
commonly held intuition that they don’t exist (in any sense at all).29
What grounds have I for claiming that our community of speakers
doesn’t take fictional entities to exist in any sense at all? The naturalness
of statements like this:
29
This addresses the question raised in the second paragraph of chapter 3, n. 13: The
use of “exists” in mathematical discourse is contrastive in what’s obviously a local discourse.
98 Truth and Ontology
Although there are cities in H.P. Lovecraft stories that exist, there are
places that don’t exist that he depicts as being located in those cities.30
Our general acceptance that fictional objects exist in no sense at all
isn’t a brute intuitive fact about fictional objects. Rather, this intuition is
an application of a more general intuition that if something is entirely
“made up” or is ontologically dependent on our linguistic practices or
psychological states, then it exists in no sense at all. “Ontologically depen-
dent” here is not understood in the sense that it’s, say, a psychological
state or a linguistic item (e.g., a word)—these things do exist—but in the
sense that it’s (part of) the content of such a thing, and this isn’t content
in the sense that an actual apple is the referent of the word “apple,” but in
the more elusive sense that a hallucination of an elf (or apple) has as its
content “an elf” (or “an apple”) that exists in no sense at all, or in the
(more elusive) sense that the phrase “that elf” (or “that apple”), when
directed at a hallucination, picks out—has as its content—nothing at all
(despite being meaningful).31 This general intuition explains not only why
we take fictional objects to exist in no sense at all, but also why we take
hallucinated or dreamt items to exist in no sense at all.
Before we go on to milk this “general intuition” for ontological im-
plications (about abstracta, in particular), I want to redescribe in what
sense determining a criterion for what a discourse commits us to, or de-
termining a criterion for what exists, has proven to be philosophically
indeterminate. The indeterminacy is twofold. First, it’s not possible to
provide reasons for choosing the (first-order) quantifiers as the textual
carrier of ontological commitment, versus, say, any of a number of inter-
preted “existence” predicates. This means, in particular, that when we
view our own scientific practice as the generation of doctrine (of a certain
useful sort), and even when we consider how that doctrine is applied,
tested, and generally brought to bear on our world, there is nothing that
rationally supports or rules out, say, Quine’s criterion for what a discourse
commits us to, versus an existence predicate approach that only admits,
say, causally efficacious entities, or ontologically independent entities, as
existing.
Next, suppose we try to force the issue by determining a criterion for
what exist—the hope being that the discovery of such a criterion, onto-
logical independence, say, will force us to an existence predicate view
(because not everything quantified over proves to be ontologically inde-
pendent of us). This last strategy is dashed because we haven’t any (ratio-
nally binding) argument for the adoption of a criterion for what exists
30
There is more to say about when to trust (and when not) that a usage of “exists”
carries ontological force in the vernacular. See chapter 5.
31
The aficionado will realize that, in trying to indicate what I’m referring to, I’m
borrowing a dangerously overused word (“content”) from philosophy of language or phi-
losophy of mind. Hopefully, the trained reader can see what I’m getting at through the
haze such terminology can arouse.
Criteria for What Exists 99
32
Vinueza 2001 attempts a metaphysical construal of psychologically dependent items:
that they possess only “representational” qualities, qualities held by virtue of being purely
psychologically generated items. I prefer to avoid a metaphysical construal of ontologically
dependent items and, more specifically, psychologically dependent items, because I presume
that we take such items, e.g., the contents of hallucinations or dreams, to exist in no sense
at all, and so, therefore, we can’t take them to have any properties either. Still, we might
attempt a version of Vinueza’s construal by shifting it to a consideration of what sorts of
statements about hallucinations or dreams we take to be true; viz., only those statements that
take the form of attributions of representational qualities to the contents of hallucinations or
dreams are ones we take to be true of such items. This won’t work, however: I can halluci-
nate pink mice. The (correct) claim that such hallucinated mice are pink is not in the form
of an attribution of a purely representational quality (because the property of being pink is
not a purely representational quality).
33
A version of this condition first appears as (*) in my 2000c, p. 227. The comments
of the next few pages are, for the most part, borrowed and modified from my 2000c. The
entire discussion can be impounded to yield conditions on what we take to exist only because
in our community of speakers we take ontologically dependent items to exist in no sense
at all.
Talk of reliability may remind readers of Field’s (1989) gloss of Benacerraf’s challenge
to the Platonist, which is to “provide an account of the mechanisms that explain how our
beliefs about these remote entities can so well reflect the facts about them” (p. 26 ). (Also
see pp. 27–30 and pp. 230–39.) At the time of my 2000c, I was convinced that the consid-
erations I was there raising, which I encapsulated in (*), were different from those raised
100 Truth and Ontology
earlier by Field, but now I’m not so sure. The main reason for my thinking the considera-
tions couldn’t be the same ones was that Field (1989, pp. 28–9) offered a response to
Benacerraf’s challenge (as he glossed it), which would be an obvious non sequitor if his
considerations (and Benacerraf’s for that matter) were the same as mine. This is that the
Platonist could argue according to Field (1989) that
The problem is that the truth of the applied mathematical statements (which is all that
the successful application of mathematics can buy us) simply won’t provide what’s re-
quested: that we have an explanation for why the properties attributed to mathematical
objects are the properties they have; and this is because the mere truth of a collection of
statements doesn’t tell us that the (mathematical) terms in those statements hold of the
items that the Platonist takes them to hold of (items ontologically independent of us). Or
equivalently, the truth of such statements doesn’t guarantee that they’re true of the items
the Platonist takes them to be true of. See example (3) in the section “Some Illustrations”
later.
The same thing happens when Field (1989, pp. 233–8) addresses Lewis’s (1986, pp.
111–2) response to Benacerraf’s puzzle. Lewis argues that all facts in the mathematical
realm hold necessarily. Again, if the reliability concern Field is concerned with is the same
as mine, then the appropriate response to Lewis (which Field doesn’t make) is that Lewis’s
response is a non sequitor: It doesn’t help to describe mathematical statements as necessary,
nor will it help even if it’s said that mathematical objects have their properties necessarily—
say whatever you want to say about these things, the issue is how the processes we use (includ-
ing whatever logic we’ve adopted) to capture mathematical truths is responsive to the prop-
erties mathematical objects have. This problem is there whether mathematical objects have
their properties necessarily or not. See example (1) in the section “Some Illustrations” later.
Criteria for What Exists 101
34
See my 1999 for a discussion of how Resnik’s views, on how the properties of math-
ematical abstracta are posited, seem to open him to the objection that, contrary to his stated
view, such abstracta are not ontologically independent of us.
35
It’s an empirical epistemic fact about us and our universe that certain purported
knowledge processes—magical ones, for example—don’t exist. Similarly, it’s an empirical
epistemic fact that we’re not Leibnizian monads prearranged to have sensory experiences
coordinated with (but not caused by) local environmental events. Were we such Leibnizian
102 Truth and Ontology
monads, then although the appearance of causality wouldn’t correspond to anything meta-
physically real, the nontriviality requirement would nonetheless be satisfied. Both the reli-
ability and the nontriviality requirements look like definitional constraints on allowing a
process to be knowledge-yielding about objects that are ontologically independent of us.
Naturalized epistemology, however, is an empirical subject matter, and so it’s an empirical
matter how those requirements are satisfied. For these reasons, I’m not convinced by Katz’s
(2002, p. 372) attempt to convict my reliability considerations of presupposing a “contact
epistemology,” or some form of causation.
36
There is a nonzero probability, of course, that the coin-flipping process gives the
right answer in every case—no matter how many times it’s done (the probability of a run
being entirely veridical diminishes but never vanishes as we let the run increase in length).
So, it’s empirically possible for some epistemic processes that we use to be only accidentally
reliable—to admit of only the trivial explanation for their reliability. It’s part of the practice
of empirical science to try to recognize and rule out such accidental reliabilities in our
knowledge-gathering activities.
Criteria for What Exists 103
is only a trivial explanation for why we know what visual sensations we’re
having (say, of a red rubber ball) when in the grip of a hallucination.37
I want to rehearse one last point before I use the reliability and non-
triviality requirements to provide an argument that mathematical ab-
stracta don’t exist. This is that the use of these requirements for meta-
physical purposes turns on the examination of how truths about a subject
matter are established. These truths, one and all, seem, as a matter of
surface grammar, to attribute properties to objects. In deciding that such
objects, after all, exist in no sense at all, it isn’t that therefore these truths
aren’t truths or that they’re somehow ill-formed, or that their grammati-
cal or logical form is different from what it appears to be. None of these
conclusions follow (as I have already shown). All that follows is that,
metaphysically speaking, there are no objects of the sort that seem to be
talked about, and so, of course, there are no objects with such and such
properties.38
Here’s an argument that we shouldn’t take mathematical abstracta
to exist any more than we take fictional items to exist: Mathematical ab-
stracta are ontologically dependent on our linguistic practices in just the
same way that fictional items are, and because—this is an important part
of the claim—the tacit conventions at work in our ontological practices
aren’t specific to fiction but to any collection of purported items that aren’t
ontologically independent of us, they should be extended to mathemati-
cal abstracta.
This argument, however, requires establishing that mathematical ab-
stracta are ontologically dependent on us in the appropriate sense. How
do we manage that? It’s easy. Mathematical objects can’t be ontologically
independent of us because then we wouldn’t be justified in claiming that
the statements we take to be true of such mathematical objects are true.
But that we’re so justified is simply a datum of mathematical practice (see
chapters 1 and 2). Suppose, therefore, that mathematical abstracta were
ontologically independent of us. Then the nontriviality requirement on
explanations for the dependability of the methods we use to establish
mathematical truths would have to be satisfied. But it can’t be satisfied in
the case of mathematical abstracta.
I can’t quite show, of course, that the nontriviality requirement can’t
be satisfied in the case of mathematical abstracta (because this requires
37
This point needs fine-tuning. I can dream about short, stocky creatures and tell
people later that I dreamt of elves. It may be pointed out to me, after I describe their visual
appearances in detail, that in fact I was dreaming about dwarves. There is therefore a non-
trivial explanation available for why we are right when we claim (correctly) that we’re dream-
ing about horses, say, although nontrivial explanations aren’t available for all the claims we
make (correctly) about such dream creatures.
38
And here I’m using “there is” in an ontologically committing way. (As the context,
I hope, and this footnote, surely, makes clear.) See chapter 5 for a discussion of how we,
in the vernacular, indicate ontological commitments.
104 Truth and Ontology
Some Illustrations
1. Suppose a philosopher argues thus: Consider a logical system S, and
suppose that S generates some truths of mathematics.39 Argument: S is a
priori; therefore, the statements about mathematics that follow from S
must be true of (certain) mathematical objects. Response: What does a
priori mean? If it means “Known to be true independently of experience”
(however that’s cashed out), then S can’t be taken to be known in this
way to hold of a collection of objects that are ontologically independent
of us unless the nontriviality requirement with respect to a priori methods
of learning about those mathematical abstracta is satisfied. But what could
be said about us, and how we’ve come to adopt S, that could possibly be
used to indicate how a priori truths constrain objects that are ontologi-
cally independent of us?
Consider, for example, the claim that the a priori nature of S is rooted
in the fact that it fixes what’s necessary and possible about anything that
exists. Response: How is that (metaphysical) claim to be established? If we
can’t imagine alternatives to S—even if alternatives to S sound like sheer
nonsense to us—that (purely) psychological fact doesn’t show—all by it-
self—that the theorems of S are metaphysically binding. For how do we
distinguish innate failure of imagination from recognition of metaphysical
limitation? Notice that even if we grant that there are such metaphysical
facts to be found out, we need to explain how a prioristic intuitions could
possibly be sensitive to such things. No such story even seems possible.
Once upon a time such philosophical stories were possible and even
respectable: Those who believed in God, and in His supernatural might,
could (respectably) think that He might be so inclined as to construct us
so that we can recognize such metaphysically binding truths, and recog-
nize that they’re metaphysically binding. There are, no doubt, questions
about how we could prove (a) God’s existence and (b) that He really is
so warmly inclined toward us (esp. given all the other evidence to the
contrary), but nevertheless, a story seems (in principle) available, given
these background beliefs. Those of us living in post-Darwinian times, and
who think we’re at best a pretty smart ape, can easily wonder what cogni-
tive mechanism could possibly explain how our intuitions about meta-
physical structures are reliable with respect to those structures.
39
Two options: (1) S is a higher-order logical system of some sort with substantial
ontological commitments; (2) S is a blend of a weaker logic—such as the first-order predi-
cate calculus—with substantial mathematical or set-theoretic axioms.
Criteria for What Exists 105
Notice that this objection applies not only to views that take mathe-
matical abstracta to be a kind of object that’s ontologically independent
of us but, nevertheless, something whose properties we can capture by
means of ratiocination of some sort; it also applies to views that take the
object construal of mathematics to be misguided, to see in the a priori
nature of mathematics not knowledge of objects but knowledge of concep-
tual linkages—purely logical knowledge that governs whatever is know-
able, or something like that.40 In this case, we still need to supply a non-
trivial explanation for the reliability of our capacity to recognize such
linkages: We need to know why such conceptual linkages must apply to
what’s ontologically independent of us—mathematical and empirical ob-
jects, for example. This Kantian-sounding move requires more of Kant’s
views than is desirable—in particular, his Idealism—because only if any-
thing we think about is preconfigured to fit an a priori conceptual struc-
ture do we know that the concepts and their linkages that we supposedly
generate a priori must fit it.41
2. This second illustration raises a problem with the applicability of
mathematics, given the claim that mathematical objects are ontologically
independent of us. Start by noting that it’s agreed by pretty much every-
one that it’s not by causal access to mathematical objects that we learn of
their properties. Mathematical objects are taken not to be in space and
time, to be acausal, independent of us, eternal, necessarily possessing their
properties; and so, access to their properties must occur by some non-
causal means. One possibility sometimes raised is that we learn of such
objects by examining empirical objects that instantiate them; for example,
40
Such a view assimilates mathematics, or that part of mathematics so assimilatable,
anyway, to logic, and grounds our knowledge of the former in the a priori perception of
truths of the latter.
41
Kant (1783) is entirely aware of this, and he’s also clear that it motivates his idealism.
He writes:
Sensibility, whose form lies at the foundation of geometry, is that upon which the
possibility of outer appearances rests; these, therefore, can never contain anything
other than what geometry prescribes to them. It would be completely different if the
senses had to represent objects as they are in themselves. For then it absolutely would
not follow from the representation of space, a representation that serves a priori, with all
the various properties of space, as foundation for the geometry, that all of this, together
with what is deduced from it, must be exactly so in nature. The space of the geometer
would be taken for mere fabrication and would be credited with no objective validity,
because it is simply not to be seen how things would have to agree necessarily with the
image that we form of them by ourselves and in advance.” (p. 39, emphasis added)
One concern surely being expressed here is that no explanation seems available for why a
priori ratiocination should guarantee that something independent of us must be the way
that ratiocination takes it to be. My interpretation: Kant is driven to a (transcendental)
nativism because he takes that position to be the only one that can explain why a priori
reasoning must apply to something; if that something is metaphysically independent of us,
no such explanation is possible, but it is possible if that something is a product of our own
“sensibility.”
106 Truth and Ontology
that, after all, there might be one sock in the drawer and another sock in
the drawer, no other socks in the drawer, and yet, the object 2 not bear
the instantiation relation to the socks in the drawer.
To make the issue perfectly plain, here’s a somewhat different exam-
ple: Consider a set of three apples. In this case, what takes the place of
the instantiation relation is the membership relation. What could be more
obvious than that each of these apples belongs to the set of these apples?
But again, if the set is to be independent of us—an item that we can’t
stipulate the properties of, then it’s (epistemically (!)) possible for the
membership relation not to be as it seems to us it must be. We can’t
stipulate this, or see it as something that follows from a dictate of “pure
reason,” unless the set itself is a “conceptual object” of some sort, an
item, therefore, that we can comfortably dictate the properties of.42
Desperate responses are always possible (bumps in rugs can always be
moved around by judicious stomping); this shows, someone might claim,
that (certain) mathematical objects are located where their instantiations
are (and so, causal relations, in particular, exist between us and such ob-
jects). This has been tried, but it doesn’t work, primarily because the
nontriviality requirement still isn’t satisfied. When epistemically available
causal relations exist between us and ordinary empirical objects, an empir-
ical study can arise of how we make mistakes and how such causal rela-
tions give rise to artifacts that mislead us.43 But no such study exists in
this case. We can be mistaken about what properties an object has, but
not in the sense that the object is exactly as it is (we haven’t made a
mistake about it), but we’re wrong about the relations it possesses to
abstracta.
Digression
The alert reader has realized that this argument applies to properties in
general. If properties are (like sets) to be seen as items that are eternal,
acausal, not in space and time, ontologically independent of us, but items
that are instantiated by their instances, we must similarly worry about the
satisfaction of the nontriviality requirement with respect to our knowl-
edge of that instantiation relation. Why couldn’t we be wrong about
something being blue—not in the sense that we think it’s blue but we’re
wrong because of bad lighting, color-blindness, or some other condition
42
With sets, it’s almost overwhelmingly tempting for a philosopher to describe the
(epistemic) situation not in terms of our perception of an abstraction relation of some sort,
but rather in terms of our recognition of analyticity; i.e., it analytically follows from the
notion of a set that its members just are the items that belong to it. What else is possible?
But, then, why describe a set as an object that’s independent of us, vs. an item that we
stipulate the properties of, and so something that is, in a perfectly clear sense, “made up”?
Notice the objection: It’s not a denial of analyticity; it’s that the notion of analyticity is
being used here to paper over an epistemic gap by fiat.
43
See chapter 6 for details on this.
108 Truth and Ontology
(and it’s actually red)—but in the sense that the object is just as we think
it to be, but the instantiation relation between the property of blueness
and the object isn’t what we think it to be? That such a possibility is so
hard to even imagine suggests that properties aren’t ontologically inde-
pendent of us and that, consequently, properties don’t exist.
It’s not a good answer, no more than in the mathematical case, to
treat the instantiation relation as an “internal relation” or an “essential
relation” of the object in order to try to stymie the possibility being
raised, because the question isn’t one about metaphysics but about episte-
mology: What’s it about us that gives us an epistemic pipeline to the
relations (necessarily or not) possessed by an object that is ontologically
independent of us—and so an object we can’t presume to know the prop-
erties and relations of? Some explanation other than the trivial one must
be given—that’s what the nontriviality requirement requires.
Consider a common explanation: We “abstract” (knowledge of) the
property from the object instantiating it. This is only to put a label on
the problem: Abstraction somehow enables our minds to follow up the
instantiation relation from concreta to the abstracta they instantiate. But
how does it do this? What magic takes us from something in space and
time to abstracta that are located nowhere?
Notice that there would be no epistemic problem if this were the
view instead: We treat objects that are similar (in some respect) as having
(in that respect) a property in common, and the property (of course) is
stipulated to hold of those objects. But if (contrary to this view) proper-
ties really are objects ontologically independent of us, then we can’t so
stipulate their properties or relations: Any claim we make about them or
their relations is subject to our supplying a nontrivial explanation for why
we know these claims.44
Staking out a position, as I’d like to, on which properties, too, exist
in no sense at all, requires more to be said that can’t be said now. So I’ll
restrict myself to the following remarks. The nonexistence of properties
doesn’t prevent us from recognizing that objects possess (specific) prop-
erties: On this view, we can still see that something is blue, determine
that electrons have charge, and so on. Further, we can understand sen-
tences that express such facts to be objectively true or false without requir-
ing objects of a certain sort—properties—the instantiation relations of
which to specific items are needed to make such sentences true (or false).
That is, it’s as objective a fact as one can wish for that snow is white,
but the sentence “Snow is white” isn’t made true by the existence of an
(instantiation) relation between two objects, snow and whiteness. Rather,
44
I should stress that, sympathetic as I am to the view I’ve just described (at least with
respect to some abstracta), it’s not my view about properties. I indicate what I think about
properties in the next paragraph. Unfortunately, an adequate (full) discussion of properties
cannot be undertaken in this book.
Criteria for What Exists 109
it’s made true by snow’s being white, that is, by one kind of thing (snow)
being a certain way (white). So the informal use of “possesses” above
doesn’t betray reference to the presence of a relation, nor does the more
tortured locution “being a certain way” betray reference to a kind of
thing. End of digression.
Of course, mathematicians don’t even try to detect the objects they posit,
since they endow them with no properties that might be detectable. But they
do recognize evidence that the theory of the posits is consistent and bears
fruitful connections to other mathematical and physical theories as counting
in favour of the existence of the posits. . . . [S]uccessful positing is measured
in terms of the success of some theory of the posits. . . . (p. 195)
This simply evades the demands of the reliability and nontriviality require-
ments on ontologically independent objects, and this is true, generally,
of any coherentist approach of this sort. The problem is that, although
establishing the truth of mathematical doctrine this way is unobjectable
and, indeed, perfectly appropriate (because the role of a mathematical
theory with respect to other theories is precisely what tells us whether it’s
indispensable or not), nothing in this procedure tells us that the posits of
the theories in question are ontologically independent of us, and conse-
quently, nothing in this procedure tells us whether such posits exist.
I’ve argued in the foregoing that we should be no more committed
to the existence of mathematical objects than we are to the existence of
fictional ones. But it’s important to stress again, despite the ontologically
weighty sound of this conclusion, how philosophically lightweight it re-
ally is. For I’ve rooted it in a communitywide tacit convention regarding
ontological dependence; that something ontologically dependent on us,
the way that dream figures and fictional objects are, is something to be
regarded as existing in no sense at all. Were this not a tacit convention of
ours, as I’ve indicated, nothing particularly dramatic would change in our
epistemic practices. The real issues, in fiction as in mathematics, about
how we recognize what’s true or false about our claims and, in addition
(in the case of mathematics), about how we successfully apply mathemat-
ics, remain on either construal of the ontology here. Nevertheless, in part
II, I take the stand that mathematical objects exist in no sense at all, and
I there turn to the question of what role mathematical posits, in the tech-
nical sense of items (1) objectually quantified over in mathematical dis-
course, although items that (2) exist in no sense at all, play in science.
110 Truth and Ontology
45
Schiffer (1996) argues that the same is true of propositions: “To say that the propo-
sition that snow is white exists in possible worlds in which there are neither thinkers nor
speakers is merely to say that the proposition would have existed (and had a truth-value)
even if there had been neither thinkers nor speakers” (p. 166, n. 5). Fictional objects be-
come mind independent on his view via the claim, which he endorses, that they are abstract
objects.
46
For an example of how this can be done in the case of mathematical languages, see
my 1994, pt. II, §§ 6–7. The terms of natural languages don’t pose any further special
problems.
Criteria for What Exists 111
those worlds, and because, in turn, I take this to amount to the truth or
falsity of sentences or propositions describing such states of affairs, and
further, because the truth or falsity of sentences or propositions describ-
ing such states of affairs, I’ve argued, is independent of the ontological
requirement that properties exist, there is no problem. Such statements
are true or false of possible worlds just as they’re true or false of the actual
world, and considerations about truth and falsity of sentences in possible
worlds don’t require the existence of properties (or other abstracta) any
more than do considerations about truth and falsity in the actual world.47
Statements about abstracta, for example, “2 + 2 = 4,” are even easier to
handle: We simply stipulate that such statements have the same truth
values in possible worlds as they do in ours.
But the reader may still suspect that a problem is lurking here: Can
sentences (or propositions) be treated as true or false, if they’re abstracta
and so, as far as the nominalist is concerned, don’t exist either? Well, if
the nominalist does want to claim that sentences or propositions don’t
exist, there isn’t a problem in doing so for reasons already given in chap-
ter 3: To say that certain sentences are true (or false), and to supply (say)
a semantics that takes such sentences to be true (or false), requires (if we
take the Tarskian route) a metalanguage in which quantifiers range over
a domain rich enough to satisfy the semantic needs of a theory of truth.
Such (objectual) quantifiers needn’t commit us to the existence of ab-
stracta—sentences, propositions, mathematical objects of various sorts,
and so on—any more so than do any other objectual quantifiers.
But, in any case, despite my nominalism, I wouldn’t want to claim
that (every) sentence (or proposition) exists in no sense at all. As I discuss
in part II, my general view is that we must distinguish the posits in pure
mathematics from the posits in applied mathematics. The posits of pure
mathematics exist in no sense at all, and if the pure mathematics in ques-
tion is the mathematics of formal languages, for example, then the notion
of a sentence or of a proposition, as such notions arise there, indicates
classes of posits of pure mathematics—these are abstracta that exist in no
sense at all. However, applied mathematics is a different matter because
(often) in application (some) of the mathematical posits of mathematical
doctrine so applied proxy for other things that do exist. My view is that
the entire semantic apparatus for truth (when applied to an actual lan-
guage) is itself a branch of applied mathematics, and so sentences (or
propositions) in such a context can proxy for something that does exist
but that, within the context of semantics, is described via abstracta (which
exist in no sense at all). Similar things happen in physics. Exactly how
mathematical posits proxy for what’s real will be the focus of part II.
47
I hope my language here won’t mislead: “States of affairs” is merely a picturesque
way to generally refer to things like snow being white, glass being transparent, etc. States
of affairs don’t exist any more than properties and relations do.
112 Truth and Ontology
48
Recall the discussion of necessary and sufficient conditions on existence earlier in
this chapter. There may be terms, i.e., that are nondenoting and so, when appended to the
existence predicate, yield falsehoods, despite it being a truth that appending such terms to
“ontologically independent of us” yields a truth.
49
I’ve argued, of course, that they range over more than everything that exists—at
least in one respectable way of understanding “range over”—but the argument I’m giving
only requires that they at least range over everything that exists.
I should add that it’s possible for ontological independence to outstrip our domain of
discourse and infect other notations in our language, i.e., predicates. For the most part, I’ve
avoided exploring the ramifications of this possibility; I will, however, raise it explicitly for
certain polemical purposes in chapters 8 and 9.
Criteria for What Exists 113
50
Compare this with Moore’s (1939) careful discussion of the notions “to be met
with in space” and “external to our minds.”
5
Ontological
Commitment
and the Vernacular:
Some Warnings
1
I illustrate this claim several paragraphs later.
114
Ontological Commitment and the Vernacular: Some Warnings 115
2
One area of discourse that can’t be used as evidence either for or against an ontologi-
cal role in the vernacular for the word “exists” is fictional discourse itself (vs. discourse
about fiction and its fictional entities). This is because even though sentences like “Although
gremlins, goblins, and orcs do exist, dragons, alas, do not” occur in stories, they fail to
indicate anything about the word “exists.” On the pretend-true view of fictional discourse
itself (to which I subscribe), the ontological force, if any, of uses of the word “exists” in
fiction is nullified; this maneuver isn’t available for mathematical discourse (by contrast)
because (applied) mathematical statements must be taken as true.
116 Truth and Ontology
ontological significance, but that this isn’t the case when no contrast is
drawn. An indication of this difference in use is our reluctance to say
straight out: “Fictional mice that talk exist” (because it sounds like an
ontological commitment to fictional mice), although there is no resis-
tance to saying “There are fictional mice that talk.”3 This suggests that
“exists”—used noncontrastively in the vernacular—carries ontological
force.
We must be careful about this claim, however, both with respect to
the supposed contrast with “there is” and with respect to the claim that
we can this easily distinguish, in the vernacular, noncommitting uses of
“exists” from committing ones. There is the point I mentioned in chapter
4 (n. 19): “Really exists” doesn’t sound redundant to the untutored ear.
(It doesn’t sound bizarre for someone to say “Numbers don’t really ex-
ist.”) This implies that rhetorically unstressed uses of “exists” need not
carry ontological force, and that ordinary speakers know this. The point
is that tone and stress play a big role—both with “there is” and with
“exists”—to indicate to speakers whether or not what’s being talked
about is something that one takes oneself to be committed to.4
Another point is that, although it’s true that “Mickey Mouse exists”
simply sounds false, it doesn’t sound false to say “Mickey Mouse exists in
the world of Disney.” Thus, in discourse about fiction, it’s not uncom-
mon to talk about what exists and what doesn’t exist in the fictional world
of some author. But surely “in the world of So-and-so” doesn’t (semanti-
cally) cancel the ontological force of “exists,” if it has any. (Lewis’s view,
that there really are possible worlds in every bit the same sense that there
3
It’s important, when harvesting intuitions about examples like these, to be clear
about the (implicit) contexts that are and aren’t being presupposed when imagining the
utterances of sentences. The stand-alone claim “Fictional mice that talk exist” will sound
natural if previous sentences (perhaps by other speakers) provide implicit contrasts. So, e.g.,
imagine someone making claims about what sorts of fictional animals there are, and igno-
rantly claiming that although there are lots of talking rats, there are no talking mice. This
might be opposed by the utterance of “Fictional mice that talk do exist.” This is different
from those contexts—which exist—where speakers are engaged in a straight-out discussion
of ontology. (The reader should not assume that cases where speakers are engaged in
straight-out discussions of ontology are always philosophical ones where “language is on
holiday.” Arguments about whether God exists or not, for example, are ones that ordinary
people, even extremely unphilosophical ones, understand the implications of very well, and are
very concerned about.)
4
Thus, the tone and stress used with “There are fictional mice,” plus context, can
indicate ontological commitment to such items. And with both idioms “there is” and “ex-
ists,” speakers have the option of being unequivocal via rhetorical enhancers: “There really
are fictional mice” and “Mice really exist.” This suggests that the intuitive difference in
ontological commitment between unstressed stand-alone uses of “There are fictional mice”
and those of “Fictional mice exist” may be due to subliminal pressure from the unconscious
inclusion of particular contexts of utterances in the one set of cases (and not in the other),
and therefore isn’t a symptom that it’s part of the semantics of “exists” (vs. “there is”) that
it carry ontological force.
Ontological Commitment and the Vernacular: Some Warnings 117
5
I’d argue that this is even true if we think “exists in the world of Disney” is short for
“exists in the fictional world of Disney.” This is because the only cogent way I know of to
make this run are along the lines of the cancellation option discussed in chapter 3, and for
similar reasons, it won’t work.
6
In this book, for example, I’ve repeatedly availed myself of rhetorical intensifiers in
order not to be misunderstood; I’ve repeatedly said, for example, that fictional objects exist
in no sense at all.
7
In the foregoing, I’ve been concerned with what—beyond the empirical objects of
science and ordinary life—we might take ourselves to be committed to (to take as “exist-
ing”). But as the history of philosophy—esp. ancient philosophy—makes clear, there are
other options. Imagine a neo-Platonist who takes “unchangingness” to be criterial of what
exists. Such a view is compatible with how “exists” works in the ordinary vernacular. Such
a person will say that ordinary physical objects don’t (really) exist; only abstracta do. A
certain style of philosopher—call him an “ordinary language” philosopher—might see neo-
Platonism (so construed) as a typical example of how (certain) philosophers try to abuse
ordinary language without paying for it. My counterclaim is that certain idioms—“exists”
in particular—are susceptible to such “abuses” precisely because of how they are used in the
vernacular.
I’m not claiming historical accuracy for the details of this example; I suspect Plato,
anyhow, derived unchangeableness as a criterion for what exists from something more basic
(i.e., the noncontradictoriness of property attributions to what exists).
118 Truth and Ontology
8
No linguistic device, no word, for that matter, is guaranteed to remain fixed in its
usage. Speakers, from generation to generation, are quite willing to put idioms to new uses
that can obliterate the possibility of a straightforward expression of what was previously
meant by that idiom. The history of word usage abounds with examples, and with writers
who are willing to be morally outraged at the loss of a particular easily expressed nuance.
9
Such a concern does arise when there’s a need to be explicit because one is in a
philosophical context, or because one is dealing with an individual with very different onto-
logical commitments: I stress again, because of how ordinary they are, conversations be-
tween the irreligious and the religious. This also contributes to an explanation for why, in
the vernacular, the idioms for making ontological commitments explicit are in thrall to what
seem to be obvious pragmatic devices: Ontological issues come up rarely not because ordi-
nary people have no real understanding of what such issues are about, but because the
ontological commitments of one’s discourse are usually understood in exactly the same way
by like-minded folk—family and friends.
Ontological Commitment and the Vernacular: Some Warnings 119
He goes on, a page later, to note that we “deal with abstract objects all
the time.” We listen to radio programs, we write reviews of books, we cor-
rect mistakes, and some of us draw triangles in the sand and write numbers
on chalkboards. To think otherwise about numbers on chalkboards,
Boolos suggests, is to be confused about the use of the word “on.”11
It seems to me entirely clear that Boolos is quite right about ordinary
language. “See,” for example, enjoys a straightforward usage extending
all the way from fuzzy colors on surfaces to baseballs and bats to neutri-
nos in the sun to various types and other abstracta. And surely (as an
embarrassingly large amount of awkward epistemological doctrine from
10
I succumbed to such a temptation, and pretty much for these reasons, at the end of
my 1998. Also, compare Austin 1962, pp. 62–77, and Bennett 1969.
11
I hear echoes of Austin here. See Austin 1962, p. 31.
120 Truth and Ontology
Concluding Remarks
Having said all this, I hasten to add that I wear my newly adopted nomi-
nalistic garb lightly. This isn’t only because I hold that nominalism is a
position that can’t be adopted for philosophically convincing reasons (but
only—at best—on the grounds that the most general intuitions prompt-
ing folk ontology tend toward nominalism), but because even with re-
spect to the latter considerations, there is still a question of exactly which
promptings of barely conscious folk-ontological intuition we should be
swayed by. Raising this issue brings us back to the Opposing Consider-
ation—specifically, to the suggestion that ordinary folk ontologically dis-
tinguish mathematical posits from their fictional kin. I’m now going to
set aside my qualms about whether “exists” carries ontological commit-
ments in contrast to “there is,” and argue that even if this is so, and even
if it’s clear because of this difference in the usage of “exists” that ordinary
people take fictional entities to exist in no sense at all but do take mathe-
matical abstracta to exist, we (philosophers) shouldn’t follow suit, despite
the fact that there are no definitive philosophical reasons for taking an
ontological stand about this, one way or the other.
I’ve claimed that ontological independence, in the appropriate sense,
should be the decisive consideration on which the existence of something
should be decided, and I’ve suggested that folk ontology is with me on
this one.12 But the Opposing Consideration suggests that my claim is
debatable: that the reason that it intuitively feels more natural to describe
mathematical abstracta as existing than it does to so describe Mickey Mouse
is because the dark promptings of folk intuition treat mathematical ab-
stracta differently from how they treat fictional entities. I concede this.
12
This doesn’t exclude, of course, other properties being empirically co-extensive* with
ontological independence in the sense that what we take to be ontologically independent
we also find to have those properties. E.g., it could be (and I think it is—see part II) that
items that are ontologically independent in the required sense are also causally efficacious
(and vice versa). My only claim is that, as a matter of folk intuition, the notion of ontological
independence appears to be more directly linked to the notion of existence than is the
notion of causal efficacy. (This is indicated by the belief held by many philosophers that
mathematical objects are ontologically independent, causally neutral, and that they exist.)
Ontological Commitment and the Vernacular: Some Warnings 121
13
This take on realism originates with Dummett and has been subsequently adopted
by Quine, Putnam, and others. See, e.g., Dummett 1963 and Quine 1981b.
122 Truth and Ontology
14
It’s likely that there is no “logic” of fiction—at least no uniform logic of fiction.
This isn’t true, of course, for discourse about fiction, or about fictional characters, because
those claims are part of our ordinary body of beliefs.
15
E.g., the elegance and ease of provable theorems, the usefulness of the augmented
system in applications, the surprising ways that results can be established, etc. The point of
the term “internal” is that considerations of how to square mathematical claims with onto-
logically independent items are not included.
Further details on how the view of mathematics I’m committed to is compatible with
semantic realism may be found in my 1994, esp. pp. 140–9, 210–4.
II
APPLIED
MATHEMATICS
AND ITS POSITS
Posited objects can be real.
W. V. Quine
This page intentionally left blank
6
Let’s begin: Posits are the purported referents of singular terms (names
and definite descriptions) wherever such terms arise in our discourse: or-
dinary life, the sciences, mathematics, discourse about fiction, and so on.
So too, and more important from the scientific point of view, among our
posits are the items picked out by quantifier claims, remarks to the effect
that “There are Ss,” even when particular Ss can’t be singled out by either
name or definite description. “Posit” is a term borrowed from Quine,
and it’s a strange term, as he uses it, because it simultaneously nods to
the world (the posits of our theories exist) and to theory construction
(posits are items we “posit” when we adopt theories). Thus, “posit”
seems stylistically tainted with a certain use/mention conflation: What’s
attributed to posits might sometimes be better attributed to the terms
purporting to refer to such things, and as commonly used, “posit” also
seems to involve diffidence about commitment (in ordinary speech, we
only “posit” what we don’t know to be true). Quine entirely disavows the
125
126 Applied Mathematics and Its Posits
latter contrast when he uses the word—at least with respect to those
posits we take to exist: We know those to exist as much as we know any-
thing we know—and harmlessly glories in the first conflation.1
The Quinean can formally distill out the distinct ways posits nod to
the world and to our theories by focusing on differences between the
quantifiers and the predicates. On Quine’s view, our quantifiers range
over what exists, not over what we merely take to exist. This is why our
theories can be wrong. (If our theories only concerned what we took to
exist, then those theories would not be falsified by what actually exists and
how it deviates from what we take to exist.) On my modification of this
view, the quantifiers are no longer restricted to what exists, but still, the
falsification of theories turns (in part) on what there is—which the range
of the quantifiers includes—not being as those theories claim it to be.
For the Quinean, this is compatible with another sense in which the
“posits” of our theories are only what we take to exist. We take frogs to
exist—this requires the predicate “frog” to be instantiated, and it may
not be, but this is a very different point from the one made in the preced-
ing paragraph, and notice how it turns on the instantiation (or not) of a
predicate.
In describing this (Quinean) view about predicates, I’m not agreeing
that the role of predicate and quantifier crisply divide in this way. They
do on the Quinean view, I think, because the extensions of predicates are
determined solely by the theories governing predicates. But on my view,
the predicate—if it’s a “criterion-transcendent term” (or a “natural kind
term” as it’s more commonly put)—is independent of theory insofar as
we may change the theory and yet regard the predicate as continuing to
pick out what it always picked out (the theory we took to govern the
predicate was wrong). (See my 2000b, pts. III and IV, for details about
how I think this goes.)
Hereafter, “posit” is a technical term. Although (1) it “refers to” the
purported references of individual noun phrases or to the items that com-
mon nouns hold of, and (2) it’s ontologically neutral, because what a
noun phrase picks out or holds of may not be taken to exist in any sense
at all (something, as I indicated in part I, that I believe to be true of
mathematical abstracta and fictional characters), nevertheless, when a dis-
course is regimented, (3) it always corresponds to the results of quantifier
commitment, the derivability of sentences of the form (∃x)Sx, where Sx
is any formula with variable x free.
Notice that Quine’s criterion determines what those committed to a
discourse (ours or someone else’s) take to exist—hence the theory relativ-
ity (“derivability”) of the existential implications. It’s a different matter
what exists on this view (which existential statements are true), but, of
course, these explicitly separate only with respect to alien discourses.
1
For the most part, Quine’s choice of terminology hasn’t misled. See, however, Col-
lins 1998.
Posits and the Epistemic Burdens They Bear 127
For Quine, and those who adopt his criterion for what a discourse is
committed to, (2) and (3) contradict each other. But the nominalist posi-
tion I’ve taken toward mathematical abstracta in part I requires my drop-
ping Quine’s criterion for an existence predicate (“ontologically indepen-
dent”) instead. Nevertheless, it’s stylistically harmless to speak of posits
even if they’re taken to exist in no sense at all, and so for smoothness
of exposition I’ll do so. In the following, I distinguish (1) quantifier
commitments, those commitments of a theory that are due to its existen-
tially prefixed implications, and that tell us what posits a theory has, from
(2) ontological commitments, those commitments that we take the theory
to have because of those of its posits that, in addition, are ontologically
independent.2
It’s important to realize that the (epistemic) distinctions I draw later
between types of posits are independent of any particular ontological po-
sition that may be taken about those types. As I’ve already indicated, one
can disagree with me about whether mathematical abstracta exist or not
and yet (1) agree entirely about the epistemic distinctions among posits
that I draw and (2) agree entirely about where mathematical abstracta
(and fictional items) fall given those distinctions. This methodological
insight about how the ontological and epistemic aspects of my position
don’t stand and fall together is facilitated by a choice of terminology
(“posit”) that I treat as ontologically neutral.
I distinguish posits by whether they have epistemic burdens and, if so,
what kinds. Two points about such burdens: First, in calling them episte-
mic burdens, what I intend to stress is that success in satisfying them
provides a prima facie reason to think that the posits exist (and, indeed,
provides the best prima facie reasons we have for thinking such things
exist). Second, such epistemic burdens can’t be translated into “analytic
entailments” that hold of the general terms designating the posits. This is
because such burdens are defeasible: It’s always open to us to change the
epistemic classification of a posit (and, indeed, the classification of all the
posits falling under a term) and thus to change whether it has epistemic
burdens or not (and which kinds).
Let’s first consider those posits that have no epistemic burdens at all.
Ultrathin posits are what I call such items, and they can be found—
although not exclusively—in pure mathematics where sheer postulation
reigns: A mathematical subject with its accompanying posits can be cre-
ated ex nihilo by simply writing down a set of axioms;3 notice that both
individuals and collections of individuals can be posited in this way.
2
As I indicated at the outset of this chapter, I’ll largely keep to the viewpoint that the
ontological commitments of a theory are to be found among its quantifier commitments.
Later, however, I’ll consider relaxing this condition on the grounds that what a theory
should be taken to be committed to—because of the ontological independence of the things
in question—may be found among other linguistic paraphernalia of a theory, i.e., its predi-
cates.
3
See my 1994, esp. pt. II. Ultrathin posits, of course, include fictional entities.
128 Applied Mathematics and Its Posits
4
The reader may wonder if there are any posits that are thin in my earlier 1994
sense—that only have Quinean rent as an epistemic burden. I don’t think so, as I argue
later—and I’m partially indebted for my views on this to a discussion with Aaron Lipeles
back in 1995 or thereabouts.
Posits and the Epistemic Burdens They Bear 129
The difference between thin posits and ultrathin posits (which live
free of charge) is striking. Should one of the former fail to pay its Quinean
rent when due, should an alternative theory with different posits do better
at simplicity, familiarity, scope, fecundity, and success under testing, then
we have a reason to deny that the thin posits, which are wedded to the
earlier theory, exist—thus, the eviction of centaurs, caloric fluid, ether,
and their ilk from the universe.5
Thick posits labor under a different epistemic burden: They’re items
to which we have “thick epistemic access.”6 An individual’s (or group’s)
thick epistemic access to a posit can be purely sensory in nature (someone
looking at something or listening to it), can involve individuals and in-
struments (microscopes, radar), or can involve a complicated social net-
work of individuals and instruments (e.g., computers) interacting (as a
group) with something, where there is a division of labor (e.g., the opera-
tion of a linear accelerator).7 Regardless, there seem to be four conditions
on thick epistemic access:
(1) The results of thick epistemic access to something are largely
independent (epistemically speaking) of what the recipient(s) ex-
pects from that access (I call this robustness).8
(2) There are means of adjusting and refining thick epistemic access
to the things being detected (refinement).
(3) Thick epistemic access to things enables tracking of them (either
in the sense of detecting what they do over time or in the sense
of taking time to explore different aspects of them) (monitoring).
(4) (Certain) Properties of the objects can be used to explain how
the kind of thick epistemic access we have to them enables
the discovery of (possibly other) properties of those objects
(grounding).
Amplifications
In what follows, I comment on conditions (1) − (4) of thick epistemic
access by considering details about the two kinds of thick epistemic ac-
cess: on the one hand, ordinary observation by the use of our naked
senses, and on the other, instrumental interventions with items otherwise
outside of our sensory range.
5
The reader may suspect that centaurs were evicted simply because we didn’t find any
to begin with. That’s usually not quite enough of a reason to rule something out if it’s an
ontological belief we’ve inherited, although it’s a strong consideration. Additional theory is
often needed to convince us that our failure to find something isn’t merely bad epistemic
luck. Here is where the Quinean virtues come into the picture.
6
See my 1997b and 2000b, p. 116. Much of the following discussion on thick episte-
mic access is drawn (and modified) from those sources.
7
See, for nice historical examples, Galison and Hevley 1992.
8
“Robustness” is a term that often comes up in philosophy of science; one shouldn’t
assume that it means here what it means elsewhere.
130 Applied Mathematics and Its Posits
On Robustness
The independence in this sense that robustness supplies includes—but
isn’t restricted to—independence from one’s theories about what one is
thickly accessing, as well as independence from one’s theories about how
that thick access itself operates.9 When thick access is a matter of pure
observation, the independence of its results from what we believe or ex-
pect arises from the nature of our ordinary evidential practices with re-
spect to observation. (To illustrate: If I want to prove the existence of
something to someone skeptical of it—e.g., moldy bread in the refrigera-
tor—and it’s possible to point at it, I can do so immediately.) This inde-
pendence is nevertheless compatible with the following facts: We can be
wrong about what we think we see; we can be trained to see things we
couldn’t see before; we may need a bit of theory in order to understand
what it is we’re supposed to be seeing, or how seriously we should take
what we’re seeing. And finally, there is no denying that there can be states
of mind—great emotion, drugs, mental illness—where the independence
of our sensory access to things from our beliefs about those things is
compromised.
When thick epistemic access operates via scientific instruments of
some sort, its robustness is (in part) due to the antecedent robustness of
observation. This is because the use of such instruments always involves
transduction of the machinations of what we’re thickly in touch with into
something we can observe. If a particular instrumental intervention with
something regularly leads to a certain result, this means that any of a
certain class of ordinary observations occurs when the device is operated.
For example, a common tool for recognizing the presence of certain acids
(in certain forms) is litmus paper, and dipping a bit of it into a liquid can
result in a change of color that we can see. So, too, more sophisticated
devices may yield flashing lights, certain sorts of visual images, sounds,
and so on. There is no requirement, however, that any particular sort of
instrumental intervention must be accompanied by a particular sort of
observation (a Geiger counter makes clicks; something else that measures
the same thing may produce a photograph); I’m only saying that when
an instrumental interaction is robust, it achieves this (in part) via a link
to observation.
But this link to ordinary observation isn’t the only source of the ro-
bustness of (instrumental) thick epistemic access to something. A source
of robustness, in all cases of such access, is an unavoidable limitation on
confirmation holism; that is, it’s due to our inability to bring scientific
theory directly to bear upon the events involved in thick epistemic access,
whether that access be a matter of pure observation or a matter of instru-
9
I must stress that this independence is never total. Theory can, often in unpredictable
ways, have an impact on our understanding of our thick epistemic access to something. The
point is that, as a matter of empirical fact, such impact is fairly limited.
Posits and the Epistemic Burdens They Bear 131
On Refinement
When it comes to ordinary observation, we’re quite familiar with the
many ways that we can (and do) instinctively adjust and refine our thick
epistemic access to something: We can squint to see things more clearly;
we can move in for a closer look; we can reach out and touch. Thus, we
have resources to help distinguish what we’re “really” seeing from what’s
only an artifact of our senses (or of how those senses track items in the
world). How do you tell that “floaters” are in your eyes and not several
inches in front of you? By engaging in various experiments that enable
you to determine that the motions of such items are due to the move-
10
These regularities are cruder than the truths of pure scientific theory in at least two
respects: First, they’re often limited in scope—applicable when instruments are used in cer-
tain specific situations but inapplicable, even with the same instruments, in other (sometimes
only slightly) different situations. Second, they’re usually deductively independent of other
such regularities; i.e., they’re usually established empirically on an individual basis, without
there being much of a hope that they can be connected (deductively) to other regularities
or, for that matter, to anything else.
11
See my 2000b, pts. I and II, for extensive discussion of gross regularities, in particu-
lar, for details on their epistemic relation to scientific doctrine, and for examples.
132 Applied Mathematics and Its Posits
ments of your eyes. Notice that these ways of refining, adjusting, and
understanding our sensory access are robust: They’re (largely) indepen-
dent of theories held about why such methods of refinement work.
Instrumental access to theoretical entities is entirely analogous: One
reason for refining our instrumental access to theoretical objects is the
achievement of robustness; the tinkering needed for this often goes on
independently of our theories about the objects accessed or of the theo-
ries about the instruments used (we simply “fiddle” to reduce “noise” of
various sorts). But, as with pure observation, crucial to this process of
refinement is also the recognition of artifacts of the instrumental process
itself, and here theory can play—although it doesn’t always—an impor-
tant role. What we see under the microscope might be an artifact of how
we dyed the sample and what effect that has on light. We may recognize
this either because we know (on the basis of a theory) that the item in
question doesn’t have the structures we seem to be seeing, or because we
can isolate this artifact by using other instrumental means of access to the
object under scrutiny. Because of the epistemic need to detect and elimi-
nate artifacts in our means of access to something, a significant amount
of experimental design is dedicated to the elimination of such artifacts.12
On Monitoring
One nice property of thick epistemic access is how it enables its recipient
to “track” the properties of something. For example, you can watch
something for several hours (a certain sort of insect, say), and in this way
determine a variety of its properties (that it moves at a certain speed, that
it can change colors, that it likes moss). Furthermore, you can determine
what it does over time; you can construct, that is, an “episodic history”
of its actions in real time.
Instrumental access to theoretical entities allows the same thing. His-
torically, in fact, the capacity to track a theoretical entity instrumentally
12
A classic discussion of this with respect to the microscope is in Hacking 1981 (re-
printed in his 1983). Two nice case studies where concern with the identification of artifacts
is also explicit are Franklin 1993 and Weiskrantz 1986. Also see my 2000b, pt. I, esp. § 6,
and also pp. 175–6. It’s worth pointing out that this sort of study doesn’t just take place
with respect to thick epistemic access to the micro-objects physicists, say, are concerned
with. The psychological literature, for example, often raises the worry about whether one
or another psychological notion has been sufficiently or correctly operationalized. What’s
innocently indicated in these contexts by the use of this (philosophically scary) word is only
the concern with whether the experiments in question have established a rigorous enough
connection, means of access, between the experimenters and the phenomenon they hope
to study; or, in other words, whether what they trust they’re measuring is, in fact, being
measured. As an illustration (there are many), see Bills 1981 on worries about the self-
concept, and Cunningham and Tomer 1990, p. 384, on fluid intelligence.
Similar worries arise in social psychology and sociology—when the phenomena in
question are macro-items that one must get to by means of statistical tools; consequently,
statistical artifacts are a notoriously ever-present danger.
Posits and the Epistemic Burdens They Bear 133
On Grounding
We can often connect certain properties of the objects we can see with
our ability to learn about their properties through sight; for example,
given our (partial) understanding of our observational means of access to
something, and given what properties we take the seen object to have,
we can understand how we’re able to recognize, by sight, its color, speed,
and so on. This need not be a subtle scientific matter (although it often
is): It’s easy to explain why I can see how fast a flock of antelopes is
moving—they’re large opaque objects that don’t travel very fast (even
when panicked). So, too, it’s easy to explain my troubles seeing how big
certain jellyfish are under certain circumstances: They’re (more or less)
transparent; but, consequently, I know I won’t have a problem seeing
how big they are if I manage to dye them first. Thick epistemic access
to theoretical items otherwise impervious to sensory detection is entirely
analogous on this point.
The four conditions on thick epistemic access aren’t all on a par as far as
their epistemic status is concerned. Grounding is different from the other
three conditions because it depends more directly on empirical assump-
tions about how thick epistemic access works. Imagine, for example, that
the old Leibnizian view is right: Causation isn’t metaphysically real; in-
stead, there is a divine coordination of perception of event with event
perceived. Even in such theologically loaded situations, thick epistemic
access would be just as epistemically useful as it is for those of us here
under more Godless circumstances. Grounding, however, would lapse:
13
Also see Brown et al. 1997. In that article, we find: “The third jet in these ‘three-
jet events’ is now recognized to be due to the emergence of an energetic gluon in a process
analogous to bremsstrahlung in QED. Although it took a few more years to make an abso-
lutely convincing case, this visual evidence for gluons was perhaps the most influential factor
in the acceptance of QCD as the correct theory of the strong interactions” (p. 21, emphasis
original).
134 Applied Mathematics and Its Posits
14
One thing that makes this hard—but not impossible—to imagine is that the possible
world depicted (not to mention the beings in it) has to be one that is itself a great deal
simpler than our world is.
Posits and the Epistemic Burdens They Bear 135
And what if these aspects of our body of beliefs didn’t cohere? There
are three possibilities. First, we could find ourselves changing our episte-
mic practices (successfully). This has happened: The (real) possibility of
generalizing observation to thick epistemic access is something we discov-
ered only around the beginning of the twentieth century. Second, we
could find ourselves changing our view of ourselves and the world. This
has happened, as well—as the history of optics from the ancient Greeks
to modern times indicates. Third, and least appealingly, we could acqui-
esce to the inconsistency.
“Inconsistency” is perhaps too strong a word. One aspect of the dis-
comfort with standard interpretations of quantum mechanics is that mea-
surement has to be interpreted within classical physics despite the funda-
mental physical theory (of the small) being nonclassical. This radical
discontinuity (“collapse of the wave function”) seems a given of quantum
mechanics as traditionally viewed, and this explains the dualistic mind/
body feel of, for example, the Copenhagen interpretation, and may also
explain the tendency of some physicists—and popular expositors of quan-
tum mechanics, as well—to sense compatibility between quantum me-
chanics and certain “spiritualistic” views. Nevertheless, there is clearly a
strong desire not to acquiesce to this state of affairs if we don’t have to.
And, views have emerged that seem capable of explaining the classical
physics of macro-objects from within a purely quantum-mechanical per-
spective. If any of these are successful, this will make our body of beliefs
coherent (in the broadly logical sense), because then the same physical
laws will govern both our naturalized epistemology and what there is at
what we (currently) take to be the most fundamental level.
Given that grounding does hold of thick epistemic access, one impor-
tant implication is that items to which we have thick epistemic access are
ones the properties and relations of which play a role in the epistemic
stories we tell about how we know about them and what means of thick
epistemic access we can have to them. I call this epistemic obligation of
thick posits their epistemic role obligation.15
The required epistemic stories, about how instrumental thick episte-
mic access from us to posits is possible, are ones told by scientists (in various
subdisciplines) about how their instruments, and their other means of
interactions with the world, function. So, for example, there’s the story
physicists tell about how the instruments they’ve developed to measure
acoustical phenomena operate.16 There’s also the story that psychologists
15
As just shown, any claim that any empirical posit whatsoever must satisfy an epistemic
role obligation must turn on grounding always holding of thick epistemic access. In any
case, it’s not true that satisfying the reliability and nontriviality conditions of chapter 4
requires a fulfillment of the epistemic role obligation.
16
In Strasberg 1991, we find an explicit discussion of a number of measurement tools
(various kinds of microphones and hydrophones, the Rayleigh disk, fiber optics, etc.), their
range, sensitivity, and other pertinent properties, in the specific context of physical theory.
I.e., a certain group of causal relations between us and certain phenomena (in this case,
fluctuations of pressure, temperature, density, etc., of matter) are studied.
136 Applied Mathematics and Its Posits
borrow from physiologists about how our visual systems work. These sto-
ries are invariably incomplete, as stories often are in the sciences, but the
important point is that they’re epistemic stories—they’re clearly designed
to indicate how we know what we know about something—how, as far
as the science of a particular time can tell us, the means of access that we
have to things operates.17 For this reason, the study of empirical phenom-
ena, and thus empiricism as a practicing method among scientists, isn’t
correctly described by merely saying observation (or, more generally,
thick epistemic access) is crucial. Rather, it’s that a systematic study is also
required of the means of access to empirical objects, along with concomi-
tant worries about artifacts of that access.18
Because posits can be both thick and thin, posits sort into four types:
Ultrathin posits, thick but not thin posits, thin but not thick posits, and
those that are both thick and thin.
17
What makes it especially clear that these stories are epistemic stories is precisely the
focus on correcting mistakes mentioned in the section titled On Refinement above. Notice
that nothing of this sort exists when it comes to the purported epistemic access to abstracta
of various kinds. This, on the one hand, illustrates how thick epistemic access is sensitive to
the requirements of the reliability and nontriviality conditions of chapter 4 and, on the
other, puts in a stark light how purported epistemic access to abstracta is not. I give more
details about this shortly.
18
Neither observation nor thick epistemic access (which generalizes upon observation)
is a metaphysical black box that achieves its epistemic deeds by magic. They’re both as
transparent to scientific inquiry as any other aspect of our universe.
Posits and the Epistemic Burdens They Bear 137
“Turtle,” for example, we take to refer to every turtle whether it has been
sighted by anyone or not. “Electron” we take to pick out electrons every-
where, regardless of whether they’re the (relatively few groups of) items
we’ve forged thick epistemic access to, the slightly larger number of elec-
trons we could have forged thick epistemic access to but didn’t, or the
numerous electrons that we’ve no hope of forging thick epistemic access
to. These posits are thin.19
Two Points
First, thick epistemic access is a relation between a community and ob-
jects during a time period. This means that not only are new instances of
the thick epistemic access relation forged between us and objects to which
we didn’t earlier have thick epistemic access, but also that the thick episte-
mic relation to items can be lost later (e.g., Roman artifacts that we dis-
cover later in time, other items—speeding atoms, rabbits bounding in
the brush—that leave the narrow domain within which we can observe
them or otherwise instrumentally detect them).
Second, aren’t all the items referred to by “turtle” either thick or thin
posits? No: We can be wrong about what we have thick epistemic access
to—we think those things are turtles, but they’re not. Second, we can be
wrong about thin posits, too: We think there are no turtles on Pluto, for
example, but we could be wrong about that. More generally, we could
be wrong about the theory of something we use for determining that
those somethings exist (out of our ken).
19
“Turtle,” of course, refers to turtles, whether we take them to exist or not. But the
thin posits that are turtles are those turtles that we take to exist on the basis of our theories
about turtles even though we haven’t got thick epistemic access to them. Are there such
turtles? Presumably; see the discussion starting several paragraphs from now.
138 Applied Mathematics and Its Posits
20
A criterion for what we take to exist, not, I stress, a criterion for what exists.
Posits and the Epistemic Burdens They Bear 139
types: We discover a new kind of frog, and then we’re convinced of the
existence of other frogs of the same kind.21
But this suggestion, too, seems open to a potential counterexample.
It’s taken to be a certainty that there are numerous kinds of viruses to
which we haven’t yet forged thick epistemic access. (Indeed—and try not
to be too upset by this—new kinds of viruses are quietly mutating on this
very page as you read.) The same is true of kinds of insects, plants, and
even mammals: We take it to be certain that there are kinds of insects,
plants, and even mammals no instance of which we’ve yet forged thick
epistemic access to.22
Why do we think this? There are two reasons. The first involves the-
ory: We know, on the basis of theory, how variable viruses, in particular,
can be, how fast they can evolve, and how widespread they are. Similarly,
we know how evolution works and that isolated species restricted to local
environments can easily develop, and that such unexplored environments,
although vanishing rapidly (e.g., the Amazon rainforests), are still with
us. The second is empirical: We read, almost on a daily basis, about dis-
coveries of new species of animal and plant. A similar point can be made
about chemical compounds: We know—without thick epistemic access
to them—that there are kinds of chemicals that we haven’t discovered
yet. So we’re firmly committed to the existence of items, even to the
existence of items of specific kinds, where we don’t have thick epistemic
access to (any of) the instances of those kinds.
The key to understanding what’s going on here is to recognize, first,
that the notions of kind and subkind implicitly at work in our positing
thin posits are ones generated by theory.23 Consequently, it’s theory that
tells us, when faced with a kind, examples of which we’ve thick epistemic
access to, whether we know enough about that kind (and its subkinds)
to take ourselves to know that there are thin posits falling under that kind
(and its subkinds). In the case of viruses, theory licenses the belief not
only that are there specific (individual) viruses that we haven’t thick epi-
stemic access to, but also that there are kinds of viruses that we haven’t
thick epistemic access to either. On the other hand, until relatively re-
cently, despite our acquaintance with the planets in our solar system, we
simply didn’t have enough theory to tell us that there are other planets
ellipsing around other stars to which we lacked thick epistemic access.
21
Although not always—and this shows how theoretical considerations are used to
convince us (or not) of the existence of thin posits. If this frog has very specific needs and
can, as a result, flourish only in very specific habitats (which, say, have been eliminated by
one or another oil company), we may not believe there are any other frogs of this type to
be found. We may take him, tragically, to be the last of his kind.
22
Could we be wrong about this? Of course—we can be wrong about anything. But
the point is that we take ourselves to know that there are kinds of things that exist no
instance of which we have thick epistemic access to, just as we take ourselves to know, e.g.,
that there are electrons.
23
See my 2000b, esp. pts. III and IV.
140 Applied Mathematics and Its Posits
24
More generally, those causal effects taking place within us and to us that we’re un-
conscious of, because, say, their effects are too small for us to perceive, are ones to which
we don’t have thick epistemic access—at least not through introspection or perception.
Posits and the Epistemic Burdens They Bear 141
25
At least, I don’t have thick epistemic access to water molecules via my unaided
senses. The point of scientific instrumentation is to change our causal relations to the world
and, as a result, to change what we can have thick epistemic access to.
I realize that what I’ve claimed, about how what we have thick epistemic access to is
individuated, is controversial: Among other things, it implies that the various indeterminacy
doctrines held by many philosophers are false—at least insofar as those doctrines require
that the causal relations we forge to the world are insufficient to distinguish, say, the differ-
ent individuation conditions associated with “rabbit” and “rabbit-fusion.” See my 2000b,
pt. III, §§ 2–3, or my 2003, where further argument is offered on behalf of causation and
how it enables us to individuate what we interact with.
142 Applied Mathematics and Its Posits
As I’ve said already, apart from the epistemic burdens posits have, it’s an
additional question what our ontological view of these things should be.
Given the arguments of part I, it’s a somewhat delicate matter to know
about things that we take to exist. This is because things that we take to
exist must not be made up out of whole cloth, and so they must satisfy
the reliability and nontriviality conditions. Nicely enough, thick posits,
items to which we have thick epistemic access, seem to easily navigate this
particular epistemic hurdle. Robustness and grounding (as defined in items
(1) and (4) in chapter 6) show that such objects can’t be like fictional
items or even like the contents of hallucinations—made up out of whole
cloth—because (a) we can’t stipulate the properties we attribute to things
to which we have thick epistemic access, and (b) the properties of things
to which we have thick epistemic access play a role in the explanations we
give for how we have thick epistemic access to them. Fictional items vio-
late both (a) and (b), and the contents of hallucinations, although they
satisfy (a), violate (b). On the other hand, the epistemic role obligations
that objects (which we can thickly access) have because of grounding, and
the refinements and adjustments of thick epistemic access that refinement
concerns, enable us to satisfy the reliability and nontriviality conditions
with respect to thickly accessed objects. Thin posits, too, are items that
satisfy (a) and (b) because the properties we attribute to them are theoret-
ically bound to the thick posits that fall under the same predicates they
fall under. Ultrathin posits, by contrast, seem nicely designed to be items
that exist in no sense at all.
143
144 Applied Mathematics and Its Posits
1
See, e.g., Putnam 1987, lecture 1, esp. p. 20.
146 Applied Mathematics and Its Posits
mains an ontological “black box.” We need to know that the entities are
ontologically independent, and all that the uniqueness of a theory tells us
is that we have no choice (for any number of reasons—including the
possible failure to have developed certain mathematical breakthroughs)
in what theory we adopt (and what that theory is quantifier committed
to). But absence of choice in theory construction—which, after all, is still
a global property of the whole theory—does not imply the ontological
independence of the entities that the theory quantifies over. Inference to
entities on the basis of a theory without thick (or thin) epistemic access
to those entities is thus illegitimate.
Here’s another application of the black box objection: Certain philos-
ophers have stressed that a powerful factor in explanation is unification.2
If a theory offers a pattern of argument that can be widely and successfully
applied to explain (and predict) phenomena, and/or if the postulation of
a “theoretical structure” (Friedman 1983, p. 236) with accompanying
existential quantification leads to a substantial unification of the kinds of
phenomena that can be described, then the theory has, in Friedman’s
words, “unifying power.” And not only is this an intrinsically appealing
property for theories to have, but if a theory U unifies various empirical
applications of more specific theories Sp1, . . . , Spn under itself, then it
garners additional confirmation for itself over and above the various more
specific theories; indeed, Sp1, for example, garners more confirmation for
itself by virtue of its deduction from U than it would all on its lonesome
(because, via its attachment to U, it additionally gains the confirmation
accrued to Sp2, . . . , Spn).
This much I won’t argue with. Friedman (1983), however, draws a
further ontological conclusion from this. Simplifying slightly,3 the claim
goes something like this: If, in order for U to unify Sp1, . . . , Spn, a kind
of entity must be co-referred to (so that it appears in each Sp1, . . . , Spn —
after possibly reconstruing some of the specific theories Sp1, . . . , Spn,
and, of course, it also appears in U), then “we have no choice but to take
it literally, to assign it a rightful place in the ‘world’ of physical reality”
(p. 250).
Friedman contrasts how a posit’s playing a role in unification in this
sense should commit us to the physical reality of that posit versus the
remaining posits in the theories in question, which are to be treated as
merely part of the mathematical formalism accompanying Sp1, . . . , Spn,
and U. But the reader should be suspicious of this move on three counts:
2
Two prominent examples are Philip Kitcher (e.g., his 1981) and Michael Friedman
(e.g., his 1983, chaps. 6 and 7, esp. pp. 236–50). Both philosophers acknowledge the
recognition of this factor as significant for explanation on the part of earlier philosophers of
science such as Hempel, Whewell, and others. See Kitcher 1981, and Friedman 1983, p.
242, n. 14. In what follows, I focus exclusively on Friedman’s discussion.
3
Friedman focuses on “structures”—such as absolute rest—not entities committed to.
For our purposes, this doesn’t matter: Such structures translate into entities committed to
as well as (certain) truths about those entities.
Posits and Existence 147
4
Friedman does seem to be helping himself to a refined version of an indispensability
argument: Terms essential to how U is linked to the theories it unifies are the ones we’re
to take as picking out real pieces of the world. But, really, this doesn’t do anything more
to show that such terms refer to what’s ontologically independent than do considerations
based on the ordinary indispensability argument.
148 Applied Mathematics and Its Posits
5
Of course, if there were numerous subatomic particles—far more numerous, I mean,
than, say, the number of hadrons—and we regularly discovered new ones on the basis of
our theory, and our theory was one that was extremely well confirmed, the epistemic situa-
tion would change. Consider, however, the following from Brown et al. 1997:
A major goal of the PETRA storage ring, and of a similar, competing collider at SLAC
named PEP . . . was to discover the top quark required by the emerging Standard
Model. At the time, phenomenological models suggested that the top quark’s mass
might fall in the 10–20 GeV range accessible at these machines, but early experiments
failed to find any evidence for its existence. So firm was the belief in the Standard
Model by 1980, however, that this failure did not lead to serious doubt about the
theory’s validity. Physicists widely supposed that the top quark had to be more massive
than could be produced at PETRA and PEP, and that it would eventually be found at
higher energies. (p. 23)
Notice that it isn’t that the existence of the top quark is, in principle, out of the range of
empirical verification: It’s that it’s out of the range of specific colliders—and in drawing this
conclusion, physicists are contravening only phenomenological models.
6
Mental objects and events, on some views of these (as chapter 3, “Some Important
Illustrations,” indicated) are posited solely via logical form; so are fictional objects (on some
views of these) and Lewis’s possible worlds. Regarding the last, it’s worth noting that our
thick epistemic access to, as it were, our world, hardly suffices to satisfy the above require-
ments as far as other possible worlds are concerned. In any case, logical-form considerations
won’t do for such ambitious ontological purposes, as I argued in chapter 3.
Posits and Existence 149
posits fall neatly into two categories: Either they’re a kind of item that
fails to have close specific theoretical ties to items we have thick epistemic
access to, and then they’re simply ultrathin posits, or they’re a kind of
item that satisfies the constraints on thick and thin posits, and then
they—the mathematical posits themselves—are thick or thin (or both).
The interesting question, to which I’ll turn in chapter 8, is which ones
are which, and how we tell.7
Ontological Identifications
From now on, I identify the following three types of posit: (i) ontologi-
cally independent posits, (ii) causally efficacious posits,8 and (iii) posits
that exist. Any of these notions could, in principle at least, diverge in its
extension* from those of the others, but I’ll argue that they all coincide.
I’ll also identify, in what follows, (i′) posits that are ontologically
dependent on linguistic or psychological processes, (ii′) posits that don’t
exist, (iii′) ultrathin posits,9 and (iv′) causally inert posits.
How are the first set of identifications motivated? What we already
have in place is the identification of ontological independence (in our
linguistic community, anyway) as the criterion for what exists. This, in
turn, leads to a characterization of how we identify what we take to be
ontologically independent of us, that is, the reliability and nontriviality
constraints of chapter 3. An excursion into (naturalized) epistemology
then reveals that the only way we have of learning about a kind of thing
ontologically independent of us is to forge thick epistemic access to some
of its instances. In turn, and this engages us in metaphysical characteriza-
tions of what we can take to be ontologically independent of us, it’s
7
I’m making a claim here that will be further substantiated during the course of part
II. After all, it may seem to some that space-time points, say, are neither thick nor thin in
my (current) senses of these terms, nor ultrathin. Later I’ll undercut the cogency of this
option.
8
The identification of type (ii) with types (i) and (iii) commits me to what has come
to be called the Eleatic Principle. See Oddie 1982, where this term was coined. Also see
Armstrong 1978 and Colyvan 2001.
9
A terminological caveat intrudes here: Consider the term “chimera”—as it is taken
to refer* to the mythological creaure—and imagine that we (or the ancient Greeks, anyway)
once believed in these things. In saying (as we now do) that “there are no chimeras—in
this sense,” we’re treating “chimera” as a thick or thin posit that has (unequivocally) failed
to live up to its epistemic obligations. We can also treat the term as ultrathin. Then we say,
“Chimeras are mythological beasts believed in by ancient Greeks.” So the distinction be-
tween ultrathin posits and thick (or thin) posits that have failed at satisfying their epistemic
burdens, is a soft one. This is no surprise because nothing at all rides on drawing sharp
distinctions in this way between kinds of posits we take not to exist in any sense at all. The
status of a term like “chimera” in usage turns on pragmatic issues—whether we’re informing
a person about truths that involve the term “chimera” or whether we’re trying to prevent
a misguided view about chimeras—as so understood—that takes them to be real. I’ll pre-
sume that any term, purporting to be thick or thin, that unequivocally fails at fulfilling its
epistemic burdens is reclassified as ultrathin—but as I’ve said, nothing much turns on this.
150 Applied Mathematics and Its Posits
determined that such items are causally efficacious—for it’s only by devel-
oping causal connections of the right sort to things that we can forge thick
epistemic access to them.10
It seems clear that the foregoing considerations aren’t strong enough
to provide the metaphysical identifications (i) − (iii), because although
ontological independence and existence should be identified if the consid-
erations in part I are correct, the identification of these with causal efficacy
needs additional argument: It seems it’s in principle possible for there to
be something ontologically independent (out there) and yet causally idle,
that is, without causal powers of any sort.
On the other hand, some philosophers may think that in adopting
the Eleatic Principle, the claim that everything that exists has causal pow-
ers, one is engaged in a kind of induction that’s commonplace in ordinary
science: Scientists are willing to presume—on the basis of quite limited
data—that such and such theoretical constraints exhaust the possibilities
of what there is. Scientists are quite sure, for example, that nothing can
move faster than the speed of light; this kind of claim, of course, is always
subject to the condition that our theories are right (one can be wrong
about anything), but that condition aside, there’s no reason why meta-
physical claims about causal efficacy should be treated differently from
other claims in science.
A way of resolving this is to consider more closely how certain sorts
of ontological conditions are ratified in science: Until now, I’ve been
almost entirely concerned with how ontological commitments in science
are established; I’ve argued, for example, that thick epistemic access to
instances of a kind of item are needed before we’re willing to claim that
items of that kind exist.11 But there is another aspect to ontology that
bears on the issue at hand, what I’ll call ontological closure conditions. The
periodic table in chemistry is an excellent example of such: Not only does
it supply a list, and rules of thumb, governing what elements there can
be, and what properties they’re likely to have, but it also excludes, quite
explicitly, the possibility of other sorts of elements. It’s easy to see how
it does this: The atoms of such elements are individuated by the number
of protons in their nuclei; because protons are discrete items, this puts
severe constraints on what elements are possible.12
Something similar happens with subatomic particles: A sort of “peri-
odic table” is also established of what sorts of properties can vary among
the fundamental constituents of these things (charge, spin, flavor, etc.),
10
Although we don’t, in practice, forge thick epistemic access to every instance of a
kind, we’re still justified, on the basis of the theory about such a kind-term, in inferring that
the remaining instances (the thin posits falling under that kind) are also causally efficacious.
11
This claim, of course, is subject to the considerations about subkinds and thin posits
discussed in chapter 6.
12
In particular, there is, at most, one element corresponding to each positive integer.
I’m not, of course, speaking of muonic or exotic atoms.
Posits and Existence 151
and this puts constraints on how many kinds of subatomic particles there
can be.13
Scientific theory is clearly used to supply these closure conditions:
Only such things can exist. Why “can” (and not “do”)? Because theory
often can’t go on to tell us that therefore such things do exist: As I’ve
argued already, what’s additionally needed is thick epistemic access to
(some of) the items of the kind in question before we’re convinced of
that.14 And it’s easy to see why theory, in general, should not be allowed
to provide ontological credentials without the involvement of thick epi-
stemic access. Apart from the fact that any theory could be wrong,15 theo-
ries don’t invariably tell us that what the theory allows to exist must exist
simply because other constraints may come in that exclude the possibility
the theory seems to allow.
Consider again the case of the periodic table: That table is simply a
(periodic) ordering of possible elements on the basis of the number of
protons in a nucleus and the corresponding number of electrons in the
shells surrounding that nucleus. But perhaps, if we worked out the funda-
mental force laws governing the contents of such nuclei, and discovered
tractable mathematics that enabled us to deduce from those laws what
nuclei are possible, we could have discovered that certain numbers of
protons in nuclei are ruled out. That is, specific theories can often be
dictated to by broader and more fundamental theories that rule out possi-
bilities the more specific theory allows.16
13
Brown et al. 1997 write: “The model is referred to as ‘standard’, because it provides
a theory of fundamental constituents—an ontological basis for describing the structure and
behavior of all forms of matter (gravitation excepted), including atoms, nuclei, strange parti-
cles, and so on” (p. 3). That this “periodic table” doesn’t include gravitational interactions
is a main motivation for the exploration of string theory. Indeed, attention is centered on
closure conditions, such as this periodic table of fundamental particles, when theories
change. The introduction of string theory allows additional sorts of objects to which (in
principle) thick epistemic access can be forged (although it still limits what’s possible). One
way, thus, to provide evidence for string theory over its more restrictive competitors is to
forge thick epistemic access to these predicted items.
14
We find in the table of contents of the September 1998 issue of Scientific American
(p. 5) under Making New Elements: “If they exist, elements 114 and beyond may prove
surprisingly stable.”
15
Does the fact that a theory can be wrong contribute to why thick epistemic access
plays a more fundamental epistemic role than do theoretical considerations in what we take
to exist? If so, why should it (since we can be wrong about what we think we have thick
epistemic access to, just as our theories can be wrong)? The reasons for this turn on condi-
tions (1) − (4) of thick epistemic access (defined in chapter 6), which make it epistemically
more reliable, in general, than scientific theory.
16
A neat example of this is supplied by evolutionary biology and chemistry. That, e.g.,
carbon-based life is possible, and that lithium-based life is not, is dictated to biology by
chemical theory. That other metal compounds can play a similar role in blood that hemoglo-
bin plays in the blood of mammals is also dictated by chemistry, and it also tells us that
these other compounds won’t be as efficient, and that therefore we’re unlikely to find them
in anything other than invertebrates.
152 Applied Mathematics and Its Posits
17
The same principle is at work in building a house well. One cannot design houses
to withstand every possible disaster—but better-built houses withstand ordinary adverse
events better than do ill-built ones.
Weinberg (1993), when considering “the final theory” in physics—the theory that will
explain everything—worries about the possibility that the question “But why is this theory
true?” will still be with us. He hopes to make the final theory more satisfying than a mere
brute fact by showing
that the final theory, though not logically inevitable, is logically isolated. That is . . .
although we shall always be able to imagine other theories that are totally different
from the true final theory . . . the final theory . . . is so rigid that there is no way to
modify it by a small amount without the theory leading to logical absurdities. In a
logically isolated theory every constant of nature could be calculated from first princi-
ples; a small change in the value of any constant would destroy the consistency of the
theory. (p. 236)
It’s hard to see why a logically isolated theory (in this sense) is more satisfying as far as
explaining why things are like this (rather than like that) is concerned. All a logically isolated
theory explains is why, given that the theory is true, we couldn’t have come up with some-
thing else (given the data and the mathematical tools we luckily had access to). But a logi-
cally isolated theory does have genuine epistemic advantages: Even with small differences in
the data, we’re still forced to the same theory. Only an epistemic catastrophe, something
that shows us that we’re radically mistaken in a fundamental way that implies, among other
things, that we can’t even preserve most of the evidence we’ve gathered in support of that
theory (or of its successors), would overthrow a logically isolated theory.
Posits and Existence 153
18
The reductionist and supervenience literature is gigantic. For details on my views
about the various kinds of reductionism, why I reject them, and some references, see my
2000b, pt. I, § 3. Also see n. 20. For an excellent presentation of the issues surrounding
supervenience, see Kim 1993.
19
Some versions of physicalism involve the view that what there is is in some sense to
be supplied (solely) by physics. I have rashly and in passing word processed remarks support-
ive of such a view (e.g., in my 2000b, pp. 45–6). I say a little more about my current views
on this in chapter 9.
154 Applied Mathematics and Its Posits
My second claim is more radical and is one that I won’t fully argue
for here: (2) that the previous condition is the only condition that can be
placed on the sciences with respect to one another, and that nevertheless,
such a condition suffices for a commitment to “physicalism,” the view
that in a perfectly respectable sense, physics, in principle, gives us (all) the
laws for everything there is.20
For our current purposes, the second claim isn’t needed because (1)
any physicalist is at least committed to my first claim (and a number of
nonphysicalist positions are committed to it, as well), and (2) in any case,
my first claim is strong enough to support the metaphysical identification
of causal efficacy with ontological independence and existence that I want
to argue for.21
Here’s how to do it (the quick and dirty way): Ontological closure
conditions from physics tell us that physical laws apply (in principle) to
everything there is. But anything causally inert won’t be amenable to
physical law; therefore, such things are ruled out.
Caveats: First, some may be worried that quantum mechanics, Bell’s
theorem in particular, directly undermines causality. It doesn’t—what it
20
Some argument for this position may be found in my 2000b, pt. I, § 3. Cosmides
et al. (1992) offer the following consistency principle on the sciences, which they call “con-
ceptual integration”: “[T]he various disciplines within the behavioral and social sciences
should make themselves mutually consistent, and consistent with what is known in the
natural sciences as well. . . . A conceptually integrated theory is one framed so that it is
compatible with data and theory from other relevant fields” (p. 4). They claim that the
natural sciences already obey this principle, but they fault evolutionary biology, psychology,
psychiatry, anthropology, sociology, history, and economics for not doing so. They write:
“As a result, one finds evolutionary biologists positing cognitive processes that could not
possibly solve the adaptive problem under consideration, psychologists proposing psycho-
logical mechanisms that could never have evolved, and anthropologists making implicit as-
sumptions about the human mind that are known to be false (p. 4).” This principle strikes
me as equivalent to the one I’ve just offered, except for the following caveat that I insist
upon: It’s perfectly acceptable for a special science to be (entirely) constituted of laws that
are “idealized” in such a way that (1) we recognize these laws (one and all) to be strictly
false (from the vantage point of the truths and laws of a more fundamental science), and
yet (2) we find the special science, as formulated, indispensable because specific application
of the broader principles (of the more fundamental science) to the specific domain of the
special science is intractable. Also, it is allowable for the laws of a special science to be
quantifier-committed to things that, strictly speaking, don’t exist. Arguably, linguistics (or
a great deal of it, anyway) is such a special science on both counts.
This caveat means, as I’ve indicated, that although one mustn’t allow such a special
science to impede the implementation of a more fundamental science to the former’s do-
main of application, one shouldn’t fault a special science merely for positing processes that
don’t exist or for invoking principles “known to be false.” So too, and for the same reasons,
one shouldn’t fault a special science merely for positing entities that fail to be definable in
terms of or supervenient upon the entities (of an underlying science) that we take to be the
only ones that “really exist.”
21
Many philosophers try to establish the Eleatic Principle not via a consideration of
the methodology of science, as I’m doing now, but by the use of purer philosophical tools
such as general considerations about the structure of inference to the best explanation or
Occam’s Razor. I’ve already (in chapter 4) given reasons for thinking this route won’t work.
Posits and Existence 155
22
Cartwright (1989, chap. 6) argues that it doesn’t undercut the common cause prin-
ciple: Other assumptions needed to draw this conclusion go first. The issue is delicate
enough and complicated enough to require much more than I can do now, as the very
large literature on this topic indicates. So I’m only staking out my position on this (without
argument).
Another way that quantum mechanics may seem to undermine causality is that mea-
surement, on the Copenhagen view, for example, requires the “collapse of the wave func-
tion.” What this means is that an otherwise completely deterministic process (when “mea-
156 Applied Mathematics and Its Posits
with what exists? Something to the effect that nothing exists that falls
outside the nomenclature of our theories. Motivating such an ontological
closure condition is quite difficult. Suppose (contrary to fact) that we had
an ontological closure condition for the kinds of items admissible in the
language of physics.23 Then we would still require (a) that the posits that
we take to fall under the nomenclature are either thick or thin, and (b)
that by virtue of (a), for every other sort of posit p (of the special sci-
ences), either (b1) we have thick epistemic access to p, or (b2) we have
enough of a theoretical grip on p so that it’s a thin posit.
This strikes me as implausible in the extreme. For example, thick
epistemic access to subatomic particles plus access to the theory of such
things doesn’t by itself suffice either for thick epistemic access to the liz-
ards composed of those subatomic particles or even for access to a biolog-
ical theory of lizards, and this is for reasons that I’ve already discussed:
(1) Thick epistemic access to something doesn’t suffice—usually—to
supply thick epistemic access to objects that thing is a part of (or vice
versa), and (2) the intractability of the theory of subatomic particles pre-
vents the deduction of theories of macro-objects.
Colyvan (2001), I should say, offers his own objection to the induc-
tive strategy for establishing the Eleatic Principle:
[T]here are many other properties that the uncontroversially real entities
share and that the uncontroversially unreal entities lack. All the uncontrover-
sially real entities are spatio-temporally located, for instance. . . . The reality
of space-time points hangs crucially on whether it’s causal efficacy or spatio-
temporal location that is the important property. Similarly, we could opt for
the property of having a positive rest mass as the mark of the real and again
the demarcation would be different. (pp. 41–2)
This gives a false picture of the options available for the common proper-
ties held by everything we take to exist. In particular, it seems quite odd
to take, for example, positive rest mass to be a criterion for what exists.
Light is hardly like the contents of hallucinations, or like numbers, for
that matter, and so it’s hard to see why it should be assigned a status on
sured”) randomly jumps to a new state in a way that’s not characterized by the dynamical
laws (Schrödinger’s equation) governing that wave function. I think this can be shown,
contrary to appearances, not to undermine causality, either, but my reasons for thinking
this also take us too far afield.
23
I did not presuppose such a thing, note, to establish the Eleatic Principle. Such a
closure condition would require a limited nomenclature in physics from which all the other
terminology of physics would be definable. I doubt this would be possible even if, say, we
discover the “theory of everything.” In brief, this is because the applications of physics to
what exists allows that application to be to items characterized in ways that fall outside
physics—such will always be necessary because of the introduction of idealizations to make
the application of physics tractable—and such idealizations can (and usually do) involve
posits that needn’t be strictly part of the “theory of everything.” See, for numerous exam-
ples of this, Truesdell 1991.
Posits and Existence 157
a par with such things.24 Space-time points are another matter: If they’re
the posits of one or another branch of pure mathematics, then, of course,
they’re like fictional objects, as I’ve already argued. But if we’re speaking
of the posits of an empirical theory, then it remains to be determined
whether they’re ontologically independent or not. Surely the simple as-
signment of a location for them won’t confer ontological independence
to them all by itself.25
The foregoing considerations about ontological closure conditions,
although of interest in their own right, have been brought on stage at
this point in order to defend the identification of causal efficacy with
existence, specifically, to rule out the possibility of causally inert objects
that nevertheless exist. But there is a second way that the identification
of causal efficacy with existence can be troubled: if Kant’s challenge26 is
brought against it. I’m not concerned with the attribution of causal effi-
cacy to fictional entities in fiction: As I mentioned when I first raised
Kant’s challenge, attributions to fictional entities, at least when they occur
within fiction, needn’t be taken to be true. The same, however, isn’t true
if such attributions are made to nonexistent entities within the context of
applied mathematics, because in that case, the attributions, as I showed
in chapters 1 and 2, must be taken to be true. I’m anticipating an analysis
that I carry out later in this book, but it’s my contention that point-
masses and space-time, for example, don’t exist. Nevertheless, it may
seem that we attribute causal powers to such things—that, for example,
space-time (in the context of general relativity) itself has energy, and that
point-masses, in the context of Newtonian physics, exert gravitational
force upon each other.
My response to this may look like a proverbial biting of the bullet
(but, I think, it really isn’t). I first claim that attributions of causation are
factive: When we describe something as the cause of something else, we
commit ourselves to the existence of the first something (and the second
something, for that matter). This is, perhaps, not particularly controver-
sial, and so I won’t dwell on it. My second claim should also not appear
particularly controversial—but it will lead to something in need of further
argument pretty quickly: This is that there aren’t any idioms in physics
that, all by themselves, anyway, betray the presence of causation. Rather,
causation is something imposed upon that formalism when we apply it. In
saying this, I’m not making the Humean-sounding claim that causation
is a projection onto the world—or onto the scientific description of that
world—and isn’t therefore in that world. Rather, I treat causation much
24
Recent experiments have been carried out to determine whether (certain types of)
neutrinos have positive rest mass. It’s odd to learn that it’s “uncontroversial” that what was
actually being determined is whether those types of neutrinos exist or not.
25
The contents of hallucinations, notoriously, are often unpleasantly located in space
and time (e.g., the malevolent trolls standing right in front of you . . . while, that is, you’re
under the effects of certain drugs), but this doesn’t confer existence on them either.
26
See chapter 4.
158 Applied Mathematics and Its Posits
27
Well, it does anyway, on my construal of how it operates (see chapter 3). There is a
close analogy here: To evaluate what others are committed to, we must focus not on what
their existential quantifier actually picks out but on what they take it to pick out (what
existential sentences they’re committed to—not which existential sentences are true). Simi-
larly, when evaluating ontological commitments via an existence predicate, we must focus
on what they take that existence predicate to pick out, not what it actually picks out.
Quine’s approach and the existence predicate approach look more disanalogous than they
Posits and Existence 159
beliefs and with our own earlier bodies of beliefs (which we’ve subse-
quently deserted), because thick access, as I’ve indicated, in our own case
is factive (and notice that the terminological choice to make thick episte-
mic access factive doesn’t cause problems with our own body of beliefs
because we take ourselves to have really observed what we think we’ve
observed). But things come apart when we describe others as having been
mistaken about what they’d thought they’d observed, because if someone
thinks she’s seen elves, then that someone is committed to elves—regard-
less of whether she’s actually seen elves.
Notice how this differs from how we evaluate inferential mistakes: If
someone makes an (invalid) inference to the existence of Bs from a body
of doctrine, we’re prone to say that such a person only thinks he’s com-
mitted to Bs—not that he’s really committed to Bs. But observation, as
I’ve just noted, is different: If someone only thinks he’s seen elves (but
really hasn’t) then he is committed to elves.28 So it’s the nonfactive use
of observation (and thick epistemic access) that’s relevant to determining
what those committed to an alien body of beliefs are committed to believ-
ing exists—what they take themselves to have observed (and instrumen-
tally interacted with), not what they’ve actually observed (and instrumen-
tally interacted with).29
Why not then adopt the nonfactive use of “observe” and “thick epi-
stemic access” and thus allow ontological commitment to be captured in
both alien and domestic bodies of beliefs by instantiations by the same
predicates (“thick” and “thin”)? The answer is because it’s important for
our kind terms—except when they’re in a restricted class of terms that
pick out artificial items of a certain sort—to be “criterion transcendent,”
to be terms that we can be mistaken about the extensions of.30
actually are only because where a kind-term is involved, one is apt to interpret it according
to how it has been translated (by our lights) rather than by what native speakers take it to
pick out.
28
Here is yet another example of how ordinary intuition treats observation as epistemi-
cally more fundamental than inference.
29
This difference between first and third person with respect to the evaluation of onto-
logical commitments corresponds to a similar difference between first and third person ob-
servation-attributions. Someone can say, when describing the effects of a drug: “The drug
is working: I’m really seeing elves,” even if her judgment is unaffected and she knows there
are no elves in front of her (and that she’s hallucinating). It’s much less comfortable to say
this in the third person: “She’s really seeing elves,” even with the stress where I’ve placed
it, is too close to ontologically underwriting the claim that there are elves. Notice we can’t
say that “she thinks she’s seeing elves,” either—that’s only appropriate if she thinks she
might be seeing actual elves. We have to say, “She’s hallucinating elves” or something like
that.
30
See my 2000b, pt. IV, § 3, for details.
8
Applying Mathematics:
Two Models
I turn now, and pretty much for the rest of the book, to a study of a
small number of restricted (but informative) examples of applied mathe-
matics. But first we should ask if a general characterization of how mathe-
matics is applied to an empirical domain can be given. I think one can,
and measurement theory is a nice initial guide.
I first present (a slightly modified) version of a description from mea-
surement theory of how mathematics is applied, give some examples, and
describe some aspects of this approach. I then offer a different character-
ization of how mathematics is applied, one that can handle the other sorts
of cases of applications of mathematics that I want to talk about.1
To this end, let’s start with an algorithmic system S (Peano arithme-
tic, hereafter PA, is a good example),2 and let’s also assume we have a
nonmathematical characterization E of some empirical domain, where
this (in practice, not axiomatized) characterization is one of a collection
of objects, as well as relations and properties holding between and of such
objects.
Taking a tip from Krantz et al. (1971), a sufficient condition for the
applicability of S to E is this: There is a (multiply valued) mapping, g,
which in general isn’t an “onto” mapping, and which (1) maps the con-
1
Some of the material to follow is drawn and modified from my 2000a, pp. 216–20.
2
The reader, however, can use any axiom system as an example of an “algorithmic
system.” Axiom systems are language based, but I understand algorithmic systems to be
broader in scope: Any system of notation, language-based, diagrammatic, etc., with a decid-
able proof-recognition procedure is an algorithmic system. See my 1994, pt. II, where algo-
rithmic systems are described, and my forthcoming(a), where Euclidean geometry is charac-
terized as an algorithmic system that is (at least partially) not language based.
160
Applying Mathematics: Two Models 161
Here are two initial examples: (a) Let E be the domain of line drawings
on a very large piece of paper supported by a (relatively) flat surface. Let
S be Euclidean geometry. (b) Let E be a set of straight rigid rods suscepti-
ble both to a “concatanation” relation (a placing of rods end to end) and
to a “less than or equal” relation (understood to apply to rods with re-
spect to their lengths). Let S be the language of addition and ordering
on whole numbers.
Elucidations
1. The reader familiar with Krantz et al. (1971) may be more struck by
the differences between their approach and mine than with the supposed
similarities. Measurement theory officially takes homomorphisms of em-
pirical domains into (intended) models of mathematical systems as its
subject matter. But this is less of a difference than it appears. A character-
ization of an empirical domain often involves a rich set of properties and
relations taken to hold of objects in that domain, some of which are
ignored when a particular branch of mathematics is applied to it. (E.g.,
when considering the measurements and sums of the lengths of rigid
rods, we ignore their colors.) This practice, of course, can be finessed in
standard measurement theory by considering only a subset of the proper-
ties and relations taken to hold of items in an empirical domain. On the
other hand, the application of a mathematical subject matter to an empiri-
cal domain often involves treating distinct items in a domain in mathe-
matically similar ways (e.g., two rigid rods can be taken to have the same
length). By mapping the empirical domain to the mathematical domain,
the standard approach need only concern itself with single-valued func-
tions that aren’t one-to-one rather than with multiply valued functions
that aren’t mapped onto the range, as with my approach. There isn’t
much of a difference, mathematically speaking.
A second apparent difference between standard measurement theory
and my approach is that measurement theory is officially an application of
model theory; that is, it’s not mappings of languages used to characterize
empirical domains to languages of (applied) algorithmic systems that are
studied in measurement theory, but mappings of the items (and proper-
ties and relations of such items) to objects (and properties and relations
of such objects) in intended interpretations of algorithmic systems. For
example, instead of studying, say, mappings of the language describing
rigid rods to the language of number theory, what are studied are map-
162 Applied Mathematics and Its Posits
pings of rods and their lengths to numbers and relations among them,
for example, the addition relation.
I’ve no objection (in principle, anyway) to this model-theoretic ap-
proach, provided it doesn’t give the (wrong) impression that using mod-
els to characterize how mathematics is applied empirically presupposes
the existence of mathematical abstracta. After all, it’s not the language
and theorems of mathematics that are described in applications (on this
approach); it’s the mathematical objects themselves (plus their relations
and properties) that are so described. I think the model-theoretic ap-
proach doesn’t presuppose mathematical objects in any case although it
certainly presupposes a (meta-)language that’s quantifier committed to
such things. As I showed in part I, pending a criterion for what a dis-
course is ontologically committed to that weds ontological commitment
to existential commitment, this won’t threaten a commitment to ab-
stracta. Despite this surety, it’s perhaps informative that an analysis of the
application of mathematics can be executed solely in terms of algorithmic
systems and not via purported mathematical objects.
2. E’s characterization of an empirical domain need not be one we
take to be “right,” or “true,” or even particularly accurate as a description
of the ontology of the phenomena to which S is applied. To illustrate,
consider an application of Euclidean geometry to ink figures on a flat
piece of paper. What’s the ontology of the empirical domain here? Not
implausibly (given current scientific views), we might take it to be one of
subatomic fields in a relativistic space-time. Leaving aside the sticky ques-
tion of exactly how the implicit unification of the quantum-theoretical
domain and that of relativistic space-time physics is to be achieved, we
obviously don’t recognize the applicability of Euclidean geometry in this
case either by way of a mapping of this poorly understood physicalist
ontology to that of Euclidean geometry (as on the first approach) or by
way of a mapping of the language of Euclidean geometry into that of the
language of fundamental physics (as on the second approach). For one
thing, we can apply Euclidean mathematics to figures on a piece of paper
without knowing anything more than that the empirical domain contains
figures on a surface (piece of paper). One way or another, we’ve done
exactly this for over 2,000 years—if not on paper, then on papyrus, or in
sand, or whatever. Apart from this, even were we much more knowledge-
able about the micro-ontology I’ve invoked, it’s not obvious that the
macro-objects (line drawings) that we treat in a Euclidean manner are
well-defined model-theoretical constructions of this underlying physicalist
ontology. I allude here to a host of familiar problems, including those
about how to individuate macro-objects in terms of subatomic items (of
any sort).3
3
I’ve described this last concern in model-theoretical terms. Of course, my own fa-
vored syntactic approach holds out even less promise in this respect—because what’s then
required are definitions of macro-terms in the language of fundamental physics.
Applying Mathematics: Two Models 163
4
Artifactual irregularities due to the dye on a sample or because of the properties of
light interacting with a microscope can nevertheless be counted, or their border geometry
described and measured.
164 Applied Mathematics and Its Posits
5
See, e.g., Friedman 1983, esp. pt. III, or Stein 1967.
6
A qualification: Velocity-sensitive forces may prove of value within the context of
Ncm, but I’m waiving this consideration in the above discussion in order to stress the
ontological irrelevance of this issue in any case.
Applying Mathematics: Two Models 167
7
Because point-masses are included among the admissible mathematical objects,
“Newtonian cohesive-body mathematics” may sound inaccurate. Regard point-masses as
degenerate examples of cohesive bodies.
8
The mass of a cohesive body is given by a mass density function defined over the
volume of that body. See Truesdell 1991 for details.
9
E.g., see Truesdell and Rajagopal 2000, where we find:
A body β is a set that has a topological structure and a measure structure. It is assumed
to be a σ-finite measure space with a nonnegative measure µ(ᏼ) defined over a σ-ring
of subsets ᏼ of β called subparts of the body. The open sets of β are assumed to be
the σ-ring of sets. The members of the smallest σ-ring containing the open sets are
called Borel sets of β. (p. 1)
and body forces;10 they’re also taken to have constitutive relations, which
govern how such bodies move and how they move and deform under the
effects of such forces.
Two points: First, we may discover that the empirical objects that
rational continuum mechanics successfully describes are ones to which con-
stitutive relations apply successfully (within a certain degree of accuracy)
because what corresponds empirically to these constitutive relations can
be described as various forces between, say, the molecules that such em-
pirical bodies are composed of and, ultimately, the forces between sub-
atomic particles. But this study—with its ultimate revelation of other
force laws in nature—necessarily moves beyond the Newtonian frame-
work.11 For my purposes, however, and staying within Ncm, we can still
regard constitutive relations as corresponding to (idealized) descriptions
of the materials that the empirical bodies (corresponding to cohesive bod-
ies) are composed of. In turn, composition out of these different materials
can be seen to induce various kinds of (resistive) (internal) forces.12 Sec-
ond, in claiming this, by no means am I claiming that a “reduction” of
continuum mechanics to the (quantum) physics of subatomic particles is
in the offing, or that, therefore, in some sense, a reduction of continuum
mechanics to the physics of point-masses, say, is also in the offing. Apart
from the fact that subatomic particles aren’t point-masses, reductions in
this broad sense are extremely rare in the sciences.
Has a family of algorithmic system been specified by the foregoing?
A family of algorithmic systems as Euclidean geometry and PA are under-
stood to be such systems? Well, apart from my being rather vague about
the specific axioms supposedly involved here, the answer is: Of course.
There is no restriction on which concepts can occur in a family of algo-
rithmic systems (no concept is “purely mathematical”). And just as group
theory has many models amenable to study, and classes of which can be
singled out by additional axioms, so too Ncm has many models or collec-
tions of models that can be similarly studied. The mathematical nature of
the latter subject matter is shown by the absence of concern over whether
any of these models is empirically “right”; that concern is to be raised in
10
Body forces are related to the mass density of such bodies taken over their volume.
Contact forces are related to the density taken over the body’s surface area.
11
It needn’t have, of course: Empirical results forced us beyond something compatible
with the Newtonian framework to quantum mechanics.
12
Not adopting this last move leads to a terminological denial of Newton’s first law.
See Truesdell 1991, p. 66. For my purposes, it really doesn’t much matter how constitutive
relations are understood because I argue in chapter 9 that, just like forces, they dissolve, in
any case, into sets of mathematical constraints on possible motions of the material points
contained in cohesive bodies—not, i.e., into something we need take ontological account
of (unless we desert the context of Ncm).
Applying Mathematics: Two Models 169
exactly the same way as the success of the application of Euclidean geom-
etry was raised earlier in this chapter.13
Now imagine an empirical world a bit more dreary than ours: There
are just (fairly) arbitrarily sized (and shaped) cohesive lumps, made up of
various materials, some more rigid than others, that travel endlessly through
space, some deforming, some not, and that occasionally break apart into
smaller chunks or deform under gravitational forces (e.g., during rota-
tion) or when they come in contact with one another. (They can also
join up, of course.) Instead of what empirically corresponds to electro-
magnetism, weak and strong forces, and the zoo of fields of subatomic
particles with their complicated laws that govern our world, there are just
what empirically corresponds to the concepts in Ncm of various contact
forces and one body force (gravity). These lumps are also made up of
different materials that affect how they react to various external and inter-
nal forces, and these correspond to various constitutive relations in Ncm.
The empirical language E contains names for some of those lumps and
some general predicates and relations that hold of them: “x is spherical,”
“x is more massive than y,” “x has more volume than y,” “x is farther
away from y than w is from z,” and so on.
The quantifier commitments of E are all and only such lumps. Ncm
is applied directly to the lumps in a way similar to how Euclidean geome-
try is applied to figures on paper: Find a useful number of categories of
cohesive bodies, composed of (idealized) kinds of materials, for example,
that we can map classes of lumps to, and then subsequently describe the
kinematic and dynamic properties of these classes.14
Applying Ncm allows us to characterize the properties of such lumps
by approximation methods, treating them as (a class of) cohesive bodies
to within a certain margin of error: approximating the amount and distri-
bution of the mass in such lumps, their volume and surface areas, and so
on, by mathematically manageable functions of various sorts,15 the ways
13
For explicit axiomatizations of (some of) analytical mechanics and (some of) contin-
uum mechanics, see Noll 1974; for references to other axiomatizations, and for some hu-
morous, and yet nevertheless nasty, polemics, see Truesdell 1980–81.
14
In Truesdell 1991, we find: “The approach is like that of Euclidean geometry, in
which, after statement of the axioms satisfied by all geometric objects, theorems characteriz-
ing and relating classes of figures are proved” (p. 200). He then adds: “Since mechanics is
a discipline vastly more subtle and sensitive than geometry, the parallel stops here and does
not extend to the theorems themselves or even to the methods of constructing proofs.” As
an illustration of this, consider Truesdell’s 1991 gloss on constitutive relations: “[Constitu-
tive relations] define materials, which are mathematical idealizations of the materials en-
countered in nature. . . . Typical constitutive classes are the bodies called rigid or solid or
fluid, isotropic or anisotropic” (p. 6).
15
“Mathematically manageable” includes, but isn’t restricted to, smooth, i.e., differenti-
able as many times as needed.
170 Applied Mathematics and Its Posits
that such lumps perform when acted on by external forces, various ways
in which the parts of the lump deform and move, by idealized constitutive
relations, and so on.
An easy example: If a lump is sufficiently far from the other lumps,
we can approximate its gravitational impact on the other lumps by ignor-
ing nearly everything about it, assuming, that is, that all its mass is con-
centrated at its center of mass. If, however, that lump strikes another
lump, we’ll have to take internal properties into account—in what detail
and how, what can still be ignored, and what not, is empirically deter-
mined by the behavior of the lumps under impact.
We can imagine an easy epistemic generalization of this case (d),
where the lumps recognized and categorized in the language of E (via
observation) are instead so recognized and categorized (at least some of
the time) by instrumentation of some sort. In this generalization, we’re
still taken to have, that is, thick epistemic access to the objects of E, and
we recognize the properties of such objects via that access.
For example, consider certain properties of the planets and other
bodies in our solar system. We have substantial access to these properties
by means of various instruments, not by naked sensory observation, and
the kinds of instrumental access available enable us to determine these
properties in great detail. This access, however, allows us to apply empiri-
cal terms to these bodies, where such terms are already active in ordinary
life.16
In both examples (c) and (d), although E has its own set of terms
for the lumps and their properties that it allows us to talk about, and
although there are methods for gaining access to the (empirical) proper-
ties and relations (characterized by E) that lumps have, to apply Ncm to
the existential commitments of E we reconstrue those commitments as
quantifier commitments of Ncm (reconstrue lumps as cohesive bodies)
and reconstrue the empirical properties and relations of lumps as the ma-
thematized properties and relations of those cohesive bodies.
There are two ways to ring changes on the Ncm examples: The first
(e) is if there are properties of lumps their reconstrual in Ncm leaves out
altogether (and not merely because their effects are minimal for purposes
of certain applications of Ncm). For example, such lumps might be col-
ored in various ways without the character of this coloring being amena-
ble to an explanation in terms of a physics based in Ncm. In this case, the
original model for the application of mathematics still applies: There is a
mapping of the language of Ncm to that of E that maps cohesive bodies
to lumps, and maps 3 to nothing at all. By the way, I don’t mean that
physical space is a void devoid of properties and that 3 is mapped to
16
E.g., we have good descriptions of the rough volumes of many of these bodies, and
their makeup—how much water is present, what gases their atmospheres are composed of
(if they’ve got atmospheres), the temperature variations on the surfaces of such bodies (if
they’ve got well-defined surfaces), etc.
Applying Mathematics: Two Models 171
that; I mean that 3 isn’t mapped to anything at all. I say more about
this in chapter 9, in the section titled “Absolute Space and Time.”
The second possibility (f) is where we have no antecedent language
E of lumps to start from. Imagine, instead, a domain to which we have
only instrumental access (not observational access), and it’s discovered
that Ncm applies to the empirical patter that our instruments detect. Sup-
pose we find that certain cohesive bodies—and notice I’m now using the
mathematical language of Ncm to describe the situation—have detect-
able properties (their “locations,” their “velocities,” and their “distances”
from each other, say),17 and we’re able to predict their future relations
and properties on the basis of initial data our instruments discover about
them plus the application of Ncm. Imagine also that the language used
is the pure language of Ncm and that we find that such mathematical
properties as location, velocity, and so on, can be correlated with instru-
mental detections. In this last case, no antecedent E is to be had. And, as
footnote 17 indicated, this is no mere thought experiment: Exactly this
kind of situation obtains with the physics of subatomic particles and their
properties.
It’s worth comparing this case with a previous one. In example (d),
although instrumental access is as essential to our determination of the
properties of the objects under scrutiny as it is here, that instrumental
access is to macro-properties to which we have (more or less) observa-
tional access in other contexts. The recent discovery of substantial
amounts of water on Mars, for example, discovered by purely instrumen-
tal means, is a discovery that may well be exploited in the near future by
Martian colonizers who will just see the stuff.
But this is false of the properties in the case under current consider-
ation: These properties that instruments enable us to access are properties
we never have observational access to. This difference is important be-
cause in the first case the (rough) scopes of the predicates denoting the
properties in question are indicated not just by our instrumental access to
their presences, but by our observational access to their presences, as well.
Why does this matter? I’ll say more later18 but, roughly, the point is
this: We generally don’t allow an instrument (or observational means, for
that matter) to define the extension of a noun phrase or predicate (to pick
out a kind of thing or a property). But the role of instrumentation with
17
Why the scare quotes? Well, what could happen is similar to what has happened with
respect to the properties of subatomic particles. “Spin,” e.g., although analogous to “spin”
as applied to macro-objects like tops, can’t be the same thing. We might similarly know
that, in fact, the objects under study can’t be massive lumps careening through space with
real “locations,” “velocities,” etc., because the so-called “lumps” are items detected via
instruments in a context that excludes this interpretation; e.g., they correspond to intensive
properties of a phenomenon and aren’t distributed (in space) at all. Nevertheless, we might
still find Ncm empirically adequate; i.e., we might find that what it predicts is empirically
verified to whatever degree of preciseness we’re capable of testing it to.
18
Actually, I’ve said quite a lot already. See my 2000b, pts. III and IV.
172 Applied Mathematics and Its Posits
19
In describing this as empirical, I’m explicitly allowing that it could change.
20
“In some way” is vague. I’m alluding to my discussion of ob- and ob*-similar sorts
found in my 2000b, pt. II, § 3. My purpose in so alluding is to forestall the thought that
the above contrast is a simple one where either instrumental access is to something that is
otherwise observable, or it’s not. Things are, as always, more complicated. Many of the
instrumental ways we have of determining the presence of various atmospheric gases, for
example, are linked “in some way” to observation, and linked so that we’re secure about
coining (projectable) terms that refer to such gases. But this isn’t because these gases are
straightforwardly observable.
Applying Mathematics: Two Models 173
21
But it mustn’t be forgotten that this does not mean that all of the distinctions the
applied mathematical terminology brings with it have been fitted with this empirical role.
On the contrary, as I show in chapter 9.
174 Applied Mathematics and Its Posits
existence of such content, and wants to claim that such nominalist con-
tents are true and the rest of the mathematical accompaniments are not,
faces the considerations raised in chapters 1 and 2.22
Nevertheless, the position on ontology developed in part I, and in
chapters 6 and 7 of part II, has resources for distinguishing ultrathin
posits from thick or thin posits even in these cases, and consequently,
such a position has resources for making pertinent ontological distinc-
tions (between which posits in such a physical theory should be taken to
exist and which not) even though the language of that physical theory
is through-and-through mathematical, and even though, therefore, the
(physical) theories in question must quantify over ultrathin posits. I illus-
trate this project explicitly in chapter 9 with respect to examples (a) − (f )
described in this chapter.
Next, I note again, in order to set up the concern with the success
of applied mathematics, that the kind-terms, the “projectable predicates,”
to use deliberately Goodmanian language, must be borrowed directly in
case (f)—and the numerous other cases like (f )—from (pure) mathemat-
ics. This may not seem to be the case in a more general sense: Ncm, after
all, uses notions such as “cohesive body,” “velocity,” and so on, which
seem borrowed from the empirical world, even if they have been modified
severely (“idealized,” as it’s put) in the process of that borrowing. In-
deed, it may seem clear that all the notions of Ncm, including that of
“space-time point,” arise fairly directly from empirical notions in this way.
So although the application of Ncm in case (f ) is to an empirical arena
from which the notions of Ncm did not originally arise, still, the notions
(ultimately) have an empirical source.23
But this needn’t be the case. Quantum mechanics, for example, uses
the notion of a complex Hilbert space, where the vectors in question
are complex wave functions. These complex-valued functions, although a
crucial part of the quantum mechanical formalism, aren’t themselves di-
rectly given a physical interpretation. Rather, it’s the results of various
mathematical operations on such wave functions (e.g., integrating the
product of the wave function, its complex conjugate, and some other
function—the latter depending on what, precisely, is being measured)
that yields, in the standard interpretation of this formalism, probabilities
of position, momentum, and so on, for particles. The latter are called,
for example, by Landau and Lifshitz (1977), “physical quantities.”
22
Such a nominalist faces another problem to be raised in chapter 9: Any restriction
of the predicates and noun phrases to what’s taken to be real results in an intractable theory:
In order to be able to say what’s true about what’s real, we have say things about more (our
terms have to refer to more) than what’s real.
23
I’m going along with this view only for the sake of argument: Long familiarity with
points makes us think of them differently from how we think of, say, the complex wave
functions in a Hilbert space that I’ll mention in a moment. But I don’t think that points,
really, are any more empirical looking than are complex wave functions. I don’t, e.g., (really)
know how to visualize one any more than the other.
Applying Mathematics: Two Models 175
24
See my 2000a.
25
Two points: First, I did not respond to this second worry in my 2000a; I’m doing
so now. Second, Steiner (1998) doesn’t always clearly distinguish the two worries.
176 Applied Mathematics and Its Posits
26
Steiner (1998, p. 10), who deliberately hints of his inclination toward theism, is
therefore inclined to think that “dumb luck” in the form of divine kindliness is involved.
Notice that the “dumb luck” worry undercuts an inductive justification for continuing to
use what we’d otherwise regard as unjustifiable strategies for designing empirically adequate
theories: Why assume that such luck will continue? (Think of Job, and what happened to
him.)
27
Of course, there is no escaping the fact that, in certain cases, success is a matter of
dumb luck, even in intellectual endeavors. But that can’t generally be the case with respect
to theory-constructing strategies, can it?
28
Steiner (1998) doesn’t couch the issue this way, but he’s out to show that scientific
methodology is “nonnaturalistic” and thus presupposes a “user-friendly” universe. That’s
equivalent to the concern as it’s put here.
29
This definition is unfortunate in seeming to imply that if we can’t describe an empiri-
cal domain in purely nonmathematical language, and therefore, if we can’t give physical
motivations for a theory in purely nonmathematical language, then (by definition) any con-
siderations about the properties of that domain will be “Pythagorean.” But Steiner’s exam-
ples make clear that he doesn’t want to press the Pythagorean charge so cheaply. He is con-
cerned with cases, as his examples show, where a theory is modified on grounds connected
to “internal” mathematical values (ease of computation, aesthetic beauty, etc.). The concern
thus turns out to be similar to the traditional charge leveled at the coherentist theory of
truth: Why should (internal) considerations of simplicity, coherence, etc., suffice to make a
theory true of what’s independent of us? The internal considerations are parochial, but what
the theory is supposedly about is not. (Also see chapter 4.)
Applying Mathematics: Two Models 177
30
See Steiner (1998, chap. 4). In giving these examples, he notes how often intuitions
of elegance, and other parochial considerations about theories, are used by physicists as
indicators of the modifications being in the right direction, empirically speaking. Steiner’s
point, I want to stress again, is not that such modifications don’t lead to empirical success,
nor is it that such successes can’t be physically explained after the fact. His claim is that
the procedure makes no sense to begin with: Epistemically, we’ve no reason to think that
modifications in theories due to notational analogies or mathematical elegance will lead to
something empirically adequate. As I said, I’ll assume for the sake of argument that his
examples pretty much illustrate his claim that such (purely mathematical) analogies are oper-
ating. (See the quote from Yang in Steiner 1998, p. 70; the discussion of how Maxwell
modified Ampère’s law, pp. 77–80; Schrödinger’s equation on pp. 81–2, etc.) However,
one must be more careful than Steiner sometimes is about whether or not a purely mathe-
matical distinction has physical content. Steiner (1998) writes: “It is convenient—and
therefore customary—to use the concept of the Taylor series in classifying phenomena as
‘first-order’, ‘second-order’, etc. That is, suppose we have a function that we feel tells the
‘whole truth’ about nature. If we expand it as a series, we can say what nature would be
like if we ignore terms beyond the first-order, second-order, etc., terms” (p. 69). He then
claims in the next paragraph:
Anyone who has studied the growth of functions realizes that if an empirical phenome-
non can be captured by a function expandable in a Taylor series (with a number of nonzero
coefficients, of course), then there can be practically no doubt that interestingly distinguish-
178 Applied Mathematics and Its Posits
able physical processes are at work, because the different terms in a Taylor series grow at
(exponentially) different rates. Such is hardly true of a classification of phenomena according
to decimal powers.
31
Steiner (1998, pp. 82–3) mentions cases where previously physically unreal solutions
to equations are reassessed: Such gives rise to the Dirac’s prediction of the positron and
Finkelstein’s black holes, for eample. Consider the following easy analogy: Imagine a rectan-
gle with two sides, x − 3 and x + 1, with area 21. Solving this problem in the obvious way
gives two solutions: x = 6 and x = −4, only the first of which is “physically real.” Imagine,
however, that we find a new empirical domain to apply “areas” to, where “areas” in this
domain are instrumentally detected in a certain way and are not spatial in nature. Given that
the same simple computational facts about areas apply to “areas,” and given that we can’t
tell a physical story for why the area rule applies to “areas,” why is it epistemically unjustified
to explore the possibility that the “physically unreal” answer has a role to play in this new
empirical domain?
Applying Mathematics: Two Models 179
his concern is with the easy and facile way physicists often have with a
mathematical formalism, manipulating and exploiting it in various ways
without possessing (sometimes any) mathematical justification for doing
so—the reasoning isn’t “mathematically rigorous,” as it’s often put.32 In
these cases, what one should be struck by isn’t that the resulting “mathe-
matics” is physically applicable but that the resulting “mathematics” can
subsequently be made into cogent genuine mathematics, because once an
epistemic justification is in place for tinkering with a mathematized physi-
cal theory (in order to produce something more empirically adequate),
there is certainly plenty of epistemic justification left for using proof-theo-
retic shortcuts and strategies in that process. Consider the Dirac delta-func-
tion: That such a thing was useful to manipulate formulas in quantum
mechanics was obvious (i.e., its empirical adequacy was obvious). That it
could be embedded in a genuine mathematical theory in a consistent
way was not. So, of course, the facile way physicists sometimes have with
mathematics poses a danger: An approach may subsequently prove math-
ematically incoherent. But this is a risk that all parties concerned are well
aware of—and it’s not enough of an epistemic danger to make facile ma-
thematizing something to be avoided.
Let me close this chapter with one last point about this matter:
Steiner claims that the methodological practice of treating, where possible,
refuted laws (or theories, for that matter) as special or limiting cases of
new laws (or theories) is “Pythagorean.” What Steiner means by this, as
he usually does, is that there is no antecedent epistemic justification for
doing so. The methodology, at least insofar as the aim is to generate an
empirically adequate theory, is no better than flipping coins.
You might have thought that the justification for this practice is easy
to establish: The problem with physical theories, always, is that it’s never
easy to apply them empirically. What’s required, always, are intricate the-
ory-specific applications of mathematical simplifications and idealizations.
That is to say, it’s not the case, when faced with a radically new theory,
and the question of how to apply it or to extract implications from it,
that one has a bag of tricks already in hand that tames the mathematics
of that theory. Now Steiner seems to notice just this point,33 but then he
writes: “[I]sn’t it just dogmatism (and certainly anthropocentric) to as-
sume that nature cares how much time we have invested learning mathe-
matical methods?” (p. 107). This misses the point. In deciding how to
modify refuted theories or how to generalize theories so that they have
greater scope, we only need epistemic justification for starting, as it were,
here rather than there. And certainly, that a particular modification of an
32
See, e.g., Steiner 1998, pp. 151, 161, 176.
33
Steiner (1998) writes: “Learning a physical theory means, above all, learning the
‘mathematical methods’ appropriate for the theory. This includes tricks for solving or ap-
proximating solutions for equations, and many other things. If a new theory yields, mathe-
matically, the old, then all of those mathematical techniques are still useful” (p. 107).
180 Applied Mathematics and Its Posits
Applied Mathematics
and Ontology
Given the ontological conclusions about ultrathin, thin, and thick posits
presented in chapter 7, the job at hand looks straightforward: We exam-
ine examples (a) − (f ) of applied mathematics (from chapter 8) and sort
the posits into the three categories. Nominalists then cheerfully commit
themselves only to the thick and thin posits. This glib summary is pretty
much right—the foregoing apparatus determines what we’re committed
to and what not when applying mathematics. Still, the route from theory
to application is never straightforward, not even in philosophy: Some of
this will be rough going because it’s sometimes controversial which posits
are which.
I start (naturally) with the easiest cases: The ontological commit-
ments that arise in these examples, those of (a) figures on pieces of paper
and (b) rigid rods, are easy to determine. We’re only committed to what
the respective languages E commit us to: straight rigid rods in the one
case, and figures on pieces of paper in the other—for, presumably, we
have thick epistemic access to both these sorts of items and take them to
be ontologically independent. The applied mathematics is a simple appli-
cation of a pure branch of mathematics, where an explanation of its em-
pirical success doesn’t require taking any of its posits as other than ultra-
thin. (The neatness of the ontological lessons of (a) and (b) is due to the
neatness of how the mathematics is applied to an empirical domain in
these cases: The language of E is logically independent of the mathemat-
ics applied to it.)
Most of this chapter addresses the first Newtonian cohesive-body math-
ematics (Ncm) example, (c), where classical continuum and point-mass
mechanics are applied to lumps careening through space. Doing this may
181
182 Applied Mathematics and Its Posits
The first Ncm example (c) is, ontologically, on a par with the (a) and (b)
figures and rods examples: We’re not committed to anything more than
E’s commitments—lumps of matter. That the applying of Ncm to E is
implemented by a reconstrual of lumps as cohesive bodies containing mate-
rial points doesn’t require there to really be such cohesive bodies contain-
ing material points or, in any case, there to really be material points. The
same is true of point-masses. Similarly, the need to use a coordinate sys-
tem of geometric points wherein the locations, velocities, and so on, of
these lumps are plotted doesn’t ontologically commit us to the existence
of any sort of purely geometric object—neither points, nor the various
sets of such points, nor the other two- or three-dimensional geometric
objects. Nor does it commit us to “absolute space,” a substantial some-
thing or other that lumps travel through. Finally, despite the presence of
Newton’s law of universal gravitation and Newton’s laws of motion, as
well as the other force laws that we apply to E, we’re not even committed
to forces: neither body forces (e.g., gravitation) nor contact forces. (In
short, the situation is as ontologically meager as any hopeful nominalist
will want.)1
A word about the argumentative strategy of this chapter: I first ap-
proach denying an ontological commitment to forces via an avoidance of
terms referring to such forces. This (linguistic) approach to ontology—
although quite widespread among philosophers—proves unsatisfactory. I
then switch to an approach motivated by the discussion in part I and in
chapters 6 and 7.
Forces
I suggested in chapter 1 that the appearance of a schematic letter appar-
ently standing for forces2 in Newton’s second law, F = ma, needn’t indi-
cate that linguistic items referring to forces occur in the sentences that
this schema stands for. Imagine first that instead of extended lumps there
are only a finite number of point-masses that the terms of E refer to (and
1
Nor can constitutive relations be construed as indicating commitments. Of course,
all this changes when we shift theories and shift our thick epistemic access to include the
internal constituents of lumps. But that’s another matter.
2
Although I primarily focus on forces in what follows, it’s easy to see that what I
claim, if right, applies to torques. It also applies to constitutive relations for the same rea-
sons—I say a little about this shortly.
Applied Mathematics and Ontology 183
that move about). In this case, instantiating Newton’s laws of motion and
Newton’s universal law of gravitation to this model results in differential
equations where no term for any force appears.3
Given the assumptions (1) that the ontological commitments of a
theory are all among the quantifier commitments of a theory (and distin-
guished from the other quantifier commitments by a special existence
predicate) and (2) that the appearance of force terms in the laws is as sche-
matic letters, one presumes to deny ontological commitments to forces.
It’s true that the differential equations describing the motions of the point-
masses are sensitive to pairwise relations between the point-masses: their
distances from each other and the amount of mass each point-mass car-
ries. To say this much is to recapitulate how Newton’s schematic law of
gravitation and his schematic laws of motion codify the differential equa-
tions governing this situation: There seems no point at which a commit-
ment to forces is made.
I shortly consider the situation if the schematic status of terms appar-
ently designating forces is denied. I should add, to avoid misunderstand-
ing, that in any case I’m open to the view that quantifier commitment in
a first-order language isn’t a necessary condition on ontological commit-
ment to something: In principle, there could be ontologically indepen-
dent items that we have thick epistemic access to but that are best codi-
fied, in a first-order theory, as predicates. In this case, I’d describe such
items as thick posits, but contrary to how the terminology of such posits
was set up in chapters 6 and 7, terms denoting them wouldn’t appear
among the quantifier commitments of the theory. I didn’t frame my dis-
cussion in those chapters to take account of this possibility,4 but for pur-
poses of analyzing the ontological commitments of a scientific theory (as
applied), it’s important to at least consider the possibility that genuine
commitments to items we methodologically treat as ontologically inde-
pendent of us are tucked notationally into the predicates of a theory as
well as into the noun phrases. I discuss this further shortly.
Let’s return to our discussion of forces. Fundamental forces, of course,
aren’t restricted to gravity. One point of schematizing Newton’s second
law is to allow it to apply to any force. But this doesn’t affect the forego-
ing considerations (as long as we remain within the context of applying
3
There are constants for the masses carried by each point-mass, for the initial values
of the distances between point-masses, and there is also a gravitational constant. There are
explicit time-dependent acceleration variables. Relative to the distinguished coordinate sys-
tem for absolute space and time, initial values for locations and velocities for those point-
masses at a particular time also appear, and by virtue of the mathematics of the calculus,
implicit time-dependent variables for velocity and position are present. That’s it. In Quinean
terms, quantification occurs over mathematical quantities, a physical constant, and several
variables concerning space and time. There is no room in any of this (first-order) language,
logically speaking, for terms that pick out forces.
4
It wouldn’t be hard to do so—but it seemed unnecessary given the arguments to be
established in chapters 6 and 7.
184 Applied Mathematics and Its Posits
5
That “forces” are “additive” amounts only to it’s being relatively easy, mathematically
speaking, to characterize the movements of cohesive bodies when more than one “force” is
involved: Each “force” acts—mathematically speaking—independently of the other
“forces.” This means that the differential equations that result from instantiating the various
force laws (plus Newton’s laws of motion) to various systems of cohesive bodies amount
only to the addition of these effects (in terms of accelerations). If “forces” were not “addi-
tive,” this alone wouldn’t force a quantifier commitment to forces; i.e., we still wouldn’t
find ourselves with instantiations of the schemas in which quantification over forces oc-
curred. All that would happen is that the instantiations would become even harder to manip-
ulate mathematically than they already are.
6
Two easy examples from first-year physics: sliding a block across a table, and a pen-
dulum.
7
See, in particular, the beginning of Truesdell’s (1991, pp. 193–4) discussion of the
motion of a free body.
8
Consider the formula for a simple pendulum cited in chapter 1: mgsinθ = ml(d 2θ/
dt 2). This is derived by transforming the “constraining forces” on the bob, among them
the force(s) that prevents the bob from breaking away from the apparatus, into geometric
constraints on the movements of that bob. This is a general phenomena: All initial references
to forces dissolve into (1) geometric constraints or (2) characterizations of how velocities,
positions, and accelerations are linked in how they change over time (because of those
constraints), or are (3) (in part or whole) ignored altogether.
Applied Mathematics and Ontology 185
9
Matters can change, of course: Scientific laws, and their applications, may be recast
so that the role of terms standing for forces changes—such terms no longer operate as
ciphers for mathematical constraints on the possible movements of (parts of) bodies. Fur-
ther, to anticipate the way I soon approach the ontological issues here, thick epistemic
access may seem to be forged to the forces themselves. It certainly looks like this is what’s
happened in the contemporary setting, because forces are characterized as the exchange of
various subatomic particles, which we, in turn, attempt to detect. I regret that I must leave
a more careful analysis of the place of forces in the contemporary non-Newtonian setting
for another time (and place).
186 Applied Mathematics and Its Posits
10
“Geometrizing away forces” is, of course, already possible in the Newtonian context
itself, and this can (and has) led philosophers to a more fine-grained analysis of the differ-
ences between what the geometry looks like in the (neo-)Newtonian context and its appear-
ance in general relativity, in order to evaluate differences in how geometric commitments
get into the two pictures.
11
I draw this example from Cohen 1999, p. 99. The example is due to George Smith.
12
It seems clear, by the way, that Newton was committed to the reality of forces, and
despite his speaking of, e.g., “centrifugal forces” (which was meant to describe what today
would be called the tangential component of curved or orbital motion; see Cohen 1999,
pp. 82–4), he didn’t take the latter to be real as he did take gravitation to be, and the other
forces he considered in the course of his thinking, and this despite the similarity of language.
Cohen (1999) writes: “Newton believed in and made use of many varieties of force in the
course of his scientific thinking. These include—among others—‘active’ and ‘passive’ forces,
chemical and other short-range or interparticle forces, and various forces associated with (or
Applied Mathematics and Ontology 187
arising from) various concepts of aether” (p. 54). Indeed, Newton would not have been so
concerned—so defensive, even—about the question of the “cause” of gravity if he hadn’t
believed in the existence of it.
13
This condition is sufficient but not necessary because, e.g., the effects of mass, al-
though present, may cancel in such a way that bodies continue to move inertially.
14
Given that constitutive relations cause deviations from the first law, it’s important
for current purposes to include them under the rubric of “impressed force.” Assimilating the
recognition of constitutive relations to the general search for force laws may look unnatural:
Constitutive relations seem much more like a temporary stopping point for such a pro-
gram—they’re to be subsequently replaced by a more fine-tuned analysis of materials that
provides “genuine” force laws from which what empirically corresponds to constitutive rela-
tions can be derived. But, regardless of the success of (reductive) attempts to replace consti-
tutive relations with “genuine” force laws, this is a fate that can befall any force law, for any
of them can turn out to be temporary stopping points in the establishment of physical laws,
and thus in the analysis of physical phenomena.
15
This programmatic assumption arises again in the discussion of absolute space and
time, and privileged frameworks. Notice the point of “at best”—being made of certain mate-
rials needn’t generate ontological commitments (e.g., the existence of massive or charged
objects) even though it helps satisfy the program of locating sources for deviations from the
first law. See what follows.
Notice also that the program places systematic constraints on ontological assumptions:
188 Applied Mathematics and Its Posits
The presence of mass, for example, affects every body in the universe—it won’t do to postu-
late a massive body B to explain the deviation of some body A from Newton’s first law, if
the result is that B can’t have gravitational impact on any other body in the universe (on
pain of empirical refutation).
16
How forces communicate from one part of a cohesive body to another? See what
immediately follows, but also notice that what’s being alluded to here is that how force
laws are to apply to the movements of material points is constrained in various ways by the
constitutive relations.
17
Exactly the same sorts of considerations motivate the adoption of the truth idiom.
See chapter 1.
18
Newton describes his first and second laws this way: (1) “Every body perseveres in
its state of being at rest or of moving uniformly straight forward, except insofar as it is
compelled to change its state by forces impressed.” (2) “A change in motion is proportional
to the motive force impressed and takes place along the straight line in which that force is
impressed.” See Newton 1726, p. 416. Also, consider his proposition 1: “The forces by
which the circumjovial planets [satellites of Jupiter] are continually drawn away from recti-
Applied Mathematics and Ontology 189
linear motions and are maintained in their respective orbits are directed to the center of
Jupiter and are inversely as the squares of their places from that center” (Newton 1726,
p. 802).
19
See the references in chapter 8, n. 13.
190 Applied Mathematics and Its Posits
the real issue is what we think there is, and that is indicated by our theo-
ries (what we say there is). No, the point is that the ontological commit-
ments of a theory should be due to how the theory links up to the
world—and that’s what notational reworkings of theories have (pretty
much) nothing to do with.20
Given my foregoing discussion of ontology, the (related) objection
is this: What should be relevant are the aspects (noun phrases, predicates,
whatever) of our theories that are connected to what they refer to (or
hold of!) in such a way that the latter must be regarded as ontologically
independent. Linguistic massagings (and remassagings) of theories so that
existential commitments are recalibrated as schematic letters (or vice versa)
drop out of consideration because of their irrelevancy to methods of rec-
ognizing ontological commitment.
Instead, what needs to be shown is that there is no reason to think
that whatever it is that’s described in one notation via first- or second-
order quantifiers, and in another notation via schemas, is something that’s
ontologically independent. And this means that no argument can be
taken seriously if it presupposes that the ontological commitments of a
theory must be among the quantifier commitments of a theory, because
when was it shown that if something is ontologically independent then
that something will always be represented in a first-order theory by noun
phrases (and quantifiers)?21
Let’s then reconsider the ontological status of forces from this other
perspective: What reasons are there to take forces to be posits that are
ontologically independent of us? Well, given that we take a category of
somethings to exist either if we see ourselves as forging thick epistemic
access to instances of it or if that category falls under a more general
category that involves thick epistemic access,22 there are lots of reasons to
think we don’t have thick epistemic access to forces: We don’t (directly, or
even indirectly) study forces. We don’t determine what the individuation
conditions for forces are, nor do we study what forces are made of.23 And,
indeed, the way that the laws of Ncm are instantiated suggests that forces
are part of a mathematical formalism within which the movement of ob-
20
Qualification: A notational reworking of a theory may reveal a tractable way of bring-
ing the theory to bear on the instrumental interaction with something. But this isn’t the
sort of case under discussion.
21
I did show (in part I) that the range of the (first-order) quantifier shouldn’t be
restricted to all and only what (really) exists, but when was it shown that anything that’s
depicted (in some way) by a theory, and that we take to be ontologically independent, must
appear within the range of the quantifiers—and thus must fall within the scope of the
existence predicate of that theory? Notice that I did argue in chapter 3 that if we want the
existence predicate to hold of all and only what exists, then that predicate must be governed
by a criterion for what exists. But once we’ve motivated a specific criterion for what exists,
the possibility becomes open to us that something codified in a language in such a way that
it doesn’t fall under a (first-order) quantifier might nevertheless fit that criterion.
22
See the discussion of this in chapter 6.
23
Again, I’m speaking strictly within the Newtonian context.
Applied Mathematics and Ontology 191
jects can be predicted, but they don’t provide an explanation for those
movements; that is, we’re not given any sense of what forces do, of how
forces percolate from one object to another, of how they work their mo-
tional magic on objects, of how they’re even involved in the process. All
this would be perfectly reasonable as a study if forces were objects of a
certain sort that we could investigate more closely to see if we could find
out how they interact with other objects. Notice the point: It’s not that
such a study is attempted unsuccessfully, and so scientists claim they don’t
understand (yet) how forces do what they do. Nothing of this sort occurs
in practice—that’s what indicates the mathematical character of forces.
Newton (1726) describes his approach to gravitational forces (more
than once) as mathematical. Here is the famous quote from the General
Scholium:
Thus far I have explained the phenomena of the heavens and of our sea by
the force of gravity, but I have not yet assigned a cause to gravity. Indeed,
this force arises from some cause that penetrates as far as the centers of the
sun and planets without any diminution of its power to act, and that acts not
in proportion to the quantity of the surfaces of the particles on which it acts
(as mechanical causes are wont to do) but in proportion to the quantity of
solid matter, and whose action is extended everywhere to immense distances,
always decreasing as the squares of the distances. (p. 943)
He adds shortly after: “I have not as yet been able to deduce from
phenomena the reason for these properties of gravity, and I do not feign
hypotheses.” And then he says: “And it is enough that gravity really exists
and acts according to the laws that we have set forth and is sufficient to
explain all the motions of the heavenly bodies and of our sea.”24
Cohesive Bodies
The next issue to discuss is the claim that the reconstrual of lumps as
moving sets of material points doesn’t require there really being such sets
of material points (that such are what lumps of matter really are). The
first remark to make is that material points—like forces (and like space-
time points, as I discuss below)—aren’t amenable to thick epistemic ac-
cess. One doesn’t, in order to apply Ncm, determine that lumps really
are made up of densely packed material points: One doesn’t empirically
determine (via subtle instrumentation, say) that lumps are thickly com-
24
For a general description of Newton’s mathematical approach to forces and its appli-
cation to the external world, see Cohen 1999, esp. pp. 148–55.
Also, see Truesdell (1991, p. 61), where we find: “Modern fundamental thought in
mechanics has reverted to the viewpoint of NEWTON and EULER: Forces are basic, a
priori concepts in mechanics.” Translation: they’re (potentially axiomatizable) mathematical
concepts imposed upon an empirical domain. They’re not interesting empirical items that
we’ve discovered and want to truly characterize the properties of.
192 Applied Mathematics and Its Posits
Even given that the functions just mentioned in the last paragraph
are empirically well defined with respect to lumps, even given, that is,
that the distribution of mass in lumps is smooth and continuous (not
discrete), and even given that the other defined properties of cohesive
bodies apply perfectly to lumps so that the predicted behavior of cohesive
bodies—how they deform and change upon impact, and so on—proves
true of lumps, this doesn’t allow an inference to the existence of the mate-
rial points upon which the mathematics of cohesive bodies depends. All
it allows (and shows) is that there is an instantiation of Ncm to a universe
of cohesive bodies (which are composed of material points) where these
cohesive bodies move the same way over time as lumps actually do.29
The reason to deny the “inference to the best explanation” to what
empirically corresponds to material points is that despite (let’s say) even
a perfect isomorphism between the movements of lumps and the move-
ments of cohesive bodies, we still have no reason to think there are (em-
pirical) points (because none of this provides thick epistemic access to what
empirically corresponds to material points). We can say that the distribu-
tion of mass in a lump is smooth, and we can capture that notion mathe-
matically via points. But this doesn’t license an ontological commitment
to points. (I discuss this again below: Cohesive bodies—some of them,
anyway—can proxy for something real: lumps. But that doesn’t require
the rest of the mathematical apparatus accompanying cohesive bodies,
e.g., material points, to proxy for anything empirical—even if that appa-
ratus is empirically indispensable.)30
That an instantiation of Ncm to a particular model of cohesive bodies
gives us exactly the same results in movement as a collection of lumps
careening through space means that (this particular instantiation of) Ncm
is (exactly) true of those lumps even though the quantifier commitments
of Ncm aren’t the same as the quantifier commitments (in example (c))
of the empirical theory E of lumps. Ncm, here, is true of lumps, but it’s
not true of what empirically corresponds to forces, points, and so on,
because there are no such things. Of course (see chapters 1 and 2), this
doesn’t stop the theory from being true simpliciter.
This conclusion raises two issues. First, one may wonder why the laws
governing the movements of lumps are approached in such an ontologi-
cally indirect way (via, i.e., cohesive bodies and material points). Well, a
precise characterization of lumps, and how they move and deform, which
doesn’t involve the axiomatizations of any terms other than those that
29
Of course, I’m speaking in an ontologically noncommitting way about cohesive bod-
ies and material points, both of which are purely mathematical entities.
30
Recall from chapter 6 that just because we have thick epistemic access to something,
it doesn’t follow that we have thick epistemic access to its parts or to what it’s part of, if
anything. This is the companion to the observation just made about applied mathematics:
The application of a mathematical system may supply a proxy role for some of the posits in
that system (a proxy role for something we’re ontologically committed to, that is). It won’t,
in general, do that for all the posits in the applied system.
194 Applied Mathematics and Its Posits
31
It’s hard to prove this, because the claim is a negative existential. But mathematical
characterizations of subject matters often employ certain concepts because we (at present)
have no other way to do it—and sometimes that’s because there is no other (tractable) way
to do it. It’s also common (this should be noted) to have alternative approaches to an
empirical subject matter that don’t employ the same mathematical characterizations, but
that employ other mathematical notions also going quite beyond the empirical subject mat-
ter itself.
32
Field (1989) writes that “the most satisfying explanations are usually ‘intrinsic’ ones
that don’t invoke entities that are causally irrelevant to what is being explained” (p. 18).
He contrasts these with “extrinsic explanations . . . (as when we explain the behaviour of a
nonhuman or non-English speaker by reference to English sentences that he or she believes
or desires)” and adds, “but it is natural to think that for any good extrinsic explanation
there is an intrinsic explanation that underlies it.” Natural or not, this last thought is ruled
out by the considerations just raised. Limitations in our very human powers to explain
things show up in the most interesting ways.
Applied Mathematics and Ontology 195
which don’t apply here). The second is ruled out because an instantiation
of such laws only to what we take to be real (only, say, to what we take
to be ontologically independent) is intractable. Our remaining choice is
to recognize that the laws are simply true as they stand, and simultane-
ously to deny that any of their existential commitments need be real.
This still allows (some of) their quantifier commitments to proxy for
what is real. In case (c), cohesive bodies proxy for same-shaped lumps in
which mass is distributed in the same way. Lumps are real; cohesive bod-
ies aren’t.
It’s also worth adding that because Ncm (but this is true of physical
theories generally) characterizes so much more than what’s in E (it brings
in, i.e., such a wealth of additional mathematical flora and fauna), it’s far
more useful in applications to E. For, as I’ve pointed out, we often can’t
apply a theory to exactly what there is because the latter is too compli-
cated and intractable to characterize directly. Instead, idealizations are
introduced, but if a physical theory has conceptual breadth beyond its
capacity to describe what there is, then it can be applied without difficulty
even to highly idealized contexts. In the case of Ncm and E, for example,
the exact geometric properties of lumps and the distribution of mass
throughout those lumps may be impossible to discover and characterize;
nevertheless, we can still apply Ncm to all sorts of cases where we idealize
the properties of lumps into something amenable to tractable physical
law. A theory that described only what there is wouldn’t be as empirically
useful.
This helps illuminate the second issue raised by Ncm’s failure to de-
scribe (only) what exists: Why, nevertheless, is it empirically adequate
when applied to E? I discussed this issue in chapter 8 (elucidation (5)),
and so it should already be clear that there can be several possible explana-
tions for the empirical adequacy of these mathematical assumptions, none
of which rely on the reality of material points, cohesive bodies, forces,
and so on, and all of which turn on empirical facts about the objects of
E. One possibility, for example, is that lumps really are infinitely divisible
in a smooth way. Were this so, we could prove that Ncm gives us exact
answers in the limit. But lumps could be granular, and then the success-
ful application of Ncm (when coupled with mathematical assumptions
about the continuous distribution of mass in lumps) would involve ap-
proximations (and in what sense these mathematical assumptions were
approximate would turn on exactly how the lumps were granular).
For example (and restricting our attention to gravitation), suppose
we could empirically establish that there are atoms, and that atoms are
indivisible spheres (of a certain fixed diameter) within which mass is
evenly distributed. In such a case, some successful applications of Ncm
would turn on material points, given the tiny size of these atoms, ade-
quately representing (at least to a certain degree of approximation) the
latter.
196 Applied Mathematics and Its Posits
33
I should warn the reader that I continue to make free use of the term “force” and
the distinction between “real forces” and “pseudo-forces.” What my ontologically innocu-
ous use of these words mean should already be clear from the foregoing.
34
Westfall (1980) notes that it’s still an open question how much Newton participated
on the sly in Clarke’s debate with Leibniz. Of course, there are considerations in support
of absolute space and time in the Principia itself and in other places in Newton’s writings.
Applied Mathematics and Ontology 197
family of frames, as I’ve characterized them, will do the job, and not a single one alone,
that shows that, mathematically speaking, what’s indispensable to the application of Ncm is
not the Cartesian product 3 × , with the standard metric, but a much weaker structure,
a four-dimensional real affine space. This particular wrinkle isn’t of concern in what follows
because, in any case, indispensability is insufficient for ontological commitment. On the
mathematical situation regarding Newton’s theory, see Stein 1967 or Friedman 1983.
39
This isn’t the whole story about possible versions of the relationist view. See Fried-
man 1983, esp. pp. 223–36. Also see Maudlin 1993. Because my only point is that the
privileged frame program isn’t a relationist view—quantification over all the apparatuses
that substantivalists help themselves to is acceptable—there’s no need to sort through this
particular tangle.
Applied Mathematics and Ontology 199
40
Stein (1967) notes that Newton applies the adjective “absolute” “in its logical sense
to concepts that are not ‘relative’ to something” (p. 189). He elucidates this claim in a
footnote:
For instance, what Newton calls “absolute quantity” of the earth’s gravity is a property
of the earth itself, which we call the earth’s “gravitational mass”; in contrast with the
“accelerative quantity” of the earth’s gravity, which varies from place to place and is a
property of a point in a given position relative to the earth—the intensity of the earth’s
gravitational field at that point; and in contrast with the “motive quantity” of the
earth’s gravity, which depends both upon the place and the body acted on—the
weight, or attraction exerted upon the body by the earth. (p. 199)
[i]n both of these passages one observes what I should characterize as a loose
and uncontrolled manipulation of verbal distinctions. What exactly do these
authors mean by “ideal entities which it is helpful to consider in theory,” or
by a notion or theory that “has a physical application,” —as opposed to
entities that “exist in reality.” . . . If the distinction between inertial frames
of reference and those which are not inertial is a distinction that has a real
application to the world; that is, if the [affine] structure [of space-time] is in
some sense really exhibited by the world of events; and if this structure can
legitimately be regarded as an explication of Newton’s “absolute space and
time”; then the question whether, in addition to characterizing the world in
just the indicated sense, this structure of space-time also “really exists,” surely
seems to be supererogatory. (p. 193)
And he adds a few lines later: “It is important to be very clear about one
point: the notion of the structure of space-time cannot, in so far as it is
truly applicable to the physical world, be regarded as a mere conceptual
tool to be used from time to time as convenience dictates. For there is
only one physical world; and if it has the postulated structure, that struc-
ture is—by hypothesis—there, once for all.”
However, even though the notion of space-time is not a mere con-
ceptual tool to be used from time to time as convenience dictates, it
does not follow that the structure is “there”: This claim presupposes an
indispensability thesis—if not the Quine-Putnam indispensability thesis,
then some indispensability thesis (or other) painfully close in import to
that one. Contrary to such a thesis, indispensable structure in an empirical
theory, and the posits that accompany it, needn’t exist.43
I now make several observations in order to take up various concerns
about the privileged frame program that the reader may have: First, I
should point out that I’m assuming that stable units of time and space
and stable ways of measuring distances and durations are available. This
assumption doesn’t require the existence of “absolute” space and time,
although, of course, it does require the positing of space and time (and
the points within them), as well as those posits of space and time having
certain properties (isotrophy, etc.). But it’s clear by now, I hope, that the
mere presence of a posit is insufficient for ontological commitment to it,
and measurement practices, turning only on the mutual consistency of the
results of measurement alone, seem to be a perfect example of a body of
practices that don’t have as their goal the squaring of the claims we make
the laws of dynamics would take the simplest form” (cited in Stein 1967, p. 192). That the
goal of the privileged frame program is to identify the set of frames with respect to which
the laws of dynamics take the simplest form is not my view: The aim is to find a frame so
that no motions of objects or their parts occur that aren’t described by Newton’s laws of
motions, the force laws, and the constitutive relations.
43
Having said this, it’s nevertheless appropriate to add that Stein’s remarks are best
seen as a challenge: Such language as “fully exists” and “exists” (or “really exists” and “ex-
ists,” as I’d prefer the contrast to go), if distinguished, must be distinguished in a principled
way. See part I, esp. chapter 5.
Applied Mathematics and Ontology 201
45
The more massive object accelerates more slowly toward the less massive object
(when the latter is fixed) than when the situation is the other way around. This is a quite
crude description of an experiment of this sort. Others (many others) much more nicely
designed are possible (and actual). See Stein 1967, pp. 190–1, for a good, although brief,
discussion on this.
46
The reader will have noticed that I often try to carefully distinguish terminologically
the mathematical items, cohesive bodies, forces, and constitutive relations, from the empiri-
cal items, lumps, sources of forces, and the materials out of which lumps are composed.
I’ve perhaps not been perfectly consistent in this respect, but I’ll be even less so in what
follows, for the sake of exposition.
Applied Mathematics and Ontology 203
exactly) applied to its side that continuously increases its rotation. Even-
tually, if the disk rotates fast enough, it will tear apart as its subbodies
move tangentially to the orbit the constitutive relations (or internal
forces) hold them in. Notice that the only factors relevant to this example
are (1) the contact force applied to the disk (and transmitted to its sub-
parts), (2) the constitutive relations (or internal forces) serving the two-
fold role of transmitting the torque through the disk, and holding its
material points in place, and (3) the (negligible) gravitational forces in-
duced by the presumed uniform distribution of mass in the disk.
If the constitutive relations are finitely powerful, and if their impact
on the disk’s cohesion with respect to inertia is measurable—which, em-
pirically, is to be expected—then one can (empirically) determine how
fast the disk is rotating because a privileged frame is fixed by when the
constitutive relations allow the disk to become distorted. If, however,
the constitutive relations of the disk make it a rigid body impervious to
distortion, then no matter how long the torque is applied to the coin,
there is no way to determine its rotational velocity empirically because
there is no way to determine what the constitutive relations are actually
doing. In such a case, we can’t determine (empirically) a privileged frame
because the infinitely powerful constitutive relations of the disk make it
impervious to empirical study (given the restricted resources in this uni-
verse). If, however, there were disks of other materials with internal forces
or constitutive relations of finite strength, the privileged frame could be
determined from them, and then by imposing that frame on the original
disk, the work being done by its constitutive relations could be deter-
mined, as well. In practice, because what materials that objects are made
of affects how they respond to forces in a systematic way, and because
the privileged frame applies to all of space and time, the study of any
spinning shape is epistemically legitimate in our search for the privileged
frame. In a universe strangely devoid of objects, such practices can’t be
carried out, but there is nothing wrong with the (ontological) assumption
that the privileged frame is given nonetheless by how the constitutive
relations (of a disk in space, say) actually are causing deviations from
Newton’s first law, and even if epistemic access to them is ruled out.47
Third observation: Imagine that it turned out that (certain) frames
differing with respect to their accelerations were compatible with the same
(relative) velocities, accelerations, distances, and so on, of (particular con-
figurations of) a set of bodies. Would this create problems with the frame
program? It’s hard to see why. It would show that empirical results about
the distribution of mass would underdetermine our choice of frames more
than it appeared at first sight. But so what? If, that is, frames really are
47
Am I shoring up my realism about the facts of accelerations in such a universe via a
kind of modal realism—by access to facts about what would be the case if there were other
objects in such (ontologically deprived) universes? I think not, although a discussion of this
takes us beyond the scope of this book.
204 Applied Mathematics and Its Posits
48
The reader may still feel that what the mathematical term “space” picks out in this
application—viz., physical space—is the culprit. I discuss why we should resist this take on
the mathematical formalism toward the end of the discussion of the fourth observation.
49
If right, this, coupled with the first observation about how stable measurements of
durations and distances are found, forecloses on the possibility of forging thick epistemic
access to space and time and thus leaves us with no reason to regard the posits of space and
time as anything more than ultrathin.
50
Some philosophers have taken these experiments as showing that the Newtonian
view attributes dynamical properties to space itself. Others (e.g., Mach) have seen them as
a challenge: to provide a frame, anchored in some concrete object or other, relative to
which the accelerative effects depicted can be seen to occur; and they have denied that such
accelerative effects will occur if such concrete objects (e.g., fixed stars) are missing. Still
others (e.g., Einstein) have used a variant of these thought experiments to run a verifica-
tionist objection to absolute space and time as they occur in Newton’s system.
Applied Mathematics and Ontology 205
stopped from spinning, the water continues to rotate, and further, the ro-
tating water recedes from the center of the bucket and up its sides. The
second (thought) experiment is of a universe with only two spheres con-
nected by a rope. If they’re rotating with respect to each other, the rope
becomes taut as they try to tangentially move from the orbit the rope holds
them to. If they aren’t so rotating, then the rope remains slack.
Do these examples show that space itself has causal properties? If one
accepts the privileged frame program, it’s hard to see why. Consider the
two spheres first: There are internal forces (or, if you wish, constitutive
relations) in the rope that can hold the spheres in orbit. Thus, if the rope
become taut, the accelerated motions involved (the spheres continuing to
circle around the center of mass of the system instead of flying tangen-
tially away from each other) are due solely to the factors within that rope.
Thus, the choice of frame is determined as one in which the spheres are
rotating rather than one in which they’re motionless. More can be said
about this example: The spheres themselves are composed of subbodies
held, let’s say, rigidly together. Were these internal forces or constitutive
relations absent or weak, it’s not just that the spheres themselves would
fly away from each other tangentially to the orbit they’re bound to by
the rope, but that they would fragment into smaller bodies that would
individually fly off tangentally.51 Recognition of these accelerated motions
(of the subbodies within the spheres) is satisfied in the privileged frame
by their source in constitutive relations or the sources of other (internal)
forces. On the other hand, if the rope remains slack, it’s clear that there
are no accelerated motions involved in the example at all, and a frame in
accord with this must be chosen instead.
The analysis of the bucket case is almost the same, except that in this
case the relevant constitutive relations and contact forces occur not in a
rope (and within the spheres) but within the fluid itself, and between the
fluid and the bucket, because it’s these factors, and the obvious conclu-
sion that it can only be them reacting against inertia that causes the fluid
to recede from the center of the bucket and up the sides of that bucket,
that force our choice of a privileged frame in which, at first, the bucket is
spinning (and the water isn’t) and then, later, the water in the bucket
continues to spin (although the bucket no longer is).52
A similar lesson can be drawn from an example that Einstein (1916)
gives:
In classical mechanics . . . there is an inherent epistemological defect which
was, perhaps for the first time, clearly pointed out by Ernest Mach. We will
51
This depends, of course, on exactly how the factors internal to the spheres operate
and how the internal factors in the rope transmit forces to the subbodies in the spheres.
52
Given the global nature of the frame to be chosen, however, an analysis of the
internal forces and constitutive relations of the twisted rope is relevant and enables us to fix
upon a frame in which, subsequently, the bucket spins, and later, the water does also. For
that matter, as I’ve mentioned, fixing the frame via something else (far away) will do the
job too!
206 Applied Mathematics and Its Posits
53
I’m assuming that the example presumes fluid bodies to be made up of (quite small)
particles that exert nonzero gravitational forces on each other rather than continuous bodies
with continuum-many mass points with zero mass but with a nonzero mass function defined
over each body. The lessons I want to draw are available in either case, but I’m trying to
stick as closely as possible to my reading of Einstein’s text.
54
And these turn out to be the same family of frames. It’s, of course, a very important
empirical assumption of the privileged frame program that the same family of privileged
frames applies to every object in the universe: a failure in this is taken (initially) to imply that
other (unaccounted for) sources of force or other factors are locally present. A more radical
proposal would be the denial that a single family of frames does apply to everything, but
this requires deserting the Newtonian context.
55
It’s clear that Leibniz is focused on contact forces, and Friedman rightly takes him
to task for this (there’s no avoiding action at a distance in Ncm). But even if we drop that
constraint, Friedman thinks the program still fails to handle accelerations arising when con-
stant rotational velocity is present; this is wrong, as I show in a moment.
56
Dainton 2001, p. 177, repeats the dismissal as curtly.
208 Applied Mathematics and Its Posits
57
Dainton (2001) writes: “A rotating solid disc will continue to rotate at a constant
angular velocity, without any need for any external force or torque to be applied: it is
maintained in motion by the preservation of angular momentum. . . . And yet, if the disc
rotates fast enough, the centrifugal forces [inertial forces!] generated by its rotation will tear
it apart” (p. 177). The last sentence is nearly on the point (speaking of inertial forces “gen-
erated” by rotation is, I think, misleading), but one must focus not on the disk being “torn
apart” but it’s remaining intact (at any rotational rate, however low): That’s where forces
come into play. (Recall the remark about the wall in the paragraph to which n. 11 is ap-
pended.)
Sklar (1976) makes a similar objection to the suggestion I’m attributing to Leibniz:
[O]ne seems to be able to imagine situations where there are inertial forces but no
causal forces inducing the motion at all. Again imagine a uniformly rotating universe
of material objects. Now imagine it to have been always in rotation, the rotation not
caused by any imposed torque at all, but simply eternally sustained by the conservation
of angular momentum. There would still be inertial forces present on any object at
rest in this coordinate frame. (p. 192)
Nothing (that’s extended), whether a disk or a universe, can rotate (from eternity or other-
wise) unless there are constitutive relations or other internal forces holding it together.
I suspect that what’s involved here is an overreaction to the fact that angular momen-
tum, like linear momentum, is conserved (without an external torque, angular momentum
remains constant). But the issue, of course, is whether nonzero angular momentum is possi-
ble without what amounts to internal forces. The answer is no, but it is yes in the case of
linear momentum. This is why “inertial forces” arise in the one case but not in the other.
Applied Mathematics and Ontology 209
applied theory exactly correspond to the physical world (there are, in the
world, objects with exactly the properties the mathematical entities are
postulated to have). But, apart from the global success of the applied
mathematics, we’ve no reason to think this is true, and because there are
lots of reasons for why an empirically adequate theory could be empirically
adequate (as the history of science indicates) apart from the entities that
theory postulates actually existing, and because, in any case, without thick
epistemic access to those posits, the theory is a black box, we’ve no episte-
mic justification for this easy solution.
When it comes to fiction, we can all agree that the properties attrib-
uted to fictional objects are in no way beholden to things independent of
us. But as the foregoing observations make palpably clear, such isn’t true
of space: The properties attributed to it are ones we have no choice about
because they impose on how objects move—and that’s not a matter for
us to dictate. The movements of objects through space, that is, are them-
selves in thrall to something ontologically independent of us and, indeed,
ontologically independent tout court.
This much is true, but we can’t say more about what this something
is, and just because, in imposing a branch of mathematics, entities are
brought along that we attribute specific properties to, and just because
we’ve no alternatives in what properties we can attribute to such objects
(and still have the applied mathematics work), and despite our inability,
say, to imagine lots of alternatives, it doesn’t follow that we’ve managed
to get our terms (“space,” say) to refer to that something that is ontologi-
cally independent of us and that is the source of the indispensability of
this application of pure mathematics. We may have, but in lieu of specific
access to the items referred to by terms in the mathematical formalism,
we’ve only got the global success of that applied mathematics as an argu-
ment for the claim that such terms refer to what’s ontologically indepen-
dent of us, and as I’ve already shown, that won’t suffice.
As an illustration of how the spatial posits are operating in a purely
mathematical way, consider again the Newtonian story about rotating
disks. Mach once suggested that inertial forces are produced when a body
accelerates with respect to the center of mass of the entire universe. Dain-
ton (2001, p. 179) suggests that this requires a mechanism by which the
center of mass can do such a thing. It’s an implication of the frame pro-
gram that, restricting our attention to gravitation, inertial effects can be
observed if a body accelerates with respect to the center of mass of the
entire universe, but the need for a mechanism is denied because it isn’t
that the center of mass is causing anything: It’s that, when all the accelera-
tive effects of the (real) forces have been accounted for, this result pops
out as a corollary.58
58
But why does it pop out as a corollary, one may wonder; why should this global
version of the frame program work? That’s to ask for an explanation where no explanation
(from within Ncm, that is) is possible. And this is always the case: We can’t reach a point
Applied Mathematics and Ontology 211
where the successful application of all the mathematics in a theory is explicable (because a
fundamental theory always employs mathematics indispensably—and we can’t explain the
success of that mathematics from within the fundamental theory itself). My point at the
moment is not so much to illustrate this simple fact of scientific life that the Quine-Putnam
indispensability thesis starts from, but rather to show how we sometimes attempt pseudo-
explanations by (what amounts to) postulating physical entities that mimic the properties
of the ultrathin posits of the applied mathematics.
59
Contrast this with the forces that affect objects moving through a fluid such as water.
Here we have access (in principle) to the mechanisms to which such forces are due—how
the molecules of water move around (and adhere to) moving objects—and what sorts of
(molecular) forces there are. Notice also that we’ve got access—thick access—to the objects
in question: not just to the objects moving through the water, but to the components of
the water itself. This blatantly contrasts with absolute space and time in the context of Ncm.
60
Nerlich (1994a) writes: “[W]ould we say, not just that we see that there is such a
hole, but that we see the hole itself, just as we see warps in clear glass?” (p. 39). Also see
Dainton’s (2001) discussion of this.
212 Applied Mathematics and Its Posits
Space-Time Points
I now turn to a discussion of space-time points themselves (largely apart
from, i.e., the question of the existence of space and time themselves).62
Field (1989) has argued that the reasons for refusing a commitment to
abstracta aren’t available for the rejection of “space-time regions,” even
if these regions are presumed to be “causally inert,” but, in addition, that
there is no reason to treat space-time as causally inert because
a field theory is most naturally construed as a theory that ascribes causal
properties (electromagnetic field values of various gravitational tensors, etc.)
to space-time points. Consequently, any theory in which talk of fields is to
be taken seriously (not as a mere shorthand for talk about physical objects)
will be most naturally construed as a theory ascribing causal properties to
space-time points . . . in such a way that which properties are assigned caus-
ally affects the behavior of objects. (p. 70)
Before taking up these considerations, let’s rehearse two previously
made points: First, it’s never enough after having established an ontologi-
cal commitment to something, to simply rest on that to carry along addi-
61
These claims, by the way, are compatible with both my hand and the table being
constituted of such things, and with such constituents existing. But notice that the existence
of such constituents is established by thick epistemic access to them, not by the mere exis-
tence of a true theory that describes the properties of tables and hands in terms of such
constituents, because such a theory could prove to be empirically adequate by virtue of its
global virtues, without our having determined that anything it was quantifier committed to
existed.
62
If space and time themselves don’t exist, there’s little reason for an ontological com-
mitment to spatial, temporal, or space-time points. Still, for philosophical reasons, some
further remarks about spatial, temporal, and space-time points, and the philosophical claims
and problems ontological commitments to such things have raised, are in order.
Applied Mathematics and Ontology 213
It should already be clear how this sins against the considerations I’ve
rehearsed, but let’s note some details. Contrast space-time regions with
electrons, and notice how, in forging thick epistemic access to electrons,
a battery of tools (and theoretical considerations) from those of Millikan’s
until today have been used to empirically verify specific properties of elec-
214 Applied Mathematics and Its Posits
trons. Not only does this include determining its mass, charge, and so
on, but it also includes the quantum-mechanical construal of it as a field:
One explores the geometric distribution of electron fields around mole-
cules. Contrast this with the visual considerations Field invokes on behalf
of space-time regions: We don’t see any of the properties of space-time
regions, not their dimensionality, not their curvature, not their composi-
tion (out of whatever).63 Indeed, it’s bizarre to say we see them at all.
And, the consideration Field raises about empirically verifying properties
of space-time (vs. verifying properties of numbers) is simply a red herring:
These properties of space-time are verified only in good Quinean fashion,
as indispensable parts of a theory empirically verified as a whole—and not
because of the presumed empirical relations between us and space-time
regions used to forge thick epistemic access to such regions.64
But Field may seem to rely on stronger grounds if he is allowed that
(1) fields are causal agents (to which, therefore, we can forge thick episte-
mic access) and that (2) such items are “most naturally construed” as
made up of space-time points or, perhaps more accurately, as simply the
exhibition of the various causal properties of space-time points them-
selves.65
Although I’ve no problem granting an ontological commitment to
fields on the grounds of their causal powers, and agree that we do indeed
forge thick epistemic access to them, as to, say, electron fields,66 it simply
doesn’t follow that we can then comfortably take ourselves to be commit-
63
And, as the preceding remarks on Nerlich’s example make clear, it’s not obvious
that we can even make sense of seeing spatial regions, vs. seeing something else that we
(theoretically construe) as spatial regions.
64
Access to electrons, of course, is deeply embedded in theory, too. But the empirical
element of thick epistemic access to the shape of the distributions of electrons around mole-
cules (their probability of location) plays a gigantic role, especially because the formulas that
govern this are so intractable.
65
Field (1989, pp. 70–1) argues against those who would distinguish a field from the
space-time it pointwise rests upon. Although the latter distinction of space-time from the
fields that live on it strikes me as “more natural” than identifying the field with space-time
(at least if ordinary practices among physicists are any indication of “naturalness”), I con-
cede this to Field because I argue it won’t help anyway.
66
Do we forge thick epistemic access to fields, or is it only that we forge thick epistemic
access to electrons that, for theoretical reasons, we take to be fields? The former—at least,
it’s the former to which thick epistemic access is being attempted. Schewe and Stein 2000
write:
[O]ne phenomenon is well established: Injecting an electron into liquid helium can
cause a tiny bubble (approximately 40 angstroms wide) to form around it. Shining
light on such bubbles, researchers expected the electrons to escape and zip out of the
helium. Not only did they fail to detect such electrons, but other experimenters ob-
served the creation of charged particles, never successfully identified.
The article goes on to summarize a radical explanation by the physicist Humphrey Maris:
[T]he light causes the electron to break up into two or more pieces, which he terms
“electrinos.” . . . This split-up, according to Maris, would arise from the quantum abil-
Applied Mathematics and Ontology 215
ity of electrons—and all matter—to act in some situations not as particles but as waves.
. . . Splitting up electron waves also has been demonstrated before . . . , but they ordi-
narily recombine and get detected as full particles.
67
For purposes of this example, I’ve allowed talk of fields to be treated as empirical
language, but, of course, “field” is a mathematical term that is itself drafted to describe
something empirical that can’t otherwise be characterized in language. See the discussion
of (f) later.
216 Applied Mathematics and Its Posits
68
E.g., the Schrödinger and Heisenberg representations of quantum mechanics, the
Lagrangian and Hamiltonian formulations of Newtonian physics, etc. There are many exam-
ples of empirical theories that differ substantially in their mathematical resources but are
seen by physicists as, nevertheless, identical theories. The approach to ontology described
in part I allows us (in principle) to make sense of the scientific take on such theories.
Applied Mathematics and Ontology 217
composition of real things out of other real things isn’t a matter of set-
theoretical constructions (which can go in any number of ways, depend-
ing on sheer mathematical taste) but a matter of how items are really put
together out of other items and, more important, how the laws that items
of one sort obey are compatible with and lead to laws governing how the
other sort of items act. That is, if the question of whether points or lines
are real is more than a mathematical question (which would be answered
simply by saying, “It doesn’t matter—do whatever is mathematically con-
venient”), then it’s important because how these things are actually put
together affects the laws that govern them.
The point can be put graphically: Can we physically tear apart lines
and find points? Or is it that points are objects composed of lines (so that
we find, upon inspection of points—upon forging thick epistemic access
to points in a way that allows us to take them apart and be sure what
those parts are like—that points are, say, classes of lines, instead)? These
considerations are, ultimately, not relevant to the question of how points
and lines are composed of one another because that question is just a
mathematical one to be stipulated by the mathematical system chosen:
Neither points nor lines are anything more than ultrathin. Composition
questions, when real objects are involved, aren’t questions to be dealt
with logically or set-theoretically, but ones to be answered by forging
thick epistemic access to such objects and their parts (as physicists have
done with atoms) and finding physically real methods of tearing things
into these parts. Although mathematics may be pertinent to the theory of
a type of real object (the way that mathematics may be pertinent to any
scientific theory), it doesn’t follow that questions of the composition of
such things are therefore to be answered by the employment of logical,
mereological, or set-theoretical tricks of various sorts.
My second example is one I officially take from Zeno’s paradoxes of
motion.69 Salmon (and Russell before him) has suggested that the mod-
ern invention of calculus has solved Zeno’s paradoxes of motion. Here is
Salmon (1977):
The solution of the arrow paradox to which Russell refers70 has been aptly
called “the at-at theory of motion.” Using the definition of a mathematical
69
Although I’m not, I should emphatically note, claiming that what I offer was a
concern of Zeno’s.
70
Russell writes:
Weirstrass, by strictly banishing all infinitesimals, has at last shown that we live in an
unchanging world, and that the arrow, at every moment of its flight, is truly at rest.
The only point where Zeno probably erred was in inferring (if he did infer) that be-
cause there is no change, therefore the world must be in the same state at one time as
at another. This consequence by no means follows. (quoted in Salmon 1970, p. 23)
Salmon describes this remark as somewhat cryptic—but it’s quite clear what Russell is imply-
ing: that there is no change, and that that’s compatible with things at one moment being
one way and at another being another way.
218 Applied Mathematics and Its Posits
71
Strictly speaking, it has to be said, Salmon’s invocation of the continuum (or, rather,
the density) to deflect this question is a non sequitor. (Imagine there were only a finite set
of points—the problem would remain!)
72
Notice that, in any case, solutions to Zeno’s puzzle like the at-at theory are highly
nonempirical. It’s not as if we’ve designed a theory of some sort about spatial and temporal
points that we can then try to design an experiment in order to test. It’s this a prioristic
approach to the problem, even on the part of philosophers otherwise quite steeped in scien-
tific method, that indicates that what’s really going on is that mathematical abstracta are
being pressed into an ontological role they’re unequipped to handle.
Applied Mathematics and Ontology 219
the arrow problem (how does something move off of a point?) is differ-
ent: It’s due to the presence of a posit that is supposed to play a role
in how something moves (it moves from point to point). Because of its
mathematical nature, that movement is impossible. Removing such a
posit allows things to simply move through space in a continuous manner
(just as you’d hope them to move).73
I’ve now come to the end of my ontological analysis of example (c).
Recall what I’ve tried to show: The application of Ncm to free-wheeling
lumps, despite the apparatus of Ncm, commits us to neither forces nor
space-time points. Nor, despite the identification (for purposes of applica-
tion) of cohesive bodies with lumps, does it commit us to either cohesive
bodies or material points. We are committed to lumps (on grounds apart
from the application of Ncm), but we’re not committed, by virtue of this
application, to lumps (really) being composed of material points.
Examples (d) and (e) don’t pose any additional issues, and although
(f ) does,74 most of the hard work has already been done in discussing
example (c); I’ll be brief, therefore, in describing the ontological commit-
ments that arise from the application of Ncm in these cases. For (d) and
(e), the situation is the same as it is for (c). That is, that we only use
instrumental tools to access the properties of lumps rather than, say, in-
struments and observation, makes no difference to our ontological com-
mitments. A robust enough epistemic link to lumps suffices to determine
73
Of course it’s an empirical question whether things do move continuously through
space or, instead, move in a more discontinuous (quantum-mechanical) way. For that mat-
ter, it’s an empirical question whether the geometry attributed to space should be the one
presupposed by this commonsense analysis of these examples. My only point here is to
indicate how the positing of points as real creates problems where otherwise no problems
would exist.
The sorts of philosophical problems that points create, when we take them to be real,
arise, of course, out of the specific mathematical properties of points. Nevertheless, treating
abstracta as real can create (different sorts of) philosophical problems, as I showed in the
earlier discussion of absolute space.
I should observe that the original Zenonian puzzle is couched not in terms of points
(at one of which, say, the tip of the arrow is located) but in terms of the place (a spatial
region) that the arrow occupies. Places in those versions of the paradox play the same role
that points do in the version of the paradox I presented. We’ve no more reason to accept
an ontological commitment to places than we have to accept one to points. Retaining either
ontological commitment forces us to peculiar positions: I’ve noted Russell’s remark about
our changeless universe already. Vlastos (1967, p. 375), for another example, supports the
category-mistake suggestion that an arrow, during a moment, is neither moving nor not
moving: Either claim is a senseless one. It’s hard to see why this should be true if a span of
time really is made up of moments.
74
Recall what these examples are: In (d) we have epistemic access to the properties of
lumps only through instrumental means. In (e) lumps have properties other than those that
can be lawfully described by Ncm (such as, say, colors). In (f) there is no antecedent (empir-
ical) language of lumps: The mathematical language of Ncm has been drafted as the only
means of description for the items Ncm is applied to in this case.
220 Applied Mathematics and Its Posits
75
It seems to me that in practice physicists can be clear on this distinction, even when
they use obviously misleading language. As an illustration, consider the following remarks of
Gell-Man (1997):
I employed the term “mathematical” for quarks that would not emerge singly and
“real” for quarks that would. . . . [T]o illustrate what I meant by “mathematical”, I
gave the example of the limit of infinite mass and infinite binding energy.
Later, in my introductory lecture at the 1966 International Conference on High
Energy Physics in Berkeley, I improved the characterization of mathematical quarks
by describing them in terms of the limit of an infinite potential, essentially the way
that confinement is regarded today. Thus what I meant by “mathematical” for quarks
is what is now generally thought to be both true and predicted by QCD [quantum
chromodynamics]. Yet, up to the present, numerous authors keep stating or implying
that when I wrote that quarks were likely to be “mathematical” and unlikely to be
“real”, I meant that they somehow weren’t there. Of course I meant nothing of the
kind. (p. 626)
Conclusion
G. E. Moore (1939) once wrote (about a much shorter work than this!)
that “I have only, at most, succeeded in saying a very small part of what
ought to be said about it” (p. 126). I’m in a similar position despite
having written so much more, now and before, on the topics broached
in this book. Despite this, I like the idea of leaving future work both for
myself, and for others. I want, by way of conclusion, to summarize what
I hope I’ve accomplished in this book, and to sketchily indicate what’s
left for the future.
Despite Quine’s widespread influence, there are very few of his claims
that don’t continue to be highly controversial. Indeed, on many issues,
almost everyone is convinced he is wrong—the disagreement comes in
on the question of how.1 His influence, that is, has largely been in setting
the tone and framework for the debate on a number of central issues
rather than in providing a philosophical foundation that other philoso-
phers have then used as a springboard for their own work.
This is not true of Quine’s criterion for what a discourse is committed
to, however. Even those philosophers he most decisively broke with on
other issues (e.g., Carnap) found themselves embracing the letter of his
view of how ontological commitment is to be identified, although Car-
nap, anyway, deflated the metaphysical implications Quine drew from that
view. In part the nearly universal explicit (and implicit) acceptance of
1
This is my take on how the profession currently feels about indeterminacy of transla-
tion, for example.
222
Conclusion 223
Quine’s criterion was because, as Quine himself realized (and was mod-
estly willing to point out more than once), something like his criterion
had already been at work among philosophers.2
Thus, I recognized (back in 1992) how radical the views were that I
was moving toward when I drew the conclusion that the philosophical
arguments Quine had used to establish his criterion were not compelling,
and that there is room in logical space for a position that simultaneously
takes a discourse to be true (which I eventually recognized was something
we had to do in the case of applied mathematics and empirical scientific
doctrine) and yet, despite existential quantification, still allows us to deny
ontological commitment to the posits in that discourse. In a series of
articles (1997a, 1997b, 1998), I argued for the philosophical indetermi-
nacy of a criterion for what a discourse commits us to, in the sense that
no good philosophical arguments (or technical ones, for that matter) are
available to establish such a criterion, and a similarly (linked) philosophi-
cal indeterminacy for a criterion for what exists.
I’ve not deserted the general position of those essays, but I have
found a way of arguing both for a criterion for what a discourse commits
us to, and for a criterion for what exists, by recognizing how ordinary
citizens of our epistemic and linguistic community take themselves to be
committed to something, and what the methods that lead to commit-
ment to something indicate about that something. The resulting criteria
are not criteria that we as a community must adopt on pain of incoher-
ence (one can easily imagine communities that don’t share our criteria):
Our criteria are simply the criteria we have adopted.
These linked criteria (something that exists is “ontologically indepen-
dent”; a discourse is committed to all items that fall under the “ontologi-
cal independence” predicate) replaces the relatively straightforward Quin-
ean syntactic criterion with a semantic one. As a result, our metaphysical
lives become much harder than they appeared to be, given Quine’s ap-
proach. I describe Quine’s criterion as syntactic in the sense that, once a
discourse has been regimented, one need only look for implied sentences
with a certain syntactic structure (they begin with an existential quanti-
fier). But once this criterion is replaced with a predicate-style one, deter-
mining what falls under that predicate is no longer a matter of the gram-
mar of the regimentation at all.
Instead, we must look beyond the letter of the theory in question to
its application to the world. We must see exactly how, in a sense, that
theory is attached to the world. Those posits of the theory that are meant
to be ones that are secured to the world via a special type of causal relation
from us to the world—a causal relation I’ve described as thick epistemic
access—are items that we take to exist. Additional items that are theoreti-
2
I think, for example (as I’ve tried to illustrate), that, in some form, it was a powerful
factor in the substantivalist/relationist debate over space and time—and this is so right from
the beginning of that debate.
224 Conclusion
cally tied to the items just mentioned in a tight enough way are also items
we take to exist.
The remaining posits of the theory subside into ultrathin status. We
don’t take them to exist in any sense at all; they’re part of the mathemati-
cal formalism that facilitates the application of our theories to the world.
It’s in this sense, and only in this sense, that I’ve made a case for
nominalism. I’ve not done so in the programmatic sense of presenting a
project of rewriting our scientific theories so that the quantifiers in their
regimentations no longer range over abstracta, but (1) in the sense of
providing an argument that mathematical abstracta (in pure mathematics)
aren’t ontologically independent and thus aren’t items that we (should)
take ourselves to be committed to, and (2) in the sense of providing an
approach to the evaluation of scientific theories, in order to sort posits so
that we can recognize which ones are ultrathin and thus not items we’re
committed to.
The exercise of examining a particular scientific theory (and its appli-
cation) in order to determine which posits are the ones we mean to exist
isn’t an easy or straightforward activity. In part this is because even if
the language describing a particular posit of a scientific theory is entirely
mathematical, it doesn’t follow that the posit itself is ultrathin. Mathe-
matical posits (from the point of view of pure mathematics) can proxy for
items that are ontologically independent of us and that therefore, despite
their mathematical clothing (within our theories), are items we’re com-
mitted to.
We must also be careful in another way. In the application of a the-
ory, we may treat ultrathin posits as items that are causally efficacious;
this is done, for example, with centers of mass and, more generally, with
point-masses. But these items are not genuinely seen as things with causal
powers because they’re not taken to exist in any sense at all. And the
proof that this is our attitude is the lack of any desire (on the part of
scientists) to forge thick epistemic access to such objects in order to verify
they have the causal powers attributed to them (just as with space-time
points).
I exhibited the exercise of evaluating the posits of a scientific theory
at fair length in the case of a small piece of Newtonian physics: I showed
how various sorts of applications of the mathematical system which I
called Newtonian cohesive-body mathematics (Ncm) should be evaluated
in terms of what their ontology commits us to.
That certain posits prove to be items we’re not committed to, given
the successful application of a particular theory doesn’t mean they’ll con-
tinue to have that ontological status later. For example, that, in the New-
tonian context, the presence of fields as posits in the theory—indeed, the
couching of that theory entirely in terms of fields—doesn’t mean we’re
committed to such, is not a lesson that need be true of later theories.
This is not merely because of the fact that action-at-distance, in other
words, the instantaneous impact of the field throughout space, is replaced
Conclusion 225
by a field that travels at a finite speed. For the latter is simply a change in
the properties of a posit (the way that the replacement of a flat space-
time by a curved space-time is). It’s the (intended) application of the
theory that tells us whether the posit is to be taken as existing or not. So
too, for example, the fact that energy is located in space-time itself in
general relativity is insufficient for a commitment to that space-time. If,
however, thick epistemic access is forged to parts of space-time, then not
merely has that posit changed its properties, but its ontological status
within the theory has changed, as well, because we’re attempting thick
epistemic access to it.
There is a sense, however, in which what I’m offering is an ontologi-
cal program, although in a much weaker sense than are nominalist pro-
grams that undertake to rewrite all of science. This is that the ontological
views I’ve argued for in this book are part of the ordinary take on ontol-
ogy, and there is no guarantee that the results they dictate sit easily with
contemporary science. That is, the results of contemporary science may
force us to revise how we think about ontology, even in the most funda-
mental ways we currently do so. (We can be wrong about anything, and
in no way are philosophical doctrines immune to surprising developments
in science.)
I noted earlier that it’s not easy to determine what the posits of a
theory are that we take to exist, and I gave one reason for that, namely,
that when posits are drawn from a mathematical formalism, it’s not given
in that formalism which ones (if any) aren’t ultrathin. Here is another:
Our view of thick epistemic access is itself part of our scientific worldview.
Thus, it’s possible to be theoretically obscure what sort of things we have
(or are trying to have) thick epistemic access to. I think, for this reason,
the ontological evaluation of, for example, the quantum-mechanical for-
malism, is highly unobvious.3
I also had to set aside any detailed discussion of the special sciences
except for the occasional, and probably obscure, hint (see chapter 7, n.
20). There are fascinating issues about how the posits in such sciences
work, and how we should sort out which of those we take to exist and
which not.
Indeed, when I first set out to write this book, I was intending to
discuss issues about the special sciences, especially linguistics, because of
3
Different views of the quantum-mechanical formalism turn in part on different views
of which aspects of that formalism should be taken as real. Some of the wilder interpreta-
tions turn on taking the complex wave function itself as existing (on the grounds, usually,
that the laws about that wave function are true). Although the ontological views I’ve argued
for in this book don’t allow us to take the just mentioned ontological view of the wave
function, there is still a great deal of uncertainty about exactly what aspects of the quantum-
mechanical formalism should be taken as items we’re to gain thick epistemic access to (and
therefore items we take to exist). Recent experiments apparently illustrating thick epistemic
access to superpositions of a particle may indicate that a purely mathematical interpretation
of the probabilistic field is not tenable.
226 Conclusion
227
228 References
Cunningham, Walter R., and Adrian Tomer, 1990. Intellectual abilities and age:
concepts, theories and analyses. In Aging and cognition, mental processes, self-
awareness and interventions, ed. Eugene A. Lovelace, 379–406. Amsterdam:
North-Holland.
Dainton, Barry, 2001. Time and space. Chesham, UK: Acumen Publishing.
Davidson, Donald, 1965. Theories of meaning and learnable language. In Inquir-
ies into truth and interpretation (1986), 3–15. Oxford: Oxford University
Press.
, 1967. Causal relations. In Essays on actions and events (1986), 149–62.
Oxford: Oxford University Press.
, 1977. The method of truth in metaphysics. In Inquiries into truth and
interpretation (1986), 199–214. Oxford: Oxford University Press.
, 1999. Reply to J.J.C. Smart. In The philosophy of Donald Davidson, ed.
Lewis Edwin Hahn, 123–5. Chicago: Open Court.
Deutsch, Harry, 2000. Making up stories. In Empty names, fiction and the puzzles
of non-existence, ed. Anthony Everett and Thomas Hofweber, 149–81. Stan-
ford, Calif.: CSLI Publications.
Donnellan, Keith, 1974. Speaking of nothing. Philosophical Review 83:3–32.
Dummett, M. A. E., 1963. Realism. In Truth and other enigmas, 145–65. Cam-
bridge, Mass.: Harvard University Press.
Einstein, Albert, 1916. The foundations of the general theory of relativity. In The
principle of relativity (1923), trans. W. Perrett and G. B. Jeffery, 109–64.
New York: Dover Publications.
Enderton, Herbert B., 1972. A mathematical introduction to logic. New York:
Academic Press.
Feynman, Richard, Robert B. Leighton, and Matthew Sands, 1963. The Feynman
lectures on physics. Reading, Mass.: Addison-Wesley Publishing.
Field, Hartry, 1972. Tarski’s theory of truth. Journal of Philosophy 69,13:347–75.
, 1980. Science without numbers. Princeton, N.J.: Princeton University
Press.
, 1989. Realism, mathematics and modality. Oxford: Blackwell.
Folger, Tim, 2000. From here to eternity. Discover 21,12:54–61.
Franklin, Allen, 1993. The rise and fall of the fifth force. New York: American
Institute of Physics.
Friedman, Michael, 1983. Foundations of space-time theories: relativistic physics
and philosophy of science. Princeton, N.J.: Princeton University Press.
Galison, Peter, and Bruce Hevley, 1992. Big science: the growth of large-scale re-
search. Stanford, Calif.: Stanford University Press.
Gell-Man, Murray, 1997. Quarks, color, and QCD. In The rise of the standard
model: particle physics in the 1960s and 1970s, ed. Lillian Hoddeson, Laurie
Brown, Michael Riordan, and Max Dresden, 625–33. Cambridge: Cam-
bridge University Press.
Gottlieb, Dale, 1980. Ontological economy: substitutional quantification and
mathematics. Oxford: Oxford University Press.
Hacking, Ian, 1981. Do we see through a microscope? In Images of science: essays
on realism and empiricism, with a reply from Bas C. van Fraassen (1985), ed.
Paul M. Churchland and C. A. Hooker, 132–52. Chicago: University of
Chicago Press.
, 1983. Representing and intervening. London: Cambridge University
Press.
230 References
235
236 Index
Cohen, I. Bernard, 186 n. 11, 186– eleatic principle, the, 149 n. 8, 150,
187 n. 12, 191n, 204 154–155, 154 n. 21, 156 n.
coherentist epistemology, 109, 23
144–145 Enderton, Herbert, 17 n. 5
cohesive bodies, ontological status of, epistemic burdens, characterization of,
192–193, 220–221 127
Collins, Arthur, 126n Euclidean geometry, 160 n. 2, 161,
Colyvan, Mark, 62 n. 26, 76 n. 51, 162, 168–169
149 n. 8, 155–157 as a backdrop for Ncm, 166–167
confirmation holism, 27–28, 76–77, epistemic role obligation, 136, 143
118 characterization of, 135
constant angular velocity, 207–208, existence. See also existence predicate;
208n vernacular “exists.”
constitutive relations, 169 n. 14, 170, necessary and sufficient conditions
182 n. 1, 182 n. 2, 184, 185, on, 87–98, 112 (See also onto-
187 n. 14, 188 n. 16, 192, logical independence)
202–203, 205 existence predicate, 4, 49, 52 n. 6, 79,
characterization of, 168 80, 82, 84, 88–89, 114, 127,
co-reference*, 70n, 78, 79n, 92, 122. 223
See also extension* extension*, 149
definition of, 61–62
Cosmides, Leda, 154 n. 20 facticity of explanation, 45 n. 24
criterion for what a discourse is com- false theories, value of, 35–36
mitted to. See also Quine, his Feynman, Richard, 23 n. 18, 47
criterion for what a discourse is fictional entities, nonexistence of, 71–
committed to 72, 83, 93–98
indeterminateness of, 98–99 Field, Hartry, 30, 43, 62 n. 26, 83 n.
criteria for what exist, 5, 50, 149 5, 84 n. 7, 99 n. 33, 194 n.
indeterminateness of, 5–6, 98–99 32, 212–215, 214 n. 65
criterion transcendence, 126, 159, finite axiomatizability, 18, 19
201n. See also vernacular “ex- definition of, 17 n. 4
ists,” as a criterion transcen- Folger, Tim, 8 n. 9
dent term folk ontology, 4, 7–9, 51, 53, 77–78,
Cunningham, Walter R., 132n 116 n. 3, 117–122. See also ver-
nacular “there is,” evaluation of
Dainton, Barry, 207 n. 56, 208n, ontological commitments of;
209, 210 vernacular “exists”
Davidson, Donald, 17 n. 5, 54 n. 9, fallibility of , 225
57–58, 144 forces
decision procedure, 17–18 additivity of, 184 n. 5
deduction, idealizations of, 17, 22 as an inevitable manner of speaking,
deflating ontology, 6–7 188
Deutsch, Harry, 71, 95 n. 26 genuine vs. pseudo, 186–187, 189
Donnellan, Keith, 74 n. 44 explanatory role of, 185–186
Duhem, Pierre, 34 n. 11, 45 n. 24 fundamental vs. nonfundamental,
Dylan, Bob, 38 184–185
Newton’s commitment to, 186 n. 12
Earman, John, 8 n. 9 ontological dependence on,
Einstein, Albert, 47, 196, 204 n. 50, 190–191
205–207 thick epistemic access to, 185n
Index 237