International Phenomenological Society

Language as Thought and as Communication

Author(s): Wilfrid Sellars
Reviewed work(s):
Source: Philosophy and Phenomenological Research, Vol. 29, No. 4 (Jun., 1969), pp. 506-527
Published by: International Phenomenological Society
Stable URL: http://www.jstor.org/stable/2105537 .
Accessed: 28/10/2012 12:53

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

International Phenomenological Society is collaborating with JSTOR to digitize, preserve and extend access to
Philosophy and Phenomenological Research.

http://www.jstor.org
 LANGUAGE AS THOUGHTAND AS COMMUNICATION
WILFRID
SELLARS

My aim in this paper is to throw light from several directionson the

intimate connectionswhich exist between conceptual thinking and the
linguisticbehavior which is said to 'express' it. The position which I
shallultimatelydelineateand defend,thoughbehavioristicin its methodo-
logical orientationis not, initial appearancesto the contrary,behavioristic
in its substantivecontentions.It can, nevertheless,be characterizedas
an attempt to give a naturalisticinterpretationof the intentionalityof
conceptualacts.

The early sections (I-IV) stress the essentiallyrule-governedcharacter

of linguisticbehavior.I arguethat a properunderstandingof the nature
and status of linguistic rules is a sine qua non of a correct interpretation
of the sense in which linguistic behavior can be said to be (and not
merely to express) conceptualactivity. The second, and larger part of
the paper (Sections Vff.) is devoted to exploringthe sense (or senses)
in which language can be said to 'express' thought. A distinction is
drawn between three differentcontexts in which the verb occurs. It is
arguedthat they involve radicallydifferentmeaningswhich, if confused,
blur the distinctionbetweenlanguageas conceptualact and languageas
means of communication,and preclude the possibility of an adequate
philosophyof language.
I

There are many interesting questions about the exact meaning or

meanings of the term 'rule' in non-philosophicalcontexts. What, for
example, is the differencebetween a 'rule' and a 'principle'?Are prin-
ciples simple 'first'rules in that they are not special applicationsof more
generalrules? Or is the primarydifferencethat rules can be arbitrary?
Or are principlesrules for choosing rules? Is the principleof induction,
for example,a higherorder rule for choosinglaw-like statements,them-
selves construedas extra-logicalrules of inference?Though these ques-

506
 LANGUAGE ANDAS COMMUNICATION
AS THOUGHT 507

tions are intrinsically interesting and relevant to the general topic of this
paper, I shall not discuss them. For however the domain of norms and
standardsis to be stratifiedand botanized,the term 'rule' has acquired
over the years a technicaland genericsense in which it appliesto general
statementsconcerningthat which ought or ought not to be done or to
be the case, or to be permissibleor not permissible-. distinctionswhich
can be put in many differentways.
For our purposes,then, a rule is roughly a general 'ought' statement.
Such statementshave been traditionallydivided into hypotheticaland
categorical'oughts,'or, as it has often misleadinglybeen put, 'impera-
tives.' The distinctionbetween hypotheticaland categoricaloughts is an
importantone, though I believe that they are far more intimatelyrelated
than is ordinarilytaken to be the case.'
Hypotheticaloughts have the form "if one wants X, one ought to
do Y." They transpose a relation of implication between a state of
affairs X and a doing of Y into an implicationappropriateto practical
reasoning.In spite of their crucial importanceto a theory of normative
discourse,I shall have nothing to say about them, save by implication.
As far as anything I have so far said is concerned, a categorical
ought is simply one that is not, in the familiarKantian sense, a hypo-
thetical ought. I shall continue my division informallyby calling atten-
tion to the most familiar variety of general categorical oughts, those,
namely, of the form
If one is in C, one ought to do A.
Notice that althoughthis propositionis conditionalin form, it is not,
in the Kantiansense, a hypotheticalought; and it is as contrastedwith
the latter that, even though it is conditional,it is called categorical.By
applicationand the use of modus ponens one can derive conclusionsof
the form
S ought to do A
which not only are not hypotheticaloughts,but are categorical(non-iffy)
statements. Notice, by contrast, that from "If one wants X, one ought
to do A," togetherwith "S wants X" it is not correctto infer "S ought
to do A."
The importantfeature,for our purposes,of generalcategoricaloughts
of the above form is that for actual existenceto conformto these oughts
is a matter of the agents to which they apply doing A when they are
actually in the specified circumstanceC; and this, in turn, a matter of
1 For an exploration of this and related issues, see my Science and Metaphysics
(London, Routledge and Kegan Paul, 1968), Chapter VI (especially sections
XIV-XVII).
508 PHILOSOPHY
ANDPHENOMENOLOGICAL
RESEARCH

their setting about doing A when they believe that the circumstances
are C.
It follows that the 'subjects'to which these rules apply must have the
concepts of doing A and being in C. They must have, to use a current
tum of phrase, the appropriate'recognitionalcapacities.'Furthermore,
for the rule itself to play a role in bringing about the conformityof
'is' to 'ought,'the agentsin questionmust conceive of actions A as what
ought to be done in circumstancesC. This requiresthat they have the
concept of what it is for an action to be called for by a certain kind
of circumstance.

II

Importantlydifferentfrom rules of the above form - which may be

called, in a straightforward sense, rules of action - are rules that specify
not what someoneought to do, but how somethingought to be. Of these
an importantsub class has the form.
Xs ought to be in state cp,wheneversuch and such is the case
The purpose of such a rule is achieved to the extent that it comes to
be the case that Xs are in state p when such and such is indeed the
case. This time, however,the conformityof actualexistenceto the ought
does not, in general,requirethat the Xs which are, in a sense, the sub-
jects of the rule, i.e. that to which it applies, have the concept of what
it is to be in state p or of what it is for such and such to be the case.
This is obvious when the Xs in question are inanimateobjects, as in
the example.
Clock chimes ought to strike on the quarterhour.
Now ought-to-be's(or rules of criticism as I shall also call them),
thoughcategoricalin form, point beyond themselvesin two ways. In the
first place they imply (in some sense of this protean term) a reason, a
because clause. The explorationof this theme would seem to take us
back to the excluded topic of hypotheticalimperatives.In the second
place, though ought-to-be'sare carefullyto be distinguishedfrom ought-
to-do's they have an essential connectionwith them. The connectionis,
roughly, that ought-to-be'simply ought-to-do's. Thus the ought-to-be
about clock chimes implies, roughly,
(Other things being equal and where possible) one ought to bring
it about that clock chimes strike on the quarterhour.
This rule belongs in our previous category, and is a rule of action.
As such it requiresthat the item to which it applies (personsratherthan
chimes) have the appropriateconcepts or recognitionalcapacities.
The distinctionbetween ought-to-do's(rules of action) and ought-to-
 ANDASCOMMUNICATION 509
ASTHOUGHT
LANGUAGE

be's (rulesof criticism)standsout clearlywhen the examplesare suitably

chosen.A possibilityof confusionarises,however,when the ought-to-be's
concern persons rather than inanimateobjects. Consider,for example,
One ought to feel sympathyfor bereavedpeople
This example is interestingfor two reasons: (1) It is a rule conformity
to which requiresthat the subjectsto which it applies have the concept
of what it is to be bereaved.In this respect it is like a rule of action.
(2) In the absenceof a clear theory of action one might think of feeling
sympathy as an action. Thus a casual and uninformedlook might lead
to the subsumptionof the example under the form
One ought to do A, if C.
It is clear on reflection,however,that feeling sympathyis an action only
in that broad sense in which anythingexpressedby a verb in the active
voice is an action.
Nor should it be assumed that all ought-to-be'swhich apply to per-
sons and concern their being in a certain state whenever a certain
circumstanceobtains are such that the conformityto them of actualfact
requires that the persons in question have the concept of this circum-
stance. The point is of decisive importancefor our problem.To set the
stage, considerought-to-be'spertainingto the trainingof animals.
These rats ought-to-bein state cp,wheneverC.
The conformityof the rats in questionto this rule does not requirethat
they have a concept of C, though it does requirethat they be able to
respond differentiallyto cues emanatingfrom C. Since the term 'recog-
nitional capacity'is one of those accordionwords which can be used
now in one sense now in another,it is a menace to sound philosophy.
On the other hand, the subjects of the ought-to-do'scorresponding
to these ought-to-be's,i.e. the trainers,must have the concept both of
the desirablestate p and of the circumstancesin which the animals are
to be in it.
If we now returnto the sympathyexample, we notice another inter-
esting feature. If we compare the ought-to-be with the corresponding
ought-to-do,
(Other things being equal and where possible) one ought to bring
it about that people feel sympathyfor the bereaved,
we see that the 'subjects'of the ought-to-be (i.e., those who ought to
feel sympathy)coincide with the 'subjects'of the correspondingought-
to-do (i.e. those who ought to bring it about that people feel sympathy
for the bereaved). It is the same items (people) who are the agent-
subjectsof the ought-to-doand the subject-mattersubjectsof the ought-
to-be.
510 ANDPHENOMENOLOGICAL
PHILOSOPHY RESEARCH

III

It is obvious, from the above considerations,that if all rules of lan-

guage, were ought-to-do'swe would be precludedfrom explainingwhat
it is to have concepts in terms of rules of language.Now many rules
of languageare ought-to-do'sthus,
(Otherthings being equal) one ought to say such and such, if in C
and as such they can be efficacious in linguistic activity only to the
extent that people have the relevant concepts. It is therefore of the
utmostimportanceto note that many of the rules of languagewhich are
of special interest to the epistemologistare ought-to-be'srather than
ought-to-do's.For only by taking this fact into account is it possible to
carry out a programaccordingto which (a) linguistic activity is, in a
primarysense, conceptualactivity; (b) linguistic activity is through and
throughrule-governed.
Much attentionhas been devoted of late to linguisticactions 2 where
the term 'action' is taken in the strict sense of what an agent does, a
piece of conduct, a performance- the practical sense of action, as con-
trastedwith the generalmetaphysicalsense in which action is contrasted
with passion. The topic of linguistic actions, whether performatory,
locutionary, illocutionary,perlocutionary3, or perhaps, elocutionaryis
an important one. Indeed, it is importantnot only for a theory of
communication,but for epistemology,for there are, indeed, linguistic
actions which are of essential interestto the epistemologist:thus asking
questionsand seekingto'answerthem. On the other hand it can scarcely
be over-emphasizedthat to approachlanguagein terms of the paradigm
of action is to make a commitment which, if the concept of action is
taken seriously, and the concept of rule is taken seriously,leads to (a)
the Cartesian idea of linguistic episodes as essentially the sort of thing
brought about by an agent whose conceptualizingis not linguistic;(b)
an inability to understandthe rule-governedcharacterof this concep-
tualizingitself, as contrasted'with its overt expression.For if thoughtis
analogousto linguisticactivityto the extent impliedby Plato's metaphor
'dialogue in the soul,' the idea that overt speech is action and its rules
rules of action will generatethe idea that all inner speech is action and
its rules rules of action, which leads to paradox and absurditywithout
end.
2 I have in mind particularly John Austin and his students. The best statement

of this approach is to be found in Austin's How to do Things with Words (London,

Oxford University Press, 1963.)
3 For an explanation and defense of these distinctions see Austin's How to do
Things with Words.
 LANGUAGE AS THOUGHT AND AS COMMUNICATION 511

I propose, instead that the epistemologist,while recognizingthat lan-

guage is an instrumentof communication,should focus attention on
language as the bearer of conceptual activity. This is not to say that
the two aspects can be separatedas with a knife. Indeed, by pointing
out that ought-to-be'simply ought-to-do'swe have already recognized
that languageusers exist at the level of agents. Roughly, to be a being
capable of conceptualactivity,is to be a being which acts, which recog-
nizes norms and standardsand engages in practicalreasoning.It is, as
Kant pointed out, one and the same reason which is in some of its
activities'theoretical,'and in some of its activities'practical.'Of course,
if one gives to 'practical'the specific meaningethical then a fairly sharp
separationof these activities can be maintained.But if one means by
'practical'pertainingto norms, then so-called theoretical reason is as
larded with the practicalas is practicalreasoningitself.

IV

Even if it be granted than many of the linguistic oughts which are

of special interest to an epistemologistare ought-to-be's,the fact that
ought-to-be'sand ought-to-do'sare conceptuallyinseparablemight be
thoughtto precludea linguisticapproachto conceptualabilities.Clearly
primaryepistemicought-to-do's(and by calling them 'primary'I mean
simply that they are not the unfolding of ought-to-be's,whether as
primarythey are categoricalor hypothetical),pertainingto the systematic
use of linguisticabilities and propensitiesto arrive at correct linguistic
representationsof the way things are, presupposethe possessionof con-
cepts by the agents to which they apply. And since all oughtto-be's
unfold into ought-to-do'swhich, in their turn, presupposeconcepts, the
outlook for linguistictheory of concepts would seem to be dark indeed.
Yet the fundamentalclues for a resolutionof the problemhave already
been given.
To fix our ideas, let us consideran examplewhich, though simplified
to its bare bones containsthe essence of the matter:
(Ceterisparibus)one ought to respondto red objects in sunlightby
utteringor being disposedto utter 'this is red.'
This ought-to-berule must not be confused with (fictitious)ought-to-do
rule,
(Ceteris paribus)one ought to say 'this is red' in the presence of
red objects in sunlight
The latter presupposesthat those to whom it applies have the concepts
of 'red' objects,' 'sunlight,'and, even more important,of what it is to
512 RESEARCH
AND PHENOMENOLOGICAL
PHILOSOPHY

say 'this is red.' In other words, they must alreadyhave the conceptual
frameworkof what it is to do somethingin a circumstance.
The distinctionbetweensaying and uttering,or being disposedto utter,
is diagnosticof the differencebetween the 'ought-to-do'and the 'ought-
to-be.' It might be objected that to use languagemeaningfullyis to say
ratherthan merely utter. But to merely utter is to parrot, and we need
a conceptwhich mediatesbetweenmerely utteringand saying.
Notice that the ought-to-dowhich correspondsto the above ought-
to-be, namely
One ought to bring it about (ceteris paribus)that people respond
to red objects in sunlight by uttering or being disposed to utter
'this is red.'
presupposesthat its agent subjectshave a conceptualframeworkwhich
includes the concepts of a red object, or sunlight, of uttering 'this is,
red,' of what it is to do or bring about something,and of what it is for
an action to be called for by a circumstance.
On the other hand, this ought-to-do does not presuppose that the
subjectsin which the dispositionto utter 'this is red' in the presenceof
red objectsin sunlightis to be broughtabout have any of these concepts.
But what of the objectionthat the subject-mattersubjectsof the ought-
to-be coincide with the agent subjectsof the ought-to-doand hence that
they too must have the concepts in question? The answer should be
obvious; the members of a linguistic community are first language
learnersand only potentially'people,'but subsequentlylanguageteachers,
possessed of the rich conceptualframeworkthis implies. They start out
by being the subject-mattersubjects of the ought-to-be'sand graduate
to the status of agent subjectsof the ought-to-do's.Linguisticought-to-
be's are translatedinto uniformitiesby training. As Wittgensteinhas
stressed, it is the linguistic community as 'a self-perpetuatingwhole
which is the minimumunit in terms of which conceptual activity can
be understood.
Furthermorethere are radicallydifferentkinds of linguisticought-to-
be's: not only word-objectought-to-be's(or, as I have called them else-
where, language entry transitions)4, but also the ought-to-be'sformu-
lated by formationand transformationrules.
4 "Some Reflections on Language Games," Philosophy of Science, Vol. 21, No. 3,
1954 (Reprinted as Chapter 11 in Science, Perception and Reality). It is important
to note that a full discussion would refer to may-be's (or permitteds) as well as
ought-to-be's - otherwise the concept of 'free' as opposed to 'tied' (stimulus bound)
linguistic activity, essential to any account of the functioning of a conceptual sys-
tem, would be left out of the picture.
 LANGUAGEAS THOUGHTAND AS COMMUNICATION 513

The oughts governingutterancesas perceptualresponses to the en-

vironment are not ought-to-do's - though, as the pragmatistshave
emphasized,perceptionas an element in enquiryoccurs in a context of
actions, epistemic and otherwise. Similarlythe oughts governinginfer-
ence are not ought-to-do's.Inferringis not a doing in the conduct sense
- which, however,by no means implies that it is not a process. Again,
as the pragmatistshave stressed, inference as an element in enquiry
occurs in the context of action, epistemicand otherwise.
A languageis a many-leveledstructure.There are not only the ought-
to-be's which connect linguisticresponsesto extra-linguisticobjects, but
also the equallyessentialought-to-be'swhich connect linguisticresponses
to linguisticobjects. There could be no trainingof languageusers unless
this were the case. Finally, there would be no languagetrainingunless
there were the uniformitiespertainingto the use of practicallanguage,
the language of action, intention, of 'shall' and 'ought,' which, as
embodyingepistemicnorms and standards,is but one small (but essen-
tial) part of the conceptualstructureof human agency.
One isn't a full-fledgedmemberof the linguisticcommunityuntil one
not only conformsto linguisticought-to-be's(and may-be's)by exhibiting
the required uniformities,but grasps these ought-to-be'sand may-be's
themselves(i.e., knows the rules of the language.)One must, therefore,
have the concept of oneself as an agent, as not only the subject-matter
subject of ought-to-be'but the agent-subjectof ought-to-do's. Thus,
even though conceptualactivity rests on a foundationof conformingto
ought-to-be'sof uniformitiesin linguistic behavior, these uniformities
exists in an ambienceof action, epistemic or otherwise. To be a lan-
guage user is to conceive of oneself as an agent subject to rules. My
point has been that one can grantthis withoutholdingthat all meaning-
ful linguistic episodes are actions in the conduct sense, and all linguistic
rules, rules for doing.
A living languageis a system of elements which play many different
types of roles, and no one of these types of role make sense apartfrom
the others. Thus, while the mere concept of a kind of vocalizingbeing
a responseby a human organismin specified circumstancesto a certain
kind of object does make sense in isolation, this concept is not as such
the concept of the vocalizingas a linguisticresponse.For to classify an
item as linguistic involves relating it to just such a system as I have
been sketching. 'Word' goes not only with 'object' but with 'person,'
'ought-to-be's,''ought-to-do's'and much, much more.
514 PHILOSOPHY
ANDPHENOMENOLOGICAL
RESEARCH

Within the frameworksketchedabove, I propose to explore the idea

that insofaras it has conceptualmeaning,languageis essentiallya means
wherebyone thinkercan express his thoughts to others. Now the term
'thought'has a wide rangeof application,includingsuch items as assump-
tions, the solving of problems,wishes, intentions,and perceptions.It is
also ambiguous,sometimes referringto what is thought, sometimes to
the thinkingof it. To limit the range of my paper, I shall concentrate
on thought as belief, and since the latter term shares the ambiguity
indicatedabove, I note that for the time being at least, I shall be con-
cernedwith believingsratherthan things believed.
The following characterizationof the state of believingsomethingwill
serve to get the discussionunder way
Jones believes that-p = Jones has a settled disposition to think
that-p.
It would be foolhardy- indeed downrightmistaken- to claim that this
formulacaptures 'the' meaningof believes, and even more so to put it
by saying that 'a belief is a settled dispositionto think that something
is the case.' For, as with most, if not all, of the words in which philoso-
phers are interested,we are confrontedwith a cluster of senses which
resembleeach other in the family way.
To say that the senses of cognate expressionsbear a family resem-
blance to one another must not be taken to imply that they present
themselvesas a family, nor even that they constitutea family. Aristotle
seems to have thought that philosophicallyinterestingconcepts present
themselvesto us as familiesin which, with a little effort, we can discern
the fathers, mothers, aunts, uncles, and cousins of various degrees. In
some cases somethinglike this may be true. But the matter is rarely so
simple, and there is more than a little truth to the idea that the families
are 'created'by reconstruction(hopefullyrational)or regimentationrather
than found.5
If the above accountof belief gets us started,it does so by confronting
us with the equally problematicconcepts of disposition and thinking
that-p. Before stepping into these quicksands,let us ostensibly make
mattersworse by turningour attentionfrom believingitself to the more
complicatedconcept of the expressionof belief. For sound philosophical
5 Equally dangerous are such mephorical contrasts as those between 'paradigm'
and 'borderline,''shadow' and 'penumbra.'All suggest a sequential strategy accord-
ing to which, once we find the thread, we know how to begin and what kinds of
difficulty to expect.
 LANGUAGE
AS THOUGHT
ANDAS COMMUNICATION 515

strategycalls for the examinationof concepts as they function in larger

contexts, rather than subjectingthem to scrutiny in splendid isolation.
By taking elusive concepts together,one may limit the degrees of free-
dom which enable them separatelyto elude our grasp. If beliefs are to
be construedas dispositions,this strategywould have us seek to relate
the sense in which beliefs are 'expressed'to the sense in which the dis-
positions of things and persons are manifestedby what they do. This
suggeststhe schema
x expressesJones'belief that-p x is a manifestationof Jones sett-
led dispositionto think that-p.
If the right hand side of this attemptedexplicationwere clearcut and
unambiguous,substantialprogresswould have been made. But it isn't;
and our only hope is that a spark of clarity may result from rubbing
unclaritiestogether.
A first unclarityconcernswhat it is for a dispositionto be 'manifested'
by a doing, and how the class of doings by which a given dispositionis
manifestedis to be delimited.If the 'disposition'is of the familiarkind
to which we refer by such expressionsas 'an angry disposition'or, per-
haps, by such a term as 'humility,'then it would seem that, depending
on circumstances,any of a wide range of episodes could be its mani-
festation.Indeed, there is a sense in which, dependingon circumstances,
any of a wide range of episodes could count as a 'manifestation'of
Jones' belief that-p. But, to characterizebelief that-p as a settled dispo-
sition to think that-p, is, if sound, to narrowthings down in an inter-
esting way. For to do so, is to introducea conceptualtie between the
designationof the disposition and the kind of episodes which can be
said, at least in a primarysense, to 'manifest'it.
For if we ask what episodes manifest a disposition to V, when 'V'
representsa verb which standsfor a doing (e.g. 'laugh')the answermust
be, in the first instance,episodes of V-ing (e.g. laughing).We have con-
sequentlycommittedourselvesto the idea that it is episodes of thinking
that-p which are, in a primarysense at least, manifestationsof Jones'
dispositionto thinkthat-p;and consequentlythat it is episodesof thinking
that-p which are, in a primary sense, manifestationsof Jones' belief
that-p. This gives us the schema
x is a primary manifestation of Jones' belief that-p -e x is a thinking
that-p.
But now our troublesreally begin. For there is a prima fadie tension
between 'beinga thinkingthat-p'and being a 'manifestation'of anything.
The latter term carries with it the implication of 'making something
manifest,' i.e., apparent,(roughly) perceptible,observable.But, we are
516 ANDPHENOMENOLOGICAL
PHILOSOPHY RESEARCH

tempted to expostulate,what need be less 'manifest'than an episode of

thinkingthat-p.
It might be thoughtthat all we need do is replace 'manifestation'by
a term which lacks this implication.And there are, indeed, such terms
at hand - thus 'realization,''actualization.'The statements
episodesof thinkingthat-p are realizationsof the settled disposition
to think that-p
episodes of thinkingthat-p are actualizationsof the settled disposi-
tion to think that-p
trip easily off the metaphysicallytrainedtongue. But they are ruled out
by our strategy.For the concept with which we are concernedis that
of the expression of a belief, and 'expression'clearly has the same
implicationof 'overtness'or 'perceptibility'as does the 'manifestation'to
which our initial intuitionshave led us.
The boulder may have slipped, but perhaps it has not rolled to the
bottom. Our task may ultimatelyprove to be like that of Sisyphus,but
perhapswe are not yet forced to make a new beginning.To continueis
to look for a way of makingcoherentthe idea that episodes of thinking
that-p are the primaryexpressions(with all that this implies) of the be-
lief that-p.
To do so within the allotted space however, I must abandon the
leisurelydialecticwhich consultsintuitionat each stage of the argument,
and instead must draw upon the familiarityof standardphilosophical
moves. In terms of this new strategy,the obvious move is to espouse a
form of logical behaviorismaccordingto which, in first approximation,
'thinkingthat-p' is, in its most episodic sense, to be equatedwith 'can-
didly and spontaneouslyuttering "p"' 6 where the person, call him
Jones, who utters'p' is doing so as one who knows the languageto which
'p' belongs. I need not remind you of all the troubleswhich beset this
move. Some of them will be taken into account as the argument moves
along. But since, in any case, my strategy remains in a broad sense
dialectical,the fact that the above equation suffers from serious inade-
quaciesneed not preventit from playingan essentialrole in the argument
The phrase 'candidly and spontaneously'is intended to sum up an
open-endedset of conditionswithout which the suggestioncan't get off
the ground.Jones' thinkingthat-p obviously cannot be a quoting of 'p'
or utteringit on the stage in the course of acting. The qualifyingphrase
also clearly rules out the case where Jones is lying, i.e. using words to
6 Similarly, 'wondering whether-p' would be equated with 'uttering "p?"', 'wish-
ing that-p' with 'uttering "would (that) p"' and 'deciding to do A' with uttering
'I shall do A.'
 ANDAS COMMUNICATION
AS THOUGHT
LANGUAGE 517

deceive. Somewhatless obviously it is intended to imply that Jones is

not choosing his words to express his convictions. He is neither lying
nor speakingtruthfully.In a sense, as we shall see, he is not using the
words at all.
According to the behavioristic position we are now considering,
thinkingthat-p is, in its primaryepisodic sense, thinking-out-loudthat-p.
As thinking-out-loud,an utteranceof 'p' is not directedto an audience.
It is not, as such, a social act. Explicit performatives(e.g. 'I promise')
are clearly out of place in utteranceswhich are, in the desired sense,
to be thinkings-out-loud.Nor is it appropriateto characterizethinkings-
out-loud in terms of the categories of illocutionaryperformance- at
least those which require an audience (e.g. 'statement,''avowal,' 'argu-
ment')7 - even though exactly similarutteranceswould, in a context of
'communication,'be appropriatelyso characterized.

VI

It is importantto realize that the ways in which we classify linguistic

expressionsare not only bound up with the jobs they do, but with the
purposesfor which the classificationis made. Since these purposestend,
for obviousreasons,to concernthe role of languageas a means of com-
munication,i.e., as that by which we give information,warn, make
statements,predict, describe, etc., we should not be surprised,our be-
havioristwill tell us, if expressionswhich, as candidly uttered in non-
communicative contexts, are thinging-out-loud, are classified in a way
which is conceptuallytied to communication,and, hence, to functionsof
quite a differentorder of complexity.One needs only think of the dif-
ference between the purely logical characterizationof 'it is not raining'
as the 'negation'of 'it is raining,' and characterizingit as the 'denial'
of the latter, or note the social implicationsof classifyinga word as a
referring expression.
Thus the ways in which common sense, and not only common sense,
classifies linguisticexpressions,and the verbs which it uses to describe
what people do with them, are heavily weightedin the directionof lin-
guistic performancesin a context of communication.That it is legitimate
to view languagein this way is not to be doubted. Indeed, it is philo-
sophicallyimportantto be clear about the categoriesin terms of which
the varietyof ways in whichlanguagefunctionsin inter-personalexchange
7 We can grant that a thinking-out-loud that-p might be a constituent of a
reasoning-out-loud or a deliberating-out-loudon a certain topic.
518 PHILOSOPHY
AND PHENOMENOLOGICAL
RESEARCH

are to be understood. But there is a danger that exclusive concern with

this perspectivewill obscurethose connectionsbetweenthoughtand lan-
guage where the latter is not functioningas a means of communication.
The point is not that there are failures of communication,e.g. the
supposedhearer may be an inanate object mistakenfor a man or a
foreigner.It is not even that there are soliloquies, if by this is meant
cases of 'talkingto oneself.' It is the more radical point that thinking-
out-loudis a form of meaningfulspeech which doesn't consist in talking
to anyone at all, even oneself, and hence is not, in any ordinarysense,
talking.

VII
But before I develop this point let me returnto the formulawe were
consideringbefore this digressionon the orientationtoward contexts of
communicationof the categoriesin terms of which common sense, lin-
guistics,and many philosophiesof languageapproachlinguisticbehavior.
The formulawas
x is a primaryexpressionof Jones' belief that-p
- x is a primarymanifestationof his settled dispositionto think
that-p (i.e. is a thinkingthat-p).
The implicationsof the term 'manifestation'(and, for that matter, of
'expression')led us in the directionof a logical behaviorismaccording
to which the relevant sense of 'thinking that-p' is 'thinking-out-loud
that-p.' Thus reinterpreted,the formula becomes
x is a primaryexpressionof Jones' belief that-p
x is a primary manifestationof Jones' settled disposition to
think-out-loudthat-p (i.e., is a thinkingout-loud that-p).
It will be rememberedthat the point of this behavioristicmove was
to assimilate the sense in which an episode is a primary expression
(implyingovertness)of a belief to the sense in which an episode of, for
example,a piece of litmus paper turningred is a manifestation(implying
overtness)of its dispositionto turn red.
It should be noted in passing that in the case of the litmus paper we
seem requiredto expand the characterizationof the dispositioninto
dispositionto turn red, if put in acid.
This generatesthe suspicionthat if we are to continuewith our strategy,
we must similarlyexpand our analysisof 'Jones believes that-p'into
Jones has a settled dispositionto think-out-loudthat-p, if.
If what?There many pitfalls here, though we can, perhaps,cover them
up temporarilywith somethinglike 'if the questionwhether-parises.'To
do so, however, would immediatelyconfront us with a more serious
 LANGUAGE AS THOUGHT AND AS COMMUNICATION 519

difficulty.For it simply isn't the case that if a person believes that-p.

he utters 'p' let alone thinks-out-loudthat-p, whenever the question
whether-parises.
Confronted by this fact, we are strongly tempted to abandon our
strategyand say that if a person believes that-p,then (otherthings being
equal) whenever the question whether-parises, he tends to think (not
think-out-loud)that-p;to which we might add that if the circumstances
are appropriatehe may express his thoughtby uttering(saying?)'p.'

VIII
On the other hand, if, however?we are to continuewith our original
strategy,we must resolutelyput aside the temptationto draw the kind
of distinctionbetween thoughtand its expressionwhich this formulation
implies, and continue with the intriguingidea that an uttering of 'p'
which is a primaryexpressionof a belief that-p is not merely an expres-
sion of a thinkingthat-p, but is itself a thinking,i.e., a thinking-out-loud
that-p.
Yet the preceding remarks do remind us that we must take into
account the fact that there is a sense of 'express'in which we can be
said to express our thoughts by using language for this purpose. Thus,
we express our thought that-p by saying 'p.' Can we sophisticate our
logical behaviorismto do justice to this fact?
Let us take a closer look at the words 'thought'and 'express.'First
the latter: it will be noticed that the referenceto observabilityimplied
by the term 'manifestation'in the context 'manifestationof the disposi-
tion to think that-p' was absorbed by the behavioristinto the phrase
which describesthe disposition. Thus, 'manifestationof the disposition
to think that-p,' became, in effect 'actualizationof the disposition to
think-out-loudthat-p.'
Thus the behaviorist'sformulabecomes, in effect,
x is a primaryexpressionof Jones' belief that-p
x is a manifestationof Jones' settled dispositionto think that-p
- x is an actualizationof Jones' settled dispositionto think-out-
loud that-p (i.e., x is a thinking-out-loudthat-p).
is
It only too clear that by pushingthis analysis of the context 'expres-
sion of belief' in this directionthe behavioristhas lost contact with the
idea that people express their beliefs by using language.The point can
be put simply - indeed bluntly - by saying that the concept of the
actualizationof a dispositionis not, as such, the concept of an action,
whereas expressingtheir beliefs is somethingpeople do.
520 RESEARCH
PHILOSOPHYAND PHENOMENOLOGICAL

The statement
Jones, by saying 'p', expressed his belief that-p
requires an interpretation of saying p as an action which is undertaken
by Jones in order to express (to someone) his belief that-p. If we suspect
that Jones is lying, we could equally describe him as saying "p", but
we would then go on to say something like
Jones, by saying 'p', pretended to believe that-p.
In neither case could Jones' saying 'p' be construed as a case of
thinking (even 'out loud') that-p. Thus were Jones speaking truthfully,
the thinking immediately involved, if any, would be of the sort described
by such formulas as
Jones thought that saying '...' would express his belief that-p
Jones intended to express his belief that-p by saying ...
or, in the case of lying
Jones intended to pretend to believe that-p by saying ....
Thus, granted the validity of the concept of thinking-out-loud, the
thinking-out-loud which, were it to occur, would be immediately in-
volved in the situation formulated by
Jones, by uttering '...,' expressed his belief that-p
would be not
Jones thought-out-loud that-p
but rather
Jones thought-out-loud that saying '...' would express his belief
that-p
or, where Jones is lying,
Jones thought-out-loud that he would pretend to believe that-p by
saying ....
Needless to say, the latter thinking-out-loud would be self-frustrating in
the presence of the audience he intends to deceive.

Ix

If we leave behaviorism aside for a moment, we can add a new

dimension to the discussion by noting that the term 'express' in contexts
pertaining to thought has two radically different senses. The difference
can be brought out by relating these senses to two different contexts,
namely,
(1) Jones expressed his thought (belief) that-p by saying ...
(2) Jones' utterance of 'p' expressed his thought that-p
I shall call the former the 'action' sense of express, and the latter,
 ANDASCOMMUNICATION 521
ASTHOUGHT
LANGUAGE

for want of a better term, the 'causal'sense. Both, as we shall see, are
to be distinguishedfrom a third sense illustratedby the context
Jones' utteranceof 'p' expressedthe thoughtthat-p
where the phrase 'the thought that-p' stands for an abstract entity, a
thoughtin Frege's sense (i.e., in one sense of this term, a 'proposition').
I shall call this the logical (or semantical)sense of 'express.'
Although my ultimateaim is to show how a logical behavioristmight
draw these distinctions,my initial move will be to discuss them in more
traditionalterms. I shall, therefore,constructa regimented(I dare not
say idealized) model accordingto which, in the course of learning to
speak a language,a child acquires the capacity to be in mental states
which are counterparts, in a sense to be analyzed, of the utterances
which come to belong to his repertoryof linguistic behavior.The idea
can be blocked out in two steps:
(a) A mental episode which is a thinking that-p is correlated,in
a certain linguistic community,with a piece of linguistic be-
havior which stands for (expressesin the logical or semantical
sense) the thought (proposition)that-p
(b) In the initial stages of the child's mastery of the language,
wheneverit has a thoughtthat-p, this thought is manifestedin
a purelyinvoluntaryway by the correspondingverbal behavior.
As our model for understandingthe sense in which the uttering of
'p' is the involuntarymanifestationof a thinkingthat-p, let us take the
instinctive connection between a pain and a piece of unlearnedpain
behavior.The fact that a connectionbetween states A and B of a child
is, in some sense, learned rather than instinctive,acquiredrather than
part of its initial equipment,by no means entails that either A or B is
under the child's voluntarycontrol. Not all learningto, do somethingin
a broad sense of 'do' consists in the addition of new behaviorsto the
stock of things that are under one's voluntarycontrol.
The key featureof our model is that the acquiredconnectionbetween
the mental act and the verbal behavior is not to be construedon the
action model of 'using the behavior to express one's thought.' Thus,
verbal behavioris not in our child's voluntarycontrol in that, although,
once the languageis learned,a necessaryand sufficientconditionof the
child saying 'p' is that it thinks that-p, the saying is the involuntary
manifestationof the thinking.
Notice that the model allows the child a rich vocabulary,including
the languageof intention and resolve as well as the languagein which
matters-of-factare stated. It also allows that the child learns to verbalize
522 RESEARCH
ANDPHENOMENOLOGICAL
PHILOSOPHY

about verbalbehaviorand even about the mentalacts of which its verbal

behavioris the involuntarymanifestation.

X
We are now in a position to weaken our model and still make our
point. We need not suppose that the child remains a chatterbox.We
can suppose it to acquire the ability to keep its thoughts to itself in
the sense that it can effectivelytell itself to keep quiet, without ceasing
to think. We can grant that to this limited extent its verbal behavior
becomes under its voluntary control. When it is thinking without
speaking,we shall say that it is in a keeping-its-thoughts-to-itself frame
of mind. When not in this frame of mind, it thinks out loud. 8 Thus,
'Thinkingout loud' remainsthe primaryform in which thinkingoccurs.
The child'skeepingits thoughtsto itself can be comparedto the opening
of a general switch which breaks (or, to mix metaphors)short circuits
the initial acquiredconnectionbetweenthoughtsand verbalbehavior.
At this stage, the child has no conception of locutionary acts (e.g.
predicting,telling) as verbal behaviorwhich can be engaged in whether
or not one is thinkingthe correspondingthoughts.It has no concept of
saying "p" without thinkingthat-p.
On the other hand, it is perfectly capable of having concepts of
actions involvingthinkingout loud. Thus, wonderingout loud about the
weather; "I shall wonder out loud about the likelihood of rain." It is
importantto see that this by no means entails that there is such a thing
as an action of thinkingout loud that-p.Even in our more sophisticated
frameworkthere is no such thing as an action of thinkingthat-p, though
there is the action of deliberating(i.e., deliberatingout loud) what to
do. By granting,as we must, that it can conceive of actions consisting
of thinkings out loud, we admit a further sense in which its verbal
behavior (as thinkingsout loud) would be under its voluntarycontrol.
The child's verbal behaviorwould express its thoughts,but, to put it
paradoxically,the child could not express them.
Notice, also, that althoughits linguisticbehaviorwould be meaningful,
and we could say of each of its utteranceswhat, specificallyit meant, e.g.
Jones' utterancemeant 'it is raining,'
It would, on our assumptions,be incorrectto say, for example
Jones, by uttering '...' meant (to convey) ..
8 The concept of 'thinking out loud' appropriate to this model should not be

equated with thinking-out-loud as construed by the behavioristic position we have

been considering. The latter does not recognize 'mental episodes' in the sense
required by the present model.
 LANGUAGE AS THOUGHT AND AS COMMUNICATION 523

For the latter supposes that Jones has the concept of an action of
uttering '....' as a piece of linguistic behavior which could exist in-
dependentlyof its being the "spontaneousverbal expression" of the
correspondingmental act. There being no such action as bringingabout
a specific mental act, there could be no such thing as bringingabout a
thinkingout loud for the purposeof conveyinga thought.
In other words, just as our regimentingfiction enables us to draw a
distinctionbetween a sense in which a mode of verbal behavior can
express thoughtswithout being used to express them, so it enables us
to distinguishbetween the context
utterance of E (in L) means ---
and the sense of 'means,'closely related to 'intends,'which involves the
context
Jones, by uttering E, means (to convey) ...
The familiarsaw that words have meaningonly because people mean
things by them is harmlessif it tells us that words have no meaningin
abstractionfrom their involvementin the verbal behavior of language
users. It is downrightmistaken if it tells us that for an expressionto
have a certainsense or referenceis for it to be used by people to convey
the correspondingthought. Rather, we should say, it is because the
expressionhas a certainmeaningthat it can be effectivelyused to convey
the correspondingthought.

XI

Let us now return to the initial accounts we gave of belief and its
expression.The first thing to note is that if we were to reformulatethem
in termsof our model we would get somethinglike the followingschema
Jones believes that-p - Jones has a settled disposition to think
that-p, if the question occurs to him
whether-p,and, indeed, to think out loud
that-p, unless he is in a keeping-his-
thoughts-to-himselfframe of mind.9
We also get the followingformulaewith respectto 'expressionof belief':
x is a primaryactualizationof Jones'belief that-p-- x is a thinking
that-p (and, indeed, a thinkingout loud
that-p unless he is in a keeping-his-
thoughts-tohimself frame of mind.)
' The 'if the question occurs to him whether-p' condition can be taken to cover

all cases in which, where the alternatives 'p' and 'not p' are relevant to his course
of thought, he thinks that-p, even if the question whether-p is not actually raised.
524 PHILOSOPHY RESEARCH
ANDPHENOMENOLOGICAL

x is a primaryexpressionof Jones' belief that-p -e x is a thinking

out loud that-p.
Thus, where Jones is in a thinkingout loud frame of mind, the verbal
behavioris both an actualizationof and, in the 'causal'sense, an expres-
sion of his belief, both a thinkingand an expressionof thought.

XII
But what will our logical behavioristsay to all this? Clearlyhe will be
unhappyabout our uncriticalacceptanceof mental acts as covert inner
episodes. What moves might he make? He may well accept our initial
formula
Jones believes that-p Jones has a settled disposition to think
that-p.
But he will emphasizethe 'settled,'which we have not yet done, and will
call attentionto the fact that it presumablycontrastswith something.It
is not obviouswhat the contrastingadjectiveshouldbe, but it, too, should
apply to dispositions.Let us, he suggests, try 'proximate,'drawingon
the contrastbetween 'settled'and 'nearthe surface.'Another appropriate
contrastwould be providedby 'short term.'
Objects, as is well known, can have causal propertieswhich are not
so to speak, immediatelyavailable. Thus iron attractsfilings, if it has
been treated in a certain way. A proximatedispositioncan roughly be
characterizedas one which is immediatelyavailable.
Our logical behaviorist,consequently,suggeststhat
Jones believes that-p = Jones has the settled dispositionto have
short term, proximate dispositions to
think-out-loud that-p, if the question
whether-parises, and he is in a thinking-
out-loud frame of mind.
In other words, our logical behavioristconstruesthe contrastbetween
fleeting thought episodes and settled beliefs as falling within the broad
categoryof dispositions,and hence construesthe 'covertness'of thoughts
as simply a special case of the covertnessof dispositions.Flammability,
he remindsus, is not a covert flame.
Many featuresof our previousdiscussioncan be fitted into this frame-
work, once its distinctivecharacteris understood.Thus, the behaviorist
substitutesfor the previous account of the child's candid and spontane-
ous verbalbehavioras the expression(in the 'causal'sense) of classically
conceivedepisodesof thought, an account accordingto which a
thinking-out-loudthat-p
is simply an 'actualization'of a
 LANGUAGE
AS THOUGHT
ANDAS COMMUNICATION 525

short term proximate, disposition to think-out-loud that-P.

In the non-behavioristic model we stipulated that the child be unable
to verbalizewithout thinking the appropriatethought, in other words,
that only if it has the mental act of thinkingthat-p does it utter 'p.' In
the behavioristicreconstructionframework,the correspondingstipulation
would be that all utterancesof 'p' be thinking-out-loudthat-p.
Both stipulationscould be formulatedin the same words, thus 'the
child utters "p" only in the course of thinkingout loud that-p.' But the
two concepts of thinking out loud are radically different. In the non-
behavioristicmodel, the phrase 'thinking-out-loud'referredto thoughts
togetherwith their verbal expression.In the behavioristicreconstruction
it is to be taken as an unanalyzedexpressionwhich means roughlythe
same as 'candid'spontaneousverbalbehavior,'but serves, by its hyphen-
ated mode of composition,to emphasizethat the basic meaningfulness
of candid, spontaneousverbal behavioris not to be construedin terms
of its being the reverberationat the tip of the tongue of covert episodes
which are thoughtsproperlyspeaking,in accordancewith the schema
x is candid, spontaneousverbal behavior = is an expression10 of
thought

XIII

It is importantnot to confuse logical behaviorismwith what might be

called logical physicalism.I mean by the latter the view which denies
that, to quote Chisholm,"whenwe analyze the kind of meaningthat is
involved in naturallanguagewe need some conceptswe do not need in
physics or behavioristics."'I Chisholmthinks that to deny the need for
such an irreducibleconcept is tantamountto trying to "analyze the
semantics ... of natural language in a physicalistic vocabulary of a
behavioristicpsychology with no undefined semantical term and no
referenceto thoughts."12
In the essay whichled to the correspondencefrom which I am quoting,
I had arguedthat the concept of meaningwhich belongs in the context
E(in L) means ---
is not to be analyzed in terms of a reference to 'thoughts.'Thus I
rejected any analysisalong either of the following lines
E(in L) means --- = candid and spontaneous utterances of F
causally express thoughts pertaining to ---
10 'Expression'in the causal sense, i.e., a manifestation at the 'surface' of a covert
process which is its cause.
11 Minnesota Studies in the Philosophy of Science, Vol. II, p. 523.
12 Ibid.
526 PHILOSOPHY
ANDPHENOMENOLOGICAL
RESEARCH

E(in L) means - - = speakers of L use E to express their

thoughts pertaining to ---
where 'thought'is to be taken as referringto classicallyconceivedinner
episodes or mental acts.
On the other hand, though I denied that 'means'in the sense appro-
priate to the context 'E(in L) means - - -' is to be analyzed (defined) in
termsof a referenceto thoughts,I also arguedthat it cannotbe analyzed
in physicalisticterms. From Chisholm'spoint of view this was a blatant
attemptto have my cake and eat it. As he saw it, to admit that "to
analyze the kind of meaning that is involved,in natural language"we
need a distinctivelysemanticalterm ('means')which cannot be analyzed
in physicalisticterms, but deny that the explicationof this distinctively
semanticalterm requiresa referenceto thoughtshas all the appearance
of paradox.
The correspondencewent on at some length, and although some
progresswas made, the issue was never really joined. As I now diagnose
the situation some ten years later, the cause of this failure was my in-
ability to clarify adequatelytwo,points:
(a) The exact nature of statements of the form 'E(in L) means -- -'
(b) The exact relationof the conceptof meaningto that of thought.
The space which remainsis too short to do anythingmore than indicate
the moves I should have made.
My basic move should have been to clarify along the lines of the
presentpaper the distinctionbetween the contexts
person expresses
and
utteranceexpresses.
My second move should have been to give a more adequateclarifi-
cation of the concept of meaningas it occurs in the context 'expression
(in L) means -- -' (as contrasted with the context 'person, by uttering E,
means - - -.') At the time of the correspondence I was unable to do
much more than offer the rather cryptic suggestionthat statementsof
this form are (a) sui generis, (b) convey (ratherthan describe)how the
subject expression is used, by exhibiting an expressionin the hearer's
active vocabularywhich has the same job - the idea being that by
rehearsinghis use of the latter, he will be able to grasp the use of the
former. As I have since argued,'3 to say what an expression means is
to classify it by the use of a sortal predicatethe applicationof which
impliesthat the expressionin questiondoes the job in its languagewhich
13 Most recently in Science and Metaphysics, Chapter m.
 ANDASCOMMUNICATION 527
ASTHOUGHT
LANGUAGE

is done in the speaker's language by an expression from which the

predicateis formed. Thus, roughly
'und' (in German)means and
has the form
'und's (in German)are 'and's
where 'and"' is a sortal predicateof the kind in question.
But above all I should have made it clear that in my view the funda-
mental concept pertainingto thinking is thinking-out-loudas conceived
by our logical behaviorists.'4This is not to say that I agree with him
in rejectingthe classical conception of thoughts as inner episodes in a
non-dispositionalsense. Rather I accept mental acts in somethinglike
the classicalsense, but arguethat the concept of such acts is, in a sense
I have attemptedto clarify, a derivativeconcept.
Finally, I should have emphasizedmy total commitmentto the thesis
that the concept of thought essentiallyinvolves that of intentionalityin
the following sense. To say of a piece of verbal behavior that it is a
thinking-out-loud,is to commit oneself to say of it that it means some-
thing, while to say of it specificallythat it is a thinking-out-loudthat-p,
is to commit oneself to say of it that it is a piece of verbal behavior
which means p.
Thus, at the primarylevel, instead of analyzingthe intentionalityor
aboutnessof verbalbehaviorin terms of its expressingor being used to
expressclassicallyconceivedthoughtsor beliefs, we shouldrecognizethat
this verbal behavioris already thinkingin its own right, and its inten-
tionality or aboutnessis simply the appropriatenessof classifyingit in
terms which relate to the linguisticbehaviorof the group to which one
belongs.

UNIVERSITYOF PITTSBURGH.

14 The priority in question, to use Aristotle's distinction, is in the order of

knowing as contrasted with the order of being. As an analogy, notice that con-
cepts pertaining to things as perceived by the senses are prior in the order of
knowing to concepts of micro-physicalparticles, whereas, (for the Scientific Realist)
micro-physical particles are prior in the order of being to objects as perceived by
the senses.