IN THE BEGINNING, there was Frege who approached the philosophical
study of language with his gaze firmly fixed upon one of those "large, vexing
subjects," the "eternal structure of thought."[2] Michael Dummett attributes the following
three theses to Frege.
". . .first, that the goal of philosophy is the analysis
of the structure of thought [that is, the objective and eternally
existing contents of thought]; second, that the study of thought is to be
sharply distinguished from the study of the psychological process of
thinking; and, finally, that the only proper method for analysing
thought consists in the analysis of language."[3]
A central aim of semantics, for Frege, is thus the elucidation of
how language expresses thought contents.
'Hesperus = Phosphorus', maintained Frege, expresses a
significant piece of information, a nontrivial thought content. Our semantic account of names must
explain how this is so. Frege
concluded that the contribution of the two names to the thought content must be
different.
Frege's sense-reference approach not only explains the contributions
of names to thought contents, it does so in a way that respects what I will call
the "intentionality intuition."
This is the powerful traditional idea that in order to be thinking about
something, one must have a cognitive fix on it, that something in one's
thought must correctly distinguish the referent from everything else in the
universe.[4] 'Hesperus' and 'Phosphorus', then, not
only contribute differently to the thought content, but, holds Frege, what each
name contributes is its distinctive mode of presentation, its own cognitive fix
on the referent. That a semantic
account of a name must make plain the cognitive perspective on the referent that
is associated with the name is another crucial feature of the Fregean
perspective, one intimately related to Frege's emphasis on thought.
Frege's outlook on the business of semantics thus eliminates semantical
accounts like Mill's "pure denotation" view of proper names, recently
championed by many of us.
Millian accounts make the semantics of the two names the same, and so
will not be able to explain what Frege took to be the very datum, that sentences
like 'Hesperus = Phosphorus' express nontrivial thought contents. The Millian approach to names, moreover,
fails to explain the speaker's cognitive fix on the referent. Even worse, that approach, at least as
it's been recently developed, implies that the speaker need not have much of a
cognitive fix, perhaps none at all.
Kripke, to mention one prominent example, takes it to be plain that one
can refer by proper name even if one has very little information about the
referent--not nearly enough to individuate it.[5] So much the worse, from Frege's point of
view, for cognitively insensitive Millian accounts.
One might, though, approach the philosophical study of language in a
radically different spirit, that suggested by Walker Percy's remarks. Let's focus for a moment not upon
language vis-a-vis thought, but upon language vis-a-vis the realm of things
language is used to talk about, or, even better, vis-a-vis our practices of
talking about things.
The social practices that constitute natural language are, after all,
pretty fascinating in and of themselves.
Articulated speech is indeed distinctively human, the first thing that
Percy's Martian notices about us.
And thought is one of Percy's "large, vexing subjects," one that might
better be approached a bit later--which is not to say that the two subjects are
not, in the end, intimately related.
My semanticist thus fixes his gaze upon language as a social,
institutional arrangement, and upon speakers as participants in a social
practice.
Speaking, it occurs to him, like other kinds of practical mastery, does
not presuppose theoretical understanding of the practice. We are indeed fortunate--God, so to
speak, has been good to us--that articulable insight is not necessary, for it is
extremely difficult to attain.
Indeed, speakers, and other practitioners, may well find their own
practices theoretically impenetrable.
Adequate theoretical characterizations of one's practices will
typically not be available to introspection. Nor will competent practitioners
typically be able even to select some correct characterization from a list
of fairly plausible candidates.[6]
The semanticist thus sees himself as engaging in an anthropological study
of the institutional arrangements that constitute natural language.[7] His charge, more specifically, is
to provide an account of the semantics of our linguistic practices. Which features of the total
communication situation do our practices count as determining the
references of proper names?
What, as our practices go, links up a particular name (or utterance) with
a particular referent? This is
the sort of question in which he is interested.[8]
Frege's sense-reference account might be seen as providing an answer to
this latter question. Contemporary
anti-Fregeans have argued that Frege's is not a good answer. It fails to accurately reflect the
character of our practices.
Perfectly competent speakers often fail to have available the sort of
information required of them by Frege's account. They often lack anything like purely
qualitative individuating beliefs about the referents of the names that they
use. The beliefs that they do have,
moreover, often correctly apply to individuals other than the referents of the
relevant names, and so on.
Frege, moreover--and this is a point of importance for distinguishing the
two conceptions of semantics--does not put forth his sense-reference account as
an answer to our anthropological semanticist's question: "What, according to actual linguistic
practice, determines the reference of proper names?" Frege's picture provides, inter alia, an
answer to this question, but this is not his focus at the beginning of "On Sense
and Reference."[9] His primary concern is with explaining
the contribution of names to thought contents. If we use 'semantics' in the
second, and non-Fregean way, we can say that Frege's interest wasn't primarily
semantical.
One with a Fregean conception of semantics, on the other hand, might well
wonder about the very relevance to semantics--in his sense--of Millian-style
accounts of names. The thesis that
the reference of a name depends upon, as Donnellan and Kripke urge, a historical
chain of communication, even if this thesis formulates some yet-to-be-classified
kind of truth about our practices with proper names, fails to answer the
specifically "semantic" questions about the contribution of names to thought
contents, and about the cognitive fix involved in the use of names. If, as we are sometimes told, a
theory of meaning is a theory of understanding, then it is far from clear that
Mill, or the Donnellan-Kripke approach, tells us anything about meaning.
Mill's contemporary sympathizers have indeed often been attacked for the
alleged cognitive insensitivity of their view. Millians have not been conscious of deep
differences between their conception of semantics and Frege's, and so they have
often been embarrassed by the apparent failure of their semantics to yield
illumination of the sort demanded by Frege. Alternatively, they have twisted
and turned to show that their semantic apparatus can be pressed into
cognitive service.[10]
The anthropological conception of semantics yields a natural response on
behalf of the Millian. The aim of
the anthropological semanticist is not, after all, to solve Frege's
problems. Nor does the
anthropological semanticist presume that his work will yield such
solutions. It is not at all obvious
that elucidating the social reference-determining conditions will explain the
cognitive dimension, for example, the informativeness of 'Hesperus =
Phosphorus'. The explanation of the
latter might, for example, turn upon, in Putnam's terms, what's in the head of
the speaker. Reference, on the
other hand--at least if the Millian is correct--has little to do with the head
of the speaker. The anthropological
semanticist, on the other hand, need not assume that his work will be of no help
in illuminating the cognitive dimension. He can adopt, as they say, a "wait and
see" attitude.
Philosophic debates in which adversaries argue at cross purposes, as in
the present case, are typically fueled by deep, unarticulated differences. My suggestion has been that contrary to
the appearances, we should not think of the two approaches as engaged in a
single explanatory project. Indeed,
their explanatory projects, while displaying considerable overlap, differ
substantially. This difference in
explanatory project, the difference in conception of the semantic enterprise,
does not, however, exhaust the deep, unarticulated differences of which I
spoke. That their semantic projects
sharply diverge is itself a symptom of a much deeper divide between Fregeans and
anti-Fregeans. My central aims here
are to call attention to differences at the level of broad philosophic
outlook, and to draw some implications for the area of intersection of
philosophy of language and philosophy of mind in which Frege and his followers
have been so interested.
The anti-Fregean view, it turns out, is far from cognitively
insensitive. That view, seen in the
context of the broad outlook to be proposed here, does provide a most natural
way of thinking about the cognitive dimension. Mill's remarks on names, and those of
Donnellan and Kripke as well, do indeed fail to provide for the sort of account
of the cognitive dimension that Frege sought. Their not providing for such an account
is indeed a virtue, for a Frege-style account, I will argue, presupposes a
Cartesian perspective that we have reason to reject.
2. FREGE'S
CARTESIANISM
Frege, like his recent critics, never does formulate--or even gesture
towards--a comprehensive philosophical outlook. His semantical work, at the same time,
is grounded in strong intuitions about intentionality, the contents of thought,
and related matters in metaphysics, epistemology, and the philosophy of
mind. There is, I submit, a big
picture lurking in the wings. I see
in Frege's work the deep influence of the Cartesian tradition. I have in mind here not so much specific
Cartesian doctrines, as a tendency of mind, a way of approaching the philosophy
of language and the philosophy of mind.
Frege's sense-reference perspective, his emphasis on the connections
between language and objective thought contents accessible to the mind, rather
than on, say, the connections between language--thought of as a public, social
institution--and the world, bespeaks this Cartesian influence.[11]
Frege might well seem, however, a most unlikely neo-Cartesian. Frege himself, as Michael Dummett
emphasizes, has played a crucial role in the 20th century anti-Cartesian
revolution. Frege's own
revolutionary contribution, as Dummett notes, consisted in making the
philosophical study of meaning, rather than skepticism and the theory of
knowledge, the starting point in philosophy. Frege, moreover, emphasized what we
might call "the publicity of thought content," the idea that the thoughts
we express with language are not in principle private to the minds of individual
thinkers, but are in the public, albeit a non-physical, domain.
Consider, however, the Fregean semantic perspective vis-a-vis the
Cartesian "mirror-of-nature" tradition in the philosophy of mind. That tradition, it is often noted, sees
the mind as set against nature, as the repository of images and conceptual
representations of things.
Pieces of language become meaningful by being associated with the
conceptual representations. Leaving aside the question of images, and the
probably related Fregean "ideas," isn't this picture at least a very close
relative of Frege's?
One difference, already noted, is Frege's anti-psychologistic
platonism, his insistence that senses do not reside in the mind, but rather in a
third, objective realm. Still,
there is for Frege a realm of representations, distinct from the things
represented and accessible to the mind, and linguistic expressions
become meaningful only by being associated with these representations.[12]
Frege's making the representations abstract and therefore public entities
may obscure, moreover, an important individualistic strain in his
view. The reference of a proper
name depends, for Frege, not upon anything like the role of the name in the
public language, but rather upon the individual's associating a particular
sense with the name.[13],[14]
Frege's view shares other characteristic features of a Cartesian
orientation. It is, Frege
tells us, the representations, senses, and not words, that refer in the
primary instance.[15] The reference of words is thus
derivative from the reference of senses.[16] Proponents of a Cartesian orientation
also characteristically emphasize the clarity and distinctness of the
representations, and make mathematical concepts the paradigm, and again,
Frege is no exception.
So much for Frege's Cartesian perspective on individual terms and the
conceptual representations with which they are associated. The same themes reappear--perhaps more
strikingly--when we consider Frege's treatment of language at the level of whole
sentences. Just as the vitality--to
use Wittgenstein's metaphor--of individual expressions derives from their
association with senses, so the vitality of whole sentences derives from
their association with sentential senses.
Just as the senses of singular terms, rather than the terms themselves,
are the things that refer in the most basic sense, so sentential senses--the
thought contents--are the primary bearers of truth and falsity. Fregean thoughts, like their
constituent senses, are well-defined and eternally existing. They constitute, Frege tells us,
the "common treasure of mankind."
Frege, although his focus is directed towards Plato's heaven and not
towards social practice, would acknowledge the platitude that natural language
is a social, institutional arrangement.
That certain sentences get correlated with certain thought contents is,
he would surely agree, an artifact of human institutional
arrangements. Frege would
insist, however, that one's thinking a certain thought content is no matter of
human convention, institutional arrangement, or anything of the like. It is a matter of one's mind grasping an
objectively existing content. More
important, then, at least in a sense more important, than the fact that
sentences express such "thoughts" is the fact that by uttering sentences we
assert them, we give voice to the thought contents that we are thinking. The traditional Fregean account of the
"propositional attitudes," for
example the idea that belief consists in a relation between a person and a
thought content, emerges directly.
3. A SOCIAL, NATURALISTIC
ALTERNATIVE
Frege, we might say, puts forth a thought-driven[17] picture of
language. Language, if we overlook
its imperfections, is thought externalized. Frege's conception contrasts
dramatically with the one I want to develop, the one I see rationalizing
the work of Frege's recent critics.
My picture shares much with, and owes a great deal to, that of
Wittgenstein.[18] Wittgenstein, although some
Fregeans are fond of claiming him for their own, gives voice to a radically
different perspective, one less representationalist, and arguably more
naturalistic. I don't want here to
engage in Wittgenstein exegesis, an even trickier business than the question of
how best to read Frege. So let me
just sketch my alternative.
The approach I have in mind, in stark contrast to the Cartesian
tradition, denies that pieces of language become meaningful by being
associated with representations, mental or objective. It is here that the connections between
Wittgenstein and contemporary anti-Fregeans emerge most clearly. Consider Putnam's slogan, slightly
adapted: Meaning ain't in, nor is
it available to, the head. Indeed,
a central lesson of the Philosophical Investigations, at least
as I read it, is that there is much less available to the head than one might
have supposed, and further that whatever is intellectually available is less
relevant to philosophers' questions about language (and even thought) than one
might have supposed.
If the vitality of linguistic expressions is not a function of associated
representations, of what is it a function?
The broadly Wittgensteinian answer is that the significance of a piece of
language is a function of its embeddedness in social, linguistic
practice. The problem for those of
us--virtually everybody--brought up Cartesian is to make this more concrete, to
somehow allow us to get the feel for how meaningfulness could be a function of
anything other than representations.
The Donnellan-Kripke historical chain picture--whether or not it provides
the last word on its subject matter[19]--can be pressed
into service here as suggesting a model of how significance might depend
upon social practice, and not upon representations. An introductory philosophy student,
quite ignorant about Aristotle and his accomplishments, asks, "Who was
Aristotle? Was he the one who
believed that everything was water?"
The name 'Aristotle' as it occurs in the student's questions,
surmises, and assertions, makes reference to Aristotle, our Aristotle, in virtue
of--as the Donnellan-Kripke sketch goes--a historical chain of
communication that stretches back to something like an original
bap-tism. Notice that the
Donnellan-Kripke account gives the name a role in the public language; it
functions as part of a public, name-using practice. The name, as uttered on these
occasions, has a conventional referent.[20] The name connects to the referent, then,
in virtue of a communal practice of using this name as a name for
him, and not in virtue of the speaker's conceptual
associations.
Wittgenstein, in the service of a social practice picture, sometimes
appears to urge that we drop talk of "meaning" in favor of talk of "use." I don't know that we need follow this
advice strictly, but there is surely something to be said for the idea. Talk of "meaning" tends, for one thing,
to suggest the very representationalist picture we are at pains to
supplant. It tends to conjure up
images of "grasping meanings," when there are--think about names from a Millian
perspective--no meanings to grasp.
And thinking about linguistic competence in terms of grasping meanings
encourages us to think about competence as involving theoretical knowledge,
to think in terms of knowing that as opposed to the more appropriate
knowing how.
It is natural enough, moreover, to speak of "meaning" both in connection
with communal linguistic practice and in connection with individual
speaker's conceptual associations.[21] So meaning-talk, instead of helping us
keep these topics distinct, encourages conflating them. Keeping them distinct
is, of course, absolutely crucial from the point of view taken here.[22]
The moral I want to draw is not that we need banish talk of meaning, but
that we handle it with care.
Wittgenstein himself at times appears to urge not that we drop talk of
meaning in favor of talk of use, but that we identify meaning with use. We want to be careful, though, not to
suggest that there isn't anything more to meaning, in any of its manifestations,
than communal linguistic practice.
We might settle for the more modest methodological exhortation that
insofar as we do talk of meaning, we give pride of place to the social,
specifically to communal practice, rather than to individual, or even
community-wide, representations.
Even when we turn to the Fregean's favorite questions about individual
cognition--and we ought not do this too quickly, so the methodological
sermon continues--our prior study of meaning as use will be focal. It is to court disaster to look first
toward what is available to the individual consciousness for the
clarification of virtually anything that comes under the rubric
"meaning."
Crucial to the outlook that I am recommending, then, is its rejection of
Frege's thought-driven conception of language.[23] One implication of the latter
conception, an implication noted above, is that it is the conceptual
representations rather than their linguistic embodiments, that, in the
first instance, refer. The real
action, as it were, takes place at a good distance from our social
practices. Such a conception makes
it natural to suppose that the first step towards understanding how words refer
is to understand how thoughts do so.[24] Nothing could be farther from the truth
according to our new picture. What
we semanticists study is not thought, but our social practices of talking
about things. Indeed, it becomes
tempting, although it is no doubt too simple, to construe silent thought on the
model of internal utterance.
So far we have the bare bones of the more socially sensitive,
naturalistic picture that I want to recommend. I will soon turn to implications for the
area of intersection of the philosophy of language and the philosophy of
mind to which Fregeans have riveted our attention. First, however, I want to more fully
sketch my picture. The additional
features of my view that I will mention are important to the overall
perspective, but they do not figure directly in the account of the cognitive
dimension that I will offer. I thus
want to mention and motivate them here, but a full discussion will not be
possible.
1) Clarity and distinctness.
Frege attributes to the third realm entities that he takes to stand
behind language, entities like senses and thought contents, an extreme, perhaps
even absolute, refinement; they are necessarily clear and distinct. Think here of Frege's comment, for which
he took some chiding from Wittgenstein,[25] that a concept
without boundaries is no concept at all.
Frege, of course, is not alone here. There is a deep tendency in the
representationalist tradition to attribute such purity to the representations, a
tendency that has consequences for one's conception of language. The ways in which the conception of
language may be affected are various.
Philosophers have sometimes seen language, or at least its more
"respectable" parts, as itself possessing a kind of clarity and distinctness--as
being associated, for example, with clear criteria that specify necessary and
sufficient conditions of application. Alternatively, language may be seen
as a mere dim reflection of the pristine realm, and the messiness of language
attributed to our all-too-human ways of getting at the clear and distinct
concepts.[26] Either way, the background
conception of perfectly refined representations plays a dominating
role.
Rejecting the representationalist picture may open the door to a
different way of thinking about linguistic practice. It is here that we anti-Fregeans have
most to learn from Wittgenstein. A
truly social, anti-representationalist picture will not merely reject the
ontology of objective or mental concepts.
It is not even enough to embrace the idea that meaningfulness is to be
cashed out ultimately in terms of linguistic practice. One might go this far, but remain in the
grip of a picture of practice and its governing rules (or RULES) that
derives from the rejected Cartesian perspective. One might well presuppose, and see
semantics as the attempt to uncover--as indeed we anti-Fregeans have--a kind of
rigidity on the part of the phenomena, a fixedness that the phenomena may
in fact fail to exhibit.
A detailed look at actual practice is out of the question here, but it
will stand us in good stead to remind ourselves that our practices with words do
not evolve from the attempt to capture in words, or communicate to others,
pre-existent concepts that, almost by definition, possess the required
refinement.
In the beginning--if the reader will indulge me one more time--there were
the primitive brutes, bumping into each other and grunting, by way of
indicating to their fellows salient features of the environment. Linguistic practice, on this
impressionistic picture, gets more articulated, more refined, as suits the
practical, social, and, eventually, intellectual needs of the brutes and
their successors. The crucial point
is that whatever precision, articulation, refinement does evolve is a
result of pressures of the sorts mentioned, and not a result of the desire to
capture in words some absolutely precise Cartesian concepts. The precision achieved is thus never
"absolute," whatever that might mean, nor is there any absolute standard--like
the Cartesian concepts--by which we might assess the precision of
usage.
To the extent one thinks that such a sketch roughly represents the
way things really go, to that extent one will be skeptical of philosophers'
tendency to impute clarity and distinctness to usage, or to
usage-when-it-is-up-to-snuff. And
to that extent, one will be sympathetic even to Wittgenstein's more radical
"lack-of-regimentation" claims.
Consider, for example, the putative "family resemblance" phenomenon. Reflection upon actual linguistic
practice with many general terms, 'game' for example, fails to reveal--urges
Wittgenstein--what the traditional picture led us to expect: a set of
features common to all games, features that are individually necessary and
jointly sufficient for something to be a game. Given our new picture, Wittgenstein's
claims don't seem outlandish. Why
indeed can't there be terms the utility of which don't depend upon there being
something substantive shared by the things to which the term applies? Might there well not be utility to
having a term that collects a range of things that are roughly similar to one
another, things that fail to share any single feature, but that share a kind of
family resemblance?[27],[28]
Consider a related "lack of refinement" phenomenon, vagueness, the
fact that, for example, there are no sharp boundaries separating games from
non-games. The Cartesian can try to
account for vagueness in any number of ways,[29] but, to be sure
this phenomenon does not fit like hand-in-glove with the Cartesian picture,
and something will have to be said.
Far from being a source of pressure to our more naturalistic, social
account, vagueness is entirely to be expected. Remember, we don't start with
pre-existent, sharply demarcated concepts, the common treasure of
personkind.[30] Our practices with words are only as
refined as they are, and there is no external standard (like how well they
capture our conceptual representations) by which to measure them. If there are cases that in practice
rarely occur, for example, borderline cases of the applicability of a term, then
it is likely that even a master of the term will not be able to say that the
term either clearly does, or clearly does not, apply. This inability reflects
not a lack of mastery, but rather reflects the fact that, as Wittgenstein
says, the borders have not yet been drawn.[31]
An important component of the anti-representationalist picture is thus
its expectation that usage will turn out to be considerably less regimented than
we have supposed. This is not to
say that we can expect Wittgenstein's family resemblance picture to provide a
generally applicable model. The
primitive brutes and their successors, after all, can become attuned to
relatively stable aspects of their natural environments--for example, the water
that they drink and bathe in--and come to have terms that apply in an at least
somewhat, and perhaps dramatically, less open-ended way. Even pieces of language that seem more
deeply institutional than do natural kind terms like 'water' may presumably vary
widely in the degree of regimentation they exhibit. Perhaps we should expect a continuum
here. The point, though, is that we cannot simply proceed as we have, taking
Kaplan's:
The reference of the first-person pronoun is the agent of the
context
as the model of what the
semanticist should aim to uncover.[32]
2) Propositional
content.
Propositions, it is often noted, have been wanted as the bearers of
truth, and as the "objects of propositional attitudes." Frege's representationalist picture
makes it natural to let the sentence-level representations, the Fregean
thoughts, play these roles. What,
though, if we reject representationalism in favor of the kind of
perspective I am advocating? What
becomes of propositions?
The topic is, however, gigantic.
I can only sketch what I take to be the natural response, that a truly
social and naturalistic conception will want to do without
propositions. I will briefly
mention two sorts of considerations that seem to me to militate against
propositions.
a) Naturalism.
Kaplan has recently objected to the Platonistic conception of the
individuation of words. We
suppose, he argues, that the distinction between word tokens and abstract types
is the basis of the individuation of words. Not so! What makes your utterance of
'Aristotle' and mine two occurrences of the same word is not the supposed fact
that both occurrences stand in a relation to some abstract stereotype that they
instantiate, or are members of, an abstract type that inhabits Frege's third
realm. The worldly occurrences need
to be linked in some more naturalistic way; their being occurrences of the
same word must be a function of our social, linguistic interactions.[33]
I want to reject, in this same spirit, the notion that thought, belief,
and assertion, for example, consist in relations to abstract entities. The anti-Fregeans' "singular
propositions" may appear to be more innocuous than Fregean thoughts, for
singular propositions are abstract entities that have been, so to speak, brought
down to earth. Among their constituents, after all, are things like you
and me. Even such propositions, however, seem to me objectionable from a
naturalistic point of view. Had we
anti-Fregeans not started with Fregean thoughts and then made the appropriate
amendments--replacing Frege's propositional constituents, senses, with
referents--would it have seemed natural to construe thought, belief, and
assertion as involving relations between minds (or even persons) and abstract
entities? The very idea that
thought, et al, involves such relations may well be a legacy of the
Fregean/Cartesian picture.
b) The messiness of
linguistic practice.
How does the propositions picture--explicated in terms either of Fregean
thoughts or singular propositions--comport with that other implication of
our social orientation, the fact that usage often fails to be sharply bounded,
that usage is only as clear as it needs be? Assume, for a moment, that Wittgenstein
was correct, and that 'game' is a family resemblance term. What proposition is expressed by 'That
is a game'? What,
specifically, are we to make of the predicate constituent? The idea that the predicate
constituent is the property of being a game surely doesn't comport well with our
"family resemblance" account.
My question is whether the apparatus of propositions is compatible in
spirit with the social orientation.
There is no question that it can be made compatible. One might insist that the proposition
does contain the property of being a game (or the sense of 'game'), it's just a
fuzzy property (or sense), or something of the like. Surely we can hang onto the notion of a
proposition, if we want to.
The question is whether the most natural way to work out a socially
sensitive conception of language involves the notion of a
proposition.
These remarks raise many questions, of course, questions about how,
without propositions, we are to do the work traditionally assigned to
them. Are we to try to do without
truth bearers? If not, what will
play that role? How are we to
construe belief, if not as involving a relation to a proposition? I cannot, of course, begin to deal with
these questions here, but I will do so elsewhere.[34]
4. INTENTIONALITY: THE MISSING COGNITIVE
FIX
I turn now to the problems that have seemed to make life so difficult for
anti-Fregeans. Let's begin with
what may appear to provide the anti-Fregean's most severe headache, the
Cartesian intuition concerning intentionality, the problem of the missing
cognitive fix. The Cartesian idea
that reference requires mental discrimination of a referent initially looks
unassailable. Indeed, I have
heard it said that this alone disqualifies the anti-Fregean's semantical
views. The Cartesian intentionality
intuition takes on a very different look, however, when reconsidered from
the vantage point of our more social, naturalistic
picture.
If we look at our actual practices, as Wittgenstein urged, rather
than think about what they must be like, the "cognitive fix" requirement
immediately begins to look suspicious, perhaps just plain wrong. People often simply don't have much of a
fix on the things to which they refer.
For one thing, as has been pointed out time and again, the beliefs that a
speaker has about the referent of a name that he is using may be very far off
the mark, and yet his reference is not affected.[35] A philosophy student may mistakenly
associate with 'Aristotle' the properties of having taught Plato and having
died of hemlock poisoning, and yet his utterances of the name, say on an exam,
count as references to Aristotle, our Aristotle.
There are many cases, moreover, in which a competent speaker doesn't
begin to possess the sort of information about his referent that would single
out that referent, that would distinguish the latter from many other
things. Felipe Alou, I know, was a
major-league baseball player. I
don't know much else about him, surely not enough to individuate him in any
serious way from many others, and yet I can use his name to say things about
him. Similarly, as Kripke points
out, all that many people know about Cicero is that he was a Roman
orator.[36] Consider also Kripke's Gell-Mann and
Feynman example in which a speaker competent with two names associates precisely
the same (meager) information with each.
Such a speaker refers to Gell-Mann when he uses the latter's name, and to
Feynman when he utters 'Feynman', despite the fact that the only distinctive
thing he knows about either of them is that each is a leading theoretical
physicist. Putnam makes the
same point with respect to the natural kind terms 'elm' and 'beech'.
Reference in the absence of an accurate cognitive fix looks miraculous, I
submit, only to the Cartesian, or the residual Cartesian strain in us. If pieces of language refer only in
virtue of their being associated with representations, then it is miraculous
that someone should refer to something in the absence of an appropriate
representation. This sense of
miraculousness fades fast, however, when we bring the conception of a public
language into sharp focus, a conception suggested by the sorts of examples
mentioned above. What connects the
student's utterance to Aristotle is not the student's cognitive fix on the
Aristotle. What connects utterance
to referent is rather the fact that the student is using a linguistic device
that, as our social practices go, refers to Aristotle. Linguistic expressions, as parts of
a public practice, attain a kind of life of their own. One who uses a proper name participates
in an institutionalized practice, and refers to the name's
conventional referent.[37] Indeed, so far from their being an
epistemological requirement of the sort supposed by the traditional picture, it
rather seems that one of the crucial functions of proper names is to allow us to
bridge great cognitive gulfs. The
public language thus makes it possible for us to speak about things even when
our beliefs about them are very scanty, confused, even badly mistaken. The examples that seem to show that
reference does not require anything like a discriminating conception should not,
then, seem astounding, or indeed at all surprising, at least not on our
social picture.[38]
One with Cartesian intuitions might admit that underscoring the social
and institutional character of natural language is the best way for the
anti-Fregean to proceed:
Emphasizing that proper names are part of a public, institutional
practice might well seem to make sense of the idea that a speaker need not have
much of a cognitive fix on the things about which he speaks. The effect of the anti-Fregean proposal,
though, is to make speech and thought radically discontinuous--for such a
cognitive fix is surely indispensible to thought--and such a radical
discontinuity is just not acceptable.
If, in the Aristotle-Socrates case above, for example, it is clear
that the descriptions the speaker would offer really do take us to
Socrates, then he surely was thinking about Socrates. If, moreover, a speaker uses a name in
the absence of a cognitive fix, then he really can't be thinking about
anyone.
So much in philosophy depends upon which phenomena one takes to be
fundamental, where one starts. The
anti-Fregean focus has admittedly not been on thought, but on language, public
language. The Cartesian, believing
as he does that this is the wrong place to begin, might well conclude that
notwithstanding the apparent counterexamples drawn from actual
communicative practice, linguistic reference, unless it is to be divorced
from thought, must require discriminating knowledge.
Alternatively, if one with Cartesian intuitions is impressed enough with
the social and institutional character of natural language and with the
anti-Fregeans' counterexamples, he might be willing to endorse the alleged
discontinuity between thought and speech.[39] "I'll give you speech," we might imagine
him saying, "but thought is quite another thing." The idea is that the correct
interpretation of speech is perhaps, in the final analysis, in the public
domain. It is a matter of the
conventions of the linguistic community, and really not a matter of what's
in the head of the speaker. What
one thinks, on the other hand, is very much a matter of what's in (or
available to) the head of the thinker.
An utterance of 'Aristotle wrote the Ethics', say in a philosophy
class, counts as a reference to Aristotle, but if the speaker takes Aristotle to
be the teacher of Plato, etc., his thought is directed upon
Socrates. What about one whose
beliefs about Aristotle are so indefinite that they fail to discriminate
anyone? Perhaps such a
speaker fails to be thinking about anyone.[40]
Emerging here again is the deep difference between a broadly Cartesian
orientation, and the more social, naturalistic view. It is no accident, no mere oversight,
that anti-Fregeans have tended to begin with the study of public practices of
communication, practices that we tend to see as, so to speak, relatively
out in the open. Thought seems to
us, or at least to me, a much more difficult and elusive topic, one with which
we might do better had we some grip on our public communicative practices.
It thus makes good sense that the Cartesian would worry that we
anti-Fregeans have distanced speech from thought in an extreme way. My worry, on the other hand, is that the
Cartesian, on the basis of his philosophical picture of thought, either denies
what seems plain about speech (for example, that reference does not require
a cognitive fix), or else grants the latter, but insists on distancing thought
from speech. I don't believe that
there is any such radical discontinuity between thought and speech, or that the
approach taken here suggests such a discontinuity. What seems dubious to me is what the
Cartesian takes to be so obvious, the idea that if we restrict our attention to
the question of about whom the speakers are thinking, it is clear that
the background descriptions provide the decisive answer.
Let's begin with examples in which speakers fail to have any sort of real
cognitive fix, examples in which the background descriptions fail to
individuate. Someone says,
"Cicero was a Roman orator," and can't identify Cicero much further. It seems very far-fetched, indeed
altogether ad hoc, to suppose that such a perfectly competent and sincere
speaker who fails to believe very much about Cicero, ipso facto fails to be
thinking about the latter. Surely, in the example above, I was thinking about
Felipe Alou. The phenomenon of reference in the absence of individuating
information is so pervasive that the supposition in question would deny
thought content to an extremely wide range of sincere utterances produced by
reflective people. This surely
seems like the proverbial philosopher's view, as opposed to what seems
plain to just about everyone.[41] Thinking about something, it would seem,
no more requires a cognitive fix than does speaking about
something.
Let's turn to the case of the speaker who has "mistaken beliefs," whose
background descriptions fit not the referent of the uttered name, but rather
someone else. Consider the
Aristotle-Socrates case above.
There are, no doubt, examples of this sort with respect to which it will
be perfectly natural to say that the speaker was thinking about the denotation
of the background descriptions.
The question is whether this is necessarily so, whether the simple fact
that the background descriptions take us to Socrates itself establishes that the
speaker was thinking about Socrates.
It seems pretty clear, at least if we take our cue from our ordinary
judgments, that the background descriptions do not play any such decisive
role. We would ordinarily say,
after all, in many such cases, that the speaker expressed a mistaken
belief, even a mistaken thought, about Aristotle.[42]
Let's turn from our ordinary judgments, and get philosophical. The Fregean, in making the background
descriptions decisive, gives a kind of cognitive priority to descriptions
over names. This is natural enough,
given the Fregean outlook, including the cognitive fix requirement, modes of
presentation that are supposed to capture the cognitive perspective, and so
on. The question is, however,
whether we ought to accept this inegalitarian treatment of names
vis-a-vis descriptions.
The situation, after all, is that the speaker actually uttered a name
that is, as our practices go, Aristotle's name. It's true that in answer to a question
about whom he was speaking he would have given a description that fit Socrates,
but how does that make it the case that he was thinking of Socrates? Why not say instead that he was thinking
about the person named by the name that he used, that is Aristotle, and the
background descriptions merely reveal his (in this case false) beliefs
about that person. In any case, the
uttered name takes us to one person, the to-be-uttered-if-asked descriptions to
another. What, other than the
disputed Cartesian picture, makes it so obvious that we should favor the
latter?
My point here, as already indicated, is not that we ought always to give
priority to the uttered name over the background descriptions, or that, more
generally, whatever singular term is uttered should furnish the key to the
referent-in-thought. Sometimes, we
ought indeed to privilege unuttered, background terms. One might, for example, utter the name
'Jones' as a mere slip of the tongue, while thinking about, and intending to say
something about, Smith.
Alternatively, one might mistakenly take 'Brown' to be the name of
Harris, and, thinking about and intending to say something about Harris, one
might use the name 'Brown'. So a
name other than the uttered name may deserve priority, and so may a background
description. Someone might, for
example, be thinking about "the most ugly, and nasty, professional wrestler,
whoever that is." Thinking that the
person in question is none other than Brutus Beefcake, the speaker may make a
remark mentioning the latter by name.
If B.B. does not really fit the description, however, there may well be
circumstances in which it would be appropriate to say that the speaker was
really thinking of "the most ugly, and nasty, professional wrestler, whoever
that is," or even that he was thinking about the Iron Sheik, if the latter is
indeed uglier and nastier than any other.
We should not, then, expect a simple formula in answer to the question of
how, in general, we are to determine which item it was about whom someone was
thinking.[43] My point here was not to provide such an
answer, one that favors, say, uttered names over background descriptions. It was merely to dispel the illusion
that at least cognitively--if not semantically--names must be backed up by
individuating concepts, and that the question of about whom someone is thinking
when he utters a name is to be resolved by reference to that background
concept. Whatever we do, in the
end, with the thorny problem of reference in thought, it is far from clear that
the Cartesian spirited Fregean idea, the contention that the
descriptions-to-be-uttered-if-asked must be decisive, has much merit.[44]
I will conclude this section with some remarks on the implications of my
view for another topic that deserves book-length treatment, that of silent
thought. It is sometimes supposed
that the Cartesian intentionality intuition has a kind of obvious
plausibility when it comes to silent thought episodes; that somehow silent
thought, as opposed to overt speech, is a most natural candidate for the
Cartesian picture. Consider again
someone who, although ignorant of Cicero's accomplishments (say other than being
a Roman orator), uses the name pretty much as we all do. We are ordinarily willing to ascribe
thoughts about Cicero to such a person when he says things like "I know who
Cicero was. He was a Roman
orator." Isn't it equally clear
that he is in a position to have silent thoughts about Cicero? Surely it's not the verbalization that
makes thought about Cicero possible.
Being a participant in a name using practice, specifically, being
competent with 'Cicero', our speaker is in a position to use the name not only
in overt speech episodes, but also in silent thought. Do names, then, actually occur in silent
thought episodes? Are there at
least some such episodes that amount to internal utterance? My point does not depend on positive
answers to these questions. Whether
or not his silent thought that Cicero was an orator amounts to his silently
rehearsing this sentence--indeed however we understand silent thought
episodes--his competence with the name, his participation in the practice, puts
him in a position to have the thought that he would express, were he to put it
into words, as "Cicero was an orator." Not only does the reference
of an utterance not depend upon one's cognitive fix, the reference of one's
silent thought similarly does not depend upon "what's in the head." This will, of course, seem preposterous
from a Cartesian point of view.
5. HOW PUZZLING IS FREGE'S
PUZZLE?
Frege's discussion of cognitive significance, at the beginning of "On
Sense and Reference," is focused upon his famous puzzle about informative
identity sentences. How, Frege
wants to know, are we to explain the difference in "cognitive value"
between the trivial 'Hesperus is Hesperus' and the informative 'Hesperus is
Phosphorus', that fact that these sentences formulate different
thoughts?
Frege's focus on identity-sentences is not altogether salutary, for
contrary to its suggestion, the fundamental problem about the cognitive
dimension of language with which Frege was centrally concerned has nothing
special to do with identity.
Frege's fundamental problem was that of accounting for the fact that
a mere change of one co-referring name for another can affect a change, indeed a
very significant change, in the thought content of a sentence. Frege might well have avoided identity,
with all its attendant perplexities, and have asked for the explanation of the
difference in cognitive value between 'Hesperus is a planet' and the
corresponding "Phosphorus" sentence.[45]
Frege's discussion--let's focus on the two sentences just
mentioned--draws attention to what I'll call "Frege's data," to cognitive
phenomena of undeniable importance:
One might understand both sentences, for example, but be willing to
assert only one of them, even emphatically deny the other. Alternatively, one might find, say, the
"Hesperus" sentence old hat, but the "Phosphorus" sentence highly
informative. Finally, one might
behave quite differently depending upon which of these sentences he
accepts.
Frege's data are both uncontroversial and uncontroversially important,
but the lessons Frege would have us learn from them are far from
uncontroversial. The most notorious lesson, quite obviously
controversial, is Frege's sense-reference approach to semantics. There is, however, a more subtle, prior
message that Frege takes the cognitive phenomena to convey, a putative
implication that can be made to seem almost like a datum itself. Notice that in formulating Frege's
puzzle in the first two paragraphs of this section, I gloss "difference in
cognitive value," as "difference in thoughts expressed." Frege's contention that the two
sentences express different thoughts, distinct propositional contents--a
natural enough contention given Frege's thought-oriented approach--is far from
uncontroversial.
One might, for example, insist, in the spirit of the Russell-Kaplan
singular propositions picture, that 'Hesperus is a planet' and 'Phosphorus is a
planet' express the same proposition.
Alternatively, one might, in the spirit of the naturalistic approach I've
been advocating, try to make do without propositions, without unified
things that are, as Strawson once put it, the upshots of assertive
utterances. Frege's quasi-datum is
surely contestable. What we cannot
contest, however, and what I want to explain, are what I've dubbed "Frege's
data," the cognitive phenomena that, for Frege, made it plain that different
thought contents were indeed expressed by the respective sentences.
Let's keep in mind that our project is not to make good on the alleged
cognitive failures of the anti-Fregean semantic approach. Having made the distinction between
semantics, on the anthropological conception, and the study of
cognitive significance, the anti-Fregean ought not to be embarrassed that,
say, Mill's remarks on names (or those of Donnellan and Kripke) do not
immediately address Frege's data.
At the same time, the output of anthropological semantics ought to cohere
with a more general account of language and thought, one that must address
Frege's data. How, then, might one
approach the cognitive phenomena if one takes a broadly Millian approach to the
semantics of proper names?
It is striking that even those who have led the revolt against the
Fregean orientation have approached the explanation of the cognitive phenomena
in way more suited to the Fregean, representationalist picture. To the representationalist,
the obvious and only way to explain differences in the cognitive roles of
expressions is in terms of differences in associated representations. Kaplan, to mention a prominent example,
tries to account for the difference in cognitive roles of indexical expressions,
by resurrecting modes of presentation, not Fregean senses mind you, but
more kosher "ways in which agents represent the references of their
terms."[46]
The appeal of the representationalist picture, here as before, is a
function of the fact that, as Wittgenstein says, we are its captives. It is difficult to so much as conceive
an alternative, even in bare outlines.
If 'Hesperus is a planet' differs cognitively for an agent from the
corresponding "Phosphorus" sentence, doesn't this have to be because he is
thinking of the referent in two different ways, because he is employing two
different cognitive perspectives.
If someone finds 'Hesperus = Phosphorus' informative, doesn't this have
to be because he associates a different way of thinking (or mode of presentation
or cognitive perspective) with the left hand side of the equation than he does
with the right hand side? How else
might we begin to explain the obvious cognitive
differences?
I have argued elsewhere against the Kaplan-Perry approach to the
cognitive significance of indexicals.[47] While I stressed there the Fregean
flavor of that approach,[48] I was not
sufficiently focused upon what I now see as, especially in Kaplan's case, the
most salient similarity with Frege, Kaplan's representationalism. He, no less than Frege, explains
cognitive differences between expressions as differences in their associated
modes of presentation.[49] Here I want to urge that we abandon
representationalism even in the study of the cognitive significance
phenomena. Modes of presentation,
no matter how liberalized and attenuated, remain spiritual descendants of the
Fregean approach. Reflection on
some of the anti-Fregean discussions of proper names, moreover, should convince
at least the discussants that no such representationalist approach, no
matter how benign its representationalism, has any future.
I have in mind here Kripke's Gell-Mann-Feynman case, and Putnam's
Elm-Beech story. The original point
of these examples was to show that reference does not depend upon what's in the
head, upon the information that the speaker associates with the expressions.[50] Don't these examples indicate, no
less forcefully, that cognitive significance doesn't depend upon the
associated information? 'Gell-Mann'
and 'Feynman', after all, can play cognitively inequivalent roles for a
speaker despite the fact that the speaker's associated "conceptual files" are
identical.
We can make the same point--that cognitive significance is not a matter
of associated information--with names that, unlike 'Gell-Mann' and 'Feynman',
co-refer. Someone might acquire the
names 'Cicero' and 'Tully', associating with them precisely the same
information, say "a famous Roman."
Still the names may differ in "cognitive value." It may never strike the speaker that
only one person may be in question, and so he may react very differently to
sentences that contain one name, that to those that contain the other. The anti-Fregean explanation of the
cognitive difference between 'Cicero' and 'Tully', then, given our own examples,
better not rely upon necessary differences in associated information, that
is upon different modes of presentation.[51]
My proposal, then, is that we forget modes of presentation, and take a
fresh look at the data to which Frege drew our attention. I urged in the last section that the
Cartesian intentionality intuition--remember how unassailable it seemed at
first--takes on a very different look when considered from the vantage point of
our more social, naturalistic picture.
The same can be said for the
highly-plausible-if-you've-been-brought-up-on-Frege idea that to explain a
difference in the cognitive roles of expressions one must appeal to a difference
in cognitive perspectives.
Let's begin with the reflection that the more epistemology one builds
into linguistic competence with names, that is, the more of a cognitive fix one
requires, the more it will seem that Frege's data present not merely
interesting and important phenomena to be explained, but a prima
facie problem, a puzzle.
Why is this? The thesis that
linguistic competence with names requires mental apprehension of their referents
induces a tension between two names co-referring, and their being
cognitively inequivalent. If in
using each of the two names one must be in cognitive touch with their single
referent, how can this identity of reference have escaped notice?
Indeed, if one raises the epistemic stakes enough, it will be impossible
to have co-referring names that differ cognitively. Imagine, to take a fanciful example,
that we required that a name user be omniscient about the referent. It would then be impossible to be
competent with two co-referring names without realizing that only one referent
was in question. Co-referring names
could then not differ in cognitive significance.
Let's go to the other
extreme. Although for present
purposes we can consider this another fanciful example, anti-Fregeans have
sometimes suggested a "no-epistemology" picture of linguistic competence. One might, on this view, possess
radically mistaken beliefs about the referent of a name, or
virtually no beliefs at all, and still be in a position to use the name as a
name for its socially determined referent.
Smith, on the periphery of a conversation between mathematicians in which
the name 'Joan' is used as a name for a theorem, mistakenly takes Joan to be a
woman. Alternatively, Smith
comes away completely unsure of who or what Joan is. In either of these cases, on the
no-epistemology view, one may still be in a position to say things about the
referent, that is the theorem, by using the name. One might, for example, speculate about
Joan's properties, or ask who or what Joan is.
If one thus doesn't need to know virtually anything about the referent,
it is very easy to see how one could pick up two co-referring names, and not
know that they had a single referent.
The no-epistemology theorist, since he doesn't think that competence
requires mental apprehension of the referent, will thus not see Frege's data as
presenting a serious and difficult problem, a real puzzle. Indeed, the explanation of the fact that
'Cicero' and 'Tully' might play different cognitive roles--and not much of an
explanation is really needed--would involve simply pointing out that competence
with two names does not put one in a position to know whether or not the names
co-refer. How could it, given that
competence requires no knowledge of the referent?
Frege advanced a view intermediate between "omniscience" and
"no-epistemology," that a name user need attach to the name a purely
qualitative concept that he takes to single out a referent. This is, of course, far from
requiring omniscience--much farther, say, than Russell's direct
acquaintance requirement[52]--and it allows
for a cognitive difference between co-referring names. Like the omniscience view, however, it
demands a substantive cognitive fix--something intellectually available to the
speaker must mentally focus him on the referent. A consequence is that, compared to the
no-epistemology view, a more substantive explanation of two co-referring names
differing cognitively is required.
If one is really mentally focused on the same thing twice, why doesn't he
know it? Frege's view yields a
natural answer, of course. One
might not know that a single thing is in question because his focus is not
direct, so to speak, as it was for Russell, but is mediated by a concept. If he is focused upon the same thing
twice, but by means of different concepts, he well may not realize that the same
thing is in question.
We are now in a position to see how the perspective I've been outlining
in this paper yields a distinctively non-Fregean approach to the cognitive
phenomena. While I have not
plumbed the depths of how little need be in the head in order to use a name--I
have not subscribed to the no-epistemology view--the use of a name, on my
account, emphatically does not involve a substantial cognitive fix, the mental
apprehension of a referent. Indeed,
one of the functions of names is to allow a speaker (or thinker) to bridge great
cognitive gaps, to allow one to speak about things in the absence of anything
like individuating conceptions.
A consequence is that, just as on the no-epistemology view, the
cognitive phenomena no longer have the air of paradox. The use of a name doesn't require mental
apprehension of its referent, and so there is no puzzle, no special
problem, about how a speaker might be competent with co-referring names and yet
not know that they co-refer. Given
how little one needs to know (or even believe) about the referent to be
competent with a name, there is no presumption that a speaker will know of
two co-referring names in his vocabulary, that they
co-refer.
I have been emphasizing the fact that the more social outlook dispels any
sense of paradox concerning Frege's data.
What seems even more important is that we now have a new form of
explanation of Frege's data, one radically different from that suggested by the
traditional representationalist account.
The powerful grip of the latter led us to suppose that the only way to
explain a cognitive difference was in terms of a difference in mode of
presentation. And
anti-Fregeans, laboring under this supposition, have been raking the leaves, as
it were, to somehow retrieve such representational differences. Notice that cognitive perspectives,
modes of presentation, have no role in our new form of explanation. We don't look into the speaker's head
and find two different conceptual files in terms of which we can now see why
'Cicero' and 'Tully' play different cognitive roles for him. We rather reflect upon the fact that
given how little need be in his head, his mere competence with the names puts
him in no position to decide the question of whether or not these names
co-refer. Indeed, as noted above,
were we to look inside his head, we might find identical conceptual files for
each of two cognitively distinct names. We might, of course, find different
conceptual files, say "being a famous Roman orator" associated with 'Cicero',
and "being a famous Roman politician" with 'Tully'. Such a difference, however, doesn't
affect our explanation of what makes it possible for him to wonder whether
Cicero was Tully. Such wonder is
not rendered more intelligible by the difference in conceptual files. A competent speaker, given how little he
need know or even believe about the references of names, might well raise the
question even if both files had included merely "is a famous Roman."[53]
I began this essay with the distinction between two conceptions of
semantics, Fregean and anthropological. The attention of the anthropological
semanticist, I argued, is not focused upon the cognitive dimension that,
under Frege's influence, has seized center stage. If it turns out that the
anthropological semanticist's work fails to provide much help in the
explication of the cognitive dimension, so be it. Much to our surprise, however, the
anti-Fregean semantical approach, at least when embedded in the sort of outlook
I've been recommending, yields a most natural approach to the cognitive
dimension. Central to this approach
is the outright rejection, argued for in Section 4 above, of the Cartesian
intentionality intuition. To put
the point in a more positive way, linguistic contact with things--reference,
that is--does not presuppose epistemic contact with them. Underscoring this deep lesson of the
anti-Fregean revolution leads, as we have just seen, to a radically non-Fregean
account of the Fregean's favorite topic, the cognitive significance of
language.[54]
ENDNOTES
