Beruflich Dokumente
Kultur Dokumente
NOAM CHOMSKY
Before entering into the question of language that they were speaking would become a
and problems of knowledge, it may be useful typical human language, though one that we
to clarify some terminological and conceptual say does not now exist. Ordinary usage
issues concerning the concepts "language" and breaks down at this point, not surprisingly: its
"knowledge" which, I think, have tended to concepts are not designed for inquiry into the
obscure understanding and to engender point- nature of language.
less controversy. Or consider the question of what are called
To begin with, what do we mean by "errors." Many, perhaps most speakers of
"language"? There is an intuitive common- what we call "English" believe that the word
sense concept that serves well enough for "livid," which they have learned form the
ordinary life, but it is a familiar observation phrase "livid with rage," means "red" or
that every serious approach to the study of "flushed." The dictionary tells us that it means
language departs from it quite sharply. It is "pale." In ordinary usage, we say that the
doubtful that the common-sense concept is speakers are wrong about the meaning of this
even coherent, nor would it matter for word of their language, and we would say this
ordinary purposes if it were not. It is, in the even if 95%, or perhaps 100% of them made
first place, an obscure sociopolitical concept, this "error." On the other hand, if dictionaries
having to do with colors on maps and the and other normative documents were de-
like, and a concept with equally obscure stroyed with all memory of them, "livid"
normative and teleological elements, a fact would then mean "flushed" in the new lan-
that becomes clear when we ask what lan- guage. Whatever all this might mean, it
guage a child of five, or a foreigner learning plainly has nothing to do with an eventual
English, is speaking-surely not my lan- science of language, but involves other no-
guage, nor any other language, in ordinary tions having to do with authority, class struc-
usage. Rather we say that the child and ture, and the like. Unless the concept of
foreigner are "on their way" to learning "community norms" or "conventions" is clari-
English, and the child will "get there," fied in some manner yet to be addressed-if
though the foreigner probably will not, ex- this is possible at all in a coherent way-one
cept partially. But if all adults were to die should be cautious about accepting arguments
from some sudden disease, and children of concerning meaning that make free use of
five or under were to survive, whatever it is such ideas, taking them to be clear enough;
This is a slightly revised version of a paper delivered at a conference in Madrid, April 28, 1986.
THE NATURE OF LANGUAGE
they are not. We understand this easily crucial and essential properties of neural sys-
enough in connection with pronunciation; tems, on a par with discussion in nineteenth-
thus to say that the pronunciation of one century chemistry of valence, benzene rings,
dialect is "right" while that of another is elements, and the like, abstract entities of
"wrong" makes as much sense as saying that it some sort that one hoped would be related,
is "right" to talk Spanish and "wrong" to talk ultimately, to the then-unknown physical en-
English. Such judgments, whatever their sta- tities. To say that the world includes elements
tus, plainly have nothing to do with the study with valence of two which therefore behave in a
of language and mind, or human biology; or certain way, or benzene rings, etc., is to say
more accurately, they have to do with some that whatever the elementary constituents of
vastly broader inquiry into the interaction of the world may be, their properties are such that
cognitive systems, some complex that is well they are correctly described in these terms at
beyond our current grasp and that we are this level of abstraction. To say that the world
unlikely ever to comprehend unless the ele- includes such abstract entities as neural nets (it
ments that enter into it are identified and is the abstract structure that we take to be
understood. The question of "error of interpre- roughly invariant through time or among
tation" or "misuse" has much the same status. individuals, not the molecules, specific orienta-
Note that a person can be mistaken about tions, etc.) or mental representations is to say
his or her own language. Thus if "livid" in fact something similar about the brain. Mentalistic
means "flushed" in my current language, and I inquiry, so understood, is justified insofar as it
tell you that it means "pale" in my language, yields insight and theoretical understanding of
then I am wrong, just as I would be wrong if I phenomena that concern us, and from another
told you, perhaps in honest error, that in my point of view, insofar as it facilitates inquiry
language "whom" is always used for a direct into brain mechanisms. Just as nineteenth-
object, not "who," or if I were to deny some century chemistry provided a guide to subse-
feature of the urban dialect that I speak quent investigations of more "fundamental"
natively. Judgments about oneself are as physical entities, so one can expect the same to
fallible as any others, but that is not what is at be true of the brain sciences, which have little
issue here. idea what to seek without some awareness of
All of this is, or should be, commonplace. the properties of the yet-to-be-discovered
Correspondingly, every serious approach to mechanisms. Mentalism, in short, is just nor-
the study of language departs from the mal scientific practice, and an essential step
common-sense usage, replacing it by some towards integrating the study of the phenom-
technical concept. The choices have generally ena that concern us into the more "fundamen-
been questionable ones. The general practice tal" natural sciences. I might add that it is
has been to define ''language" as what I have generally pointless to demand too much clarity
called elsewhere "E-language," where "Emis in these matters. As the history of physics and
intended to suggest "extensional" and "exter- even mathematics shows, clarity about founda-
nalized." The definition is "extensional" in tional issues (e.g., in mathematics, the notions
that it takes language to be a set of objects of of limit or even proof) develops as a result of
some kind, and it is "externalized" in the inquiry and is not a necessary preliminary to it;
sense that language, so defined, is external to foundational questions and questions of con-
the mindlbrain. Thus a set, however chosen, is ceptual clarity are often premature, and can
plainly external to the mindlbrain. often be approached and settled only as re-
As a side comment, let me say that I will use search progresses without too much concern
mentalistic terminology freely, but without any about exactly what one is talking about.
dubious metaphysical burden; as I will use the A typical formulation of a notion of E-
terms, talk about mind is simply talk about the language is the definition of "language" by
brain at some level of abstraction that we the distinguished American linguist Leonard
believe to be appropriate for understanding Bloomfield as "the totality of utterances that
THE PHILOSOPHY OF LANGUAGE
can be made in a speech community," the it has commonly been argued, most notably by
latter another abstract entity, assumed to be W. V. Quine, that choice of grammar is a
homogene~us.~ Another approach, based ulti- matter of convenience, not truth, like the
mately on Aristotle's conception of language choice of "a grammar" for the well-formed
as a relation of sound and meaning, is to sentences of arithmetic in some notation. But
define "language" as a set of pairs (s, m), now we face real questions about the subject
where s is a sentence or utterance, and m is a matter of the study of language. Clearly, there
meaning, perhaps represented as some kind is some fact about the mindlbrain that differen-
of set-theoretical object in a system of possi- tiates speakers of English from speakers of
ble worlds, a proposal developed by David Japanese, and there is a truth about this
Lewis among others. There are other similar matter, which is ultimately a question of
proposals. biology. But sets are not in the mindlbrain, and
Under any of these proposals, a grammar grammars can be chosen freely as long as they
win be a formal system of some kind that enumerate the E-language, so the study of E-
enumerates or "generates" the set chosen to language, however constructed, does not seem
be "the language," clearly an infinite set for to bear on the truth about speakers of English
which we seek a finite representation. and Japanese; it is not, even in principle, part
The concept "E-language" and its variants of the natural sciences, and one might argue
raise numerous questions. In the first place, that it is a pointless pursuit, a kind of chasing
the set is ill defined, not simply in the sense after shadows. Many philosophers-W. V.
that it may be vague, with indeterminate Quine, David Lewis, and others-have con-
boundaries, but in a deeper sense. Consider cluded that linguists must be in error when they
what are sometimes called "semi-grammatical hold that they are concerned with truths about
sentences," such as "the child seems sleep- the mindlbrain, though clearly there are such
ing." Is this in the language or outside it? truths about language for someone to be
Either answer is unacceptable. The sentence concerned with; they also hold that puzzling
clearly has a definite meaning. An English philosophical problems are raised by the claim
speaker interprets it in a definite way, quite that grammars are "internally represented" in
differently from the interpretation that would some manner. Others (Jerrold Katz, Scott
be given by a speaker of Japanese. Hence it Soames, and others) have held that linguistics
cannot simply be excluded from the set "E- is concerned with some Platonic object that we
English," though it is plainly not well formed. may call "P-language," and that P-English is
But speakers of English and Japanese will also what it is independently of what may be true
differ in how they interpret some sentence of about the psychological states or brains of
Hindi-or for that matter how they will speakers of English. One can see how these
inttrpret a wide variety of noises-so then all conclusions might be reached by someone who
languages and a vast range of other sounds begins by construing language to be a variety of
also fall within English, a conclusion that E-language.
makes no sense. It is doubtful that there is any There is little point arguing about how to
coherent solution to this range of problems. define the term "linguistics," but it is plain
The fact is that a speaker of English, Japa- that there is an area of investigation, let us call
nese, or whatever, has developed a system of it "C-linguistics" (cognitive linguistics) which
knowledge that assigns a certain status to a is concerned with the truth about the mind/
vast range of physical events, and no concept brains of the people who speak C-English and
of E-language, nor any construct developed C-Japanese, suitably idealized. This subject
from it, is likely to be able to do justice to this belongs strictly within the natural sciences in
essential fact. principle, and its links to the main body of the
A second problem has to do with choice of natural sciences will become more explicit as
grammar. Evidently, for any set there are the neural mechanisms responsible for the
many grammars that will enumerate it. Hence structures and principles discovered in the
THE NATURE OF LANGUAGE
The source of all of these problems resides language (the "grammar," in one of the two
in an inappropriate choice of the basic concept senses in which this systematically ambiguous
of the study of language, namely "language." phrase has been used), that poses philosophi-
The only relevant notion that has a real status cal problems, which are probably not worth
is what is usually called "grammar." Here trying to solve, since the concept is of no
again we find an unfortunate terminological interest and has no status. It may, indeed, be
decision, which has undoubtedly been mislead- pure "folly" to construct and discuss it, to ask
ing. Guided by the misleading and inappropri- what formal properties E-languages have, and
ate analogy to formal languages, I and others so on. I suspect it is. In particular, the analogy
have used the term "language" to refer to to formal systems of arithmetic and so on is
some kind of E-language, and have used the largely worthless, and should be discarded,
term "grammar" with systematic ambiguity- though other analogies to arithmetic and
a fact that has always been spelled out clearly, logic, as systems of mentally represented
but has nevertheless caused confusion: the knowledge, are quite definitely worth pursu-
term "grammar" has been used to refer to the ing, and raise quite interesting questions, yet
linguist's theory, or to the subject matter of to be seriously explored. The debates of the
that theory. A better usage would be to past generation about these matters seem to
restrict the term "grammar" to the theory of me a classic example of the philosophical
the language, and to understand the language errors that arise from misinterpreting concepts
as what we may call "I-language," where "I" is of ordinary language-in this case, developing
to suggest "intensional" and "internalized." a useless, perhaps quite senseless concept, and
The I-language is what the grammar purports assuming erroneously that it is the relevant
to describe: a system represented in the mind/ scientific notion that corresponds to, or should
brain, ultimately in physical mechanisms that replace, some concept of ordinary language-
are now largely unknown, and is in this sense a source of philosophical error that was clearly
internalized; a system that is intensional in that exposed in the eighteenth-century critique of
it may be regarded as a specific function the theory of ideas, if not earlier, and has
considered in intension-that is, a specific more recently been brought to general atten-
characterization of a function--which assigns tion by Wittgenstein.
a status to a vast range of physical events, Let us now use the term "language" to refer
including the utterance "John seems to be to I-language, and the term "grammar" to
sleeping," the utterance "John seems sleep- refer to the theory of an I-language. What
ing," a sentence of Hindi, and probably the about the term "universal grammar," recently
squeaking of a door, if we could do careful resurrected and given a sense that is similar to
enough experiments to show how speakers of the traditional one, but not identical, since the
English and Japanese might differ in the way entire framework of thinking has been radi-
they "hear" this noise. cally modified? The term "universal gram-
As contrasted with E-language, however mar" has also been used with systematic
construed, I-languages are real entities, as real ambiguity, to refer to the linguist's theory and
as chemical compounds. They are in the mind, to its subject matter. In keeping with our
ultimately the brain, in the same sense as effort to select terms so as to avoid pointless
chemical elements, organic molecules, neural confusion, let us use the term "universal
nets, and other entities that we construct and grammar" to refer to the linguist's theory
discuss at some appropriate abstract level of only. The topic of universal grammar is, then,
discussion are in the brain. They are what they the system of principles that specify-what it is
are, and it is a problem of science to discover to be a human language. This system of
the true account of what they are, the gram- principles is a component of the mindlbrain
mar for the speaker in question. The story prior to the acquisition of a particular lan-
presented by many philosophers is entirely guage. It is plausible to suppose that this
backwards. It is the E-language, not the I- system constitutes the initial state of the
THE NATURE OF LANGUAGE
language faculty, considered to be a subsys- Let us now consider the question of knowl-
tem of the mindlbrain. edge. The language a person has acquired
This initial state, call it So, is apparently a underlies a vast range of knowledge, both
common human possession to a very close "knowledge-how" and "knowledge-that." A
approximation, and also appears to be unique person whose mind incorporates the language
to humans, hence a true species property. It is English (meaning, a particular I-language that
what it is, and theories concerning it are true falls within what is informally called "English")
or false. Our goal is to discover the true theory knows how to speak and understand a variety
of universal grammar, which will deal with the of sentences, knows that certain sounds have
factors that make it possible to acquire a certain meanings, and so on. These are typical
particular I-language and that determine the cases of knowing-how and knowing-what, ordi-
class of human I-languages and their proper- nary propositional knowledge in the latter
ties. Looked at from a certain point of view, case, and this of course does not exhaust the
universal grammar describes a "language ac- range of such knowledge. It seems entirely
quisition device," a system that maps data into reasonable to think of the language as a system
language (I-language). A theory of universal that is internalized in the mindlbrain, yielding
grammar, like a particular proposed grammar, specific cases of propositional knowledge or
is true or false in whatever sense any scientific knowledge how to do so and so. We now have
theory can be true or false. For our purposes, to consider at least three aspects of knowledge:
we may accept the normal reaalist assump- (1) the internalized system of knowledge of the
tions of the practicing scientist, in this connec- language, (2) knowing how to speak and
tion. Whatever problems may arise are not understand, and (3) knowledge that sentences
specific to this enterprise, and are surely far mean what they do (etc.).
better studied in connection with the more It is common among philosophers, particu-
developed natural sciences. larly those influenced by Wittgenstein, to hold
Crucially, (I-)languages and So are real that "knowledge of language is an ability,"
entities, the basic objects of study for the which can be exercised by speaking, under-
science of language, though it may be possible standing, reading, talking to oneself: "to know
to study more complex abstractions, such as a language just is to have the ability to do these
speech or language communities; any further and similar thing^,"^ and indeed more gener-
such inquiry will surely have to presuppose ally knowledge is a kind of ability. Some go
grammars of (I-)language and universal gram- further and nold that an ability is expressible in
mar, and always has in practice, at least tacitly, dispositional terms, so that language becomes,
even when this is explicitly denied, another as Quine described it, "a complex of present
confusion that I will not pursue here. An I- dispositions to verbal behavior." If we accept
language-henceforth, simply "a languagen- this further view, then two people who are
is the state attained by the language faculty disposed to say different things under given
under certain external conditions. I doubt very circumstances speak different languages, even
much that it makes any sense to speak of a if they are identical twins with exactly the same
person as learning a language. Rather, a history, who speak the same language by any
language grows in the mindlbrain. Acquiring sensible criteria we might establish. There are
language is less something that a child does so many well-known problems with this concep-
than something that happens to the child, like tion that I will simply drop it, and consider the
growing arms rather than wings, or undergoing vaguer proposal that knowledge of language is
puberty at a certain stage of maturation. These a practical ability to speak and understand
processes take place in different ways depend- (Michael Dummet, Anthony Kenny, and oth-
ing on external events, but the basic lines of ers, in one or another form).
development are internally determined. The This radical departure from ordinary usage
evidence seems to me overwhelming that this is is, in my view, entirely unwarranted. To see
true of language growth. how radical is the departure from ordinary
THE P H I L O S O P H Y O F LANGUAGE
usage, consider the consequences of accepting incapable of exercising it, say a swimmer who
it, now using "ability" in the sense of ordinary cannot swim because his legs and arms are
usage. In the first place, ability can improve tied. But that is surely an entirely different
with n o change in knowledge. Thus suppose kind of case than the one we are now
Jones takes a course in public speaking o r in considering, where the ability is lost but the
composition, improving his ability to speak knowledge is retained.
and understand, but learning nothing new T o sustain the thesis that knowledge is
about his language. The language that Jones ability, we would have to invent some new
speaks and understands is exactly what it was concept of ability, call it "K-ability," which we
before, and his knowledge of language has not understand in the sense of knowledge. Then
changed, but his abilities have improved. we could say that Jones, who improved his
Hence knowledge of language is not to be ability to speak with n o change in his knowl-
equated with the ability to speak, understand, edge of English, retained his K-ability to
etc. speak (etc.) without change; and Smith fully
Similarly, ability to use langauge can be retained his K-ability while entirely losing his
impaired, and can even disappear, with no loss ability to use English, in the normal sense of
of knowledge of language at all. Suppose that "ability." Plainly this is pointless. The in-
Smith, a speaker of English, suffers Parkin- vented concept K-ability is invested with all
son's disease, losing entirely the ability to the properties of knowledge, and diverges
speak, understand, etc. Smith then does not radically from the quite useful ordinary con-
have "the ability to d o these and similar cept of ability. It is true that knowledge is K-
things," and therefore does not have knowl- ability, since we have defined the novel
edge of English, as the term is defined by invented term "K-ability" to have the proper-
Kenny, Durnmett, and others. Suppose that ties of knowledge, but that is hardly an
use of the chemical L-Dopa can restore Smith's interesting conclusion.
ability completely, as has been claimed (it does Exactly this tack is taken by Anthony
not matter whether the facts just noted are Kenny, in the face of conceptual arguments
accurate; since we are dealing with a concep- such as those just reviewed. Thus in the case
tual question, it is enough that they could be, as of the patient with Parkinson's disease, Kenny
is certainly the case). Now what has happened says that he did indeed have the ability to use
during the recovery of the ability? O n the the language when he had no ability to use the
assumption in question, Smith has recovered language, thus shifting to "K-ability ," plainly,
knowledge of English from scratch with a drug, since the ability was totally lost.5 Crucially, K-
after having totally lost that knowledge. Curi- ability diverges radically from ability, and is
ously, Smith recovered knowledge of English, like knowledge, as we can see from the fact
not of Japanese, though n o evidence was that a person may have entirely lost the ability
available to choose between these outcomes; to speak and understand while entirely retain-
he regained knowledge of his original English ing the K-ability, can improve the ability with
with no experience at all. Had Smith been a the K-ability unmodified, etc.
speaker of Japanese, he would have recovered Kenny also assumes that there is a contradic-
Japanese with the same drug. Evidently, some- tion between my conclusion concerning the
thing remained fully intact while the ability was person who has lost the ability while retaining
totally lost. In normal usage, as in our technical the knowledge and my statement elsewhere
counterpart to it, we would say that what (which he accepts) that there might in princi-
remained fully intact was "possession of the ple be a "Spanish pill" that would confer
language," knowledge of English, showing knowledge of Spanish on a person who took
again that knowledge cannot be reduced to it. There is no inconsistency. The issue in
ability . 4 connection with aphasia or Parkinson's dis-
Note that there are cases where we would ease has nothing to d o with a pill for acquiring
say that a person retains an ability but is a certain language; rather, the point is that the
THE NATURE OF LANGUAGE
there is surely no reason to suppose that every though here too the options are narrowly
trait is specifically selected. constrained by general principles. Beyond
The phenomena of the languages of the that, it seems that the principles allow for a
world appear to be highly diverse, but, increas- limited range of variation. That variation is
ingly, it has been shown that over a large and limited has often been explicitly denied. The
impressive range they can be accounted for by leading American linguist Edward Sapir held
the same principles, which yield highly varied that languages can vary "without assignable
results as the properties of lexical items vary limit," and Martin Joos put forth what he
from language to language. Thus in Spanish, called the "Boasian" view, referring to Franz
there are clitic pronouns, including the reflex- Boas, one of the founders of modern linguis-
ive, while in English there are not, so that the tics: namely, that "languages could differ from
forms of English and Spanish, say in causative each other without limit and in unpredictable
constructions, look quite different. But the ways." Such views echo William Dwight
principles that govern them appear to be Whitney, who greatly influenced Ferdinand
essentially the same, their consequences differ- de Saussure, and who emphasized "the infi-
ing by virtue of a lexical property of the nite diversity of human speech."
pronominal system: in Spanish, but not in Such views perhaps appeared tenable in
English, there is a system of pronouns that are some form if one regarded language as a
lexically marked as affixes, and therefore habit system, a network of practical abilities,
must attach to other elements. The manner in a complex of dispositions, and the like. In
which these affixes attach, and the targets to that case, language would be constrained
which they adjoin, are determined by the very only by whatever general conditions constrain
same principles that determine the formation the development of abilities and habits in
of complex syntactic constructions such as general, by what are sometimes called "gener-
operator-variable constructions and others, so alized learning mechanisms," if these exist.
it now appears. But this conception does not allow one even
In other languages, many more items are to approach the essential features of normal
identified in the lexicon as affixes, and the language use, as has been demonstrated
same syntactic principles determine complex beyond reasonable doubt in my view; and as
morphological forms that reflect in another already noted, the conception has been en-
way the same underlying and near universal un- tirely unproductive.
derlying stru~tures.~Thus in Japanese, the caus- Assuming without further discussion that
ative element is not a verb, like Spanish hacer this conception must be abandoned, the ques-
or English make, but rather an affix, so a verb tion of language variation will take on a new
must move from the embedded clause to attach form in the context of a general revision of the
to it, yielding what appears to be a monoclausal framework of inquiry into problems of natural
causative as distinct from the English-Spanish language. A conceptual change of this nature
biclausal causative; in Spanish too there is a re- was proposed about thirty years ago, reviving
flection of the same process when se raises to in a new form some long-forgotten approaches
the main verb in the sentence "Juan se hizo to the study of natural language. This rather
afeitar," as if hizo-afeitar were a single word. sharp conceptual change underlies the research
The point is that as lexical items vary, the very program that has been given the name "genera-
same principles determine a wide range of su- tive grammar," referring to the fact that the
perficially different complex phenomena in ty- grammar-or as we are now more properly
pologically quite different languages. calling it "the languagem-generates an un-
The principles of universal grammar are bounded range of specific consequences, as-
fixed as constituent elements of the language signing a status to every expression and thus
faculty, but languages plainly differ. How do providing the mechanisms for the creative
they differ? One way has already been noted: aspect of language use. The central questions
they differ in properties of lexical items, of the study of language, conceived along the
THE PHILOSOPHY OF LANGUAGE
lines of the earlier discussion, now become the already briefly discussed. What alternative,
following: then, can we propose? I will keep to the
terminology suggested above, departing from
1.
earlier usage.
(i) What is the system of knowledge
The first proposal was that a language is a
attained by a person who speaks
rule system, where the kinds of rules and their
and understands a language?
interrelations are specified by universal gram-
(ii) How is that knowledge acquired?
mar. In one familiar conception, the rules
(iii) How is that knowledge put to
included context-free rules, lexical rules, trans-
use?
formational rules, phonological rules (in a
The last question has two aspects, the produc- broad sense), and what were misleadingly
tion problem and the perception problem. called "rules of semantic interpretation" relat-
The second question, how language is ac- ing syntactic structures to representations in a
quired, is a variant of what we might call system sometimes called "LF," suggesting
"Plato's problem," raised for example when "logical form" but with certain qualifications.
Socrates demonstrated that a slave boy with This term "rules of semantic interpretation" is
no training in geometry in fact knew geome- misleading, as David Lewis among others has
try, perhaps the first psychological (thought-)- pointed out, because these rules relate syntac-
experiment. The problem is not a trivial one: tic objects, mental representations. They re-
people know a great deal more than can late syntactic structures and LF-represen-
possibly be accounted for in terms of the tations, which are syntactic objects. The term
standard paradigms of epistemology (or per- "semantics" should properly be restricted to
haps more accurately, what they know is the relation between language and the world,
different from what one might expect in these or to use of language, some might argue. The
terms), language being a striking example. criticism is accurate, but it applies far more
The production problem might be called broadly. In fact, it applies in exactly the same
"Descartes's problem," referring to one of the form to what Lewis and others call "seman-
central Cartesian criteria for the existence of tics," where "meanings" are set-theoretic
other minds: namely, when experiment dem- objects of some sort: models, "pictures,"
onstrates that another creature that resembles situations and events, or whatever. These are
us exhibits the creative aspect of language use, mental representations,1° not elements of the
then it would only be reasonable to attribute world, and the problem arises of how they are
to the creature a mind like ours. In more related to the world. It is often assumed that
recent years, a similar idea has been called the relation is trivial, something like incorpo-
"the Turing Test." This problem, one aspect ration, so that it is unnecessary to provide a
of more general problems concerning will and justification for these particular systems of
choice, remains beyond the scope of serious mental representation, but it is easy to show
human inquiry in fact, and may be so in that this cannot be true unless we trivialize
principle, rather as Descartes suggested. In our conception of what the world is by
any event, having nothing to say about it, I restricting attention to something like what
will put it aside, keeping just to the perception Nelson Goodman calls "versions," all men-
problem, or what is sometimes called "the tal representations, abandoning (perhaps as
parsing problem" (restricting attention to meaningless) the question of why one collec-
certain computational aspects). tion of "versions" is jointly acceptable or
These questions were posed as constituting "right" and others not, that is, not pursuing
the research program of generative grammar the common-sense answer: that certain ver-
about thirty years ago, along with an argu- sions are jointly "right" because of their
ment to the effect that prevailing answers to accord with reality. But if we take this tack,
them in terms of habit systems and the like which I do not suggest, semantics disappears
were completely unacceptable for reasons and we are only studying various systems of
THE NATURE OF LANGUAGE
mental representation. In fact, much of what (4) [that Y lies] CAUSE problems
is called "semantics" is really the study of the (5) Y CAUSE-lie problems
syntax of mental representations. It is a
curious fact that those who correctly call their Subject-object asymmetries of this sort are
work in this area "syntax" are said to be found very widely in language. They reflect
avoiding semantics, while others who incor- in part the fact that subject-verb-object sen-
rectly describe their studies of syntax as tences are not treated in natural language as
"semantics" are said to be contributing to two-term relations as is familiar in logical
semantics. l 1 analysis, but rather in the more traditional
Adopting this conception of language, a terms of Aristotelian logic and the universal
language is a complex of rules of the permitted grammar of the premodern period, as
format, interconnected in a way permitted by subject-predicate structures with a possibly
universal grammar. In contrast to the concep- complex predicate. In part, the asymmetries
tion of language in terms of habit systems or appear to follow from a newly discovered
abilities, this was an extremely productive principle governing empty categories of the
idea, which led quickly to a vast increase in sort illustrated earlier. But whatever the
the range of phenomena brought under investi- explanation, problems of this nature abound,
gation, with many discoveries about facts of and an approach in terms of rule systems
language, even quite simple ones, that had leaves them unsolved, except in a rather
never been noted or explored. Furthermore, superficial way. From another point of view,
the array of phenomena discovered and inves- there are simply too many possible rule
tigated were made intelligible at some level, systems, even when we constrain their form,
by providing partial rule systems that ac- and we thus do not achieve a convincing
counted for their properties. The depth of answer to our variant of Plato's problem.
explanation, however, could never really be Recognition of these facts has been at the
very great. Even if appropriate rule systems core of the research program of the past
could be constructed, and even if these sys- twenty-five years. The natural approach has
tems were found to be restricted in type, we been to abandon the rules in favor of general
would always want to know why we have these principles, so that the question of why we have
kinds of rules and not others. Thus, languages one choice of rules rather than another simply
typically have rules that allow the direct object does not arise. Thus if there are no rules for
of a verb to function as its subject, though it is the formation of passive constructions, or
still being interpreted as the object; but the interrogatives, or relative clauses, or phrase
converse property does not exist. Or consider structure, and no rules that change grammati-
again causative constructions, say, the form cal functions such as causative and others,
that we can give in abstract representation as then the question why we have certain rules,
(2), where the element CAUSE may be a not others, does not arise. Increasingly, it has
word as in Spanish-English or an affix as in become clear that rules are simply epiphen-
Japanese: omena, on a par with sentences in the sense
that they are simply "projected" from the (I-)
( 2 ) problems CAUSE [that Y lies]
language, viewed in a certain way. But as
The principles of universal grammar permit a distinct from sentences, which exist in mental
realization of this abstract form as something representation and are realized in behavior,
like (3), where CAUSE is an affix, or with there is no reason to believe that rules of the
CAUSE-lie associated in a closely linked verb familiar form exist at all, they have no status
sequence as in Spanish: in linguistic theory and do not constitute part
of mental representation or enter into mental
(3) problems CAUSE-lie Y
computations, and we may safely abandon
But the form (4) does not underlie a possible them, so it appears. We are left with general
realization as (5): principles of universal grammar.
THE PHILOSOPHY OF LANGUAGE
If there were only one possible human Koopman, Lisa Travis, and others, but we may
language, apart from lexical variety, we would put these further complexities aside.
then have a simple answer to our variant of A crucial fact about the head parameter is
Plato's problem: universal grammar permits that its value can be determined from very
only one realization apart from lexicon, and simple data. There is good reason to believe
this is the language that people come to know that this is true of all parameters; we must deal
when they acquire appropriate lexical items with the crucial and easily demonstrated fact
through experience in some manner. But that what a person knows is vastly un-
clearly the variety of languages is greater than derdetermined by available evidence, and that
this, so this cannot be the complete story- much of this knowledge is based on no direct
though it is probably closer to true than has evidence at all. Empty categories and their
been thought in the past. Thus in languages properties provide a dramatic example of this
such as English or Spanish, verbs and preposi- pervasive phenomenon, almost entirely ig-
tions precede their objects, and the same is true nored in earlier work. Thus a person is
of adjectives and nouns, as in such expressions provided with no direct evidence about the
as "proud of Mary" (where "Mary" is the position and various properties of elements
object of "proud" with a semantically empty that have no physical realization. There is
preposition of introduced automatically as a little doubt that this problem of "poverty of
kind of case-marker for reasons determined by stimulus" is in fact the norm rather than the
universal grammar) and "translation of the exception. It must be, then, that the values of
book" with a similar analysis. The categories parameters are set by the kinds of simple data
noun, verb, adjective, and preposition (more that are available to the child, and that the
generally, adposition) are the lexical catego- rich, complex, and highly articulated system
ries. The general principles of universal gram- of knowledge that arises, and is shared with
mar determine the kinds of phrases in which others of somewhat different but equally
they appear as heads. The lexical entry itself impoverished experience, is determined in its
determines the number and category of the basic features by the principles of the initial
complements of these heads and their semantic state So of the language faculty. Languages
roles, and the general principles of phrase may appear to differ, but they are cast in the
structure determine a limited range of other same mold. One might draw an analogy to the
possibilities. biology of living organisms. Apparently, the
There is, however, an option left underde- biochemistry of life is quite similar from yeasts
termined by the principles of universal gram- to humans, but small changes in timing of
mar. English and Spanish settle this option by regulatory mechanisms of the cells and the
placing the head invariably before its comple- like can yield what to us seem to be vast
ments. We may say that they choose the "head- phenomenal differences, the difference be-
initial" value of the "head parameter." In tween a whale and a butterfly, a human and a
Japanese, incontrast, verbs, adpositions, adjec- microbe, and so on. Viewed from an angel's
tives, and nouns follow their complements. point of view, with numerous other possible
The range of phrase structures in the two though not actual physical worlds under con-
languages is very similar, and accords with sideration, all life might appear identical apart
quite general principles of universal grammar, from trivialities. Similarly, from an angel's
but the languages differ in one crucial choice of point of view, all languages would appear
the head parameter: the language may choose identical, apart from trivialities, their funda-
either the "head-initial" or the "head-final" mental features determined by facts about
value of this parameter. In fact, this is only the human biology.
simplest case, and there is a very limited range The language itself (again, as always, in the
of further options depending on directionality sense of I-language) may be regarded as
of assignment of abstract case and semantic nothing more than an array of choices for the
roles, a matter that has been explored by Hilda various parameters, selected in accord with
THE NATURE OF LANGUAGE
whatever options universal grammar permits. biological nature and sufficiently similar expe-
Since there are a finite number of parameters, rience with simple utterances.
each finite-valued (probably two-valued), it Turning to Plato's problem, a language is
follows that there are a finite number of acquired by determining the values of the
possible languages. One can see at once why parameters of the initial state on the basis of
questions concerning the formal properties of simple data, and then the system of knowledge
natural languages are largely irrelevant; there is represented in the mindlbrain and is ready to
are few questions of mathematical interest to function-though it might not function if the
raise concerning finite sets. person lacks the ability to use it, perhaps
Here a qualification is necessary. We are because of some brain injury or the like. As for
separating out the lexicon (to which I will the parsing problem, it presumably should be
briefly return), a system that in principle can solved along such lines as these: the hearer
extend without bound though with sharp identifies words, and on the basis of their
constraints in many languages (thus in En- lexical properties, projects a syntactic structure
glish, we may always add another monomor- as determined by principles of universal gram-
phemic name of arbitrary length), and we are mar and the values of the parameters. Connec-
considering only what we might call "core tions and associations among these elements,
language," to be distinguished from a "periph- including the empty categories that are forced
ery" of marked and specifically learned excep- to appear, are determined by other principles
tions; irregular verbs, idioms, and the like. of universal grammar, perhaps parametrized.
These may presumably vary without bound Thus given the sentence "a quiCn se hizo Juan
apart from time and memory limitations, afeitar," the mind of the speaker of Spanish
though surely in a manner that is sharply automatically assigns a structure with two
constrained in type. It is the core language empty categories, one the subject of "afeitar,"
that is nothing other than an array of values another its object. Principles of universal
for parameters. I assume, of course, that the grammar then produce a contradiction, in the
distinction between core and periphery is a manner informally described earlier, and the
real-world distinction, not a matter of con- sentence receives no coherent interpretation,
venience or pragmatic choice, except insofar though of course it has a status; thus the
as this is true of theories in chemistry and Spanish speaker assigns to it a lexical and
other branches of natural science, a consider- syntactic structure, and might even be able to
ation irrelevant here. For obvious reasons, the "force" a certain meaning, if the sentence were
periphery is of much less interest for the basic produced by a foreigner, by me for example. A
psychological-biological questions to which monolingual speaker of English will also assign
linguistics is directed, if conceived along the a certain status to this expression, at least in
lines of the previous discussion, and I will some kind of phonetic representation, very
ignore it here. likely considerably more.
Keeping to the core, then, there are finitely The abandonment of rule systems in favor
many possible languages. What a person of a principles-and-parameters approach,
knows, when that person speaks and under- which has been gradually developing over the
stands a language, is a vocabulary and a past twenty-five years and has been achieved
particular array of values of parameters: an I- to a substantial extent only in the past half-
language. Once the parameters are set and dozen years, has been extremely productive.
lexical items acquired, the entire system func- It has, once again, led to a vast leap in
tions, assigning a status to a vast range of empirical coverage, with entirely new empiri-
expressions in a precise and explicit manner, cal materials discovered in well-studied lan-
even those that have never been heard or guages, and with languages of great typologi-
produced in the history of language (and well cal variety incorporated within essentially the
beyond, as noted earlier). Others understand same framework. The depth of explanation
what we say, because they have the same has also advanced considerably, as it has
THE PHILOSOPHY OF LANGUAGE
become possible to explain why there are can produce large-scale phenomenal differ-
processes described by certain rules but not ences, yielding different species of organisms.
others. The principles now being developed In general, a slight change in the functioning
yield very sharp and surprising predictions of a rigidly structured and intricate system can
about languages of varied types, predictions yield very complex and surprising clusters of
which sometimes prove accurate, and some- changes as its effects filter through the sys-
times fail in highly instructive ways. My guess tem. In the case of language, change of a
is that we are at the beginning of a radically single parameter may yield a cluster of differ-
new and highly productive phase in the study ences which, on the surface, appear discon-
of language. nected, as its effects filter through the invari-
The shift of perspective from rule systems ant system of universal grammar. There is
to a principles-and-parameters approach reason to believe that something of the sort is
might be regarded as a second major concep- correct. Thus, among the Romance lan-
tual change in the development of generative guages, French has a curious status. It differs
grammar, the first being the conceptual from the other Romance languages in a
change noted earlier as part of the so-called cluster of properties, and it appears that these
"cognitive revolution," from a conception of differences emerged fairly recently, and at
language as a system of habits or abilities to a about the same time. It may be that one
centalistic approach that regards language as a parameter was changed-the null subject
computational system of the mindlbrain-a parameter that permits subject to be sup-
step towards integrating the study of language pressed, some have speculated-yielding a
to the natural sciences, for the reasons dis- cluster of other modifications through the
cussed earlier. The second shift of perspective mechanical working of the principles of univer-
is more theory-internal than the first, but is in sal grammar, and giving French something of
a sense a much sharper break from the the look of a Germanic language. At the same
tradition, for two reasons. One is that the time, French and Spanish share certain fea-
"cognitive revolution" of the 1950s was in tures distinguishing them from Italian, and
many respects a rediscovery in different terms there are numerous other complexities as we
of ideas and insights that had been developed look at the actual languages, or "dialects" as
long before, both in psychology and the study they are called. Similarly, we find most
of language, during the seventeenth century remarkable similarities among languages that
"cognitive revolution." A second is that the have no known historical connection, suggest-
rule system developed in early generative ing that they have simply set crucial parame-
grammar were in certain respects a formaliza- ters the same way. These are essentially new
tion, in a different framework, of tradi- questions, which can now be seriously formu-
tional notions about the way sentences are lated for the first time and perhaps addressed.
constructed and interpreted. The shift to As conceptions of language have changed
a principles-and-parameters approach intro- over the years, so has the notion of what
duces ideas that have only a remote resem- counts as a "real result." Suppose we have
blance to those of the traditional or modern some array of phenomena in some language.
study of language, and the basic notion of the In the era of structural-descriptive linguistics,
discipline and the ways in which problems are a result consisted in a useful arrangement of
formulated and addressed take on a consider- the data. As Zellig Harris put it in the major
ably different form as well. theoretical work of structural linguistics, a
The principles-and-parameters approach grammar provides a compact one-one repre-
yields a rather new way of thinking about sentation of the phenomena in a corpus of
questions of typology and comparative-his- data. Some, for example Roman Jakobson,
torical linguistics. Consider again the analogy went further in insisting on conformity to
of speciation in biology. Apparently, small certain general laws, particularly in phonol-
changes in the way fixed mechanisms function ogy, but in very limited ways.
THE NATURE OF LANGUAGE
Under the conception of language as a rule of a word, and such meanings have great
system, this would no longer count as a intricacy and involve the most remarkable
significant result; such a description poses assumptions, even in the case of very simple
rather than solves the problem at hand. concepts, such as what counts as a possible
Rather, it would be necessary to produce a "thing." At peak periods of language acquisi-
rule system of the permitted format that tion, children are "learning" many words a
predicts the data in question and in nontriv- day, meaning that they are in effect learning
ial cases, infinitely more. This is a much words on a single exposure. This can only
harder task, but not a hopeless one; there are mean that the concepts are already available,
many possible rule systems, and, with effort, with all or much of their intricacy and
it is often possible to find one that satisfies structure predetermined, and the child's task
the permitted format, if this is not too is to assign labels to concepts, as might be
restricted. done with very simple evidence.
Under the more recent principles-and- Many have found this conclusion com-
parameters approach, the task becomes harder pletely unacceptable, even absurd; it certainly
still. A rule system is simply a description: it departs radically from traditional views.
poses rather than solves the problem, and a Some, for example Hilary Putnam, have
"real result" consists of a demonstration that argued that it is entirely implausible to sup-
the phenomena under investigation, and count- pose that we have "an innate stock of notions"
less others, can be explained by selecting including carburetor, bureaucrat, etc.12 If he
properly the values of parameters in a fixed and were correct about this, it would not be
invariant system of principles. This is a far particularly to the point, since the problem
harder problem, made still more difficult by arises in a most serious way in connection with
the great expansion of empirical materials in simple words such as "table," "person,"
widely differing languages that have come to be "chase ," "persuade," etc. But his argument
partially understood, and to which any general for the examples he mentions is not compel-
theory must be responsible. Where the prob- ling. It is that to have given us this innate stock
lem can be solved, we have results of some of notions, "evolution would have had to be
depth, well beyond anything imaginable ear- able to anticipate all the contingencies of
lier. It is an important fact that the problem is future physical and cultural environments.
now intelligibly formulable, and that solutions Obviously it didn't and couldn't do this." A
are being produced over an interesting range, very similar argument had long been accepted
while efforts to pursue this inquiry are unearth- in immunology; namely, the number of anti-
ing a large mass of new and unexplored gens is so immense, including even artificially
phenomena in a wide variety of languages that synthesized substances that had never existed
pose new challenges, previously unknown. in the world, that it was considered absurd to
This discussion has been based on the suppose that evolution had provided "an
assumption that lexical items are somehow innate stock of antibodies"; rather, formation
learned and available, suggesting that apart of antibodies must be a kind of "learning
from parameter-setting, language acquisition process" in which the antigens played an
as well as parsing and presumably the cre- "instructive role." But this assumption has
ative use of language (in the unlikely event been challenged, and is now widely assumed
that we can come to understand anything to be false. Niels Kaj Jerne won the Nobel
about this matter) are to a large extent Prize for his work challenging this idea, and
determined by properties of the lexicon. But upholding his own conception that an animal
acquisition of lexical items poses Plato's "cannot be stimulated to make specific anti-
problem in a very sharp form. As anyone bodies, unless it has already made antibodies
who has tried to construct a dictionary or to of this specificity before the antigen arrives,"
work in descriptive semantics is aware, it is a so that antibody formation is a selective
very difficult matter to describe the meaning process in which the antigen plays a selective
THE PHILOSOPHY OF LANGUAGE
and amplifying role.13 Whether or not Jerne is Smith to confess by torture, we are using the
correct, he certainly could be, and the same term ironically. Since these facts are known
could be true in the case of word meanings, the essentially without evidence, it must be that
argument being quite analogous. the child approaches language with an intu-
Furthermore, there is good reason to itive understanding of concepts involving in-
suppose that the argument is at least in tending, causation, goal of action, event, and
substantial measure correct, even for such so on, and places the words that are heard in a
words as carburetor and bureaucrat, which, in nexus that is permitted by the principles of
fact, pose the familiar problem of poverty of universal grammar, which provide the frame-
stimulus if we attend carefully to the enor- work for thought and language, and are
mous gap between what we know and the common to human languages as conceptual
evidence on the basis of which we know systems that enter into various aspects of
it. The same is true of technical terms of human life.
science and mathematics, and it is quite Notice further that we appear to have
surely the case for the terms of ordinary connections of meaning, analytic connections,
discourse. However surprising the conclusion in such cases as these; we have a rather clear
may be that nature has provided us with an distinction between truths of meaning and
innate stock of concepts, and that the child's truths of fact. Thus, if John persuaded Bill to
task is to discover their labels, the empirical go to college, then Bill at some point decided or
facts appear to leave open few other possibili- intended to go to college; otherwise, John did
ties. Other possibilities (say, in terms of not persuade Bill to do so. This is a truth of
"generalized learning mechanisms") have meaning, not of fact. The apriori framework of
not, to my knowledge, been coherently for- human thought, within which language is
mulated, and if they are some day formu- acquired, provides necessary connections
lated, it may well be that the apparent issue among concepts, reflected in connections of
will dissolve. meaning among words, and more broadly,
To the extent that anything is understood among expressions involving these wsrds. Syn-
about lexical items and their nature, it seems tactic relations provide a rich array of further
that they are based on conceptual structures of examples. It appears, then, that one of the
a very specific and closely integrated type. It central conclusions of modern philosophy is
has been argued plausibly that concepts of a rather dubious: namely, the contention, often
locational nature, including goal and source of held to have been established by work of Quine
action, object moved, place, etc., enter widely and others, that one can make no principled
into lexical structure, often in quite abstract distinction between questions of fact and ques-
ways. In addition, notions like actor, recipient tions of meaning, that it is a matter of more or
of action, event, intention, and others are less deeply held belief. Philosophers have, I
pervasive elements of lexical structure, with think, been led to this dubious conclusion,
their specific properties and permitted interre- which is held by some (e.g., Richard Rorty) to
lations. Consider, say, the words chase or have undermined centuries of thought, by
persuade. Like their Spanish equivalents, they concentrating on an artificially narrow class of
clearly involve a reference to human inten- examples, in particular, on concepts that have
tion. To chase Jones is not only to follow him, little or no relational structure: such sentences
but to follow him with the intent of staying on as "cats are animals." Here, indeed, it is not
his path, perhaps to catch him. To persuade easy to find evidence to decide whether the
Smith to do something is to cause him to sentence is true as a matter of meaning or fact,
decide or intend to do it; if he never decides or and there has been much inconclusive debate
intends to do it, we have not succeeded in about the matter. When we turn to more
persuading him. Furthermore, he must decide complex categories with an inherent relational
or intend by his own volition, not under structure such as persuade or chase, or to more
duress; if we say that the police persuaded complex syntactic constructions, there seems
THE NATURE OF LANGUAGE
little doubt that analytic connections are to be dubious at best. There is, it seems clear,
readily discerned. a rich conceptual structure determined by the
Furthermore, the status of a statement as a initial state of the language faculty (perhaps
truth of meaning or of empirical fact can and drawing from the resources of other geneti-
must be established by empirical inquiry, and cally determined faculties of mind), waiting to
considerations of many sorts may well be be awakened by experience, much in accord
relevant; for example, inquiry into language with traditional rationalistic conceptions and
acquisition and variation among languages. even, in some respects, the so-called "empiri-
The question of the existence of analytic cist" thought of James Harris, David Hume,
truths and connections, therefore, is an empiri- and others.
cal one, to be settled by empirical inquiry that I think we are forced to abandon many
goes well beyond the range of evidence commonly accepted doctrines about language
ordinarily brought to bear. Suppose that two and knowledge. There is an innate structure
people differ in their intuitive judgments as to that determines the framework within which
whether I can persuade John to go to college thought and language develop, down to quite
without his deciding or intending to do so. We precise and intricate details. Language and
are by no means at an impasse. Rather, we thought are awakened in the mind, and
can construct conflicting theories and proceed follow a largely predetermined course, much
to test them. One who holds that the connec- like other biological properties. They develop
tion between persuade and decide or intend is in a way that provides a rich structure of
conceptual will proceed to elaborate the struc- truths of meaning. Our knowledge in these
ture of the concepts, their primitive elements, areas, and I believe elsewhere-even in
and so on, and will seek to show that other science and mathematics-is not derived by
aspects of the acquisition and use of language induction, by applying reliable procedures,
can be explained in terms of the very same and so on; it is not grounded or based on
assumptions about the innate structure of the "good reasons" in any useful sense of these
language faculty, in the same language and notions. Rather, it grows in the mind, on the
others, and that the same concepts play a role basis of our biological nature, triggered by
in other aspects of thought and understanding. appropriate experience, and in a limited way
One who holds that the connection is one of shaped by experience that settles options left
deeply held belief, not connection of meaning, open by the innate structure of mind. The
has the task of developing a general theory of result is an elaborate structure of cognitive
belief fixation that will yield the right conclu- systems, systems of knowledge and belief,
sions in these and numerous other cases. One that reflects the very nature of the human
who holds that the connection is based on the mind, a biological organ like others, with its
"semantic importance" of sentences relating scope and limits. This conclusion, which
persuade and decide or intend (i.e., that these seems to me well-supported by the study of
sentences play a prominent role in inference, language and I suspect holds far more
or serve to introduce the term persuade to the broadly, perhaps universally in domains of
child's vocabulary, and thus are more impor- human thought, compels us to rethink funda-
tant than others for c o m m ~ n i c a t i o n ~faces
~) mental assumptions of modern philosophy
the task of showing that these empirical and of our general intellectual culture, includ-
claims, which appear to lack any plausibility, ing assumptions about scientific knowledge,
are in fact true. The first task seems far more mathematics, ethics, aesthetics, social theory
promising to me, but it is a matter of empirical and practice, and much else, questions too
inquiry, not pronouncements on the basis of broad and far-reaching for me to try to
virtually no evidence. The whole matter re- address here, but questions that should, I
quires extensive rethinking, and much of what think, be subjected to serious scrutiny from a
has been generally assumed for the past point of view rather different from those that
several decades about these questions appears have conventionally been assumed.
THE PHILOSOPHY OF LANGUAGE
syntax, but it deals with questions that others Columbia University Press, 1980), 136f.; and
incorrectly term "semantic," and it is, I suspect, Jerne's Nobel Prize lecture, "The Generative
one crucial way to study semantics. Grammar of the Immune System," Science
12. See Putnam, "Meaning and Our Mental Life," 229.1057-9, September 13, 1985.
manuscript, 1985. 14. The proposal of Paul M. Churchland, Scientific
13. For discussion in a linguistic-cognitive context, Realism and the Plasticity of Mind (Cambridge
see my Rules and Representations (New York: University Press, 1979; 1986, 51f.).