Linguistics, Psychology, and the Ontology
of Language
Noam
Chomsky’s well-known claim that linguistics is a “branch of
cognitive psychology” (Chomsky 1972, 1) has generated a great deal of
dissent—not from linguists or psychologists, but from philosophers. Jerrold Katz, Scott Soames, Michael
Devitt, and Kim Sterelny have presented a number of arguments, all intended to
show that the Chomskian subfield hypothesis is incorrect—there is a
significant distinction between the disciplines of linguistics and psychology.
On both sides of this debate, two
distinct issues are often conflated: (1) the ontological status of language and
(2) the evidential relation between psychology and linguistics. The ontological issue is, I will argue,
not the relevant issue in the debate.
The aforementioned philosophers have provided several reasons to think
that the objects in the domain of linguistics—what linguistics
studies—are not within the domain of psychology. If these arguments are sound, Chomsky
is incorrect to think that the only object of study in linguistics is the
I-language, which is “some element of the mind of the person who knows
the language” (Chomsky 1986, 22).
Even if this Chomskian position on the
ontology of language is false, linguistics may still be a subfield of
psychology if the relevant evidence in linguistic theory construction is
psychological. Two options are
open to the philosopher who denies Chomskian conceptualism: linguistic
nominalism or linguistic Platonism[1]. The former position holds that
syntactic, semantic, and phonological properties are primarily properties, not
of mental representations, but rather of public language sentence tokens; The
latter position holds that the linguistic properties are properties of public
language sentence types. I will
argue that both of these positions are compatible with Chomsky’s claim
that linguistics is a branch of psychology, and the arguments that have been
given for nominalism and Platonism do not establish that linguistics and
psychology are distinct disciplines.
In
“Some Notes on What Linguistics is About,” Jerry Fodor takes a
similar approach to this debate.
The question of whether to count a particular theory as part of
linguistics rather than psychology or mathematics is, as Fodor notes, not
really of interest. “The
question what it is for a true theory to be linguistic is a boring question” (Fodor 1981, 198). The question at issue in this debate,
the interesting
question, is “what is it for a linguistic theory to be true.” The answer to this latter question is a
matter of what evidence confirms or disconfirms linguistic theories[2].
The
source of the ontological/evidential conflation can be found the writings of
Noam Chomsky. Chomsky makes a
distinction between the I-language, the part of the brain of the competent
speaker responsible for knowledge of language, and the E-language, an
“‘externalized language’...understood independently of the
properties of the mind/brain” (Chomsky 1986, 20). It is far from clear what the
E-language is. The E-language is not “essentially Platonic,” as
Michael Devitt has claimed (Devitt unpublished, 18). This could not correct, given that Chomsky’s
paradigmatic example of a theory based on the notion of an E-language is
Bloomfieldian linguistic nominalism[3].
The
term E-language is intended to cover a variety of different notions of
language. “We can define
‘E-language’ in one way or another or not at all, since the concept
appears to play no role in the theory of language” (Chomsky 1986,
26). This
characterization—or lack thereof—of “E-language” makes
it quite hard to figure out what the target of Chomsky’s critique is supposed
to be. Perhaps the best way of
approaching E-language is to define E-language as any linguistic object is not
part of the “I-language”—any notion of language outside of
what is in the head. This may be
the only way of understanding certain claims made by Chomsky; For example, he
has written that “there is nothing in the world selected by such terms as
‘Chinese’ and ‘German’” (Chomsky 2000, 155). On this understanding, both the
linguistic nominalist and the Platonist are committed to a view on which the
E-language does play a significant role.
Chomsky
writes as if the relevant issue in this debate is an ontological issue: the
question is whether or not the E-language actually exists. The reason Chomsky denies the reality
of the E-language is that such a notion is purportedly unnecessary in a
well-developed linguistic theory.
The objects in a theory are taken “to be real insofar as they
enter into explanatory theories that provide insight and understanding”
(Chomsky 1991, 5). This approach
to the question of the reality of the E-language reflects Chomsky’s
commitment to a Quinean account of ontology. There is no need to quantify over E-languages in presenting
a theory of linguistics—the only relevant objects in the theory are the
I-languages. An exhaustive account
can be given of the semantic, syntactic, and phonological properties of all
languages qua properties of the I-language. The linguistic properties of the E-language are derivative
from the properties of the I-language.
In this sense, E-languages do not exist in the domain of the theory,
because a discussion of E-languages would be redundant, not providing any
“insight and understanding.”
Philosophers
have criticized the claim that E-languages are theoretically superfluous,
pointing out the possibility of divergence between the E-language and the
I-language. It is important to
note that the Chomskian claim is an empirical conjecture, to be proven or
falsified by forthcoming psychological evidence. It is also a very strong claim in support of the thesis that
linguistics is a subfield of psychology. I will label this claim the strong
subfield thesis.
It should not be assumed, as some
philosophers have, that it is necessary to claim that the E-language is
theoretically superfluous in order to claim that linguistics is a subfield of
psychology. Making this assumption
would conflate the ontological issue and the evidential issue. Even if the E-language and I-language
do diverge, there is still reason to claim that there is no distinction between
these two fields if the relevant evidence in characterization of the E-language
is psychological. I will label
this claim the weak subfield thesis.
In other
scientific fields, there are distinct domains that nonetheless are considered
part of a single science. Consider
the relationship between psychology and one of its other subfields, social
psychology. The domain of
psychology in general includes the brain and the nervous system of humans and
other animals, usually described at a fairly high level of abstraction. The domain of social psychology
includes objects outside of this domain—factors regarding specific
political systems and historical events, for example. Although the domains of psychology and social psychology
differ in this respect, the latter is nevertheless a subfield of the
former. Social psychology does not
fulfill the requirements of a strong subfield thesis; However, it is very
likely that the weak subfield thesis is true in this case. Evidence for psychological theories
will certainly bear on theories in social psychology—consider the bearing
of psychological evidence demonstrating that homosexuality has a neural basis
on approaches to gender relations in social psychology. The burden of proof for a philosopher
attempting to show that linguistics and psychology are distinct fields is to
show that both the strong and weak subfield theses are false.
Jerrold
Katz clearly states a position on which there is no distinction between the
ontological and evidential issues—not only are linguistics and psychology
concerned with different domains, psychological evidence does not bear on
linguistics at all. According to
Katz, “the nature of the objects which constitute the subject-matter of a
science determines the nature of a science” (Katz 1996, 282). The natural science of biology has in
its domain only concreta such as humans, frogs, cells, and the golgi
apparatus. The formal science of
geometry has in its domain abstracta such as squares, lines, and points. These differences in domain mark the
distinction between formal sciences such as geometry and natural sciences such
as biology. Katz claims that
formal sciences, unlike natural sciences, are nonempirical.
According
to Katz, the domain of linguistics consists of abstracta, not concreta. Linguists are concerned only with
sentence and word types, not tokens.
In several books and articles, Katz has presented an overwhelming number
of arguments for this claim, too many to be considered in a paper of this size. The most recent argument for this
claim, which Katz feels is the clearest and most convincing[4],
relies on a well-known fact about natural languages. If the effects of performance limitations are set aside, the
rules of natural languages allow for the construction of infinitely many
sentences. There are not infinitely
many linguistic tokens, so linguistics must be concerned with types. Types are abstract objects, so
linguistics is analogous to geometry, not biology—it is a formal science.
Katz contends further that, pace Chomsky, linguists cannot be concerned
with psychological entities, given that mental states are physical objects, not
abstract types. “Given the
finiteness and discontinuity of matter, of brain matter in particular, there
can’t be an infinity of mental/neural objects” (Katz 1996, 278). A difficulty quickly arises for this
claim—the same line of reasoning could be used to argue that psychology
is formal science. Thought is a
productive capacity: idealizing away from performance factors, there are
infinitely many possible thoughts[5]. A psychological nominalist could not
account for this infinite number of thoughts, given that there are too few
brain states to account for them, so psychologists must become Platonists. Psychology is a formal science, not a
natural science, and there is indeed no difficulty for a Chomskian who wishes
to claim that linguistics is a subfield of psychology. Surely this line of reasoning regarding
psychology leads to a reductio of Katz’s initial argument for Platonist
linguistics.
The
mistake in Katz’s reasoning is a failure to notice a shift in modal
context. Claims regarding the
infinite sentence of natural languages and the productivity of thought abstract
away from the limitations of human minds in the actual world. In discussing the possibility of infinitely
many sentences and thoughts, it is useful to see these appeals as descriptions
of other possible worlds. Such
descriptions need not commit us to an infinite ontology for language or thought
within the actual world.
Katz argues against this line of
reasoning, contending that unactualized possibles have to be construed as
abstract objects. Katz cites
Quine: “Certainly it is hopeless nonsense to talk thus of unrealized
particulars and try to assemble them into classes” (Quine 1960, 34, cited
in Katz 1996, 279). This
acceptance of Quinean skepticism about modality is presented without argument
by Katz; Katz overlooks a good deal of more recent work. In particular, utilizing the approach
to necessity developed by Saul Kripke, there is a framework for understanding
how talk of infinitely possible sentences and thoughts could be true without
carrying any ontological commitment whatsoever. Katz’s argument, based on an illegitimate appeal to
types, does not succeed in distinguishing psychology from linguistics.
Even if Katz’s argument had
succeeded in establishing that linguists are concerned with types, not tokens,
it is far from clear that this would establish that linguistics is a
nonempirical science. It is not
clear that having abstracta in the domain of a science is enough to make a
science formal and nonempirical.
Physics may have to appeal to abstracta in order to explain certain
phenomena, but surely physics is nonetheless a natural, empirical science. Katz also does not have a clear
response to Quine’s claim that “the abstract entities which are the
subject of mathematics” are posits on the same footing as those in other
sciences, subject to revision if necessary (Quine 1953, 45). If this Quinean claim is correct, even
those sciences whose domains consist entirely of abstracta are empirical.
As Jerry Fodor points out, there is in
fact no incompatibility between linguistic Platonism and the dependence of
linguistics upon psychological evidence.
“Indeed, a reasonable Platonist might want to endorse
[representation of grammars],” in order to have some explanation of
linguistic intuitions (Fodor 1981, 205); A representational theory will reveal
which Platonist grammar is the grammar represented by human language users. The real interest, Fodor contends, is
not in the assimilation of linguistics to the formal sciences, but seeing which
linguistic theory is true of the speakers of a language.
Scott
Soames contents, as Katz does, that psychology and linguistics are concerned
with distinct domains of objects, but unlike Katz, Soames does not base the
distinction between the domains on the difference between abstracta and
concreta. Soames is also a
linguistic Platonist, but his arguments do not rely on this position on the
ontology of language[6].
Soames argues that the fields of
linguistics and psychology are conceptually distinct. Conceptual distinctness is defined as follows: “the
are concerned with different domains, make different claims, and are established
by different means” (Soames 1984, 155). This definition is a clear example of conflation of the
ontological and the evidential issues; It is an ontological claim that the
domains of linguistics and psychology are distinct; It is an evidential claim
that evidence in favor of psychological theories does not bear (directly) on
linguistic theories.
Soames attempts to establish the
conceptual distinctness of these fields by showing that the linguist and the
psychologist are concerned with two different sets of questions. The questions that concern the linguist
are the “leading questions.”
These leading questions concern linguistic properties that individuate
languages:
the linguistically significant
properties... grammaticality, ambiguity, synonymy, entailment, analyticity,
contradiction, and so on. These
properties and relations are characteristics which define languages and serve
to identify and distinguish them (ibid., 159).
Such
properties are, according to Soames, the concern of linguistics proper, not
psychology.
As
noted above, Chomsky would contend that these properties are primarily
properties of the I-language. If
this claim is correct, there is no reason to think that a purely psychological
answer to the leading questions cannot be given. In a response to this
Chomskian hypothesis, Soames claims that there are two different sets of facts
involved—for example, facts regarding grammaticality in linguistics, and
facts regarding judgments of grammaticality in psycholinguistics. An exhaustive psychological account can
be given of the latter but not the former. In distinguishing these sets of facts, Soames is just
arguing from his own position. The
important question is whether a representation of syntactic properties is
involved in speakers’ judgments of grammaticality. If this is so, then a wholly
psychological theory can provide a complete characterization of both
grammaticality and judgments of grammaticality. Soames responds by claiming that this representational
thesis is unlikely—“no theorist would assume in advance that things
would turn out this way” (ibid, 169).
In
support of this rejection of the representational thesis, Soames offers an
analogy between linguistics and mathematics. There are a number of different axiomatizations of mathematics,
all producing the same elementary number theory. There is no need for a mathematician to suppose any of these
axiomatizations must be represented in the minds of the arithmetically
competent. Such a representational
thesis would be implausible given the complexity of the axiomatizations. In the case of linguistics, there are a
number of different grammars, each of which provides identical accounts of the
“linguistically significant properties.” There is no need for the linguist to concern herself with
the question of which of these grammars, if any at all, is represented in the
minds of the linguistically competent. Simpler, more general grammars involve
the need for a great deal of memory storage and
computation—“economies in grammatical principles will be pursued
regardless of computational costs” (ibid., 170). The complexity of such
grammars is, for Soames, the source of the implausibility of the
representational thesis.
The
complexity argument provides reason to think that the strong subfield thesis is
false—the domain of linguistics diverges from that of psychology, because
linguistic properties cannot be explained entirely in terms of the properties
of the I-language. Is there good
reason to accept the complexity argument?
Would psychologists accept, as a general principle, the claim that
limitations in memory storage and computational capacity make Chomskian
computational theories unlikely?
It is very common in one well-developed subfield of psychology, the
study of vision, to posit highly complex mechanisms in order to explain certain
facts. The research program that
began with the work of David Marr has been highly successful explaining
phenomena such as the derivation of structure from objects in motion. Explanations of these phenomena involve
the claim that visual input modules carry out fairly complex computations on
visual representations. If one
could not claim in general that fairly complex computation is involved in
psychological processing, such a research program must be rejected. Soames’ complexity argument is
not at all convincing.
As noted
above, anyone seeking to distinguish linguistics from psychology must deny both
the strong and weak subfield thesis.
Soames’ unsuccessful arguments against the representational thesis
do not provide reasons to deny the weak subfield thesis. Soames has a separate argument against
the evidential dependence of linguistics upon psychology. Soames contends that evidence that does
not provide answers to the leading questions only plays an indirect role as
evidence for linguistic theories.
At some points, Soames writes as if psychological evidence plays no role
whatsoever: “There is a theoretically sound... conception of
linguistics...considered in abstraction from the cognitive mechanisms”
(ibid., 157). At other points,
Soames makes the evidential relation clearer: “the relevance of such
psycholinguistic data to theories in linguistics is limited to this indirect
[evidential] role” (ibid., 160).
Does psycholinguistic data play such a limited
role? In his paper “Is Linguistics a Branch of Psychology,” Stephen
Laurence notes several examples of psycholinguistic data that does play a
direct evidential role. In
particular, experiments by Fodor, Bever, and Garrett demonstrated how a certain
aspect of the grammar, phrasal boundaries, are to be assigned to
sentences. Errors resulting from
the placement of click sounds revealed the location of constituent breaks:
“There was a significantly greater error for location of clicks not
objectively placed at the major boundary than for those which objectively
occurred at the boundary” (Fodor, Bever, and Garrett 1974, 252). As Laurence points out, these
experiments have a clear explanation in the Chomskian theory: “Such data
can be made sense of on the assumption that the internalized grammar is one
that assigns phrasal boundaries in a way consistent with data from these
experiments” (Laurence unpublished, 5).
The click experiment presents a dilemma
for Soames. Soames must either
accept that psycholinguistic evidence does bear directly on linguistic theory,
or deny that phrasal boundaries are truly an aspect of the grammar. The latter position is simply
false—Soames would have to hold that all of the various grammars that
produce the same results regarding “linguistically significant
properties” such as grammaticality are equivalent. A difference in phrasal boundaries is
certainly a difference in the grammar—any grammar that located the
boundaries at different locations is a different grammar. If Fodor, Bever, and Garrett had
discovered that errors resulted from placement of clicks at locations that were
not previously thought to be major structure boundaries, a revision of the
linguistic theory would be required.
Soames also has not provided a clear reason to reject the weak subfield
thesis—psychological evidence does bear directly on linguistic
theories.
Michael
Devitt and Kim Sterelny offer a nominalist position—the linguist is concerned
with the semantic, syntactic, and phonological properties of tokens,
“datable, placeable, parts of the physical world” in the language,
not Platonic abstract types (ibid, 515).
Devitt and Sterelny contend that the prevailing Chomskian views on the relationship
between linguistics and psychology rest on a conflation. “The transformational linguists
conflate two distinct theoretical tasks: one concerned with linguistic symbols
and the other concerned with linguistic competence” (Devitt and Sterelny
1989, 499). The linguistic symbols
are the output of the competence, the competence itself is a psychological
process that leads to those outputs.
On the interpretation of the term
“E-language” offered above, Devitt and Sterelny are claiming that
there is a theoretical interest in study of the E-language. Not only is this a proper object of
study, pace Chomsky;
The study of the properties of public language tokens is prior to the study of
competence in a language. We
cannot construct a psycholinguistic theory without first constructing the
separate theory of linguistics proper.
In order to understand the competence involved in language use, we must
first understand what the syntactic, semantic, and phonological properties of
the outputs are.
The contention that the study of
properties of the output of the competence is prior to the study of the
competence is a position that shares a problem with Soames’s. Namely, if one must characterize the
outputs of a competence first, psycholinguistic data concerning the competence
cannot bear directly on the characterization of the outputs. This is simply not true—as noted
above, the Fodor, Bever, and Garrett click experiments use psychological
evidence regarding competence in order to establish claims about the nature of
the output of the competence. In
this case, the study of the competence is prior to the study of the outputs of
the competence.
The claim that outputs are prior to competence is an evidential claim, and one that is false. To deny that Devitt and Sterenly have the correct view of the evidential relation is not to deny that their ontological position could be correct. One could still hold that phrasal boundaries, among other linguistic properties, are properties of tokens of the language, although the evidence that reveals the nature of properties such as phrasal boundaries is psychological. Even if linguistic nominalism is true, as Devitt and Sterelny contend, the weak subfield thesis can also be true, and linguistics ought still to be considered a subfield of psychology.
References Cited
Chomsky, Noam. Language
and Mind. Enlarged edition. New York: Harcourt Brace
Jovanovich,
1972.
___. Knowledge
of Language. Westport, Conn.: Praeger, 1986.
___.
“Linguistics and Adjacent Fields: A Personal View.” In Kasher, Asa,
ed. The
Chomskian
Turn. Oxford: Basil Blackwell,
1991.
___. New
Horizons in the Study of Language and Mind. Cambridge, U.K.: Cambridge
University
Press, 2000.
Devitt, Michael.
Ignorance of Language. Manuscript.
Devitt, Michael
and Kim Sterelny.
“Linguistics: What’s Wrong with ‘the Right
View.”
Philosophical
Perspectives, 3, 1989: pp. 497-531.
Fodor, Jerry.
“Introduction: Some Notes on What Linguistics is About.” In Block,
Ned,
ed.
Readings in the Philosophy of Psychology, Volume 2. Cambridge, Mass.:
MIT
Press, 1981.
Fodor, Jerry,
T.G. Bever, and M.F. Garrett. The Psychology of Language. New York:
McGraw-Hill,
1974.
Katz, Jerrold. Language
and Other Abstract Objects. Totowa, New Jersey: Rowman and
Littlefield,
1981.
___. “The
Unfinished Chomskian Revolution.” Mind and Language, vol. 11, No
3,
September
1996: pp. 270-294.
Katz, Jerrold,
and Paul Postal. “Realism vs. Conceptualism in Linguistics.” Linguistics
and
Philosophy, 14, 1991: 515-554.
Laurence, Stephen.
“Is Linguistics a Branch of Psychology?” Manuscript.
Quine, W.V. From
a Logical Point of View. Cambridge, Mass.: Harvard University
Press,
1953.
___. Word and
Object. Cambridge, Mass.: MIT Press, 1960.
Soames, Scott.
“Linguistics and Psychology.” Linguistics and Philosophy, 7,
1984: 155-
179.
___. “The
Necessity Argument.” Linguistics and Philosophy, 14, 1991: 575-80.
[1] These characterizations of the positions
on the ontology of language are due to Katz (Katz 1981). There are two difficulties with using
these terms: (1) They suggest that linguistics is analogous to mathematics,
which begs the question in favor of Katz’s position, on which linguistics
is a formal science, and (2) There is only superficial similarity between
Chomsky’s position on psychological reality of language and the picture
of mathematical objects produced by intuition one finds in the works of
mathematical conceptualists such as Brouwer.
[2] In particular, Fodor argues against the
“Wrong View,” according to which the only relevant evidence in
linguistic theory construction is the intuitions of speakers. Further evidence that Fodor has the
distinction between the evidential issue and the ontological issue in mind is
his argument that the “Right View” is consistent with Katz’s
linguistic Platonism.
[3] Another indication that Chomsky does not
think E-languages are essentially Platonic is his coining of a separate term
intended to cover Platonic languages, “P-languages” (Chomsky 1986,
33).
[4] “...My earlier arguments did not
make clear the striking fact that the inadequacy which Chomsky exploited to
overthrow Bloomfieldian structuralism is also an inadequacy of Chomsky’s
position” (Katz 1996, 272).
[5] The word ‘thought’ is
ambiguous—it is used to refer to the psychological states, and to the
propositional objects of those psychological states. I am using the word in the former sense.
[6] In a reply to Katz and Postal (1991), Soames claims that his Platonism is consistent with considering linguistics an empirical science: “I don’t think the refutation of Chomsky’s ‘Conceptualism’ depends on [linguistic Platonism]; nor do I think characterizing languages in that way requires one to view linguistics as non-empirical” (Soames 1991, 580).