CLS 29
Volume 2: The Parasession
What We Think, What We Mean,
and How We Say It:
Papers from the Parasession on
the Correspondence of Conceptual,
Semantic and Grammatical Representations
1993
Compiled and edited by
Katharine Beals
Gina Cooke
David Kathman
Sotaro Kita
Karl-Erik McCullough
David Testen319
Of dynamos and doorbells: semiotic modularity and the
(over}determination of cognitive representation
Michae} Silverstein
‘The University of Chicago
It is somewhat as though a ay-
namo capable of generating e~
nough power to run an elevator
Were operated alnost exclu-
sively to feed an electric
doorbell.
--E. Sapir (1921(1949}:24)
Linguists ought especially to take account of what
Sapir pointed out in the passage of Language from which the
quotation above is taken. In it, he tries to relate our
Tinguistic analyses of grammar to the "gamut of psychic
uses" with which the realtime "flow" of linguistic tokens is
bound up. Thus, Sapir points out, simplifying somewhat,
“the typical linguistic element labels a concept," under our
usual assumptions about the abstract objects called
grammars. The generated pairings of linguistic form and
content that constitute the putative mental representations
of our grammars, Sapir says, work from "the highest latent
or potential content of speech, the content that is obtained
by interpreting each of the elements in the flow of language
as possessed of its very fullest conceptual value,"
notwithstanding the fact that "(w]e are not in ordinary life
So much concerned with concepts as such as with concrete
particularities and specific relations" (1921(1949}:14-15)-~
meaning here the contextual ization of linguistic performance
in textualized reference and modalized predication, and in
the social situations in which this takes place. And in
many ways, as Sapir points out, "the state of mind...in
which abstract concepts and their relations are alone at the
focus of attention and which is ordinarily terned
reasoning," is the imagined (and imaged) embodiment of our
grammatical working assumptions as they would emerge in
realtine performance, were that psychologically possible.
Now Sapir’s point is, of course, that language-in-use
might look, shall we say, cognitively "degenerate" from the
point of view of our granmarian’s account of the maximal
conceptual power implicit in forn-content pairings as
cognitive representations relevant to language processing.
Yet even the maximally degenerate use of language, has, in
part, its own order of factuality, being the "symbolic’...
rend{ition of situation) in the grooves of habitual
expression." Thus for Sapir’s example of someone's saying
in a characteristic social situation "I had a good breakfast
this morning," he observes "Each element in the sentence320
defines a separate concept or conceptual relation or both
combined, but the sentence as a whole has no conceptual
significance whatever" (1921(1949}:14).
At first a seeming paradox, the point becomes clear, I
think, if we understand the framework in which Sapir/s
observations were made as part of our more contemporary view
of the semiotic modularity of linguistic signals, their
potentially multiple framings as meaningful signs in
semiotie-functional terms. In such a view, achieving at
least one consequential realtine mapping that interrelates
the semiotic modalitios constitutes the psychological
problem of language-in-use. lence, we face such questions
as the degree of determinacy of cognitive representations
necessary in any one or more of the semiotic modalities
Implied by the structure and use of language so as to enable
various consequences, cognitive, interactional, etc. In
particular, we face the question of the correctness of the
almost universal assumption -- certainly anong linguists ~~
that at the heart of language production and comprehension
is a psychological mechanism for achieving a full, multi-
planar formal ropresentation of a sentence or equivalent
formal unit, and its pairing with some one, computibly
determinate cognitive representation of its “meaning,” its
“logical form," its "semantic interpretation," or whatever
euch a putatively complete corresponding object is termed.
one can be agnostic on the question of how this pairing of
two such objects, form and meaning, is precisely
describable; the’connonly held belief is that at the heart
of any "use" of Language -- performing-in-realtine -- is the
achievement of this kind of double representation, by
accessing knowledge of linguistic form-meaning pairings -- a
particular kind of competence -- though the details of the
actual process involved differ among specific accounts, to
the extent they are addressed at all.
Such a realm of representation Seens to be what Sapir
is pointing out as "the highest latent or potential content.
of speech," and I wish here to clarify first his view of
language structure in terms of the problen of cognitive
representations, before going on to see its role in a
semiotic-nodular view of language-in-use. 1 shall tern
Sapir’s view of language structure (as also Bloomfield’s
view, and indeed the view of every pre-post-Bloomfieldian
Linguist in America or in Europe one could cite) the
(axammaticosemantic) ceding view of structure, and first
develop the outlines of this framework so as to reach
certain conclusions about the psychological utility of
putatively complete forn-meaning pairings
‘The coding view of structure
Like every major tradition of grammatical analysis, the
coding view tries to establish a systematic relationship,
that is, a specific mapping, between contextually~
abstractable and systematically scheatized linguistic forme
321
and whatever the existence of these forms entails about
differential reference-and-modalized-predication. Such
entailments are themselves commonly spoken of as “conceptual
schemata,“ as note Sapir's use of the terms, (separate)
soncept, conceptual, conceptual relation (=concept with
certain nonlocalized coding correlates in linguistic form)
Notwithstanding this usage, we need not be confused: we
need not take Use of this term as an index of our belief in
naive conceptualism -- Quine's (1968:186) “museum” theories
of meaning -- that locates quote-concepts-unquote in a
mental realm unpermeable to social-level facts, in all their
messy historical and dialectical contingency. Such a
Cartesian world, or actually such an inner Platonic world,
is familiar to us in much cognitivist projection. As we
will develop below, the direct analogy taken from
philosophical discourse -- probably via Carnap and Chomsky -
= of natural language sentence-forms as uninterpreted
logical-syntactic formulae, and a complete "semantic
representation" or equivalent of those linguistic formulae
as their structure of truth-conditions relative to the
compositionality of formulae, is not an essential commitment
of the coding view.
‘The critical methodological question in the coding view
is, "How is such-and-such differentiable concept -- of
course, specified as differentiable from among others in a
particularly-structured universe of possibilities -- how is
such a concept specifically and differentially signaled in
language form?" That is, what is the formally-
characterizable expression of such-and-such conceptual
differentiation? Note that the coding relationship we are
after may be multiple in both directions, with many
different means of differential coding for a given
differential conceptual value, and vice-versa. Indeed, the
whole first part (chapters II-VI) of Sapir's little book is
aimed at showing the independence of "form" and "meaning" in
the specific sense of the multiplicity of meanings that any
formal category-type or construction-type bears both within
and across languages. And contrariwise the book shows the
multiplicity of formal expression, both within and across
languages, that any differential concept has. This is the
coding view's "take" on the matter of relative arbitrariness
in language, the limits to which in the realm of grammatical
structure it is happy to encounter as empirical fact, and
attempt to explain with grounded theoretical constructs and
methodological expectations.
‘The asyumetry of coding. As it turns out, it is
impossible to carry out the program of the coding view
stated in these terms, since it is impossible to know
anything about the nature of differential conceptual values
independent of some coding relationship(s). Hence, if we
see coding as a structure linking our ultimate goal of
methodologically *independent" variables of differential
conceptual values to methodologically “dependent” variables
of differential linguistic forms, the only way we know of322
conceptual values is through the regularity of the codings,
this regularity becoming the fundamental assumption, ontic
commitment about, and object of inquiry. We thus posit the
existence of structures of grammatical categories, the
internal organization and interrelationships of which are
the center of interest.
Two types of categorial structure. Cutting to the
finish, and incorporating the contributions to the
enterprise of Saussure and Bloomfield in particular, several
specific conclusions and projects emerge here.
First, ronal catsoaRies are categories of form, that we
can hold to the criterion of characterizability purely in
terms of mutual distributional relations. Such notions as
defining the "Subject" of a larger syntactic construction
terms limited to facts of configurationality of hierarchical
constituency of a distributional position with respect to
other, mutually distributed categories, are formal
definitions in this sense, definitions determining formal
categories purely in terms of arrangements of them in
relation to other formal categories. This is the practical
face of the thesis of the radical autonomy of form, or
radical asymmetry of the forin-meaning distinction, in which
realizing that we cannot study the putative *concepts™ or
*meanings" or "truth conditions” directly, some scholars
have proposed that in principle we must study only form
(akin to the priority of logical formulae within Carnapian
syntactic calculus, which is in principle autonomous with
respect to its "interpretive" logical semantics and even
more to its pragmatics).
It was Bloomfield who first articulated the correlative
structural-functional conclusion, that if language can be
completely and consistently described in strictly formal-
categorial terms, then the “function* of any linguistic for
is given by its distributional possibilities, however
complexly these are to be described within the whole
language. And of course to the extent that we cay so
describe the structure not only of some particular language,
but in principled terms, the possible structures of ali
languages, we have a complete "formal" account ~~ i.e., an
account of linguistic form purely in terms of linguistic
form-categories -- that might even be formalized in the
sense intended by the term "generative* grammar. We might,
however, have to fall back on a partial structural-
functionalism, in which we try to separate -- or modularize
== those aspects of linguistic structure that seem to be
amenable to purely formal-categorial characterization, fron
various other types of undeniably formal patterning, but
patterning the regularity of which seems to require other
kinds of explanatory strategies. In the current jargon,
such other possibilities have been termed * functional”
explanations, though of course structural-functional
explanations are no less functional; they merely locate the
function in the criterion of distributional structure under
certain assumptions about parsing formulae.
323
We are in fact familiar, from the current literature,
with the problens involved in trying to formalize a purely
formal-categorial account of grammar. There are trade-offs
necessary in the positing of formal categories whose
distributional properties are driven by -- or at least a
function of -- assumptions about restrictions on types of
constituency (for example, uniquely binary-branching and
headed) and on significant classes of variant constituencies
(as are generated by "movement rules," for example). Many
formal categories necessary to such an account have turned
out to be unrealized in any surface distribution (so-called
Wenpty categories," for example), and are the clearest cases
of Bloomfield’s structural-functional hypothesis, since
‘their distributions are only indirectly manifested in the
distributions of what we might term, with Bloomfield and his
successors, lexical forms. (Observe that the concept of
lexical form is a grammatical term, involving non-null
surface phrasal projections of specific formal categories;
the lexicon of a language, let us recall, is a determinate
part of ite grammar.) lence note the sense of theoretical
triumph surrounding the well-argued-for empty category as a
part of “universal formal grammar, since it implies that to
the extent its formal-categorial basis is a mental reality,
thie mental structure is at the very least actively
constructed in language acquisition, probably on specific
structural-functional principles.
‘Types of grammatical categories. It is important not
to confuse such formal categories with the GRAMMATICAL
CATEGORIES of languages, which are recognized by how
specific formal expression-types (sometimes partly or wholly
characterizable in formal-categorial terms) can be regularly
mapped onto differentially communicable values in structured
semiotic universes. Even with default structural-
functionalist methodological commitments, we look for
gramnatical categories with additional criteria. We
presuppose, for example, that there obtain, by degrees,
universality and comparability across the attested languages
of the world (anong which we always make at least implicit
typological comparisons) that ground any specific linguistic
analysis in a structured universe of possibly-distinguishea
gramnatical categories recognizable by their "signatures" in
formal categorial structure, the structure of linguistic
form. We expect that certain grammatical category types
cluster in recurrent arrangements of formal categorial
structure, which become diagnostic of them by virtue of the
veracity of such expectations. In many attemptedly formal-
categorial (and self-styled "formalist") accounts of
Linguistic structure, grammatical categories are in fact
informally and nonchalantly snuck in in the choice of labels
for particular nodes in constituency-diagrams and their
equivalent, or in various principles of putatively formal-
categorial arrangement, or in the de facto limitation of
certain principles and'rules of distribution to only certain
Classes of categorial organization. (That many contemporary324
Linguists do not realize that they are confused about formal
and grammatical categorization probably is a result of the
post-Bloomfieldian -- specifically 2. Harris (1942;1945). --
and ever-after misreading of Language (e.g., 1933:218 on so-
called secondary morpheme words).) Such smuggling of a
theory of grammatical categories into a purportedly
formalist theory of formal categories does neither aspect of
the problem of linguistic coding well
In fact, it is on the very relationship between formal
categories and grammatical categories that the coding view
attempts to theorize. (Sapir calls the latter "grammatical
concepts," by the way, emphasizing both poles of the mapping
between grammar ~- form -- and differential sense-structure
== "meaning.") There are several different kinds of
relationships that can immediately be identified, each with
its own kind of implication for a theory of cognitive
representations and their intentional vs. automatic
generation and processing by speakers/hearers of languages
(and note the distinct sense of "generate" here, speaking of
actual real time mental happenings, not structural
conditions on the autonomy of formal categorial relations)
In the easiest situation to recognize, a grammatical
category is always isolable as such in a particular lexical
form, for example an affix or clitic or other type of
syntactically-distributed single-lexical form. Observe that
formal phrasal head-types, so-called “lexical categories" in
some theoretical discourses, are also organized by ~~
generally cross-cutting -- grammatical categories insofar as
their formal distributions are a correlative function of
such class-membership. Such a grammatical category is
highly localizable with respect to a system of formal-
categorial organization based on the principles of
hierarchical concatenative constituency relations, and, if
always bearing its unique coding relationship wherever its
corresponding formal category occurs, it is robust across
all distributional possibilities with respect to other
grammatical categories, independent of them in formal
categorial coding. In’practice, there is a trivial sense of
localizable grammatical categorization, in that every
recurring lexical form in a language constitutes ite own
idiosyncratic (and very semantically complex!) grammatical
category-complex, with questionable grounding in universal
comparability, absent further assumptions
In the normal linguistic situation, where we recognize
complex constituency relationships anong formal categories,
it turns out that grammatical categories interact one with
another in characteristic ways, both in terms of which
categories characteristically interact, and what formal-
categorial expression of these interactions can be
recognized. Morphological paradigns most vividly express
one such interaction, simultaneous multivariate coding in
single lexical forms (syncretism of grammatical categories),
where sets of distinct lexical forms, e.g., affixes, are, in
effect the dependent coding values filling the cells of
325
dimensionalized natrices of granmatical categorial arrays,
which diagram the structure of semantic variables in this
case. Here, of course, systematic mutual neutralizations
occur with characteristic regularities (much of the earlier
typological literature has been concerned with elaborating
such regularities in the world’s languages), and illuminate
the relationships of grammatical categories. Observe that
syntactic-plane paradigms of grammatical categories also
For all other kinds of coding, we need to invoke the
distinctness of formal categories and granmatical categories
in a more specific way. First, we must recognize that
morphological pieces of lexical form can, from their
(structural-) functional roles, be codings of grammatical
categories that must be “attached to," in the sense of
functionally associated with the grammatical categories of,
hierarchically superordinate or syntagnatically disjoint
lexical scopes or projections, notwithstanding the formal
categorial distributional appearance. (Though also,
following Harris and others, completely mixing up the two
kinds of category-types ~~ glaringly, for example, in the
analysis of verbal auxiliation -- Chomsky’s (1957) very
originary "transformational" assault-by-denonstration on
Markov and phrase-structure grammars, which he must have
attributed to all other theorists, rested completely on
facts of this sort in English grammar.) Note that so-called
‘Noun of Agency’ and related affixes in many languages are
formally affixed to Verb Stems or Themes, though they are a
grammatical category of Clause vs. Phrase derivational
relationship. (Note how all kinds of formal categorial
relations of the clause are carried over intact, or in
determinately-derived form, in the corresponding phrase.)
In effect, everything except the affix in such an example
constitutes a formal categorial configuration, which may or
may not be a fornally-identified "constituent in any
reasonable way even if one might want to argue that the Verb
gets the phrasal Affix because it gets the inflectional
machinery of the Clause (but in many fornal-categorial
accounts is not the head of the clause, note). But the
important point is that the formal-categorial structure
internal to the configuration, and the formal categorial
labels that define its formally superordinate lexical scope
or projection according to formal-categorial analysis, are
both involved in specifying that this is a derivational
paradigm (Clause and corresponding Phrase), and hence are
all necessary to describing the coding relationship here of
a grammatical category. Conversely, it is uniquely the fact
of grammatical categoriality of, say, Noun of Agency, that.
seems to justicy this analysis, notwithstanding the fact
that what is COMMON material in both Clause-level fornal-
categorial structure and Phrase-level fornal-categorial
structure docs not form a fornal-categorial constituent of
any sort justifiable on purely formal-categorial grounds
(other than the facts of derivation that is, to be precise).