146 4. Code Model Linguistics: Patch or Abandon?
How did Chomsky turn the attention of the discipline from the langue of the group to the mental process of the individual? The move involved two major steps:
1. Restricting the definition of language and elevating grammar 2. Abridging the classic code model
These shall be addressed in turn.
4.4.1.1.1. A restricted definition of language and associated elevation of grammar
The manner in which Chomsky came to restrict the definition of language and elevate the notion of grammar is related to his use of information theory. As mentioned
previously, Noam Chomsky’s early work in syntax was in part a response to Hockett, who had attempted to incorporate certain elements of Shannon’s and Weaver’s infor-
mation theory in accounting for human language. Hockett had clearly expressed his intention to incorporate an information theoretic approach in his account of phonology
and syntax. Chomsky, however, did not as clearly address the relationship of his own early work in syntax to developments in information theory, therefore comparatively few
contemporary linguists seem aware of that influence. While the relationship of Chom- sky’s work to information theory could be established simply on the basis of the analo-
gies he builds, there is no need for speculation; Chomsky specifically addresses Shannon and Weaver in several of his early papers and builds upon their work in the development
of his ideas concerning models of language. This is especially clear in Chomsky
1956 ,
“Three Models for the Description of Language,” a paper discussing mathematical models of language, which was presented at a conference on information theory.
Interestingly, Claude Shannon presented a paper at the same conference. See Shannon
1956 .
Chomsky’s 1956
“Three Models” paper laid the foundation for his more famous 1957 “Syntactic Structures” see
1957 :18–25. In these two papers, Chomsky established
that a “communication theoretic” approach to grammar was inadequate for the description of human language
Chomsky 1957 :24, 34, 49. He was, of course, referring
to the theory more commonly known today as “information theory.”
75
This rejection and the weight of his argument were to have a broad influence. As mentioned previously,
George Miller and others later abandoned the information theoretic approach in psycho- linguistics, largely, it seems, as a result of Chomsky’s influence
Miller 1964 :n1;
Miller and Chomsky 1963
; Palermo 1978
:110–113. It is crucial to note, however, that Chomsky’s argument concerning information theo-
ry did not constitute a total rejection. His statement of rejection refers specifically to the theory of probability
Shannon and Weaver proposed, and Hockett’s applications of
75
‘Communication theory’, a label based upon Shannon’s original title, A Mathematical Theory of Communication 1948
, has occasionally been used to refer to Shannon’s theory. ‘Information theory’ is the label more commonly used, and is therefore the label used in this study.
4. Code Model Linguistics: Patch or Abandon? 147
Markov processing in particular see Weaver 1949b
:117. At this point it is important to distinguish between the mathematical applications of Shannon’s theory and the model of
communication Shannon offered, for Chomsky did not reject Shannon’s model of communication. He continued to find Shannon’s model useful for its analogic value, as is
evident in his use of information theoretic terminology in later works e.g.,
Chomsky and Halle 1968
:3, to be discussed. Nor did Chomsky reject certain conditions relevant to the engineering problem of communication as established by Shannon
1948 ,
1949 and later
modified by Weaver 1949a
, 1949b
. In general, the focus Chomsky gave to syntactic and phonological components is
reminiscent of the focus Shannon and Weaver gave to the concern Weaver called “Level A,” the “technical problem” of communication, that is, “How accurately can the symbols
of communication be transmitted?” Weaver 1949b
:96. A major element in that concern was the structuring of the signal. If the signal were consistently structured, with enough
distinction between elements and with sufficient redundancy, then that signal could be confidently received and “decoded.” That is, having read the signal, the receiver could
produce a constructed message and measure the statistical confidence that the constructed message was equivalent to the primary message.
This required, of course, that the transmitter and receiver identify a circumscribed set of possible structures, that is, the structures which the transmitter had the capability of
handling in rendering the transmission. The structures available within that set might be finite, as in the three points “0,” “.5,” and “1,” or they might be infinite, as in the infinite
number of points between “0” and “1,” but regardless, the actual message assembled had to be a possible message. That is, if messages could include any point between “0” and
“1,” then “0.23432” could be a message, but “1.1” would be excluded. It is for this reason that Shannon writes, “The significant aspect is that the actual message is one selected
from a set of possible messages.
The system must be designed to operate for each possible selection, not just the one which will actually be chosen since this is unknown at
the time of design” Shannon 1949
:31; italics added. The information theoretic concern with the mathematical structuring of the signal
obviously influenced Chomsky’s early interest in mathematical structuring of syntax in language systems, and in human language in particular. As was addressed in chapter
3 ,
Shannon’s approach to the “technical problem” of structuring the signal involved certain logical conditions. To the extent that Chomsky’s approach followed Shannon’s, it shared
those same logical conditions.
As should be evident, Chomsky’s notion of “well-formedness” is directly related to Shannon’s constraints regarding possible messages. An information theoretic perspective
regarding the set of possible messages and the message and signal alphabets was expressed in Chomsky
1957 as follows:
… I will consider a language to be a set finite or infinite of sentences, each finite in length and constructed out of a finite set of elements. All natural languages in their spoken or written
form are languages in this sense, since each natural language has a finite number of phonemes
148 4. Code Model Linguistics: Patch or Abandon?
or letters in its alphabet and each sentence is representable as a finite sequence of these phonemes or letters, though there are infinitely many sentences.
Chomsky 1957 :13
As might have been anticipated, readers unfamiliar with Shannon’s theory were somewhat perplexed by this means of expression and the conditional statements it
entailed. In contrast, readers familiar with Shannons theory will see immediate resemblance.
Similar expressions regarding the set of possible messages, the signal structure, and the correspondence between structure and meaning appear in Chomsky and Halle
1968 ,
where they write, “We may think of a language as a set of sentences, each with an ideal phonetic form and an associated intrinsic semantic interpretation. The grammar of the
language is the system of rules that specifies this sound-meaning correspondence”
Chomsky and Halle 1968 :3; italics added.
The grammar is described here as the system of rules specifying the correspondence of sound and meaning.
76
This is, of course, a clear point at which Chomsky and Halle cross the line into code model territory. Yet, in spite of this description of grammar, the
generative tradition has in general given relatively little attention to the issue of semantics. Within the generative view it is considered necessary to isolate semantics and
structure, so that structure may be defined independently of any semantic or pragmatic context in which it might appear, for language was presumed to represent the structural
repertoire
of the set of possible messages. The manner in which sentences may be related to meaning and employed in actual performance is considered a separate issue altogether.
It may be noted that Shannon had employed a similar strategy in isolating structure and semantics. Shannon writes: “Frequently the messages have meaning; that is they
refer to or are correlated according to some system with certain physical or conceptual entities. These semantic aspects of communication are irrelevant to the engineering
problem”
Shannon 1949 :3. Chomsky followed, stating: “… the notion of “gram-
matical” cannot be identified with “meaningful” or “significant” in any semantic sense” Chomsky 1957
:15. He continues: “Despite the undeniable interest and importance of semantic and statistical studies of language, they appear to have no direct relevance to the
problem of determining or characterizing the set of grammatical utterances. I think that we are forced to conclude that grammar is autonomous and independent of meaning ….”
Chomsky 1957 :17.
While it is not an exact parallel with Weaver, the modular semantic component Chomsky introduced in his
1965 Aspects of the Theory of Syntax is in some respects
76
As with generative linguistics, readers may also note some similarity between Kenneth L. Pike’s notion of a form- meaning composite and the notion of signs espoused by Saussure. Pike, however, is careful to distinguish his own
view from that of Saussure Pike 1971
:63. It should also be noted that, in contrast to Saussure and Chomsky, Pike is intensely interested in context. Indeed, he defines the key component of his theory, the tagmeme, as a “unit in
context.”
4. Code Model Linguistics: Patch or Abandon? 149
reminiscent of Weaver’s independent semantic receiver and semantic noise input.
77
Weaver writes:
One can imagine, as an addition to the diagram [see figure 3.6
], another box labeled “Semantic Receiver” interposed between the engineering receiver which changes signals to messages
and the destination. … Similarly one can imagine another box in the diagram which, inserted between the information source and the transmitter, would be labeled “semantic noise” ….
Weaver 1949b :115–116
Chomsky’s description of language via the conditions spelled out in Shannon’s information theory effectively served to restrict the definition of language and elevate the
notion of grammar. Whereas earlier linguists had concerned themselves with language as the langue of respective speech communities, in the generative tradition language was
simply the product of grammar. Concerning this restricted definition, Simpson writes:
A particular use of the phrase ‘a language’ is that of transformational-generative grammar where a language is held to be an infinite set of sentences, each sentence being finite in length
and constructed out of a finite set of elements. This view … is thus able, it is claimed, to give a precise definition of a language; for a language is seen to be the output of the grammar that
can construct the set of sentences constituting the language: consequently, a given language can be defined as that which a given grammar produces.
Simpson 1994 :1896–1897
In code model terms, then, generative linguistics views the grammar as the code i.e., algorithm used to generate the language. More importantly to the history of linguistics,
the code as grammar is made the focus of attention.
4.4.1.1.2. An abridged model of communication
