3. The Code Model Decoded 71
Unfortunately, many communication scholars have failed to differentiate Weaver’s suggestions from the theory proposed by Shannon. Quoting Ritchie
1986 , Rogers
writes:
One source of confusion on the part of communication scholars is “the habit of citing ‘Shannon and Weaver’ when it is Weaver’s speculations that are being quoted, under the
assumption that they are somehow supported by Shannon’s mathematics.” When the Shannon and Weaver book is cited by communication scholars, they are usually referencing ideas from
Weaver’s part … rather than from Shannon’s part ….
Rogers 1994 :425, n. †
3.2.3.3. The diffusion of information theory into linguistics
While Shannon’s model obviously had some “first” point of introduction into lin- guistics, the fact that his model and information theory were so broadly employed in
other disciplines helped to insure that multiple points of entry were established.
39
Nevertheless, it is useful to note some of the early introductions.
40
The social and intellectual atmosphere of the day seemed to have greatly influenced the reception offered to Shannon and his theory. Having just come through World War II,
many in the scientific community were intensely interested in overcoming the perceived chaos stemming from that conflict. Many scientists turned their energies toward
providing systematic and mathematical explanations for social and communicative phenomenon. Information theory provided something of an answer to their question, for
in addition to its obvious applications in the fields of electrical and computational engineering, the theory initially seemed to offer great promise as a tool for analyzing the
social effects of communication.
Within this historical context, Norbert Wiener, a mathematician at the Massachusetts Institute of Technology, proposed a theory which, at the time, seemed to offer the hoped-
for explanation. Wiener called his theory “Cybernetics,” publishing a book by the same name, subtitled Control and Communication in the Animal and the Machine. The book
was published in 1948, the same year as the publication of Shannon’s initial article, “A Mathematical Theory of Communication.”
41
Cybernetics was to provide a broad context in which Shannon’s information theory could be viewed, and interest in cybernetics was to play an important role in the diffusion
39
It should be noted that having multiple points of entry may have contributed to acceptance of Shannon’s model of communication within linguistics. The point is that with several major theoreticians eagerly supporting the model,
and doing so independently, the concerns of any skeptics were easily overpowered by the positive support.
40
In his study on the diffusion of information theory, Dahling 1962
:130 maps several early exchanges of the theory as they appeared in various types of literature. Unfortunately, some of Dahling’s citations numerically cross-
reference an unpublished bibliography assembled by the Stanford University department of communication studies, rather than providing full bibliographic information. Of course this limits the usefulness of Dahling’s publication,
particularly since it seems that Stanford’s department of communication studies no longer maintains a copy of that bibliography Donald Roberts 1999, personal communication.
41
It is interesting to note that Shannon was a colleague of Wiener’s and acknowledges him in the 1948 paper and 1949 book
Shannon 1949 :81. Weaver also notes Wiener’s influence
Weaver 1949b :95, n. 1. Wiener similarly
acknowledges Shannon in his own writings.
72 3. The Code Model Decoded
of information theory into linguistics. For example, Charles Hockett writes the following concerning his early review of information theory which appeared in Language
1953 :
In 1948, Norbert Wiener’s “cybernetics” burst forth on the postwar intellectual scene, and Claude Shannon’s information theory made its more subdued entrance. These struck me as
genuinely new and as important. There had to be implications for linguistics and anthropology, and I set out to discover what they were. When I found enough to warrant writing this [1953]
review, I wrote it. Subsequently I found more—indeed, the information-theoretical approach became part of my standard intellectual equipment, fitted neatly into the physicalist frame of
reference ….
Hockett 1977 :19
While Hockett’s review was one of the more visible introduction points for infor- mation theory, it was not the first. For example, Oliver Straus, of the Research
Laboratory of Electronics at the Massachusetts Institute of Technology, had already published a brief article on “The Relation of Phonetics and Linguistics to Communication
Theory”
Straus 1950 . It appears, however, that Straus’ paper went largely unnoticed by
the community of linguists Dahling 1962
:130.
42
George Miller and J. A. Selfridge also published on certain applications of information theory in psycholinguistics in 1950. In
contrast to Straus’ paper, Miller and Selfridge 1950
was cited in several later linguistic publications
Dahling 1962 :130. In a footnote within his original
1953 review, Hockett
himself mentions several early references to information theory that occurred within linguistics prior to his own review:
References to information theory have already appeared in linguistics discussions: J. Whatmough, presidential address before the Linguistics Society, December 1951; C. F.
Hockett, Language 27.337, 445 1951; R. Jakobson, C. G. M. Fant, and M. Halle, Preliminaries to Speech Analysis, Technical Report No. 13 January and May 1952 of the
Acoustics Laboratory, MIT. Hockett 1953
:70, n. 2
43
In a comment which would seem to have foreshadowed the future now history of information theory within linguistics, Hockett also wrote, “It is not certain that all of
these references are based on adequate understanding of the theory” Hockett 1953
:70, n. 2; italics added.
Many linguists of the current generation are unaware of the significance cybernetics and information theory had within their historical context. Accordingly, they easily miss
the role these theories played in the development of contemporary linguistics. John Horgan provides some useful background material concerning the spread of Cybernetics
and information theory:
Cybernetics was created largely by one person, Norbert Wiener …. He proclaimed that it should be possible to create a single, overarching theory that could explain the operation not
only of machines but also of all biological phenomena, from single-celled organisms up through the economies of nation-states. All these entities process and act on information; they
42
Dahling 1962
:130 also mentions several other early articles which were similarly overlooked. Unfortunately, the bibliographic information for these others is lost. See again note 40 of the present chapter.
43
Readers interested in researching Hockett’s citation of his 1951 article in Language 27 will find the cited material to be less than helpful. It is possible that the citation in Hockett
1953 involves a misprint. Nevertheless, that citation is
correctly quoted in the present study.
3. The Code Model Decoded 73
all employ such mechanisms as positive and negative feedback and filters to distinguish signals from noise.
… Closely related to cybernetics is information theory, which [was developed by] Claude Shannon, …. Shannon’s great achievement was to invent a mathematical definition of
information based on the concept of entropy in thermodynamics. Unlike cybernetics, information theory continues to thrive—within the niche for which it was intended. Shannon’s
theory was designed to improve the transmission of information over a telephone or telegraph line subject to electrical interference, or noise. The theory still serves as the theoretical
foundation for coding, compression, encryption, and other aspects of information processing.
By the 1960s information theory had infected other disciplines outside communications, including linguistics, psychology, economics, biology, and even the arts.
Horgan 1996 :
207– 208
In contrast to Shannon’s theory, which only proposed to handle the transmission of information, Wiener’s theory purported to explain the way in which information was
used in purposeful control. “In this sense, cybernetics may be regarded as the study of information handling”
Bothamley 1993 :132.
In addition to cybernetics and the mathematical study of social situations, much post- war research was poured into the idea of machine translation. Shannon’s theory seemed
to offer great potential in that area of study. Several key players in the code model drama seem to have come upon the theory by that course. Noam Chomsky, whose Syntactic
Structures
strongly reflects the influence of information theoretic constraints, alludes to the impetus of such projects in comments regarding the machine translation project
headed by Victor Yngve:
What happened is that [Victor] Yngve’s project was continually hiring very good linguists. I’d made clear even before I was appointed that I didn’t think the project made any sense. Others
Lees, Matthews, Lukoff … had varying views about the matter, and did work on aspects of it. But gradually they all reached the same conclusion, and began to concentrate more and
more on straight linguistics, then in a real ferment at MIT. Yngve wasn’t happy about it: he was dedicated to machine translation. Chomsky in personal communication to Barsky;
Barsky 1997
:90; brackets and ellipsis in original
Shannon’s information theory was also quickly adopted into behavioral psychology, which at the time had a strong relationship to linguistics. Pollack notes, “Shannon’s
development … was enthusiastically grasped very early by a handful of psychologists, primarily those associated with the Psycho-Acoustics Laboratory at Harvard University”
Pollack 1968 :333. Psychologist George A. Miller recounts that “initial enthusiasm was
stirred by the realization that Shannon had provided a tool for describing discrete events that at the same time was compatible with the continuous Fourier analysis of commu-
nication systems, with which the group was already acquainted” Miller in personal communication to Pollack;
Pollack 1968 :333. Chomsky’s position as published in
Syntactic Structures was to influence Miller such that he modified his early views
regarding human language and information theory; see Miller 1964
, n. 1, Miller and
Chomsky 1963 .
Miller also recounts some of his early exposure to linguistics, which occurred in the context of a series of work conferences on selected research problems see
Sebeok
74 3. The Code Model Decoded
1960 :1. In his closing statement offered at one of the conferences held in 1958, Miller
offers some interesting insight on the role these conferences played in the diffusion of information theory and, ultimately, in the development of the code model. He
comments:
Several times during this conference I had occasion to remember another conference held at the Massachusetts Institute of Technology about ten years ago, when Mr. Locke had the happy
notion that it would be good to get people together to talk about speech, language, and communication. So he began the conferences on speech analysis, which were attended by
communication engineers, mathematicians, psychologists, and linguists. … The linguists talked about some mystic entity called the “phoneme,” an abstraction that had never been seen
on the face of an oscilloscope. … The engineers had a fantastic idea that they could measure information, but the probabilities and logarithms involved made it certain that nobody really
understood what they were saying. And the psychologists kept insisting that the important thing was the perception of speech, which did not impress anybody but the psychologists. …
Those conferences continued, however, and eventually we managed to educate each other. I can now talk about information theory, and the phoneme does not frighten me quite the way it
did. Roman Jakobson will now use words like “source” and “receiver” and “message” and “code.” And many engineers are now deeply worried about the acoustic basis of phonemes
and speech perception.
Miller 1960 :386–387
While a few linguists did attempt to employ the theory of probability Shannon proposed for example, see
Hockett 1955 , most adopted the model only for analogic
purposes, giving considerably less attention to the theory it addressed see Pollack 1968
. By the mid-1960s, Shannon’s theory had been given certain modifications that contrib-
uted to such use. The most significant of these was the substitution of the terms encoder and decoder for Shannon’s earlier transmitter and receiver. The notion of encoding and
decoding had been employed in Shannon’s initial paper, but were used in reference to processes,
rather than mechanical components. For the purposes of linguists employing early versions of the code model, the change in terminology was convenient; encoder and
decoder were more easily personified than were the earlier transmitter and receiver, terms that spoke too clearly of their technical origins. In the process, Shannon’s essential
distinctions between the information source and transmitter and the receiver and destination were neglected. Contemporary linguistic appeals to the model invariably use
the adjusted terminology.
Having been adopted by linguists, Shannon’s characterization of communication began to serve the role of a true model. This is an important distinction, for Shannon had
not offered a model, per se. In his graphic representation of the model, for example, he had offered a diagram, a schematic, and discussed the workings of actual electronic
apparatus. For his purposes, he could have drawn pictures of telegraphic keypads, wires, and receivers, or even used a photograph, for that matter. A generic diagram simply
allowed him to address several types of apparatus with a single illustration.
In contrast to how they were used in engineering, within linguistics Shannon’s drawing and explanation were being used as an analogy, modeling a complex real world
process. As addressed in chapter 2
, models depend upon generalization; they can represent only a portion of the character of the real world phenomenon they are used to
3. The Code Model Decoded 75
describe. When such models are based upon an analogy, the strength of this character- ization is further limited by the quality of the comparison made. Shannon had recognized
human communication to be a complex process, but for his purposes had factored out elements which were not pertinent to his study of electronic transmission and reception.
Linguists adopting Shannon’s description without its associated theory were, at best, inadvertently subjecting the problem of human communication to reductionism.
3.3. Integrating the constituent models
Chapter 2
cited a selection of appeals to the code model. This section addresses the integration of the three constituent models by documenting several types of appeals:
1. Quotations which inadvertently appeal to two of the constituent models in addressing a single issue
2. Quotations which address one of the constituent models while employing terminology from another
3. Quotations which purport to discuss one or another of the constituent models, but do so via an appeal to the code model
Each of these types of appeal demonstrate willingness to fuse the models. In addition, some appeals demonstrate a naïveté regarding the distinctive qualities of the constituent
models. This does not mean, however, that all of the theoreticians quoted are even aware that they are appealing to multiple models.
3.3.1. Saussure’s speech circuit and the conduit metaphor
It should be clear at this point in the discussion that Saussure employed a translation model in developing the speech circuit model of communication. In addition to its being
a translation model, Harris notes that it is also a transmission model. Accordingly, it borrows apparent plausibility from the conduit metaphor. Harris writes:
The first point to note is that the speech-circuit model is a transmission model. It represents communication as involving passage through a succession of phases arranged in linear
progression along a track or pathway. In this succession there are no gaps. The process is envisaged as a continuous journey or transfer of information from one point in space to
another point in space: that is, from a location in A’s brain to a location in B’s brain or, in the reverse direction, from B’s brain to A’s. Now a model of this kind undoubtedly receives much
support from numerous expressions used in everyday speech to describe the processes of com- munication. For example, ideas are said to be put into words; words are exchanged; verbal
messages are put across or got over, sent or passed on; and eventually received and taken in. This way of talking about communication as transmission has been described as ‘the conduit
metaphor’
Reddy 1979 , and it is a metaphor with extensive ramifications in various
European languages. The influence of this metaphor in predisposing us to accept any transmission model of speech as plain ‘common sense’ is not to be underestimated.
Harris 1987
:213–214; underscore added
It is interesting to note that Locke, upon whom Saussure relies, actually employed the term conduit in describing communication. He also extends the basic conduit