The conduit metaphor and information theory

76 3. The Code Model Decoded metaphor, speaking of an “ill use of language” as a breaking or stopping of “the pipes.” Locke writes: For Language being the great Conduit, whereby Men convey their Discoveries, Reasonings and Knowledge, from one to another, he that makes an ill use of it, though he does not corrupt the Fountains of Knowledge, which are in Things themselves; yet he does, as much as in him lies, break or stop the Pipes, whereby it is distributed to the publick use and advantage of Mankind. Locke 1975 :Book III, Chapter 11, Section 5; italics added

3.3.2. The conduit metaphor and information theory

Before addressing the integration of the conduit metaphor and information theory within the code model, it is important to note that information theory is in no way an extension of the conduit metaphor, nor vice versa. As addressed in section 3.2 , the literature occasionally supports some confusion regarding this issue e.g., Berge 1994 :614. No doubt, this confusion stems from the fact that Reddy’s landmark article on the conduit metaphor also discusses information theory Reddy 1979 :302–306. It is important to read Reddy’s discussion carefully, however, for in no way does he suggest that information theory is derived from the conduit metaphor. Rather, he is addressing the inclusion of conduit metaphorisms in Shannon and Weaver 1949 . As Reddy concurs, when properly understood, the conduit metaphor and information theory do not support agreement, but rather a conflict in perspective. Via integration, the two find a relationship in the code model; but nevertheless, there is disparity between the conduit metaphor and information theory themselves, rather than a dependency. Although Reddy discusses information theory, he is not concerned with information theory, per se; he simply finds information theory a particularly useful example of the way in which conduit metaphorisms can hinder effective understanding of a theory. He writes: Indeed, the most convincing documentation [of the subtle pressure of the conduit metaphor] one could wish for is to be found in the historical development of mathematical information theory. For here, if ever, with both a concept-free algebra of information, and working machines to use as models, the effect of the conduit metaphor should have been avoided. But, in fact, it was not. And the conceptual basis of the new mathematics, though not the mathematics itself, has been completely obscured by the semantic pathologies of the conduit metaphor. Reddy 1979 :302 Of primary concern for Reddy is the way in which the conduit metaphor introduces “semantic pathology” into discussions of information theory, and thereby into appli- cations of that theory. Reddy borrows “semantic pathology” from Stephen Ullman, who explains that it arises “whenever two or more incompatible senses capable of figuring meaningfully in the same context develop around the same name” Ullman 1957 :122, in Reddy 1979 :299. To illustrate such pathology, Reddy offers example sentences employing two senses of the word “poem,” which he distinguishes in notation as POEM1 and POEM2 . POEM1 refers to poems as written texts and POEM2 refers to poems as mental abstractions: 3. The Code Model Decoded 77 POEM1 a The poem was almost illegible. b The poem has five lines and forty words. c The poem is unrhymed. Reddy suggests that examples a through c refer to a written text 1979 :299. Examples d through f, however, are a bit more ambiguous, suggesting an underlying abstraction 1979 :300. POEM2 d Donne’s poem is very logical. e That poem was so completely depressing. f You know his poem is too obscene for children. Of these, Reddy writes, “The most probable referent of the word is not the text, but rather the concepts and emotions assembled in the reading of the text” Reddy 1979 :300. Reddy then offers an additional example g, demonstrating a semantic pathology in which “poem” can be easily read as either POEM1 or POEM2 . POEM1 OR POEM2 g Martha’s poem is so sloppy Poems are certainly not the only items that are conceptually and expressively handled in this way. Concerning such semantic pathology, Reddy writes: It is easy to see that this ambiguity of the term “poem” is intimately related to the conduit metaphor. If the words in language contain the ideas, then POEM1 contains POEM2 , and metonymy, a process of meaning extension second in importance to metaphor, takes over. That is, when two entities are always found together in our experience, the name of one of them—usually the more concrete—will develop a new sense which refers to the other. Just as ROSE1 = the blossom developed ROSE2 = the shade of pinkish red by metonymy, so POEM1 gave rise to POEM2 . For, in terms of the conduit metaphor, the two are seen as existing together, the second within the first, and all the conditions of metonymy are met. Reddy 1979 :300; italics added Within information theory, a similar semantic pathology is demonstrated in the terms employed, both by the theoreticians who developed the theory and by others, particularly linguists. This is especially noteworthy concerning some uses of the term “message,” which, like “poem,” can be identified as MESSAGE1 and MESSAGE2 . The first is in reference to the signal sent by the transmitter, while the second is used in reference to pre-transmission material. As addressed previously, this occurs in Shannon’s 1948 paper and the 1949 reprint primarily in his decision to use common terminology e.g., infor- mation, message , even though he was using the terms in technical senses that did not closely follow the common usage. In Weaver, however, the threat posed by the conduit metaphorisms are more subtle and more problematic Reddy 1979 :304–305. For 78 3. The Code Model Decoded example, Weaver writes, “The transmitter changes this message into the signal …” Weaver 1949b :98; italics in original. 44 Concerning this issue, Reddy writes: If [Shannon and Weaver] did not [recognize the ambiguity and semantic pathology inherent in their terminology], I believe it is because their thought processes were responding to the biasing effect of the conduit metaphor. Weaver, it seems, could not hold the theory clearly in mind when he spoke of human communication, and used conduit metaphor expres- sions almost constantly. “How precisely,” he asked, “do the transmitted symbols convey the desired meaning?” p.4 Or he compared two “messages, one of which is heavily loaded with meaning and the other of which is pure nonsense.” p.8 In truth, it seems that he still thought of the MESSAGE2, the repertoire members, as being sent across the channel, even though this destroys the notion of information as selective power. Weaver hedges significantly when he describes the action of the transmitter. It “changes,” he says, “the message into the signal [italics Weaver’s].” p.7 Really, this is a strange description. A code is a relationship between two distinct systems. It does not “change” anything into anything else. It merely preserves in the second system the pattern of organization present in the first system. Marks or sounds are not transmuted into electronic pulses. Nor are thoughts and emotions magically metamor- phosed into words. Again, this is conduit-metaphor thinking. There is no justification what- soever in information theory for talking about communication this way. It is worth noting that Shannon, who actually originated the mathematics, may have had a more coherent understanding than Weaver. At some points in his own exposition, Shannon used exactly the right ordinary language terms. He wrote, “The receiver ordinarily performs the inverse operation of that done by the transmitter, reconstructing the message from the signal” p.34. But it still does not seem that he perceived the damage done to the paradigm by his own and Weaver’s conduit metaphorisms. Reddy 1979 :305; italics Reddy’s except where otherwise noted It is interesting to note that Reddy quotes Shannon and Weaver 1949 from the second edition, which reverses the order of the two articles so that Weaver’s article occurs first. Of course that change affected the pagination of the two articles. If that were the only effect generated by the change, it would be insignificant. It is likely, however, that the change was indeed significant. With Weaver’s article coming first in the book, that article served as an introduction to information theory, shaping readers’ under- standing accordingly. The first edition, which placed Shannon’s article first, more accurately placed Weaver’s article in the position and role of a commentary. Reddy notes that a similar problem occurs in information theoretic terminology that has been developed subsequent to Shannon 1948 and Shannon and Weaver 1949 . Certain of these terms misrepresent information theory, all the while supporting the conduit metaphor quite well. For example, the terms “encode,” “decode,” and “infor- mation content” all suggest that the message has somehow been inserted into the signal. As Reddy reminds his readers, “Signals do something. They cannot contain anything” Reddy 1979 :306. 44 Reddy 1979 :305 also mentions this quotation, referencing Shannon and Weaver 1949 :7, although from a different edition than that quoted in this study. 3. The Code Model Decoded 79

3.3.3. Information theory and Saussure’s speech circuit