CHAPTER IX

Intonation and Phonesthesia

The reader may have noticed that during the course of the foregoing chapters I have treated intonation contours as both lexical and phonological elements. This ambivalence is no accident; it is, I believe, intimately related to the question of the connection between intonation and phonesthesia or ideophonic meaning. The phonesthetic nature of intonation has been mentioned by others (e.g., Bolinger 1947, also 1949), and is discussed at some length by Liberman (1978), who attempts to state rigorously what is involved in ideophonic meaning. Since topics like phonesthesia have long lain at the fringe of respectable linguistic inquiry, Liberman’s attempt is valuable, and his explicit claim that intonation and phonesthesia share some specific semantic characteristic is, as we shall see, an important insight. But I believe his description of the relation between the two is seriously in error, and in this chapter I present an alternative view.

At the heart of Liberman’s discussion of phonesthesia is the postulation of two different “modes of lexical structure,” ideophonic and morphemic. This is not a rigid dichotomy—“it is possible for a given word to have both ideophonic and morphemic analyses independently, and most lexical systems have this character to some extent”—but the distinction is basic to his characterization of the intonational lexicon, which in his view “has a fundamentally ideophonic structure” (97). The two modes exhibit a cluster of attributes as follows:

ideophonic: iconic, metaphorical, not clearly segmentable;

morphemic: arbitrary, referentially precise, generally clearly segmentable.

Liberman explicitly states (93-97, passim) that iconicity and arbitrariness are the important features of the two modes, and that the other characteristics follow naturally from them.

It seems to me that his whole discussion suffers from a misunderstanding-more specifically, a gross overextension—of the concept of iconicity. He refers to the iconicity of echoic or ideophonic words as “the metaphorical relationship of the sound of a word to a non-linguistic sound” (94), and defines iconic meaning as a “mode of meaning in which the signifié is a general metaphorical extension of some intrinsic property of the signifiant” (96). This appeal to metaphor is his way of explaining the obvious fact that some ideophones are less iconic than others:

Liberman has failed to distinguish metaphor from what I will call conventionalization. Any iconic symbol naturally involves some conventionalization of the relation between one percept (the symbol) and another (the referent). “At the lowest size-level of most, or all, iconic systems,” say s Hockett, “one finds a layer of arbitrariness. Thus a road-map means the territory it represents iconically down to a certain level, but there is no precise correlation between the width of the line representing a road or a river and the actual width of the road or river—these features are not representfed to scale” (1958:577). This conventionalization is simply the extent to which the form of a symbol diverges from the form of the thing symbolized in order to make the symbol fit the exigencies of the symbolic medium. Human noises are not exactly like cows’ noises, and the closest we English speakers come is our conventional word moo.

But this conventionalization is not, pace Liberman, the same as metaphor, and it is important to distinguish the two. No metaphor is involved in getting from the sound a cow makes to the English word moo, only conventionalization. But when a child uses this word moo to mean ‘cow’ rather than the noise the cow makes, then that is not part of the conventionalization, but a metaphorical extension of meaning. In the same way, the fact that click meaning ‘sudden insight’ or crash meaning ‘financial collapse’ are at bottom iconic does not mean that click is some how an iconic representation of the form of a sudden insight, or crash an approximation of what it sounds like when an economic system fails. Iconically, click represents only the sound of, say, two parts of an assembly falling into place together, and crash the sound of a heavy object falling or colliding with something. It is metaphor that does the rest.

Naturally, such metaphorical extensions are themselves so often conventionalized that it may be difficult to separate literal from metaphorical, and I am obviously not proposing a way to draw the line between the two in specific cases. The point of the distinction, rather, is to escape the necessity of concluding, as Liberman does, that metaphor is the essence of iconicity. That conclusion is surely to be avoided, for it is beyond question that thoroughly arbitrary symbols, not just iconic ones, are subject to metaphorical extension as well. Cold can mean ‘unfriendly’. Bright can mean ‘intelligent’. Sharp can refer to knives, but also to cheeses and repartees. Liberman’s assumption that metaphorical extension of abstract meaning is a natural concomitant of non-arbitrary symbolism, and that “for obvious reasons,” the arbitrary signs in the “morphemic mode of lexical structure” are characterized rather by “referential precision” (97) is a most extraordinary view, one which cannot possibly withstand serious scrutiny.

Equally untenable is the related claim that all ideophonic meaning is iconic. While it is evident that there is some similarity of sound and meaning in words like moo, bang, knock, etc., I see no basis for the claim that the combination of sounds /gl-/ has anything but an arbitrary relationship with vague emissions of light. In the same way, the widespread assumption (which Liberman shares; cf. p. 96) that gestural meaning is iconic is also lacking in empirical support. It is difficult, for example, to imagine what sort of non-arbitrary connection might be found between up-and-down head movements and the meaning ‘yes’, or between sideways head movements and ‘no’, especially in light of the fact that in the Middle East the connections work the other way around.¹

And yet there does seem to be some basis for asserting that gesture, phonesthesia, and intonation somehow exhibit a more direct connection between form and meaning than other segmental morphemes. Liberman assumes that this is because their meaning is iconic, and I suspect that many would agree. But if the term ‘iconic’ is not to be reduced to meaninglessness, I think it must be reserved for those cases where, in Hockett’s words (1958:577), “there is some element of geometrical similarity between the [word] and its meaning.” I think that we can express what is special about phonesthesia with reference not to iconicity, but to its be havior in relation to the design feature of language that Hockett (e.g., 1958) has called ‘duality of patterning’ and which is also known especially among European linguists as ‘double articulation’. This is, roughly, the fact that the huge number of meaningful chunks of any language (pierernes) are arrangements of one or more of a relatively small number of meaningless chunks (cenemes).²

Since language has duality of patterning, words have an ambivalent nature: they are simultaneously strings of cenemes and configurations of sound. Iconicity exploits this ambivalence: in iconic words, the phonemes function not only as cenemes, but also as sounds. That is, bang is distinguished from ban, bag, pang, bung, etc. by the cenematic—message-differentiating—function of the phonemes of which it is composed. But independently of the purely linguistic structure of the phonology, there is also a more direct connection between the sound of the word bang and its meaning. From the point of view of the design features of language, the peculiarity of iconicity lies in the fact that there is a sound-meaning link which is independent of cenematic structure.

This property of bypassing the cenematic structure is the hallmark of phonesthetic or ideophonic meaning in general. Iconicity is only the simpiest case—where the direct sound-meaning link is due to similarity in form between the sound of the symbol and the sound of the referent. But if we define phonesthesia as a sound-meaning link independent of cenematic structure, then there is no need to equate it with iconicity. Iconicity is a special case of phonesthesia, but not all phonesthesia is iconic. For example, there is no need to claim that the sound-meaning relationship in the English words gleam, glimmer, glisten, glow, etc. is in some mysterious metaphorical way ‘not arbitrary’, but only that the effect of this initial /gl-/ is an effect of the sound, independent of the cenematic structure.

Such a treatment accounts for the fact that phonesthetic effects have long frustrated the segment-minded analyst’s attempts to assign them to specific morphemes. It is a well-known problem of morphemic analysis that words like gleam appear to contain a ‘recurrent partial’ /gl-/, which makes a recognizable (if vague and abstract) contribution to their meaning, yet if we segment /gl-/ as a morpheme, we are left with forlorn would-be morphemes -int, -immer, -isten, etc., which do not recur. But if we view phonesthetic meaning in the terms I have suggested, we will not attempt to segment these words. Rather, we will say that the sound of /gl-/ at the beginning of a word has acquired an arbitrary meaning in English, a meaning independent of the function of /g/ and /1/ as cenemes, and thus without the segmentability characteristic of ordinary morphemes.

Such a definition of phonesthesia also explains why effects like those in gleam and glisten seem somehow secondary. The phonemes of which such words are composed function primarily as cenemes, and only secondarily as actual sounds. That is, the lexical structure of English would in some sense not be affected if the word bang were replaced by, say, foss, or if the gleam series were replaced by steke, forstle, warken, sugg, etc. The same meanings in the lexical structure of English could be attached to the new creations, and all that would be lost would be the phonesthetic meaning associated with the gl- initial of the originals.

By saying “all that would be lost,” I do not mean to belittle the power of such ‘secondary’ associations. Obviously, they are exploited fully by poets and other sensitive users of the language. Moreover, there can be little doubt that the word-formation processes of a language are influenced to some extent by the phonesthetic intuitions of native speakers. (See, e.g., Bolinger 1940, 1950). The point is simply to characterize the sense in which they are secondary and the sense in which they truly involve sotmd-symbolism. The phonesthetic effects exist, as it were, by the grace of the actual sounds which manifest the cenematic structure of the language and the actual string of cenemes in a given word.

An analogy to written language will help make this clearer. A certain bagel store in Ithaca has a flashy sign advertising HOT BAGELS. The sign is painted so that a picture of a bagel is used for the О in HOT. Now this is the visual analog of the phonesthesia in, say, gleam. It is entirely irrelevant to the primary message of the sign HOT BAGELS that the actual shape of the letter (ceneme) О is like the shape of a bagel, but the sign painter exploited the similarity in shape to add a vague secondary meaning to the message of the sign. If the word for ‘hot’ were heisse, or the shape of the letter O’ were 5, then the sign would still convey exactly the same primary message with the form HEISSE BAGELS or H5T BAGELS; but the secondary effect would be absent. Again, this is not to belittle the power of the ‘secondary’ association, for no doubt it sells a lot of bagels on a grey winter’s day. The point is simply that the primary function of the letter О is as a ceneme, a letter of the alphabet, and not as a shape similar to that of a bagel. In its cenematic function, О could theoretically take on any form; the secondary association in the sign is based on the shape it actually has.

This characterization of phonesthesia, incidentally, is independent of any element of universality that may be involved. There may well be universal associations, such as those that appear to exist between high front vowels and the notion of smallness, and indeed, there may be a whole network of human phonesthetic associations at whose existence we can scarcely yet guess. On the other hand, many such associations—e.g., the English /gl-/ case—may be idiosyncratic in individual languages. Our knowledge is simply inadequate to say for sure. But in either event, the structural peculiarity of phonesthesia is the bypassing of duality of patterning, the fact that sounds are associated with meanings solely by virtue of their sound and independently of their function as the realization of cenemes.

The connection of phonesthesia to gesture should by now be readily apparent: since gestures are not sequences of cenematic segments but perceptual gestalts, no cenematic structure is involved in their meaning. Again, the question of iconicity is irrelevant; gestural meaning is different from ordinary lexical meaning not because it is iconic—it may or may not be—but because it exhibits no duality of patterning.

And if the arguments presented for a contour rather than a level analysis are valid, then in exactly the same way, intonation exhibits no duality of patterning either. The smallest meaningful elements—e.g., nuclear tones—are not sequences of cenematic segments; the sound-meaning link is direct. In the terms we have been using, the exceptional aspect of intonation, phonesthesia, and gesture which distinguishes them from normal modes of lexical meaning is not iconicity, metaphor, or universality, but the fact that they do not involve cenematic structure. Phonesthesia is a sort of halfway-house, because the sounds used phonesthetically still constitute phonemes at another level—as in the English /gl-/ example. With intonation and gesture, on the other hand, cenematic structure is not bypassed; it does not exist.

Obviously, because Liberman chooses to perpetuate the analysis of contours as sequences of pitch levels, i.e., to postulate a cenematic structure for intonation, he cannot express the similarity between phonesthesia and intonation in the terms I have used here. I have argued that the explanations he does offer are based on confusion about the nature of iconicity and metaphor. But notwithstanding the confusions, his intuition that intonational and ideophonic meaning share some fundamental characteristic must be reckoned a significant insight. I feel that the foregoing discussion, in conjunction with the arguments in Chapter 8 on levels vs. configurations, preserves Liberman’s basic insight, while discarding some undesirable implications which are largely a result of his assumption that intonation must be expressed in terms of static tones.