CHAPTER Χ

Conclusion

By now, I hope, it should be clear to the reader in what sense this work has been a discussion of intonational meaning and not simply an analysis of English intonation. If there is a single most important point to what I have written, it is the following: insofar as past investigators of intonation have failed to produce satisfactory analyses, it is because they have failed to consider in what ways intonational meaning is structured like segmental meaning and in what ways it is different. They have gone into their study with implicit preconceptions about the organization of intonation and have forced suprasegmental phenomena into inappropriate molds.

The two most widespread—and yet contradictory—such preconceptions are that intonation, like the rest of language, is organized into all-or-none contrasting segments, but that functionally it is somehow around the edge; a third, consequent preconception is that anything that does not fit into the all-or-none categories is paralinguistic or ‘emotional’ (Chapters 5 and 6). I have shown how these preconceptions have made it possible for one investigator to analyze as all-or-none contrast what another ignores as emotional variation: since all is implicitly considered to be peripheral anyway, there is little basis for a decision about which parts are to be assigned to the explicitly peripheral—i.e., paralinguistic—domain.

I have argued that one of the principal ways in which suprasegmental phenomena differ from the rest of language is that they involve the systematic use of gradience. It should be emphasized that this is an aspect of meaning and not merely of form. In the segmental lexicon we assume that meaning is conveyed by contrasts between categories; in intonation, we must also assume that semantic continua are expressed by unsegmented formal continua within contrasting categories. I have shown (Chapter 5) how the explicit recognition of certain gradient dimensions relieves us of the need to see suprasegmental phenomena as either all-ornone or paralinguistic. Specifically, I have suggested that pitch range—both overall height relative to the speaker’s voice, and relative width or steepness of pitch movement—is an independently meaningful dimension of gradience, along which segments can vary without in some sense des troy ing their identity as segments; for example, steep fall and shallow fall are not two different contours, but are both instances of a single category fall, together with a difference of pitch range which makes a separate contribution to the meaning of the contour.

Note, however, that gradience, so defined, implies the existence of segments. I have argued (Chapter 7) that intonation proper includes a lexicon of contrasting contours; intonational meaning is thus lexical as well as gradient. Two claims are subsumed under this rubric. First, there is, as most past analysts have assumed, some part of the intonational system of English that is organized into all-or-none contrasting categories, like more familiar segmental phenomena. Second, and more controversial, the choice of intonational category is not specified by the grammar of the sentence of which it is a part, but makes an independently meaningful contribution to the interpretation of the whole sentence, a contribution not unlike that made by modal particles like doch and etwa in German.

It is with regard to this second claim, perhaps, that I have strayed furthest from my announced pretheoretical’ aim of merely providing a basis for further discussion of intonation. Naturally, no scientific work exists in a theoretical vacuum, and my opinions about what it is important to observe have been shaped to a considerable extent by the ‘anti-deterministic’ views of Bolinger and others. My conclusions in Chapters 4, 7, and 8 in particular reflect my view that as much as possible of what a speaker says should be seen as representing an independently meaningful choice.

I have shown in Chapter 8 that evidence from the use of certain ‘stylized’ tones argues against any phonological analysis in which contours are seen to consist of sequences of pitch level phonemes, and that the lexical segments of English intonation are to be considered phonologically unitary. This is a significant claim in two respects. First, it explicitly recognizes gradience as a separate parameter of intonational meaning; pitch level analyses are in principle unable to distinguish the more-or-less nature of certain pitch differences from the all-or-none nature of others. Second, and perhaps more important, it provides a possible explanation for the widespread preconception that intonation is somehow peripheral—specifically, that it is gesture-like—and for Liberman’s contention that intonational meaning is somehow like the meaning of ideophonic words: I have suggested (Chapter 9) that the essence of phonesthesia and gesture is that they do not involve duality of patterning, and that intonation contours, by being phonologically unitary, thus convey meaning phonesthetically as well.

Finally, I have devoted a good deal of space to exploring the implications of the rhythmic-relational view of stress developed by Liberman and Prince. I have showed how this concept enables us to interpret certain instrumental phonetic data which fit poorly into a ‘stress-level. or an ‘accent’ analysis (Chapter 2), and how it enables us to understand the phenomenon of deaccenting and to unify a good deal of mysterious accent-placement data (Chapters 3 and 4). Further, I have suggested that a similar approach might prove fruitful in discussing the ‘phrasing’ function of intonation (Chapter 7). It should be pointed out that accent seems to be exclusively a relational phenomenon, and does not involve segments—except, of course, the segmental items being related—at all. By contrast, the relational aspects of ‘intonation proper’ involve (at least in English) relations between intonational segments; the role of intonation cannot be expressed in terms of either segments or structural relations alone.

Intonational meaning, in summary, is lexical, gradient, phonesthetic, and relational.

This means that the first—essentially taxonomie—task of the analyst is to separate out lexical, gradient, and relational effects from among the observable semantic distinctions of intonation. The analyst will need to present not only an inventory of the lexical segments (meaningful contours) of intonation, but also a list of the dimensions of gradience along which the meaningful contours may vary (and the semantic effects of such variation), as well as a discussion of the role of relative height of nuclear pitch peaks and of relative pitch at boundaries.

I have outlined the general form of such a taxonomy, basing myself largely on past work. The most important features of the taxonomy, ineluding the separation of accent (rhythmic structure) from intonation (pitch movement), the internal structure of intonational tunes, and the inventory of nuclear tones, are presented in Chapter 1. However, there are many phenomena I have not discussed, such as: the classification and use of different heads (see Crystal 1969a); the uses of the dimensions of gradience; the semantics of focus and the status of compound stress; and the kinds of phenomena involved in intonational phrasing signals. Thus much work remains to be done at this taxonomie level, which is another reason I regard the book not as an analysis of English intonation, but only a discussion of preliminaries.

Once a taxonomy is better established, the phonologist will naturally be interested in questions like: what the distinctive features of intonation contours are, and how similar they are to tone in tone languages; how the domain of pitch contours is to be specified relative to the segmental string; what implications the simultaneous nature of intonation contours has for phonological theory. The phonetician, meanwhile, will want to investigate the acoustic cues by which we identify the different contours, looking, among other things, for evidence of categorial perception; much phonetic research also needs to be done on the nature of relational cues to structure. Finally, the grammarian will want to test different theories of syntax and semantics, not only in terms of segmental grammar, but also in terms of suprasegmental phenomena. As I noted above, it is here that I have let my theoretical viewpoint show through most strongly. It remains true, however, that my primary goal has not been that of testing models of language, but only of discussing how intonation must be ineluded in whatever model we construct.