NOTES

1. General Introduction and Review of Past Work

1. Excellent reviews of the literature are to be found in Crystal (1969a, Chs. 1-3)-the best synthesis of past work—and Gibbon (1976a, Ch. 3)—the most complete and up-to-date author-by-author synopsis available.

2. Sledd’s evidence was principally the ‘scooped’ intonations (see below, Ch. 2, Sec. 1), as in

As Sledd pointed out, this would have to be written /²won³derful¹#/, which frustrates Hockett’s criterion for identifying the intonation center.

3. There can be no doubt that ‘nucleus’ and ‘sentence stress’ are simply different names for the same phenomenon (see, e.g., Crystal 1969a;210).

4. In the following discussion I will use the term ‘pitch obtrusion’ rather than ‘pitch prominence’ (Bolinger 1958a seems to use the two more or less interchangeably), in order to avoid any possible confusion with ‘prominence’, my impartial cover term for what is variously known as stress, accent, etc.

5. O’Connor and Arnold, however, use the term ‘tone group’ for what we are here calling ‘tune’.

6. For example:

7. Bolinger (1966:690) writes: “I doubt . . . that there is that much difference between the dialects. All that I have ever heard a British speaker say has sounded normal to me. He merely has the annoying habit of using certain intonations too much and others too little.”

The following anecdote may help to explain our impression of substantial differences between the dialects. I was once in a conversation with an American anthropologist who had just returned from Nigeria, where for two years his only contact with native speakers of English had been with British missionaries. While he was catching up on presidential politics, someone said to him, “Did you know Carter was a nuclear engineer when he was in the Navy?” Somewhat surprised, the anthropologist replied:

I made a mental note of this apparent British influence on his intonation. Thinking about it later, however, I realized that what an American would have said in the same context is:

That is, in both dialects such echo questions (conveying perhaps a kind of restrained surprise or even skepticism) have falling-rising intonation; the difference is that the British use question syntax, while the Americans do not. The British influence was on his syntax, not his intonation. My initial reaction, however, suggests that there is a tendency to attribute dialect differences to funny intonation when they should actually be ascribed to funny syntax or funny lexical choice.

8. I grant that the concept of ‘intonational lexicon’—the term is Liberman’s—is rather novel; the discussion here and in Ch. 7, Sec. 2 is intended to show that the idea is not as far-fetched as it might seem.

9. Bolinger, it is true, rejected the whole notion that continuously changing pitch contours were appropriately described in terms of levels; see Ch. 8.

10. As I was making final corrections to the manuscript, Janet Bing drew my attention to the fact that Pike’s terms precontour and primary contour cannot be equated with the British terms as straightforwardly as I have done in Fig. 1. As she points out, Pike s taxonomy shares significant characteristics with Bolinger s accent analysis, in that each (Bolinger-style) accented syllable could be associated with a complete precontour-plus-primary-contour.

The same is true, I might add, for American analyses in the Trager-Smith tradition, in which this sentence could be transcribed ²De³ Tocqueville² | ²studied at ³Princeton¹ #, and thus in Hockett’s (1958) terms would consist of two macrosegments each composed of a pendant and a head.

While Bing’s observation that Fig. 1 is oversimplified is certainly correct, the problem in this instance is not so much a matter of how the terms of one tradition match up to those of another, but how the terms of any tradition are to be applied to the data. Specifically, both the Pike and the Trager-Smith analyses have trouble in practice in deciding whether the little variations in pitch that accompany stressed syllables are simply ‘allophonic variation’ or another pitch level phoneme, though this decision determines whether a given stressed syllable is part of a precontour or can claim its own primary contour. The same practical difficulty—in a different theoretical disguise—plagues the British tradition in its decisions on the location of tone group boundaries, which determines whether a given syllable is seen as a nucleus or only part of a head. In shorter sentences, where the practical uncertainties are less severe, then the correspondence between British and American is indeed as shown in Fig. 1:

The “common denominator” implied by the partial correspondence between the two traditions seems to be that all accented syllables have certain intonational characteristics in common; this is the basis for the American analyses and for the Kingdon/Schubiger use of head to refer only to the main accented syllable preceding the nucleus. However, one accented syllable of a sentence or tone group—usually the last—also has peculiar characteristics of its own, which is the basis of the British tradition. The reason for the indeterminacy in application of both traditions lies in the very indeterminacy of such notions as pitch phoneme, accented syllable and tone group.

11. Note that Vanderslice and Ladefoged’s attention to the intonation contours that follow the intonation center, while virtually ignoring what precedes it, implicitly involves the division between head and nucleus. Cf. Fig. 1.

12. Interestingly, however, Bolinger notes (19580:55) that “after an A accent there seems to be an all-or-none difference between a level and a rise, but a gradient difference between a level and a fall”—i.e., the very distinction between fall-rise and fall.

13. Jackendoff (1972, Ch. 6) misses this point. While borrowing Bolinger’s terminology and appearing to borrow his concepts, he is actually using ‘A accent’ and ‘B accent’ to refer to units like the British tones, as can be seen from his definition: “the intonation on an emphatically stressed syllable plus the intonation following until the next emphatically stressed syllable or the end of the sentence, whichever comes first” (Jackendoff 1972:258, emphasis added).

14. This implies some reanalysis of full and reduced vowels, which he undertakes in Bolinger (1976).

15. Since Lieberman’s emphasis is on refuting the Trager-Smith analysis, with its claim that stress levels are manifested by different degrees of loudness and are perceptible independently of context, it is not always clear how his proposals compare to others’. For example, his 1965 experiment involved only a single short sentence (They have bought a new car) and his data are presented in very condensed form, so it is impossible to tell whether the Trager-Smith ‘secondary stresses’ were perceived in the processed utterances as stressed or unstressed, or inconsistently. Moreover, to judge from his use of the feature [+Ps] (‘prominence’) in other parts of his 1967 book, it appears that that feature does not correspond to the syllable prominence consistently perceived by the linguists in the 1965 experiment. Clarification would be helpful.

16. Nowhere in the article Stockwell cites (Bolinger 1958fr) does Bolinger introduce the phrase ‘morphological stress’ as a technical term; the sentence from which Stockwell appears to have taken this is the following: “I use accent now on the syntactic side, and reserve stress for the morphological” (82). In the context of the rest of his work—and indeed, in the context of the passage (79-83) in which the sentence just quoted appears, it is clear that what Bolinger means by ‘morphological stress’ is not a level of stress at all but the “potential for prominence (accent)” discussed above. As we have seen, in Bolinger’s original scheme the only actual distinction of prominence is between accented and unaccented.

17. Hockett, in a seminar discussion of the origins of the phonemic principle (Spring 1977), compared the establishment of an important new idea to starting a car in winter—just as the car will sputter and die several times before catching, so the idea will surface and disappear several times independently before finally being put in a form which captures the attention of a whole field. Liberman and Prince’s article seems to indicate that the idea of stress-as-rhythm is finally running smoothly.

18. In this connection, Gunter observes wryly that class arguments over how to mark a certain type of falling-rising contour in the Trager-Smith system “are settled by professorial fiat, a kind of authority that must have settled many such arguments over the past fifteen years” (1972:212η).

19. The role of rhythm in stress has also been discussed by a few British linguists, notably Abercrombie and Halliday. The notion of ‘rhythmic foot’, which Halliday (1967) incorporates into his analysis, is taken over from traditional metrics into linguistic work by Abercrombie (1964); one of Abercrombie’s concerns is formalizing the elusive notion of ‘stress timing’ first suggested by Pike (1945) (see Halliday 1967:12). As we saw, Halliday’s distinction between salient and weak syllables is based on position in the foot, and would thus seem to be a purely rhythmic phenomenon. Yet an ambiguity appears in the analysis when he talks (e.g., in the passage quoted above in Section 5a) of acoustic correlates of stress like pitch movement and syllable duration. Do these physical properties define the salient syllables, which in turn define the foot boundaries, or does the rhythmic structure of the feet define the salient syllables, whose physical properties then follow from their salience? (This latter position would be roughly the equivalent of Liberman and Princes.) Halliday does not say, which is presumably the source of Crystal’s skepticism (1969a:202-203, 1969c: 382) about the whole idea of rhythmic feet. To my knowledge, there has been no other British development of the notion of stress as a rhythmic phenomenon.

20. Level tone, as can be seen from Fig. 5, is the subject of much disagreement and uncertainty; in Ch. 8 I propose a new analysis which shows that the level tones do differ from the others on the chart in that they are part of a subsystem of ‘stylized intonation’. As for the taxonomy of heads and preheads, the best that can be said is that there is far less agreement—and for that matter far less discussion—than there is on the taxonomy of tones. No attempt has been made to treat heads and preheads here; the reader is referred to Crystal (1969a) for the most complete discussion to date. See also n. 10 above and Ch. 3, Sec. 3 below.

2. Evidence for the Rhythmic Nature of Prominence

1. This ceteris paribus condition might be seen unsympathetically as reducing Vanderslice and Ladefoged’s definition to near-circularity. At the very least, we should expect a little more precision about what other factors are taken to be cetera and why, especially since the rhythm hypothesis seems to offer a consistent account of how the phonetic nature of nearby syllables can affect the timing of a syllable and its perceived prominence.

2. The term ‘scoop’ in this sense goes back at least to Chao (1932), and has been used by Pittenger et al. (i960) and Vanderslice and Ladefoged (1972). Gunter (1972) calls these intonations ‘humped descent’. These were the contours which Sledd (1956) used as evidence of the inadequacy of Hockett’s definition of intonation center (see above, Ch. 1 n. 2), though Sledd did not use the term ‘scoop’. Householder (1957:239n) uses ‘scooped’ to refer to the Liberman-Sag ‘contradiction contour’, i.e., a sequence of high-falling head and low-rising nucleus; I do not know if this usage is widespread.

3. In British terms, this major pitch jump marks the end of the prehead and the beginning of the head.

4. Bolinger and Gerstman would not disagree with this point: crediting both Householder and J. D. O’Connor, they note (85-86) that the difference between lighthouse keeper and light housekeeper is functionally the same as the one seen in high-line voltage and high line-voltage, even though in the latter examples only added length, not actual silence, is involved. However, they take this as refining their understanding of the notion ‘disjuncture’, which they define not as ‘separation of syllables’ but as ‘separation of syllable centers’. They are then free to see disjuncture, so understood, as indicating constituent structure.

5. In this context we can see again the error of Stockwell’s identifying accent with primary stress, commented on above in Ch. 1, Sec. 5b.

6. This solution involving the feature [+emph] was suggested to me in a letter from Vanderslice.

3. The Phonology of Deaccenting

1. Compare Jakobson and Halle’s remarks (1971:33-38) on the ‘relative’ nature of prosodie features as compared to ‘inherent features’.

2. Thirty years ago, Nida (1948:269, п. 44) suggested that “Junctures are to be treated on the same level as order, which is the other formal feature of arrangement”-that is, junctures are not chunks, but features of the arrangement of chunks.

3. The ideas in this chapter and the next were developed before I had seen Liberman (1978) or Liberman & Prince (1977), and an earlier written version has appeared as Ladd (1979). This chapter especially has been rather extensively rewritten to link the exposition more explicitly to the relational concept of accent.

4. Schmerling admits as much and attempts to head off such criticism with the following apologia:

But it should be clear that any notation system carries with it a load of theoretical implications that cannot be so easily wished away. In particular, in the context of the treatment of deaccenting proposed here, it can be seen that the notion of ‘absence of stress’ is quite meaningless. More specifically, Schmerling’s separation of ‘topic—comment’ sentences from the rest of deaccenting, which we shall mention presently, is purely an artifact of her notation.

5. The cases of ‘prenuclear’ deaccenting discussed in the next several pages raise two related questions: (1) Why does the accent shift right instead of left in some cases of deaccenting? (2) In certain of the ‘normal’ or ‘out of the blue’ versions, why is the accent not on the rightmost content word, as predicted by almost every ‘normal stress’ rule ever proposed? The reader is asked to postpone these questions until the next chapter, where we discuss the whole question of how accent ends up where it does; only the phonological and relational nature of deaccenting is at issue here.

6. This is often claimed to be common in women’s speech, but is actually found in both men and women in such contexts, to signal hesitance, tentativeness, or uncertainty.

7. Vanderslice and Ladefoged’s concept of deaccenting—changing [+accent] to [—accent]—does not handle these cases any more convincingly, since both a pretonic accent and an A accent involve a ‘pitch obtrusion’ and must thus, by their definitions, be considered [+accent]. Vanderslice’s (1977) example Jean ate her soup at the smorgasbord does appear to treat the difference between accented (i.e., A-accented)Jean and pretonic-accented (i.e., deaccented) Jean simply by changing the value of the feature [Accent]:

However, this example succeeds only because Jean begins the utterance with a strong syllable, and there is thus no pitch jump (see Ch. 2 Sec. 1). If we changed the name to Janine, we could still have the contrast between accented (A-accented)Janine and pretonic-accented (deaccented)Janine, but this time both would involve a pitch jump to the syllable-nine:

In the second case Janine is functionally deaccented just as is the second exampie of Jean, but because of the pitch obtrusion of the pretonic accent Vanderslice and Ladefoged must label it [+accent]. Cf. the comments on Bolingers explanation of rhythmic stress shift, above page 38.

4. The Grammar of Accent Placement

1. As Schmerling argues and as Bolinger himself has pointed out repeatedly, it is generally not possible to attach an explicit meaning to the term ‘contrastive stress’ anyway. This is discussed further in the next section.

2. Schmerling (1976:76) rejects Chomsky’s use of focus in this context largely because his attempt to formalize it involves what she sees as an inappropriate use of the notion of presupposition. In one of Chomsky’s formulations, if the focus of a sentence is replaced by a variable (e.g., an indefinite), the result is an expression of one of the presuppositions of the sentence. Thus in John writes poetry in the garden, the focus is poetry and the presupposition is John writes something in the garden or John writes X in the garden. Jackendoff (1972) makes some further attempts along the same lines to formalize the notion of focus (see Ch. 7, Sec. 3 below).

3. My analysis thus separates the accentual and intonational aspects of ‘contrastive stress’. Narrow focus is signalled solely by the location of the accent; various intonational characteristics such as greater volume and widened pitch range can also be used to signal what might be called ‘emphasis’. It is quite possible to have narrow focus without emphatic intonational cues, and equally possible to have emphatic intonational cues without narrow focus, as in

But it is quite true that the two frequently occur together, especially if a narrow focus is intended on an item which would receive ‘normal stress’ anyway, e.g., sh

(Cf. also n. 5 just below). Something like the separation of accentual from intonational cues is suggested by Cutler and Isard (1978) in their discussion of contrast.

The fact that the two types of cues can be used in combination should not be permitted to obscure their separateness, though it has certainly tended to confuse the whole question of ‘contrastive stress’ even more than it is to begin with: more often than not contrastive stress is considered in theory to depend on the intonational cues but is defined in practice, as we have seen, on the basis of accentual cues, as the converse of normal stress.

4. The idea is hinted at in a few places. Bolinger (1958a:52) gives an exampie of a sentence with the accent on a preposition and notes that in the context any other accent position would be ‘contrastive’. Schmerling (1976:89) also mentions accent placement by default in connection with ‘topic-comment’ sentences; but as we have seen, she considers deaccenting to involve actual absence of stress, and does not develop the notion of default accent.

5. Several friendly critics have objected that it is possible to disambiguate these two readings with intonation. Thus:

would likely be interpreted as deaccenting languages, while

would tend to imply focus on speak. But while it is possible to disambiguate, it is not necessary; moreover, the correlations between certain intonations and certain interpretations are, I think, only probabilities. From the point of view of where the accent goes and not how it relates to the intonation—i.e., purely from the point of view of the rhythmic structure, not the intonation—the structure is ambiguous.

6. This brings up the interesting question of the interaction of the word order with accent, which is discussed in some detail by Danes (1967) and Schubiger (1964).

7. Halliday (1967b) also presupposes such a distinction; see especially p. 206.

8. The greater accentability of content words than function words is part of any description of English sentence accent and I presume needs no justification here. As for the greater accentability of nouns than other content words, the reader is asked to suspend the amassing of counterexamples until subsection d. below, where some of the details of this idea are discussed.

9. Notice that this view of ‘contrastive stress’ (in Bolingers sense) actually argues against the distinction that Bolinger makes between contrastive accent and contrastive stress. Focus on part of a word is seen here merely as one end of a spectrum, not as something essentially different from focus on a word or constituent.

10. Instances of such accent shift in noun compounds are surprisingly common in everyday speech, as one quickly learns when listening for them. Note that the examples here have been chosen so that the ‘contrastive’ argument is not possible. In, say, We’ve got lots of fuel púmps, but no fuel filters, one could argue that pumps and fílters take the accent away from the normally accented fuel because of the contrast. But in We’ve got lots of bóoks, but we haven’t got any bookcáses, in order to explain the accent on cases as contrastive, one would have to say that it contrasts with zero—which is true of anything.

11. In this case there is another possible solution, namely, not to deaccent at all. (We’ve got lots of bóoks, but we haven’t got any bóokcases). Compounds seem to vary in their tolerance for having their internally determined accent pattern disrupted by deaccenting.

12. Note that in all these examples George must be interpreted as deaccented, as in the other cases of deaccented names discussed above.

13. Thanks to Eric Brook for stretching my interpretive component with regard to these sentences.

14. This is not to imply, of course, that this general view is uniquely Bolinger’s; it is espoused by a variety of (broadly) ‘functionalist’ writers. Bolinger, however, has been most active in applying this outlook to suprasegmental analysis.

5. Paralanguage and Gradience

1. For the intonation of yes-no questions to which Bolinger refers in this quote, see Fries (1964) and Lee (forthcoming).

2. I am taking for granted here that dimensions such as loudness, tempo, and various differences in ‘tone of voice’ are properly considered “less linguistic” or paralinguistic and are organized in more or less the way described by Crystal (1969a, Ch. 4; see also 1975).

3. But Nash and Mulac, concentrating on the pitch movement on thought rather than on the overall pitch contour of the utterance, call the difference between these two contours the contrast between Bolinger’s A and В accents. This shows, among other things, the difficulty of applying Bolinger’s taxonomy.

6. Around the Edge of Language?

1. From Wooden Ships, by David Crosby and Stephen Stills, copyright 1969 by Gold Hill Music and Guerrilla Music.

2. Bolinger (1964:20-21) make comparable observations about the universality of certain paralinguistic cues.

3. To the extent that Osser was eliciting judgments of paralinguistic emotional signals and not of linguistically systematized intonation contours, this result is exactly what we would expect.

4. In fact, my colleague Louis Mangione informs me that compared to other European works of the same period, Hillier’s book is actually quite enlightened.

7. Intonation and Grammar

1. Pike is obviously concerned here with distinguishing intonation from lexical tone in tone languages, but the fact that that is his emphasis does not alter his fundamental distinction between intonational and segmental (‘lexical’) meaning.

2. Reacting to the foregoing section, Cutler writes (personal communication) that some as yet unpublished research by herself and others suggests that ‘closed class’ or function words are stored rather differently in memory from ‘open class’ or content words. If true, this is obviously an important finding, and to some extent deflates my assertion above that Cutler would “certainly” include closed class words in the lexicon. It seems to me nevertheless that the possibility of significant subdivisions within the lexicon does not affect my basic point, which is that intonational meaning should be assumed to be like lexical meaning until proved otherwise. We can, if research like Cutler’s warrants, change this to “intonational meaning should be assumed to be like one significant subdivision of lexical meaning—say, closed class words—until proved otherwise”; the point is essentially the same.

3. An earlier version of the fall-rise analysis presented here appears in Ladd (1977).

4. Jackendoff incorrectly borrows Bolinger’s terminology, calling fall ‘A accent’ and fall-rise ‘B accent’. We have already seen (Ch. 1, Sec. 3) that in Bolinger’s system fall and fall-rise are both A accents; В accent is very different. See Fig. 2, and also Ch. 1 n. 12.

5. Further evidence for this analysis is presented in Ch. 8.

6. The nucleus of the contradiction contour, of course, can move around freely:

This seems to provide further evidence that the contradiction contour consists of a low-rise nucleus and a high-falling head; as we said in Ch. 1, however, this does not preclude interpreting the combination of the two as some sort of ‘holistic’ contour as Liberman and Sag suggest.

7. However, this is not to convey the impression that the taxonomy of fallingrising contours in English is clear and simple. Many writers distinguish fall-rise (with nucleus early in the sentence) not only from the contradiction contour, but also from fall-rise with nucleus late in the sentence. Lee (1956b), for example, distinguishes the falling-rising tune (roughly = contradiction contour), falling-rising sequence (nucleus early in the sentence), and falling-rising tone (nucleus late in the sentence). Halliday likewise separates his tones 13 and 53 (nucleus early in the sentence) from tone 4 (nucleus late in the sentence); Crystals notion of ‘nuclear tail’ (1969a:01.5) deals with the same problem. However, the distinction between the falling-rising sequence and the fallingrising tone is much more elusive than that between the fall-rise and the contradiction contour, and Kingdon, for example, explicitly treats fall-rise early in the sentence (‘Divided Tone ΙΙI’ or one IIID as a type of fall-rise (Tone III). (This treatment is implicit also in Palmer 1922.) Moreover, even analysts who insist on a distinction between the falling-rising sequence and the falling-rising tone show by their sometimes inconsistent transcriptions that this distinction is somehow less critical than others; these inconsistencies will be the subject of a separate paper. In any case, I do not feel that my failure to make finer distinctions among fall-rise tones represents a serious deficiency of my analysis, for the discussion in the next section seems equally applicable no matter where the nucleus occurs.

8. Here and throughout the rest of the chapter we will use the widely used British ‘tonetic’ notation first developed by Kingdon (1939). A ^̌ placed directly before the nuclear syllable indicates a fall-rise; a ⁴ placed directly before the nuclear syllable indicates a fall. Thus I fed the ^̌cat is to be interpreted as

and In Ithaca, maybe is to be read

9. Actually, this is an oversimplification. With a fall-rise tone the speaker could also say In Massa ^̌chusetts, maybe (. . . but he doesn’t even live in New York). Here the set being referred to is something like ‘states of the U.S.’, represented in the context by New York. This recalls the situation remarked on above, where the set ‘animals’ is taken to be in the context on the basis of speaker A’s mention of dog.

10. Chafe (1976) implies the same kind of analysis of double-nucleus sentences as JackendofFs: he talks of ‘pairings’ of nouns, and implies that the pairing is what is being asserted.

11. Something like the interaction with quantifiers and superlatives seems to be the explanation for another specific fall-rise nuance pointed out by Cutler (1977):

The use of fall-rise in such contexts—with a wide variety of words expressing mild approval, like OK, all nght, decent— sets up a hierarchy of possible evaluations and explicitly puts the evaluation given somewhere between the possible extremes. Hence the nuance of mediocrity, which is absent from not `bad, О `Κ, all `right, `decent. These latter make no reference to other possible evaluations, set up no hierarchy and do not downgrade the mild approval. Note in this connection the unlikelihood of using fall-rise with evaluative adjectives that are clearly at one end of the scale or the other:

12. Halliday, in a most distressing choice of terms, calls these three aspects tonicity (nucleus placement), tonality (boundary placement), and tone (pitch contour selection).

13. Cf. Gage (1958:128): “No source has ever given a satisfactory description of the phonetic nature of the terminals it assumes. The situation with regard to /|/ is particularly bad, since it is often defined no more than it would be by saying: ‘end of a phrase known not to end in either /||/ or /#/.’” This problem is not merely of historical interest, but seems likely to plague Liberman as well once he attempts to transcribe longer stretches of speech than are discussed in his dissertation. Since his tunes contain only a single pitch peak, two adjacent pitch peaks will therefore belong to separate tunes (cf. Ch. 3, Sec. 3 above), and will need to be separated by a boundary with optional boundary tones.

8. Stylized Tones and the Phonology of Intonation

1. A slightly different version of this chapter appears as Ladd (1978).

2. The difference between Liberman’s and Leben’s view and Fox’s view is a specific instance of the tune-tone controversy. For Liberman and Leben, the whole configuration is a ‘tune’. In British terms, on the other hand, the lowpitched stretch that may precede the stepping-down pitches is the head, and the stepping-down pitches themselves constitute the nucleus (plus tail). Fox’s term ‘step-down tone’ expresses the structural parallel between the calling contour and the other nuclear tones. Crystal (1969b), taking only the final pitch into account, denies such a structural parallel, and claims that the true analog to the other nuclear tones is what he calls ‘level’ tone; Fox (1970), however, shows that Crystal’s analysis makes incorrect predictions about nucleus placement. See Crystal (1969b) and Fox (1970) for details.

3. Pike (1945:71): “The same call can be given rapidly, but preserving the chanting character.” Crystal (1969b:34f) makes related observations.

4. This distinction goes unnoticed and unaccounted for in the warning/calling analysis.

5. Dwight Bolinger (personal communication) offers a more general explanation for the shift in this example, namely that stylized intonation would seldom be used to answer questions (i.e., in this case the question What?). This makes sense in terms of the stylized analysis, since a speaker who asks a question signals that the answer is not predictable from the context. As Bolinger points out, stylized intonation in answer to a question has overtones of Tve told you a million times’:

6. The extent to which gradience is involved here was suggested to me by both Dwight Bolinger and Duncan Gardiner (personal communications). Gardiner finds a “hierarchy of melodic forms”, from the least formalized (normal speech intonation) to the most formalized (accompanied singing), with the spiels of talking blues performers, street vendors, auctioneers, etc., on some sort of scale in between. Bolinger also suggests that there must be some connection between music and the steady level pitches of stylized intonation, and speculates that it may have to do with repetition. He makes the striking observation that one feature of poetry, music, and artistic utterance in general that makes it different from ordinary speech is that it is designed to be repeated. Hearing a song or poem for the twentieth time can be a profoundly moving experience; hearing the daily news for even the second or third time is a crashing bore. Bolinger suggests that the constant subtle ups and downs of pitch—which are levelled in both music and stylized intonation—are the “accompaniment of informative utterance.” While the functions of music and stylized intonation are obviously quite different, they both nevertheless convey a sense of familiarity or repetition.

7. Or perhaps more precisely, “level tone stepped up to” and “level tone stepped down to.” This, of course, is reminiscent of the basis on which Bolinger distinguishes В and С accent.

8. In the same way, it seems likely that the only difference between Liberman’s ‘warning/calling tune’ and ‘surprise/redundancy tune’ is that one ends in a plain fall and the other in a stylized fall but that both have the same type of head.

9. Woo (1969) distinguishes two broad traditions of tone language analysis, the Orientalist’ and ‘Africanist’, the main difference being that the Africanists have had to deal with systems making considerable use of level tones, while the Orientalists have been confronted with languages having a great many contour tones. This has, understandably, colored their theoretical outlooks, and Woo’s distinction is very useful. As I suggest below, the autosegmental approach to pitch phenomena seems to have been much influenced by its development in the analysis of Af ricanist tone languages.

10. Writing the stylized tones in Trager-Smith notation is rather difficult, the difficulties arising principally from theoretical restrictions on the use of the juncture /|/, and from descriptive waffling on the acoustic nature of the juncture /#/ when it follows anything other than pitch level /¹/. I asked a half-dozen members of the Cornell linguistics faculty, all well-trained in Trager-Smith notation, how they would transcribe the stylized contours, and got wildly differing responses—all three terminal junctures were suggested at least once.

The Trager-Smith transcriptions here are based on discussions with C. F. Hockett. It is probably true that many linguists would prefer to use the /|/ terminal juncture for stylized contours rather than /#/; Hockett’s use of /|/ (as explained for example in Hockett 1958:37) is somewhat more restricted than in the original Trager-Smith system. But even if we wrote all the stylized tones with /|/, the relation to the plain tones would still not be very obvious.

Liberman’s system is not much clearer. My choice of representation for plain and stylized fall is based on his discussion, pp. 98-104, and is surely the way he would handle the contrast. But the transcriptions of the rising tones must be extrapolated from various examples; possibly he would choose to represent them not as sequences of two ordinary tones, but as sequences of a tone and a ‘boundary tone’. In any case, my observations here would still be applicable.

11. Psychoacoustic experiments discussed by Howie (1976: see especially pp. 222 and 242-243) suggest that Mandarin second and third tones (roughly high-rise and low-rise, though they differ in other respects as well) are more readily confused in isolation than other pairs of tones, and that native speakers rely to some extent on the pitch of adjacent (especially following) syllables for information about the tone on a given syllable.

9. Intonation and Phonesthesia

1. The details vary, and in no case that I know of is a Middle Eastern gesture identical to the European one with the opposite meaning, but in any case up-and-down movements are used for ‘no’ and sideways ones for ‘yes’.

2. Hockett borrows from Hjelmslev the terms piererne and ceneme (based on the Greek roots for ‘full’ and ‘empty’, respectively) for use in his discussion of design features of symbolic systems, mainly because their applicability is not restricted to human language. For language, the pleremes are morphemes and the cenemes phonemes.