1. The Concept of Relations

So far I have emphasized the perceptual aspect of the rhythm hypothesis: the acoustic nature of prominence may be explained in part by assuming that the physical realities are organized cognitively into a rhythmic structure, which in turn affects our perception of the physical realities. A more significant aspect of the idea for linguistic theory, as Liberman and Prince emphasize, is its focus on the concept of relations. Because of the success of segmentation in linguistic analysis, linguists have always attempted to see stress in segmental terms—either as actual segments (Trager-Smith stress ‘phonemes’), or as features of segments (Jakobson and Halle’s prosodic features’, Chomsky-Halle stress levels). The point of the rhythm hypothesis is, in Liberman’s words (1978:169), that “prosodie features are features of structure, rather than features of segments.” The essence of stress and stress patterns is to be found in the relations between segments or constituents, and not in the segments themselves. Older analyses of stress have in effect reified the audible manifestations of these abstract relations, in order to incorporate them into a theory that knows only segments. Liberman and Prince propose instead that linguistic theory must be enriched to include relations among its primes.

This is their fundamental point, and it is one we do not easily assimilate after thinking in terms of segments for so long. Schane, for example, in a paper entitled “The Rhythmic Nature of English Word Stress” (1977), adopts Liberman’s terms ‘weak’ and ‘strong’, but he uses these labels as features of individual syllables, features that can be assigned to a given syllable without regard to the features of any other syllable; the stress level of a given syllable is seen to depend on a bundle of prosodic features (strong or weak, heavy or light, etc.). Schane’s system thus misses the most important insight of Liberman and Prince’s work, which is that ‘weak’ and ‘strong’ cannot be defined on single syllables alone. Segmental features—say, features of vowel quality—are paradigmatic, and a given vowel can be high or front or round independently of other vowels. But ‘weak’ and ‘strong’ have meaning only syntagmatically; it is meaningless to speak of a given constituent as weak, except in relation to a paired constituent that is strong.¹

An analogy may help clarify this point. In the couplet

(1) I think that I shall never see
A poem as lovely as a tree

we say that tree rhymes with see. The rhyme is part of the phonological aspect of the couplet, but it is not a segment, nor is it a feature of any segment or set of segments. We do not account for our feelings about the sound of (1) by positing a segment /R/ at the end of each line, nor do we assume that the vowels of see and tree have (in addition to their other distinctive features) a feature [+rhyme]. Rather, rhyme is a relation between two items, based on certain of their phonological characteristics and their positions in the rhythmic structure of the couplet. We can describe the phonological characteristics of tree which make us say that it rhymes with see, but we cannot, on the basis of those characteristics, say that “tree rhymes.” In the same way, we can describe the phonological and rhythmic features that make us perceive one constituent as weaker than another, but we cannot, on the basis of those characteristics, say that that constituent has a particular stress level. Liberman and Prince’s theory suggests that talking about, say, tertiary stress, independently of context, is exactly as meaningful as asking, “Does tree rhyme?”

It seems likely that the concept of relations will prove rich and productive far beyond Liberman and Prince’s original application of it. The whole question of phonological boundaries, for example, cries out for a relational solution; the neo-Bloomfieldian ‘plus juncture’ is surely a case of reifying the audible manifestations of an abstract relation (cf. Chapter 7 Section 5 below).² Moreover, it is presumably no accident that grammatical relations are currently receiving new attention (in, e.g., Shibatani 1977, Kac 1978, and the various writers on ‘relational grammar’); it seems to me that a unified notion of linguistic relation—both phonological and grammatical—will emerge from this preliminary work. While it is beyond the scope of this book to develop the formalism for dealing with rhythmic relations any further than Liberman and Prince have already done, we can show the applicability of the idea to a specific problem of English intonation. This chapter presents a relational analysis of the widely misunderstood phenomenon of deaccenting.³

2. The Relational Nature of Deaccenting

Deaccenting is seen in the behavior of accent placement in coordinate sentences and sentences in connected discourse. It goes by such names as deaccenting, destressing, anaphoric destressing, deaccentuation, etc., or is lumped under the all but useless cover term ‘contrastive stress’. It is illustrated in the following examples.

(2) A: Has John read Slaughterhouse-Five?
B: No, John doesn’t réad books.

 (3) A bill was sent to Congress today by President Carter which would require peanut butter sandwiches to be served at all government functions. At a press conference today, a group of Senators led by Republican Barry Goldwater of Arizona denóunced the measure. Goldwater said. . .

(4) Harry wants a VW, but his wife would prefer an Américan car.

In each of these examples the deaccented noun has somehow been referred to or alluded to earlier in the discourse. Thus in (2) the discussion is about a book; in (3) the measure is coreferential with a bill . . . earlier in the paragraph (this is a case of what might be called ‘journalistic соreference’, an essentially stylistic device that permits a reporter or newscaster to avoid repetition of a word); in (4), the sentence is about a car.

Observations of this sort are, of course, plentiful in the literature. The phenomenon has been noted, discussed, and alluded to in a variety of contexts by a variety of investigators with a variety of points of view (e.g., Bierwisch 1968, Bolinger 1958a, 1972b, Chafe 1970, 1973, 1974, 1976, Gunter 1966, 1972, 1974b, Halliday 1967b, Hultzén 1956, Schmerling 1976). Most of these writers talk about the deaccenting of ‘repeated’ or ‘presupposed’ material or ‘given (old) information’ or items which are ‘already in the discourse’ or are in some way ‘predictable’. Chafe has developed an approach to these matters based on what the speaker takes to be in the addressee’s consciousness. Schmerling concludes that deaccented items are treated as ‘insignificant’ in some way by the speaker. Gunter (1974b) is an excellent essay pointing out that whatever may be going on here, it is not ‘predictability’. We shall return to this question at the end of Chapter 4.

Since most writers have been concerned with the (broadly) grammatical aspects of deaccenting, its phonological nature has not received much attention. As a result, many discussions of the subject are filled with inconsistencies and unwanted implications. A conspicuous example is Schmerling (1976), who indicates deaccenting by marking /˘/ on the most prominent vowel of the deaccented word. She glosses /˘/ as indieating items “which fail to be assigned stress,” (75) but distinguishes these from syllables with no stress mark at all (“I have here ignored the ‘reduced stress’ frequently found on certain auxiliaries, conjunctions, and so forth, for which the ˘ is perhaps more traditionally employed.” [5]) On the other hand, she also distinguishes /˘/ from three higher levels /ˊˆˋ/, though she notes that she is doubtful about the difference between /˘/ and /ˋ/, and also (citing Vanderslice and Ladefoged) about the difference between /ˋ/ and /ˆ/ when they occur after /ˊ/. This presumably implies that, after /ˊ/, the other three levels /ˆ/, /ˋ/, and /˘/ may all actually be the same thing. For example, if the stress pattern on what she would write as . . . Barry Goldwater denóunced the mĕasure [from our example] and on Barry Goldwater retíres tomòrrow are actually the same, and if what she writes as There’s a cár côming might just as well be written Theres a cár côming, then it follows that the stress sequence /ˊ˘/ could be equivalent to /ˊˆ/. And in fact, it is not hard to accept this conclusion if we compare examples like her Jóhnson diêd with what she might well write as Trûman dièd last mónth, and now just yesterday Jóhnson diĕd. In short, Schmerlings implication that deaccenting is a special level of stress is fraught with inconsistencies, and her /˘/ notation must be interpreted as signifying simply “this syllable is deaccented, whatever that may involve phonologically.”⁴

Similarly untenable is Chafe’s vague conception of deaccenting as a phenomenon of low pitch (e.g., 1974, passim). This covers plenty of obvious cases, but it is far too specific. That is, low pitch does signal deaccenting in, say, John doesn’t réad hooks, but high pitch works just as well, as in the following:

(5) A: John was mad because he got nothing but books for Christmas.
B: Oh, doesn’t John réad books?

The common basis of descriptions like Schmerlings and Chafe’s is fairly easy to see. The most obvious cases of deaccenting are those (like John doesn’t réad books) where the deaccented item occurs in the tail of a falling nuclear tone. The impression of greatly reduced ‘stress’ (Schmerling) and of markedly lower pitch (Chafe) are both simply consequences of postnuclear position. Vanderslice and Ladefoged (1972:828) recognize this, and propose an apparently simple account of deaccenting as actual deletion of the feature [+accent]. In the example John doesn’t réad books, they would see contextual factors changing books from [+accent] to [−accent], which would then automatically, given their rule of intonation center assignment, shift the intonation center back to read. By comparison to John doesn’t read bóoks, the deaccented version of books has a markedly lower level of stress ([+heavy −accent] instead of [+heavy +accent +intonation]); it also, assuming a falling intonation contour, has lower pitch as well. Vanderslice and Ladgefoged see these phonetic characteristics as a consequence of the deletion of the feature [+accent], not, like Chafe and Schmerling, as the phonological essence of deaccenting.

Vanderslice and Ladefoged are certainly on the right track. While the most obvious cases of deaccenting are certainly the postnuclear ones on which Chafe’s and Schmerling’s analyses are based, it is not hard to demonstrate with evidence from accent placement in dialogue that the phenomenon extends to prenuclear deaccenting as well.⁵ For example:

(6) a. i.  [‘out of the blue’, parents accented]
        What time are you meeting your párents?

a. ii. [in context of talking about parents’ visit, parents deaccented]

A: Boy, I really have to get moving. My parents are coming today and I’ve still got to finish this paper and get the place cleaned up and get some laundry done and. . . Wow, I don’t know if I’m going to make it.

B: Oh, yeah, I forgot about that. What time are you méeting your parents?

b. i. [‘out of the blue’, parents accented]

     What time are your párents coming?

b. ii. [in context, parents deaccented]

A: Boy, I really have to get moving. My parents are coming today and I’ve still got to finish this paper and get the place cleaned up and get some laundry done and. . . Wow, I don’t know if I’m going to make it.

B: Oh, yeah, I forgot about that. What time are your parents сóming?

Similarly:

(7)    a. i.     A: What’s the matter?

        B:  I’ve forgotten how to make French Tóast.

a. ii.  A: Why don’t you have some French Toast?

        B: I’ve forgotten how to máke French Toast.

b.i.    A: What’s the matter?

        B: There’s nothing to make French Tóast out of.

b. ii.  A: Why don’t you have some French Toast?

        B: There’s nothing to make French Toast óut of.

In all the ii sentences, parents or French Toast is deaccented, in the sense that the accent placement seen in the i cases is changed. In the a cases, the accent shifts to the left, and the ‘level of stress’ on parents or French Toast accordingly appears quite reduced. This is the typical case of deaccenting. But in the b cases the accent shifts to the right and the ‘stress’ on parents or French Toast would normally be written /ˆ/ or level 2—that is, in traditional terms, the stress is reduced very little. The same kinds of contextual factors are present in both cases; the only difference is the relative position of the deaccented item and the accent. But Schmerling’s /ˇ/ notation or Chafe’s ‘low pitch’ make it impossible to relate these cases to postnuclear deaccenting.

Unfortunately, because of their narrow view of sentence stress discussed in the previous chapter, Vanderslice and Ladefoged cannot account simply for these cases either. Their analyses of the ‘normal’ (nondeaccented) versions of When are your párents arrivingΡ and There’s nothing to make French Tóast out of would presumably be:

Deletion of the [+accent] on parents or Toast should, according to their intonation center assignment rule, shift the intonation center to the left—as in the case of John doesn’t réad books. The rightward shift that actually occurs in these cases involves, in Vanderslice and Ladefoged’s terms, a reaccenting of arriving and out of. Thus the Vanderslice-Ladefoged framework has no way of treating deaccenting as a simple phenomenon after all: sometimes it appears as accent deletion on the deaccented syllable, and sometimes as accent insertion elsewhere.

The relational nature of the rhythm hypothesis, however, provides us with a simple way of viewing deaccenting as a unified phenomenon (cf. Liberman’s comments pp. 167ff). Roughly speaking, what postnuclear and prenuclear deaccented syllables have in common is the presence of the nucleus elsewhere. Put this way, the observation seems almost tautological, but it is nevertheless the key to the problem: the deaccenting of a syllable can best be understood as a relative weakening of its hierarchical rhythmic position. This is seen in the following diagrams:

This relative weakening will naturally have different phonetic effects depending on how the resulting rhythmic structure matches up with the intonation contour. In both of these examples, the deaccenting involves switching a single pair of node-labels from weak-strong to strong-weak, or vice-versa; but in John doesn’t réad books, books falls after the nucleus, i.e., in a very weak position, while in When are your parents arríving, parents precedes the nucleus and keeps ‘secondary stress’. Deaccenting is perceived in both cases by inferring a rhythmic structure in which the deaccented item is weaker than it would be if it were not deaccented. It does not depend on anything as straightforward as ‘failure to be assigned stress’ or ‘low pitch’.

Failure to recognize this has led at times to the postulation of special sentence-types to account for deaccenting. Bresnan (1972) proposed a distinction between what she called ‘initiatory’ and ‘elicitory’ questions, in order to account for the different accent placement in the following:

This is identical to the accent shift in When are your párents arriving? and When are your parents arríving? just discussed. Bolinger (1972b) was quick to point out that ‘elicitory questions’ are merely an instance of deaccenting:

Similarly, the sentences that Schmerling (1976:89-98) discusses as ‘topic/comment’ sentences are also deaccenting in disguise. Compare:

(13) [‘out of the blue’, parents accented]
My párents called.

(14) [in context, parents deaccented]
A: Maybe we ought to call your parents and tell them.
B: My parents cálled--they already know.

(15) [‘out of the blue’, gray hair accented]
I had a gray háir fall out.

(16) [Schmerling’s ‘topic/comment’]
I had a gray hair fall óut.

Schmerling writes the stress marks on (16) as J had a gray hâir fall óut, and thus does not equate it with deaccenting, but she points out that it implies the speaker has numerous gray hairs (95)—that gray hair is somehow ‘in the context’. Indeed, Schmerling points out that the subjects of ‘topic-comment’ sentences are in some sense ‘old information, and makes many sensitive observations about the special assumptions involved in such sentences—observations that could be set almost verbatim into a discussion of deaccenting. If we do not assume that deaccenting implies a special ‘level of stress’ but only a relative weakening, we need posit no special sentence-type here; the more general rubric of deaccenting does the job.

In Chapter 4 I will take the relational nature of deaccenting for granted, and return to discuss in more detail the grammar of accenting and deaccenting. The remainder of this chapter is devoted to a detailed exegesis of Bolinger’s work on the discourse functions of Accents A and B, showing how those discourse functions are also related to deaccenting. This discussion, in addition to illustrating yet another disguise that deaccenting may adopt, will provide still more support for the rejection of Bolinger’s accent taxonomy already proposed in Chapters 1 and 2.

3. Pretonic Accent and Deaccenting

Bolinger observes that in phrases and short sentences, where there are often two accent peaks, the accent on the first peak tends to be A (falling) if the accented word is contrastive or important or new to the discourse, and В (step-up-to or rising) if the accented word merely qualifies the following Α-accented word, or if it is not new to the discourse. He has numerous examples of this in a number of articles (especially 1958a, 1957a, 1965b).

Again, in (1957a:62), he notes that:

Dozens of similar examples are given in Bolinger (1965b, Sec. V), including ‘quantifying and degree modifiers’:

‘enhancing modifiers’:

and more ‘intermediate compounds’ (here he notes that these are “more or less stereotyped combinations”):

With regard to the latter example he observes that

A final type where BA is prevalent is

It should be emphasized that Bolinger is not claiming a one-to-one grammatical correlation between certain types of constructions and the BA accent pattern. He goes on to say that

Now clearly Bolinger has identified a general phenomenon here, which any suprasegmental analysis must be able to account for. Moreover, his ‘accent’ framework seems to make it easy to equate sentence contours which are superficially quite different. Thus

are both BA despite the different pitch levels involved. But there is a quirk in Bolinger’s taxonomy that has made it possible for him to point out this phenomenon without observing that it goes beyond the cases he has described.

The quirk in his taxonomy has to do with Accent B. Recall that Bolinger (1958a) posits two separate shapes for the B, the common characteristic being ‘upmotion’; let us number these two for easy reference:

Now, all the BA examples we have been discussing so far have involved only B1. Moreover, the semantic contrast that Bolinger identifies between В (i.e., B1) and A is paralleled by the same contrast between B1 and B2. That is, the extra punch or emphasis added to really by changing

is paralleled exactly by changing

The distinction that Bolinger’s taxonomy allows him to write simply as BA vs. AA is exactly paralleled by the distinction between B1-B2 and B2-B2, but Bolinger’s three-accent system leaves us with no ready way to write this one. So there seems to be a good case for revising the taxonomy to make B1 and B2 separate entities.

Moreover, the same sort of contrasts that support dividing В accent into B1 and B2 can be found with С accents, and we may divide Bolinger’s С into C1 and C2. For example, really can have punch or emphasis added to it in the same way as in the examples we have already discussed:

Now, B1 and C1 seem to have a number of things in common which set them apart from other accents, including B2 and C2. First, they obviously share the semantic function of deemphasizing the prominent syllable to which they apply, and it is often nearly impossible to detect meaning differences between them. This is certainly not true of B2 and C2, which contrast sharply, for example, as answers to WH-questions:

With a B2 the answer might mean “have you heard of it?” or “is that good enough?”,⁶ whereas with a C2 we may get a meaning like “what are you going to make of it?” or “can’t fault me there.” But the difference between B1 and C1 in analogous contexts is often impossibly subtle and sometimes even hard to hear:

Second, B1 and C1 are defined solely in terms of pitch movement to the accented syllable, up or down respectively. This contrasts with A and B2, which are explicitly defined in terms of pitch movement during or after the accented syllable. (C2, I believe, is also to be defined this way, though this is not so clear from Bolinger’s definitions. I have equated C2 with the British low-rising tone, and thus consider the rise to be one of its distinguishing characteristics. Bolinger does note [1958a:50] that in C-accent “a further fall seems to be avoided.”)

Third, all of Bolinger’s examples of the contrast between В (i.e., B1) and A involve contrasts on accented syllables which precede other accented syllables. B1 does not seem to occur as the final accent of a sentence—the one which in more conventional accounts would often be identified as the sentence stress or nucleus—but only in earlier positions. This is true of C1 as well.

From all of this it is reasonable to conclude that B1 and C1 must be lumped together as ‘pretonic’ accent, a jump up or down to a prominent syllable in a pretonic contour—that is, a rhythmically strong syllable in the head. Pretonic accent signals a prominence that is somehow subordinate to other accents, as shown by its use in signalling discourse relationships and emphasis. The distribution of B1 and C1 is determined, I think, not by any semantic difference between В and С (as Bolinger’s taxonomy would imply), but simply by the shape of the head of which the accented syllable is a part. For example, in the two versions of Marie Smith’s brother told me in (35), the tunes are grossly

The fact that we have a B1-C1(-C2) sequence in one and a B1-B1 (-C2) sequence in the other should be attributed, I believe, to the occurrence of a single phenomenon of pretonic accent in different overall pitch contours.

The general range of labels—’expected, stereotyped, commonplace’—which Bolinger suggests for the meaning of ‘pretonic’ accent (i.e., his ‘B’ accent in the passages quoted above) is strikingly like the meanings that have been attached to deaccenting. There is good reason to believe that deaccenting and pretonic accent are somehow the same phenomenon.

For example, it is well known that epithets are usually deaccented:

(36) A: How did your operation go?
B: Don’t talk to me about it--I’d like to strángle the butcher.

The deaccenting of butcher signals that it is an epithet, which is intended to refer to the doctor whose presence in the context is somehow understood. If butcher is accented, it seems to refer to a literal butcher, and B’s reply sounds like incoherent psychopathic raving:

(37) A: How did your operation go?
B: Don’t talk to me about it--I’d like to strangle the bútcher.

Similar observations could be made for bastard, cat, and various others whose literal meanings are quite different from their meanings as epithets. Thus, in What’s the mátter with that cat? the referent of cat is readily taken to be human, given an appropriate context, while in What’s the matter with that cát? it is likely only to be feline.

The interaction of pretonic accent and epithets is identical. Consider the following dialogue:

(38) A: Everything OK after your operation?
B: Don’t talk to me about it! The butcher charged me a thousand bucks!

With a pretonic accent (B1) on butcher, this makes perfect sense:

But with an A accent we have a literal butcher oikc more, and Incoberence:

Or again with cat:

(41) He wasn’t a damned valet and lie wasn’t a damned bodyguard. He was a working road manager. [He] took care of money, airplane reservation, getting the band to where they were supposed to go. The cat worked hts ass off [Rolling Stone no. 221, 9 Sept. 1976, PP. 13-14]

The last sentence must be

not

‘journalistic coreference,’ mentioned earlier, is another device tiiat works equally well with either deaccenting or pretonic accent, as can be seen from the following examples:

Notice the sequence of A accents on A notorious . . . , while Quang Phuc Dong has pretonic accents on all but the last accented syllable. Compare the version with postnuclear deaccenting:

(45) A notorious Communist propagandist and scatolinguist was arrested today after being picked up on a rubber raft in Lake Ontario; the Coast Guard captain who rescued Quang Phuc Dong said. . .

The purpose of ‘journalistic coreference’ as it is used in broadcast news seems to be to add information about the subject of the news item without resorting to a separate sentence exclusively for background information. The first half of the sentence is presented as ‘news’—with a succession of A accents—while the additional information in the second half is ‘deemphasized’—with either pretonic accent or deaccenting—as if to say “This is nothing new, it’s the same guy we were just talking about.”

Finally, there are dozens of adverbs and similar items which most often occur deaccented at the end of the sentence, but which take pretonic accent when the sentence is rearranged:

In short, in numerous types of sentences we see the parallel use of deaccenting and Bolinger’s pretonic accent, which justifies equating them.

Indeed, some of Bolinger’s pretonic accent examples can readily be explained in terms of the analysis of deaccenting already presented earlier in the chapter. For example:

with a pretonic accent on bomb, will be seen as representing

in which bomb is relatively weaker than in the ‘normal’ rhythmic structure

Deaccented bomb has a pretonic accent because of the pitch jump on the first long syllable of the sentence (see Chapter 2, Section 1), but the rhythmic phenomenon involved is exactly like other cases of deaccenting.

Viewing deaccenting as relative rhythmic weaking thus makes it possible to subsume a number of rather different phonetic effects under the same functional rubric—effects, like Chafe’s ‘low pitch’ and Schmerling’s ‘absence of stress’ and Bolinger’s ‘B accent’, which have been observed and described piecemeal for many years. As Liberman suggests (1978:167), such a view also makes it possible to give a coherent account of the function of accent, with deaccenting and contrastive or focus accenting seen as opposite sides of the same coin. This is the task undertaken in the next chapter. Before moving on, however, we should mention some residual problems posed by certain cases of pretonic accent.

Pretonic accents like the one in A bomb had wrecked it, which cause a shift in the location of the nucleus, are easily handled by the relational view of deaccenting, but others, like the distinction between pretonic accent and nuclear A accent on really in It’s really good, are more difficult. They seem to be more purely intonational, rather than rhythmic, and it is not at all clear what sort of different rhythmic structures might underlie the pretonic vs. nuclear distinction in these cases.⁷

There are two principal alternatives, both involving formal modifications already discussed by Liberman (1978). The first is that the distinction might reflect different metrical bracketings; these are discussed by Liberman (1978:110ff and 170ff) in connection with the behavior of clitics and other rhythmically weak syllables. For example, we might assume that the rhythmic structure

would be associated with pretonic accent on really, while the special prominence conveyed by the extra nuclear Α-accent on really would reflect a rhythmic structure

But we have no real way of telling what might constitute motivation for associating a particular rhythmic structure with a particular intonation, especially in longer sentences like The butcher charged me a thousand bucks, where the number of conceivable rebracketings is enormous. The line is thinly drawn between insight into the workings of language and the clever manipulation of formalisms.

However, a second, and, I think, more promising possibility for explaining the use of pretonic accent in terms of rhythmic structures is to assume that each nuclear tone is associated with a coordinate constituent (T) of a rhythmic structure. Thus:

(butcher deaccented, with pretonic accent, used as epithet)

There are at least two reasons why this proposal is worth considering.

First, it seems fairly clear that we will in any case have to account for what Liberman (1978:119ff) calls complex tonal patterns and what Crystal (1969a:244-252) treats as subordinate tone units by means of rhythmic structures which are more complex than the ones we have illustrated so far. Liberman, for example, discusses intonational tags—like the vocative in Sam struck out, my friend—dividing the rhythmic tree first into ‘Stem’ and ‘Affix’. Thus no expansion of the formalism would be involved in dividing the tree into coordinate constituents.

More importantly, such a formal device emphasizes the purely intonational (as opposed to rhythmic) nature of the contrast. We will in any case have to allow for utterances with more than one nucleus (e.g., the ‘parallel’ examples in Chapter 4 Section 2c, or the sentences with a fall and a fall-rise treated in Chapter 7). My proposal suggests that each Τ node defines the domain of an intonational tune of head + nuclear tone. That is, the butcher, in the sentence where it has an Α-accent and its literal meaning, would not be in the head at all, but in a separate tune of its own with the nucleus on butch-. Obviously, work remains to be done on this problem.