1. The Role of Intonation: Two Approaches

It is time to treat in detail the question of how intonation affects the meaning of sentences. There are two general approaches to this problem, which we might label ‘syntactic’ and ‘lexical’. The ‘syntactic’ approach involves the assumption that the intonation of an utterance depends in some way on certain grammatical facts about the utterance—everyday notions such as ‘question intonation’, ‘declarative intonation’, and ‘sentence-final intonation’ belong to this view. However, it is almost impossible to find consistent associations between grammar and intonation—as we saw above (Chapter 5, n. 1), generalizations like “yes-no questions have rising intonation” are simply not supported by data. As a result, recent syntactic characterizations of intonation’s role, like Lieberman (1967), have tended to be in terms of fairly abstract deep structures. Lieberman’s view is that the ‘linguistic’ uses of intonation reflect certain configurations of deep structure syntactic markers and/or phrase boundaries; similar ideas surface in later generative work like Downing (1970), Pope (1972), Jackendoff (1972), Hirst (1974), Hirst and Ginésy (1974). One obvious problem with this view, as we saw in Chapter 5, is that it assumes a sharp distinction between grammatical and affective uses of intonation which is difficult to justify.

The ‘lexical’ approaches to the role of intonation all assume in one way or another that intonation contours have semantic force independent of the grammar of the sentence of which they are a part. The oldest and most widespread version of this general point of view is Pike’s idea that the special function of intonation is the expression of speaker’s attitude, which we have already discussed in detail. A second, somewhat idiosyncratic formulation is that of Gunter (1972), who sees intonation as signalling the relevance of utterances to their contexts; as we saw, he explicitly separates ‘grammatical’ from ‘expressive’ uses of intonation—thereby pre sumably excluding speakers attitude from relevance’—but he also explicitly argues against a sentence-grammar approach like Lieberman’s. Finally, somewhere in between Gunter and Pike is the recent, rather more stylish idea that intonation has a specific ‘pragmatic’ function, such as indicating illocutionary force, or signalling the violation, suspension, etc., of Gricean conversational postulates (e.g., Liberman and Sag 1974, Sag and Liberman 1975, Gibbon 1976b, Glenn 1977.)

Notice that both these general approaches may involve the assumption that the task of the investigator is to discover “the” function of intonation. This idea is simply around-the-edge-of-language in a new guise, especially insofar as the roles proposed for intonation are often quite idiosyncratic. For example, one thing left unexplained in Lieberman’s approach, as Gunter (1976) points out, is why a heterogeneous collection of deep structure characteristics like phrase boundaries and question markers should be singled out for a special mode of surface realization. Similarly, while there are a great many grammatical and lexical ways of connecting sentences to their context—pronominalization, for instance—we do not find discussions of “the” function of pronominalization; why then does Gunter feel compelled to identify a single such function for intonation, especially when it involves leaving so much obvious data (the ‘expressive’ uses) unaccounted for?

The approach I take here is a lexical one, similar in most respects to that of M. Liberman (1978). I will argue in the next section that such a view makes no unnecessary edge-of-language assumptions and is not, as some have suggested, a priori unreasonable. In Sections 3 and 4 I present a detailed analysis of the use of the fall-rise tone in English, showing how we can account for both ‘grammatical’ and ‘affective’ aspects of its use by positing a single meaning common to all occurrences. In Chapter 8, I further illustrate my lexical approach with an analysis of the ‘calling contour’ of English.

2. The Intonational Lexicon

Liberman (1978) proposes that intonation is organized into a ‘lexicon’ of ‘intonational words’. He compares the mode of meaning of these words to ‘ideophonic’ effects in language, especially as seen in languages (like Korean and Bahnar, which he cites) which have recognizable ideophonic components to their segmental lexicon. He notes (94) that

He then suggests that English intonational meaning should also be viewed in just this way. That is, the elusive effects of intonation can be explained by assuming that each contrasting contour has an abstract context-free element of meaning, which produces specific nuances in specific contexts.

Liberman’s view has two important elements:

1. Intonation is not assigned to utterances by rules of syntax, but represents an independent lexical choice from an intonational lexicon; in context, the general meaning of the intonation contour interacts with the other meanings in the sentence to produce specific nuances.

2. The mode of meaning of intonation is like that of ‘ideophonic’ segmental words.

For Liberman, of course, the ‘ideophonic’ nature of intonation is manifested in the way the general meaning interacts with other meanings to produce specific nuances; thus from his point of view I have separated two different aspects of a single phenomenon. I have done so deliberately, however; as discussed in Chapter 9,1 think the ideophonic or phonesthetic nature of intonation is attributable to something very different. On what is to me the separate issue of how intonation interacts with grammar, I think that Liberman’s hypothesis is potentially very productive, and in its broad outlines correct. The analyses of the fall-rise and calling contours to be presented below show, if anything, that Liberman has not taken his hypothesis far enough.

Certainly the notions of ‘intonational lexicon’ and ‘intonational word’ raise a lot of eyebrows, but it seems to me that this is due largely to the automatic assumption that intonation is around the edge of language. In this section the reader is asked to suspend that automatic assumption long enough to consider the merits of Liberman’s proposal and of the criticisms that have been directed at it. For ‘intonational lexicon’, as we saw in Chapter 1 Section 4, need imply no more than an inventory of meaningful contours: as such, a notion of lexicon is implicit in much British work, and explicit in such American works as Pike (1945) and Gage (1958). We must, after all, concede that in some sense intonation contours are ‘meaningful’: Pike’s observation that “a change of pitch contour will change the meaning of the sentence” (20) is obviously correct. Even Liberman’s statement of the way an abstract general meaning underlies the specific nuances of a contour in context is not especially new: observing the range of usages of a particular word and trying to figure out what they all have in common is, after all, a time־honored method of dealing with intonational meaning or any other unfamiliar semantic territory.

Nevertheless, the idea of context-free intonational meanings is hard for many linguists to accept. Gunter, as we saw, finds the task of assigning a meaning to “an intonation in the abstract, divorced from words,” to be “baffling.” Cutler (1977:106) likewise maintains that “effects exercised by intonation contours can be shown to be context-dependent to such a degree that the attempt to extract from them an element of commonality valid in all contexts must be reckoned a futile endeavour.” It may well be that this counsel of despair is well founded, but it seems to me that there are two issues here which have not been kept entirely separate—largely because of the edge-of-language notion. The first is whether the positing of abstract general meanings is a reasonable way to account for any context-dependent meaning, intonational or otherwise; the second is the extent to which intonational meaning is like segmental meaning.

The first question—often under the heading pragmatics’—is of course a matter of considerable interest nowadays, and one widely adopted approach has been to explain contextual effects with reference to a contextfree meaning. At the level of whole sentences, this gives us the notion of illocutionary force. If I say You son of a bitch to a good friend it generally conveys a very different attitude than if I say it to a stranger; while there can be no doubt that in some loose sense you son of a bitch ‘means’ different things in these two contexts, it is not outlandish or unusual to account for this in terms of the effects of different contexts on ‘the same’ sentence.

Below the level of the sentence, it would generally be agreed that clearly referential lexical items like nouns have an element of context-free meaning, and are given more specific interpretations in context—such as the distinction between sitting on a chair and sitting in a chair cited earlier. Far more controversial is the proposal that we account for grammatical meaning like case and tense in the same way: this is the approach of the ‘form-content’ and ‘Jakobsonian’ schools (e.g., Garcia 1975, Waugh 1976, Reid 1976). Somewhere in between the extremes of referential and grammatical meaning lies the sort of meaning most often labelled pragmatic’—the meaning of syntactic and lexical devices that serve to anchor utterances in their discourse context and to signal things about the speaker’s presuppositions and his relationship to his hearer and to society. Here a variety of linguists have taken tentative steps in the direction of talking about context-dependence in terms of abstract context-free meanings, some more tentative than others. Gary (1976), for example, describes the effect of fronting prepositional and participial phrases in terms of a common meaning of ‘contrary to expectations’, while R. Lakoff (1972) proposes that a variety of effects of the English modal must can be explained with reference to a variety of culture-bound contextual changes rung on a single semantic common denominator ‘obligation’. Lakoff mentions particles and honorifics as other linguistic phenomena that might be explained in such a way.

It seems to me, then, that unfavorable reaction to the notion of context-free intonational meaning must be seen as part of the larger debate over how to account for context-dependence in general. Specifically, I would contend that the problems of accounting for intonational meaning are in this respect no different from the large number of other problems of ‘pragmatics’—in fact, Cutler contends that the interpretation of intonation contours is to be treated in exactly that way. And it should be clear that whether or not intonational interpretation is ‘pragmatic’ has nothing whatever to do with the question of an intonational lexicon. Elements like German doch or English the or Japanese wa certainly have pragmatic’ effects, and we may suspect that Cutler would find the attempt to extract from their varied contextual effects an element of commonality valid in all contexts to be a futile endeavor. But certainly she would also include them in the lexicon of their respective languages.

More informally, we might say that just because you assume the existence of words doesn’t mean you have to assume that the words have an abstract general meaning. As we just saw, the question of context-free meanings is unresolved for many types of segmental lexical items as well. Cutler’s contention that the context-dependence of intonation cannot be explained in terms of abstract general meanings has no direct bearing on the more fundamental issue of intonational lexicon. If, as I argued in Chapter 6, there is nothing peculiarly intonational about the semantic work that intonation does, then we should actually expect it to be organized into some kind of lexicon—just as segmental meaning is. To start with any other assumption is to guarantee that intonational meaning will appear mysterious—the self-confirming edge-of-language idea again.

But this brings us to the most basic question of all:is intonational meaning different from other meaning? Pike (who, incidentally, was a strong proponent of the notion of abstract context-free intonational meanings) put forth the view that intonational meaning is a phenomenon unto itself (1945:21):¹

Liberman takes an important step past Pike’s position when he makes an explicit comparison between intonational meaning and the meaning of ‘ideophonic’ segmental words. To my knowledge, this is the first suggestion that intonational meaning need not be seen as entirely sui generis, and to my mind it is this suggestion—not the notion of abstract meanings—which constitutes the truly radical nature of Liberman’s ‘intonational lexicon’. Nevertheless, Liberman still clearly sets ‘ideophonic’ (including intonational) meaning apart, as quite different from ordinary ‘morphemic’ meaning. That is, neither writers like Pike and Liberman, nor critics of the abstract meaning approach, have ever really considered what seems to me to be the simplest hypothesis: that intonational meaning is like segmental meaning.

Here again the edge-of-language notion has been self-confirming. This is illustrated particularly well in Cutler’s reaction to Liberman’s views:

If those are indeed the implications for a psycholinguist, then Cutler may well be right to reject the idea of intonational lexicon. But why does she insist, for example, that contours be processed “independently of the interpretation of the text on which they are imposed’? Why must we assume that the meaning of the contour is “combined” with the meaning of the text? I question whether hypotheses about the meaning of any lexical item could meet these standards.

Suppose, for example, that we hypothesized that the use of the Japanese particles wa, ga, ni, etc. could best be accounted for by positing for them abstract context-free meanings which produce specific grammatical and semantic effects in specific contexts. Would Cutler then require us to produce a model of sentence comprehension performance in Japanese in which the meanings of the particles are processed independently of the text in which they are included, and then combined with the meaning of the text? It seems very unlikely that anyone could successfully design such a model—or that anyone would ever try. But I do not see how such a model is the inevitable implication of the abstract-meaning hypothesis. Rather, Cutlers inference that intonation would have to be processed separately and then meshed with the rest of the text proceeds not from the concept of intonational lexicon, as she claims, but from the unexamined assumption that intonation is around the edge of language.

Again, consider the analogy to lexical tone. We do not assume that speakers of tone languages process the tone and the segmentals independently and then combine them to arrive at the correct identification of the lexical item at hand; experiments on the perception of lexical tone are assumed to bear on tone discrimination. The psycholinguist dealing with intonation in the framework of an intonational lexicon would likewise interpret experimental data simply as evidence of strategies for contour discrimination, not contour interpretation. Interpretation of the sentence is in terms of all the lexical items, both segmental and intonational. The intonational lexicon hypothesis gives us no reason to assume—indeed, it gives us reason not to assume—that interpretation of the sentence proceeds in two separate stages.²

3. Preliminaries to the Analysis of Fall-Rise
a. Two Approaches to the Problem

For a clear illustration of the notion that intonation is not merely specified by the syntax but makes an independent lexical contribution to the meaning of the sentence, we will turn to the fall-rise tone.³ There is a rather extensive literature on falling-rising contours in English (e.g., Hultzén 1959, Jackendoff 1972, Lee 19560, 1956b, Liberman and Sag 1974, Schubiger 1956, 1958, Sharp 1958); part of the reason for this is that such contours have many striking ‘grammatical’ effects. Thus they provide a good proving ground for hypotheses about how intonation works.

Among the best-known effects of intonation on grammar are the distinctions in scope of negation in pairs of sentences like:

In the fall-rise sentences, the negation applies to the focus constituent (All and because hes unhappy); the sentences tend to be interpreted as ‘Not all the men went’ and ‘It is not because he’s unhappy that John drinks’. In the fall sentences, on the other hand, the negation applies to the verb, and we get ‘None of the men went’, and ‘It is because he’s unhappy that John doesn’t drink’. The grammatical effects (on the two versions of All the men didn’t go) have most recently been discussed by Jackendoff (1972) and Liberman and Sag (1974), and their treatments are a good point of departure.

Jackendoff exemplifies the ‘syntactic’ approach to intonational meaning. Specifically, Jackendoff claims that the fall-rise reflects an underlying two-variable focus presupposition.⁴ In sentences with a single nucleus, and with falling intonation, Jackendoff formalizes the focus presupposition using a single-variable lambda set: the logical structure of John writes poetry in the garden is roughly:

Presupposition: ‘λ x [x writes poetry in the garden] is well-formed’, or, more crudely, ‘there is somebody who writes poetry in the garden’.

Assertion: ‘x = John’ (‘John is the one’).

In sentences with two nuclei, Jackendoff extends this logical system to allow for two-variable lambda sets. The constituent with the fall-rise (‘B accent’) is the independent variable (x), and that with the fall (‘A accent’) is the dependent variable (y). Thus, in

the logical structure of B’s reply would be

Presupposition: ‘λ x, y [y ate x] is well-formed’;

Assertion: χ = the beans, y = Fred.

Compare:

Presupposition: ‘λ χ, y [χ ate y] is well-formed’;

Assertion: x = Fred, y = the beans.

However, this leaves Jackendoff with the implication that there must always be a second focus-variable in sentences with a fall-rise, which immediately causes problems in sentences with a single fall-rise nucleus, like (1). What is the dependent variable in such cases? Jackendoff seeks to rescue his analysis by arguing that the fall-rise still represents an independent variable, and that the dependent variable is what he calls the “affirmation-negation distinction.” Jackendoff’s justification for introducing this variable ex machina is at best unconvincing, however; Liberman and Sag dismiss this solution as “an ad-hoc circumvention of self-created problems” (419). The reader is referred to Jackendoff (1972:258-265) for further details.

The analysis put forth by Liberman and Sag differs from Jackendoff’s in two important respects. First, they identify the intonation of (1) as an instance of what they call the ‘contradiction contour’; they illustrate this with their example:

which they punctuate Spanish-style, with fore-and-aft exclamation points, thus:¡Elephantiasis isn’t incurable! Second, their general approach to intonational meaning exemplifies Liberman’s ‘intonational lexicon’ idea: they describe the contradiction contour as a holistic contour whose general meaning is of contradiction or questioning the addressee’s assumptions, and they explain the effects of this contour on All the men didn’t go as follows (422-423):

While I fully agree with Liberman and Sag’s rejection of JackendofFs logic-based approach to intonational meaning—we will return to this question presently—I would argue that their taxonomy of falling-rising contours is seriously in error. Jackendoff’s analysis presupposes a ‘lexical segmentation in which the falling-rising contour in (1) is identified with the falling-rising contour on beans in (5). Jackendoff, as we saw, calls both В accent; in our terms, both are fall-rise nuclear tone. Liberman and Sag, on the other hand, while not denying the existence of a contour like the fall-rise nuclear tone, identify the falling-rising intonation in (1) not with the fall-rise, but with the contradiction contour. They conclude, in effect, that Jackendoff not only came up with an inadequate solution, but didn’t even pose the problems properly.

As I suggested in Chapter 1, there is little sense in trying to explain linguistic phenomena before we agree on what the phenomena are. Jackendoff’s taxonomy is compatible in all significant respects with the British-style analysis adopted in Chapter 1 (see Figure 5). However, since Liberman and Sag’s is based on computerized pitch tracings, we should consider their critique seriously before we proceed with our discussion of the semantics of the fall-rise. But it will become clear that Liberman and Sag have proved their own point that “it’s difficult not to be misled by similarities in pitch contour, apparent or actual” (424), and that Jackendoff’s posing of the problems is essentially correct.

b. Taxonomy of Falling-Rising Contours

Liberman and Sag’s principal reason for their grouping is that in the ‘contradiction contour’ the terminal rise is distinctly heard, whereas in (5) it may not be. Since the terminal rise in (1)is distinct, they argue that we should equate this with the contradiction contour rather than (5). But they themselves (425) discuss acoustic evidence that with emphasis on a final word there is a “concomitant reduction in the terminal rise”; that is, terminal rises immediately preceded by falls may be reduced. What this means for the fall-rise tone is that the closer its nucleus is to the end of the sentence, the less distinct its terminal rise will be. In (1) the nucleus is early in the sentence, and the terminal rise is not reduced. In (5), it is at the very end of the sentence, and the terminal rise may be indistinct.

To be sure, this is no argument that Liberman and Sag are wrong, only that they cannot base their argument on the distinctness of the terminal rise alone. We have shown that it is possible to maintain our analysis in terms of fall-rise tone, but we need more evidence to show that it is necessary. Such evidence emerges from their discussion. They handle what we are calling a fall-rise tone early in the sentence by saying that “the contradiction contour can coexist with contrast.” Their examples 16a and 16b, in their notation, are as follows:

They describe these two as follows:

Now as we saw in Chapter 4, ‘contrastive stress’ consists primarily in 10-eating the nucleus or sentence accent in a ‘non-neutral’ or ‘narrow focus’ position. In 16a, the position of the nucleus signals a narrow focus on I and the tail of the falling-rising tone covers the whole rest of the sentence. In 16b, on the other hand, the nucleus is on school— which is the ‘neutral’ or ‘broad focus’ placement—and the part preceding it is the head. The contradiction contour is thus not a fall-rise at all, but a low-rise with a high-falling head.⁵

The significance of this becomes clearer when a polysyllabic word is involved. Compare:

In (8) there is a low-rise nucleus on incurable, and the pitch in the head rises and falls without regard to word-stress in elephantiasis. But in (9) there is a fall-rise tone with its nucleus on elephantiasis, and the pitch prominence is located on the stressed syllable -ti-.

The semantic difference becomes clear in these examples, too. In (8) we do have something like a holistic ‘contradiction’ or questioning of speaker A’s assumptions (‘Whaddya mean you’re going to die—elephantiasis isn’t incurable.’) On the other hand (9) narrowly focuses on elephantiasis, and says that it, at least, is not incurable. (Since the discourse seems only to provide a choice between elephantiasis and rabies, the implication is hard to escape that rabies is incurable.) Switching responses in these contexts, we get a change:

In (10) the speaker is again focusing on elephantiasis, and the force of the reply is something like Ί know that elephantiasis, for one, is not incurable, though Гт not saying you don’t actually have something else that is.’ (This response sounds odd to some informants, but given the proper speaker assumptions it would be perfectly normal: suppose В knows A has been having bizarre symptoms and has gone to the doctor to find out what is ailing him.) In (11), on the other hand, we could have a comedy routine, for there is a mismatch between segmentals and intonation. The intonation implies that the reply is a relevant contradiction, yet the segmentals contradict only half of A’s sad report, leaving us to wonder just what В thinks of the other half. From the TV show “All in the Family”:

The difference between the two intonations also emerges in a simple syntactic test. As Liberman and Sag note, the contradiction contour is unembeddable:

But the fall-rise tone can be embedded freely:

If the fall-rise tone were simply contrast overlaid on the contradiction contour, then (14) should be impossible too. The behavior of the fall-rise and contradiction contour relative to the segmental syntax is thus incorrectly predicted by the Liberman-Sag taxonomy, whereas in terms of the nuclear-tone analysis it can be simply explained. The ‘beginning’ of the fall-rise contour—the high-pitched part—depends on the location of the nucleus, and can thus go anywhere the nucleus can go. The ‘beginning’ of the contradiction contour, on the other hand, is the beginning of the head; since the head is defined roughly as the pitch contour on that part of the sentence that precedes the nucleus, then almost by definition we will not expect to find it starting anywhere but at the beginning of the sentence.⁶

Finally, there are cases of actual ambiguity between the fall-rise and the contradiction contour, a sure clue that we are dealing with two linguistically separate categories. This is seen in an example we used in Chapter 2:

This can be interpreted either as the contradiction contour (‘John’s not in BOSTON—what are you talking about; he’s right in the next room watching the tube.’) or as the fall-rise with nucleus on John (‘JOHN’s not in Boston—it was Henry’s turn to go this time.’) Since the nucleus of the fall-rise is on the monosyllable John rather than on a long polysyllabic word like elephantiasis, the distinction between the two contours may be neutralized, which helps explain Liberman and Sag’s confusion of the two. But the semantic distinction remains sharp despite the phonetic identity.

All the evidence just presented shows that the contradiction contour is to be kept separate from the fall-rise tone with nucleus early in the sentence. This very point has been treated before in the literature, and my position seems to be supported by Lee (1956b) and Schubiger (1958: 26n). The remainder of this chapter assumes this distinction, and makes no further mention of the contradiction contour.⁷

4. A Semantic Analysis of Fall-Rise
a. Fall-Rise in Single-Nucleus Sentences

The use of the fall-rise tone is exemplified in the following dialogues:⁸

(16)A: Did you feed the animals?
B: I fed the ˇcat.

(17) A: What would you think of getting a dog?
B: A ˇcat, maybe.

(18) A: You have a VW, don’t you?
B: I’ve got an ˇOpel.

(19) A: Do you want a glass of water?
B: I’ll have a ˇbeer.

(20) A: Harry’s the biggest fool in the State of New York.
B: In ˇIthaca, maybe.

(21) A: My roommate and I are always arguing about food costs because she buys a lot of junk food.
B: ˇThat’s one hassle I don’t have with my roommate.

In all of these sentences there is a narrow focus, but the fall-rise tone adds the information that the variable of the focus presupposition is a member of a set which is in the context. The meaning of fall-rise is thus something like focus within a given set. It picks something out of a set of possibilities and focuses on it, but it specifically notes the connection of the set of possibilities to the context. In the following paragraphs this analysis is applied to the examples just given.

The first is a straightforward illustration of the meaning of fall-rise as we have defined it:

(16) A: Did you feed the animals?
B: I fed the ˇcat.

Speaker В clearly implies that he didn’t feed the dog (or whatever). ‘Of the group that you asked about, yes, I did feed the following member of that group.’ Like the fall, the fall-rise signals a focus: speaker B presup-poses ‘I fed something’. But a fall focus is simply new information—rhematic, unpredictable—whereas the fall-rise focus also signals a connection to the context. I fed the ˇcat thus means ‘I fed something [focus presup-position] from a set of things in the context [fall-rise nuance] and it was the cat [assertion].’

The hierarchy implied in I fed the ˇcat is like this:

In the next example the situation is slightly different:

(17) A: What would you think of getting a dog?
B: A ˇcat, maybe.

Here we have:

Thus the higher element in the hierarchy need not be explicitly mentioned in order to be in the context. Indeed, without specification of shared assumptions between speakers, the higher element—and thus the nature of the set-in-the-context signalled by fall-rise—may be unclear:

(24) A: What would you think of getting a dog?
B: A “stove, maybe.

This sounds absurd out of context, but if we know that the speakers are a couple on a tight budget, or maybe a pair of ascetics, then we can get something like this out of it:

(This illustrates that ‘context’ must include at least facts about speakers and their shared assumptions. One could make a case that dog and cat are inherently ‘animals’ in the lexicon, but dog and stove are ‘things we can afford’ only for the nonce.)

In the next two examples the fall-rise is used to achieve a kind of polite reply:

(18) A: You have a VW, don’t you?
B: I’ve got an ˇOpel.

(19) A: Do you want a glass of water?
B: I’ll have a ˇbeer.

In (18) speaker В seems to be saying ‘You asked about foreign cars (German cars, small cars, etc.)—is this close enough?’ In the same answer with a fall, I’ve got an ՝Opel, he would simply be saying no. Similarly, in (19)speaker В says: ‘You offered me something to drink—how about a beer instead of water?’ In comparison, I’ll have a ՝beer would have a rather abrupt feel, because in addition to saying no it makes a specific request without any other polite softening.

In the following examples involving superlatives, the situation seems reversed, or at least quite different: the fall is used for enthusiastic agreement, while the fall-rise signals disagreement or at least amendment.

(20) A: Harry’s the biggest fool in the State of New York.
B: Inˇ Ithaca, maybe.

Here speaker В disagrees with A’s statement, and says that he will accept that Harry is the biggest fool in x only if x is Ithaca and not New York State. But if В replies In the whole ՝world, maybe, he agrees that Harry is the biggest fool in New York State, and says that he might even be consider ed the biggest fool in the whole world. The same sort of relationship is found in the following examples:

(26) A: Harry’s the biggest liar in town.
B: The biggest ˇfool, maybe (..,but I think he means what he says)

(27) A: Harry’s the biggest liar in town,
B: The biggest ‘fool, maybe (--not only does he tell lies, but they don’t even make sense)

In (26) the speaker disagrees with liar, and implies that fool is a lesser evil. But in (27) the speaker agrees with liar and goes on to say that Harry may also be the biggest fool, which is to be considered an even greater evil—implied is something like ‘it’s all right to be a liar as long as you’re a clever one.’

If we were to compare the polite replies I’ve got an ˇOpel and I’ll have a ˇbeer with these examples involving superlatives, without reference to any other part of the intonational system, we might find the facts mysterious and contradictory. In the first cases fall-rise is polite, and fall is abrupt, while in the second cases fall-rise disagrees, while fall agrees. But the differences can be reconciled through the analysis of fall-rise proposed here. In the polite replies, the attitude conveyed depends on the meaning of the fall-rise tone. In both cases, the contradiction is mitigated by the speaker’s deliberate use of intonation to relate the focused item to his addressee’s question. Opel is picked out of the set of foreign cars (German cars, small cars, etc.) which the speaker takes to be in the context. Beer is focused on within the set of things to drink which the speaker takes his addressee to have offered him. If the same sentences have a fall, no special connection is made between the utterance and its context, and the sentences come out as bald contradictions.

In the cases with superlatives, the fall-rise tone likewise relates the focused item to a set in the context. But because a superlative is involved, the set is specifically hierarchical. That is, the nature of superlatives is such that if one is the biggest fool in x, one is also the biggest fool in y (where y is less than x), and not necessarily the biggest fool in 2 (where z is greater than x). Thus the answer In ˇIthaca, maybe focuses on Ithaca as a member of the set named in A’s original statement, namely, the State of New York; because the superlative establishes a hierarchy, a member of the set can only be a subordinate or proper subset. So В is narrowing the area in which he considers A’s statement valid and, hence, partially disagreeing with him.⁹ The same intonational phenomena are found with comparatives:

(28) A: That new military base is going to be as big as Texas!
B: Well, New ˇJersey, maybe,

(29) A: This is a bigger atrocity than My Lai.
B: Worse than ˇAuschwitz, maybe.

Returning once more to our original set of examples, we can use our analysis of fall-rise to explain a nuance between two different accent placements:

(21) A: My roommate and I are always arguing about food costs because she buys a lot of junk food.
B: ˇThat’s one hassle I don’t have with my roommate.

(30) A: My roommate and I are always arguing about food costs because she buys a lot of junk food.
B: That’s one hassle I don’t have with ˇmy roommate.

The reply in (21), much more than in (30), seems to cry out for completion with ‘but boy, there sure are a lot of others.’ In (30), the fall-rise focuses on the constituent my roommate and relates my roommate to a set of roommates in the context. In (21), on the other hand, the set in the context to which the speaker relates the focused item is a set of hassles; hence the different implication.

Our analysis of fall-rise can be applied without modification to sentences with two nuclei or accentual peaks. These are the examples on which Jackendoff bases his logical analysis of fall-rise (‘B accent’).

(5) [context: discussion of who ate the various leftovers after a big pot-luck dinner]

A: I know who ate the cabbage, but what about the beans?
B: ‘Fred ate the ˇbeans.

Speaker В focuses on the beans as a member of a set in the context—the leftovers. We can see more clearly what it means for him to do so by comparing his answer with another possible answer: suppose he had instead deaccented the beans:

(31) A: I know who ate the cabbage, but what about the beans?
B: ‘Fred ate the beans.

In this case, В adds no extra emphasis to the beans; the deaccenting relates it to the immediately preceding sentence, and the force of his answer is something like ‘Fred is the one who did what you were asking about.’ In (5), however, В specifically anchors his reply in the larger context of the discourse, which is about who ate the various leftovers; by the fall-rise on the beans, he says something more like: ‘As for the beans, in their capacity as one of the leftover dishes under discussion, Fred is the one who ate them.’

The fall-rise in this double-nucleus sentence thus means the same thing as the single fall-rise in our first group of examples, which shows that its presence or absence is independent of the fact that there is another accent in the sentence. It is this independence that cannot be expressed by Jackendoff’s logical system, and that is the crux of the madequacy of his analysis. For one thing, the two-variable function forces him to choose one variable as independent and one as dependent, and his choice does violence to any reasonable interpretation of those terms. Calling the beans the independent variable and Fred the dependent variable seems inconsistent with the fact that the beans (or any item with a fall-rise) necessarily has some connection with the context, while Fred in this case is unpredictable, new information. Jackendoff’s own statements that the independent variable must be “chosen freely” and that the dependent variable “is a value chosen not freely, but rather in such a way as to make the sentence true” (263) fit poorly with the discourse-level facts.

Moreover, by emphasizing the difference between these two possible replies, JackendofFs system leaves us with no way of talking about their similarity. In ՝Fred ate the beans, Jackendoff says that Fred is asserted to be the value of the single focus-variable. (Presupposition: x ate the beans; assertion: x is Fred.) In ⁴Fred ate the ˇbeans, on the other hand, Fred is said to be the value of a dependent variable related to an independent variable elsewhere in the sentence. (Presupposition:y ate x; assertion; if χ is the beans, then y is Fred.) But as we saw, what speaker В is actually doing when he uses or fails to use the fall-rise on the beans is to set the sentence into the discourse in two different ways. The difference between the two depends on the way in which the beans is linked to the context;pace Jackendoff, it has nothing at all to do with the way Fred is related to the rest of the sentence. In both cases, in fact, Fred plays exactly the same role, namely, that of the new information, the rheme of the answer: this can be seen from the fact that the elliptical sentence ՝Fred would be just as appropriate a reply. And the appropriateness of any of these three possibilities in the context of A’s question contrasts sharply with the strangeness of the following exchange:

(32) A: I know who ate the cabbage, but what about the beans?
B: ˇFred ate the ‘beans.

In this case, B’s reply is akin rather to the elliptical sentence The ‘beans, which obviously does little to answer A’s question. In other words, ˇFred ate the ˇbeans and ˇFred ate the beans are considerably more similar to each other than either is to ‘Fred ate the ՝beans. But for Jackendoff, in ‘Fred ate the beans there is a one-variable function, while in the other two there is a two-variable function, and all three are essentially different. Nothing in his logic reveals the fundamental discourse connections.¹⁰

Finally, recall that Jackendoff is required to resort to the device of calling the dependent variable the ‘affirmation-negation distinction’ in order to explain the use of fall-rise in simple sentences with a single nucleus, like I fed theˇcat. His system, in other words, is not merely unrevealing, but unworkable as well. By treating fall-rise independently of fall, as proposed here, we can readily describe sentences with only a fall-rise, and we are able to describe the discourse relationships it expresses far better than a system which treats both tones as two variables of a single function.

c. Fall-Rise and Scope of Negation

The chapter would be incomplete without some discussion of one of the central points at issue in JackendofFs and Liberman and Sag’s work. This is the interpretation of the scope of negation in sentences with accented quantifiers, like (1) and (2). The analysis of fall-rise proposed so far gives a relatively simple account of the scope effects discussed by Jackendoff and by Liberman and Sag which is at least as convincing as either of their analyses.

Fall-rise, as we have seen, permits us to put things in loose sets like ‘foreign cars’ and ‘things we can afford’. But it can also be used in situations that are grammatically or lexically more tightly structured, to signal set and proper subset, hypernym and hyponym—for example, the way the fall-rise interacts with the hierarchies established by superlative constructions. The following examples show that the use of fall-rise with quantifiers is similar to that with superlatives.

(33) A: Come on into the lounge and meet some of our people--there’s probably a few you’ve never met, but it shouldn’t be any problem.

B: [looks around, sees only a few familiar faces] I’ve never met ՝Tnost of them!

(34) B: The party was terrible—I’d never met any of the people there.
A: Oh, come on—Joe and Julia were there, weren’t they?
B: Well, I’d never met “most of them.

In (33), the quantified set in the context is a few of them; most of them is not a member or subset of that set. In (34) the set in the context is all of them (from any + Neg.), and most of them is a subset. The intonation implies a hierarchy all > most > few, just as in our superlative examples it implied the whole world > New York State > Ithaca.

Another example: I was leaving an exhibition on the ecology and economics of feeding the world, and glanced at a table near the exit where there were books for sale on vegetarianism, the politics of food, nutrition, and other related topics. The following exchange took place between the salesperson and me:

(35) A: Have you seen our books?
B: I’ve ˇgot a ˇcouple of them.

Here a couple of them is being focused on as a subset of ‘our books’. But suppose I had bought two or three of their books and then met a friend on the way out. Then we could have had:

(36) A: I see you bought one of their books.
B: I bought a ‘couple of them.

Here a couple of them is not a subset of the set in the context, namely one of their books.

The two meanings of All the men didn’t go present no problem when we see how quantifiers interact with fall-rise. In ˇAll the men didn’t go,the quantifier is tagged as a subset—but there is no quantifier of which all is a subset, so the quantifier associates with the negative, and we get not all, which of course can be a subset of all. With the fall, on the other hand, all is not identified as a subset, and the negative does not associate with it.

As Jackendoff notes, there is not a one-to-one correspondence between fall-rise or fall and a particular type of negative + quantifier interpretation (see his discussion 352ff). This is exactly what our analysis would predict. For example, in the following examples the intonation tends to affect the meaning, but the semantic distinctions do not seem to have to do with the scope of negation.

(37) A: Do you see anyone here you haven’t met?
B: I haven’t met ‘one of them (=any of them).

(38) A: Do you see anyone here you haven’t met?
B: I haven’t met ˇone of them (=there is one I haven’t).

(39) I don’t want to talk to ^anyone (=no one).

(40) I don’t want to talk to ˇanyone (=just anyone).

And in the following examples, the logical and propositional meaning of the sentence remains the same with the two intonations, and only the way it is connected to the discourse changes.

(41) A: I suppose it was pretty rough meeting all those linguists, wasn’t it--there were probably a few who wouldn’t talk to you.

B: Hah! ‘Most of them wouldn’t talk to me! (most > few)

(42) A: It wasn’t so bad meeting all those linguists, was it. You didn’t think anyone would talk to you, but it seemed like Beveldown and Wandervogel were being pretty friendly.

B: Well, ˇmost of them wouldn’t talk to me. (most > all)

Finally, our other classic fall-rise examples:

(3) John doesn’t drink because he’s unˇhappy.

(4) John doesn’t drink because he’s un’happy.

show that the effects of fall-rise on scope of negation are not merely a function of quantifiers or even of particular quantifiers, but can be obtained any time the meaning of the fall-rise, the negative, and the focused item fit together in an appropriate way. In (3)because hes unhappy is focused on as one reason out of a set of possible reasons; the combination of the negative, the focus on one reason, and the reference to a set of other reasons causes us to interpret the negative as associated with the focused reason, and we infer that John drinks, but not because he’s unhappy. In (4), on the other hand, there is no reference to other reasons and we have no cause to interpret the scope of the negative as being outside its clause.

That is, our analysis of fall-rise puts the association of negation and focus in a different perspective. The semantic effects discussed by Jackendoff are real enough, but they are secondary pragmatic’ effects. The priтагу message of fall-rise is focus within a given set; the logical relation of the negative and the focus is not part of the ‘deep structure’, but only the result of such lines of inference as “all can’t be a subset so it must mean not all.”¹¹

d. Conclusion

The analysis of fall-rise presented here shows the potential value of the intonational lexicon hypothesis. By positing a single abstract general meaning of ‘focus within a given set’, we have explained a wide variety of semantic effects of the fall-rise tone, including narrowly grammatical uses (as in ˇAH the men didn’t go or John doesn’t drink because he’s unvhappy), broadly attitudinal effects (the polite feel of I’ve got an ˇOpel), and nuances that lie somewhere in between (the different effects with comparatives and superlatives). This shows that there can be no sharp line drawn between grammatical and attitudinal uses in our analysis of intonational form; that is, we cannot say, like Lieberman (1967), that such-and-such a formal distinction signals grammatical meaning, while some other distinction is purely emotional.

Moreover, the fall-rise analysis exemplifies again the pitfalls of ‘syntactic determinism’ discussed at the end of Chapter 4. Jackendoff assumes that the logical relation of negation to quantifiers is specified in ‘deep structure’, and that deep structure configurations “trigger” one intonation contour or another. Because of this assumption, Jackendoff must construct an elaborate logico-syntactic device to deal with the fact that in many cases the distinction is irrelevant, or even worse, unclear: scope of negation is not always specified. What is specified is the message of focus within a given set, and it is inferences based on this meaning that give us clear scope differences in certain cases. This suggests that we should approach the complexity of semantics not by assuming that all sorts of relations are specified in deep structure and determine various surface phenomena, but by examining what is specified on the surface, in its own terms, and then seeing how speakers use that information to make inferences about deeper connections.

5. Intonation and Phrasing

The reader may have noticed the absence, in the discussion so far, of any reference to what might be called phrasing cues— the role of intonation in signalling the organization of utterances into phrases, sentences, parentheses, and the like. While the lexical analysis just illustrated does give detailed accounts of certain intonational effects, I believe it is futile to analyze the phrasing cues in the same way. Obviously, these cues are an important part of intonation’s grammatical role, but I believe they are systematized quite differently from the ‘lexical’ contours so far discussed. My purpose in this section is not to present a detailed treatment of phrasing—the subject certainly merits a book of its own—but simply to show how such a treatment must be integrated into the framework developed so far, and to provide evidence for considering the two intonational functions to be formally distinct.

a. Pitch Contours and Boundaries in American Work

Most American investigators of intonation have assumed that phrasing is partly a function of pitch, and in their analyses they do not separate the uses of pitch in phrasing from the specification of meaningful pitch contours. For example, Lieberman’s marked and unmarked breath group depend on the assumption that phrasing is one of the functions of contours: specifically, Lieberman claims that the rising pitch which distinguishes You’re coming? from You’re coming represents a linguistic unit [+BG] (marked breath group), the placement of which also allows us to distinguish [I’ll move on] [Saturday] from [I’ll move] [on Saturday].

In the Trager-Smith analysis, ‘terminal junctures’, which are primarily seen as cues to phrasing, are defined in terms of pitch movement, tempo, and volume, and the specification of any pitch contour has to inelude a terminal juncture at the end. Thus while /||/ might be seen as a signal of a phrasing break in ²Are you ³finished³ || ³he ³askecd³ ||, it is also considered to be an integral part of the intonation contour /^{2 33}||/ on the first half of the sentence. M. Liberman’s ‘boundary tones’ apparently play a similar role.

In the clearest statement known to me of the general point of view that pitch and phrasing are intimately related, Gardiner (1977) states that intonation

The morphemes (not phonemes) of intonation are pitch levels not unlike the Trager-Smith levels (though Gardiner sees the levels not as purely relative but as organized into musical intervals). These morphemes have meanings like emphasis, finality, and non-finality; contours are seen as sequences of pitch morphemes:

Gardiner adds (personal communication) that this view “does not explain the conventionalized types” such as the contradiction contour or the stylized fall (see Chapter 8), which he sees as “analogous to idiomatic phrases.”

b. Boundary Phenomena as Relational

The essence of the American view, then, is that the specification of the pitch at boundaries is part of the larger problem of specifying pitch contours in general. The point of this section is to take issue with this view, and to present evidence that the pitch contours of the intonational lexicon function independently of phrasing cues. Before discussing the evidence, we may once more turn to Chinese tone for a helpful analogy. It is indisputable that Chinese uses pitch contours for something other than grammatical phrasing, yet there are cues to phrasing in Chinese as well. While their basis is not well understood—presumably pause timing, pitch relations between adjacent phrases, and modification of tone contours before pause all play a role—the principle that a language may have both significant pitch contours and separate phrasing cues is established beyond a doubt. This is the principle I propose to apply to the analysis of English intonation.

The most obvious evidence for this idea—noted by Bolinger (1961a, 1970)—is parenthesis. A parenthesis is set off from the sentence in which it is inserted both by pauses before and after, and—perhaps more importantly—by an overall lowering of pitch within the parenthesis:

What is significant is that within the parenthesis, the function of pitch contours (in Bolinger’s terms, the pitch accents) remains unchanged. Both the function of the parenthesis and the intonational cues which identify it are independent of the tones or pitch contours involved.

Better evidence for the independence of phrasing from tone is seen in cases where a phrasing pause actually interrupts the tone, separating part of the tail from the rest. This is seen most readily in quotationattributions:

Notice that in the last example the identification of the tone as fall-rise depends on the rise on said.

What these data suggest is that phrasing cues involve the use of pause and pitch relations, but that pitch contour— tone—is a separate aspect of intonation. That is, the parenthesis is identified not by anything about the shapes of the pitch contours within it, but by the pauses preceding and following it, and by the pitch of the contours within it relative to the pitch of the matrix sentence. Similarly, the phrase boundary pause in the quotation-attribution sentences seems to interrupt the lexical contour, which suggests either that the pause is the only phrasing cue here, or at least that any pitch movement involved in identifying the boundary is systematized differently from pitch in lexical contours.

There is an obvious kinship between this proposal and the typical British view. While admittedly short on phonetic detail, the British in general see the placement of the nucleus, the placement of intonational boundaries, and the pitch contour as three independent aspects of the prosodie system, with three quite separate functions.¹² The tone-unit boundaries are seen to coincide with syntactic boundaries, and to be signailed by various devices of timing. The idea of phrase boundaries interrupting tone contours, however, does not to my knowledge appear anywhere in the British literature; any stretch of pitch bounded by phrasing pauses by definition contains a tone-nucleus. This is a subject for further investigation.

In order to see the relation between phrasing and the lexical uses of intonation, let us return to consider briefly an important aspect of the stress-as-rhythm hypothesis. The most striking characteristic of Liberman and Prince’s conception of stress is that it does not involve phoneme-like or morpheme-like elements, but is exclusively a function of the hierarchical organization of the segmental sounds. The rhythmic structure is not seen as a sequence of stresses, but only as a representation of the relations between elements of the segmental string. Constituents are interpreted as focused or deaccented according to their position in a structure, not because they are accompanied by some morpheme of accent or phoneme of stress. Traditional accounts of stress fail because they reify notions like ‘secondary stress’ and are then unable to state the acoustic or articulatory correlates of such element. Liberman and Prince’s treatment proposes, in effect, that we reify instead the rhythmic structure and take actually occurring variations of pitch, intensity, and timing not as meaningful or contrastive in themselves, but as cues for identifying a structure.

It seems to me that we must explicitly recognize the relational nature of phrasing as Liberman and Prince have proposed we do for stress. Notions such as ‘terminal juncture’ or *boundary tone’, like ‘secondary stress’, can be seen as attempts to treat as segments phenomena that are better handled relationally. The phrasing function of intonation is to be expressed not in terms of contrastive elements of pitch or juncture, but in terms of relational cues which permit us to infer a structure.

The most obvious problem solved by this view is that of boundaries which find no realization as measurable pause or pitch movement, but which nevertheless seem to “be there” somehow. This problem was always particularly acute for the Trager-Smith analysts, whose rule that there could only be one primary stress per phonemic clause’ led to the fairly arbitrary insertion of the juncture /|/ at syntactic boundaries, even in the absence of pause or pitch movement, simply to permit the assignment of two primary stresses in what would otherwise have been a single intonational stretch (cf. Householder 1957). Stating the acoustic correlates of the terminal junctures, especially /|/, often thus demanded a fair leap of faith, as even Trager-Smith followers were prepared to admit.¹³

Consider, though, the similarity of the problem of ‘terminal junctures’ to that of stress levels. Trager and Smith, bound to a theory of phonology which demanded that all audible, meaningful differences be expressed as segments, analyzed rhythmic relations as strings of ‘stress phonemes’; by explicitly allowing relations as elements of our phonological theory, we can provide a better account of the complexity of stress and rhythm. Rhythmic relations are inferred on the basis of the whole structure, and it is futile to look for acoustic correlates of the ‘stress level’ on a particular syllable. In the same way, confronted with pitch relations between phrases and pitch movements at pauses, Trager and Smith reified the relations as segments—terminal junctures—which could be arranged into strings like other phonemes. But if we extend the relational concept to pitch phenomena, we will not posit boundary segments and then worry about their acoustic correlates, but will see pauses and pitch relations as cues to structure. Syntactic boundaries are present in any case, an inherent part of the structure inferred by the hearer; the perception of boundaries, like the perception of stress, is not a matter of hearing a particular acoustic cue or set of cues, but of understanding the structure of which they are a part.

Naturally, this view will require the development of a formalism for representing pitch relations, an undertaking which is beyond the scope of this book. Conceivably, such a formalism might involve something like three ‘pitch levels’: baseline, plus relatively higher and relatively lower; see Crystal (1975: Chapter 4) for related speculations. The point here has been simply to present empirical and theoretical justification for distinguishing the lexical and relational functions of intonation.

c. Phrasing and the Grammatical-Affectiv e Distinction

By separating relational and lexical uses we may incidentally attain new insight into the often-made distinction between grammatical and affective uses of intonation (cf. Chapter 5 Section 2b). Since phrasing functions are—almost by definition—narrowly syntactic, and since (as we saw in Sections 3 and 4 above) the meaningful contours of intonation can have both grammatical and expressive effects depending on context, we may speculate that part of the basis for the grammatical-affective distinction has been an intuitive understanding of the relational-lexical difference, together with factors like the proximity to paralanguage, discussed in Chapters 5 and 6.

The grammatical-affective distinction may also be based in part on the intonation of other European languages, notably French, in which most of the grammatical uses seem to involve relative pitch, and where the utility of positing any contours at all is open to question. For example, Delattre (1966) treats the distinction between what he calls major and minor continuation as a distinction between two different contours, but it is just as possible to view it as purely relational, a function not of contour shapes, but of the relative height of phrase-final pitch peaks:

The relative pitch of œufs and frais signals a more major boundary after frais (major continuation) than after œufs (minor continuation); contours play no role.

In any case, it seems indisputable that the English intonation system is vastly richer than that of French (cf. Schubiger 1965:175), and it is likely that the reason French writers on intonation (e.g., Bally 1941, Delattre 1963, 1966, 1972, Martin ms.) tend to equate the grammatical function of intonation only with phrasing is that English-style intonational lexical segments, with their sometimes grammatical, sometimes affective nuances, play no significant role in French. Unfortunately, their native point of view carries over to their studies on English, and even in the work of careful writers like Delattre we find statements like the following: “The contrast between minor continuation and major continuation is much more marked and much more frequently observed in [French, German and Spanish] than it is in English; only very good English speakers observe it consistently” (Delattre 1963:198). Perhaps by recognizing the relational-lexical distinction, we will be more successful in comparing English intonation to that of other languages.