“The Structure of Intonational Meaning” in “The Structure Of Intonational Meaning”
The Grammar of Accent Placement
1. Syntactic vs. Semantic Approaches
Treatments of the placement of accent can be divided into two basic approaches, which we might call ‘semantic’ and ‘syntactic’ or, perhaps more accurately, Bolinger’s and all the rest. The discussion generated by Joan Bresnan’s article “Sentence Stress and Syntactic Transformations” (1971) is a good illustration of the dichotomy. Bresnan attempts to remedy the problems with Chomsky and Halle’s Nuclear Stress Rule (NSR) by suggesting that it must apply cyclically to underlying Ss and NPs. We will give an example here, since it will be useful to return to it in subsequent discussion.
(1) Helen left directions for George to follow.
(2) Helen left directions for George to follow.
(3) I have instructions to leave.
(4) I have instructions to léave.
(The second pair is from Newman 1946.) Bresnan explains the meaning difference (and subsequently the accent placement) in (1) and (2) by arguing that in (1), follow has an underlying object (namely directions) in the subordinate clause (i.e., George was to follow the directions), while in (2) there is no object of follow (i.e., the directions said George was to follow). In (1), directions in the lower sentence in deep structure is stressed on the lower cycle, reducing the stress on George and follow, then is deleted on the higher cycle; the higher directions must now receive the stress, since George and follow have been reduced. But in (2), follow is stressed in the lower cycle and retains the stress throughout the derivation. Similar processes would apply to (3) and (4). In their replies to Bresnan’s article, G. Lakoff (1972), Berman and Szamosi (1972), and Stockwell (1972) all come up with numerous examples in support of their own theories, examples which show that Bresnan cannot be right, at least not in detail. But both Bresnan and her critics accept the fundamental premise that accent placement is determined by syntax.
This premise is not shared by Bolinger (1972b). Like Lakoff and the rest of Bresnan’s critics, Bolinger adduces numerous counterexamples to her specific proposals, but unlike the others, he goes on to argue—as he has elsewhere—for a ‘semantic’ as opposed to ‘syntactic’ theory of accent placement. Accent, independent of syntax, goes on the ‘point of information focus’ in the sentence, or on items of contrast or ‘emotional highlighting’; “the speaker adjusts the accents to suit his meaning” (635). Copious examples back up this claim, which is summarized in the title, “Accent Is Predictable (If You’re a Mind-Reader).”
This basic idea—that accent is not predictable from syntax—is taken in a different direction in Schmerling (1976). Schmerling sees two sides to Bolinger’s arguments. What she calls his “negative claim” is that accent does not depend solely or even principally on syntax; the “positive claim” is that it does depend on the point of information focus. Schmerling finds Bolinger’s arguments for the first position convincing, and herself devotes an entire chapter (Chapter 4) to discussion of additional evidence for this view. But the second point, she says, leaves a great deal unexplained, and the major proposals in her book are an attempt to pin down more specifically the rules—whatever they may depend on—governing accent placement. In effect, Schmerling accepts that you have to be a mind-reader to predict accent, but argues that even if Bolinger were a mind-reader, he would not get very far with his specific suggestions about ‘point of information focus’ and ‘relative semantic weight’.
Two examples will suffice to show that even when we accept unpredictability of accent as a fact of life, we find that Bolinger has identified a problem rather than a solution. The first example is taken from Schmerling (1976:41-42):
A particularly telling pair of examples showing the inadequacy of a theory which attempts to account for stress in this way can be found in the recent reports of the deaths of two former presidents as I heard them; the examples are worth discussing in some detail. In December of 1972 former president Truman was hospitalized in critical condition. He remained in the hospital for some time, and daily reports concerning his now critical, now serious, now critical condition were given in the news media. Because of the seriousness of Truman’s condition and his advanced age, it could reasonably be assumed that he would not survive this crisis and that it was just a matter of time before he would die. At the time when Truman finally did die, I was visiting my parents; one morning I came downstairs to breakfast, and my mother, who had gotten up earlier and listened to the news, announced to me:
(91) Trûman diéd.
A few weeks later I was back at my job at the University of Texas. One afternoon my husband drove to campus to pick me up when I was finished working for the day, and as I got into the car he announced:
(92) Jóhnson diêd.
Though Johnson’s health had been in the news some time previously, he was apparently recovering from the heart attack he had had, and his condition had ceased to be newsworthy. (In fact, the most recent news concerning Johnson had been his attendance at a civil rights conference at the LBJ library in Austin.) Johnson’s health was not on people’s minds as Truman’s had been, and when his death came it was a surprise.
What is significant for the present discussion is the difference in the contexts in which these two reports were uttered: Truman’s death was expected; Johnson’s was not. Bolinger’s theory would appear to suggest, however, that the mention of Truman in the relevant context should have suggested ‘death’ and, therefore, that died in (91) should not be stressed. On the other hand, the mention of Johnson in the relevant context should not have suggested ‘death’ any more than anything else one might have wanted to say about him, and therefore died in (92) should be stressed. Bolinger’s theory would thus appear to predict stress contours opposite to the ones which actually occurred.
The second example is from Bolinger himself. He lists a number of examples with a NP-Infinitive sequence (like instructions to leave or directions for George to follow) which violate Bresnan’s rule and illustrate his point that the accent placement in these depends on ‘relative semantic weight’ including:
(5) I can’t finish in an hour--there are simply too many tòpics to elúcidate.
(6) I need a lìght to réad by.
(7) I need a tòol to wríte with.
His comments on the last of these are illustrative of the problem he has created. It is true, as he says, that “in toòl to wríte with, tool is a relatively empty word; Get me a péncil to write with [not intended as contrastive Bolinger’s footnote] has a semantically richer noun, one to which write with is relatively incidental.” But he goes on to say: “In fact, Get me a tóol to write with would probably be contrastive, suggesting that one intended to write with a hammer or a screwdriver instead of a pencil” (636). This fascinating observation goes unexplained. I don’t believe that tóol to write with must be ‘contrastive’ in any explicit sense1 (e.g., addressee brings speaker a pencil, speaker says no, bring me a tool to write with) for the meaning of ‘hammer’ or ‘screwdriver’ to be possible. Is tool then simply the point of information focus? If it is, why the shift in meaning from general ‘implement’ to specific ‘implement used for certain building and repairing tasks, e.g., hammer, screwdriver’? Notice that there is no comparable shift in the meaning of directions in the two versions of directions for George to follow, or in any of Bolinger’s other examples. ‘Point of information focus’ leaves this unexplained.
It seems to me that Schmerling is correct, then, in pointing out that Bolinger’s observations are more programmatic than definitive. But in following Bolinger’s program, Schmerling very early takes a wrong turn by disposing of ‘normal stress’. Such a concept is, I believe, central to an understanding of the whole problem, and it is treated separately in the following section.
For all their considerable differences, the two general approaches to accent placement-Bolinger’s and the syntactic approach-both presuppose a distinction between contrastive and normal. The syntactic approach makes this distinction explicit: its rules attempt to account for ‘normal stress’ and not for any ill-defined ‘contrastive’ intonation. Bolinger’s contention that the “accent goes on the point of information focus” appears to preclude the existence of a syntactically defined normal accent position, yet the notion of a semantically neutral accent is there, at least implicitly in references to context-free intonations and to accent placements with and without presuppositions, as well as in references (like the one above in tool to write with) to ‘contrastive’ accent.
There are, as Schmerling points out, a lot of hidden assumptions involved here which surely won’t stand the light of day, and in her Chapter 3 (substantially the same work as Schmerling 1974), she claims that ‘normal stress’ is a concept which can only mislead us, not enlighten us. She suggests and rejects two possible characterizations of ‘normal stress’, one more or less syntactic, one more or less semantic, both more or less familiar. The semantic definition is essentially “the accent placement used in a neutral context,” but Schmerling argues rightly that any context at all involves assumptions which may affect accent placement. This is even true, as she shows, for ‘citation forms’ (1976:50-52). Since these forms have been the last bastion of those who seek a purely ‘syntactic’ definition of ‘normal stress’, the fact that even citation creates a context of its own reduces syntax-based stress rules to circularity: “My rules account for normal stress, and ‘normal stress’ means the stress my rules predict” (48).
But Schmerling’s line of reasoning amounts to the following: “Here are two possible characterizations of ‘normal stress’ which are implicit in the literature. They don’t work. Therefore, normal stress must not exist.” There is, of course, another possibility, which I think is closer to the truth, namely that some sort of neutral accent placement exists but has not been characterized correctly. It seems to me that recent work on the concept of ‘focus’ points the way to understanding ‘normal stress’, and that interpreting normal stress in terms of focus will allow us to integrate the insights of the ‘semantic’ and ‘syntactic’ approaches.
a. A Characterization of Normal Stress
The interaction of focus and accent placement is discussed in Halliday (1967b), Chomsky (1971), and Jackendoff (1972). The most important point that emerges from these works is that while most of the possible accent placements in a sentence signal a narrow focus, one leaves the focus broad or unspecified. While focus is hardly a well-defined concept, its effects in dialogue provide hard data for those unsatisfied by intuitive definitions. Halliday’s examples will illustrate. In
(8) John painted the shéd yesterday.
the focus can be the shed, or painted the shed, or the whole sentence, etc. Thus (8) could be used to reply to a range of questions like What’s new, What did John do, What did John paint yesterday, etc. By contrast, other possible accent placements narrow the focus, so that for example
(9) Jóhn painted the shed yesterday.
could only answer the question Who painted the shed yesterday, and
(10)John painted the shed yésterday.
could only answer When did John paint the shed. Chomsky’s article and its example Was he warned to look out for an ex-convict wearing a red shirt make the same point: ‘normal stress’ can signal multiple foci, while ‘contrastive stress’ narrows the possible range of foci to a particular constituent or small set of constituents. Jackendoff incorporates Chomsky’s observations into his discussion of focus and the scope of words like even.
‘Normal stress’, then, appears to mean the accent placement that leaves the focus broad or unspecified, and on the face of it, this phenomenon makes trouble for the ‘semantic’ approach. Schmerling, of course, simply contests the relevance of focus in this context2 and denies the existence of normal stress. Bolinger, however, talks of focus himself, and admits the existence of context-free intonations. Unfortunately, his explicit statement—what Schmerling calls his “positive claim”—that accent goes on the point of information focus applies only to cases where the focus is a single point. If the focus is broad or unspecified, accent cannot go on the ‘point’ of focus; in such cases one item is somehow chosen to bear the accent, and this is the phenomenon which the syntactic analyses have attempted to describe.
Indeed, even though Schmerling says she does not believe in ‘normal stress’, this unmarked-focus accent is in general what she has tried to account for. This is seen most clearly in her discussion of sentences like Jôhn is a wônderful mán (accent marks as shown by Schmerling 1976:43). She is puzzled by the fact that, as she puts it, there is “primary stress on an item which appears to be conveying little if any information at all. . . . Clearly, [this sentence] involves an assertion concerning John’s character; man seems to add hardly any information.” In the original version of her dissertation she proposed an explanation that she admitted was unsatisfactory; in the published version, this explanation has been deleted and the puzzle is left to stand unsolved.
The point is not that she fails to come up with an explanation of the accent placement in such sentences, but rather that she fails to realize that what she is trying to account for is the very ‘normal stress’ pattern whose existence she denies. It is obvious that there is nothing inherent in such sentences that prevents them from being accented in the way that Schmerling seems to expect:
(11) A: What kind of a man is John?
B: Oh, he’s a wónderful màn!
But it is also obvious that the difference between the two accent placements is that this version focuses on wonderful in a way that Schmerling’s version does not: these intuitions about focus are the basis of the traditional judgments ‘normal. and ‘contrastive’. Schmerling examines the notions of normal. and ‘contrastive’ and finds them wanting, but as a native speaker of English she still has the intuitions about focus, and intuitively she directs most of her efforts to accounting for the accent on the cases with ‘unmarked focus’ even while denying any validity to the concept ‘normal stress’.
However, if we make explicit that what we mean by ‘normal stress’ is the accent placement that signals an unmarked focus, then we are in a position to profit from the insights of Bolinger’s and Schmerling’s semantic approach. For there is a problem with the concept of normal stress, namely, the implication that “every sentence has a ‘normal’ pronunciation, and that any special prosodie properties can be described in terms of deviations from this norm” (Schmerling 1976:49). Defined this way, normal stress is indeed, as Schmerling maintains, a mirage. Her examples of sentences that “must have contrastive stress” show this clearly:
(12) Even a twó-year old could do that.
(13) John was killed by himsélf.
If we take sentence accent as a signal of focus, noting that focus can be unmarked, then these sentences present no problem. It is no surprise that a word like even precludes an unmarked focus; its very function is to focus on some part of the sentence (cf. Jackendoff 1972:247). If we want to phrase it paradoxically, we can say that there is no normal stress pattern possible for sentences like Even a two-year-old could do that, but this is a paradox only insofar as we insist on having a syntactic ‘norm’ against which to measure ‘deviations’. If instead normal stress is taken to be the accent pattern that signals an unmarked focus, then the absence of such an accent pattern in sentences like these is exactly what we would expect. In this sense the point against the semantic approach is a small one: our definition of normal stress is based on the semantics of focus. Bolinger’s formulation might be changed from “accent goes on the point of information focus” to “accent goes on the point of information focus, unless the focus is unmarked, in which case the accent goes in a location determined by the syntax.” But focus remains a semantic phenomenon; we have simply added an unmarked case.
b. Syntax of Focus
Once we understand the role of accent in such terms, we see that the investigator’s task is not to write rules by which ‘normal stress’ can be assigned to any sentence, but to discover the principles by which the meanings of broad and narrow focus are signalled. Here the work of the ‘syntactic’ approach represents a solid basis for further work; for an opening into the problem we can simply reinterpret the Trager-Smith-Chomsky-Halle normal stress rule (Chomsky and Halle’s ‘compound and nu- clear stress rules’) in terms of focus. Instead of saying that normal stress goes on the rightmost accentable item in the sentence, we will say that accent—in general—goes on the rightmost accentable item of the focus constituent. If the focus constituent is the whole sentence, we get ‘normal stress’; if not, we get a narrow focus on the constituent identified by the placement of the accent.
This formulation is similar, of course, to those found in other work on focus. Thus:
The tonic falls . . . on the last accented syllable of the item under focus. [Halliday 1967b:207]
If a phrase P is chosen as the focus of a sentence S, the highest stress in S will be on the syllable of P that is assigned highest stress by the regular stress rules. [Jackendoff 1972:237]
But it is important to point out a subtle difference between Jackendoff’s and Halliday’s statements which is the essence of the difference between the straight ‘syntactic’ approach and the view adopted here. Jackendoff considers his focus rule to be a principle that interacts with the ‘regular stress rules’, i.e., syntactic rules that assign normal stress. In my opinion Halliday’s formulation is more nearly correct: there are no ‘regular stress rules’. The focus rule is the accent placement rule. There is no essential difference between normal stress and any accent placement that signals narrow focus; broad and narrow are simply points on a spectrum.
This can be shown with an example involving even.
(14) Even a nineteenth-century professor of clássics wouldn’t have allowed himself to be so pedantic.
Suppose I say this to take a friend to task for a pedantic remark. Assuming the friend has nothing to do with classics, is not a professor, and is more or less contemporary, then we could reasonably say that the focus of the sentence is nineteenth-century professor of classics, that being the most pedantic sort of person I could think of in the heat of the moment. Suppose, however, I say the sentence while discussing a particularly fine example of nineteenth-century pedantry. Then we might reasonably say that the focus is only professor of classics; I am trying to come up with the most pedantic sort of nineteenth-century writer I can think of. Similarly, if the discourse is even more particularly about pedantry in the work of nineteenth-century professors, then the focus could be identified as simply classics (of classics?), my implication being that of all the pedants among nineteenth-century professors, classicists were likely to have been the worst.
Even though from the point of view of the whole sentence there is a narrow focus (‘contrastive stress’) associated with the presence of even, there are nevertheless multiple focus interpretations possible. Thus breadth or scope of focus is not an either-or dichotomy as implied by ‘normal’ vs. ‘contrastive’: focus can apply to constituents of any size. Normal stress is simply the accent placement that permits the broadest possible focus interpretation—focus on the whole sentence. Other accent placements signal foci which are narrower than the whole sentence, but which, as we have just seen, may also have internal structure of their own and thus permit multiple focus interpretations as does normal stress.
c. ‘Contrastive Stress’
One of the advantages of the approach just sketched is that it frees us from trying to understand things in terms of the catch-all notion ‘contrastive stress’. The existence of the term implies the existence of an identifiable, unified semantic phenomenon with identifiable phonological manifestations, but in fact it is almost invariably used as nothing more than the syntactic converse of normal stress—anything not ‘normal’ is ‘contrastive’. For example, the accent in Did John paint the shed yésterday? would usually be termed contrastive, because normal stress falls on shed; yet semantically the best we can do is say that yesterday contrasts with ‘any other day’, which tells us nothing. It is just this empty use of the notion ‘contrastive’ against which Bolinger (esp. 1961b) and Schmerling (1976:60-66) have convincingly argued. Yet it also seems clear that there is something special about the accent placement in this example, something which underlies everyone’s intuitions that it is different from ‘normal’ stress—namely that it focuses on yesterday in a way in which Did John paint the shéd yesterday? does not focus on shed. ‘Contrastive stress’ is nothing more than accent placement that signals narrow focus, and narrow focus can be used for reasons other than explicit contrast.3
Various phenomena which have never made much sense in terms of contrast make more sense in terms of narrow focus. For example, Chomsky (1971:205) notes that “in ‘parallel constructions’, in some sense of this notion that has never been made quite clear, contrastive intonation is necessary.” He illustrates with:
(15) John is neither éager to please, nor éasy to please, nor cértain to please. . .
Now, these accents are ‘contrastive’ only in the sense that they are not ‘normal’; they can better be seen as focusing on the points of difference in otherwise identical phrases.
Lest the reader think that I am offering the same product under a new label, however, I should reemphasize that focus is an independent semantic phenomenon. By trying to predict accent patterns on the basis of syntax, Chomsky puts the cart before the horse. He assumes that ‘parallel’ is a syntactic phenomenon, which somehow ‘triggers’ contrastive stress. In fact, however, I would argue that ‘parallel’ is a semantic notion signailed by the paired foci. Chomsky picks his examples of parallel constructions on the basis of the accent pattern and then argues that the syntax is the deciding factor. We have here an illustration of Bolinger’s precept that “the error of attributing to syntax what belongs to semantics comes from concentrating on the commonplace” (1972b:634). That is, in Chomsky’s example just cited, it is undoubtedly true that eager is felt to be parallel to easy and certain partly because of the syntax. But the syntax is not decisive; ‘parallel’ is an independent semantic notion, as the following example shows.
Suppose A and В are in a conversation about the difficulty of staying in touch with relatives.
(16) A: When was the last time you saw any of your relatives?
B: My mother called me yesterday—does that count?
There are three equally plausible accent patterns for B’s remark, with three distinct meanings. In each case the tag question does that count? asks, in effect, whether the focus signalled by the accent is acceptable to the other members of the conversation. If В says My móther called me yesterday, he wants to know whether parents are to be included among the relatives being discussed. If he says My mother cálled me yesterday, he is asking whether telephonic contact counts. But if he says My móther called mé yesterday, he is specifically setting up a parallelism between (relative) (get in touch with) (member of conversation) and (member of conversation) (get in touch with) (relative), and wants to know if that parallelism is acceptable. That is, he is asking whether situations in which the contact is initiated by the relative are to be considered as fulfilling the family obligation under discussion. It should be clear that the syntax in no way determines the accent placement: if, and only if, the speaker wants the accent pattern that says ‘parallel’, he uses it.
I should note that I am uncertain whether ‘parallel’ is to be regarded as a specific accent pattern, or whether the semantic nature of focus is such that one possible meaning of paired foci is ‘parallel’. In either case, one specific use of ‘parallel’ is ‘reciprocal’ or ‘vice-versa’, as we have just seen. Another example is:
(17) A: Hey, come hére.
B: No, yóu come hére.
The reciprocal meaning also seems to be at work in the old standby:
(18) Jóhn hit Bíll and then hé hit hím.
Once we get away from the idea that the accents on he and him are ‘contrastive’, and see instead paired narrow-focus accents combining to form the accent pattern meaning ‘parallel’, then this example is no longer mysterious.
d. Summary
In summary, the dispute between the ‘semantic’ and ‘syntactic’ approaches to accent placement reminds us again of the story of the blind sages and the elephant: everyone has a piece of the truth. I have argued in this section for integrating the two approaches as follows: treating accent as a signal of focus, treating focus as a semantic notion involving no inherent normal-contrastive dichotomy, and assuming that focus can apply to any constituent and that accent falls on the rightmost accentable item of that constituent. However, this particular elephant is a fairly complicated creature, and the last statement—that accent falls on the rightmost accentable item of the focus constituent—is the source of numerous difficulties which are the subject of the remaining sections of this chapter.
Not all cases of what would traditionally be called ‘contrastive stress’ can be considered narrow focus accents; rather, they are a kind of byproduct of deaccenting which I will call default accent. If deaccenting is seen as the relative weakening of an item or constituent, then necessarily—because of the relational nature of rhythmic structure—it must also involve the concomitant strengthening of some other item or constituent. This strengthening can be seen as a kind of accenting by default. Such a notion expresses the effect of the accent in two of our earlier examples, repeated here:
(19) A: Has John read Slaughterhouse-Five?
B: No, John doesn’t read books.
(20) 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...
The accent in (19) is in no sense ‘contrastive’, as it is often said to be: the meaning of B’s reply is not the explicit contrast ‘John doesn’t read books, he writes (reviews, collects, burns, etc.) them.’ Rather, the point of the accentual pattern is that books is deaccented; the focus is broad, but the accent falls on read by default. Similarly, in (20), the accent on denounce does not signal any sort of focus, but is there by default, because of the deaccenting of the measure.
The idea of default accent applies to scores of examples.
(21) I can’t imagine what it would be like to be a dentist--but I’m awfully glad there are guys who want to be dentists.
(22) A: Man, it’s hot! Doesn’t feel like it’ll cool off till tomorrow at least.
B: Yeah, they sáid it would be hot all day.
(23) What happens to male spiders, anyway? Are they like a lót of male insects [...in that the female eats them after mating?]
In (23) the speaker has implied that spiders are insects. By accenting insects, the opposite would be implied.
One of the puzzles in Akmajian and Jackendoff (1970) suggests a solution in terms of default accent:
(24) John washed the car. I was afraid someone élse would do it.
(25) John washed the car. I was afráid someone else would do it.
As they point out, in (24) the implication is that I wanted John to wash the car, while in (25) I wanted to do it myself. In terms of ‘contrastive stress’, Akmajian and Jackendoff have no explanation for these nuances. But if we see someone else in (25) as deaccented, to signal preferentiality with John, then the accent falls on afraid by default.
There are a number of reasons that the phenomenon of default accent seems not to have been noticed before.4 First, of course, is the existence of the term ‘contrastive stress’: since default accent is not ‘normal’, then by the traditional definition it must be contrastive. Second, the idea of default accent emerges clearly only from an understanding of the relational nature of deaccenting. As long as writers (like Schmerling and Chafe) treated deaccenting as an actual level of stress, low pitch, etc., there was no reason to suspect that the accentual behavior of any other item in the sentence might be aifected by it. Once we see deaccenting as a relative weakening, the discovery of concomitant relative strengthening comes as no surprise. Finally, however, there are many cases where the ‘contrastive stress’ explanation seems quite satisfactory. For example, in
(26) Harry wants a VW, but his wife would prefer an Américan car.
it does seem reasonable to talk of ‘contrast’ on American; the need for a separate notion of default accent in cases like these is obscured.
The existence of such cases should not be taken as casting doubt on the notion of default, however, for there is evidence for the distinction between default and narrow focus accent in the form of actual ambiguity. Consider the question linguists have all been asked many times: How many languages do you speak? While the neutral accent placement seems to be on languages, I have also been asked the question this way: How many languages do you spéak? This version has two possible interpretations, one corresponding to rather complete ignorance of linguistics and one corresponding to a somewhat sophisticated understanding of the field. The ignorant speaker has deaccented languages to convey the implication ‘I know that linguistics means studying languages.’ The knowledgeable speaker, on the other hand, has accented speak because he knows that many linguists know a great deal about certain languages without necessarily being able to speak them, and is interested only in those languages of which his addressee has a speaking knowledge. We might represent the ambiguity this way:
(27) How many languages do you speak? [deacc]
(28) How many languages do you speak? [focus]
In (27), the accent on speak is a default accent; in (28), it is a narrowfocus accent.5
The distinction is thus clear in principle; it becomes muddied only in some contexts, like (25), where we cannot say for certain if the accents are placed by default, or as narrow-focus accents, or for some intermediate reason. The structure allows either interpretation, and in some cases the context may not require or even permit us to distinguish. But this indeterminacy should not trouble us; it is implied by the relational nature of stress patterns. The rhythmic structure for American car will be
This configuration of strong and weak can result from relative strengthening of American (for narrow focus) or from relative weakening of car (for deaccenting), but there is no reason to assume that it cannot also result from both at once. Bolinger has suggested to me that “perhaps accents go where they do both in order to highlight what they are on and to cast in the shade what they are not on.” This aptly sums up the semantic force of accent, and captures in metaphor its relational nature: if we shine the spotlight on one actor, everything else on stage is in shadow by comparison. In deciding how to accent a sentence, a speaker must take account of where he wants light and where he wants shadow; the decision can involve an enormously complex tradeoff between conflicting semantic forces. But in many cases the various forces are all more or less in harmony. Indeed, speakers probably cast most sentences so as to assure that the forces will be in harmony. These are the cases like Américan car, where we want American accented and car deaccented. One accent placement is compatible with both intentions.6
One danger of tying the ‘focus rule’ to traditional normal stress and focus rules, as we have so far, is that we inherit their inadequacies. I think their most significant inadequacy—the unexamined assumption that is the source of most of their difficulties—is the concept of accentability. The Trager-Smith-Chomsky-Halle tradition assumes that content words are inherently accentable and function words are inherently not accentable.7 This is how they account for the accent placement data in sentences like the following:
(30) I’m leaving for Créte tomorrow.(neutral)
(31) I’m leaving for Crete tomórrow. (focus on tomorrow)
According to the focus rule or the traditional normal stress rule, neutral accent falls on Crete because it is the rightmost accentable item. Tomorrow is somehow not accentable.
It is clear that there is a difference of accentability between Crete and tomorrow; the same difference is reflected in the narrower foci in the following:
(32) Even a guy who’s leaving for Créte tomorrow shouldn’t look so pleased with himself.
(multiple focus interpretations possible within the constituent a guy who’s leaving for Crete tomorrow)
(33) Even a guy who’s leaving for Crete tom6rrow shouldn’t look so pleased with himself.
(focus on tomorrow)
What is not so clear is that this difference is a simple dichotomy: content word vs. function word, accentable vs. non-accentable. While most of the accent placement data examined so far would support such an interpretation, at least two types of sentences, discussed in the previous chapter, disturb the generalization. These are the types of sentences which have given rise to Bresnan’s distinction between initiatory and elicitory questions and Schmerling’s notion of topic-comment sentences: sentences with subject and intransitive predicate, and WH-questions ending with a verb. What is exceptional about these is that the neutral accent placement is on the last noun:
(34) МУ párents called. (neutral)
A: Maybe we ought to call your parents and tell them.
(35) B: My parents cálled--they already know.
(parents deaccented)
(36) How many lánguages do you speak? (neutral)
(37) How many languages do you spéak?
(languages deaccented, or speak focused)
(For more examples see Chapter 3 Section 2.) Since many sentences have a noun as their last content word, a rule putting the accent on the last content word and one putting it on the last noun will be equivalent in many cases. But cases like those just cited provide a way of distinguishing the two rules, and the one referring to the last noun seems to account for more of the data. In sentences without nouns, like I háte him or She didn’t want to léave, the rule referring to the last content word still holds.
This suggests not a dichotomy between content words and function words, but rather a hierarchy of accentability, where content words are more accentable than function words—this is surely uncontroversial—and where nouns are more accentable than other content words.8 That is, lexical items are not, as the Trager-Smith-Chomsky-Halle tradition implies, either accentable or not accentable; accentability is a matter of degree. The most important implication of this is that accent placement within the focus constituent can no longer be seen as the mechanical selection of the ‘rightmost accentable item’. Rather, accent placement depends somehow on the interaction of position in the constituent and some poorly understood hierarchy of parts of speech. The focus rule must be revised:
Revised Focus Rule: Accent goes on the most accentable syllable of the focus constituent.
It would be possible, of course, to formulate the traditional normal stress rule this way: the difference lies in what is subsumed under ‘most accentable’. Only the view presented here makes allowance for degrees of accentability: accentability is not absolute, as in the traditional view, but relative within the focus constituent. By treating accentability this way, we can, in addition to accounting for the difference between nouns and other content words, begin to incorporate a good deal of other data into a unified explanation of accent and focus. The remainder of this section is devoted to a brief discussion of a few such topics.
a. Compounds
Within noun compounds, accentability is determined by internal factors. This does not affect the application of the focus rule. Suppose we modify an earlier example:
(38) Even a nineteenth-century clássics professor wouldn’t have allowed himself to be so pedantic.
In order to focus on the whole constituent nineteenth-century classics professor, we have to put the accent on classics. In effect, we have put the accent on the constituent classics professor, which is farthest to the right; within that constituent, classics is more accentable than professor for reasons having to do with the syntax and semantics of compounds. While accentability within compounds is not well understood, its principies are clear enough to the native speaker that accent placement can take them into account. Thus
(39) Even a nineteenth-century classics proféssor wouldn’t have allowed himself to be so pedantic.
would likely be interpreted as a narrow focus (professor as opposed to student, perhaps), because we know that normally classics would be more accentable than professor.
b. Bolinger’s ‘Contrastive Stress’
Notice that the focus rule says that the accent goes on the most accentable syllable of the focus constituent. This fits with Bolinger’s notion of stress as a potential for accent. Given accent on a certain word, the lexical stress determines which syllable of the word is actually accented. This is exactly the same as the application of the rule to compounds that we just discussed: given accent on a certain compound, the internal principles of accentability within the compound determine which word of the compound is actually accented. And just as we could infer narrow focus in (39) by knowing the principles of accentability in compounds, so we can infer ‘contrastive stress’—in the restricted sense of Bolinger (1961b) — by knowing the usual stress placement in a word. In Bolinger’s example, This whisky wasn’t éxported, it was déported, there is a real sense in which we are focusing on one syllable (or morpheme) of the word. We are signalling to the hearer that the whole point of the sentence lies in the contrast between these two possible syllables. Given the constituent structure
the de- is the most accentable syllable of only one constituent—itself. Hence we infer narrow focus—contrastive stress—on de-. But we can do this only because we know that word stress rules would put the accent on -port- if the focus were any broader.9
c. Deaccenting
With a concept of relative rather than absolute accentability, we can return to our discussion of deaccenting and include it as an integral part of an account of accent rather than a separate phenomenon. If deaccenting is to be seen as a lowering of the degree of accentability of an item or constituent, then, like compounding or lexical stress, deaccenting will interaet with the focus rule. This explains the phenomenon of default accent. Consider again example (19):
(19) A: Has John read Slaughterhouse-Five?
B: No, John doesn’t réad books.
If we consider this a broad focus in the reply, and note that books, by being deaccented, is less accentable in the focus constituent than read, then the accent will fall on read quite normally by the focus rule. The noteworthy feature of the accent placement in this sentence is precisely that it does not signal narrow focus on read; that is why we proposed the notion of default accent. The focus rule, together with the concept of deaccenting as a lowering of accentability, makes it clear how the focus remains broad.
One of the puzzles in our discussion of deaccenting was the question of why deaccenting sometimes shifts the accent to the left and sometimes to the right (Chapter 3 n5). As we saw, this phenomenon makes problems for an analysis like Vanderslice and Ladefoged’s, which explains deaccenting in terms of the deletion of the feature [+accent] in a linear sequence of accents. Vanderslice and Ladefoged assume that deaccenting is a function of the shifting of the intonation center, which they locate by rule at the last [+accent] syllable in the sentence. This predicts that deletion of [+accent] could produce only a leftward, never a rightward shift. The analysis proposed here, however, which incorporates both reference to constituent structure (not just linear sequence) and to degrees of accentability (not just [±accent]), explains and predicts accent shifts in deaccenting correctly.
An especially interesting case of rightward shift in deaccenting is seen in nominal compounds, as in the following examples:10
(41) We’ve got lots of books, but we haven’t got any bookcáses.
(42) [To a couple who has just returned from a dance floor, said by someone who was impressed by their dancing ability]: Did you two ever take dancing léssons?
(43) [Student in phonology class discussion of Jakobson’s theory explaining the rarity of certain types of speech sounds in terms of acoustic distance from other types]: Maybe the reason clicks are so rare is just that they’re difficult to produce. [Professor]: Well, speakers of click lánguages wouldn’t think so.
(44) Come on down and take advantage of our oil change special this week. We’re selling [brand] 10W-30 all-weather óil for just 52c a quárt, and we’ve got [brand] oil fílters for only a dollar ninety-fíve.
(45) A: Hasn’t the faculty voted on that yet?
B: No, it’s coming up at the next faculty méeting.
In each case, the normally accented member of the compound—book-, dancing, click, oil, faculty—has a reason in the context for being deaccented. The contextual factors in the deaccenting in all these cases are similar to those we have seen in others; what is different is that the accent shifts to the right. But we can explain this in terms of our discussion so far: the focus principle puts the accent on the compound, and within the compound the usual hierarchy of accentability is reversed by the deaccenting. Notice what would happen if accent shifted to the left in any of these cases: the focus would be narrowed because the accent would fall in a much smaller constituent (e.g., We’ve got a lot of books, but we haven’t gót any bookcases11).
There are many other examples of rightward shift; I cite only two more. In I looked the answer up, in order to deaccent the answer, the accent must fall on the verb looked up, the next most accentable item in the sentence. Within this constituent, lexically determined accent patterns make up more accentable than looked, hence the accent ‘shifts right’ and falls on up. Shift left to looked would be interpreted as a narrow focus on looked (say, looked up instead of wrote up), because, as in the case of deported, looked is the most accentable item of only one constituent, itself. Similarly, in How many languages do you speakP, ‘leftward shift’ from languages would give How mány languages do you speak?, with a narrow focus on how many. Given the degrees of accentability within the focus constituent—i.e., the whole sentence—speak is more accentable than how many, so once languages is deaccented, the accent shifts all the way right to speak. All these examples make clear that a treatment of deaccenting in terms of linear sequence alone is inadequate.
The account presented here also makes it possible to explain Schmerling’s observations (1976:55-56,67-70) that context plays a role in determining normal stress. It is true that in certain contexts, a particular accent placement may be interpreted as neutral even though in another context it might be interpreted as signalling a narrow focus. Her example:
(46) This is the dóctor I was telling you about.
(presumed ‘normal’).
(47) This is the doctor I was télling you about.
(‘normal’ if spoken in the context of a hospital or medical convention).
Schmerling uses examples of this sort as further evidence against the notion of ‘normal stress’. Yet it is clear that both of these sentences do convey a broad focus on the entire predicate NP: this is x. In the hospital context, doctor can be deaccented; its accentability as a noun is lowered and the verb telling is the most accentable item in the focus constituent. It is exactly because it is easy (in the hospital context) to interpret doctor as deaccented that (47) is interpreted as a broad focus, i.e., ‘normal’. Outside the hospital, where there would be no apparent reason to deaccent doctor, it would be more accentable than telling, and would receive the accent. Here an accent on telling would leave the hearer puzzling to find a connection between doctor and the context, or trying to understand why the speaker was insisting on telling.
A similar case is seen in the common pronunciation among linguists of language acquisition with the accent on acquisition. This is an exception to the usual pattern in compounds expressing an object-verb semantic relation, compounds such as básket weaving, aúto repair, oíl production, cróp rotation, wísh fulfilment, and so on. The explanation is simple: in countless contexts where linguists would have reason to speak of language acquisition, there would also be reason to deaccent language, as in the following:
(48) A: What’s his dissertation about?
B: Something to do with language acquisítion.
A linguist’s dissertation is likely to be about some aspect of language—so language can safely be deaccented. But the linguist’s proud parents, who are not linguists, would tell their friends, who are also not linguists, that their son’s dissertation was about lánguage acquisition. For them language would not be deaccented, and the usual relationships of accentability within the compound would apply.
d. The Accentability of Nouns
Finally, I should return to discuss various doubts and questions concerning the idea that nouns are more accentable than other content words. Unlike the difference in accentability between content words and function words, which is universally assumed, the primacy of nouns is suggested in only a few places. Gunter (1966:178) speculates that in a truly contextfree sentence, all the nouns are accented; a similar principle is incorporated into the accent rules for speech synthesis devised by Coker et al. (1973). Schmerling states the idea quite explicitly in her Principle II: “The verb receives lower stress than the subject and the direct object, if there is one; in other words, predicates receive lower stress than their arguments, irrespective of their linear position in surface structure.” (Though Schmerling’s Principle refers to predicates and arguments rather than nouns and verbs, all the evidence she cites is compatible with either interpretation.) I believe these formulations can best be expressed by the hierarchy of accentability proposed here.
Yet numerous types of apparent counterexamples abound in the literature and in everyday speech, such as these two:
(49) Jesus wépt.
(50) I found a flý in my soup.
I label these ‘apparent’ counterexamples, because I think they will be shown to follow regular rules once the details of deaccenting and of relative accentability are better understood. Here I simply present briefly the kinds of explanations that I think will be found to apply to these cases.
Jesus wépt probably involves the deaccenting of names. Names seem to be much less accentable than other NPs in identical contexts; the explanation I propose is that names are commonly deaccented to refer them to the context in a special way, something like “This name refers to the one that you and I know, rather than to any of the other possible referents of the same name.” Because of this, we are quite ready to interpret Jesus wépt—with Jesus, in effect, deaccented—as ‘normal’. This, it seems to me, is also the explanation for Schmerling’s problem with the citation pronunciation of John diéd with the accent on died (1976:51-52). Schmerling is right when she says that “not all citations are really contextless” and “putting an example sentence in a context forces an informant to consider what the sentence means.” But she is exactly wrong, I think, when she compares John diéd to the usual citation form My phýsics professor died and concludes: “Apparently it is easier to imagine having a physics professor than it is to imagine knowing this John of linguistic example-sentence fame.” Rather, it is John who is assumed to be in the context somehow—implying “You know John, we’re talking about the same guy, right?” The physics professor, on the other hand, is not taken to be a part of the hearer’s contextual frame of mind. (Berman and Szamosi 1972:317 also cite several examples with deaccented names as ‘normal’; Bolinger 1972b:640 objects to this on the grounds that the names would have to be ‘presupposed’, but I think Berman and Szamosi have at least a piece of the truth.)
I found a flý in my soup involves both the deaccenting of locatives and differences of accentability between definite and indefinite NPs. Locatives, like names, are often deaccented to suggest to the hearer that the location is familiar, close at hand, ‘in the context’ somehow, to signal that no real ‘news’ is involved in the specification of the location. Compare the following sentences containing accented prepositional phrases, one with a ‘news’ locative, one with a metaphorical locative, and one with an instrumental phrase:
(51) 1 found a piccolo in my máilbox.
(52) You haven’t got a military bone in your bódy.
(53) John carved an elephant with his Swiss knífe.
(in (52), the other possible accent placement You havent got a military bóne in your body deaccents the locative and refers to the physical context, i.e., the addressee’s literal body, thereby conveying the surrealistic implication ‘Everyone was issued one; where’s yours?’) The accentability of nouns also seems to depend to some extent on whether they are definite or indefinite; not surprisingly, indefinite nouns are more accentable than definite. In a sentence like (50), this means that if the complement NP is definite and the locative NP indefinite, the accent is more likely to fall on the locative, as in
(54) I left my car in a tów zone.
What the cases discussed in this subsection make clear—like Schmerling’s example This is the doctor I was telling you about— is the extent to which deaccenting, in the sense of relatively lower accentability, is an integral part of many examples which linguists have been inclined to see as exhibiting normal stress. If we are sensitive to this possibility, we may find that the answer to many puzzles about accent placement lies in the principle that nouns are in general more accentable than other content words, but that they may also be deaccented—or lowered in relative accentability—for quite a variety of reasons. Also, these cases remind us that focus and deaccenting are in the same linguistic pot with such things as newness, definiteness, word order, and contextual reference, and that as we make progress in understanding those phenomena in the languages of the world, we will be in a position to understand better the details of noun accentability in English.
This leads finally to a detailed discussion of the meaning of deaccenting. Up till now I have merely exemplified deaccenting, glossing over an explanation of when it occurs with rough intuitive notions like ‘given’ or ‘already in the discourse’. If I have demonstrated that there must be a unified relational phenomenon of deaccenting, I have scarcely touched on the semantic reasons underlying its use. Why do we deaccent? What does it mean when we do?
For an opening into the problem, let us return to example (20), in which deaccenting seems specifically to signal coreference.
(20) 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 denounced the measure. Goldwater said...
The lack of accent on measure says something like “this noun is interpretable only by reference to some other part of the discourse”; the measure here means specifically “the bill that was sent to Congress today by President Carter which would require peanut butter sandwiches to be served at all government functions.” Obviously, “legislative act or proposal” is only one of many dictionary meanings of measure, and obviously, too, no dictionary would ever list “bill sent to Congress today by President Carter which would require peanut butter sandwiches to be served at all government functions” as a meaning for measure; yet just as obviously, that is exactly how measure is interpreted in this context.
But coreference is only one reason for deaccenting; most examples cannot be so simply explained. Even in our early examples like Américan car and John doesn’t réad books there is no strict coreference, and in other examples the nature of the connection to the context is even murkier. What we need is a more general description of the function of deaccenting, which will cover both coreference and the less specific uses. Such a general rubric is provided by Jakobson’s notion (1971) of shifter, an item which can be fully interpreted only by reference to the context. (The most obvious examples are pronouns.) Deaccenting, I would argue, is a signal that the deaccented word has become a shifter in this sense. In the context of (20), the pronoun it (which is inherently a shifter and thus has far less inherent semantic marking) would have risked being unintelligible, its referent being unclear; by using the measure, which is not inherently a shifter and has some lexical meaning of its own, the speaker “fleshes out” the shifter, so to speak, to make clear just where in the discourse the hearer is to seek the referent; but the shifter-like quality, the interpretability only with reference to something else in the context, is clearly signalled by the lack of accent.
Such a description of the function of deaccenting is attractive, because it ties together the ‘obvious’ cases of deaccenting like John doesn’t réad books with other cases like those cited at the end of the previous section, which are not generally recognized as involving deaccenting. Moreover, it provides a simple solution to the tool to write with problem mentioned at the very beginning of this chapter. As we said, Bolinger notes that tool to write with would be the usual accentual pattern, and that tóol to write with “would probably be contrastive, suggesting that one intended to write with a hammer or screwdriver instead of a pencil.” In this latter case, tool, accented, stands on its own, semantically speaking, and is likely to receive its most usual specific meaning of “implement used for certain building and repairing tasks, e.g., hammer, screwdriver,” or some such. In the other case, tool, deaccented, is stripped down to its semantic bare bones, ‘implement’, and reaches out in the context for something to flesh out its meaning—just as the measure did in the previous example.
Notice that the principles of accentability presented in the last section predict that neutral accent in a construction like tóol to write with will fall on the noun, and that therefore the accent on the verb in tool to wríte with is a default or focus accent. By showing the sense in which tool is referred to its context for its full interpretation, we not only provide justification for considering deaccenting to be a single semantic process, but also confirmation of the accentability principle which predicted that tool in tool to write with would in fact be felt as deaccented.
It should be emphasized that the meaning just given for tool to write with is not the only possible one. Given only this context, tool does seem to depend on to write with; but given a richer context, tool may find sematic support elsewhere. We can imagine a comedy routine:
(55) A: What did you bring me this for? I asked for a tool to wríte with, not a tool to drive scréws with.
B: I don’t care what you do with it—you can scratch your nose with it for all I care. You asked for a tool, I brought you a tool.
Speaker A has deaccented tool in each construction with reference only to the immediate grammatical context, fleshing out the meaning within the construction itself. Speaker В replies as if tool were deaccented with reference to the broader context, taking the contextual referent in each case to be the screwdriver at hand.
Similar observations could be made for the case of Helen left directions for George to follow. The two versions are repeated here:
(1) Helen left diréctions for George to follow. (George was to follow the directions.)
(2) Helen left directions for George to fóllow. (The directions said George was to follow.)
Bresnan’s point about these is that both are ‘normal stress’ readings, and therefore she takes the syntax to be decisive. Indeed, in both cases we do have a focus Helen left x— this is the sense in which both are ‘normal’. But we can see the difference of accent placement as a difference of deaccenting, not syntactic relations. Within the constituent represented by x, directions is most accentable if no deaccenting is involved,12 and we get Helen left diréctions for George to follow. But if directions is deaccented, then follow becomes most accentable, and we get Helen left directions for George to follow. The difference in meaning signalled by the deaccenting is rather analogous to that between restrictive and nonrestrictive relative clauses. In (1), the for-to phrase is nonrestrictive, giving merely supplementary information about the directions. In (2), the for-to phrase is restrictive, giving the hearer essential information which narrows the possible range of referents. In the first case, the directions are simply directions; they could be directions to a theater or directions for making moussaka. In the second case we are told not only that they are directions, but that they are directions saying such-and-such. It is in this sense that the meaning of directions in (2) is fleshed out in the context, and it is in this sense that we can talk of directions as deaccented.
The syntactic relations that are the basis of Bresnan’s argument are simply inferred from the semantic relations signalled by the deaccenting or lack of it. Without any other context, we seek to interpret the deaccenting of directions within the sentence, and accordingly make the syntactic inferences that Bresnan describes. But given a richer context, we may be able to interpret the accent placement in a different way, and the syntactic relations may depend on other factors. That is, the syntactic relations are independent of the accent placement; given suitable contexts, it is possible to get both syntactic interpretations with both accent placements:13
(56) A: George didn’t have any idea he was supposed to follow Helen.
B: Whaddya mean--Helen left directions for George to follow!
(57) A: George feels pretty badly about ruining dinner, but the package didn’t have any directions.
B: I know, but--gee, didn’t you look over on the counter by the toaster? Helen left directions for George to fóllow.
In these cases, for George to follow in (56) and directions in (57) are deaccented to refer them to the broader context (i.e., speaker A’s sentence), and the accent shifts accordingly. But we still infer in (57) that George is supposed to follow the directions, while in (56) he is just supposed to follow, even though the accent placement is the reverse of Bresnan’s examples. Accent, as Bolinger maintains, is independent of syntax. Berman and Szamosi (1972:313) make a similar observation of an ambiguous sentence where “both readings are obtainable with either stress pattern.”
There is, to be sure, much for the skeptic to wonder about in the foregoing discussion. For example, I said that the measure in (20) had, in effect, become a shifter—referring to the bill sent to Congress . . . earlier in the paragraph—and I used that to argue for my characterization of deaccenting as a signal of contextual reference. But it would be difficult to maintain that all deaccented words are shifters in any obvious sense. In the case of John doesn’t réad books, for example, books still means ‘books’, and its interpretation can hardly be said to depend on the context. Yet the deaccenting still signals a kind of contextual reference; it implies a special relevance or relation of the deaccented part of the answer to the addressee’s question. “With reference to your question whether John has read Slaughterhouse-Five,” the speaker is saying, “you might want to know the following: John doesn’t read books.”
Or consider Schmerling’s account of the examples about the deaths of Truman and Johnson, quoted at length at the beginning of the chapter. When her mother announced Truman diéd, the deaccenting of Truman said: “This noun has some special relevance to the context.” Since (by Schmerling’s account) this was the first thing her mother said to her one morning, the reference to the context had to be something other than any immediately preceding discourse. But since Truman’s illness had been in the news, any hearer who had been listening to the news would, in Schmerling’s situation, assume that that was the contextual reference. Her mother was saying in effect: “I know that you know that Truman has been critically ill: well, he died.” Speaker and hearer were thus able to make sure that their fund of common knowledge and assumptions was still operative.
But suppose Schmerling had been busy finishing her dissertation and had not been following the news; then she might have been unaware of Truman’s condition. Then when her mother said Truman diéd, Schmerling’s reaction might have been “Oh, was he expected to?” or “Was something wrong with him?” or (depending on how early in the morning it was) perhaps just “Huh?” Schmerling would have interpreted Truman as deaccented, but would have been unable to figure out what its special contextual relevance was. Her mother, seeing her puzzlement, would have realized that she had presupposed too great a common context between herself and her daughter and would have backed up one square: “You knew he was sick, didn’t you?” This kind of misunderstanding happens all the time. The well-known phenomenon of cryptic conversations between married persons or friends who have known each other for a long time is based in part on the phenomenon of deaccenting to refer to a context to which the outsider has no access.
Now, in the case of John doesn’t réad books, the relevance signalled by the deaccenting of books is so obvious that the explanation seems forced. “Everyone knows Slaughterhouse-Five is a book,” the skeptic might say; “books is deaccented because the conversation is about a book.” But there are two replies to the skeptic. First is the obvious fact that books need not be deaccented in this context. The speaker could just as well have replied John doesn’t read bóoks, which seems to convey greater finality, or perhaps a nuance of surprise that the hearer was unaware of John’s reading habits. The conversation is no less about a book; but the speaker simply chooses to connect answer to question in a slightly different way. Second, consider a possible reply like John doesn’t réad trash. The conversation is not “about trash” in any meaningful sense. Rather, deaccented trash is referred to the context; the hearer is told “this word has some special relation to the context.” Given the context, of course, the special relation can only be the implication of the speaker’s critical evaluation of Slaughterhouse-Five. The grammatical device of deaccenting and its function are the same as in John doesn’t réad books; the difference is that in one case the special relation inferred by the hearer verges on the inappropriate (just answer the question, please), whereas in the other case it passes virtually without notice. I think that ‘automatic’ deaccenting of the sort in John doesn’t réad books plays a considerable role in keeping discourses on the track and helping interlocutors monitor potential misunderstandings—as did the deaccenting of Truman in our hypothetical embellishment of Schmerling’s example just discussed.
In any case, we have here another illustration of Bolinger’s dictum that “the error of attributing to syntax what belongs to semantics comes from concentrating on the commonplace.” It is easy enough to look at a dialogue like the Slaughterhouse-Five exchange, see the obvious fact that the dialogue is about a book, and conclude that there is a syntactic rule that deaccents books in the context. But the situation is more subtle than that. The rule—if we could call it that—says rather “deaccent something in order to signal its relation to the context.” Speakers use the rule in obvious cases to let the hearer know that they share certain basic knowledge about the world and that the conversation is running smoothly. But they also use it in other cases—or fail to use it—when they have some special effect in mind. The deaccenting ‘rule’ in the more usual sense of the word rule is simply the focus rule—accent goes on the most accentable syllable of the focus constituent—together with our knowledge of the relative accentability of different items in different contexts. These details in the operation of the rule, important though they may be, should not be confused with the reasons for which the speaker chooses to operate the rule in the first place.
The foregoing three chapters make two broad points of theoretical significance. First, I have discussed a variety of phenomena which can be seen as part of a single unified system once we accept a rhythmic/relational view of prominence like that proposed by Liberman and Prince (1977). Such a theory accounts for a wider range of experimental data than the simpler ‘accent’ view, which tends to look for cues to prominence on a syllable-by-syllable basis, and is forced to treat timing as a separate phenomenon; in a rhythmic account of prominence, timing and pitch obtrusion are among the concrete cues to a more abstract cognitive structure. Moreover, if we understand prominence relationally, we can unify a variety of phonetically disparate phenomena under a single rubric of ‘deaccenting’, which provides us with significant insights into the placement of accent. For one thing, it relieves us of the necessity of giving duplicate explanations for different manifestations of deaccenting (e.g., Schmerling’s ‘topic-comment sentences’ or Bresnan’s ‘elicitory questions’); for another, it enables us to see deaccenting and focus as opposite sides of the same coin if we treat accentability as relative within constituents.
The second important point, raised throughout Chapter 4, has to do with the balance between syntax and semantics in our grammatical descriptions. In a paper entitled “Free Will and Determinism in Describing Languages” (Bolinger, 1978b), Bolinger makes the point that one of the goals of much linguistic work has been to explain as much as possible of what speakers say in ‘deterministic’ terms: as the automatic, syntactically-specified consequence of other choices. Bolinger challenges this view (as he has elsewhere; e.g., 1972b, 1977), saying that it can be maintained only if we “concentrate on the commonplace.” By forcing ourselves to treat in our analyses the kinds of subtle distinctions that speakers are capable of making when appropriate (if unusual) circumstances arise, we find that many phenomena which seem, at first glance, to be syntactically determined, actually reflect some element of meaningful free choice on the part of the speaker.14 Specifically, in Chapter 4 I have treated accent placement from Bolingens general point of view. I have shown how the deaccenting of an item or constituent signals that it has some special relation to the context. The nature of this special relation is not spelled out by the grammar, however, but must be inferred by the hearer on the basis of the linguistic context, the situation, and the fund of shared assumptions between speaker and hearer. More generally, I have argued that neither focus nor deaccenting is a consequence or output of a syntactic rule. Rather, they are part of the meaning of the sentence intended by the speaker. The syntax tells us how to convey a certain focus or deaccenting, but it does not specify when.
We use cookies to analyze our traffic. Please decide if you are willing to accept cookies from our website. You can change this setting anytime in Privacy Settings.