ISSUES REGARDING UTTERANCE STRESS ASSIGNMENT (USA) IN ENGLISH.

 

INTRODUCTION.

 

   Nobody likes to read introductions and yet they feel obliged to do so, so as not to miss something of importance. I will therefore limit myself to a few words which I feel to be VERY important in that they will make this dissertation easier for the reader to understand.

   Before I began to study the subject matter that would enable me to write this dissertation, I had little previous knowledge on USA in English. However, it was exactly for this reason that I chose to write about USA. I felt that a English linguistic student should have a working knowledge of this subject since English is a intonational language and that this was even more important for me since I hope to teach English (as a foreign language).

   Therefore, I began this dissertation not in the position of having already studied in the field for a number of years but as a relative newcomer and this is reflected in the form of the dissertation.

    The idea was not to produce some extremely technical piece of work but to gain a sound understanding of the issues involved in USA.

    The first part of the dissertation is more general in nature and could be followed by someone with a minimun of linguistic training. It is in this section that I air for the mostpart my own views as they developed during the writing of the dissertation.

    The second part of the dissertation is much more theoretical in nature and begins by reproducing some of the more pertinent arguments as to why USA and its resulting dividing of a sentence into presupposition and focus is a surface phenomena. From the results of this discussion, we are in a position to go on to look at how certain surface constituents can be +focus, and how negation can focus on a particular constituent. In this section, we also look at other phenomenon related to USA: namely how we read a sentence with regards to co-referentiality.  

    Clearly then, this dissertation handles more than one single idea or observation regarding USA. Because of this, it makes reference to various terms often used by more than one linguist in slightly different ways. I have tried to be consistent but where I quote others, I didn't feel at liberty to change their words and so a certain amount of flexability is called for. The reader is credited with already being familiar with such terms and their variant usages and finer distinctions between two or more terms e.g. sentence v utterence, are only given where it is felt to be essential to the point in hand.  

 

 


 

SENTENCE V UTTERANCE STRESS

 

  Before entering into the main body of this dissertation, a few words must be said regarding the difference(s) between sentential and utterance stress.

  My initial ideas having read a number of papers was that sentential stress was a purely theoretical construct not capable of being proved (or perhaps more importantly disproved) since it is never given physical realization:the case being that the moment the sentence is said, it is no longer a sentence but an utterance.

  Sentential stress did allow one to speak of "the null context", but such a concept isn't ruled out by adopting utterance stress either. Even in an utterance, everything can be new.

   The idea of sentence stress also allows linguists to talk of "default accenting" and "focus accenting" from a different angle. It is inherent in the very idea of default accent, for example, that it does not go on the same item as a sentence accent would.

   Ladd seems to be of this persuasion although he doesn't explicitly say as much. See section on Deaccenting and Default Accent for example. 

   However, to even talk of an abstract entity being stressed was questionable for me at first and it was only quite late in the day that Noel Burton Roberts pointed out to me that since word stress was stored in the abstract representation, so too could sentence stress be.

  Utterance stress on the other hand looked a far more interesting and useful project. Different theories could be empirically tested and disproved and the context could be investigated so as to tell us why a speaker chose to stress one word rather than another.

  Therefore, I have chosen to pursue a topic that is utterance orientated as suggested by the title of this dissertation.

   Until recently, I would have said that no mariage was theoretically possible between models of sentence and utterance stress because of their very nature: abstract v contextual. However, I have now seen an attempt to link them within the same theory and I look forward to reading this paper in its final form. But for the time being, I have assumed that the situation is as I have said above. Utterance stress is an altogether more interesting topic to study.

 


PART 1

 

SEMANTIC V SYNTACTIC THEORIES OF UTTERANCE STRESS ASSIGNMENT (USA) ENGLISH.

 

 

   Chomsky and Halle's NSR and Bresnan's follow up of their work had in common that they predicted USA by syntactic means. 

   Indeed, at first glance, Bresnan's work looked promising since it apparently correctly explained well known exceptions to the Chomsky and Halle NSR, such as

 

   1) Helen left DIRECTIONS for George to follow.

 

   2) Helen left directions for George to FOLLOW.

 

   Bresnan maintained that in 1), "directions" is the underlying direct object of "follow" and therefore receives stress in the lower cycle with stress on "George" and "follow" consequently being reduced.  When "directions is raised, it is then the only contender for primary stress since the other potential contenders for primary stress have already been reduced.

   In 2), "directions" is quite simply not the direct object of "follow", and hence it is "follow" as the rightmost content word in the deep structure that receives primary stress in the first cycle and maintains it throughout the derivation.

   However, Bolinger soon pointed out that Bresnan had been misled by the commonplace and that counter-examples did exist.

 

   3) I have passages to MEMORIZE

 

   4) I can't finish in an hour. There are simply too many

      topics to ELUCIDATE.

 

   5) I need a light to READ by.

 

 

   On the basis of such examples, Bolinger favoured a semantic theory of USA.

 

 

   "The Chomsky-Halle Nuclear Stress Rule and its modifications by Bresnan, and to some extent the criticisms that have been levelled at it, have in common an attempt to account for accent in terms of syntax. Instead, accent should be viewed as independent, directly reflecting the speakers intent and only indirectly the syntax. Accented words are points of information focus.(Bolinger 633:01)

 

   But Bolinger is able to dispense with syntax so readily because of the nature of his work. For the most part, he merely collects "counter-examples" to syntactic theories and uses them as evidence for a Semantic theory of USA. (The assumption being that if counter-examples exist, then his case is proved)  Were he to look more closely at other related phenomena such as the focusing of constituents or the scope of indefinite pronouns and negation, then he would find that syntax is not a tool that can be dispensed with entirely since it facilitates description. Therefore, later on, we shall see that linguists who do indeed favour a Semantic approach to USA such as Jackendoff and Ladd, nevertheless make reference to the syntax. To understand their position, and to forward my own views on the matter, I wish to examine here the relationship between syntax and semantics in relation to USA.

   The advocates of Syntactic USA rules seem to me to wrongly beleive that to say that stress is placed on for instance Subject NP's or verbs rather than prepositions or pronouns is to give a reason. This is clearly not so. All it is is an observation. Since we can find no physiological reason why this is so, we must look elsewhere for explanations. This is what Bolinger was doing when he remarked

 

   "Bresnan 1971 carries the syntactic argument one step

    further by excepting pronouns as such from items that

    carry "primary stress". A semantic theory accounts for

    this by saying that pronouns are formal deictic elements

    that are semantically empty. (Bolinger 636:24)

 

   In the following discussion, it will soon become apparent that I am not using the word "semantics" in its narrower linguistic sense of "truth conditional" or even in its everyday sense of "meaning in general". I have often used the word "psychological" instead of "semantics" and this should be taken as an indication by the reader as to how a semantic theory of USA should be interpreted when studying USA. Semantic theories of USA include psycholinguistic factors.

   I believe that an understanding of the relationship between syntax and semantics in relation to USA as I propose above is essential for an understanding of what is to follow. Otherwise the linguists quoted may be accused of fluctuating between syntactic and semantic statements and consequent inconsistency.

    The difference between semantic and syntactic theories of the USA are better understood if we consider the ideas of another linguist.*(see note in references). He remarks that no syntactic category can arise if that category does not already have psychological reality.

   Therefore, Subject Noun Phrases, Pronouns, Prepositions etc can all be differentiated both syntactically and psychologically. But with regards to the question in hand of USA, I believe that we must turn to semantics/psycholinguistics for explanations. Why is this so?

   There are a number of reasons. Cross linguistically, stress (and other related phenomena such as intonation and tones) do seem prone to indicate different semantic meanings.

   Within English, there is a large amount of apparently systematic alternation of stress patterns, and this intuitively has some sort of semantic correlate.

   This is not to say that all alternations have semantic corollaries. We could look at the alternations in pronunciation of the word "economics" or the more fluid German word order where there is apparently no difference in the semantics. Each case must be taken on its merits.

   Where there is no alternation such as in the position of prepositions in English (or very little anyway since in the SS, we do find abandoned prepositions word finally although without apparent semantic import), it is easier to describe the rule syntactically. However, even in such cases, if we were to insist on an explanation as to why prepositions in English nearly always precede the noun, we would have to retreat to psychological reasons such as the facilitation of the processing of information with fixed word order. In the final analysis, syntactic explanations alone will never give us access to the deeper motivations behind human speech patterns; only the tools for describing a situation that exists.

    So when linguists argue whether the lack of stress on sentence final pronouns is due to the syntax or the semantics, the debate is largely pointless. All words can be referred to both syntactically and semantically and therefore incorporated into syntactic and semantic theories. But for the reasons alluded to above, I believe that with regards to USA, we would do better to advocate a semantic theory and keep syntax as a descriptive tool. Ladd reaches a similar conclusion

 

   "Our definition of neutral accent placement is based on the SEMANTICS of focus. Bolinger's formulation must 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." (Ladd 104:12)

 

   However, I would like to modify his one but last sentence to "... in which case the accent goes in a location most easily described syntactically".

 

 

 

 

NORMAL STRESS AND POSSIBLE SYNONYMS

 

   Whether we advocate a syntactic or semantic theory of USA, the question still remains as to whether we can talk of "normal" stress.

   Since I favour a semantic theory of USA, I will give here only possible semantic definitions  of normal stress.

 

   "Normal stress is the accent placement that would be used in the null context were this not a contradiction (since utterances can't be uttered in null contexts)

 

   However, such a definition is problematic since even if we take the simplest of sentences, write it down, and ask someone to read it, the reader is known to assign a context to that sentence and there is ample evidence that the assignment of context to the sentence depends largely on the readers individual beliefs and presuppositions. Gussenhoven demonstrates this and comes to the conclusion

 

   "What this means is that the concept of normal stress

    cannot reasonably be a part of a linguistic theory of

    Accent Assignment as it necessarily involves a prior

    interpretation of semantic material as either background

    or variable (-focus or + focus ) (Gussenhoven 387:10)

 

   Another possible way out is to say that

 

   "Normal stress is that stress pattern assigned to a sentence (which simultaneously becomes an utterance) which we assume holds entirely new information for us. i.e. all + focus".

 

   But Gussenhoven points out the inadequacies of such an approach.

 

   "Many sentences are excluded from having a full focus

    interpretation because their semantic material is too

    obviously part of the background. Also, sentences that

    include a focus governor cannot be given a full 

    (+ focus) interpretation either". (Gussenhoven 388:07)

 

  Therefore, the question we will ask here is whether there are any viable alternatives left defining "normal stress".

  Whether linguists expound a syntactic or a semantic theory of USA, they are essentially engaged in the same task; they are trying to stipulate the rules by which speakers assign stress to sentences just as phonologists try to stipulate the rules by which we realize phonemes as phones.

  Now, in phonology, no one would dispute that a speaker is quite capable of making a mistake e.g.

 

   "I need to gno [   ] the answer"           instead of

 

   "I need to know the answer"

 

  Drawing a parallel, let us imagine that certain rules for the USA do exist and may be blundered by the speaker. This doesn't seem counter-intuitive since we can think of examples ourselves where the sentence never seems acceptable e.g.

 

   Was it an ex-convict with a red shirt that he was warned

   to look out FOR. (Cleft sentences are remarked on later).

 

 If normal stress is somehow to be taken as synonymous with "grammatically correct" stress as defined by the rules, then it does exist. Examples as those above are hard to create since our imagination is such that we can create contexts that do accommodate what at first appears as totally unacceptable USA under any circumstances. Indeed, some readers may object to the above example on the grounds that they have indeed found a suitable context. (Here I can only imagine one example where the speaker is echoing, but this suffices.)

   Therefore, another possible definition of normal stress might be

 

  Normal stress is that stress pattern assigned to a sentence which (the stress pattern) is befitted to the context in general and the beliefs and presuppositions of the speaker"

 

   Hence, the second line in the dialogue below would appear deviant as per our new definition of normal stress.

 

  "Do you have six?"

  "No, I HAVE ten".  

 

   Therefore, if normal stress is taken instead as "context and presuppositionally" correct stress, it most certainly exists and examples proving this are easier to come by than those that show the existence of "grammatically" correct stress.

   But all this might seem a rather trivial point. If any rules for USA exist at all, then it is evident that they can be broken.

    Does the postulation of normal stress or some such similar notion allow us to gain anything else?

     Ladd seems to think that this is the case. Speaking of a sentence which other linguists call "contrastive", Ladd wishes to deny that the concept of contrastive stress is valid but yet wants to maintain that certain stress patterns intuitively are marked in some way.

 

   "Yet it is also obvious that there is something special

    about this accent placement which underlies everyone’s

    intuitions that it is different from "normal stress"

    (Ladd 102:01)

 

   The point is that one cannot talk of contrastive stress or any other sort of "special" stress without automatically implying a non-contrastive or non-special stress.

    Ladd certainly needs some concept of normal stress, otherwise he will not be able to talk about his special cases of "deaccenting and default accent" and "narrow focus".

   Perhaps the problem lies in the very word "normal". To some, it might suggest statistical prevalence. But it is dubious whether the majority of sentences are in fact uttered with normal stress if "normal" is to be interpreted statistically. Worse still, the concept of normal suggests that any deviance from normal stress is "abnormal" (the everyday antonym of normal). However, we clearly do not want to suggest that a speaker who does not utter a sentence with normal stress is in some way abnormal.

   Ladd chose the label "Neutral Accent Placement" for a situation that other linguists might call "normal stress" and summarizes this as follows.

 

   "If the focus constituent is the whole sentence, the

    accent may be said to be neutral and the focus unmarked.

    This case corresponds to "normal stress" in traditional

    descriptions,...." (Ladd 129:7)

 

   "Neutral Accent Placement" does seem a better title. The placement of the stress on the rightmost content word allows the sentence to be compatible with the greatest number of possible contexts (as is shown later with the sentence "Did Maxwell kill the judge with a hammer"). Yet, of the various contexts with which the sentence is felt to be compatible, one context does not stand out as being more likely than the others as long as we only consider the stress and not the presence of proper nouns etc and the actual context in

which the sentence was uttered. Neutral Stress Placement truly is neutral with regards to context.

  And of course the earlier undesirable connotations with statistics and/or abnormality are lost.

   Further points should be made. Sometimes, Ladd speaks of "broad focus"(Ladd 104:21) instead of "Neutral Accent Placement". This is because it allows a "broad" interpretation" i.e. the greatest number of possible contexts with which it is compatible. The antonym of "broad" is "narrow" and when a sentence is uttered with non-neutral accent placement it does indeed display something that could be described as "narrow focus". Sentences uttered with "narrow focus" always have a narrower range of possible contexts with which they could be compatible.

 

 

DEACCENTING AND DEFAULT ACCENT.

 

  In this section, I would like to take up another issue debated by Ladd in his chapter titled "Deaccenting and Default Accent" (Ladd 104).

 

   "To my knowledge, only Bolinger and Schmerling come close

    to formulating a corollary of deaccenting, which is that

    in order for an item to be perceived as deaccented, the

    accent must fall elsewhere, and it is thus possible for

    a word to be accented as it were by default".(Ladd

    105:25)

 

   So that we have something to work with, let us look at one of his examples (Ladd 104:29)

 

   A)  Has John read Slaughterhouse-FIVE?

 

   B)  No, John doesn't READ books.

 

   He remarks of this example

 

   "...we can say that the accent is on "read" merely to

    deaccent "books". This is not a focus accent but a

    default accent. We are not focusing on read; the

    positive, marked aspect of the accentual pattern is not

    that "read" is accented, but that "books" is

    deaccented".(Ladd 105:29)

 

   Let us reflect on both "deaccenting" and "default" accent in turn.

   Deaccenting can only take place if we are prepared to admit that another word carried the accent but "lost" it so to speak. However, there is never any evidence for such an event in English to my knowledge and such a theory is therefore unfalsifiable and consequently uninteresting. Those who favour it probably would make reference to "sentence stress" and "utterance stress" as two separate phenomena and where there is a supposed  difference, they would attribute it to the deaccenting of a word in the sentence and the appearance of stress elsewhere in the utterance.

   Default accent by contrast is always an utterance phenomenon since it implies that the context causes deaccenting of a word and that the accent is then placed elsewhere by default. It is this idea that I wish to challenge. To do so, I will assume that deaccenting does exist but this is theoretically questionable since one could object to an accent (a physical thing) which is context sensitive being assigned to a sentence (an abstract thing) which is removed of all context.

   If "books" is deaccented for the sorts of reasons normally forwarded by linguists e.g. repeated information, predictable, presupposed, given etc, then surely "read" should be deaccented for the same reasons. Indeed more so since the verb "to read" appears in the question (although in a different tense), while "books" has to be inferred from the question as the superordinate of the hyponym "Slaughterhouse-Five".

   The idea that a speaker deaccents a word for a certain reason(s) only to then place a default accent on another word which should be deaccented for the same reason(s) is intuitively weak.

   But it must be said that sentence B) doesn't sound unnatural in this discourse and so its accenting remains to be explained.

   The answer I believe is that "read" IS a focus accent and focuses on the constituent "doesn't read" while deaccenting "books" which is given/presupposed. (How stress on a word allows certain phrases to be interpreted as in focus is explained later but for now, the reader should intuitively feel that this is possible).

   The astute reader might then ask why B) doesn't then place the accent directly on "doesn't" since "read" is also given or presupposed. Briefly, there seems to be a rule in English that doesn't allow a negative to be stressed except when making a counter-assertion c.f. the example above with

 

   A)  I know that John reads books.

 

   B)  That's not true. John DOESN'T read books. 

 

 Furthermore, such a theory leads to a certain amount of ambiguity as Ladd himself noticed. Of the sentence

  

   "Harry wants a VW, but his girlfriend would prefer an

    AMERICAN car" (Ladd 109:10)

 

he says that it seems reasonable to talk of a focus on American but that

 

   "...this will not explain the deaccenting of "car". (Ladd

   109:15).

 

   Quite clearly though, "car" is given or presupposed in some way by virtue of "VW" while "American" isn't and this observation is enough to explain away the example without recourse to a default accent (whether you admit to deaccenting or not in principle). Of another sentence

 

   "Harry wants a Buick, but his girlfriend can’t STAND

   American cars". (Ladd 109:13)

 

Ladd suggests that we cannot say if "stand" is accented because it is the speakers wish to focus on this constituent (focus accent) or because "American" and "cars" are deaccented (default accent).

   I however, feel that such utterances are explainable solely with the concept of a focus accent. "Stand" is focused on because it opposes "wants" while "American" and "Car" are in some way given or presupposed and therefore poor contenders for stress.

  Ladd’s confusion led him to make a rather strange remark.

 

   "But armed with the concept of default accent, it is not

    hard to demonstrate with evidence from dialogue that the

    phenomenon of deaccenting includes both occurrence after

    the sentence accent - which gives the impression of

    greatly reduced "stress"- and occurrence before it- which

    does not".(Ladd 107:27)

 

  But how can this be so? Either an item is perceived as deaccented or not; or are we to allow various degrees of deaccenting?

   What Ladd has actually said if I may rephrase his words above, is that if the accent is moved to the right, the deaccented word is heard as less deaccented than if the accent is moved to the left. Why might this be?

   The answer fits in well with my earlier remarks on focus and constituents. If the accent is moved to the right, the original word that would have had the accent in the null context  may still be in focus while this is not the case if the accent goes leftwards.

   His conception of the utterance and its stress is really one of refocusing (broad and narrow focus) and not default accent. This explains why he perceives a difference between what he believes to be a default accent appearing to the left and right of the deaccented word. If default accent was truly disparate of focus, then it wouldn't matter where the default accent should occur. This however is clearly not the case.

   Finally, we should ask whether there is any similarities between the mechanisms that focus on a certain constituent and those that choose to deaccent a word.

   As said before, it is not here that we will try to define

the exact mechanisms but Ladd did allude to a list of possible reasons for deaccenting such as "repeated", "presupposed", "given", "predictable" etc.

   What then might be the reasons for focusing on a constituent? The reasons are the antonyms of the above terms i.e. "new", "unpredictable" etc. Clearly, there is a relationship here which can be expressed without recourse to two separate types of stress (focus and default).

   Therefore, I conclude here that there is no work for Ladd’s default accent  and that the idea of focus accent suffices. Cf.

 

   "From this, we can conclude that there are two

    fundamentally different types of non-neutral accent

    placement - narrow focus accent and default accent."

    (Ladd 106:24)

 

 

DEGREES OF ACCENTABILITY

 

   My remarks in this section will be surprisingly short considering the amount of debate that this question has caused. This is because the groundwork has already been covered in previous sections, especially "Semantic v Syntactic Theories of USA".

   There, I argued that words or phrases could be referred to semantically or syntactically and supporters of both semantic and syntactic theories of USA have indeed formulated their rules and ideas in such a way that certain types of words/phrases are more likely to be stressed than others. There seems to be general agreement though that prepositions, articles and pronouns are less prone to receive stress. However, the question as to whether there is a hierarchy between NP's (particularly in subject position) and VP's as to which receives stress is not so clear.

   Briefly reiterating what was said before, we must firstly note that there is a lot of alternation in English as to whether the NP or VP is stressed and it apparently has repercussions on the meaning. Therefore, if we are to seek explanations, and not just make observations, we would do better to look towards the semantics (psycholinguistics) and not the syntax. But this still leaves the hierarchy question unanswered.

  Schmerling (1972) believes that NP's do have a tendency to be stressed in preference to VP's.(although she believes the motivation to be syntactic.) She bases her argument on such examples as

 

   The KETTLE'S boiling

   The ICE has melted

 

   My KEYS have disappeared

   The PROFESSOR rang 

   

   The CHIMNEY-POT'S fallen off

   The BRAKES failed

 

  However, Allerton and Cruttenden didn't agree with her

 

   "The syntactic structure has a part to play if only

    a subsidiary one:it specifies the subject and the verb

    as candidates for the main accent and in the unmarked

    case the main accent will come on the verb as the right-

    most constituent (Allerton and Cruttenden 153:05)

 

   Cruttenden, by not apparently having the kind of understanding that I have outlined above between syntax and semantics, finds himself wanting the best of both worlds.

 

   "It may, as Bolinger suggests, be true that ultimately

    nucleus placement comes down to speaker choice, but

    there are too many regular correlations with syntax to

    say that it is totally irrelevant".(Cruttenden 86:31)

 

 Ladd for his part believes that nouns do attract stress rather than verbs.

 

   "This suggests not a dichotomy between content words and

    function words, but rather a HEIRARCHY 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. (Ladd 116:09)

 

  One of the problems for the uninitiated is that such statements presuppose that we can recognize the stressed item and the only task is to give a rule that forecasts/explains the sentential stress. But are we presupposing to much?

  

   "We noted that there is no one-to-one relation between acoustics and perception." (Cruttenden 5:28).

 

  Such a statement is not reflected in the examples throughout the literature which mark sentences (the sources of which are often left unrevealed) as having this or that USA.

It may well be that certain sentences said in certain contexts do not have a clearly definable stress pattern. Why then, have linguists been deceived into thinking that they have?

 

   "In particular, different types of segment directly

    influence the acoustics of a prosodic feature and such                                                                                            

    segmental effects have to be discounted by listeners

    interpreting the meaning of a longer prosodic

    pattern.(Crutenden 5:28)

 

  I fail to see why Cruttenden limits this statement to "longer prosodic features". It may be that short intransitive sentences as listed above hold their own pitfalls. Indeed, I can envisage certain problems for the speaker when uttering a short utterance rather than a long one.

   In uttering a long sentence with USA on the last content word, the speaker can clearly mark the beginning of the utterance as unstressed by a series of relatively level tones. In a short sentence, this device is not available to him. Even if the speaker intends the first word to be unstressed and attempts to attach pitch to the intransitive verb, the subject NP may be interpreted as being the onset of the pitch movement due to its close proximity. Underneath, I have listed a series of intended pitch movements on the verb and how they might be misinterpreted by the hearer.

 

 

   INTENDED                             MISINTERPRETED AS

 

 

 

 

PRE-NUCLEAR LOW + HIGH                        RISE

 

 

 

 

 

 

 

PRE-NUCLEAR LOW + HIGH FALL                  RISE FALL

 

 

 

 

 

                  

 

PRE-NUCLEAR MID + LOW RISE                   FALL RISE

 

 

 

 

 

 

 

PRE-NUCLEAR MID + RISE FALL                  FALL RISE FALL

 

 

   The emerging idea should be clear although I would rather not make claims that such patterns do exist. Some of them must however and many more combinations are possible. And the above represents just one possible point of confusion in the interpretation of intransitive sentences. I have yet to see any argument proving that all sentences must have a primary stress.

   In view of such considerations as above, it may not be surprising that a small residue of problems is left unsolved. While we cannot give a definitive definition of what constitutes stress, it seems ambitious to expect to always be able to recognize it.                    

                   

 

REASONS FOR FOCUSING.

 

   In a earlier section "Deaccenting and Default Accent", we briefly mentioned some of the possible motivations for the placement of focus accent although we never reached any conclusions.

   In this section, we will examine the question more closely, using a paper by E.F.Prince as our guiding light since she sets out the issue quite clearly and discusses a number of difficulties with various proposals.

   Princes article assumes that focus assignment has something to do with the "given/new" dichotomy. She offers three, more precise possible definitions of exactly how "given" and "new" can be defined and argues for the inherent superordinacy of one of the definitions over the others. As well as adding some of my own remarks and examples to the conversation, I have introduced a fourth possible definition of "given" and "new" which I felt should be treated since it is often cited in the litterature as a possibility.

 

Given as Predictability and Recoverability.

 

  Prince begins by saying that in speaker/hearer terms, this idea may be worded as

 

   "The speaker assumes that the hearer CAN PREDICT OR COULD

    HAVE PREDICTED that a PARTICULAR LINGUISTIC ITEM will

    or would occur in a particular position WITHIN A

    SENTENCE" (Prince 226:04).

 

    This idea seems to have come from some of Halliday’s earliest work where he makes no great distinction between the two words, treating them as two sides of the same coin.

However, I find it more correct to treat the two phenomena separately although they are interrelated.

   I shall interpret "predictability" in the sense of "clairvoyancy" and "recoverability" as "the replacement of elliptic items". Of the latter, we cant say very much since elliptic items cannot be stressed. However, ellipsis is used as a test for deciding whether an item is predictable and latter on I hope to convince the reader that this is theoretically wrong. But first, we shall examine predictability.

  By defining "predictability" in the sense of "clairvoyancy" pure and simple, I am trying to rule out examples of anaphoric and cataphoric reference. The basic difference between predictability and reference is that the former is not dictated while the latter is. That is to say that where predictability is concerned, you may predict item x but item y could have occured whereas with recoverability and ellipsis item x must occur c.f.

 

   I saw Robert leave        as opposed to arrive, laugh etc

 

   Robert and Brian left     Robert left (and no other inter

                                          pretation is poss)

                                                 

 

Of course, predictability must nevertheless be based on past discourse otherwise it becomes impossible to predict almost anything, but the difference between the terms should now be clear.

   However, the problem with "predictability" is that

due to the linear nature of utterances, such a model can  only really concern itself with items appearing to the right in a phrase  or clause. This is because it is normally impossible to predict how a speaker will begin a phrase/clause but once he has embarked on it, we may be able to make strong predictions as to how the speaker will terminate it. Brown and Yule speak of "Top Down Processing" with regards to the relationship between sentences (Brown and Yule 234) but the concept can also be applied intersententially.

   Hence we would expect that any USA rules based on predictability would tend to place the focus stress on the initial word of a sentence, this being rarely predictable. This is clearly not so.

   Another weakness of predictability a la clairvoyancy is its apparent inability to explain such sentences as

 

   Your TROUSERS are on fire

 

where the state of being on fire is rather unusual with regards to trousers and therefore may be said to be unpredictable.

   Let us now look at recoverability (and ellipsis). The idea here seems to be that although ellipted items can never be stressed, they can be used as a test for seeing if an item is given.

   Since ellipsis is normally anaphoric or cataphoric, the ellipted item must occur in the discourse or text. In this sense then, it is given. Conversely, the idea is that if an item cannot be elliptic, it cannot be "given"

Prince quickly shows how such a test leads to incredibly strange results. In both sentences (Prince 226) immediately below, "John" appears and is therefore given in some sense

 

   a)  Mary paid John and he bought himself a new coat.

 

   b)  John paid Mary and he bought himself a new coat.

 

However, only in sentence b1) can John be elliptic

 

   a1) *Mary paid John and bought himself a new coat

 

   b1)  John paid Mary and bought himself a new coat.

 

Since ellipsis is a test for giveness, Prince asks if we are to conclude that John is given in b) but not in a), since only sentence b1) allows ellipsis.

   The problem is that the constraints on ellipsis are syntactic and not dictated by anything to do with the given/new dichotomy. By definition, anaphoric and cataphoric reference seeks out items actually occurring in the discourse and in this sense such words are always given.

   A possible attempt to save the recoverability test is to say that sentence a1) does have a change of subject in the two clauses and it is this change which is unpredictable.

However, sentences such as the one below do not display any such change. "An account" is object in both clauses and yet ellipsis is blocked. 

 

   c) *I'll open an account if you'll open

 

Here, the reason can only be syntactic.

 

 

Given as saliency

 

Expanding on her first tentative proposal and its inherent weaknesses, E.F.Prince proposes a second possible definition of given.

 

   "The speaker assumes that the hearer has or could appropriately have some particular thing/entity/...in his/her CONSCIOUSNESS at the time of hearing the utterance". (Prince 228:30) The question now is how exactly an item can be deemed to be in the consciousness of the hearer at a certain moment.  In its most simplistic form, one could simply check back through the preceding discourse to see if a certain item has already appeared. If it has, it is "given". H.A.Halliday adopts this view with regards to certain grammatical processes.

 

"In any information unit that is non-initial in a discourse, recoverable information tends to be represented anaphorically, by reference, substitution or ellipsis. Ellipsis involves systematic  features having no realisation in structure and therefore having no potentiality of association with information focus: what is unsaid  cannot than otherwise be taken for granted. By <reference> is meant here  the anaphoric use of what are essentially items of extralinguistic, situational reference such as pronominals and demonstratives, as in does John rent this house?-no, he's bought it. Substitution involves those items that are essentially text-referring  like one and do: Has anyone seen the play?- I think John has done.  From the point of view of information structure WH- items are reference items when interrogative and substitutes when relative." (Halliday (Journal of Linguistics 1967-68 206.)

 

Since we are now working with a definition that demands that a “given" item is assumed to be in the CONSCIOUSNESS of the hearer by the speaker, it is important to ask "How long does givenness last?"(Chafe 32:22) Chafe comes to the rightful conclusion that

  

 

    "...givenness is a status decided on by the speaker, and that it is fundamentally a matter  of the speakers belief that the item is in the addressee's  consciousness...."  (Chafe 32:36)

 

 

    This in turn would be influenced by such factors as "the number of intervening sentences" (Chafe 32:48) or "the effect of such discourse boundaries as a change of scene." (Chafe 33:02)

    Topic entities, by simple repetition, may achieve a higher degree of givenness as observed in the quote below.

  

 

 "Alternatively, in other types of discourse, there may be an overwhelming tendency to reserve pronouns for reference to <topic entities> in the discourse" (Brown and Yule 221:26)

 

 

    Other factors seem also intuitively to be possible. Perhaps NP's remain given longer than VP's etc, while explicitly formulated NP's may retain their given status longer than implied NP's.         

 

Given as theme

 

Unlike the above proposals, this definition is "clause bound" in that any clause can be assigned a theme/rheme status irrespective of past discourse and/or predictability.

 

    " Basically, theme is what comes first in the clause; and while this  means that there is in the unmarked case an association of the theme  with the given, the two are independent options.  The difference can  perhaps best be summarized by the observation that while <given> means  <what you were talking about> (or <what I was talking about before>),  <theme> means <what I am talking about> (or <what I am talking about now>); and, as any student of rhetoric knows, the two do not necessarily  coincide. (H.A.HALLIDAY 212;05)

 

   Such a model allows some interesting observations. Hence, D.J.Allerton has pointed out that the speaker, in choosing a certain NP  to appear at the beginning of a clause, is granting it the status of  'psychological subject' (i.e. theme). He also observes that Greenberg’s 1st Universal is 'subject before object' c.f. SVO, VSO, SOV (all unmarked in the surface). Therefore, the 'theme' and the 'subject' of a clause are often the same.  But beyond this simple observation, the model runs into serious difficulties. The crux of the matter seems to be that  theme is mapped  onto structure and the latter is moulded by other factors e.g. syntax, transitivity etc (as discussed by Halliday 218-223)

   Therefore, we may find our choice of "theme" dictated at times by such considerations. This becomes apparent when we look at languages other than English where the first morpheme, word, phrase or clause cannot occur initially in the corresponding English sentence. If the first item truly represents the "psychological" starting point, then we seem justified in inferring that all humans do NOT have the same general "psychological make-up."  We can do two things at this point to extricate ourselves from this quandrum. We can state that "theme" has as much (if not more) to do with syntactic structure of sentences than psychological processes as does Halliday in the above quote, or we can state that languages do indeed mould our "psychological" processes. Neither question is tackled here. The general point has been to show that the debate regarding "theme/rheme" has shed but little light on the question of USA. The "theme/rheme" model has a closer association with the "topic/comment" dichotomy. ( I take the  latter model here as that described by Charles N. Li and Sandra A.  Thompson (Subject and Topic). 

 

 

Given as shared knowledge (presupposition)

 

  Prince labelled her final definition as "shared knowledge" but I will henceforth treat it as "presupposition" since this title is more common in the literature.

 

   "The speaker assumes that the hearer "knows", assumes, or can infer a particular thing (but is not necessarily thinking about it) (Prince 230:34)

 

 

   Jackendoff offers us a more exact definition of what this may mean in reality and how it is obtained (although the model below concerns itself with true assertive statements only, alterations having to be made for imperatives and interrogatives.  His procedure works in three steps, the first of which is

 

   ".... focus assignment derives two formal objects from the otherwise determined semantic representation SR of a sentence S. The first, the focus, consists of that semantic material associated with surface structure nodes dominated by the marker F. The second is a one-place predicate presupposition (x) formed by replacing the focus by an appropriate semantic variable x in SR." (Jackendoff 245:14)

 

   His second step is designed to rule out false presuppositions and ties his formalism to "the particular universe of discourse" of which the speaker might be talking.

 

    "The second step in the derivation constructs another formal object, the presuppositional set, defined as the set of values which, when substituted for x in presupposition (x), yield a true proposition".(Jackendoff 245:29)

 

   His third step is the merging of the presupposition and focus to give us the assertion of the sentence. Of course, the focus must be a member of the presuppositional set and the presuppositional set itself is constrained in certain ways e.g. is a coherent set and amenable to discussion.

   The rather heavy formalisms above can be exemplified as chose Jackendoff to do. Taking the sentence

 

    John LIKES Bill

 

the presupposition seems to be

 

    the relation between John and Bill is (x)  

  

(by replacement of the focus by an appropriate semantic variable (x) in SR.

 

and the presuppositional set (which is represented by the lambda notation  ) is

 

   

          the relation between John and Bill is x  is 

 

and as said, the focus must be a member of the presuppositional set

 

  

     like    x  the relation between John and Bill  is x         

 

   Returning to Prince's definition, it is noticeable that her final definition (shared knowledge) is more extensive than the others. Prince herself noticed this.

 

   "Although different from one another, predictability, saliency, and "shared knowledge" are not mutually independent: If a speaker assumes that the hearer can predict that some particular item or items will occur in some particular position within a sentence, then the speaker must assume that it is appropriate that the hearer have some particular thing in his/her  consciousness. And if the speaker assumes that the hearer has some particular thing in his/her consciousness, then the speaker must assume that the hearer has some assumption or can draw some inference". (Prince 231:29)

 

   Her conclusion is that her first two definitions are somehow incorporated by "shared knowledge"

 

   "Thus an understanding of giveness in the sense of "shared knowledge"........ is germane (and, perhaps , prerequisite ) to an understanding of giveness in the other two senses....(Prince 232:17)

 

  Consequently, all items that aren't "presupposed" (assumed by the speaker to be shared knowledge ) are liable to take USA. In the final analysis, this is in my opinion demonstratably wrong.

   An infinite number of sentences are possible where a "presupposed" item does receive USA in order to indicate some sort of emotional state, opinion etc.

  

    She RAN to the shops    (said mockingly of a fat woman)

 

    ROBERT couldn't have cooked this meal (said  

    disparagingly of Robert)

 

    He's going to see JANE tonight ( JANE being said in such

    a way as to indicate that he loves her)

 

   That emotionally charged words are stressed shouldn't surprise us. It seems to be a universal fact that the pitch in the human voice is raised when the speaker is in an emotional state. The problem is incorporating the phenomenon indicated above by the sentences into a coherent and correct theory of USA: something which Prince's definitions (and others) don't do. 


 

PART 2.

 

THE RELATIONSHIP BETWEEN DEEP STRUCTURE (DS), SURFACE STRUCTURE (SS), USA, AND FOCUS/PRESUPOSITION.

 

         Having now sketched the outline of the presupposition/focus model, I wish to look at various aspects of it in greater detail.  The most useful debates seem to have stemmed from Chomsky in his article "Deep   Structure, Surface Structure and Semantic Interpretation".                      He begins by asking whether presupposition/focus is present in the Deep Structure (DS) as well as in the Surface Structure (SS). Since, up until recent times, the DS and perhaps the intermediate Underlying levels (UL) were believed to account for a great deal of the semantic import of an SS, it seemed only reasonable to expect to find presupposition/focus in the DS. However, on the other hand, we have just seen that presupposition and focus in the SS are apparently demarcated by stress assignment. The latter is of course a phonological phenomenon and can't be represented  in the DS which is syntactic and autonomous to the phonological  component.  If presupposition/focus is to be represented in the DS, it would have to be done syntactically. Chomsky used a number of famous examples to develop the argument including

 

3)   Was it  a) an ex-convict with a red SHIRT It     

     wasn't

             b) a red shirted EX-CONVICT

              

             c) an ex-convict with a shirt that is RED

 

      that he was warned to look out for?

 

   and its proposed US

 

3US)  The one he was warned to look out for was X

 

 

If it’s DS that determines presupposition/focus, then from the presupposition of all the sentences in 3 i.e.

 

   3 P)   He was warned to look out for someone.

 

 

 we can state that the foci of 3a ,b & c are respectively 

 

 

   4)   a)    An ex-convict with a red SHIRT

        b)    A red shirted EX-CONVICT

        c)    An ex-convict with a shirt that is RED

 

Moreover, 3a,b, & c can all be answered by the response

 

   5)   No, he was warned to look out for an AUTOMOBILE 

        salesman.

 

This isn’t surprising since 4a,b & c are paraphrases of each other.

 

But now Chomsky invites us to consider the possible responses

 

   6a) No, he was warned to look out for an ex-con with a

       red TIE.

 

   6b) No, he was warned to look out for a red-shirted

       AUTOMOBILE salesman

 

   6c) No, he was warned to look out for an ex-convict with

       a shirt that is GREEN.

 

Clearly, 6a) is the only natural reply to 3a), 6b) is the only  natural reply to 3b) and the same holds for the relationship between 6c)  and 3c). Chomsky concludes that the above pairings and their lack of  interchangeability are due to "the alternative conception of  presupposition and focus as determined by the intonation centre of surface structure. (Chomsky 201:26) Sometimes however, there is a relationship between DS, the USA and presupposition/focus. R.S.Jackendoff says

 

   "The strongest statement that we can make is that in cleft sentences, the focus must be included in the clefted phrase, and that the presupposition includes the lowest clause." (Jackendoff 233:04)

 

and this is illustrated by an example taken from Chomsky and modified by  Jackendoff.

 

7)  It was an ex-convict with a red shirt that he was warned 

    to look out for

 

I have marked the possible foci conforming with Jackendoff's  statement. Although then, there is a tenuous relationship between DS, SS, the USA rule and Presupposition/Focus in English, Chomsky points out that  equally, there are other sentences where the DS clearly has nothing to do with the presupposition/focus in the SS and that it is the USA rule and SS together that decide the presupposition/focus exclusively. This is well illustrated where a focused phrase would not correspond to a phrase in the DS at all. Taking the sentence

 

   8)  Is John certain to win?

 

and the natural responses

 

   9a)  No, John is certain to LOSE

 

   9b)  No, John is likely not even to be NOMINATED

 

   9c)  No, the election will never take PLACE

 

9b) would seem to be an answer to a question where "certain to win” is taken as the focused phrase.  But if we accept the DS of 8) as

 

8DS)   [  John win  ]    is certain

 

we can see that "certain to win" is not a DS constituent. In essence then, what is needed is a formal rule stating the relationship between SS, the USA rule and presupposition/focus. Jackendoff suggests the following.

 

"One artificial construct is required: a syntactic marker F which can be associated with any node in the surface structure".(Jackendoff 240:18)

 

He continues

 

"Two systems of rules will make use of the marker F, one in the semantics and one in the phonology. The former system contains at least the rule of focus assignment, which we will present here in preliminary form.

 

(Focus assignment, first approximation)

 

The semantic material associated with surface structure nodes dominated by F is the focus of the sentence.  To derive the Presupposition, substitute appropriate semantic variables for the  focused material". (240:30)

 

 

THE FOCUSING OF CONSTITUENTS     

 

Let us now turn away from the discussion of SS, DS & Semantic interpretation in order to discuss another separate but related topic which, up to now has been implicitly assumed whenever necessary in the examples.

   This is the relationship between USA, syntax and the focusing of constituents. We have not as yet given a formal statement as to how a certain stress assignment can mark one part of an utterance as +focus and consequently the rest as presupposition (-focus). Jackendoff states that

 

  "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 237:09)

 

   Let us now see how this rule of stress assignment functions. Using the phrase structure diagram produced below

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  Jackendoff points out that if the speaker wants to stress NP1 or 2, then, since the

 

    "...articles and the preposition are treated as

     proclitic by the stress rules..."(Jackendoff 238:06)

 

   the stress will fall on the main stress of whichever noun he wishes to stress.

  Should the speaker wish to stress VP1 or S1, "hammer" will receive stress in both cases since

 

  "..the Nuclear Stress Rule puts [ 1 stress ] on the final

   constituent."(Jackendoff 238:12).

 

   The emerging pattern is clear to see but difficult to put into words. Stressing the rightmost content (as opposed to grammatical word) in a P, causes that P to be in focus. Furthurmore, we can move up the phrase structure to superordinate Ps (reaching even S) and say that these Ps are also in focus so long as they don't govern any material that appears to the right of the word that was stressed.

   The ambiguity that this might cause as in the case where "hammer" is stressed and consequently NP3 alone, or NP3+PP1, or NP3+PP1+VP1,or NP3+PP1+VP1+S (ie S ), is partly disambiguated in Jackendoffs mind by

 

   "...further subtleties of stress and intonation"   

       (Jackendoff 238:17)

 

   One more observation needs to be added here. If we were to stress the word "judge" in the above example, then the focused P(s) could be NP2, or NP2+VP1. But the second option is less likely. Why?

 

   "... that when expressive or contrastive stress shifts intonation centre, the same principle applies as in normal cases for determining focus and presupposition, but with the additional proviso that naturalness declines far more sharply as larger and larger phrases containing the intonation centre are considered as a possible focus". (Chomsky 204:44).

 

 

 

 

STRESS, FOCUS, AND NEGATION.

 

  We have now seen a number of arguments which suggest that certain semantic interpretations (meanings) are expressed only in the SS. We can now look at further evidence which supports this conclusion while making us aware of yet another phenomena involving stress. Consider the sentence

 

   Maxwell didnt kill the judge with a silver hammer.   

   (Jackendoff 257:25)

 

   and the possible conditions under which this denial would be true i.e.

 

   Not Maxwell &/or not kill &/or not silver &/or not (the) judge.

 

   If any of the above situations hold, then the sentence is true. This is why some linguists would say that negative sentences are vague rather than ambiguous.

   However, if the sentence is said with main stress on the word, let us say, Maxwell, then we seem justified in inferring

  

   Someone killed the judge with a silver hammer.

 

   It wasnt Maxwell that killed the judge with a silver

   hammer. (Not Maxwell)

    

 

   Similar inferences can be made if the stress should fall on another word instead. The stressed constituent seems to come under the focus of negation to the exclusion of all others.

   Why should we believe that such differences in meaning are not represented in the DS ?

  Well firstly, we would have to have as many DS's as there are surface interpretations.( At least five for the above sentence).

   Secondly, the original ideas behind DS meant that we should be able to formulate a clefted construction expressing the + focus and -focus(presupposition) constituents clearly. This is not always possible though.

 

   *It wasnt(a)hammer that Maxwell killed the judge with a

    silver (one(of)).(Jackendoff 255:14)

 

   The above arguments then not only confirm our earlier conclusions on DS,SS and Semantic Interpretation; that its SS and USA that carries presupposition and focus, but they also illustrate something new for us. Namely that certain linguistic phenomena, (negation in this case), can focus on certain words or constituents.

   As before, we can find both syntactic and semantic constraints on what words/constituents can come under focus.

In a sentence such as

 

  "I know that your sister doesn't like the chocolates"

 

  we never seem able to infer

 

  a) not "I"         and/or

 

  b) not "know".

 

  however we choose to stress the utterance. The material in the superordinate clause can never come within the focus of "not" which appears in the subordinate clause and this seems best described as a syntactic constraint.

   More interesting though, Jackendoff seems to have found a type of example where the focus of negation is constrained by semantic considerations rather than syntactic ones. He points out that sentences with focused instrumentals such as

 

   Maxwell didn’t kill the judge with a silver HAMMER

 

  cannot have a reading where the instrument is focused syntactically (by clefting) but is not associated with the negation (because it appears to the left of the neg) e.g.

 

  *It was with a HAMMER that Max didn’t kill the judge.

 

cf  It was with a hammer that Max killed the judge.

 

  He concludes that the ungrammaticality of the above sentence is because it makes little sense to talk of using an instrument in order for not having done something. That the ungrammaticality of this sentence is pragmatic rather than syntactic is suggested by the fact that the similar sentence without negation is acceptable. Such initial observations lead to other findings. If the focused instrumental must be associated with negation, then the ungrammaticality of

 

   *With a hammer, Max didn’t kill the judge.

 

 is no doubt due to the fact that the preposed constituent is outside the scope of negation (for syntactic reasons alluded to above).

 

cf   With a hammer, Max killed the judge.

 

where preposing of the PP is allowed. Hence, Jackendoff has shown that

 

   a) Semantic interpretations depend on SS as well as DS

 

   b) That negation focuses on certain words and

      constituents and that this focusing is constrained by

      certain syntactic and semantic considerations.

 

 

STRESS AND COREFERENTIALITY.

 

  In the previous section, we saw that the assignment of USA could affect the reading of a sentence with regards to the focus of negation.

  We shall now see that the assignment of USA can affect the reading of a sentence with regards to the coreferentiality of nouns.

  However, the question of coreferentiality of nouns in English isn't settled merely by USA. Other components of the grammar often have an overriding influence in constraining or suggesting the interpretation of a sentence and the function of USA in connection with coreferentiality seems is the disambiguation of sentence strings where these other grammatical components do not make the choice for us.

   Before giving examples of cases where USA indicate for us which reading of a sentence is the speakers intended one, let me just illustrate for the reader cases where the semantic component of the grammar or pragmatics constrain or strongly suggest the intended reading (the difficulties of distinguishing between semantic and pragmatic inferences is of little importance generally in this paper but I believe the first two examples beneath to be semantic and the last two pragmatic)

 

   1)  The cup fell on its microchip.  

 

       ( "Its" can't refer to "cup" since cups don’t have

         microchips)                                              

 

 

   2)  The woman has just given birth to her first child.

 

       ("Her" refers to "the woman" since you cant give

        birth to another persons child, matters of surrogate

        mothers etc left aside)

 

  

   3)  She took her trousers off.

 

       ("Her" refers to "she").

 

   4)  During the fight, he took a knife to his throat

 

       ("His" refers to someone other than "he")

 

   Akmajian and Jackendoff (A&J) in a short squib and discussion suggest that "contrastive stress" prohibits coreferentiality and furnish us with many examples divided into three cases. The first case is where "contrastive stress on either a pronoun or noun will prohibit co reference" (sentence 5-9)

 

 

    5)   After he woke up, JOHN went to town.

 

    6)   After HE woke up, John went to town.

 

    5) & 6) are understod as not being co referential c.f.

 

    7)   After he woke up, John went to town. (ambiguous)

  

    8)   John hit Bill and then George hit him (him = Bill)

 

    9)   John hit Bill and then George hit HIM (HIM = John)

 

   The second case is where "contrastive stress" is necessary to get any reading at all"

 

    10)   John hit Bill and then HE was hit by Ira

         (He = John) c.f.

 

    11)  * John hit Bill and then he was hit by Ira

 

  Before examining the third case, let us discuss the data up to now.

  Starting with the second case, we can find syntactically parallel sentences where there is no need for contrastive stress to be assigned in order to derive a grammatical sentence e.g.

 

    11a)   John hit Bill and then he was arrested by the

           police

 

  The pragmatics of the sentence seems to be enough to allow correct coreferentiality to take place (He = John) without further aid from contrastive stress. Indeed, if the above sentence had stressed "he" as sentence 10, it probably would have been interpreted as "HE" = Bill e.g.

 

    10a)   John hit Bill and then HE was arrested by the

           police 

 

   The reasons for the "discrepancies" between 10,11 & 10a,11a respectively is that contrastive stress does not indicate coreferentiality directly but via the intermediary of pragmatics.

   Grice’s Maxims of Quality and Relevance apply and all other things being equal, we will apply them in order to get

a certain reading from a sentence (hopefully, the one intended by the speaker since he too assumes Grice’s Maxims to be operative).

   But when there is an indication that the Maxims are momentarily dispensed with, (by the placement of contrastive stress as in 10 where an active subject becomes a passive subject) or there is an indication that our normal pragmatic reasoning should be momentarily dispensed with (as in 10a where the victim is arrested ) we read the sentence contrary to normal pragmatic procedure.

   In their first case, the normal pragmatic procedure is to read the sentence co-referentially but stress appearing on either the noun or pronoun warns us that this norm is suspended. However, where the sentence has only one noun and pronoun e.g. 5 or 6, it is impossible to construct counter-examples where contrastive stress suggests a co-referential reading.

   With regard to Akmajian and Jackendoff's third case, I shall say very little since I'm not sufficiently familiar enough with theories of pronominalization and the construction of cleft sentences. However, their observations that pronominalized items and proper nouns in certain circumstances can't be co-referential 

 

   12)  *It was John's dog that he bit.  * Denotes non-

                                           Coreferentiality

  

are by no means without doubt attributable to constraints on pronominalization and/or coreferentiality exclusively. The grammaticality of the sentence below (13) suggests that the constraint might rather have something to do with theories of relativization as proposed by Keenan and Comrie 1972

since 13) is exactly like 12) in that the pronoun is both preceded by and commanded by a proper noun and yet only 13 can be co-referential.

 

   13)   It was John's dog that bit him.

 

   With regards to sentences such as

 

   14)  *In his apartment, Mary kissed John.  (* A&J)

 

   15)   In his rotting hovel of an apartment on Scrac

         Street Mary finally broke down and kissed him.

 

   I'm not convinced by Akmajian and Jackendoff's explanation.

 

   "We think that the relevant principle in these cases is that as the sentence becomes longer, the relative stress level of the NP's and pronoun decreases with respect to the clauses containing them. This would seem to relate to Case 1, where contrastive stress  precluded coreferentiality".(Akmajian and Jackendoff 126:22)

 

  Firstly, I have shown that contrastive stress doesn't directly give us a reading of non-coreferentiality but rather warns us against assigning a normal pragmatic reading to a sentence.

  Secondly, Akmajian and Jackendoff seem to be proposing a multi-level stress system with their comments above. This is problematic and has been dismissed by other phoneticians when discussing intonation.

  Finally, I can make a tentative proposal as to why 14) is ungrammatical and 15) isn't. The fronting of certain constituents of a sentence that aren't normally clause initial is often carried out to "highlight" them and is therefore accompanied by receiving a focus stress. (This process is often referred to as "topicalization" or as a marked theme) Sentence 14) immediately becomes grammatical if we assign such a stress focus e.g.

 

   14a)  In his APARTMENT Mary kissed John.

 

   We could equally assign focus stress to "his", but without a context to help, the reading is ambiguous since there is a conflict of interest between the pragmatic rules I have referred to above and the rule of focus stress.

   The pragmatic rules would say that stressing "his" warns us not to read the sentence co referentially as would normally be the case i.e.

 

    14b)  In HIS apartment Mary kissed John. ("His" is

          someone other than John who is in the discourse).

 

   But the rules of Topicalization allow "His" to be stressed without putting any such constraints on coreferentiality. e.g.

 

    14c)  In which apartment did they kiss?

        

          In HIS apartment Mary kissed John. (John's)

 

   The more we find that we need recourse to pragmatics, the more certain it is that Akmajian and Jackendoff's examples cannot be solved by rules of pronominalization and coreferentiality alone. A major problem on finding an ungrammatical sentence is deciding EXACTLY why that sentence is ungrammatical. If we can subsume any reasons under more general principles, then we are probably more correct in our analysis than were those before us. USA does have a bearing on co-referentiality but probably via the intermediary of pragmatics. 

 


 

LIST OF REFERENCES

 

AKMAJIAN A. AND JACKENDOFF R. (1970) Coreferentiality and  

                                     Stress (sqibs and

                                             discussions)

                                     LI 1  

                                    

 

ALLERTON D.J. AND CRUTTENDEN A. (1978) Three reasons for

                                       accenting a definite

                                       subject JL 15-16

 

BOLINGER D. (1972) Accent is predictable (if you're a mind

                   reader LG 48

 

BRESNAN J.W. (1971) Sentence Stress And Syntactic

                    Transformations

                    LG 47

 

BROWN G. AND YULE G. (1983) Discourse Analysis CUP                                                                    

 

CHAFE W.L. (1976) Giveness, Contrastiveness, Definiteness,

                  Subjects, Topics, and Point of View.

                  Subject and Topic ed Li C.N.

                 

CHOMSKY N. (1971) Deep structure, surface structure, and

                  semantic interpretation

                  ed Steinberg D.D. and Jakobovits L.A.

                  CUP

 

CRUTTENDEN A. (1986) Intonation CUP

 

 

GUSSENHOVEN C. (1983) Focus, Mode and the Nucleus JL 19       

 

HALLIDAY M.A.K. (1967) Notes on transitivity and theme in

                       English (Part 2 ) JL 3-4

 

JACKENDOFF R.S. (1972) Semantic Interpretation in Generative

                       Grammar MIT Press

 

LI C.N. AND THOMPSON S.A.(1976) Subject and Topic: A New

                                Typology of Language

                                Subject and Topic ed Li C.N.

                                Academic Press

                        

PRINCE E.F. (1981) Toward a Taxonomy of Given-New

                   Information

                   Radical Pragmatics ed Cole P. Academic

                                                 Press

 

SCHMERLING S.F. (1972) A Re-examination of "Normal Stress"

                      LG 50

 

* The remarks regarding "psychological reality" and categorization are not my own but something that I came across while reading on another topic. Unfortunately, I haven't been able to retrace the work to give the author full credit.