Contrastive Focus vs. Discourse-New: Evidence from Phonetic Prominence in English
1. Introduction This article presents evidence from English that the theory of grammar makes a distinction between the contrastive focus and discourse-new status of constituents. The evidence comes from a phonetic investigation which compares the prosody of all-new sentences with the prosody of sentences combining contrastive focus and discourse-new constituents. While the distribution of pitch accents and phonological phrase organization in these different sentence types is the same, their patterns of phonetic prominence duration, pitch and intensity-- vary according to their composition in terms of contrastive and/or new constituents. These distinctions in phonetic prominence are plausibly the consequence of distinctions in the phonological representation of phrasal prosodic prominence (stress) for contrastive focus and discourse-new constituents in English. 1.1 Background Certain central generalizations about the patterns of pitch accenting found in sentences with different types of information structure in English are widely agreed upon. For example, in most standard varieties of English, in an all-new, out-of-the-blue utterance spoken in a context with no shared prior discourse, all noun phrases necessarily carry pitch accents, regardless of position in the sentence (see Gussenhoven 1983 et seq, Selkirk 1984 et seq, Rochement 1986 and others). Imagine a situation where family members have just sat down to an evening meal together and one asks whether anything newsworthy has happened that day. Responses like those in (1) would contain constituents that are discourse-new1 and the pitch accenting shown would be appropriate:
All-new sentences like these which contain no (contrastive) focus within them have been widely referred to as broad focus sentences. For reasons to be explained in section 1.3, this term (due to Ladd 1980) is not used in this paper. 1
Has anything newsworthy happened today?

A. B. C.

Elza miled the cramels. I rad that the Tmes is offering nwspaper subscriptions to the por. Wttgenstein spke to nscombe at the fculty meeting.
(In this paper obligatory pitch accents in the sentences are marked by acute accents, optional pitch accents by grave accents.) Its been proposed that the locus of obligatory pitch accent in such cases is the position of phrasal stress prominence (Ladd 1980, 1996, 2008, Selkirk 1984, 1996, Truckenbrodt 1995, Calhoun 2006, 2010, Fry and SamekLodovici 2006, Kratzer and Selkirk 2007), though other scholars adhere to the view that there is no such thing as phrasal stress prominence and see pitch accent distribution within the sentence as defined independently (Bolinger 1961, 1972, Gussenhoven 1983, 1991). The empirical picture in English is radically different when sentences like those in (1) are uttered in other types of discourse contexts. Consider the case where the sentence in (1A) is uttered as a correction to a previous statement, as in (2BDF). (2) A. Srah miled the cramels. B. No, Elza mailed the caramels. C. Elza te up the cramels. D. No, Elza miled the caramels. E. Elza miled the pster. F. No, Elza miled the cramels. The correction sentences (2BDF) have semantic/pragmatic properties that are quite distinct from the all-new, out of the blue sentence (1A). Each of the corrections

Focus+given scenario: In a conversation on the topic whether American academics take normal vacations, the college personnel officer who is present remarks: Only AmndaFoc takesG a normalG vacationG. (Why is she different?)
In both (5) and (6) Amanda carries an obligatory pitch accent; the presence of this pitch accent reflects the contrastive focus status of Amanda, which is associated with the focussensitive particle only. In the case of (6), though, there is no pitch accent within the following predicate phrase, due to its discourse-given status (indicated with G-marking, and by italics). In (5), the discourse-new predicate phrase receives the pitch accenting it would normally receive in a simple all-new sentence lacking contrastive focus. Compare (5) to the all-new sentence in (7): (7) New-new scenario: First line in a short story about Amanda. Amnda tkes a nrmal vaction. She takes two weeks in the summer, either at the ocean or on a lake. The distribution of pitch accents in (7), which consists only of discourse-new constituents, is identical to that in (5), where discourse-new material is preceded by a Focus constituent. We will see, though, that a Focus-new sentence type, illustrated here with (5), is distinguished from a new-new sentence type, like that in (7), in patterns of phonetic prominence. The paradigm of sentences for our experiment includes minimal triplets of the type schematized in (4). The members of the minimal triplet differ in the Focus vs. new status of two syntactic phrases in a sentence: Focus-new, new-Focus and new-new. As shown below in sections 4-6, there is a three-way distinction in patterns of phonetic prominence
in these three sentence types, so these findings provide support for the hypothesis that the grammar does indeed make a distinction between contrastive focus and discourse-new. 1. 3 Other experimental investigations of the Focus-new distinction in English The current experiment is by no means the first to compare the prosodic properties of allnew sentences with sentences containing contrastive focus in English. But earlier studies differ from ours in employing sentences in which the material surrounding the (putative) contrastive focus is discourse-given, as in responses to alternative questions (Cooper et al 1985), correction statements (Breen 2007, Breen et al 2010) and responses to whquestions (Eady and Cooper 1986, Xu and Xu 2005, Breen 2007, Breen et al 2010). The sentence types compared in this work have been of the following schematic types (analogous to the all-new sentence (1a) vs. contrastive focus sentences (2BDF) or (3BDF) above): (8) a. b. c. d. I new II new III new

Focus Given Given Given Focus Given Given Given Focus
The studies cited above report differences in duration, pitch and intensity between the Focus constituents found in contrastive focus sentence types (8bcd) and the discoursenew constituents found in the all-new sentence type (8a) in all sentence positions (I, II, III). Focus constituents had longer duration than the corresponding new constituents in all-new sentences. The Focus constituents also showed greater pitch protrusion and greater post-Focus pitch compression. But the design of these experiments requires us to interpret these results with some caution. One shortcoming of the design of these earlier experiments is that fully controlled comparison of the prosody of (putative) Focus and new constituents is not possible, since
a (putative) Focus constituent in contrastive focus sentence types and its corresponding new constituent in the all-new sentence type do not appear in prosodically identical environments within their respective sentences. For example, we know that in a sentence of type (8c) in English, the (putative) Focus in medial position would be followed by accentless material, whereas the medial new constituent in (8a) would be followed by accented material; comparable distinctions in surrounding prosodic context appear with the other Focus sentence types. Such a difference in prosodic context itself could potentially be responsible for differing phonetic properties, rather than a putative syntactic difference in Focus vs. new status (and a consequent difference in phonological representation). This shortcoming in design is of particular importance in the study of pitch patterns, which are context-dependent. So, for example, the fact that the sentencemedial constituent in (8c) has significantly higher F0 than the corresponding medial constituent in (8a) could be the consequence of the putative contrast in information structure status (Focus vs. new) of that constituent. But it could also result from a difference in the preceding prosodic context. If the obligatorily pitch-accented (and phrase-stressed) new constituent in initial position in (8a) triggers a downstepping or lowering of pitch range on what follows, as in all-new sentences in other Germanic languages (van den Berg et al 1992, Truckenbrodt 2004), then the new constituent in medial position will accordingly be lowered in pitch, while still bearing pitch accent (and phrasal stress). In sentence type (8c), on the other hand, the medial Focus is preceded by a Given constituent, and because that initial Given constituent lacks the obligatory pitch accent (and phrasal prominence) of a new constituent, it may fail to trigger the same degree of downtrend on what follows. As a consequence the medial constituent in (8c) the putative Focus--would show higher F0 than the corresponding medial new constituent in (8a). This means that in a comparison of the pitch values of the medial Focus and new constituents here, its not certain one can attribute the difference in F0 to the Focus-new distinction. The Given-new distinction in the preceding context and its consequent phonological and phonetic effects could just as well be responsible. Which is to say that the prosodic differences displayed could be consistent with a common F-marking representation for the contrastive focus and discourse-new medial constituent.

For these reasons, we need to compare sentences containing sequences of (putative) Focus-new, new-Focus and new-new constituents in order to provide crucial evidence for the existence of a contrastive focus vs. discourse-new distinction in grammar. The experiment we report here constitutes a first attempt to investigate a paradigm of sentence types like (4) which controls for sentence-internal context, with a mind to determining whether a systematic greater prominence for Focus-marked constituents, as compared to merely discourse-new constituents in the same sentence, does in fact emerge. At this point some terminological clarification is in order. In phonetic studies examining sentences belonging to the types schematized in (8), the term narrow focus sentence has typically been employed to refer to sentences types that contain an internal contrastive focus, as in (8bcd), while the term broad focus sentence has been used to refer to the all-new sentence type (8a). This terminology, due originally to Ladd 1980, does not adequately characterize the sentences types at issue. The term narrow focus has come to be associated with Focus constituents that appear in Given contexts within the sentence, as in (8bcd). This practice may indeed be partly responsible for a general failure to recognize that a Focus may also co-occur with discourse-new constituents in a sentence, or with some combination of Given and new, even though these latter sentence types contain narrow focus too. As for the term broad focus, it has typically been used to refer to sentences which do not contain a narrow contrastive focus, and which moreover consist of constituents that are all discourse-new. Such sentences are assumed by Ladd 1980, following Jackendoff 1972, to be Focus constituents themselves. But the notion that every sentence must contain at least one Focus (if not narrow then broad) is questionable on semantic/pragmatic grounds (Kratzer and Selkirk 2010, in preparation). And even if an entire sentence were itself a Focus constituent, this would imply nothing about its internal composition in terms of Given vs. new. In other words, the broad/new focus terminology is descriptively inaccurate. In this paper, the term contrastive focus sentence refers, depending on the circumstances, either to a sentence that contains a Focus constituent within it or to the particular class of contrastive focus sentences examined in our experiment, which combine sentence-internal Focus and
discourse-new constituents. We use the term all-new sentence to refer to sentences lacking any (internal) Focus and which moreover contain no Given material. 1. 4 Whats ahead In the next section, we lay out details of the design of our experiment, including the sentence types examined. Section 3 reports data on the distribution of pitch rises and falls in the intonational contours of these sentences and motivates the assumptions that we will make concerning the distribution of pitch accents and phonological phrase organization. Sections 4-6 report on the findings concerning phonetic prominence, specifically duration (section 4), pitch (section 5) and intensity (section 6). In these, reports of the phonetic findings are followed by discussions which develop our understanding of the implications of these findings for the semantics, syntax and phonology of the (putative) Focus vs. discourse-new distinction, leading to a summary in Section 7. 2. The design of the experiment 2.1 Experimental materials The goal of this study is to compare the pronunciations of sentences which are essentially identical except for the Focus vs. new status of key constituents. Ensuring that these sentences are indeed produced with the intended information structural meanings in mind poses a methodological challenge. Just how can the experimenter induce a speaker to pronounce a sentence with the desired information structure in a way that approximates natural usage? Elicitation of sentences combining Focus and new constituents requires materials that are somewhat more complex than is typical in studies of standard narrow focus sentences. These latter have usually been elicited in wh-question-and-answer dialogues like (3) or dialogues with correction statements like (2), making up experimental paradigms like (8).

2a. Bill chooses the most awful companions. He was dating that horrible lawyer last year, and then there was Kate, who we all hated. He even took [Minnie]FOC to a [Mariners game]. And shes insufferable. 2b. Bill is a sports freak. Hell go to any kind of sporting event, regardless of the team thats playing. He even took [Minnie] to a [Mariners game]FOC. And they havent been in contention for years! 2c. Bills had a pretty busy week. He had meetings all through the weekend, and then he went to Seattle for a conference. He took [Minnie] to a [Mariners game]. I bet that was fun. 2.2 Recording the materials Stimulus materials consisted of minimal triplet discourses, as described above. There were 18 such triplets, for a total of 54 discourses. A full list of materials is in Appendix A. Each subject participated in three sessions. Each session consisted of six discourses in each condition, for a total of 18 discourses. Two members of a minimal triplet never cooccurred in the same recording session. Sessions were scheduled at least two days apart. During each session, subjects completed the eighteen discourses first in one random order, then in another. The repetition ensured a full set of recordings from each subject in the case of errors in reading or recording, or difficulties in acoustic analysis. Subjects were seated in a soundproof recording booth and read the sentences into a condenser microphone positioned on a desk in front of them, slightly to one side. Each stimulus consisted of a paragraph presented on a computer screen, generally three to five sentences in length, as shown in the examples in section 2.1. The first several sentences (the context) were presented in plain type. The last two sentences were in bold type. The first bold sentence was the target sentence, and the last sentence in the paragraph was designed to reinforce the desired information structure for the target sentence. Subjects were instructed to silently read each paragraph several times, to get a

Selkirk (1984, 1995, 2000), Beckman and Pierrrehumbert (1986), Ladd (1996, 2008), Gussenhoven 1990 and in the ToBI transcription system for English (Beckman and Ayers Elam 1993, Veilleux et al 2008) 24
Moreover, we expected that the post-verbal complements in these utterances might occupy distinct phonological phrases in prosodic structure; prior work on English (and other languages) shows that phonological phrase breaks have a strong tendency to appear between two verbal complements in all-new utterances (see, e.g. Selkirk 1986, 2000, 2011). If so, a tonal reflex of that phrasing should appear-- in the form of a L tone coinciding with the right edge of each complement, creating a phrase-final pitch fall from the H* pitch accent of each complement. That phrase-edge L tone is analyzed as a peripheral L- phrase accent by Beckman and Pierrehumbert 1986 and in the English ToBI system (Beckman and Ayers Elam 1993, Veilleux et al 2008); as part of a phrase-final H*+L pitch accent in Gussenhoven 1990. The intonational contours in our materials strongly suggest that these expectations regarding (i) the presence of a H* pitch accent in both complements and (ii) the presence of L tone at the right edge of both complements are borne out. This distribution of pitch accents and edge tones and the consequences for the prosodic structure representation of the various sentence types are discussed in sections 3.1 and 3.2.
Figure 2 Schematic F0 contours for each condition, measured at seven points in the sentence and averaged across subjects
Figure 2 shows schematic F0 contours for each focus/new condition. The conditions represent differences in the focus/new status of the verbal complements: A = Foc-new; B = new-Foc; C = new-new. The m3 and m6 measurements correspond to F0 peaks on the final pitch accent within Complement 1 and Complement 2 respectively. The peak at m1 corresponds to the F0 peak found in the focus-sensitive particle (e.g. only) in the contrastive condition A and B sentences and to the peak within the verb in all-new sentences. The valley measurements m2, m4, m5, m7 were taken at the left- and rightmost edges of the word containing the pitch accent. 3.1 Pitch accents on Complements 1 and 2 The contours in Fig. 2 reflect the generalization that, in all but a tiny number of cases, there is some sort of F0 peak in both complements in all conditions. In particular, 97% of the complements show an identifiable preceding rise and 99% of complements show an identifiable following fall. The high F0 mark in each complement (m3, m6) is on average 0.55 ERB (22 Hz) higher than the values at the left edge of the accented word (m2, m5) and 0.86 ERB higher (34 Hz) than values at the right edge of the accented word (m4, m7). The histograms below, pooled across position and condition, show that the vast majority of all utterances feature a preceding rise and a following fall in F0 surrounding the pitch accented syllable. This can be seen in the fact that the vast majority of the distribution lies to the right side of the zero marker on the X axis.

Figure 6. Histograms for each condition showing the amount of downtrend from m4 to m5. Negative values indicate uptrend. Y axis shows the number of data points falling within each bin (range of values). These data provide evidence for a L tone target at m4the right edge of Complement 1-in all conditions. As Figure 6 shows, data for pitch movement between the two low measurement points generally include a mix of positive and negative values, with modes
close to zero. Although the median case in each condition shows a small amount of downtrend, there are many cases where the F0 value at m4, located at the right edge of Complement 1, is equal to or even lower than the following measurement: such cases constitute 35-43% of the data points in each condition. These data point to either a Vshaped valley with a single L target at m4, or a U-shaped valley in which the first of the L tones that define the U-valley floor appears at m4. As for the cases where m4 is higher than m5, most can also plausibly be analyzed as U-shaped valleys, showing varying degrees of downtrend between m4 and m5. Most cases of downtrend between the two low points are small in magnitude. Median values range from 0.06 to 0.18 standard deviations depending on the condition. These values are an order of magnitude smaller than, for instance, the pitch fall following complement 1, which ranges from 0.79 to 1.17 standard deviations across conditions. In sum, data from all three conditions suggest the presence of a sentence medial L tone that is final in Complement 1. This in turns supports the hypothesis of a phonological phrase break between Complement 1 and Complement 2, regardless of the Focus or new status of the complements, as indicated in (13). Consider an alternative hypothesis concerning the pitch valley between the complements, one which posits not a L tone at the right edge of Complement 1, but instead a sequence of a H* pitch accent in Complement 1 followed by a L+H* pitch accent on Complement 2. This hypothesis may seem to be consistent with the averaged data presented in the schematic pitch contours in Figure 2, though only if the low pitch values at position m4 preceding the m5 minimum can be ascribed to an asymptotic phonetic interpolation between a H* tone and a following L+ tone, one which would create the radically concave fall shape seen in the m3-m5 sequence in Figure 2. But Figure 6 shows that a H* plus L+H* sequence cant make sense of the full range of data; the H* plus L+H* proposal (even with asymptotic interpolation) is inconsistent with the cases in Fig. 6 where m4 is lower than m5. These cases require the postulation of an intervening phraseedge L tone at m4. This L edge-tone is also consistent with the cases where m5 is lower than m4, if one assumes that there is variation possible in the amount of downtrend between phrases. In sum, the data favor the hypothesis that a L tone is present at the right
edge of Complement 1 over the hypothesis that a L tone is absent in that position. They therefore support the phonological phrase analysis given in (13), for all sentence types. 3.3 The tone(s) preceding Complements 1 and 2 Sentences from conditions A, B, C do differ in the part of the intonational contour that precedes the complements. In Condition C, a subject pronoun and a verb precede the first and second verbal complements; the initial F0 measurement of the sentence (m1) was taken at the main-stressed syllable of the verb, which would bear any pitch accent that might appear. Inspection of pitch tracks in these cases indicates that the verb may or may not carry a pitch accent. In the utterance in Figure 3 the verb carries a pitch accent. The sentences in Condition A and B included a focus-sensitive particle like only, situated between the subject pronoun and the verb. The initial high pitch peak of the sentences in Conditions A and B typically fell on that particle; in these conditions, the initial F0 measurement of the sentence (m1) was taken on the stressed syllable of the particle. In Figures 4 and 5 we have transcribed the pitch peak on the particle as H*. That initial H* peak is followed by a gradual (nonasymptotic) descent to a low target immediately preceding the H* pitch accent of Complement 1; this contour speaks to the absence of any additional L tone immediately following the initial H* on the particle. Note that, as Figure 2 shows, the m1 pitch values on the focus particles in Conditions A and B were, on average, quite a bit higher than the m1 pitch value on the verb in Condition C sentences. The authors impression in listening to the recordings was that speakers were choosing values for the pitch of the H tone on the focus particles that were especially high in their pitch range. 3.4 Summary On the basis of the pitch contours of sentences in the three different Focus/new conditions in our study, we have established that the Focus and new complements share two essential phonological properties. The vast majority show a pitch peak, represented

here as a H* pitch accent. And this H* immediately precedes a phrase-final L tone, indicating that the complement phrases in our materials, whether sentence-medial or sentence-final, are each final in a phonological phrase. The fact that (putative) Focus and new complements both show these phonological properties in all conditions means that any differing patterns of quantitative phonetic prominence exhibited by the Focus-new, new-Focus and new-new sequences cannot be explained in terms of presence vs. absence (or type) of pitch accent or in terms of medial vs. final position within a phonological phrase. Rather, they must be due to the Focus vs. new status of the complements themselves, or to some other aspect of phonological representation which would mediate between this Focus/new representation and its phonetic realization. 4.0 Focus Duration We now turn to the question of whether a Focus constituent is distinguished from a new constituent by a greater degree of phonological or phonetic prominence. In this section we ask whether a Focus constituent displays greater duration when a comparison is made of Focus-new (A), new-Focus (B) and new-new (C) sentences. Our materials also allow us to ask whether the presence or location of Focus in a sentence has any effect on the duration of surrounding constituents in the same sentence. We are able to make a controlled investigation of these questions because the constituents preceding and following the putative Focus in our materials are discourse-new and so have the same information structure status as the corresponding context constituents in the all-new sentences (cf. discussion in section 1), and because these context constituents also have the same prosodic structure across all conditions (as shown in section 3). Section 4.1 reports on the duration findings themselves: a three-way distinction in phonetic duration patterns across the three sentence types. These findings support the claim that the theory of grammar (i) makes a distinction between contrastive Focus and discourse-new and (ii) allows for that distinction to be reflected in phonetic interpretation. In section 4.2 it is suggested that differences in the phonological
representation of phrasal prominence (stress) in the three different sentence types are responsible for the three-way durational distinctions found in the phonetics. 4.1 Duration Findings Shown below in figure 7 are the average durations of the measured stretch of each complement in each condition.
DURATION (ms) CONDITION Complement 1 A B C 326 Complement 294 268
Figure 7: Average duration of measured stretches of first and second verbal complements in three information structure conditions: A = Foc-new; B = newFoc; C = new-new. Whiskers represent 95% confidence intervals. Recall that the measured stretch of each complement consisted of a string containing the pitch accented syllable and one or two adjacent syllables. For the sentences of any 35

minimal triplet, the measured string was identical. (For details see section 2 and Appendix B.) In the statistical analysis that follows Complement 1 and Condition C are the baseline levels for position and condition that other levels of those variables are compared to. As Figure 7 shows, Complement 1 is on average 35-40 ms. longer in condition A, where it is a Focus, than in the other two conditions, where it is new. This amounts to an 11% lengthening effect of Focus over new. The difference in Complement 1 duration between A and the control condition C is significant: = 0.51, p < 0.001. Note that the difference in duration of the discourse-new Complement 1 in B and C, 2% of a standard deviation, is not significant. The relative pattern between conditions changes at Complement 2. Conditions A and C, which were significantly different at Complement 1, are roughly equal at complement 2. Conditions B and C, which were roughly equal at Complement 1, are different at Complement 2. Complement 2 is 25-30 ms. longer in condition B, where it is a Focus, than in either of the other two conditions, where it is new. This amounts to a Focus lengthening effect of about 10%. These differences in relative duration by complement and condition are reflected in significant interactions between position and condition: for B vs. C, = 0.40, p < 0.001; for A vs. C, = -0.55, p < 0.001. A central finding, clearly visible in Figure 7, is the greater duration of the Focus constituent relative to new constituents in the corresponding positions in other conditions. This finding supports the claim that the grammar treats Focus constituents differently from new constituents. The magnitude and direction of these effects do not differ significantly by subject or item, showing that they are statistically-robust generalizations. As we noted earlier in section 1.2, the hypothesis that a Focus element receives greater phonetic prominence than a merely discourse-new element entails that the withinsentence relationship between the 1st and 2nd complement differs across the three conditions. This is exactly what the significant interactions between position and condition show. The difference in duration between the first and second complements is largest in condition A, intermediate in condition C and smallest in condition B.
Overall, the durations at Complement 2 are smaller than those at Complement 1: = 0.84, p < 0.001. This simply reflects the fact that the strings measured in Complement 2 were shorter than those measured in Complement 1, because the utterance-final syllable was generally excluded from the measured sequence (cf. section 2). The exact size of this difference differs between speakers and between items; incorporating those differences into the model significantly improves fit: 2 (2) > 25, p < 0.001 for both effects. To summarize, our experiment has revealed a three-way distinction in duration patterning that correlates with a three-way distinction in Focus/new status, as seen in Figure 7. It is the significantly greater duration found with Focus constituents that has the result that the three sentence types Focus-new, new-Focus and new-new are distinguished in this aspect of their prosody. The duration of corresponding new constituents in the different sentence types is virtually constant, not affected by the presence or absence of a Focus constituent in the same sentence13. 4.2 Implications of Focus duration findings for the grammar 4.2.1 Syntactic representation of contrastive focus vs. discourse-new If the theory of grammar made no distinction between contrastive Focus and discoursenew and instead represented both of the complements in our sentence types in the same fashion, e.g. as F-marked, there would be no possible way to derive the observed three way distinction in duration patterning. Theories of focus and focus prosody which posit a

(Discourse-new constituents have been left syntactically unmarked in (15), as suggested in section 1.1.) The grammar of English then gives a specific characterization to the relation between the distinct syntactic representations in (15) and the phonetic durational patterns that correspond to them (cf. Figure 7), as envisaged in general terms by (II). 4.2.2 A phonological distinction between Focus and discourse-new Consider now the assumption made widely in generative grammar that phonetic interpretation is syntax-blind and accesses only phonological representation (Chomsky & Halle 1968). This assumption narrows considerably the set of hypotheses that can be entertained about the nature of the possible relation(s) between syntactic representations like (15) and their phonetic interpretation. It divides the theory of that relation into two parts: (IIa) a theory of the relation between syntactic and phonological representations, one that allows differences in phonological form to be assigned depending on the Focusmarked status of a syntactic constituent, for instance; and (IIb) a theory of phonetic interpretation that is sensitive to those (and other) differences in phonological representation. In what follows we pursue the consequences of adopting the specific assumption that phonetics is syntax-blind and that a mediating phonological representation does enter into the definition of the Focus-phonetics relation. The question, then, is what the character of that mediating phonological representation is. In section 3, we showed that pitch accents are located on both complements in all three conditions, and that the pitch accented word in both the complements is the last one in a phonological phrase. So neither presence of pitch accent or phrase-final position can be the aspects of phonological representation that are responsible for the duration difference between Focus and new constituents. But positing differing patterns of prosodic prominence (stress) in the phonological representation could predict the durational differences amongst the pitch accented sequences in our sentence types. This explanation seems plausible, given that stress at lower levels in English is reflected in duration (see Fry 1955, Huss 1977, Beckman 1986, Okobi 2006, inter alia for the effects of stress on syllable duration in English). Moreover, there is ample precedent for assuming a
phonological representation of phrasal prominence patterns in English and in particular for assigning a Focus constituent greater prosodic prominence (stress) than any nonFocus within the same domain (see references in first paragraph of section 1.2). We need to posit three distinct prominence representations in the phonology that would mirror the different syntactic distributions of Focus in (15), in accordance with IIa. The phonetic component of English is hypothesized to assign greater duration to elements that bear greater abstract phonological prominence (stress)14, in accordance with IIb. In this way, the quantitative phonetic patterns of Focus-related duration in English that are seen in Figure 7 would be accounted for. Assuming, then, that a phonological representation of prosodic prominence is the basis for the durational patterns in Figure 7, the specific prosodic prominence representations in (16abc) could be posited for the three distinct cases. The Focus constituent in first and second complement position in (16a) and (16b), respectively, carries greater prominence than a new constituent in the same sentence. In (16c)Condition C-- the two new constituents in sequence bear the same, lower, level of prominence. (16) Prosodic Prominence Representations in Conditions A,B, C: A three-way contrast a. Condition A: Foc-new ( ( (x H* x x )(x) ( x H* ) ( ) Lx ( x H* ) ) ) LIntonational Phrase Phonological Phrase Prosodic Word

Applying this principle to sentential representations like (16abc), the within-sentence relations in phonetic prominence that were observed are predicted. We began this paper with a research question relevant to the theory of semantics and pragmatics: Does data from the phonetics of English provide support for making a grammatical distinction between the contrastive focus vs. discourse-new status of constituents of a sentence? In examining the evidence that it does, we have charted out a further research hypothesis: That the patterns of phonetic prominence that have been found to accompany the distribution of Focus and new in English (and other languages) are to be explained in terms of an interface relation between phonetic prominence and patterns of prosodic (stress) prominence in phonological representation.
References Baayen, R., D. Davidson & D. Bates. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language 59, 390412. Bates, D. (2007). lme4: An R package for fitting and analyzing linear, nonlinear and generalized linear mixed models. Software application. Beaver, David, and Clark, Brady. 2008. Sense and sensitivity: How focus determined meaning. Malden MA: Wiley-Blackwell. Beckman, Mary E. 1986. Stress and Non-Stress Accent. Dordrecht: Foris. Beckman, Mary E. and Pierrehumbert, Janet. 1986 Intonational structure in Japanese. Phonology Yearbook 3, 225-309. Beckman, Mary, and Ayers-Elam, Gayle. 1993. Guidelines for ToBI labelling. URL: www.ling.ohio-state.edu/~tobi Berg, Rob van den, Gussenhoven, Carlos, and Rietveld, Toni. 1992. Downstep in Dutch:Implications for a model. In Papers in laboratory phonology II: Gesture, segment and prosody, eds. D. Robert Ladd and Gerald Docherty, 335-358. Cambridge: Cambridge University Press. Bolinger, Dwight. 1961. Contrastive accent and contrastive stress. Language 37, 83-96. Bolinger, Dwight. 1972. Accent is predictable (if youre a mind-reader). Language
48(3), 633-644. Breen, Mara. 2007. The identification and function of English prosodic features, Ph.D. thesis, Brain and Cognitive Science, MIT. Breen, Mara, Federenko, Evelina, Wagner, Michael and Gibson, Edward. 2010. Accoustic correlates of information structure. Language and Cognitive Processes Volume 25, (Issue 7&9), 1044-1098. Bring, Daniel. 2007. Intonation, semantics and information structure. In The Oxford Handbook of Linguistic Interfaces, eds. Gillian Ramchand and Charles Reiss. Oxford: Oxford University Press. Bring, Daniel. 2008. Been there, Marked that: A theory of second occurrence focus. In Proceedings of NII Workshop on Focus-2005, eds. Makoto Kanazawa and Christopher Tancredi. Tokyo. To appear in published version of proceedings. Byrd, Dani and Saltzman, Elliot. 2003. The elastic phrase: modeling the dynamics of boundary-adjacent lengthening. Journal of Phonetics 31, 149-180. Calhoun, Sasha. 2006. Information structure and the prosodic structure of English: a probabilistic relationship, Ph.D. dissertation, University of Edinburgh. Calhoun, Sasha. 2010. The centrality of metrical structure in signaling information structure: A probabilistic perspective. Language 86(1), 1-42. Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and points of view. In Subject and Topic, ed. Charles Li, 25-56. New York: Academic Press. Choi, Jeung-Yoon, Hasegawa-Johnson, Mark, and Cole, Jennifer. 2005. Finding intonational boundaries using acoustic cues related to the voice source. Journal of the Acoustical Society of America 118(4): 25792587. Chomsky, Noam, and Halle, Morris. 1968. The Sound Pattern of English. New York: Harper and Row. Cinque, Gugliemo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24:239-297. Cooper, William E., Eady, Stephen, and Mueller, P. 1985. Acoustical aspects of contrastive stress in question-answer contects. Journal of the Acoustical Society of America 77:2142-2156. de Lacy, Paul. 2007. The interaction of tone, sonority and prosodic structure. In The Cambridge handbook of phonology, ed. Paul de Lacy, 281-307. Cambridge: Cambridge University Press.

Hyman, Larry. 2006. Word-prosodic typology. Phonology 23:225-258. It, Junko, and Mester, Armin. 2007. Categories and projections in prosodic structure. In OCP4 [Old World Conference in Phonology]. Rhodes, Greece. It, J. and Mester, A. (2010) Recursive prosodic phrasing in Japanese, in Proceedings of the 18th Japanese/Korean Conference, 147-164, CSLI, Stanford. It, J. and Mester, A. (2011) Recursive prosodic phrasing in Japanese, in Prosody Matters: Essays in Honor of Elisabeth Selkirk (eds T. Borowsky, S. Kawahara, T. Shinya, and M. Sugahara), Equinox Publishers, London. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge MA: MIT PRess. Kahnemuyipour, Arsalan. 2004. The syntax of sentential stress, Ph.D. dissertation, Linguistics, University of Toronto. Kahnemuyipour, Arsalan. 2009. The syntax of sentential stress. Oxford: Oxford University Press. Keating, P. (2003). Phonetic encoding of prosodic structure. Proceedings of the 6th International Seminar on Speech Production, Sydney, December 7 to 10, 2003. Keating, P. & Shattuck-Hufnagel, S. (2002). A Prosodic View of Word Form Encoding for Speech Production. UCLA Working Papers in Phonetics 101, 112-156. Kiss, Katlin. 1998. Identificational focus vs. information focus. Language 74:245-273. Koch, Karsten. 2008 Intonation and Focus in Nle?kepmxcin (Thompson River Salish). Ph.D. dissertation, University of British Columbia. Kratzer, Angelika. 1991. The representation of focus. In Handbuch Semantik/Handbook Semantics, von Stechow, Arnim and Wunderlich, Dieter, eds, 825-834. Berlin, New York; de Gruyter. Kratzer, Angelika. 2004. Interpreting focus: presupposed or expressive meanings? Theoretical Linguistics (30(1), special issue on Interpreting Focus, 123-136. Kratzer, Angelika, and Selkirk, Elisabeth. 2007. Phase theory and prosodic spellout: The case of verbs. The Linguistic Review 24, 93-135. Kratzer, Angelika, and Selkirk, Elisabeth. 2010. Distinguishing contrastive, new and given information. Talk delivered at the International Conference on Information Structure, sponsored by SFB 632: Informationsstruktur. Potsdam, July 2010. Kratzer, Angelika and Selkirk, Elisabeth. (in preparation) Distinguishing contrastive, new and given information. Manuscript, UMass Amherst.
Kratzer, Angelika and Shimoyama, Junko. 2002. Indeterminate pronouns: The view from Japanese. In Proceedings of the 3rd Tokyo Conference on Psycholinguistics, Otsu, Yukio, ed., 1-25. Tokyo: Hituzi Syobo Krifka, Manfred. 1992. A Compositional Semantics for Multiple Focus Constructions. In Informationsstruktur und Grammatik, Sonderheft der Linguistischen Berichte, Joachim Jacobs, ed. Also in Proceedings from Semantics and Linguistic Theory 1, Cornell Working Papers in Linguistics 10, 1991 Krifka, Manfred. 2008. Basic notions of information structure. in Acta Linguistica Hungarica 55 (2008), 243-276. Also in Interdisciplinary Studies of Information Structure 6, C. Fery and M. Krifka, eds. Potsdam, 2007. Kgler, Frank & Skopeteas, Stavros. 2006. Interaction of Lexical Tone and Information Structure in Yucatec Maya. TAL 2006 International Symposium (Tonal Aspects of Language), 83-88. La Rochelle, France, April 27-29, 2006. Kgler, Frank & Skopeteas, Stavros. 2007. On the universality of prosodic reflexes of contrast: The case of Yucatec Maya. In Proceedings of the XVI. International Congress of Phonetic Sciences, Trouvain, Jrgen & Barry, William J. (Eds.). Saarbrcken, Germany, 4-10 August 2007. Ladd, D. Robert. 1980. The structure on intonational meaning: evidence from English. Bloomington IN: Indiana University Press. Ladd, D. Robert. 1984. Declination: a review and some hypotheses. Phonology Yearbook 1:53-74. Ladd, D. Robert. 1988. Declination 'Reset' and the hierarchical organization of utterances. Journal of the Acoustical Society of America 84:538-544. Ladd, D. Robert. 1996. Intonational Phonology. Cambridge: Cambridge Univ. Press. Ladd, D. Robert. 2008. Intonational phonology. 2nd ed. Cambridge: Cambridge University Press. Ladd, D. Robert and Schepman, Astrid. 2003. Sagging transitions between high pitch accents in English: experimental evidence. Journal of Phonetics 31(1), 81-212. LeGac, David. 2002. Tonal alternations and prosodic structure in Somali. Speech Prosody 2002, Aix-en-Provence. Liberman, Mark, and Sag, Ivan. 1974. Prosodic form and discourse function. Proceedings of the annual meeting of the Chicago Linguistics Society 10:416-427. Liberman, Mark, and Prince, Alan. 1977. On stress and linguistic rhythm. Linguistic Inquiry 8:249-336.

4a. The Red Sox had an exhibition game for charity, and they gave the players various bright-colored uniforms. Bill Mueller and Nomar Garciaparra have really played well this year. But they only gave Manny the yellow one. Thats the one thats reserved for the most valuable player. 4b. The Red Sox had an exhibition game for charity, and they had special bright-colored uniforms made for the occasion. There were a lot of different colors; a couple of the jerseys were orange, one was purple. But they only gave Manny that yellow one. That was a lousy color. 4c. The Red Sox had an exhibition game for charity, and they gave all the players crazy bright-colored uniforms to wear for the occasion. The whole thing was pretty funny to watch. They gave Manny the yellow one. It was so ugly. 5a. There are so many competing football teams in California. There are three or four popular big college teams, and then there are the Raiders and the Chargers in the NFL. It seems they only watch the 49ers in Salinas. This might be the most popular team in the state. 5b. California is a big football state, with millions of fans. Each town has its own special fan population. In Los Angeles, the fans all care about how they look on TV. Theyre organized in competing clubs in San Diego. People only watch the 49ers in Salinas. Its an unpopular team everywhere else. 5c. Central Coast California is a pretty interesting place for sports. Theres a lot of amateur sports, but also a lot of professional team activity. They watch the 49ers in Salinas. Theres also a minor league ballpark there. 6a. Nora is a compulsive shopper. Shes always running off to Target. She makes frequent passes through T.J. Maxx. And she usually goes to Wal-Mart on Monday. Her housemates complain about all the stuff she brings home. 6b. Nora doesnt manage her schedule very well. On Tuesday, she always has a lot of free time. And on Thursday, too. But she usually goes to Wal-Mart on Monday. It would be easier if she picked a better day. 6c. Noras been looking for a special kind of espresso machine that she saw in Gretchens house. She cant seem to find it anywhere. She tried everywhere in Northampton. And she went to Wal-Mart on Monday. That didnt seem to work out, so shell keep looking.
7a. The U.S. Department of Defense just released their annual weapons inventory. Theyve got computer-guided precision bombs, and robot missiles. But they still make use of more conventional arms. For instance, they mostly store mines in Idaho. They really dont keep any sophisticated weapons there. 7b. The U.S. Department of Defense just released their annual report on the military infrastructure and its geographical distribution. The vast majority of military installations are in the South. There are facilities scattered in other places, though. For instance, they mostly store mines in Idaho. Its pretty sparsely populated there. 7c. The U.S. Department of Defense just released their annual report on the storage of arms and hazardous materials. I read the section on the installations in the West, which was pretty interesting. For instance, they store mines in Idaho. Probably because theres not many people there. 8a. Janes a really controlling chef. She yells at Michael and is constantly interfering with Larry. But she usually lets Lena prepare the mayonnaise. She thinks anyone else would overbeat it. 8b. Janes a controlling chef. She tries to do all the work herself. She slices the vegetables, browns the meat, and prepares the stock for sauces. Oddly enough, she always lets Lena prepare the mayonnaise. Mayonnaise is really tricky. 8c. Jane works as a chef. She is in charge of a number of cooks, and one of the hardest parts of the job is knowing their abilities and supervising the kitchen activities. She lets Lena prepare the mayonnaise. Shes an efficient worker, and frees up time for Jane. 9a. Our soccer team went to compete in European tournaments last year. It was pretty intimidating, because the level of play was so high. But we surprised everybody with a number of second- and third-place finishes. We even wound up winning in Mannheim. That was an incredible experience. 9b. Our soccer team went to compete in European tournaments last year. In Madrid we played really well, and in Manchester, too. We even wound up winning in Mannheim. And thats the toughest tournament in Europe. 9c. Our soccer team went to compete in European tournaments last year. It went better than we expected, and we had a good time. We wound up winning in Mannheim. They gave us a big trophy, and we keep it in the clubhouse now. 10a. DeBeers is the largest ore corporation in the world. They employ thousands of people, including engineers, detonation teams, and earth-moving experts. They primarily use miners for extracting aluminum. Its an amazing range of jobs.