Printer Friendly

Spanish evidence for pitch-accent structure *.

Abstract

This paper considers the ability of the standard view of the pitch accent and two alternative views to account for the tonal-alignment data of two bitonal rising pitch accents in Spanish (one of which is used in cases of broad focus and the other in cases of contrastive focus). It is shown that there is a difference in the relationship between the tones from one pitch accent to the other. The standard view of the pitch accent, which does not assume that pitch accents have an internal structure, is unable to account for such a difference. Postulating that pitch accents have an internal structure allows for an analysis of the two Spanish pitch accents that is able to account for the different relationship between the tones in the different pitch accents. Two possible pitch-accent structures have been mentioned in previous literature, and these are considered in order to see which, if either, is able to account for the Spanish tonal-alignment data. The Spanish tonal-alignment data lead to the proposal that pitch accents have a hierarchical structure, and this structure is employed to propose an analysis of the two Spanish rising pitch accents.

1. Introduction

In the late 1970s and 1980s much work on intonation was dedicated to the development of a phonological theory of intonation (Liberman 1975; Bruce 1977; Pierrehumbert 1980; Beckman and Pierrehumbert 1986; Pierrehumbert and Beckman 1988). The resulting autosegmental-metrical (AM) model of intonational phonology (Ladd 1996) characterizes pitch movements as a string of tonal events in which rises in the intonation contour are made up of a sequence of a low tone (L) and a high tone (H) associated with particular points in the speech stream (e.g. stressed syllables of prominent words, phrase edges). Falls, similarly, consist of a sequence of H plus L. It has always been assumed within the AM model that pitch accents (i.e. the tonal event associated with a particular unit, the stressed syllable in Spanish) may contain two tones. Little consideration has been given, however, to the relationship between the two tones. In the present study I show that the standard view of the pitch accent is unable to account for the relationship between the tones of Spanish pitch accents. I further argue that postulating that pitch accents have a hierarchical internal structure allows for an accurate analysis of the relationship between these tones.

The remainder of this paper is organized as follows: in the following section I present the three different views of pitch accents that have been proposed in previous research, making note of the predictions that each structure makes for tonal-alignment patterns. Section 3 explains the methodology of the experiment carried out for the present study. In section 4 I present the Spanish tonal-alignment data obtained from the experiment described in section 3. Section 5 contains an analysis of the data that argues for the postulation of a hierarchical internal structure for pitch accents. Section 6 contains the conclusions that can be drawn from the results of the present study and considers their implications for the AM model of intonational phonology.

2. Pitch-accent structure

Pierrehumbert and Beckman (1988) represent bitonal pitch accents as two tones linked to one slot on the tone tier, as in (1).

(1) Pierrehumbert and Beckman's (1988) view of the bitonal pitch accent

[ILLUSTRATION OMITTED]

In this representation the strong (s) branch contains the tone that has "priority in establishing the alignment of the pitch accent" (Pierrehumbert and Beckman 1988:125). This is the tone that is commonly marked with a * in AM notation, meaning that the pitch accent in (1) would be represented as [H.sup.*] + L. The [H.sup.*] indicates that the H, rather than the L, is the portion of this pitch accent that aligns with, or anchors to, a particular point in the speech stream, which in Spanish is the stressed syllable. Three other bitonal pitch accents are possible as well (i.e. H + [L.sup.*], [L.sup.*] + H, L + [H.sup.*]) depending on whether the first or second branch is strong and on whether a H or L is on that branch. To exemplify, (2) demonstrates schematically the alignment of the two bitonal rising pitch accents (i.e. [L.sup.*] + H and L + [H.sup.*]) with a stressed syllable (S).

(2) Distinct alignment of two L + H pitch accents (S = stressed syllable)

[ILLUSTRATION OMITTED]

Pierrehumbert and Beckman's (1988) representation of bitonal pitch accents, as seen in (1), makes certain predictions about the nature of pitch accents. First, the unstarred tone (i.e. the tone on the weak branch) of every bitonal pitch accent has the same structural relationship to the starred tone (i.e. the tone on the strong branch). This predicts that where more than one bitonal pitch accent exists in a language, all unstarred tones should align similarly with respect to the starred tones. (1) This is represented in (2), where, despite a difference in the alignment of the rise with the stressed syllable, the two rises are identical. Second, due to the linking of both tones to one slot on the tone tier, it predicts that the two tones should function as a single unit rather than as two independent units. While the representation of pitch accents in (1) is the most commonly accepted representation and is assumed by the majority of researchers working on intonation, it is not the only possible representation.

Grice (1995b) offers two alternatives to the representation in (1), and each makes its own set of predictions. The first alternative is what she calls a flat-structured pitch accent, due to there existing no intermediate level between the tonal root node (i.e. the equivalent of the tone node on the tone tier in [1 ]) and the individual tones. This flat structure, which is based on Yip's (1989) work on Chinese and African languages, is, in part, like Pierrehumbert and Beckman's (1988) representation as shown in (1). The difference is that there is also the possibility for the two tones to be linked to separate tonal root nodes. These two possibilities are shown in (3).

(3) Flat structure adapted from Grice (1995b)

a. Cluster

[ILLUSTRATION OMITTED]

b. Unit

[ILLUSTRATION OMITTED]

In the cluster in (3a), each tone corresponds to its own tonal root node. While one of the two tonal root nodes is specified (by the *) as obligatorily linking to the tone-bearing unit, the other tonal root node is independent and links later either to an open (i.e. unlinked) syllable or, if there are no available open syllables, to a syllable already linked with another tonal root node. These two possible tonal associations of the cluster, (3a), are shown in (4).

(4) Two possible associations of the tones in a cluster

a. Association to different syllables

[ILLUSTRATION OMITTED]

b. Association to the same syllable

[ILLUSTRATION OMITTED]

The flat structure makes certain predictions about tonal alignment. First of all, it predicts that there could exist two rising or two falling pitch accents with different tonal-alignment patterns. In the unit, (3b), the two tones would be dependent on each other in their alignment since they are both dominated by the same tonal root node, which is the part of the pitch accent marked to align with the tone-bearing unit. In the cluster, (3a), the tone contained by the unstarred tonal root node would be independent of the starred tone and its alignment could be affected by other factors such as whether there is an unlinked syllable to associate with, the distance before the next tonal specification, etc.

The other alternative to Pierrehumbert and Beckman's (1988) view of the pitch accent that Grice (1995b) offers is what she calls the hierarchical structure, which is based on the structure used by Nespor and Vogel (1986) for the prosodic word. Grice (1995b) calls this structure hierarchical because some tones may occur within an optional part of the pitch accent ([[tau].sub.w]) while others occur within the core of the pitch accent ([[tau].sub.s]). The hierarchical structure is shown in (5).

(5) Grice's (1995b) hierarchical pitch-accent structure

[ILLUSTRATION OMITTED]

In (5) the pitch-accent node (PA) is the equivalent of the tonal root node in (3). This dominates an intermediate level that Grice (1995b) labels the supertone level. The supertones ([tau]), in turn, dominate the tones (T, which refers to any tone, either H or L) themselves. As shown in (5), it looks as if a tritonal pitch accent would be possible, though Grice (1995b) specifies that branching may take place maximally at one level within a hierarchical pitch accent, either at the PA node or at the strong supertone node, but not both. The strong supertone node and the strong tone are obligatory portions of the pitch accent, while their weak counterparts are optional. Grice (1995b) assumes that there is language-specific parameter setting for headedness at each level to determine whether the strong branch is on the right or the left. In (5) the supertone level is right-headed while the tone level is left-headed, as indicated by the location of the strong nodes at each level.

Within this hierarchical structure, then, there are two possible resulting structures for bitonal pitch accents in each language. With the parameters set as in (5), the two resulting structures are as in (6).

(6) Example of bitonal pitch accents within the hierarchical structure

[ILLUSTRATION OMITTED]

The possible pitch-accent structures in (6) make predictions about tonal alignment just as did the structures in (1) and (3) above. The pitch accent in (6b) is basically the same as that in (3b), and the predictions are similar. The weak tone will align with respect to the strong tone since they are both dominated by the same supertone node and therefore are closely related structurally. This close relationship should be reflected in the alignment of the tones with the speech stream. Where the predictions for the pitch accent in (6b) differ from those for the pitch accent in (3b) is that in (6b) one tone at the tone level is marked as strong, and it therefore has priority over the weak tone for alignment with the stressed syllable. In (3b) neither tone has priority over the other, since alignment priority is determined at the tonal root level.

The pitch accent in (6a) is different from those presented above. It is similar to the pitch accent in (3a) in that the strong and weak tones are independent of each other, but there is a crucial difference. In (3a) the two tones are contained by separate tonal root nodes, making them completely independent units. In (6a), on the other hand, the two tones are independent of each other only in that they are not dominated by the same supertone node. They are, however, both dominated by the same PA node, which means that they are not totally independent. The prediction is that the tone contained by the weak supertone node in (6a) will show more freedom from the tone contained by the strong supertone node in its alignment patterns than will the weak tone from the strong tone in (6b); but it is also predicted that this freedom will be limited since the two tones are structurally related in (6a), unlike those in (3a).

The three types of structure seen in the above figures are each grounded in previous research. The structure in (1) has been a part of the AM theory for some time. The flat structures in (3) and (4) have been motivated strongly by data from tone languages but still need to be evaluated with regard to intonation languages. The hierarchical structure in (5) and (6) has been used for certain types of prosodic structure, and therefore extending it to other prosodic structures is worth investigating.

Very little work has been done on pitch-accent structure. Grice (1995b), who presents the two possible alternatives to Pierrehumbert and Beckman's (1988) view of the pitch accent, argues for the flat structure. Her argument, however, is not based on the predictions that the two competing structures make. Rather she argues for the flat structure based on the ability to use an intermediate stage in the derivation of a pitch accent to trigger English downstepping. Downstepping is the process by which a peak in fundamental frequency (F0) is produced substantially lower than the preceding peak. While analyzing English downstepping has proven problematic within the AM model, and it would be nice if an analysis would fall neatly out of something such as pitch-accent structure, this cannot be used to motivate one structure over another since neither makes any predictions whatsoever about downstepping. Interestingly, while she did not set out to directly examine pitch-accent structure, Frota (2000) notes that her European Portuguese tonal-alignment data seem to support the hierarchical structure, contra Grice's (1995b) argument for the flat structure. What is clearest from the small amount of past consideration given to pitch-accent structure is that more investigation of this topic is needed before a conclusion can be reached.

3. Methodology

It has been observed at least since Navarro Tomas (1944: section 20) that Spanish stressed syllables are accompanied by a rising F0 with the peak generally occurring in the post-tonic syllable. Examples of this are shown in Figures 1 and 2.

[FIGURES 1-2 OMITTED]

In these figures, the F0 rise begins near the beginning of the stressed syllable (delimited by the vertical lines marked with s > and S >) of the word termino `he finished' and continues into the following unstressed syllable. The same F0 pattern is not found associated with the stressed syllable of the word banana `banana', but this is due to its being the last word of the utterance (Face i.p.; Prieto et al. 1996).

De la Mota (1997) has shown that for words in narrow focus the F0 peak generally occurs within the stressed syllable, and Face (i.p., 2001) has shown that this is due to a contrastive-focus pitch accent in Spanish with early F0 peak alignment. Examples of the same utterances as in Figures 1 and 2, but with contrastive focus on the word termino `he finished' are given in Figures 3 and 4.

[FIGURES 3-4 OMITTED]

An experiment was designed for the present study to examine the tonal-alignment patterns of these two rising pitch accents in more detail for the purpose of investigating whether they motivate a particular internal pitch-accent structure.

The corpus used for the experiment consisted of sets of sentences with either two or three content words containing a stressed syllable, as well as additional unstressed words. The first two stressed syllables in each sentence were the target syllables under investigation. Within each set of sentences, the first target syllable of each sentence was segmentally identical to the first stressed syllable of the other sentences. Likewise, the second target stressed syllable of a sentence was segmentally identical to the second target stressed syllable of the other sentences within the same set. Within each set of sentences, the number of intervening unstressed syllables was varied so that there were between 0 and 5 unstressed syllables between the two target syllables. For the purposes of the present study, the second target syllable of those sentences with only two content words will be ignored, since this syllable is the final stressed syllable of the utterance. Face (i.p.) shows that the distinction between the pitch accents under consideration in the present study does not occur on the final stressed syllable of an utterance.

Each sentence included in the experiment was the answer to three different setup questions. One question forced a broad-focus reading of the sentence, one forced a reading with narrow focus on the word containing the first target syllable, and the last forced a reading with narrow focus on the word containing the second target syllable. An example of a sentence answering each of these types of question is given in (7).
(7) Example question and answer pairs with different focal contexts.
 (Italics mark target stressed syllables, capitals mark contrastive
 focus.)

 a. A: [??]Que dijo Pepe?
 what said 3rd sg. Pepe
 `What did Pepe say?'

 B: Que le daban el numero pertinente.
 that to him give-impert. 3rd pl. the number relevant
 `That they were giving him the relevant number'

 b. A: [??]Dijo Pepe que le pedian el
 said-3rd sg. Pepe that to him request-imperf. 3rd pl. the
 numero pertinente?
 number relevant
 `Did Pepe say that they were asking him for the relevant
 number?'

 B: No. Que le DABAN el numero
 no that to him ask-imperf. 3rd pl. the number
 pertinente.
 relevant
 `No. That they were GIVING him the relevant number.'

 c. A: [??]Dijo Pepe que le daban el
 said-3rd pl. Pepe that to him give-imperf. 3rd pl. the
 documento pertinente?
 document relevant
 `Did Pepe say that they were giving him the relevant
 document?'

 B: No. Que le daban el NUMERO
 no that to him give-imperf. 3rd pl. the number
 pertinente.
 relevant
 `No. That they were giving him the relevant NUMBER.'


By making use of sentences that are lexically and grammatically identical but vary only in the focal context, I was able to examine cases where only intonation differed from one sentence to the next. And since the stressed syllables are associated with an F0 rise, as has been shown above in Figures 1-4, and since the alignment of that rise differs between broad focus and narrow focus, these question-and-answer pairs allow a detailed examination of the differing tonal alignment between the broad-focus and contrastive-focus pitch accents. That is, on words produced in contrastive focus, an F0 peak is expected within the boundaries of a stressed syllable (cf. Figures 3 and 4), while when a word is produced in broad focus the F0 peak is expected in a following syllable (cf. Figures 1 and 2).

A total of 108 question-and-answer pairs were included in the experiment (see the Appendix). The questions were recorded by a male native speaker in his mid-twenties from Madrid, Spain. The recorded questions were played to the subjects, who, upon hearing the question, responded by reading the answer as they would respond in conversation to the particular question asked. These responses were recorded using a Sony MZ-R90 mini-disc recorder and a Sony ECM-MS907 microphone. A total of 20 subjects, both males and females, participated in the experiment. The subjects were all students at the Universidad Autonoma de Madrid and were between the ages of 22 and 29 years old. All subjects were born in Madrid and had never lived in another part of Spain. In addition, none of them spoke another language natively, though many had studied foreign languages in school. After the recording sessions were completed, the data was were digitized and analyzed using the PitchWorks software designed specifically for intonation studies. Measurements were made of the location of the beginning and the end of each rise in F0 in relation to the stressed syllable, and these measurements are the basis for the analysis presented in this paper.

4. Spanish tonal-alignment data

We have seen (Figures 1-4) that the broad-focus and the contrastive-focus pitch accents in Spanish are characterized by a rise in F0. These rises are attributable to pitch accents containing a sequence of two tones: a low tone (L) followed by a high tone (H). Thus, tonally, both of these pitch accents can be analyzed as L + H, where the L accounts for the valley at the beginning of the rise in F0 and the H accounts for the peak in F0 at the end of the rise. The difference between the two pitch accents is in the way in which this F0 rise aligns with the stressed syllable.

We have seen above that the location of the F0 peak is the biggest difference between the broad-focus and contrastive-focus pitch accent, with the peak of the broad-focus pitch accent occurring after the stressed syllable (cf. Figures 1 and 2) and the peak of the contrastive-focus pitch accent occurring within the stressed syllable (cf. Figures 3 and 4). For this reason, I will begin my quantitative examination of the alignment of the tones in these two pitch accents by considering the location of the F0 peak (i.e. the phonetic manifestation of the H) with respect to the end of the stressed syllable with which the pitch accent is associated.

While in Table 1 the results are given separately for initial position and medial position within the sentence, the results are similar. (2) The F0 peak resulting from the H of the broad-focus pitch accent occurs on the average between 52 ms and 60 ms after the end of the stressed syllable, while that resulting from the H of the contrastive-focus pitch accent occurs on the average between 45 ms and 48 ms before the end of the stressed syllable. In both initial and medial position an analysis of variance (ANOVA) shows that the type of pitch accent has a highly significant effect on the location of the F0 peak. (3)

While there is a clear difference in the location of the F0 peak between the two rising pitch accents under consideration, there does not seem to be a corresponding difference in the location of the F0 valley (cf. Figures 1 and 3; Figures 2 and 4). The average location of the F0 valley resulting from the L of the two pitch accents is shown in Table 2.

Table 2 shows that there is a much smaller difference in the location of the F0 valley resulting from the L between the two pitch accents than that found above (Table 1) for the location of the F0 peak. In both initial and medial position, the difference is only 8 ms. Despite this small difference, an ANOVA shows that the difference is statistically significant in initial position. This indicates that while the difference in the location of the F0 valley is small, it is also very consistent.

An ANOVA shows that the difference in the location of the F0 valley in medial position is not statistically significant, but this is a result of the experimental design. In the experiment conducted for the present study, the target syllable in medial position immediately followed the controlled number of unstressed syllables (see section 3). When there are fewer unstressed syllables between the two target syllables to which pitch accents are associated, the tones of the two pitch accents are closer together. This creates cases of tonal crowding, where multiple tones occur in a short period of time. Several previous studies on the intonation of different languages have shown that in cases of tonal crowding, the alignment of the tones is temporally adjusted (Arvaniti et al. 1998; Grice 1995a; Silverman and Pierrehumbert 1990; Face i.p.; Prieto et al. 1996 for Spanish). If the H of the preceding pitch accent becomes too close to the L of the pitch accent in medial position, it would push the L to the right, cutting into the difference in the location of the L between the two Spanish rising pitch accents. This is, indeed, the case. When the most extreme cases of tonal crowding (i.e. those cases with zero or one unstressed syllable intervening between the two target syllables) are removed from consideration, the difference in the location of the F0 valley between the broad-focus and contrastive-focus pitch accent is statistically significant. This result is shown in Table 3.

I have shown in this section that there is a statistically significant difference between the broad-focus and contrastive-focus pitch accent in the location of both the F0 peak and the F0 valley. The difference in the location of the F0 peak is quite large, with it being realized a considerable distance after the end of the stressed syllable in the broad-focus pitch accent and a considerable distance before the end of the stressed syllable in the contrastive-focus pitch accent. The difference in the location of the F0 valley is much smaller, yet consistent. The valley seems to be more strongly aligned to the stressed syllable in the broad-focus pitch accent than in the contrastive-focus pitch accent, though in both cases it is realized near the beginning of the stressed syllable. (4)

5. Analysis

In considering whether or not the Spanish tonal-alignment data reported in the preceding section motivate postulating an internal structure for pitch accents and, if so, what the nature of that structure is, the relationship between the two tones of the two Spanish rising pitch accents must be considered. It has always been accepted within the AM model of intonational phonology that pitch accents may contain two tones, but, as discussed in section 2, different views of the pitch accent make different predictions about the relationship between the two tones.

In the preceding section we saw that there is a statistically significant difference in the alignment of both tones between the two Spanish rising pitch accents. We also saw, however, that there is a much larger difference in the alignment of the H than in the alignment of the L. In the broad-focus pitch accent, the L is realized near the beginning of the stressed syllable (Table 2), and the H is realized a considerable distance after the stressed syllable (Table 1). In the contrastive-focus pitch accent, the H is realized within the stressed syllable (Table 1), but the L does not precede the stressed syllable in a fashion comparable to that in which the H of the broad-focus pitch accent follows it. That is, there is a different relationship between the two tones in the Spanish broad-focus pitch accent than there is between the two tones of the Spanish contrastive-focus pitch accent. This different relationship between the tones cannot be accounted for within the Pierrehumbert and Beckman (1988) view of the pitch accent shown in (1). The only way to account for this different relationship between the tones is to propose that the Spanish broad-focus and contrastive-focus pitch accents have different structures. Even so, it must be determined what the nature of that structure is. Does the flat structure, shown in (3), or the hierarchical structure, shown in (5) and (6), best account for the Spanish tonal-alignment data?

Both the flat structure in (3) and the hierarchical structure in (5) and (6) are capable of capturing differences in tonal alignment between pitch accents, such as the one that exists between the Spanish broad-focus and contrastive-focus pitch accents. The task at hand is to determine which of these structures best accounts for the difference in tonal alignment between the Spanish broad-focus and contrastive-focus pitch accents, taking into account the predictions that each structure makes for tonal alignment (see section 2).

The contrastive-focus pitch accent, with both the L and the H seemingly aligned with the stressed syllable, is represented nearly identically by these two competing structures. In both (3b) and (6b), both tones are dominated by the same node at the higher levels (i.e. the tonal root node in [3b] and the supertone in [6b]). This accounts for both the L and the H of the Spanish contrastive-focus pitch accent being aligned with the stressed syllable. In the case of the flat structure in (3b), the tonal root node containing both tones is marked to align with the stressed syllable, and therefore the two tones it contains should align with that syllable. In the case of the hierarchical structure in (6b), the dominance of both tones by the same supertone means that they have a strong relationship structurally, and this strong relationship should be reflected in the two tones being aligned closely to each other. Since one must be the head of the pitch accent (i.e. the strong tone) and align with the stressed syllable, and since the two tones are closely related and this should be reflected in their alignment, the prediction is that the weak tone should also be realized in or very near the stressed syllable. Thus, both the flat structure and the hierarchical structure are able to account for the alignment of the two tones in the Spanish contrastive-focus pitch accent.

With regard to the broad-focus pitch accent, both the flat structure in (3a) and the hierarchical structure in (6a) represent the independence of the two tones. In both cases the two tones are less strongly related structurally than in the other bitonal pitch accent within the same structure (i.e. [3b] and [6b]). As discussed in section 2, the crucial difference between (3a) and (6a) is that in the flat structure in (3a), the tones pertain to different root nodes and are completely independent of each other. In the hierarchical structure in (6a), the two tones pertain to different supertones, but these supertones are dominated by the same pitch-accent node, which is equivalent to the tonal root node. Therefore while the tones are somewhat independent of each other in both (3a) and (6a), in the hierarchical structure in (6a) they maintain a certain degree of dependency since they are dominated by the same pitch-accent node while in the flat structure in (3a) they are completely independent of each other.

We have already seen that in the broad-focus pitch accent, the L is aligned at or near the onset of the stressed syllable (Table 2) and the H is realized after the stressed syllable (Table 1). What remains unclear, however, is the degree of independence of the H from the L. Is the H completely independent of the L, or is it merely dependent to a lesser degree than in the contrastive-focus pitch accent? In order to answer this question, I examined the data for the alignment of the H in initial position when it immediately precedes the controlled number of unstressed syllables (see section 3). Based on what is known about tonal crowding (see section 4) and the temporal adjustment of tones realized in a short period of time in Spanish (Face i.p.; Prieto et al. 1996), if the L and the H of the broad-focus pitch accent are completely independent of each other, the L should crowd the H and push its realization to the right, meaning that it would be realized near the middle of the string of intervening unstressed syllables. If, on the other hand, the two tones are not completely independent of each other, the prediction would be that the H would remain near the L even when there are numerous intervening unstressed syllables where the H could potentially be realized.

The data for the alignment of the H depending on the number of following unstressed syllables are given in Table 4. In this table, as more unstressed syllables follow the stressed syllable, the H of the broad-focus pitch accent is realized further and further beyond the end of the stressed syllable. However, it is important to note that the amount of increase beyond the end of the stressed syllable decreases. This could indicate that the H is reaching a limit on how far beyond the end of the stressed syllable, and therefore how far from the L, it is willing to be realized. While this would provide evidence for the dependence of the two tones in the broad-focus pitch accent, and therefore for the hierarchical structure, this evidence is not convincing on its own. There is further evidence within Table 4, however, that the H is dependent on the L in the Spanish broad-focus pitch accent.

When there are five unstressed syllables following the stressed syllable, Table 4 shows that the H is realized on the average 109 ms after the end of the stressed syllable. Even though there are five unstressed syllables upon which the H could be realized, it is realized on the average only 109 ms after the stressed syllable, which in the vast majority of cases places it in the very first of the five unstressed syllables. If the H were independent of the L, the tonal crowding effects of the L should force the H to be realized much further into the string of unstressed syllables. Since this is not the case, it must be concluded that the H of the Spanish broad-focus pitch accent is not completely independent of the L. Since, then, the two tones are dependent on each other in both of the Spanish rising pitch accents, the hierarchical structure more accurately accounts for the Spanish tonal-alignment data, as within this structure the two tones are always structurally related. (5)

The findings presented for Spanish tonal alignment support the proposal of an internal pitch-accent structure that is able to account for the tonal-alignment differences between pitch accents. In addition, the lack of complete independence of the H from the L in the broad-focus pitch accent motivates postulating that the hierarchical structure more accurately accounts for the Spanish tonal-alignment data than does the flat structure.

Based on the above findings, I propose that the two Spanish rising pitch accents under consideration have the internal structures shown in (8).

(8) Spanish pitch-accent structures

a. Broad focus

[ILLUSTRATION OMITTED]

b. Constrastive focus

[ILLUSTRATION OMITTED]

In the broad-focus pitch accent, represented in (8a), the L and the H are somewhat independent of each other and therefore must be dominated by different supertones. Furthermore, since it is the L that is aligned with the stressed syllable, the supertone level must be left-headed in Spanish.

In the contrastive-focus pitch accent, represented in (8b), both tones seem to be fairly strongly aligned with the stressed syllable. This alignment is accounted for by their close structural relationship, with both tones being dominated by the same supertone. In the contrastive-focus pitch accent, the H is clearly more strongly aligned with the stressed syllable than in the broad-focus pitch accent (see Table 1), but this could be the result of its being the head of the pitch accent or the result of its being dominated by the same supertone as is the head. The statistically significant difference in the alignment of the L between the two pitch accents (see Table 2 and Table 3) demonstrates that the L is slightly less strongly aligned with the stressed syllable in the contrastive-focus pitch accent than it is in the broad-focus pitch accent of which it is the head. This less strong alignment of the L, in combination with the strong alignment of the H, motivates analyzing the H as the head of the contrastive-focus pitch accent, and therefore the tone level in Spanish as right-headed.

While the hierarchical structures proposed above for the two Spanish rising pitch accents accurately account for the tonal-alignment data, the question of how to represent these pitch accents in the standard AM notational scheme must be considered. The standard AM notation can serve as a convenient shorthand for the actual structure and in this role is an invaluable convenience. The Spanish broad-focus pitch accent has been analyzed as [L.sup.*] + H, indicating the alignment of the L, but not the H, to the stressed syllable (Face i.p., 2001; Hualde 1999; Sosa 1995, 1999). While previous analyses have taken into account tonal alignment without considering pitch-accent structure, this representation is straightforward and well motivated. And since the L is the head of this pitch accent in the analysis proposed here, this [L.sup.*] + H notation is compatible with the findings for pitch-accent structure.

The more difficult case is that of the contrastive-focus pitch accent. In those analyses that have postulated a difference between a broad-focus and a contrastive-focus pitch accent, the contrastive-focus pitch accent has been analyzed as L + [H.sup.*] (Face i.p.; Hualde 1999; Sosa 1999) and [(L + H).sup.*] (Face 2001). The L + [H.sup.*] analysis as used in these previous studies inaccurately predicts the lack of alignment of the L with the stressed syllable. The use of [(L + H).sup.*] was an attempt to account for both tones being aligned with the stressed syllable, but it presents the theoretically undesirable possibility of a language having three phonologically distinct rising pitch accents: [L.sup.*] + H, L + [H.sup.*] and [(L + H).sup.*]. Once the pitch-accent structure proposed here is assumed, however, the difficulty disappears. If the * marks the head of the pitch accent, the contrastive-focus pitch accent can be notated as L + [H.sup.*]. It is crucial, however, that the * marks the head of the pitch accent, without making any prediction about the alignment of the unstarred tone, which is determined by the overarching pitch-accent structure.

6. Conclusion

In this paper I have presented the tonal-alignment data for two Spanish rising pitch accents in order to determine whether these data motivate postulating that pitch accents have an internal structure and, if so, what the nature of that structure is. I have shown that the Spanish broad-focus pitch accent is characterized by a rise that begins near the onset of the stressed syllable and continues beyond the end of that syllable. The Spanish contrastive-focus pitch accent is also characterized by a rise that begins near the beginning of the stressed syllable, but the rise reaches its maximum within the stressed syllable.

The relationship between the two tones has been shown to differ between the two pitch accents, and this difference cannot be accounted for within the standard view of the pitch accent held by Pierrehumbert and Beckman (1988). However, if internal pitch-accent structure is postulated, the two pitch accents can simply be analyzed as having different internal structures, accounting for the different relationship between the tones. Two possible internal structures that Grice (1995b) has proposed as possible alternatives to the standard view of the pitch accent have been considered: (1) a flat structure based on Yip's (1989) work on Chinese and African languages, and (2) a hierarchical structure based on Nespor and Vogel's (1986) structure of the prosodic word.

Either of these two possible alternatives to the standard view of the pitch accent is able to account for the alignment of the tones in the Spanish contrastive-focus pitch accent. Only the hierarchical structure, however, is able to account for the alignment of the tones in the Spanish broad-focus pitch accent. While the two tones in the broad-focus pitch accent are more independent of each other than are the tones in the contrastive-focus pitch accent, they are not completely independent. The hierarchical structure makes exactly this prediction, while the flat structure predicts total independence. On these grounds it is argued that it is the hierarchical structure that most accurately represents the Spanish tonal-alignment data.

While the proposed hierarchical structure provides the most accurate representation of the data, it is also of interest to provide a simple notation within the standard notational scheme of the AM model of intonational phonology. If the head of each pitch accent is marked with *, the Spanish broad-focus pitch accent can be represented as [L.sup.*] + H, since the L is its head, and the contrastive-focus pitch accent can be represented as L + [H.sup.*], since the H is its head.

The analysis presented in the present study has potential implications for the analysis of the intonation of other languages. I have shown for Spanish that pitch-accent structure can account for the particulars of the alignment of the tones within a pitch accent. This may allow the analyses of other languages to be simplified in that tonal alignment can be attributed to pitch-accent structure rather than to merely descriptive phonetic-implementation rules. In addition, work on pitch-accent structure using data from languages other than Spanish will be important in order to determine whether the proposed hierarchical structure is employed universally or whether different types of structure are found in different languages.

University of Minnesota
Received 24 January 2000
Revised version received
8 March 2001


Appendix

The question-and-answer pairs used for the present study are listed below. The versions including all contents resulted in the sentences with three stressed words. The versions with the parenthetical content excluded resulted in the sentences with two stressed words.

Set A

No intervening unstressed syllables

A: [??]Que dijo Manuel?

B: Que le da numeros (pertinentes).

A: [??]Dijo Manuel que le pide numeros (pertinentes)?

B: No. Que le DA numeros (pertinentes).

A: [??]Dijo Manuel que le da documentos (pertinentes)?

B: No. Que le da NUMEROS (pertinentes).

One intervening unstressed syllable

A: [??]Que dijo Ana?

B: Que le daba numeros (pertinentes).

A: [??]Dijo Ana que le pedia numeros (pertinentes)?

B: No. Que le DABA numeros (pertinentes).

A: [??]Dijo Ana que le daba documentos (pertinentes)?

B: No. Que le daba NUMEROS (pertinentes).

Two intervening unstressed syllables

A: [??]Que dijo Pepe?

B: Que le daban el numero (pertinente).

A: [??]Dijo Pepe que le pedian el numero (pertinente)?

B: No. Que le DABAN el numero (pertinente).

A: [??]Dijo Pepe que le daban el documento (pertinente)?

B: No Que le daban el NUMERO (pertinente).

Three intervening unstressed syllables

A: [??]Que dijo Maria?

B: Que le dabamos el numero (pertinente).

A: [??]Dijo Maria que le pediamos el numero (pertinente)?

B: No. Que le DABAMOS el numero (pertinente).

A: [??]Dijo Maria que le dabamos el documento (pertinente)?

B: No. Que le dabamos EL NUMERO (pertinente).

Four intervening unstressed syllables

A: [??]Que dijo Ignacio?

B: Que se lo daba para el numero (pertinente).

A: [??]Dijo Ignacio que se lo pedia para el numero (pertinente).

B: No. Que se lo DABA para el numero (pertinente).

A: [??]Dijo Ignacio que se lo daba para el documento (pertinente)?

B: No. Que se lo daba para EL NUMERO (pertinente).

Five intervening unstressed syllables

A: [??]Que dijo Pilar?

B: Que se lo dabamos para el numero (pertinente).

A: [??]Dijo Pilar que se lo pediamos para el numero (pertinente)?

B: No. Que se lo DABAMOS para el numero (pertinente).

A: [??]Dijo Pilar que se lo dabamos para el documento (pertinente)?

B: No. Que se lo dabamos para EL NUMERO (pertinente).

Set B

No intervening unstressed syllables

A: [??]Que dijo Victor?

B: Que lo termino Nana (ayer).

A: [??]Dijo Victor que lo comenzo Nana (ayer)?

B: No. Que lo TERMINO Nana (ayer).

A: [??]Dijo Victor que lo termino Pablo (ayer)?

B: No. Que lo termino NANA (ayer).

One intervening unstressed syllable

A: [??]Que dijo Sonia?

B: Que lo termino la nana (de los ninos).

A: [??]Dijo Sonia que lo comenzo la nana (de los ninos)?

B: No. Que lo TERMINO la nana (de los ninos).

A: [??]Dijo Sonia que lo termino la abuela (de los ninos)?

B: No. Que lo termino LA NANA (de los ninos).

Two intervening unstressed syllables

A: [??]Que dijo Raul?

B: Que termino la banana (de la chica).

A: [??]Dijo Raul que vio la banana (de la chica)?

B: No. Que TERMINO la banana (de la chica).

A: [??]Dijo Raul que termino la manzana (de la chica)?

B: No. Que termino LA BANANA (de la chica).

Three intervening unstressed syllables

A: [??]Que dijo Carmen?

B: Que termino con la banana (de la chica).

A: [??]Dijo Carmen que jugo con la banana (de la chica)?

B: No. Que TERMINO con la banana (de la chica).

A: [??]Dijo Carmen que termino con la manzana (de la chica)?

B: No. Que termino con LA BANANA (de la chica).

Four intervening unstressed syllables

A: [??]Que dijo Eduardo?

B: Que termino lo de la banana (de la chica).

A: [??]Dijo Eduardo que comenzo lo de la banana (de la chica)?

B: No. Que TERMINO lo de la banana (de la chica).

A: [??]Dijo Eduardo que termino lo de la manzana (de la chica)?

B: No. Que termino lo de LA BANANA (de la chica).

Five intervening unstressed syllables

A: [??]Que dijo Manuela?

B: Que termino con lo de la banana (de la chica).

A: [??]Dijo Manuela que comenzo con lo de la banana (de la chica)?

B: No. Que TERMINO con lo de la banana (de la chica).

A: [??]Dijo Manuela que termino con lo de la manzana (de la chica)?

B: No. Que termino con lo de LA BANANA (de la chica).

Set C

No intervening unstressed syllables

A: [??]Que dijo Jose?

B: Que los comparara Mar (en enero).

A: [??]Dijo Jose que los leera Mar (en enero)?

B: No. Que los COMPARARA Mar (en enero).

A: [??]Dijo Jose que los comparara Pilar (en enero)?

B: No. Que los comparara MAR (en enero).

One intervening unstressed syllable

A: [??]Que dijo Nieves?

B: Que los comparara Mar (en enero).

A: [??]Dijo Nieves que los leyera Mar (en enero)?

B: No. Que los COMPARARA Mar (en enero).

A: [??]Dijo Nieves que los comparara Pilar (en enero)?

B: No. Que los comparara MAR (en enero).

Two intervening unstressed syllables

A: [??]Que dijo Antonio?

B: Que los comparara con Mar (en enero).

A: [??]Dijo Antonio que los leyera con Mar (en enero)?

B: No. Que los COMPARARA con Mar (en enero).

A: [??]Dijo Antonio que los comparara con Pilar (en enero)?

B: No. Que los comparara con MAR (en enero).

Three intervening unstressed syllables

A: [??]Que dijo Isabela?

B: Que los comparara para Mar (en enero).

A: [??]Dijo Isabela que los leyera para Mar (en enero)?

B: No. Que los COMPARARA para Mar (en enero).

A: [??]Dijo Isabela que los comparara para Pilar (en enero)?

B: No. Que los comparara para MAR (en enero).

Four intervening unstressed syllables

A: Que dijo Pablo?

B: Que los compararamos para Mar (en enero).

A: [??]Dijo Pablo que los leyeramos para Mar (en enero)?

B: No. Que los COMPARARAMOS para Mar (en enero).

A: [??]Dijo Pablo que los compararamos para Pilar (en enero)?

B: No. Que los compararamos para MAR (en enero).

Five intervening unstressed syllables

A: [??]Que dijo Rosa?

B: Que los compararamos para Lamar (en enero).

A: [??]Dijo Rosa que los leyeramos para Lamar (en enero)?

B: No. Que los COMPARARAMOS para Lamar (en enero).

A: [??]Dijo Rosa que los compararamos para Pilar (en enero)?

B: No. Que los compararamos para LAMAR (en enero).
Table 1. Location of F0 peaks resulting from high tones by
pitch accent type and position within the sentence

Pitch-accent type Location of F0 peak (a)
 initial position medial position

Broad focus +60 ms +52 ms
Contrastive focus -48 ms -45 ms
Probability (ANOVA) p < 0.0001 p < 0.0001

(a.) With respect to the end of the stressed syllable. A positive
value means that the F0 peak is realized after the end of the
stressed syllable while a negative value means it is realized
within the stressed syllable.
Table 2. Location of F0 valleys resulting from low tones by
pitch accent type and position within the sentence

Pitch-accent type Location of F0 valley (a)
 initial position medial position

Broad focus +4 ms +16 ms
Contrastive focus -12 ms +8 ms
Probability (ANOVA) p < 0.0001 p = 0.0689

(a.) With respect to the beginning of the stressed syllable. A
positive value means that the F0 peak is realized within the
stressed syllable while a negative value means it is realized
before the stressed syllable.
Table 3. Location of F0 valleys in medial position (two or
more intervening unstressed syllables)

Pitch-accent type Location of F0 valley
 in medial position (a)

Broad focus +2 ms
Contrastive focus -8 ms
Probability (ANOVA) p = 0.0005

(a.) With respect to the beginning of the stressed syllable. A
positive value means that the F0 peak is realized within the
stressed syllable while a negative value means it is realized
before the stressed syllable.
Table 4. Location of F0 peaks resulting from high tones in the
broad-focus pitch accent in initial position by the number of
following unstressed syllables

Following unstressed Location of F0 peak (a)
syllables

0 -5 ms
1 +35 ms
2 +61 ms
3 +86 ms
4 +94 ms
5 +109 ms

(a.) With respect to the end of the stressed syllable. A positive
value means that the F0 peak is realized after the end of the
stressed syllable while a negative value means it is realized
within the stressed syllable.


Notes

* Correspondence address: Department of Spanish and Portuguese Studies, University of Minnesota, 34 Folwell Hall, 9 Pleasant Street SE, Minneapolis, MN 55455, USA. E-mail: facex002@tc.umn.edu.

(1.) If Pierrehumbert and Beckman's (1988) representation of the pitch accent is not considered to be a structure, but rather a mere representation of the sequencing of the tones, then no predictions are made with respect to tonal alignment. This, however, is problematic in dealing with consistent tonal-alignment differences between the pitch accents of a language, such as those that I discuss in section 5.

(2.) Initial position and medial position here refer to the first and second stressed syllables of a sentence, respectively. The initial stressed syllable, however, is not the very first syllable of the sentence.

(3.) De la Mota (1997) and Face (2000, 2001) have shown that stressed syllables lengthen when produced in contrastive focus. This opens the possibility that the occurrence of the F0 peak within the stressed syllable is merely the result of syllable lengthening, and that there are not two phonologically distinct rising pitch accents. Face (2001), however, shows that while the stressed syllable lengthens in contrastive focus, the rise is also significantly shorter. Thus, in spite of syllable lengthening, there are indeed two phonologically distinct rising pitch accents.

(4.) One question that could be raised is whether the difference in the F0 rise is attributable to a phonological distinction between pitch accents or to a more canonical (i.e. less sloppy) pronunciation in cases of contrastive focus. The statistically significant stronger alignment of the valley to the stressed syllable in the broad-focus pitch accent is evidence that these are actually two phonologically distinct pitch accents. While the F0 peak is less strongly aligned in broad focus, the F0 valley is more strongly aligned. This supports the evidence presented by Face (i.p., 2001) and Sosa (1999) that the differences between these two rises are attributable to their being the result of phonologically distinct pitch accents.

(5.) I have based the selection of the hierarchical structure solely on its ability to account for the Spanish tonal-alignment data. It should also be pointed out that it is preferable on other grounds as well. If the two tones were completely independent of each other, as the flat structure predicts, it would have to be explained how two independent tones are generated. If they are part of a single unit, as in the hierarchical structure, this is not an issue, as it is then the single pitch-accent node that is generated, leading to a one-to-one correspondence between pitch accents and stressed syllables.

References

Arvaniti, Amalia; Ladd, D. Robert; and Mennen, Ineke (1998). Stability of tonal alignment: the case of Greek prenuclear accents. Journal of Phonetics 26, 3-25.

Beckman, Mary E.; and Pierrehumbert, Janet B. (1986). Intonational structure in Japanese and English. Phonology Yearbook 3, 255-310.

Bruce, Gosta (1977). Swedish Word Accents in Sentence Perspective. Lund: Gleerup.

De la Mota, Carme (1997). Prosody of sentences with contrastive new information in Spanish. In ESCA Workshop on Intonation: Theory, Models and Applications, 75-78. Athens: ESCA, University of Athens.

Face, Timothy L. (2000). Prosodic manifestations of focus in Spanish. Southwest Journal of Linguistics 19, 45-62.

--(2001). Focus and early peak alignment in Spanish intonation. Probus 13, 223-246.

--(i.p.). A phonological analysis of rising pitch accents in Castilian Spanish. Hispanic Linguistics 11.

Frota, Sonia (2000) Prosody and Focus in European Portuguese. New York: Garland.

Grice, Martine (1995a). The Intonation of Interrogation in Palermo Italian. Tubingen: Niemeyer.

--(1995b). Leading tones and downstep in English. Phonology 12, 183-233.

Hualde, Jose Ignacio (1999). Basic intonational contours in Spanish. Paper presented at the First Sp-ToBI Workshop, Columbus, Ohio, 1-3 October.

Ladd, D. Robert (1996). Intonational Phonology. Cambridge: Cambridge University Press.

Liberman, Mark (1975). The intonational system of English. Unpublished doctoral dissertation, MIT.

Navarro Tomas, Tomas (1944). Manual de entonacion espanola. New York: Hispanic Institute in the United States.

Nespor, Marina; and Vogel, Irene (1986). Prosodic Phonology. Dordrecht: Foris.

Pierrehumbert, Janet B. (1980). The phonology and phonetics of English intonation. Unpublished doctoral dissertation, MIT. (Distributed by the Indiana University Linguistics Club in 1987.)

--; and Beckman, Mary (1988). Japanese Tone Structure. Cambridge, MA: MIT Press.

Prieto, Pilar; Shih, Chilin; and Nibert, Holly (1996). Pitch downtrend in Spanish. Journal of Phonetics 24, 445-473.

--; van Santen, Jan; and Hirschberg, Julia (1995). Tonal alignment patterns in Spanish. Journal of Phonetics 23, 429-451.

Silverman, Kim E. A.; and Pierrehumbert, Janet B. (1990). The timing of prenuclear high accents in English. In Papers in Laboratory Phonology I, John Kingston and Mary E. Beckman (eds.), 72-106. Cambridge: Cambridge University Press.

Sosa, Juan Manuel (1995). Nuclear and pre-nuclear tonal inventories and the phonology of Spanish declarative intonation. In Proceedings of the XIIIth International Congress of Phonetic Sciences, Stockholm. Sweden, 646-649. Stockholm: University of Stockholm.

--(1999). La entonacion del espanol. Madrid: Catedra.

Yip, Moira (1989). Contour tones. Phonology 6, 149-174.
COPYRIGHT 2002 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2002 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Face, Timothy L.
Publication:Linguistics: an interdisciplinary journal of the language sciences
Date:Mar 1, 2002
Words:8755
Previous Article:Le: from pronoun to intensifier *.
Next Article:Is there a Spanish imperative intonation revisited: local considerations *.
Topics:


Related Articles
Focus and topic in VP-anaphora constructions(*).
La entonacion del espanol.
Government must act now; Comment.
Aculab unveils text-to-speech software. (New Products).
Why quantitative (1)?
Is there a Spanish imperative intonation revisited: local considerations *.
Focus as perspectivation (1).
Focus in double object constructions *.
Aspects of the prosody of Kuot, a language where intonation ignores stress (1).
Solving speech mystery.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters