Non-local pitch range relationships in read and elicited speech Alejna Brugos

Một phần của tài liệu Workshop on Prosody and Meaning Barcelona September 17-18, 2009 (Trang 48 - 53)

Boston University abrugos@bu.edu

The Strict Layer Hypothesis (Selkirk,1984; Nespor & Vogel, 1986) stipulates that all utterances are parsed into Full Intonational phrases (IPs), which in turn are parsed into one or more intermediate phrases (or equivalent). Theory-based annotation systems, such as ToBI (Beckman et al., 2005), further provide definition of tonal and timing-related properties of intermediate phrases and full IPs. The resulting system leaves little theoretical space for an intonationally defined grouping of prosodic units that are analyzed as Full IPs. Many researchers have observed, however, that tone relationships appear to extend beyond the boundaries of a local prosodic domain, especially in cases where an utterance consists of multiple intermediate phrases, i.e. where the syntactic and/or semantic content of a sequence of phrases is tightly grouped (Ladd 1988, 1992; van den Berg et al. 1992, inter alia).

In experiments with British speakers, Ladd (1988, 1992) showed that in multi- clause utterances, the degree of partial reset (measured via the f0-scaling of clause-initial accent peaks) was sensitive to whether clauses were grouped more tightly, via the coordinating conjunction and, or less tightly, as set off by but (see Figure 1 and Table 1). Truckenbrodt & Féry (2003) and Féry & Truckenbrodt (2005) found similar results for German, further suggesting also that reset values of sentences with the order but/and are comparable to those of sentences without additional branching structure, as with clause-internal downstep, or sentences with clauses with only and-type conjunction.

This study seeks to reproduce results from Ladd, and Féry & Truckenbrodt, and to test the hypothesis that these patterns will hold when additional syntactic and prosodic complexity is introduced. It is hypothesized that speakers will be capable of producing patterns sensitive to structure that extend beyond the confines even of Full IPs, testing the limits of stricter versions of the Prosodic Hierarchy. Pilot data from 4 speakers of American English, who produced sentences of 3 conjoined clauses of comparable syntactic structure and length.

Sentences varied by conjunction order (and/but vs. but/and, or and only) and by length (simple or branching subject, simple or branching object) (see Table 2).

The introduction of branching subject and object nodes was intended to increase the likelihood of prosodic breaks within the clauses, as well as between clauses.

Subjects read sentences aloud and produced sentences comparable to those in the reading task using a picture-based elicitation. The f0 of clause-initial accent- related peaks for 221 tokens was measured.

Comparison of f0 values from clause-initial peaks of the first and third clauses (A and C, see Figure 1) support results from previous studies: the average between-peak difference is smaller for the and/but order than for but/and in all length conditions (Table 3) in both read and picture-elicited productions (Table 4).

This pattern held in spite of varying degrees of prosodic complexity within the clauses, as more complex sentences (especially those with both subject and object branching) were more likely to include additional prosodic breaks.

Sentences produced via the picture elicitation also contained frequent

disfluencies (e.g., within-clause pauses), yet the same reset pattern was observed.

These results support the notion that the phonetic implementation of utterances is sensitive to syntactic complexity and meaning relationships between phrases. The notion that these sentences are produced as a single complex prosodic unit, containing multiple Full IPs, calls for reanalyzing aspects of prosodic hierarchies (Selkirk, 1984; Nespor & Vogel, 1986), such as with recursion of prosodic levels, as suggested by Féry (2008) and Wagner (2005). Such results also suggest a need for expansion of prosodic annotation systems to allow for non-local tone relationships.

47

Figures and Tables

Figure 1: The structure of 3-clause sentences with varying conjunction orders

a. but/and b. and/but c. and only

A but [ B and C ] [ A and B ] but C A (and) B and C Table 1: Examples from Ladd (1988)

Conjunctions/structure sample sentence but/and

A but [B and C]

Ryan has a lot more money, but Warren is a stronger campaigner, and Allen

has more popular policies.

and/but

[A and B] but C

Allen is a stronger campaigner, and Ryan has more popular policies, but

Warren has a lot more money.

Table 2: Examples of sentence types from this study:

conjunction complexity sample sentence

but/and

A but [B and C]

Simple subject, simple object

The bunny got a lemon, but the rhino got an onion and the mummy got a flower.

And/but [A and B] but C

Branching subject, branching object

The baby in the bunny costume won a flower and a bunch of balloons and the baby in the rhino costume won a ribbon and a box of crayons, but the baby in the mummy costume won a lemon and a bowl of berries.

Table 3: Average difference in semitones between A1 (f0 of initial accent-related peak of the first clause) and C1 (f0 of initial accent-related peak of the third clause), shown by sentence length and conjunction order, for all 4 subjects.

and only but/and and/but # of

tokens

Simple subject/simple object 5.00 4.94 4.65 N=55

Simple subject/branching object 3.98 4.19 2.71 N-55

Branching subject/simple object 3.81 3.19 2.56 N=56

branching subject/branching object 2.11 3.09 1.41 N=55

TOTAL 3.72

(N=93)

3.86 (N=64) 2.83 (N=64) N=221

Table 4: Average difference in semitones between A1 (f0 of initial accent-related peak of the first clause) and C1 (f0 of initial accent-related peak of the third clause), shown by elicitation method, for all 4 subjects.

and only but/and and/but # of tokens

Read 3.87 3.78 2.85 128

Elicited 3.38 3.96 2.82 93

TOTAL 3.72 (N=93) 3.86 (N=64) 2.83 (N=64) N=221

References

Beckman, M., Hirschberg, J & Shattuck-Hufnagel, S. (2005) The original ToBI system and the evolution of the ToBI framework. In Prosodic Typology: The Phonology of Intonation and Phrasing. Jun, S., ed. OUP: 9-54.

van den Berg, R., Gussenhoven C. & Rietveld T. (1992) Downstep in Dutch: Implications for a model. In Docherty & Ladd (eds.) Papers in Laboratory Phonology II: Gesture, segment, prosody. Cambridge: CUP. 335-367.

Fery, C. (2008) Syntax, Information Structure, Embedded Prosodic Phrasing and the Relational Scaling of Pitch Accents, ms.

Féry, C. & Truckenbrodt, H. (2005) Sisterhood and Tonal Scaling. Studia Linguistica.

Special Issue “Boundaries in intonational phonology”. 59.2/3. 223-243.

Nespor, M. & I. Vogel (1986) Prosodic phonology. Dordrecht: Foris.

Ladd D. R. (1988) Declination ‘reset’ and the hierarchical organization of utterances.

JASA 84, 530–544.

Ladd, D. R. (1990) Metrical representation of pitch register. In Papers in Laboratory Phonology I: Between the grammar and physics of speech. ed. J. Kingston & M.

Beckman, 35–57. Cambridge: Cambridge University Press.

Pierrehumbert, J. (1980) The phonology and phonetics of English intonation. PhD Dissertation, MIT, Cambridge, MA.

Selkirk, E. (1984) Phonology and Syntax: The Relation between Sound and Structure.

Cambridge: MIT Press.

Truckenbrodt, H. & Féry, C. (2003) More on hierarchical organization and tonal scaling.

Ms.

Wagner, M. (2005) Prosody and recursion. Doctoral Dissertation, MIT, Cambridge, Mass.

49

Một phần của tài liệu Workshop on Prosody and Meaning Barcelona September 17-18, 2009 (Trang 48 - 53)

Tải bản đầy đủ (DOC)

(166 trang)
w