Paragraph-Declination in Read Speech of Native and Non-Native Speakers of English

Contents

I. Introduction

II. Theoretical part
II.1 Tone languages and intonation languages
II.2. Declination in prosody
II.3. External and internal definitions of large-scale prosodic domains
II.3.1. Paratones
II.3.2. Paragraph declination or supradeclination

III. Empirical part
An analysis of pitch patterns of read speech across paragraphs;
A Chinese speaker of English and a native speaker of English in comparison
III.1. Method: Data collection and analysis
III.2. Findings
III.3. Discussion

IV. Summary

V. References

VI. Appendix

I. Introduction

Declination in prosody refers to the fact that the pitch of the voice is most commonly on average lower at the end of a prosodic unit than it is at the beginning, especially in controlled (read) speech. The main concerns of this paper are the analysis and comparison of the paragraph- declination of read speech of native and non-native speakers of English. More precisely, the analysis and comparison of the paragraph-pitch-patterns in the English of a native speaker and of a speaker with the tone language Chinese as her native language. Do distinct differences between the pitch-curves (i.e. the curves of the F0-formants) exist, and, if so, what might be the reasons for those differences? Does the particular feature of the tone language Chinese, i.e. the word-meaning-defining pitch, in any way affect paragraph-declination in L2-English? Or does this feature have no effect on paragraph-declination at all, maybe for physiological reasons? For this purpose, two speech samples of read speech will be divided into paratones and the pitch-curves of the paratones of the intonation and tone language will then be measured and compared. But beforehand, a brief outline of the main differences between tone and intonation languages will be provided. A short definition of declination in prosody and two ways of defining prosodic domains follow.

II. Theoretical Part

II.1 Tone languages and intonation languages

Tone languages are languages where the use of pitch determines word-meaning. Tone is thus a feature of the lexicon and is being described in terms of prescribed pitches for syllables or sequences of pitches for morphemes or words. A change of tone can then, with or without the use of affixes, change the meaning of a word. Cruttenden (1997: 9) gives the following example of minimal pairs from Szechuanese, a variety of Chinese:

illustration not visible in this excerpt

Most of the world’s languages are tone languages, among them are Chinese languages as e.g. Cantonese or Mandarin, African languages as Hausa or Kanuri or Indoeuropean languages as Punjabi. Intonation languages as e.g. English and most of the European languages in contrast use tone or pitch, respectively, only on the level of phrases or sentences. Thus, a change of word-pitch does not produce a distinctive word-meaning.1

II.2. Declination in prosody

Declination in prosody, also referred to as 'drift', describes ‘the natural tendency of syllables in a tone-unit to descend gradually in pitch’(Couper-Kuhlen, 1986: 82). Declination may not always be perceived conciously as hearers probably make unconscious adjustment for declination in their perception of intonation. Listening tests with synthesized material without declination have shown that such speech sounds unnatural,

1 Intonation on a sentence level in the tone language Bejing Chinese is also possible and is achieved by the use of at least four distinct intonational devices: channelling, tempo, focusing and the use of intonation carriers. They give the „final shape to the prosody of strings of syllables of breath group size which are shaped already by primary and secondary prosodic modifications“ Focusing, for signalling contrastive stress for example, is a momentary, gradual enlargement of a channel. (See Kratochvil in Hirst & Di Cristo 1998: 417 for further details).

thus the previous supposition is substantiated (cf. Couper-Kuhlen, 1986: 82). Furthermore, declination has been attested in a wide variety of unrelated languages and is, according to Couper-Kuhlen, a ‘universal tendency of physiological origin’ (cf. Couper-Kuhlen, 1986: 83), namely the gradual decrease in sub-glottal air- pressure which occurs automatically during the exhalation phase of respiration regulated for speech. The lower the air pressure, the lower the vibration rate of the vocal chords (in Hz) and the lower the pitch.

illustration not visible in this excerpt

Wichmann (2000: 103) presents two main approaches to modelling (sentence) declination: one being the overlay model, the other one being the downstep model (see Figure 1):

Figure 1. (a) F0 contour of an isolated, neutral statement, showing canonical declination patter; (b) illustration of declination: overlay model and downstep model. Wichmann (2000: 103)

Explaining those two main approaches, Wichmann cites a musical analogy from Ladd:

‘In one view declination could be compared to a musical tune which consists of a descending sequence of notes. In the other view it could be compared to a tune performed by an unaccompanied singer who gradually goes flat. In the first model, the falling pitch is part of the tune; in the second it is not. In the first, the ‘same’ notes will be at the same pitch; in the second, each successive occurrence of the ‘same’ note will be at a slightly lower pitch, usually imperceptible to the listener.‘ (Wichmann 2000:103)

However, there does not seem to be full agreement on the question which of the two models describes the phenomenon more accurately. At least when regarding global declination, i.e. declination of bigger prosodic units such as paragraphs or paratones (see section II.3.1 ), Ladd (1996: 331) states that it is widely accepted that the downstep model delivers a more accurate description. Trying to analyse paragraph declination, I will stick to this model in this paper.

Cruttenden (1986: 167) points out that ‘declination which occurs in neutral sentences (i.e. sentences said specially for experimental purposes) is represented by a slightly declining base line and a more steeply declining top line (see Figure 2 below). Thus, a narrowing of pitch range is produced as the intonation group progresses. The baseline and the top line are then ‘reset’ at the beginning of the next group; the top line may not, however, be reset on the same level as at the beginning of the previous group because intonation-groups may themselves be linked together by key.’ One has to note that declination rarely occurs in such a neat way in natural data, i.e. conversation, due to several obscuring factors which will be dealt with in section II.3.2.

II.3. External and internal definitions of large-scale prosodic domains

Trying to model paragraph intonation, two possibilities exist. The first one

is to define large-scale prosodic domains with the help of external citeria, i.e.

boundary features, the second one is to define them on internal critera, as e.g.

coherent patterns of internal systematicy. Both approaches will be used in the

empirical part of this paper.

II.3.1. Paratones

The term ‘paratone’ stands in analogy to the term ‘paragraph’ in writing and describes ‘intonational patterns associated with a series of related utterances.’(Wichmann 2000: 105). Paratones are much shorter than paragraphs in writing, especially in spontaneous speech. In read speech they often align with written paragraphs but this is not obligatory. Paratones are defined purely in terms of external boundary features such as: ‘a low pitch and a lengthy pause followed by a high sentence onset’(Wichmann 2000: 106) and also by a preceding low F0-formant. Consequently, a distinction is made between larger speech units, major paratones and smaller units, minor paratones. Major paratones are characterized by very high peaks and follow topic pauses, minor paratones begin with lower peaks and follow contour pauses. (cf. Wichmann 2000: 105).

II.3.2. Paragraph declination or supradeclination

The other way of defining paragraph intonation is to examine internal, global pitch trends across paragraphs. Here, the term s upradeclination describes a ‘gradual lowering of pitch over a topic unit (or paragraph)’ (Wichmann 2000: 107). This gradual lowering is, according to Wichmann (2000: 107) sometimes expressed in terms of the average F0, the initital onset height or the onset height of both topline or baseline. Evidence for this kind of pitch trend behaviour comes from controlled, experimental studies using specially constructed texts read aloud. Wichmann (2000: 107) cites a study from Sluijter and Terkens (1993) where a five-sentence paragraph was specially designed so that the same sentence could occur in four out of five positions:

‘Averaging over a number of speakers they found that while the sentence finished at the same pitch regardless of its position in the paragraph, the initial values of both the topline and the baseline were on average higher, the earlier the sentence occurred in the paragraph.’

(Wichmann 2000: 107) (see Figure 2 below).

illustration not visible in this excerpt

Figure 2: Average starting and ending pitch values of both topline and baselineof sentences in initial, medial and final position in a read paragraph

(from Wichmann 2000:17, after Sluijter and Terken 1993)

However, this ideal pattern can rarely be found in naturally occurring data because several factors, such as ‘textual or situational effects’ (Wichmann 2000: 108), obscure this underlying orderliness:

Information structure affects the pitch height of the accents associated with non-focal information and delays the ‘normal’ sentence-initial reset until later. A topic phrase shift within one sentence might be a reason for a delay of the highest peak. Also, the occurrence of ‘given’ items at the beginning of a sentence and with it its reduced pitch height can delay the highest peak. Wichmann (2000: 109) lists a whole set of structures which may cause such a delay: reporting clauses, adverbial expressions of time or place, existential ‘there‘, cleft sentences, initial subordinate clauses and noun phrases, particularly with a definite determiner. Figure 3 (overleaf) displays such a delaying effect of initial, non-focal information on the position of the highest F0 value.

[...]