PWorkshop Archives: Spring Term 2001

09 Jan 2001

Wing Chiu Ivan Yuen

Declination has been widely reported in both tonal and non-tonal languages. It is modelled as a baseline against which each phonological accent/tone is realised. A production experiment was conducted on Cantonese to test the presence of such baseline and whether the baseline interacts with different tones.

16 Jan 2001

Journal Update

23 Jan 2001

Journal Update

30 Jan 2001

Journal Update

06 Feb 2001

Dr Mits Ota

In this talk I will re-examine the so-called minimal word hypothesis, which states that early words are single binary feet. After a quick review of some of the previous supporting arguments and their empirical problems, I will present new types of evidence from child Japanese, which exhibits a period where, in accordance with the hypothesis, lexical words are minimally bimoraic and maximally disyllabic.

13 Feb 2001

Prof. Mark Steedman

20 Feb 2001

Susana M Cortes Pomacondor

This piece of research aims at investigating the acquisition of English sounds by Catalan speakers from Barcelona and from Majorca. Specifically, it analyses the production and perception of two English phonemes: the voiced labiodental fricative and the voiced dental fricative. The voiced labiodental fricative was studied due to the presence of this sound as a phoneme in Majorcan Catalan and as a minor phonetic detail in Central Catalan (Catalan spoken in Barcelona.) The voiced dental fricative, in turn, was chosen because it is a phoneme in English whereas it is an allophone of the voiced dental stop in intervocalic position in both Catalan dialects. The two segments were investigated in intervocalic position ( post +continuant position) to study the effects of the L1 sounds on their acquisition.

Two test were implemented in this project: a production and a perception test. The production test consisted of different tasks to check subjects' performance across different speech styles. An identification test was carried out to analyse the relation between production and perception.

27 Feb 2001

Dr. Anne Wiltshire (University of New Mexico)

Condon (1986) hypothesized that a speaker moves precisely with the articulatory changes of speech, and that these movements are detectable at the micro level using contrastive frame sound film analysis. This phenomenon has been termed synchrony. If the hypothesis is correct, then it is important to determine how synchrony is structured and to demonstrate that speech and movement are unified systems, distinguishable but not separable. If synchrony exists, then it has several implications and applications for the study of language and human behavior across many dimensions, implying a link between linguistic, perceptual, and sensory motor consciousness. This study will therefore contribute to the validity of the original claim and add information about the phenomenon itself.

06 Mar 2001

Bert Remijsen

Ma`ya and Matbat are two small Austronesian languages (approx. 4000 and 1000 speakers, respectively). They are spoken in the Raja Ampat archipelago, off the west coast of New Guinea. Both of these languages feature a lexical tone system, and in this talk I present the results of acoustic studies on these tone systems.

Matbat features five lexically contrastive tonemes. While lexical tone in se is rare in the Austronesian family, such a complex tone system has only been reported for Utsat, an Austronesian language adjacent to a Chinese language (Maddieson and Pang 1993). I claim that, like Utsat, Matbat (and also Ma`ya) have developed lexical tone under the influence of non-Austronesian tone languages.

Ma`ya features three lexically contrastive tones. I will present the results of a study into variation between the tone systems of three dialects of of this language. Interesting results include a push-chain change in tonal realization in one dialect, and an unusual boundary tone in another. Another interesting phenomenon of Ma`ya word prosody is that, next to tone, it features lexically contrastive stress (Remijsen [under review]). If time allows, I will discuss this as well.

The word-prosodic systems of Matbat and Ma`ya are introduced here.

13 Mar 2001

Dr. Corine Astésano

Traditionally, French and English have been described as representing the two prototypes of respectively syllable-timed and stress-timed languages (Pike, 1945). In addition, the French accentual system is said to be characterized by a non-lexical final accent marked by lengthening, and congruent to prosodic boundaries. These characteristics have led some authors to consider French as a 'language without accent' (Rossi, 1980) or a 'boundary language' (Vaissière, 1991; Beckman, 1992). This (traditional) view stems from the fact that most phonological theories are based on laboratory speech (mostly isolated words), emphasizing the relationship between prosody and the syntactic properties of read speech. This dichotomy between syllable-timed and stress-timed languages is not supported by more recent investigations on spontaneous, natural speech. Indeed, many authors now reject this dichotomy and show that syllabic and accentual timing tend to coexist to various degrees in all languages. Other authors also point out the tendency for the 'traditional' final accentuation to coexist in French with a (melodic) word-initial accentuation, comparable to that of English (see Astésano, 1999 for references).

We subscribe to this latter conception of French prosody, because our research indicates that laboratory speech yields an incomplete view of the characteristics of French. Though difficult to study in a controlled way, continuous speech in natural situations of communication offers us a broader view of the realizations of linguistic categories, their natural functions, and the changes in the language. We thus propose to combine both a functionalist approach, encompassing various speaking styles (read, spontaneous and journalistic speech) and a generativist approach which posits a core linguistic system: in this view, the variability observed during the process of enunciation in various situations of communication reflects alloform expressions of this core system (Astésano, 1999). We adopt a phonological model of French prosody, which, while proposing a metrical frame of analysis, takes into consideration the descriptions of natural connected speech to account for the complexity of the French accentual system (Di Cristo, 1999). The model proposes that words are marked, in the underlying phonological structure, by an initial and a final accent, whose surface realization depends on grouping and rhythmic rules, as well as those pragmatic rules which account for nuclear and emphatic accent assignment.

Our analyses have revealed the existence of a relevant phonological unit for the prosodic characterisation of French: the Complex Prosodic Word (CPW), constituted of 2 Metrical Feet (MF), the first ending in a Word-Initial accent, the second ending in a final accent. This phonological unit is a cohesive rhythmic 'figure' and a reference unit for read speech as well as spontaneous speech, insofar as its duration is less variable than that of the syllable. This tendency is particularly marked for spontaneous speech, and expresses a tendency towards stress-timing, which appears to be even more salient in French than syllable-timing. We have thus shown that initial accentuation not only belongs to the accentual system of French but that it is also central to a description of the metrical properties of this language. Moreover, our results on infra-syllabic duration indicate that the CPW is phonetically clearly defined: the (lexical) initial accents are characterized by a significant lengthening of the Onset element over the Rime, whereas the reverse is true for the final accents. This tendency was also described for English (Campbell, 1992).

These results have important implications not only for the development of phonological theories but also for their psycholinguistic consequences, particularly as regards to the issue of speech segmentation and lexical access. Indeed, taking into consideration the initial accent in the metrical structure of French allows us to suggest that stress-timing should also play a role for lexical access in French. That is, instead of assuming that every syllable has equal potential as word-beginning for the listener (Dumay et al., 1999), we suggest that the syllables bearing a word-initial accent are most likely to play this role. In fact it may be that the CPW defined by an initial and a final accent, and not the Prosodic Word delimited by a sole final accent, is the basic metrical pattern in French, thus contradicting the traditional conception of speech segmentation in French (Cutler, 1999; Banel et al., 1998). The present project aims at testing this latter hypothesis. We designed a set of experiments to further test the relevance of the CPW as an encoding unit in French (Production priming experiment) and to pinpoint its function in the decoding processes (Gating and word-monitoring paradigms). Constant reference to equivalent research in English is made in this talk, although a systematic comparison between the two languages will come as a second step in the project. Indeed, a better understanding of the French phonological system needs first to be reached in order to allow a more insightful comparison between those two 'prosodically distant' languages.

20 Mar 2001

John Rye (Dolphin Oceanic Ltd)

Text to speech (TTS) synthesis is nowadays an integral part of many PC systems whether intended as an access aid for visually impaired users or not. Sophisticated concatenative techniques can provide more natural sounding speech at an ever lower total cost. Also, some formant synthesis techniques are still useful. This paper will discuss these and some other aspects of TTS, which can be important when TTS is being used as an access aid by visually impaired computer users, for instance in the workplace. The particular subjects discussed are the environment and the nature of the texts to be spoken and the need for a quick response and fast intelligible speech. Some of the consequences for TTS system design are also discussed.

27 Mar 2001

Korin Richmond, Joe Frankel, Alan Wrench