PWorkshop Archives: Autumn Term 2002

22 Oct 2002

Viktor Tron

Past tense suffixation in Hungarian has been long recognized as a rather intricate subsystem which proved a challenge to linguistic analysis (Rebrus 2000, Siptar and Torkenczy 2000). Through presenting a novel approach to describe Hungarian past tense suffixation, the talk aims to strengthen the case for a functionalist approach to systematicity in language.

If we abstract away from vowel harmony the suffix used for marking the past tense in Hungarian shows a three-way allomorphy (i.e., ~t/~tt/~Vtt when V stands for a mid vowel). The facts that these surface variants share the alveolar stop /t/ and that the alternants can be related by otherwise motivated processes such as gemination and vowel epenthesis encouraged many to assume a common (underlying) representation for the variants. Since it is beyond doubt that phonotactic constraints are at play in conditioning the allomorphy, it is reasonable to propose an analysis based on phonotactic (syllable structure) constraints. Such an account, however, turns out to be extremely hard to tune in when it comes to details. In particular, cases abound where the local phonological environment of the verb stem is of no help in predicting the attested past tense forms. These cases are usually relegated to the realm of exceptions and idiosyncrasies and are denied the possibility of being systematic in some other way.

In addition to an account of the productive rule-like patterns of the allomorphy, the analysis proposed here is trying to capture the additional systematicity in irregular items as well as to explain the motivation behind idiosyncrasies and hesitations. This is made possible by a holistic view on the problems of morphophonology, which conceives of grammar as an emergent network of richly represented individual lexical items (Bybee 2001). Instead of abstract rules and representations, relations between surface forms such as uniformity and contrast (Kenstowicz 2000) are rendered the driving force in the explanations. In particular the relations between past and present tense forms of verbs appear to have a crucial role. The analysis we present can be cast in an Optimality Theoretical framework.

29 Oct 2002

Mits Ota

This study examines whether lexical and phrasal aspects of pitch phonology are acquired simultaneously or separately. In Japanese, word accent is realized as a downfall in pitch, and its location is lexically specified. In addition, phrase-initial short syllables are lowered in pitch unless they are accented, and this creates a rise from the first to second syllable of the phrase. F0 excursion data obtained from spontaneous speech produced by Japanese-speaking children (1;3-1;9) show consistent falls at the location of the target lexical accent but variable contours on non-accented phrase-initial sequences. I argue that this is due to the different realizations of lexical and phrasal features in the input data, which make phrase-initial lowering less accessible to the learner than lexical accent. The results provide a new explanation for the previous finding that 1-year-olds frequently produce falling contours for target words with rising contours (Hallé, de Boysson-Bardies & Vihman 1991).

12 Nov 2002

Thais Christofaro-Silva (Universidade Federal de Minas Gerais)

This paper will consider three ongoing sound changes in tautosyllabic clusters in Brazilian Portuguese: the loss of the liquid, the change in the obstruent and the change in the liquid. The loss of the liquid yields a CV syllable (ou[tr]o > ou[t]o 'another one', exem[pl]o > exem[p]o 'example' (cf. CRISTOFARO-SILVA (2000)). The change in the obstruent involves the sequence of coronals [tl] when the alveolar stop is replaced by a velar stop (A[tl]etico > A[kl]etico 'Atletico club' (FREITAS (2001)). The change in the liquid occurs when a lateral is replaced by a tap ([pl]ano > [pr]ano 'plan' GOMES (1987)). Potential motivation for these sound changes will be addressed. Considering that in the vast majority of cases sound changes are said to be phonetically motivated the role of articulation will be considered (BROWMAN and GOLDSTEIN (1992); MOWREY & PAGLIUCA (1995)). Then auditory-perceptual properties will be addressed as the motivation for the sound changes under consideration (OHALA (1989, 1993)). Finally, the role of type frequency will be addressed as the motivation for the phenomena considered in this paper (BROWN (1999), BYBEE (2001), CRISTOFARO-SILVA & OLIVEIRA (2002)). It will be shown that all three cases involving ongoing sound changes in tautosyllabic clusters in BP may be understood as motivated by a lower type frequency moving towards a higher type frequency. This proposal---which also incorporates articulatory and perceptual properties---offers a more insightful analysis for the phenomena considered.

19 Nov 2002

Dan Silverman (University of Illinois & University of Edinburgh)

Pre-aspirated stops, known to be quite rare in the world's languages, are shown herein to be significantly rarer still. Their aspiration component is typically reinforced by an oral gesture that is influenced by the following stop, and/or the preceding vowel. Alternatively, the aspiration component weakens to zero, and is replaced by vowel length. In this typological study I document the phonetic details of so-called pre-aspirates, and offer phonetic explanations for their rarity and their diachronic instability.

26 Nov 2002

Rob Clark

This paper addresses the problem of generating a full range of appropriate intonation contours particularly for concept-to-speech systems where there is a specific requirement to produce different meanings and contrasts through intonation.

03 Dec 2002

Simone Ashby

It has been suggested that infant-directed speech (IDS) may serve as a set of early language instructions, and that prosodic bootstrapping -- in the form of exaggerated pitch contours, heightened pitch range, melodic repetitiveness, et al. -- may enable the infant to begin to segment the speech stream, thereby gradually acquiring and making sense of discrete lexical items.

Assuming prosodic bootstrapping is a real phenomenon and prosodic modification in IDS facilitates an initial segmentation of the continuous audio stream, those prosodic cues which identify particularly IDS-like speech should also flag parts of the stream to which the infant should pay particular attention.

The question is how then do we go about characterising IDS in order to accurately investigate those prosodic cues? Can such properties be quantified? Or is quantification confounded by other important variables, such as voice quality and affective intent? Further, how might we distinguish IDS from different varieties of hyperarticulated speech, whereby the speaker is similarly constrained by output-orientated goals for effective communication?

The aims of this research are twofold: (1) to reassess the prosodic identity of English IDS both quantitatively, and with regard to other carefully articulated speech forms; and (2) to contribute to the understanding of hyperarticulated speech by comparing IDS, where it is assumed the addressee has scant familiarity with the ambient language, against a backdrop of other uniquely motivated hyperarticulated speech registers, including speech to foreigners, speech across a noisy communication channel, and human-machine interaction.

Research background and preliminary findings will be presented.

10 Dec 2002

Wing Chiu Ivan Yuen

It has been generally accepted in the literature that pitch is relative. However, there is no explicit characterisation of what pitch is relative to. The talk will examine what a tone is relative to, when its acoustic correlate F0 is contiuously varied over time as a result of declination.