KINEMATIC AND ACOUSTIC CORRELATES OF QUANTITY IN SWEDISH AND WOLOF : A CROSS-LANGVAGE STUDY

The main question addressed here is: does syllable structure affect the robustness of vowel quantity kinematic and acoustic phasing patterns? This investigation is an attetnpt to answer this question by examining data from two unre1ated languages, and also to unveil regularities related to the phonotactics of the languages studied. INTRODUCTION Some languages use quantity for lexical and grammatical distinctions. This investigation focusses on the use of vowel quantity for lexical distinctions in two unrelated languages, Swedish and Gambian W olof. Swedish has a centrast between two lengths, and this centrast occurs only in lexically stressed syllables, where the vowel is either short or long. If the vowel is long, it is either word final or followed by a short consonant, VVC; if the vowel is short, it is followed by a long consonant, VCC (Eiert, 1964; Engstrand & Krull, 1994). This linguistic specificity has been reffered to as the complementary distribution of the quantities (i.e. shortlong vs. long-short). Wolof also has a centrast between two lengths and the centrast occurs only in the first syllable which is the lexically stressed syllable. Here also the vowel is either short or long. However, unlike Swedish, Wolof does not show a complementary distribution of the quantities, as both short and long vowels are followed by a single consonant, i.e. VC vs. vvc. The main question addressed here is: does syllable structure affect the robustness of vowel quantity linguistic distinctions? Thus speech rate is varied to test for this robustness, and the ernerging acoustic and kinematic phasing patterns will be analyzed and discussed in terms of linguistic constraints. METHOD One male speaker of Wolof and one male speaker of Swedish uttered a series of CVCV words that were chosen to vary vowel quantity and quality in the two languages. Each token was embedded in a carrier sentence: The speakers were instructed to produce the randomized Iist of utterances at a normal conversational rate, and at a selfselected fast rate, at least fourteen times in each speech condition. There were 5 conditions in all: 2 speech rates (conversational or normal and fast) X 2 vowellengths (short and long) X 3 vowel types (i, a u) X 2 voicing contrasts (voiceless and voiced) X 3 places of articulation (p, t, k). The movement data were recorded using, for W olof, a three-coil transmitter system, with five transducers, the Electro Magnetic Articulograph (AG 100, Carstens Medizinelektronik), and for Swedish, another three-coil transmitter system, the Electro Magnetic Midsagittal Articulometer (EMMA Haskins), as described by Perkeil et al. (1992).