| dc.description.abstract |
In the recent past, Ultrasound imaging has been used extensively as a tool to measure coarticulation. The extent of coarticulation is the magnitude of distance between tongues contours of adjacent phonemes that varies based on the articulatory dynamics of each phoneme. This lack of invariance is language-specific which is related to the articulatory constraint properties of a particular language. In this context, the present study aimed to analyse the magnitude of distance between vowel and consonant in a VCV syllable across two languages. Sixty adult speakers, including an equal number of native speakers of Malayalam and Hindi, participated in the study. The stimuli consisted of VCV sequences with C corresponding to voiced/unvoiced counterparts of dental (/t?/, /d?/) or retroflex (/?/, /?/) or velar stop consonants (/k/, /g/), in the context of vowels /a, i, u/. Tongue contours of each vowel and consonant (V to C) were obtained using Mindray 6600 Ultrasound module and the distance between tongue contours was calculated using Articulated Assistance Advanced (AAA) based on Root Mean Square (RMS) procedure. Extent of coarticulation (EC) is the amount of influence of one phoneme on a neighbouring phoneme which is depicted as the magnitude of distance between the tongue contours. To find the EC of one token, tongue contour of each repetition of each phoneme was plotted. On an average, tongue contours representing 10 repetitions were obtained for each phoneme in the workspace to minimize the variation. Averaged C spline and V1/V2 spline were considered as an analysis pair. These pairs of mean and standard deviation splines were further evaluated using the function “Diff”. This function “Diff” works based on 2 tailed t-tests using the Welch-Satterthwaite equation which is inbuilt in the AAA software. This helps to compare two mean splines and provides Root Mean Square (RMS) distance. The resulting RMS distance values were considered as EC since it is the distance between the analysis pair. Results indicated that across languages, the extent of coarticulation was significantly different for dentals, velars, and retroflexes. There was a significant difference in coarticulation across languages in the preceding vowel context. The magnitude of coarticulation was high for both voiced and unvoiced counterparts of retroflexes and velars in Malayalam than Hindi in the entire three vowel contexts. But there was a limited difference between languages in the following vowel context. A significant difference was noticed for a-t? and a-? with high mean values for Hindi compared to Malayalam. Though the phonological system was different across languages, the disparity of the pattern of coarticulation was trivial. This similarity can be accounted for as exposure of Hindi on Malayalam speakers during their schooling. But this exposure is limited to academic level, and not as a communication mode which reduces the similarity across languages. This study highlights the role of both the dynamic properties of articulators in coarticulation and the weight of language-specific characteristics in speech production. |
|