KOARTIKULACIJSKI PRITISAK I KOARTIKULACIJSKI OTPOR: ULTRAZVUČNO ISTRAŽIVANJE

Recent coarticulation models can be divided into two large groups: 1. those that see features as basic units of speech production (for example feature spreading theory, window model of coarticulation) and 2. those that have gestures (articulatory structures) as the focal point of speech production (for example articulatory phonology, degree of articulatory resistance theory). Both groups of models successfully predict some coarticulatory events, but in some cases they predict different outcomes for the same articulatory structure. The first goal of this paper was to test ultrasound as a physiological method of investigating articulation. The second goal was to test which of the two groups of coarticulatory models more successfully predicts the behaviour of tongue dorsum during a non-dorsal consonant between two high vowels (for example, tongue dorsum during /upu/). Speech material consisted of 8 VCV nonsense words, where V represented /u/, /a/, while C represented /p//t/, /n/ and /m/. Acoustic and ultrasound data were collected from two healthy Croatian speakers. Segmentation and annotation were performed by means of the Ultra-CATS software, while the synchronization of the acoustic and ultrasound signal was confirmed by Praat. The results showed that the first group of coarticulatory models did not successfully predict the coarticulation of a non-dorsal consonant between the two high vowels. What is more, the movements of the tongue dorsum proved to be different for different types of non-dorsal consonants, which the feature-based models also fail to explain. The second group of models predicted the results of this study much more successfully. The paper discusses possible implications of such results on phonetic and phonological explanations of coarticulation.