An investigation into the relationship between electropalatography and ultrasound using the vowels /a/ and /i/ in typical adult speakers

Abstract

This study investigated the relationship between electropalatography (EPG) and ultrasound through comparison of quantitative measures derived from these two techniques. The measures used were two established EPG measures: WT and COG, and emergent ultrasound measures: DEI, TCPI and LOCa-i. A pre-existing corpus of simultaneous EPG and ultrasound recordings was used in which four adult speakers of Scottish English produced /CV#C/ sequences with the vowels /a, i/ and the consonants /p, f, t, s, k/. The quantitative measurements were taken at the acoustic mid-vowel point using the corresponding ultrasound and EPG images. Within technique, comparisons were made to establish the extent to which each measure was able to quantify the large gestural differences between the two vowels. Correlations were run between WT and DEI, and COG with TCPI and LOCa-i to allow comparisons of the extent to which the coupled measures quantify a similar aspect of tongue behaviour. Additionally, consonant-on-vowel coarticulatory effects were analysed using the measures WT and DEI. All five measures quantified aspects of tongue shape and consequential tongue-palate contact as different for the two vowels, with /i/ having larger values than /a/ on all measures. For /i/, there was a significant positive correlation between WT and DEI, and a significant negative correlation between COG and TCPI/LOCa-i. When individual speaker results are considered the relationship is less clear. EPG identified consonant on-vowel coarticulatory effects on /i/, whereas ultrasound identified consonant-on-vowel coarticulatory effects on /a/. The two techniques provide comparable information in their ability to quantify the large differences required in articulatory gesture for the vowels /i/ and /a/. When finer articulatory detail is concerned, the measures provide complementary information. The correlations suggest that the measures quantify slightly different aspects of movement. The results from the analysis on coarticulation show how the ultrasound can identify coarticulation in the two vowels when EPG does not, and vice versa. Adding information on tongue shape, information that EPG does not provide, these ultrasound measures have potential for use in both research and clinical settings, particularly in sounds where palatal contact is minimal, such as /a/.