Klaske van Sluis

The acoustic contrast between the Dutch consonants /t/ and /d/ 75 for the pre- and post-operative speech conditions. No statistically significant difference was seen for the parameters prevoicing and burst duration between pre- and post-operative condition. An increase in vowel duration following the initial consonant /t/ was seen for the post-operative speech condition ( p =.017). Table 4.2: Mean and standard deviation, minimum and maximum value and statistical testing of the acoustic values for the /t/ pre- and post-operative n =17. * p ≤ .05 tested with WMPSR test Acoustic Mean SD Min Max p -value Parameter Prevoicing Pre 1.3 3.7 0 11.1 (%) Post 1.3 5.4 0 22.2 1.00 Burst dur. Pre 37.2 10.9 25.2 64.2 (ms) Post 38 5.5 29.7 47.7 .747 Vowel dur Pre 93.9 16.6 73.6 138.1 (ms) Post 107.5 32 68.2 174 .017* Table 4.3 shows the results with acoustic features of the consonant /d/ and outcomes of statistical testing for the pre- and post-operative speech conditions. The presence of prevoicing for the consonant /d/ decreases significantly in the post-operative speech condition ( p =.003). Burst duration of /d/ increases significantly in the post-operative speech condition ( p <.001). No difference was found for the vowel duration for the vowel following the initial /d/ ( p =.120). Table 4.3: Mean and standard deviation, minimum and maximum value and statistical testing of the Acoustic Values for the /d/ pre- and post-operative n =17. * p ≤ .05 tested with WMPSR test Acoustic Mean SD Min Max p -value Parameter Prevoicing Pre 57.5 36.7 0 100 (%) Post 14.4 23.1 0 77.8 .003* Burst dur. Pre 24.5 5.9 15.9 37.4 (ms) Post 34.9 6.5 23.6 47.3 < .001* Vowel dur Pre 101.5 12.6 77.3 129.9 (ms) Post 109.1 21.2 79.8 150.1 .120 To obtain a rough estimate of the sizes of the effects of time (pre/post- treatment) and phoneme identity (t/d) on the acoustic measurements, an anal- ysis of linear mixed effect models was performed using maximal random factors (see Table 4.4, Figure 4.1) [11]. Models using random phoneme slopes did not converge and only random slopes for time were used. Interactions between phoneme identity and time were significant for prevoicing and burst duration.