Klaske van Sluis

76 4.4. Discussion Table 4.4: Linear mixed effect models analysis Y ∼ phoneme*time + (1 + time|speaker) + (1 + time|word). p determined assuming t follows Z-test (Nor- mal) statistics. † : t = 2.43, p = 0.015 Y: Prevoicing (%) Burst dur. (ms) Vowel dur. (ms) Intercept 1.31 37.21 93.23 Phoneme 56.21 -12.66 8.3 Time 0.77 0.99 12.79 † Phoneme:time -44.87 9.44 -4.91 t(Phoneme:time) -8.24 3.28 -1.07 p 1.7 10-16 0.001 0.28 Figure 4.1: Visualization of the linear mixed effect models based on the estimates of Table 4.4 4.4 Discussion Several studies have compared differences between healthy laryngeal speech and speech after total laryngectomy [1]. After total laryngectomy the pharyngo- esophageal segment functions as the new voice source. The resulting tracheo- esophageal speech is has a lower F0 and high noise components [1, 3]. In this study, post-treatment values of F0, %V, HNR, and MVD worsened compared to the pre-operative values. The voices of the participants included in this research are deviant from healthy laryngeal voices. In the pre-operative situa- tion, the voices are distorted due to either tumor presence on the vocal cords, earlier treatment with (chemo) radiotherapy and/or presence of a trachea can- nula. These acoustic changes in the vowel /a/ suggests an effect on acoustic