Klaske van Sluis

The acoustic contrast between the Dutch consonants /t/ and /d/ 77 features of consonants as we studied specific in the /t/ and /d/ tokens. This study has shown that the acoustic contrast of the /t/ and /d/ in word ini- tial position reduces after total laryngectomy (Tables 4.2 and 4.3). Prevoicing and burst duration values for /t/ do not change significantly after surgery. For the /d/, presence of prevoicing decreases and Burst duration increases post- operatively. Values for /d/ become more like /t/ in the post-operative speech condition. Vowel duration following the initial consonant /t/ was higher in the post-operative condition compared to pre-operative condition (Tables 4.2 and 4.4). The duration of the vowel following the initial consonant /d/ also increased but this increase was not significant. An explanation might be an overall slowing of the speech rate. With help of linear mixed effect models, the combined effects of phoneme identity and treatment were analyzed (Ta- ble 4.4, Figure 4.1). There was a strong interaction of phoneme identity and time (pre- / post-treatment), that confirmed the conclusion that the effect of treatment was limited to the initial /d/. And thus, that the difference between /t/ and /d/ became smaller after treatment. For tracheo-esophageal speakers, the reduced acoustic contrast between /t/ and /d/ in initial position could lead to intelligibility issues. In Dutch /t/ and /d/ in phonology are marked as phonemes. Earlier research on tracheo-esophageal speech has shown that an intended /d/ was more often misheard as /t/ than that an intended /t/ was misheard as /d/ ([6], Table 4.3 confusion matrix). Our study provides evidence that prevoicing and burst duration changes for /d/ might explain at least part of these intelligibility issues. The current study has some strengths and limitations. An advantage of the current research is that our analysis is performed on running speech. Recording running speech leads to most natural speaking conditions. Another strength of the study is the pre- and post-treatment within subject design. To our knowl- edge, there have not been studies that compared pre- and post-operative voice characteristics within the same individual. In this approach, changes in voice and speech can be spotted at an individual level. Limitations of the current study include the use of different microphones and that the used text was not phonetically balanced. Only /t/ and /d/ in initial position were frequently present in the text which left out other phonemes with voicing distinction for analysis. Therefore, our data did not contain enough tokens containing other plosives (/b/ and /p/; /g/ and /k/) to investigate the effect of treatment on other places of articulation. The aerodynamics of voicing and the size of the air cavity before the constriction, suggest that these effects might be different for /g/ and /k/, for /b/ and /p/ than for /t/ and /d/. 4.5 Conclusions The acoustic features of initial consonants /t/ and /d/ do move closer together, with /d/ becoming more like /t/ in tracheo-esophageal speech. This could ex- plain results from earlier research that showed asymmetric confusion between