This study addresses voice disorder assessment. After several years of research about perceptual an- alysis of dysphonia, our work is focused on the phonetic influence responsible to pathological events emergence. The phonetic labeling method allows a perceptual evaluation of each phoneme separately, but needs to be automated. This work proposes an original methodology involving an automatic clas- sification system as well as knowledge of both pathological and machine learning experts. This meth- odology aims to bring a better understanding of phonetic phenomena related to dysphonia. Firstly, the automatic system was validated on a dysphonic corpus (80 female voices), rated following the GRBAS perceptual scale by an expert jury. Then, an automatic phonemic analysis underlined the significance of consonants and, more surprisingly, of voiceless consonants for the same classification task. Submitted to pathological experts, it appears that the onset of the vocal fold vibration could explain these findings. These observations led to a manual analysis of voiceless plosives, which highlighted a lengthening of VOT according to the dysphonia severity, validated by a preliminary statistical analysis.