Vowel recognition threshold as a function of temporal segmentation