This is an extension of the investigator's previous work [J. Acoust. Soc. Am. Suppl. 171,S105; and72,S16 (1982)]. A series of spoken /sɪsi,sisɑ/ tokens was modified using a PCM speech editor. The /s/‐noise segments were excised. The original intervocalic /ɪ‐i,ɪ‐ɑ/ contexts and the /ɪ,i,ɑ/ vowels in isolation were assembled into a perceptual identification test and low‐pass filtered at 3 kHz with an attenuation slope of 48 dB/oct. As a consequence, the noise and transient energy coextensive in time but higher in frequency than the vocalic formant structures was removed more completely than before. Subjects identified the consonants heard adjacent to the vocalic segments. There was better agreement as to thelabialstop frequently heard in the intervocalic /ɪ‐i/ context than previously. As before,dental‐alveolarstops were frequency heard in /ɪ‐ɑ/. These results reflect perceptual integration over the arresting and releasing transition and silent interval duration cues. [Haskins Laboratories provided valuable assistance in the preparation of stimuli on NIGHD Contract No. NIH‐71‐2420.]