In two experiments on normals we presented CV nonsense syllables both dichotically and monotically, with onsets of the syllables separated by 0, 15, 30, 60, and 90 msec (first experiment) and 0, 90, 180, 250, and 500 msec (second experiment). We found that when one of the CV's trailed the other by 30–60 msec, the trailing CV became more intelligible than when it was given simultaneously; the leading syllable's intelligibility dropped from its “simultaneous” level when leading by 15 and 30 msec. The leading message was more intelligible between 15 and 250 msec when the two channels were mixed monotically. In the dichotic simultaneous conditon, voiceless consonants were more intelligible than voiced, especially in voiced‐voiceless pairs. When the voiced CV trailed the voiceless CV, the former became almost as intelligible as its voiceless counterpart. A left hemisphere “speech processor” was postulated, with suppression of information from ipsilateral sources during contralateral stimulation. The postulated “speech processor” may be involved in acoustic‐signal‐vocal‐tract control functions.