When dichotic nonsense syllables are presented to temporal lobectomy patients at equal intensities, the ear contralateral to the site of the lesion performs more poorly than the ipsilateral ear; if the ipsilateral ear is stimulated below threshold, the contralateral ear performs near 100%. However, if the intensity ofspeechin the ipsilateral ear is increased above SRT, the contralateral scores drop markedly as intelligibility increases in the ipsilateral ear. This “trade off” does not occur when noise is the competing stimulus. This phenomenon is interpreted as a sign of “speech identification” and is suggested as a potential technique for differentiating speech from nonspeech elements.