Identification ratings were taken from English monolingual subjects on two different acoustic continua: a VOT (voice onset time) speech continuum and a TOT (tone onset time) nonspeech continuum. Both continua were made by varying the temporal asynchrony between two component acoustic events. An improved signal detection model was employed for fitting the results. The model could only fit the nonspeech data. Further analyses suggested that the differences between speech and nonspeech processing in humans could lie on the form of the signal densities they evoke. Nonspeech signal densities seem to be gaussian whereas speech signal densities do not.