This paper describes an investigation of the effect of sound duration on the perception of a linguistic characteristic, the recognition of phonemes. A pair of single syllable words was selected which consisted of identical phoneme sequences, except for the final phoneme which was either the voiced or unvoiced version of the same articulation. Different versions of these words were synthesized with all possible combinations of four different vowel durations and five different final consonant durations. In all cases the sound wave corresponding to the final consonant was inharmonic and identical except for duration. These synthesized words were recorded and presented to a group of 33 listeners who had to judge whether they heard the final consonant as a voiced or as an unvoiced sound. The results clearly show that the relative durations of vowel and final consonant can be used as a cue for hearing the final sound as voiced or unvoiced. This demonstrates that, in certain circumstances, nonspectral characteristics such as duration can serve as the basis for phoneme recognition. It also gives an example of a cue for sound recognition which, due to its relative nature, can be perceived only when it appears in a sequence of phonemes and not when the sound is pronounced in isolation.