Individuals vary their speaking rate, and listeners use the speaking rate of precursor sentences to adjust for these changes [Kidd, G. R., ‘‘Articulatory-rate context effects in phenone identification,’’ J. Exp. Biol.: Human Percep. Perfor. 15, 736–748 (1989)]. This speaking rate adjustment is not limited to speech from a single talker [Sawusch & Newman, ‘‘Perceptual normalization for speaking rate II: Effects of signal-discontinuities,’’ Percep. Psychophys. 62, 285–300 (2002)]; if one speaker stops talking and another begins, listeners will use rate information from the first talker to interpret speech in the second. We examined how long this cross-talker effect lasts. A male talker produced a carrier sentence at fast and slow speaking rates. A female talker then produced one or more subsequent words at a constant speaking rate. The final word in the female voice contained a duration-based contrast. We examined how much intervening speech in the female voice was necessary before the male voice’s speaking rate no longer influenced perception of the final target word. Results suggest that listeners require approximately 1 s of intervening speech or of silence before they stop using rate information from the male talker. This suggests that rate normalization may be based on a 1-s temporal window, with any speech information that occurs within 1 s in advance of a target phoneme influencing perception of that phoneme.