Kunisaki and Fujisaki [Ann. Bull. RILP11, 85–91 (1977)] have reported that perception of fricatives from a synthetic [s]‐[∫] continuum depends on the nature of the following vowel: Listeners give more [s]responses when the vowel is rounded than when it is unrounded. After replicating their basic finding using 100‐ms fricative noises followed by either [a] or [u](with appropriate formant transitions), we have begun to investigate the limits of this contextual dependency. Our experimental manipulations include (1) extending the duration of the fricative noise, (2) introducing silent intervals of varying length between the fricative noise and the vocalic portion (so that a fricative‐stop‐vowel sequence is heard), and (3) preceding the fricative‐stop‐vowel stimuli with another vowel (thereby introducing a syllable boundary between the fricative and the stop). We have also preceded and/or followed the fricative noises with steady‐state vowels, and we have systematically varied the formant transitions of the following vowel, in order to dissociate contextual effects due to the transitions themselves from those due to the vowel. By comparing our perceptual results with measurements of natural speech, we will examine to which extent contextual dependencies in perception (and their changes in the various experimental conditions) parallel coarticulation effects in the production of analogous utterances. [Work supported by NICHD and BRSG.]