Fig. 2.

Experiment showing the sensitivity of speech categorization to nonspeech context (Holt, 2005). The schematic shown in (A) illustrates the stimuli used: 22 tones varying in frequency formed a sequence of sounds that preceded a syllable drawn from a series that varied perceptually from ga to da. The tones had either a high-frequency distribution or a low-frequency distribution, as shown by two samples plotted on time-by-frequency axes in (B). When listeners identified the speech syllables in these tone-sequence contexts, the distribution of tone frequencies had an effect on speech perception (C). Listeners more often identified the consonants as ga when high-frequency sequences of tones preceded the syllables. The same syllables were more often reported to be da when low-frequency sequences preceded them.