The perceptual magnet effect has been cited as evidence for the existence of phonetic prototypes in speech perception. According to Kuhl (1991), discrimination accuracy is lower among more prototypical instances of a phonetic category than among less prototypical instances. This paper reports four experiments examining the magnet effect in the vowel /i/. Experiment 1 showed that subjects rated some examples of /i/ as more prototypical than others. A vowel with an extreme F2 value was selected as the best instance in the test set. In Experiment 2, a perceptual magnet effect was not obtained with materials based on Kuhl's tokens of /i/ or with items normed for each subject. Furthermore, subjects' ratings of the best h1/ were not stable across changes in acoustic context. In Experiment 3, a modified 4IAX procedure was used to assess the perceptual magnet effect with greater sensitivity. Discrimination accuracy in the prototype and nonprototype conditions did not differ significantly, contrary to prediction. In Experiment 4, subjects categorized the vowels developed from Kuhl's test set as either "/i/" or "not /i/." Many vowels in the nonprototype /i/ condition were labeled as "not /i/." This finding suggests that the perceptual comparisons required in Kuhl's original study spanned different phonetic categories. Therefore, claims regarding the perceptual magnet effect as evidence of prototypes in speech perception must be taken with strong reservations. Taken together, the present results suggest that the qualitative notion of a phonetic prototype may not have great explanatory power in describing human speech perception because the internal structure of phonetic categories varies with changes in context, prototypicality does not influence the discriminability of category members, and the nature of the memory representation for phonetic prototypes is poorly defined.