What effects do within‐ and between‐category variations have on the perception of speech? Using audio‐morphing and the Straight channel vocoder (Kawahara, 2004), we produced 320 high‐quality phonetic continua varying in place, manner and voicing including word/word (blade/glade), word/pseudo (blouse/glouse), pseudo/word (bown/gown) and pseudo/pseudo (blem/glem) pairs. A 2AFC task confirmed the category boundary shift for word/pseudo and pseudo/word pairs (Ganong, 1980), equivalent for onset (bench/gench) and offset (flad/flag) pairs. This suggests that lexical influences on categorical perception are not produced on‐line but rather occur post‐perceptually, consistent with top‐down effects. Sensitivity to within‐ and between‐category phonological variation was investigated using sparse fMRI in a paired auditory repetition priming paradigm. Minimal pairs (48 across the 4 stimulus groups) were presented to participants who listened in the context of a semantic monitoring task. Between‐category pairs with a phonological change produced a greater neural response compared to within‐category same pairs with the same magnitude of acoustic difference. This response to phonologically different pairs provides a neural correlate of categorical perception in left middle temporal, inferior frontal and pre‐central regions. These responses in inferior frontal regions may contribute towards top‐down influences on categorical perception of speech (cf Ganong Effect).