A Mismatch Negativity (MMN) study was performed to investigate whether pre-attentive vowel perception is influenced by phonological status. We compared the MMN response to the acoustic distinction between the allophonic variation [ε–e] and phonemic contrast [e–i] present in a Southern-Italian variety (Tricase dialect). Clear MMNs were elicited for both the phonemic and allophonic conditions. Interestingly, a shorter latency was observed for the phonemic pair, but no significant amplitude difference was observed between the two conditions. Together, these results suggest that for isolated vowels, the phonological status of a vowel category is reflected in the latency of the MMN peak. The earlier latency of the phonemic condition argues for an easier parsing and encoding of phonemic contrasts in memory representations. Thus, neural computations mapping auditory inputs into higher perceptual representations seem ‘sensitive’ to the contrastive/non-contrastive status of the sounds as determined by the listeners’ knowledge of the own phonological system.