Minimal pairs are defined as pair of words in a particular language which differ in only one phonological element and have a different meaning (Roach, 2000). Several authors argued their relevance in the treatment of phonological disorders (for instance, Barlow and Gierut, 2002). In this study we investigate the nature of minimal pairs showing that a subtype of them entails a peculiar form of processing. In many languages bound morphemes used to mark inflection generate minimal pairs. In English, the present third person singular morpheme -s and the past tense morpheme -ed generate in most cases minimal pairs, such as “asks / asked”. Several authors (Stemberger and MacWhinney, 1986, Bertram et al, 2000) have argued that inflected forms may be stored in the lexicon as units, i.e. together with the bound morpheme. If inflected forms are stored as units in the lexicon, discriminating lexical minimal pairs and morphosyntactic minimal pairs should not be different processes. Elements should be stored similarly in the lexicon, and then compared phonologically when the subject is presented with a minimal pair. In this study we addressed this question presenting 20 monolingual native speakers of English with lexical and morphosyntactic minimal pairs (30 per condition, frequency differences not significant), and with pairs of identical words (leading, thus, to 120 trials). Participants were asked to press “white” if words were different and “black” if words were identical. Conditions were matched on word length. Results show that subjects are significantly faster in discriminating words generating a lexical minimal pair, such as “back / badge” than words generating a morphosyntactic minimal pair, such as “asks / asked”, t (19) = -4.486, p < .001. A third condition was also present to deepen our understanding of the processing of morphosyntactic minimal pairs. In this condition subjects were presented with morphosyntactic minimal pairs generated by very infrequent verbs. Unexpectedly, minimal pairs generated by infrequent verbs revealed to be faster recognised (19) = 2.120, p < .05 than the other morphosyntactic minimal pairs. Even if this may be interpreted as a consequence of attention arousal for unexpected stimuli, the result is problematic if we assume inflected forms to be stored in the lexicon as units. Together, these results suggest that inflected forms are not stored as units and that the discrimination of morphosyntactic minimal pairs relies on the discrimination of inflectional morphemes. As such, we suggest that increasing the sensibility to morphosyntactic minimal pairs in people with a morphosyntactic disorder, such as children with Specific Language Impairment (SLI), should improve their language performance.