The present study examines the nature of the Spanish rhotics experimentally, investigating the possibilities that the trill is best understood as deriving from an underlying geminate or that the tap and trill may be separate phonemes. In order to explore these two theories, a behavioral task was designed exploiting the absence of native words in Spanish containing a final trill onset that carry antepenultimate stress. The geminate view of Spanish rhotics argues that this pattern is absent because it is prohibited by the synchronic phonological grammar, while two-phoneme approaches consider it a diachronic gap. A nonword naming task tested the representation of this nonextant word shape by comparing trisyllabic nonce forms containing trills to other controlled sequences in theoretically banned (antepenultimate) and permitted (penultimate) stress conditions. The results of the experiment are interpreted as evidence against a geminate analysis of Spanish taps and trills. The data also suggest gradience in phonemic categorization and are considered within an emergent conception of the lexicon.