Previous research in the authors' laboratory has shown that bilingual exposure has specific consequences on the phonological detail represented in early words (Ramon-Casas, Swingley, Sebastián-Gallés, and Bosch, 2009). When tested in a word recognition task involving a Catalan-specific vowel mispronunciation, Catalan-Spanish simultaneous bilingual toddlers (18- to 24-month-olds) differed from Catalan monolinguals in their ability to detect the wrong pronunciation in four target words. One possible factor related to bilinguals' failure to react to the vowel mispronunciation could be the cognate status of the items in the test material. It has been shown that cognate and non-cognate words have a different status both in language production and perception (Kroll and Stewart, 1994). Cognate words, differing in vowel quality when produced in one or the other language in bilingual environments, are more subject to variability, and this factor may affect the degree of specificity in their vowel representation. As a consequence, bilingual toddlers might find certain mispronunciations to be more acceptable in cognates than in non-cognate words. In order to clarify this issue, a familiar word recognition experiment was run in Catalan, using four non-cognate targets and a similar methodology to the previous study with cognate words. Two groups of 18- to 24-month-old infants participated in this experiment, from Catalan monolingual and Catalan-Spanish bilingual homes, respectively. Results indicated that both groups successfully detected the vowel mispronunciation in these non-cognate items, thus suggesting that cognate status does affect the detail encoded in the representation of words in bilinguals' early lexicons. It is suggested that phonological form differences between non-cognate words may facilitate the building of more stable representations and preserve them from the effects of input variability.