Spanish–English bilingual lexical organization was investigated using masked cognate and non-cognate priming with the lexical decision task. In Experiment 1, three groups of bilinguals (Spanish dominant, English dominant and Balanced) and a single group of beginning bilinguals (Spanish) were tested with Spanish and English targets primed by cognate and non-cognate translations. All the bilingual groups showed cognate but not non-cognate priming. This cognate priming effect was similar in magnitude to the within-language repetition priming effect; it did not vary across participants who had different second-language acquisition histories, nor was the size of the priming effect modulated by the direction of the translation. The beginning bilingual group only showed cognate priming when the primes were in Spanish (L1) and the targets in English (L2). In Experiment 2, both form-related and unrelated word baselines were used with a single group of bilinguals. The results were the same as Experiment 1: cognate priming and no non-cognate priming. Experiment 3 examined the cognate priming effect with reduced orthographic and phonological overlap. Despite this reduced form overlap, it was found that the cognate effect was the same size as the within-language repetition effect. These results indicate that cognate translations are special and ways of modifying models of bilingual lexical processing to reflect this were considered.