Recent research has shown that cognate word processing is modulated by variables such as degree of orthographic and phonological overlap of cognate words and task requirements in such a way that the typical preferential processing observed in the literature for cognate words relative to non-cognate words can be annulled or even reversed (Comesaña et al., 2012; Dijkstra, Miwa, Brummelhuis, Sappelli, & Baayen, 2010). These findings beg the question about the precise representation and processing of identical cognates (e.g., "plata-plata," silver in Spanish and Catalan, respectively) and non-identical cognates (e.g., "braç-brazo" [arm]). The aim of the present study was to further explore this issue by manipulating for the 1st time cross-linguistic similarities of identical and non-identical cognate words as well as stimuli list composition. Proficient balanced Catalan-Spanish bilinguals performed a lexical decision task in Spanish. In Experiment 1 identical and non-identical cognates along with non-cognates made up the experimental list, whereas in Experiment 2 identical cognates were excluded from the list. Results showed modulations in cognate processing as a function of their degree of orthographic and phonological overlap. These results confirm prior findings regarding the processing of cognates when cross-linguistic similarities are taken into account. Most important, the direction of the cognate effect was affected by the stimuli list composition (i.e., the preferential processing for cognate words was restricted to the list containing identical cognates). Results have important implications for the Bilingual Interactive Activation Plus model (BIA+; Dijkstra & van Heuven, 2002), especially regarding identical and non-identical cognate word representation.