In the process of language learning, individuals must acquire different types of linguistic knowledge, such as the sounds of the language (phonemes), how these may be combined to form words (phonotactics), and morphological rules. Early and late bilinguals tend to perform like natives on second language phonological tasks that involve pre-lexical processes (e.g., categorization or identification of isolate phonemes) but their performance decreases when the tasks tap into lexical processes (e.g., accessing to the lexicon storage and selecting the appropriate words). This graded performance across phonological processes is consistent with the processing hierarchy proposed by different models of (native) speech perception. The relation between the first and second languages is further evidenced by the correlation between native and non-native phoneme discrimination of both early and late bilinguals. This correlation provides evidence that individual differences in the command of non-native sounds are caused by a language-specific capability. We propose that the study of second language processes combined with the assessment of individual differences is a relevant field for the understanding of the speech perception architecture. (Contains 2 tables.)