Models of the development of speech perception often include the notion that infants learn to automatically focus attention on language-specific relevant cues in speech (e.g., Jusczyk, 1997). Studies using behavioral methods have revealed that infants stop discriminating certain non-native speech contrasts between 6 and 12 months of age. However, behavioral studies with adults show that sensitivity to the cues needed to discriminate these non-native contrasts is not lost (e.g., Logan, et al., 1989). Specifically, if their attention is directed to the relevant cues, they can make use of them. These findings lead to an interesting hypothesis concerning the nature of the speech perception deficits observed in children with specific language impairment (SLI). It is possible that focusing attention on the relevant cues in the native language is not automatic for SLI children. This possibility is supported in a study from our lab in which children with SLI failed to show mismatch negativities to the vowel contrast in “bed” versus “bid”, even though they showed good behavioral discrimination of these vowels. The finding that MMN is reduced or absent to subtle phonetic differences that are not phonemic in a speaker’s language in a number of studies supports the claim that focusing on relevant cues in a native language has become automatic for speakers with typical language. I will discuss how electrophysiological investigations of infant speech perception could help in determining the role of attention in the development of speech perception and in language-learning disorders.