Background Deficits in procedural memory have been proposed to account for the language deficits in specific language impairment (SLI). A key aspect of the procedural deficit hypothesis (PDH) account of SLI is that declarative memory is intact and functions as a compensatory mechanism in the acquisition of language in individuals with SLI. The current study examined the neural correlates of lexical-phonological and lexical-semantic processing with respect to these predictions in a group of adolescents with SLI with procedural memory impairment and a group of chronologically age-matched (CA) normal controls. Methods Participants completed tasks designed to measure procedural and declarative memory and two ERP tasks designed to assess lexical-semantic and lexical-phonological processing in the auditory modality. Procedural memory was assessed using a statistical learning task. Lexical-semantic processing was assessed using a sentence judgment task modulating semantic congruency and lexical-phonological processing was assessed using a word/nonword decision task modulating word frequency. Behavioral performance on the tasks, mean amplitude of the cortical response, and animated topographs were examined. Results Performance on the statistical word-learning task was at chance for the adolescents with SLI, whereas declarative memory was no different from the CA controls. Behavioral accuracy on the lexical-semantic task was the same for the adolescents with SLI and CA controls but accuracy on the lexical-phonological task was significantly poorer for the adolescents with SLI as compared to the CA controls. An N400 component was elicited in response to semantic congruency on the lexical-semantic task for both groups but differences were noted in both the location and time course of the cortical response for the SLI and CA groups. An N400 component was elicited by word frequency on the lexical-phonological task for the CA controls not for the adolescents with SLI. In contrast, post hoc analysis revealed a cortical response based on imageability for the adolescents with SLI, but not CA controls. Statistical word learning was significantly correlated with speed of processing on the lexical decision task for the CA controls but not for the adolescents with SLI. In contrast, statistical word learning ability was not correlated with the modulation of the N400 on either task for either group. Conclusion The behavioral data suggests intact semantic conceptual knowledge, but impaired lexical phonological processing for the adolescents with SLI, consistent with the PDH. The pattern of cortical activation in response to semantic congruency and word frequency suggests, however, that the processing of lexical-semantic and lexical-phonological information by adolescents with a history of SLI may be supported by both overlapping and nonoverlapping neural generators to those of CA controls, and a greater reliance on declarative memory strategies. Taken together, the findings from this study suggest that the underlying representations of words in the lexicons of adolescents with a history of SLI may differ qualitatively from those of their typical peers, but these differences may only be evident when behavioral data and neural cortical patterns of activation are examined together.