This experimental study investigated three components of the lexical-acquisition process: fast mapping, word learning, and word extension. Thirty preschool-age children with specific language impairment (SLI) and 30 age- and gender-matched normal language (NL) controls participated. Two types of low-frequency words were used to name objects: phonologically simple and phonologically complex. Two types of objects were used: semantically familiar and semantically unfamiliar. Comprehension and production were probed across components, with trials to criterion calculated for word learning. During the word-extension task, recognition was also assessed. In addition, norm-referenced receptive and expressive vocabulary tests were administered. On tasks where group differences were found, many children with SLI performed as well as controls. The fast-mapping task revealed no group differences (p On this and the word-learning task, comprehension exceeded production for both groups. On the word-learning task, the SLI group comprehended and produced fewer words than did controls. In contrast to the NL group, the SLI group produced more words for semantically unfamiliar than familiar objects. Overall, the SLI learned more phonologically simple than complex words, but the NL group showed no difference. Although the SLI group required more trials than controls to comprehend words, no group differences were found in trials to production. On the word-extension task, the SLI group scored lower than controls only for production. Vocabulary-test scores did not accurately identify children with SLI or predict number of words learned; but these scores did predict a small amount of variance for fast-mapping and word-extension performance. Fast mapping performance accounted for 25% of word-learning variance; fast-mapping performance and word-learning performance combined accounted for 54% of word-extension variance. Different predictor variables were found for each language group. Findings suggest that some, but not all, children with SLI demonstrate poor word-learning and word-extension performance. Overall, the SLI group had greater difficulty than controls on production measures and required more trials to achieve learning criterion for comprehension of words than the NL group. The semantic familiarity of the target objects affected productive word learning by the NL group but did not appear to have a similar effect on the SLI group.