This investigation compared the effectiveness of two language treatment methods, modeling versus modeling plus evoked production, in promoting productive vocabulary in three toddlers identified as late talkers. A single-subject alternating treatments design was employed in this study in which different sets of words were taught under the two treatment methods during group and individual instruction. Some evidence of differential patterns of response to the treatment types was found for two subjects, but the subjects differed as to which particular treatment method was associated with better performance. Neither treatment method was effective for the third subject. Dynamic assessment measures were only marginally useful in predicting overall lexical learning potential in these subjects, and did not predict response to the two teaching methods. Implications of these results are discussed with regard to the role of language intervention for late talkers.