To investigate task performance and the use of private speech and to examine the effects of a cognitive training approach, 30 learning disabled (LD) and 30 nonLD Ss (7 to 8 years old) were given a 17 piece wooden puzzle rigged so that it could not be completed correctly. Six variables were measured: (1) proportion of private speech that was task relevant; (2) proportion of private speech that was task irrelevant; (3) rate of private speech; (4) number of strategies used; (5) time on task before reaching the rigged piece; and (6) persistence time. Three significant differences were found between LD and nonLD Ss: LD Ss had a significantly lower proportion of task relevant utterances; time on task before reaching the rigged piece was significantly longer for LD Ss; and normally achieving Ss had significantly longer persistence times. Ss in the cognitive training group (who had watched a videotape of a young boy completing the task and learned to think out loud) had significantly higher proportion of private speech and task relevant private speech, used significantly more strategies, and had significantly longer persistence times. (Author/CL)