An alternating treatments design was used to measure the differential effects of two error-correction procedures (word supply and word analysis) and a no-training control condition on the number of oral-reading errors made by four moderately mentally retarded children. Results showed that when compared to the no-training control condition, both error-correction procedures greatly reduced the number of oral-reading errors of all subjects. The word-analysis method, however, was significantly more effective than was word supply. In terms of collateral behavior, the number of self-corrections of errors increased under both intervention conditions when compared to the baseline and no-training control conditions. For 2 subjects there was no difference in the rate of self-corrections under word analysis and word supply but for the other 2, a greater rate was achieved under word analysis.