Each of two experimenters taught one set of 32 prekindergarteners to discriminate four different letter combinations (R-P, Y-V, C-G, and K-X). Each set of children was randomly selected and assigned to two treatment conditions. The treatment consisted of three warm-up trials, 10 actual training trials, and four post-test trials on a match-to-sample task. The errorless discrimination training group (EDT) was not given any feedback for each trial, while the reinforcement-extinction group (RE) was told after each trial if his responses were correct or incorrect. Error numbers during training trials and criterion trials were separately analyzed. The mean number of errors was 4.06 for the EDT group and 16.89 for the RE group. None of the other main effects or interactions were significant. The mean error during the post-test trials was 2.95 for the EDT group and 6.53 for the RE group. The sequence main effect and the interactions were insignificant. However, there was a main effect for experimenter; the author felt that the difference could be accounted for by the increased number of errors made by the RE group in Set II. Table and reference are included. (AW)