Errorless discrimination training is a technique in which the discriminative stimulus is supplemented with a salient cue which is gradually removed or faded during the course of training. In this study errorless discrimination training was used to teach preschool children the distinctive features of letters of the alphabet that are difficult to discriminate. To test the hypothesis that the success of EDT depends on whether or not stimulus control is transferred from the obvious cue used during training to the relevant dimension of the discriminative stimulus, two EDT groups were used. For group EDT-1, the obvious cue was superimposed over the feature of the letter differentiating it from its paired comparison. For group EDT-II, the obvious cue did not specifically enhance the distinctiveness of the relevant dimension. A third group, R-E, was taught by the traditional reinforcement-extinction approach. Ss were 108 four- and five-year-olds in pre-kindergarten classes. Six letter combinations, R-P, Y-V, G-C, Q-O, M-N, and K-X, were used in a match-to-sample format, with 10 training trials for each letter combination. Five posttest trials and five delayed posttest trials one week after training were given. The EDT-I group made fewer errors on the two posttests than either of the other groups. The EDT-I group also made significantly fewer errors during training than the R-E group. (KM)