Abstract
Within Gibson's framework of teaching children to discriminate the distinctive features of letters, 28 preschool children were trained in letter discrimination using three different types of matching-to-sample alternatives: (1) high-confusion alternatives; (2) low-confusion alternatives; or (3) a sequence of low-, middle-, and high-confusion alternatives. On posttest 1, given after a criterion of two consecutive errorless training days, only the high-confusion and sequence groups significantly improved over their pretest scores as compared with the no-treatment control group. On posttest 2, however, given after a standard number of 20 training days, the performance of all three training groups, including low-confusion, was superior to the control group; no differences were found among the three training groups. The implications of these and other findings for letter discrimination training were discussed.