When teaching discriminations, many researchers and practitioners recommend presenting multiple examples of both the correct and incorrect stimuli. To test this suggestion, we compared two procedures for presenting multiple examples. In one, multiple examples across trials (ME, Across), one correct (S+) and one incorrect (S-) stimulus were presented each trial; examples then changed across trials. In another procedure, multiple examples within trials (ME, Within), three stimuli (either 2 S+'s and 1 S-, or 1 S+ and 2 S-'s) were presented each trial; examples again changed across trials. Two experiments were conducted to test these procedures. The first procedure used a non-fading program to teach discrimination; the second used a fading procedure. In the first experiment, we taught 10 persons to identify words under these two procedures. The former procedure was superior in acquisition; the latter procedure, however, was better under generalization for most participants. In the second experiment, we presented the two procedures within a fading paradigm. The results replicated those in Experiment 1: ME, Across was better for acquisition, but ME, Within was better for generalization. Results were discussed and follow-up studies suggested.