The significance of massed and distributed practice in discrimination learning to criterion by goldfish was studied. All animals were trained to discriminate between 6 horizontal black and white stripes (positive stimulus) and 2 vertical black and white stripes (negative stimulus). 5 groups of animals, 12 Ss each, got different training programs. Groups varied in number of trials per day (30, 10, 5 successive trials) and in distribution of trials over the day (30 trials given in blocks of 5 with ITI of 1 h or 5 trials, each spaced by an ITI of 1 h). In general animals with a small number of massed trials/day were superior in early acquisition period. When trials/day were spaced the method proved to be more efficient. The training method showed no effect concerning retention.