Abstract : Octopuses, unlike other invertebrates tested are capable of definite improvement over a series of discrimination reversals; one animal showed one trial reversal. One cause of improvement over a series of reversals is that during later reversals animals are able to ignore irrelevant cues, by which in early reversals they are hindered. A similar result has been found in rats by Lawrence and Mason (1955). Adaptation of responses to irrelevant cues might come about in either of two ways. Animals might either learn through experience of a particular irrelevant cue that it is irrelevant, or they might learn that a particular cue is relevant, and thus come to pay attention to this cue and ignore others. The first alternative is suggested by Harlow (1959) in his theory of Error Factor elimination; the second by Sutherland's model of discrimination learning (Sutherland, 1959). In the present experiment, all animals had equal opportunity to learn that brightness was the relevant cue, but only Group HV had the opportunity to learn that orientation was irrelevant. In spite of this, the three subjects in Groups O confronted with the second irrelevant cue from R sub 9 onwards, both learned subsequent reversals as fast as Group HV, and showed no greater tendency to respond to orientation. The results are consistent with those obtained in a previous experiment with rats (Mackintosh, 1963), and support Sutherland's theory of discrimination learning. (Author)