Sunzmary.-This experiment was designed to test implications of all-or-none vs incremental positions on associative learning. These contrasting implications were viewed in terms of t t e original acquisition of the association, further learning of the association, and acquisition of a new response to the original stimulus. At acquisition an all-or-no2e theory was favored by (1) chance Pr of correct on Test 2 given incorrect on Test 1, with latencies similar to those of incorrect rather than correct responses, ( 2 ) the large decrease in latency during acquisition rather than prior to or after acquisition, ( 3 ) latencies on acquisition consistent regardless of trial, (4) similarity of acquisition of reversal Rs to original learning. An incremental theory was supported by ( 1 ) proportion of CRs acquired per trial and( 2 ) latency decreases after acquisition. Estes' concept of chance CRs was supported, but the "neutral" test trials were not.Increments in habit strength with practice is a learning concept widely held since Thorndike's first learning experiments. This view of the nature of learning has been supported by common sense and the group learning curve. Task performance seems to improve gradually over trials, but while performance provides the data from which learning is inferred, it has been common in the last quarter century to differentiate berween the learning of an associative habit and the performance of such habit. It has been shown rather conclusively that motivation and reward affect associative learning and performance differently, indicating the necessity of discriminating between learning and performance in theoretical discussions (Deese, 1958).T h e concept of incremental associative learning has recently been challenged by certain Es in the field of learning. It has even been suggested that the idea of incremental habit strength over trials is an artifact of the neth hod of averaging group data (Estes, 1960).T h e purpose of this thesis was to present dacn relevant to this current issue in learning theory. T h e issue involves two positions on the nature of associative learning, the all-or-none or one-trial position and a gradual strengthening of response tendencies or an incremental position. The all-or-none position states that eirher an association is made to full strength or not a t all on a given trial. T h e incremental viewpoint maintains that the learning process involves gradual increments in associative strength over repeated training trials, increasing the probability of the reinforced response to asymptote for the experimencal condition.T w o major designs heve been developed co test contrasting predictions of 'Based on a doctoral dissertation presented to the Universiry of Minnesota.