Abstract: Discriminative learning abilities were studied in 12 mature, malnourished horses. All horses initially received a condition score (CS) between 2 and 4 on a scale of 1 (poor) to 9 (extremely fat). Then horses were assigned to one of 3 treatments based on their eventual, rehabilitated CS during discrimination testing: thin, CS 1-3; moderate, CS 4-6; and fat, CS 7-9. The discrimination learning task was performed for 14 days with a maximum of 20 trials per day. Daily criterion was set at eight consecutively correct trails. Total trials to first criteria and total errors during testing were recorded. Analysis of variance showed that treatments did not differ (P>0.05) in total trials to first criterion, however horses on the fat treatment did have higher total error scores (P<0.05) than horses on the thin or moderate treatments. This difference was probably owing to lack of motivation in the fat treatment horses, rather than to true learning ability differences. The sex of the horse did not significantly affect either learning score.