The ability of normal cats to perform delayed matching- and nonmatching-to-sample with trial-unique stimuli was investigated both in a modified Wisconsin General Testing Apparatus requiring manipulatory responses and in a Nencki-type testing room requiring locomotor responses. Cats trained in the WGTA learned the two tasks at about the same rate, on average, as that reported for monkeys. However, unlike monkeys, whose strong preference for novelty facilitates their learning of the nonmatching rule and retards their learning of the matching rule, the cats learned the two different rules at about the same rate, suggesting that cats do not share the monkey's strong preference for novelty. In contrast to their relatively rapid learning of the manipulatory versions of the two tasks, cats learned the locomotor versions only slowly or even failed to learn. Experimental analysis indicated that a major source of the cats' difficulty on these locomotor versions was interference from a strong tendency in the large testing room to use visuospatial strategies. Nevertheless, once the matching or nonmatching rule was learned at short delays, whether in the WGTA or the testing room, the cats performed at criterion levels without further training even at delays of 10 minutes, indicating that this species, like monkeys, has a highly developed long-term recognition memory ability.