Pigeons were trained on a modified three-key matching-to-sample procedure, in which only one comparison key (rather than two) was lighted after an observing response to the centerkey standard. Pecks on keys of matching comparison hues were reinforced. When nonmatching hues appeared as the initially lighted comparisons, the nonmatching hue terminated, and the matching hue appeared on the other side key only if the pigeon did not peck the nonmatching comparison for 4.8 sec. Pecks to the nonmatching hue reset the 4.8-sec delay interval. Three hues were used during acquisition. During transfer tests, two novel hues were substituted individually or together for one or two of the training hues. Latencies to the novel side-key hue were shortest when a novel matching hue appeared as the standard on the center key, and were essentially identical to baseline matching latencies. In contrast, when a novel hue appeared as either a standard or comparison in a nonmatching combination, latencies increased with increasing separation between the novel hue and the nonmatching hue. These transfer data demonstrate the concept of hue matching.Key words: matching-to-sample, transfer, concept of matching, hue dimension, key peck, pigeons Cumming and Berryman (1961) reported that pigeons trained to high levels of accuracy on a three-key simultaneous matching-to-sample task did not exhibit transfer to a novel hue. Farthing and Opuda (1974) showed that pigeons do not exhibit transfer even if they have a history of reinforcement for pecking the novel hue outside the matching context. These data, together with related findings , suggest that pigeons in a matching-to-sample experiment learn only a set of specific "SD rules" (for example, "if red on center, peck red on side", "if green on center, peck green on side", "if blue on center, peck blue on side"