Four experimentally naïve white Carneau pigeons learned to match three colors to each other in a variant of an Identity matching-to-sample procedure with an FR20 on samples and a response-initiated FI8-s on comparisons. In Experiment 1, the extent to which subjects were matching on the basis of identity was assessed by presenting, in extinction, test trials comprising novel stimuli serving as the sample (and matching comparison) or as the nonmatching comparison. The results from Experiment 1 suggested intermediate or little to no transfer on the basis of identity. Experiment 2 reassessed transfer on the basis of identity with differential reinforcement on the test trials. Under these conditions, two of the four birds demonstrated substantially better than chance levels of performance. These data imply that while the extended response requirements may be necessary, other procedural aspects may be responsible for generalized identity matching in the pigeon.