According to the discrimination hypothesis (White, 2002), remembering is a delay-specific discrimination made at the time of retrieval. In the present experiment the delay-specific nature of the discrimination was made explicit by making correct choices in a delayed matching-to-sample task performed by pigeons conditional on whether the retention interval was short or long. Retention interval was varied over several durations in a maintained generalization test without reinforcement for correct matching responses. Opposing gradients demonstrated generalization of delay-specific remembering, consistent with the view that the temporal dimension of the retention interval can be treated in the same way as non-temporal dimensions of the sample stimulus.