Pigeons acquired a successive delayed matching‐to‐sample task at delay intervals ranging from 2.5 to 7 seconds. Test sessions were conducted during which delay‐interval illumination conditions were changed from those illumination conditions that prevailed during the baselines. Compared to baseline delayed matching performance, changing delay‐interval illumination disrupted matching. This disruption occurred whether the change in delay‐interval illumination represented an increase or a decrease, relative to the baseline, and whether there was or was not a change in illumination during the test session. It was concluded that illumination per se introduced during delay intervals of delayed matching tasks does not interfere with pigeon short‐term memory. Rather, a change in delay‐interval illumination, relative to the baseline, appears to retroactively interfere in pigeon shor