In a two‐alternative forced‐choice frequency discrimination task observers were instructed to respond by indicating the presentation interval in which the higher of the two tones had appeared. The frequency difference between the two tones was fixed at 15 cps. The independent variable manipulated was the time between the two presentation intervals within each trial. Discrimination was measured at nine separation times from 30 msec to 9.6 sec. The data indicate a peak in discrimination, varying among the four observers, with about 1‐sec separation. The data are accounted for by two “memory” mechanisms. A short‐term effect causes the first of two closely spaced signals to interfere with the second. When the signals are separated by longer intervals a decaying memory of the frequency of the first makes comparison with the second signal less precise, lowering discrimination. [This work was supported by the U. S. Air Force under a contract monitored by the Air Force Office of Scientific Research.]