A signal was presented in each observation interval of two‐alternative, temporal‐forced‐choice trials, and 12 observers each selected the louder of the two signals. The observers were informed that the signals of each trial would be of 1000 Hz and either 2 or 4 dB different in amplitude. Some 70% of the trials did in fact present such pairs of 1000‐Hz signals. Surreptitiously, on infrequent and randomly determined trials, one of the signals presented, a probe signal, was of a frequency other than 1000 Hz. The probe‐signal method thus applied to the nominal amplitude discrimination task produced results clearly demonstrating differential performance as a function of the signal frequency, for 10 of the 12 observers. For those 10 observers, the relation between performance and signal frequency appears asymmetrical about 1000 Hz.