Models which attempt to account for our ability to discriminate the pitch of pure tones are discussed. It is concluded that models based on a place (spectral) analysis should be subject to a limitation of the type Δf⋅d⩾ constant, where Δfis the frequency difference limen (DL) for a tone pulse of durationd. The value of this constant will depend on the ability of the system to resolve small intensity differences. If a resolution of 1 dB is assumed, the value of the constant is about 0.24. In principle, a mechanism based on the measurement of time intervals could do considerably better than this. Frequency DLs were measured over a wide range of frequencies and durations. It was found that at short durations the product of Δfanddwas about one order of magnitude smaller than the minimum predicted from the place model, except for frequencies above 5 kHz. A “kink” in the obtained functions was also observed at about 5 kHz. It is concluded that the evidence is consistent with the operation of a time‐measuring mechanism for frequencies below 5 kHz, and with a spectral or place mechanism for frequencies above this.