Initially, psychometric functions were measured for the detection of amplitude modulation (AM) or frequency modulation (FM), using a two-alternative forced-choice (2AFC) task. Carrier frequencies were 125, 1000, and 6000 Hz, and modulation rates were 2, 5, and 10 Hz. For the two lower carrier frequencies, FM detection tended to be best at the lowest modulation rate while AM detection was best at the highest rate. For the 6000-Hz carrier, both AM and FM detection tended to be poorest at the lowest modulation rate. Then, pairs of values of AM and FM were selected that would be equally detectable, and psychometric functions were measured for the discrimination of AM from FM, again in a 2AFC task. For carrier frequencies of 125 and 1000 Hz, the ability to discriminate AM from FM was always poorest at the highest modulation rate (10 Hz); at this rate some subjects were essentially unable to discriminate AM from FM when the detectability of the modulation was relatively low (d′ of 1.16 and below). For a modulation rate of 2 Hz, and when the detectability of the modulation was moderate (d′ up to about 2), some subjects discriminated the type of modulation as well as they detected the modulation. For a carrier frequency of 6000 Hz, the effect of modulation rate varied across subjects, but there was still a trend for poorer discrimination of modulation type at the highest modulation rate. It is suggested that FM detection at a 10-Hz modulation rate is based largely on changes in excitation level for all carrier frequencies. For a 2-Hz modulation rate, and for the two lowest carrier frequencies, an extra mechanism, possibly based on phase locking, may play a role in the detection and discrimination of FM. This mechanism may be ineffective at modulation rates above about 5 Hz because the stimuli spend insufficient time at frequency extremes. To check on this, psychometric functions were measured for the detection of FM and AM using quasitrapezoidal modulation with a rate of five periods per second and carriers of 250, 1000, and 6000 Hz. This produced improvements in performance relative to that obtained with 5-Hz sinusoidal modulation and, for the two lower carrier frequencies only, the improvements were markedly greater for FM than for AM detection. This is consistent with the idea that the use of phase-locking information depends on the time that the stimuli spend at frequency extremes.