Ten rats were trained to press a right lever following two successive sounds and to press a left lever following four successive sounds. Three sound sequences were used in initial discrimination training, such that reliable classification could not be based on the duration of each sound, the interval between sounds, or the total duration of the sound sequence. Classification of seven novel sound sequences suggested that the animals were either using number or the sum of sound durations in a sequence as the relevant cue. When this total sound duration was put in conflict with number, rats classified by number. The conclusion was that rats can discriminate number, even when temporal cues are controlled.The problem was to determine whether the number of successive events (stimuli or responses) could serve as an effective discriminative stimulus. Although earlier research on fixed-ratio (FR) performance (e.g., Ferster & Skinner, 1957) involved variation in the number of responses necessary for reinforcement, Mechner (1958) was the first to attempt to identify the number of responses as an effective discriminative stimulus. In this procedure the press of a left lever (left leverpress) by a rat was reinforced only after at least N successive right leverpresses, with values of N over blocks of sessions at 4, 8, 12, or 16 responses. The probability of a left leverpress as a function of the number of prior right leverpresses was fairly symmetrical on a linear scale, with a maximum probability near N. As N increased (4,8, 12, 16), variability of the distribution increased. The effective variable in Mechner's experiment might have been how long the animal pressed the right lever, not the number of right leverpresses. If the animal pressed on the right lever at a fairly steady rate, as usually occurs in FR schedules, the number of right leverpresses would be completely confounded with the time elapsed since the first right leverpress. Functions similar to those obtained by Mechner occur when reinforcement is contingent on a response after a particular time in the peak procedure (Roberts, 1981). Therefore, although counting may account for Mechner's data, a timing mechanism is also plausible.One way to separate time and number is to use a treatment that changes response rate. Mechner This research was supported by NSF Grant BNS 79-04792. We thank Warren H. Meck for assistance and advice on this experiment. Reprints may be obtained from Russell M. Church, Department of Psychology, Brown University, Providence, Rhode Island 02912.and Guevrekian (1962) found that an increase in water deprivation increased response rate but did not affect the relation between the probability of a left leverpress and the number of preceding right leverpresses. Thus, they concluded that the left presses were controlled by number, not time. Even if rapid learning of a new temporal criterion is ruled out, however, subjective time could be the determiner of the left leverpress. An increase in deprivation might increase the speed of an internal clock,...