Lever pressing was maintained by a concurrent token production schedule in rats. Token deliveries (i.e., ½" steel ball bearings) were assigned probabilistically to the right or left lever (on 1:3 or 1:9 ratio for Experiment 1 and 1:6 ratio for Experiment 2). The location of the rich lever remained constant within session, but varied across daily sessions according to a pseudorandom binary sequence. Once assigned to a lever, token delivery was arranged by a random interval 15 s schedule. Transition to token exchange was determined by varying exchange-production schedules that varied after each condition (FR 1 for Experiment 1; and RR 2, FR 2, RR 3, and FR 4 for Experiment 2). During token exchange, depositing a token in the receptacle was reinforced with access to sweetened condensed milk. The generalized matching law provided an adequate description of the session wide ratio of left to right lever presses, detecting considerable undermatching across all conditions. Sensitivity to the token ratios were best described by the current session's reinforcement ratio in comparison to the prior two session's reinforcement ratios. One of four subjects showed increased response rates across successive token deliveries in the exchange-production schedule while the remaining subjects showed the opposite effects of decreased response rates as tokens were earned. Similarly, sensitivity to token-production ratios increased across successive token deliveries for one of four subjects. Differences between subjects can be explained by the duration of elicited sign tracking of tokens, measured by duration of postreinforcement pausing and direct observation of rats during sessions.