Systematic variations in the rate and temporal patterns of responding under a multiple concurrent-chains schedule were quantified using recurrence metrics and selforganizing maps to assess whether individual rats showed consistent or idiosyncratic patterns. The results indicated that (1) the temporal regularity of response patterns varied as a function of number of training sessions, time on task, magnitude of reinforcement, and reinforcement contingencies; (2) individuals showed heterogeneous, stereotyped patterns of responding, despite similarities in matching behavior; (3) the specific trajectories of behavioral variation shown by individuals were less evident in group-level analyses; and (4) reinforcement contingencies within terminal links strongly modulated response patterns within initial links. Temporal regularity in responding was most evident for responses that led to minimally delayed reinforcers of larger magnitude. Models of response production and selection that take into account the time between individual responses, probabilities of transitions between response options, periodicity within response sequences, and individual differences in response dynamics can clarify the mechanisms that drive behavioral adjustments during operant conditioning.Keywords Behavior dynamics . Individual variations . Matching law . Neural networks . Reinforcement schedules A basic assumption of several theories of operant conditioning is that group-level measures of performance are sufficient for understanding the mechanisms of learning within individuals (Pear, 2001;Rachlin, 2000;Staddon & Cerutti, 2003;Williams, 1990). Although there is good evidence that this is often not the case (Gallistel, Fairhurst, & Balsam, 2004;Gallistel et al., 2007), what kinds of behavioral descriptions might best address such complications is unknown. Possibly, modeling behavior at the level of the individual would suffice. But, if the mechanisms engaged during learning vary significantly across individuals, then fitting a single uniform learning model to individual performances might be just as inappropriate as assuming that group-level models adequately describe behavior (e.g., see Worthy, Hawthorne, & Otto, 2013). Gaining a clearer understanding of which aspects of learned behavior are prevalent within and across individuals, and the extent to which those properties are predictable, is thus critical to choosing measures of behavior that most clearly reveal the nature of the mechanisms that drive long-term changes in behavior. The purpose of the present study was to quantitatively assess the regularity 1 of responses produced by individual subjects performing in a relatively complex operant conditioning task to determine which measures of behavior best characterized changes in subjects' response patterns as a function of training conditions.Early in the history of animal learning research, it was discovered that training animals within constrained environments often led subjects to perform stereotyped actions. For instance, Gut...