International Journal of Comparative Psychology

Cumulative records, which show individual responses in real time, are a natural but neglected starting point for understanding the dynamics of operant behavior.   To understand the processes that underlie molar laws like matching, it is also helpful to look at choice behavior in situations such as concurrent random ratio that lack the stabilizing feedback intrinsic to concurrent variable-interval schedules.  The paper identifies some basic, nontemporal properties of operant learning: Post-reinforcement pulses at the beginning of FI learning, regression, faster reversal learning after shorter periods, and choice distribution on identical random ratios at different absolute ratio values.  These properties suggest that any operant-learning model must include silent responses, competing to become the active response; and response strengths that reflect more than immediate past history of reinforcement.  The cumulative-effects model is one that satisfies these conditions.