An Integrated Trial-Level Performance Measure: Combining Accuracy and RT to Express Performance During Learning

Memory researchers have studied learning behavior and extracted regularities describing learning and forgetting over time. Early work revealed forgetting curves and the benefits of temporal spacing and testing for learning. Computational models formally implemented these regularities to capture relevant trends over time. As these models improved, they were applied to adaptive learning contexts, where learning profiles could be identified from responses to past learning events to predict and improve future performance. Often times, past performance is expressed as accuracy alone. Here we explore whether a model's predictions can be improved if past performance is expressed by an integrated measure that combines accuracy and response times (RT). We present a simple, data-driven method to combine accuracy and RT on a trial-by-trial basis. This research demonstrates that predictions made using the Predictive Performance Equation improve when past performance is expressed as an integrated measure rather than accuracy alone.