The role of prior knowledge in incremental associative learning: An empirical and computational approach

Abstract

Abstract Our experiences are encoded in relation to existing knowledge, and learning of new information is influenced by what has already been learned. Although learning is often an incremental process spanning multiple repetitions, the influences of prior knowledge have thus far been investigated primarily in one-trial learning. Incremental learning studies have generally not taken prior knowledge influences into consideration. Aiming to fill this gap, we examined the contribution of prior knowledge to incremental learning. Prior knowledge was manipulated using famous and novel faces. Participants viewed repeated pairs of faces: either a famous face paired with a novel face, or a pair of novel faces. Prior knowledge facilitated processing, as evidenced by decreased reaction times, in a task that was unrelated to previous knowledge. This decrease only emerged with repetitions and was maintained throughout the experiment, demonstrating a continuous influence of prior knowledge on learning. Enhanced processing was also related to learning of specific-face information, beyond response learning. Interestingly, decreased reaction times were observed during learning even when participants did not explicitly remember pair-associations in a final memory test. Computational modelling suggests that delayed facilitation may be attributed to prior knowledge, which allows for the existence of a stable representation into which new information can be assimilated via back-propagation learning. Focusing on incremental learning and using behavioral measures and computational modelling, the current study suggests that the influence of prior knowledge on learning and memory may span multiple learning processes.