Computational models of perceptual learning across multiple auditory tasks: Modeling daily learning limits as memory decay

Humans have a remarkable ability to adapt their perceptual acuity to the task at hand, commonly referred to in the literature as perceptual learning. Understanding this ability at a computational level may have important implications across a wide variety of different psychological phenomena. There is evidence suggesting this ability plays an important role in speech comprehension, mathematics, and perceptual expertise, for instance. Computational models of perceptual learning have largely focused on hypothesizing how one or more mechanisms might explain the observed perceptual learning for a single task. Here we explore how a single model might explain the learning curves across two auditory perceptual learning tasks. Our results suggest that an ideal observer model with noisy input can predict learning when daily limits are not reached, and that daily limits on learning can be modeled by a decay of memory for trials observed on the current day of practice.