Overlearning of a visual task makes the learning rapidly hyper-stabilized to protect it from being overwritten by training on a new task –A new role of overlearning since 1885–

Abstract

Overlearning refers to the continued training on a stimulus/task even after the learning has plateaued. Although research on overlearning has focused on how it affects retention of learning since its discovery in 1885 by Ebbinghaus, here we report a novel role of overlearning: Extensive overlearning dramatically changes the learning status from being plastic/unstable to hyper-stable to prevent the learning from being overwritten by new and different learning. First, we examined the effect of overlearning on stability of visual perceptual learning (VPL) using an interference paradigm. Training on a detection task with one orientation was immediately followed by the second training on another orientation. After VPL by the first training became plateaued, a small (slight) or large (extensive) amount of overlearning was conducted with a different group of 12 subjects. As a result, after slight overlearning VPL by the first training was retrogradely interfered with by the second training, whereas extensive overlearning made VPL by the first training anterogradely interfere with VPL by the second training. We further examined how excitatory/inhibitory neural processing is involved in overlearning by measuring the ratio of the concentrations of excitatory (glutamate) to inhibitory (GABA) neurotransmitters (E/I ratio) in the early visual areas using magnetic resonance spectroscopy. After slight overlearning, the E/I ratio became significantly higher than before training, whereas extensive overlearning made the E/I ratio significantly lower. These results suggest the following aspects. Training first makes the learning status plastic/unstable with an increased E/I ratio in the early visual areas. However, extensive overlearning makes the status hyper-stabilized with a rapid E/I ratio decrease. Such hyper-stabilization leads to anterograde interference and prevents the first learning from being overwritten by the second learning. Hyper-stabilization is different from typical stabilization which occurs over hours of time passage and allows for VPL of both the first and second trainings.

Meeting abstract presented at VSS 2016