Perceptual learning: toward a comprehensive theory

Abstract

Visual perceptual learning (VPL) is long-term performance increase resulting from visual perceptual experience. Task-relevant VPL of a feature results from training of a task on the feature relevant to the task. Task-irrelevant VPL arises as a result of exposure to the feature irrelevant to the trained task. At least two serious problems exist. First, there is the controversy over which stage of information processing is changed in association with task-relevant VPL. Second, no model has ever explained both task-relevant and task-irrelevant VPL. Here we propose a dual plasticity model in which feature-based plasticity is a change in a representation of the learned feature, and task-based plasticity is a change in processing of the trained task. Although the two types of plasticity underlie task-relevant VPL, only feature-based plasticity underlies task-irrelevant VPL. This model provides a new comprehensive framework in which apparently contradictory results could be explained.

Keywords: early visual cortex; location specificity; reweighting; task-irrelevant perceptual learning; task-relevant perceptual learning; transfer.

Figure 1

The Seitz & Watanabe model of VPL. The model assumes that VPL results from interactions between spatially diffusive reinforcement signals that are triggered by successful task performance and bottom-up signals from a feature presented during training, irrespective of whether the feature is task-relevant or task-irrelevant. Attention enhances bottom-up signals from a task-relevant feature, whereas it decreases or suppresses signals from a task-irrelevant feature. A red arrow represents excitatory stimulation and a blue arrow inhibitory stimulation. Task-irrelevant VPL occurs only when task-irrelevant feature signals are so weak that the signals fail to be detected and therefore are not suppressed by an attentional system. However, it is noteworthy that if a task-irrelevant feature is presented in an environment optimal to the feature, task-irrelevant learning can occur if the feature is supra-threshold. Task-relevant VPL and task-irrelevant VPL are location specific due to constraints from the bottom-up signals from a feature presented during training.

Figure 2

Schematic illustration of the dual plasticity model. In this model, there are two types of plasticity, feature-based plasticity and task-based plasticity. Feature-based plasticity results from presentation of a feature during training, irrespective of whether the feature is task-relevant or task-irrelevant. Feature-based plasticity reflects changes in feature representation and constrains VPL to the trained feature and location. Task-based plasticity arises as a result of involvement of a given task during training. Task-based plasticity reflects changes in task processing and constrains VPL to the trained task. Task-relevant VPL (R-VPL) consists of feature-based plasticity and task-based plasticity. Task-irrelevant VPL (I-VPL) consists of only feature-based plasticity.

Figure 3

Model of the mechanism of feature-based plasticity. While the Seitz & Watanabe model assumes that this is the entire mechanism of both task-relevant and task-irrelevant VPL, here we propose that the mechanism of the Seitz & Watanabe model is only for feature-based plasticity. Task-relevant VPL is subserved by this feature-based plasticity mechanism and also by task-based plasticity that results from performing a given task during training and is also specific for the trained task. A red arrow represents excitatory stimulation and a blue arrow inhibitory stimulation.

Figure 4

Distinction between what is involved during training on a task and what is changed in association with VPL of the task. What is involved is not necessarily the same as what is changed.