Abstract
Introduction: Measured human perceptual learning (or lack of) for a given task is generally attributed to properties of the human perceptual system. However, as in other domains (e.g. object recognition, Tjan and Legge, 1998), the amount of task-inherent stimulus information to be learned will also influence the amount of human perceptual learning. Here, we investigate the effect of stimulus information on human learning in a localization task where the observer learns about: 1) the orientation of the signal, 2) orientation and polarity. Ideal observer analysis is used to objectively assess the maximal amount of stimulus information that can be learned. Method: We used a rapid-perceptual learning paradigm (Abbey et al., 2001). A learning set consisted of 4 trials. One out of 4 signals was randomly chosen and used throughout a set. On each trial, the signal appeared randomly in 1 out of 8 image locations embedded in image noise. Three observers had to localize the target on each trial and identify it on the 4th trial of the learning set. The signal set consisted of elongated Gaussians with 4 different orientations (0°, 45°, 90°, 135°). In the orientation (ORI) condition the signals had the same polarity, while in the orientation/polarity (ORIPOL) condition the 0° and 90° signals were Gaussian increments and the 45° and 135° signals were Gaussian decrements. Observer participated in 1200 learning sets per condition. Result: Human learning performance (averaged across observers) was 2.8% (±0.85%) and 7.63% (±1.17%) across the 4 learning trials for the ORI and ORIPOL condition, respectively. Ideal observer learning performance was 6.0% and 14.9%, respectively, suggesting that the greater human learning in the orientation/polarity condition reflects stimulus information. Conclusion: In any learning task, the stimulus information needs to be taken into account before attributing human perceptual learning (or lack of) to properties of the perceptual system.