The mechanisms of perceptual learning are analyzed theoretically, probed in an orientation-discrimination experiment involving a novel nonstationary context manipulation, and instantiated in a detailed computational model. Two hypotheses are examined: modification of early cortical representations versus task-specific selective reweighting. Representation modification seems neither functionally necessary nor implied by the available psychophysical and physiological evidence. Computer simulations and mathematical analyses demonstrate the functional and empirical adequacy of selective reweighting as a perceptual learning mechanism. The stimulus images are processed by standard orientation- and frequency-tuned representational units, divisively normalized. Learning occurs only in the "read-out" connections to a decision unit; the stimulus representations never change. An incremental Hebbian rule tracks the task-dependent predictive value of each unit, thereby improving the signal-to-noise ratio of their weighted combination. Each abrupt change in the environmental statistics induces a switch cost in the learning curves as the system temporarily works with suboptimal weights.

Perceptual learning, the improvement in performance with practice, reflects plasticity in the adult visual system. We challenge a standard claim that specificity of perceptual learning depends on task difficulty during training, instead showing that specificity, or conversely transfer, is primarily controlled by the precision demands (i.e., orientation difference) of the transfer task. Thus, for an orientation discrimination task, transfer of performance improvement is observed in low-precision transfer tasks, while specificity of performance improvement is observed in high-precision transfer tasks, regardless of the precision of initial training. The nature of specificity places important constraints on mechanisms of transfer in visual learning. These results contribute to understanding generalization of practiced improvements that may be key to the development of expertise and for applications in remediation.

The role of feedback in perceptual learning is probed in an orientation discrimination experiment under destabilizing non-stationary conditions, and explored in a neural-network model. Experimentally, perceptual learning was examined with periodic alteration of a strong external noise context. The speed of learning, the performance loss at each change in external noise context (switch cost), and the asymptotic accuracy d' without feedback were very similar or identical to those with feedback. However, lack of feedback led to higher decision bias (error responses matching the external noise context). In the model, the stimulus representations are constant, whereas the read-out connections to a decision unit learn by a Hebbian plasticity rule that may be augmented by additional feedback input and criterion control of decision bias.

A memory-based scaling model--ANCHOR--is proposed and tested. The perceived magnitude of the target stimulus is compared with a set of anchors in memory. Anchor selection is probabilistic and sensitive to similarity, base-level strength, and recency. The winning anchor provides a reference point near the target and thereby converts the global scaling problem into a local comparison. An explicit correction strategy determines the final response. Two incremental learning mechanisms update the locations and base-level activations of the anchors. This gives rise to sequential, context, transfer, practice, and other dynamic effects. The scale unfolds as an adaptive map. A hierarchy of models is tested on a battery of quantitative measures from 2 experiments in absolute identification and category rating.

Eye movements are an important data source in vision science. However, the vast majority of eye movement studies ignore sequential information in the data and utilize only first-order statistics. Here, we present a novel application of a temporal-difference learning algorithm to construct a scanpath successor representation (SR; P. Dayan, 1993) that captures statistical regularities in temporally extended eye movement sequences. We demonstrate the effectiveness of the scanpath SR on eye movement data from participants solving items from Raven's Advanced Progressive Matrices Test. Analysis of the SRs revealed individual differences in scanning patterns captured by two principal components that predicted individual Raven scores much better than existing methods. These scanpath SR components were highly interpretable and provided new insight into the role of strategic processing on the Raven test. The success of the scanpath SR in terms of prediction and interpretability suggests that this method could prove useful in a much broader context.

Pupil size is correlated with a wide variety of important cognitive variables and is increasingly being used by cognitive scientists. Pupil data can be recorded inexpensively and non-invasively by many commonly used video-based eye-tracking cameras. Despite the relative ease of data collection and increasing prevalence of pupil data in the cognitive literature, researchers often underestimate the methodological challenges associated with controlling for confounds that can result in misinterpretation of their data. One serious confound that is often not properly controlled is pupil foreshortening error (PFE)—the foreshortening of the pupil image as the eye rotates away from the camera. Here we systematically map PFE using an artificial eye model and then apply a geometric model correction. Three artificial eyes with different fixed pupil sizes were used to systematically measure changes in pupil size as a function of gaze position with a desktop EyeLink 1000 tracker. A grid-based map of pupil measurements was recorded with each artificial eye across three experimental layouts of the eye-tracking camera and display. Large, systematic deviations in pupil size were observed across all nine maps. The measured PFE was corrected by a geometric model that expressed the foreshortening of the pupil area as a function of the cosine of the angle between the eye-to-camera axis and the eye-to-stimulus axis. The model reduced the root mean squared error of pupil measurements by 82.5 % when the model parameters were pre-set to the physical layout dimensions, and by 97.5 % when they were optimized to fit the empirical error surface.

Recent reports of training-induced gains on fluid intelligence tests have fueled an explosion of interest in cognitive training—now a billion-dollar industry. The interpretation of these results is questionable because score gains can be dominated by factors that play marginal roles in the scores themselves, and because intelligence gain is not the only possible explanation for the observed control-adjusted far transfer across tasks. Here we present novel evidence that the test score gains used to measure the efficacy of cognitive training may reflect strategy refinement instead of intelligence gains. A novel scanpath analysis of eye movement data from 35 participants solving Raven’s Advanced Progressive Matrices on two separate sessions indicated that one-third of the variance of score gains could be attributed to test-taking strategy alone, as revealed by characteristic changes in eye-fixation patterns. When the strategic contaminant was partialled out, the residual score gains were no longer significant. These results are compatible with established theories of skill acquisition suggesting that procedural knowledge tacitly acquired during training can later be utilized at posttest. Our novel method and result both underline a reason to be wary of purported intelligence gains, but also provide a way forward for testing for them in the future.

No abstract