Level and mechanisms of perceptual learning: Learning first-order luminance and second-order texture objects

Bayesian approaches to smooth pursuit of random dot kinematograms: effects of varying RDK noise and the predictability of RDK direction

Smooth pursuit eye movements respond on the basis of both immediate and anticipated target motion, where anticipations may be derived from either memory or perceptual cues. To study the combined influence of both immediate sensory motion and anticipation, subjects pursued clear or noisy random dot kinematograms (RDKs) whose mean directions were chosen from Gaussian distributions with SDs = 10° (narrow prior) or 45° (wide prior). Pursuit directions were consistent with Bayesian theory in that transitions over time from dependence on the prior to near total dependence on immediate sensory motion (likelihood) took longer with the noisier RDKs and with the narrower, more reliable, prior. Results were fit to Bayesian models in which parameters representing the variability of the likelihood either were or were not constrained to be the same for both priors. The unconstrained model provided a statistically better fit, with the influence of the prior in the constrained model smaller than predicted from strict reliability-based weighting of prior and likelihood. Factors that may have contributed to this outcome include prior variability different from nominal values, low-level sensorimotor learning with the narrow prior, or departures of pursuit from strict adherence to reliability-based weighting. Although modifications of, or alternatives to, the normative Bayesian model will be required, these results, along with previous studies, suggest that Bayesian approaches are a promising framework to understand how pursuit combines immediate sensory motion, past history, and informative perceptual cues to accurately track the target motion that is most likely to occur in the immediate future.NEW & NOTEWORTHY Smooth pursuit eye movements respond on the basis of anticipated, as well as immediate, target motions. Bayesian models using reliability-based weighting of previous (prior) and immediate target motions (likelihood) accounted for many, but not all, aspects of pursuit of clear and noisy random dot kinematograms with different levels of predictability. Bayesian approaches may solve the long-standing problem of how pursuit combines immediate sensory motion and anticipation of future motion to configure an effective response.

Current directions in visual perceptual learning

The visual expertise of adult humans is jointly determined by evolution, visual development, and visual perceptual learning. Perceptual learning refers to performance improvements in perceptual tasks after practice or training in the task. It occurs in almost all visual tasks, ranging from simple feature detection to complex scene analysis. In this Review, we focus on key behavioral aspects of visual perceptual learning. We begin by describing visual perceptual learning tasks and manipulations that influence the magnitude of learning, and then discuss specificity of learning. Next, we present theories and computational models of learning and specificity. We then review applications of visual perceptual learning in visual rehabilitation. Finally, we summarize the general principles of visual perceptual learning, discuss the tension between plasticity and stability, and conclude with new research directions.

Visual Perceptual Learning and Models

Visual perceptual learning through practice or training can significantly improve performance on visual tasks. Originally seen as a manifestation of plasticity in the primary visual cortex, perceptual learning is more readily understood as improvements in the function of brain networks that integrate processes, including sensory representations, decision, attention, and reward, and balance plasticity with system stability. This review considers the primary phenomena of perceptual learning, theories of perceptual learning, and perceptual learning's effect on signal and noise in visual processing and decision. Models, especially computational models, play a key role in behavioral and physiological investigations of the mechanisms of perceptual learning and for understanding, predicting, and optimizing human perceptual processes, learning, and performance. Performance improvements resulting from reweighting or readout of sensory inputs to decision provide a strong theoretical framework for interpreting perceptual learning and transfer that may prove useful in optimizing learning in real-world applications.

The Functional Form of Performance Improvements in Perceptual Learning

The functional form of performance improvements has been extensively studied in speeded cognitive and motor tasks; in such tasks, reductions in response times have been characterized by the ubiquitous power law of learning or by a simpler exponential form. Performance improvements in perceptual capacities are also important in expertise, but their functional form is unknown. This study investigated the functional form of perceptual learning. For individual observers, reductions in thresholds were best described by an exponential function, rather than a power or compound exponential and power (apex) function. Learning was specific to orientation, a result that supports the perceptual locus of the learning, and was decoupled in high and low external noise, a result that reflects separable learning mechanisms in the two conditions. The simple exponential form of learning implies a constant relative rate of learning throughout practice; there was no evidence supporting multilevel hypotheses, such as serial reverse hierarchical and parallel-learning models, that posit multiple processes of learning characterized by different rates.

Perceptual learning improves neural processing in myopic vision

Visual performance is jointly determined by the quality of optical transmission of the eye and neural processing in the visual system. An open question is: Can effects of optical defects be compensated by perceptual learning in neural processing? To address this question, we conducted a perceptual learning study on 23 observers with myopic vision, targeting high frequency deficits by training them in a monocular grating detection task in the non-dominant eye near their individual cutoff spatial frequencies. The contrast sensitivity function and visual acuity in both eyes (without optical correction) were assessed for all the observers in the training group before and after training, and for all the observers in the control group twice with a 10-day interval between the tests. In addition, the threshold versus external noise contrast function was measured for five observers in the training group before and after training. We found that (a) training significantly improved contrast sensitivity at the trained spatial frequency, visual acuity, and contrast sensitivity over a wide range of spatial frequencies in both eyes; (b) training did not lead to any significant refractive changes; (c) the mechanism of improvements was a combination of internal additive noise reduction and external noise exclusion; and (d) the improvements in visual acuity and contrast sensitivity were almost fully retained for at least four months in the three observers tested. These results suggest that perceptual learning may provide a potential noninvasive procedure to compensate for optical defects in mild to modest myopia.

Involvement of the TPJ Area in Processing of Novel Global Forms

The neuropsychological syndrome “simultanagnosia” is characterized by the inability to integrate local elements into a global entity. This deficit in Gestalt perception is mainly apparent for novel global structures administered in clinical tests or unfamiliar visual scenes. Recognition of familiar complex objects or well-known visual scenes is often unaffected. Recent neuroimaging studies and reports from simultanagnosia patients suggest a crucial involvement of temporoparietal brain areas in processing of hierarchically organized visual material. In this study, we investigated the specific role of the TPJ in Gestalt perception. On the basis of perceptual characteristics known from simultanagnosia, we hypothesized that TPJ is dominantly involved in processing of novel object arrangements. To answer this question, we performed a learning study with hierarchical stimuli and tested behavioral and neuronal characteristics of Gestalt perception pre- and posttraining. The study included 16 psychophysical training sessions and two neuroimaging sessions. Participants improved their behavioral performance for trained global stimuli and showed limited transfer to untrained global material. We found significant training dependent neuronal signal modulations in anterior right hemispheric TPJ regions. These activation changes were specific to trained global stimuli, whereas no systematic neuronal response changes were observed for recognition of untrained global stimuli, local elements and regular objects that served as control stimuli. In line with perceptual characteristics in simultanagnosia, the results argue for an involvement of TPJ in processing of novel global structures. We discuss the signal modulations in the context of a more efficient or different neuronal strategy to process familiar global stimuli.

Broad bandwidth of perceptual learning in second-order contrast modulation detection

Comparing characteristics of learning in first- and second-order systems might inform us about different neural plasticity in the two systems. In the current study, we aim to determine the properties of perceptual learning in second-order contrast modulation detection in normal adults. We trained nine observers to detect second-order gratings at an envelope modulation spatial frequency of 8 cycles/° with their nondominant eyes. We found that, although training generated the largest improvements around the trained frequency, contrast sensitivity over a broad range of spatial frequencies also improved, with a 4.09-octave bandwidth of perceptual learning, exhibiting specificity to the trained spatial frequency as well as a relatively large degree of generalization. The improvements in the modulation sensitivity function (MSF) were not significantly different between the trained and untrained eyes. Furthermore, training did not significantly change subjects' ability in detecting first-order gratings. Our results suggest that perceptual learning in second-order detection might occur at the postchannel level in binocular neurons, possibly through reducing the internal noise of the visual system.

Processing deficits of motion of contrast-modulated gratings in anisometropic amblyopia

Several studies have indicated substantial processing deficits for static second-order stimuli in amblyopia. However, less is known about the perception of second-order moving gratings. To investigate this issue, we measured the contrast sensitivity for second-order (contrast-modulated) moving gratings in seven anisometropic amblyopes and ten normal controls. The measurements were performed with non-equated carriers and a series of equated carriers. For comparison, the sensitivity for first-order motion and static second-order stimuli was also measured. Most of the amblyopic eyes (AEs) showed reduced sensitivity for second-order moving gratings relative to their non-amblyopic eyes (NAEs) and the dominant eyes (CEs) of normal control subjects, even when the detectability of the noise carriers was carefully controlled, suggesting substantial processing deficits of motion of contrast-modulated gratings in anisometropic amblyopia. In contrast, the non-amblyopic eyes of the anisometropic amblyopes were relatively spared. As a group, NAEs showed statistically comparable performance to CEs. We also found that contrast sensitivity for static second-order stimuli was strongly impaired in AEs and part of the NAEs of anisometropic amblyopes, consistent with previous studies. In addition, some amblyopes showed impaired performance in perception of static second-order stimuli but not in that of second-order moving gratings. These results may suggest a dissociation between the processing of static and moving second-order gratings in anisometropic amblyopia.

Binocular combination of second-order stimuli

Phase information is a fundamental aspect of visual stimuli. However, the nature of the binocular combination of stimuli defined by modulations in contrast, so-called second-order stimuli, is presently not clear. To address this issue, we measured binocular combination for first- (luminance modulated) and second-order (contrast modulated) stimuli using a binocular phase combination paradigm in seven normal adults. We found that the binocular perceived phase of second-order gratings depends on the interocular signal ratio as has been previously shown for their first order counterparts; the interocular signal ratios when the two eyes were balanced was close to 1 in both first- and second-order phase combinations. However, second-order combination is more linear than previously found for first-order combination. Furthermore, binocular combination of second-order stimuli was similar regardless of whether the carriers in the two eyes were correlated, anti-correlated, or uncorrelated. This suggests that, in normal adults, the binocular phase combination of second-order stimuli occurs after the monocular extracting of the second-order modulations. The sensory balance associated with this second-order combination can be obtained from binocular phase combination measurements.

A Weber-like law for perceptual learning

What determines how much an organism can learn? One possibility is that the neural factors that limit sensory performance prior to learning, place an upper limit on the amount of learning that can take place. We tested this idea by comparing learning on a sensory task where performance is limited by cortical mechanisms, at two retinal eccentricities. Prior to learning, visual performance at the two eccentricities was either unmatched or equated in two different ways (through spatial scaling or visual crowding). The magnitude of learning was equivalent when initial levels of performance were matched regardless of how performance was equated. The magnitude of learning was a constant proportion of initial performance. This Weber-like law for perceptual learning demonstrates that it should be possible to predict the degree of perceptual improvement and the final level of performance that can be achieved via sensory training, regardless of what cortical constraint limits performance.

Beyond multiple pattern analyzers modeled as linear filters (as classical V1 simple cells): useful additions of the last 25 years

This review briefly discusses processes that have been suggested in the last 25 years as important to the intermediate stages of visual processing of patterns. Five categories of processes are presented: (1) Higher-order processes including FRF structures; (2) Divisive contrast nonlinearities including contrast normalization; (3) Subtractive contrast nonlinearities including contrast comparison; (4) Non-classical receptive fields (surround suppression, cross-orientation inhibition); (5) Contour integration.

Visual perceptual learning

Perceptual learning refers to the phenomenon that practice or training in perceptual tasks often substantially improves perceptual performance. Often exhibiting stimulus or task specificities, perceptual learning differs from learning in the cognitive or motor domains. Research on perceptual learning reveals important plasticity in adult perceptual systems, and as well as the limitations in the information processing of the human observer. In this article, we review the behavioral results, mechanisms, physiological basis, computational models, and applications of visual perceptual learning.

Specificity of perceptual learning increases with increased training

Perceptual learning often shows substantial and long-lasting changes in the ability to classify relevant perceptual stimuli due to practice. Specificity to trained stimuli and tasks is a key characteristic of visual perceptual learning, but little is known about whether specificity depends upon the extent of initial training. Using an orientation discrimination task, we demonstrate that specificity follows after extensive training, while the earliest stages of perceptual learning exhibit substantial transfer to a new location and an opposite orientation. Brief training shows the best performance at the point of transfer. These results for orientation-location transfer have both theoretical and practical implications for understanding perceptual expertise.

Modeling mechanisms of perceptual learning with augmented Hebbian re-weighting

Using the external noise plus training paradigm, we have consistently found that two independent mechanisms, stimulus enhancement and external noise exclusion, support perceptual learning in a range of tasks. Here, we show that re-weighting of stable early sensory representations through Hebbian learning (Petrov et al., 2005, 2006) can generate performance patterns that parallel a large range of empirical data: (1) perceptual learning reduced contrast thresholds at all levels of external noise in peripheral orientation identification (Dosher & Lu, 1998, 1999), (2) training with low noise exemplars transferred to performance in high noise, while training with exemplars embedded in high external noise transferred little to performance in low noise (Dosher & Lu, 2005), and (3) pre-training in high external noise only reduced subsequent learning in high external noise, whereas pre-training in zero external noise left very little additional learning in all the external noise conditions (Lu et al., 2006). In the augmented Hebbian re-weighting model (AHRM), perceptual learning strengthens or maintains the connections between the most closely tuned visual channels and a learned categorization structure, while it prunes or reduces inputs from task-irrelevant channels. Reducing the weights on irrelevant channels reduces the contributions of external noise and additive internal noise. Manifestation of stimulus enhancement or external noise exclusion depends on the initial state of internal noise and connection weights in the beginning of a learning task. Both mechanisms reflect re-weighting of stable early sensory representations.

New approaches to visual rehabilitation for cortical blindness: outcomes and putative mechanisms

Cortical blindness is a chronic loss of vision following damage to the primary visual cortex (V1) or its postchiasmal afferents. Such damage is followed by a brief period of spontaneous plasticity that rarely lasts beyond 6 months. Following this initial phase, the visual deficit is thought to be stable, intractable, and permanent. Cortically blind subjects demonstrate spontaneous oculomotor adaptations to their deficits that can be further improved by saccadic localization training. However, saccadic training does not improve visual sensitivity in the blind field. In contrast, recent studies by a number of independent groups suggest that localized, repetitive perceptual training can improve visual sensitivity in the blind field, although mechanisms underlying the observed recovery remain unclear. This review discusses the current literature on rehabilitative strategies used for cortical blindness with emphasis on the use of perceptual training methods. The putative mechanisms that underlie the resulting, training-induced visual improvements are then outlined, along with the special challenges posed to their elucidation by the great variability in the extent and sometimes nature of the V1 damage sustained in different individuals.

MECHANISMS OF PERCEPTUAL LEARNING

What is learned in perceptual learning? How does perceptual learning change the perceptual system? We investigate these questions using a systems analysis of the perceptual system during the course of perceptual learning using psychophysical methods and models of the observer. Effects of perceptual learning on an observer's performance are characterized by external noise tests within the framework of noisy observer models. We find evidence that two independent mechanisms, external noise exclusion and stimulus enhancement support perceptual learning across a range of tasks. We suggest that both mechanisms may reflect re-weighting of stable early sensory representations.

Sensory noise explains auditory frequency discrimination learning induced by training with identical stimuli

Thresholds in various visual and auditory perception tasks have been found to improve markedly with practice at intermediate levels of task difficulty. Recently, however, there have been reports that training with identical stimuli, which, by definition, were impossible to discriminate correctly beyond chance, could induce as much discrimination learning as could training with different stimuli. These surprising findings have been interpreted as evidence that discrimination learning can occur in the absence of perceived differences between stimuli and need not involve the fine-tuning of a discrimination mechanism. Here, we show that these counterintuitive findings of discrimination learning without discrimination can be understood simply by considering the effect of internal noise on sensory representations. Because of such noise, physically identical stimuli are unlikely to be perceived as being strictly identical. We show that, given empirically derived levels of sensory noise, perceived differences evoked by identical stimuli are actually not much smaller than those induced by the physical differences typically used in discrimination-learning experiments. We suggest that findings of discrimination learning with identical stimuli can be explained without implicating any fundamentally new learning mechanism.

Learning to identify near-threshold luminance-defined and contrast-defined letters in observers with amblyopia

We assessed whether or not the sensitivity for identifying luminance-defined and contrast-defined letters improved with training in a group of amblyopic observers who have passed the critical period of development. In Experiment 1, we tracked the contrast threshold for identifying luminance-defined letters with training in a group of 11 amblyopic observers. Following training, six observers showed a reduction in thresholds, averaging 20%, for identifying luminance-defined letters. This improvement transferred extremely well to the untrained task of identifying contrast-defined letters (average improvement=38%) but did not transfer to an acuity measurement. Seven of the 11 observers were subsequently trained on identifying contrast-defined letters in Experiment 2. Following training, five of these seven observers demonstrated a further improvement, averaging 17%, for identifying contrast-defined letters. This improvement did not transfer to the untrained task of identifying luminance-defined letters. Our findings are consistent with predictions based on the locus of learning for first- and second-order stimuli according to the filter-rectifier-filter model for second-order visual processing.

Double dissociation between first- and second-order processing

To study the difference of sensitivity to luminance- (LM) and contrast-modulated (CM) stimuli, we compared LM and CM detection thresholds in LM- and CM-noise conditions. The results showed a double dissociation (no or little inter-attribute interaction) between the processing of these stimuli, which implies that both stimuli must be processed, at least at some point, by separate mechanisms and that both stimuli are not merged after a rectification process. A second experiment showed that the internal equivalent noise limiting the CM sensitivity was greater than the one limiting the carrier sensitivity, which suggests that the internal noise occurring before the rectification process is not limiting the CM sensitivity. These results support the hypothesis that a suboptimal rectification process partially explains the difference of LM and CM sensitivity.

Identification of contrast-defined letters benefits from perceptual learning in adults with amblyopia

Amblyopes show specific deficits in processing second-order spatial information (e.g. Wong, Levi, & McGraw (2001). Is second-order spatial loss in amblyopia explained by the loss of first-order spatial input? Vision Research, 41, 2951-2960). Recent work suggests there is a significant degree of plasticity in the visual pathway that processes first-order spatial information in adults with amblyopia. In this study, we asked whether or not there is similar plasticity in the ability to process second-order spatial information in adults with amblyopia. Ten adult observers with amblyopia (five strabismic, four anisometropic and one mixed) were trained to identify contrast-defined (second-order) letters using their amblyopic eyes. Before and after training, we determined observers' contrast thresholds for identifying luminance-defined (first-order) and contrast-defined letters, separately for the non-amblyopic and amblyopic eyes. Following training, eight of the 10 observers showed a significant reduction in contrast thresholds for identifying contrast-defined letters with the amblyopic eye. Five of these observers also showed a partial transfer of improvement to their fellow untrained non-amblyopic eye for identifying contrast-defined letters. There was a small but statistically significant transfer to the untrained task of identifying luminance-defined letters in the same trained eye. Similar to first-order spatial tasks, adults with amblyopia benefit from perceptual learning for identifying contrast-defined letters in their amblyopic eyes, suggesting a sizeable degree of plasticity in the visual pathway for processing second-order spatial information.

Perceptual learning of motion direction discrimination in fovea: separable mechanisms

Dosher and Lu (1998) [Perceptual learning reflects external noise filtering and internal noise reduction through channel reweighting. Proceedings of the National Academy of Sciences of the United States of America, 95 (23), 13988-13993.] proposed three mechanisms of perceptual learning: stimulus enhancement, external noise exclusion, and multiplicative noise reduction. In this study, we used pre-training as a manipulation to evaluate the separability of these mechanisms as a key test of the theoretical framework. Observers were trained in identifying the motion direction of a moving sine-wave grating in fovea with varying amount of superimposed external noise across trials, after receiving no pre-training, pre-training in high external noise, or pre-training in zero external noise in the same task. We found: (1) Without pre-training, perceptual learning significantly reduced contrast thresholds by about the same amount across all the external noise levels. (2) Both types of pre-training significantly reduced contrast thresholds in the corresponding conditions. (3) Pre-training in high external noise greatly reduced subsequent learning in high external noise, accounting for 64.6% of the total (pre-training + subsequent) improvements in that condition. On the other hand, the amount of subsequent learning in low external noise conditions was essentially the same as the total (pre-training + subsequent) amount of improvements in high external noise, suggesting that pre-training in high external noise had mostly only improved performance in noisy displays. (4) Pre-training in zero external noise practically eliminated or left very little additional learning in all the external noise conditions. We concluded that the two mechanisms of perceptual learning, stimulus enhancement, and external noise exclusion, can be trained independently in motion direction discrimination in fovea; training in low noise suffices to improve observer performance over all the external noise conditions.