Perceptual learning of bisection stimuli under roving: slow and largely specific

In perceptual learning, performance often improves within a short time if only one stimulus variant is presented, such as a line bisection stimulus with one outer-line-distance. However, performance stagnates if two bisection stimuli with two outer-line-distances are presented randomly interleaved. Recently, S. G. Kuai, J. Y. Zhang, S. A. Klein, D. M. Levi, and C. Yu, (2005) proposed that learning under roving conditions is impossible in general. Contrary to this proposition, we show here that perceptual learning with bisection stimuli under roving is possible with extensive training of 18000 trials. Despite this extensive training, the improvement of performance is still largely specific. Furthermore, this improvement of performance cannot be explained by an accommodation to stimulus uncertainty caused by roving.     

Interleaving bisection stimuli – randomly or in sequence – does not disrupt perceptual learning, it just makes it more difficult

Presenting stimuli of two or more stimulus types randomly interleaved, so called roving, disrupts perceptual learning in many paradigms. Recently, it was shown that no disruption occurs when Gabor stimuli were presented interleaved in sequence, instead of randomly. Here, using bisection stimuli, we found the opposite pattern of results. Presenting bisection stimuli in a sequence disrupted perceptual learning, whereas we found improvement under roving conditions. A meta-analysis showed that parts of this deviation from previous studies is possibly caused by the initial performance level of participants. These results do not prove previous results wrong, they just show that multiple factors play a crucial role in perceptual learning which cannot always be easily controlled for.     

Perceptual learning of motion discrimination by mental imagery

Perceptual learning can occur when stimuli are only imagined, i.e., without proper stimulus presentation. For example, perceptual learning improved bisection discrimination when only the two outer lines of the bisection stimulus were presented and the central line had to be imagined. Performance improved also with other static stimuli. In non-learning imagery experiments, imagining static stimuli is different from imagining motion stimuli. We hypothesized that those differences also affect imagery perceptual learning. Here, we show that imagery training also improves motion direction discrimination. Learning occurs when no stimulus at all is presented during training, whereas no learning occurs when only noise is presented. The interference between noise and mental imagery possibly hinders learning. For static bisection stimuli, the pattern is just the opposite. Learning occurs when presented with the two outer lines of the bisection stimulus, i.e., with only a part of the visual stimulus, while no learning occurs when no stimulus at all is presented.     

Perceptual learning,roving and the unsupervised bias

Perceptual learning improves perception through training. Perceptual learning improves with most stimulus types but fails when certain stimulus types are mixed during training (roving). This result is surprising because classical supervised and unsupervised neural network models can cope easily with roving conditions. What makes humans so inferior compared to these models? As experimental and conceptual work has shown, human perceptual learning is neither supervised nor unsupervised but reward-based learning. Reward-based learning suffers from the so-called unsupervised bias, i.e., to prevent synaptic “drift”, the average reward has to be exactly estimated. However, this is impossible when two or more stimulus types with different rewards are presented during training (and the reward is estimated by a running average). For this reason, we propose no learning occurs in roving conditions. However, roving hinders perceptual learning only for combinations of similar stimulus types but not for dissimilar ones. In this latter case, we propose that a critic can estimate the reward for each stimulus type separately. One implication of our analysis is that the critic cannot be located in the visual system.     

Reverse feedback induces position and orientation specific changes

To investigate the mechanisms of perceptual learning, we recently introduced a paradigm in which incorrect, reverse feedback followed after some but not all vernier presentations. This feedback paradigm exerted a strong effect on performance that seemed to bias decisions rather than to yield perceptual learning. Here, we show that observers can develop independent decision biases for different stimulus orientations as well as for different visual field positions. Our results demonstrate that the effects of incorrect, reverse feedback are surprisingly specific.     

Co-learning analysis of two perceptual learning tasks with identical input stimuli supports the reweighting hypothesis

Perceptual learning, even when it exhibits significant specificity to basic stimulus features such as retinal location or spatial frequency, may cause discrimination performance to improve either through enhancement of early sensory representations or through selective re-weighting of connections from the sensory representations to specific responses, or both. For most experiments in the literature, the two forms of plasticity make similar predictions (Dosher and Lu, 2009, Petrov et al., 2005). The strongest test of the two hypotheses must use training and transfer tasks that rely on the same sensory representation with different task-dependent decision structures. If training changes sensory representations, transfer (or interference) must occur since the (changed) sensory representations are common. If instead training re-weights a separate set of task connections to decision, then performance in the two tasks may still be independent. Here, we performed a co-learning analysis of two perceptual learning tasks based on identical input stimuli, following a very interesting study of Fahle and Morgan (1996) who used nearly identical input stimuli (a three dot pattern) in training bisection and vernier tasks. Two important modifications were made: (1) identical input stimuli were used in the two tasks, and (2) subjects practiced both tasks in multiple alternating blocks (800 trials/block). Two groups of subjects with counter-balanced order of training participated in the experiments. We found significant and independent learning of the two tasks. The pattern of results is consistent with the reweighting hypothesis of perceptual learning.     

A comparison of the effects of ageing upon vernier and bisection acuity

While most positional acuity tasks exhibit an age-related decline in performance, the effect of ageing upon vernier acuity continues to be the subject of some debate. In the present study we employed a stimulus design that enabled the simultaneous determination of bisection and vernier acuities in 36 subjects, aged between 22 and 84 years. This approach provided a means for directly testing the hypothesis that ageing affects bisection acuity but not vernier acuity by ensuring that differences in stimulus configuration and in the subject's task were kept to an absolute minimum. Optimum thresholds increased as a function of age for both bisection and vernier tasks. Inter-subject threshold variability also increased with age. Issues surrounding the comparison of absolute vernier thresholds across different studies are discussed and two important methodological factors are identified: the precise statistical method used to estimate thresholds, and the magnitude, in angular terms, of the smallest spatial offset of the elements of the vernier stimulus which can be displayed. Comparison with previously published data indicates that the discrepancy between this study and most previous investigations with respect to the effect of age upon vernier performance can be at least partly accounted for by differences in the minimum displayable vernier offset. Vernier thresholds do increase with age. The increased variability of vernier thresholds in older subjects would appear to limit the diagnostic value of the test as a means of enabling normal ageing to be distinguished from visual loss due to pathology of the eye or visual system.     

Learning of interpolation in 2 and 3 dimensions.

We investigated learning of spatio-temporal interpolation in depth and its relation to spatio-temporal interpolation in two dimensions by means of a vernier discrimination task. Performance improved with training but improvement did not or only partially transfer between opposite directions of motion in depth. Improvement was also at least partly specific for the eye and for the direction of two-dimensional motion used during monocular training. This specificity might explain the apparent specificity of interpolation in three dimensions. Training with stimuli moving in two dimensions increased performance for a stimulus moving in depth. The results indicate that improvement in spatio-temporal interpolation occurs at least partly on a stage preceding stereoscopic vision, and are a rare example for transfer of improvement between different perceptual tasks.     

Perceptual learning with spatial uncertainties

In perceptual learning, stimuli are usually assumed to be presented to a constant retinal location during training. However, due to tremor, drift, and microsaccades of the eyes, the same stimulus covers different retinal positions on sequential trials. Because of these variations the mathematical decision problem changes from linear to non-linear (). This non-linearity implies three predictions. First, varying the spatial position of a stimulus within a moderate range does not deteriorate perceptual learning. Second, improvement for one stimulus variant can yield negative transfer to other variants. Third, interleaved training with two stimulus variants yields no or strongly diminished learning. Using a bisection task, we found psychophysical evidence for the first and last prediction. However, no negative transfer was found as opposed to the second prediction.     

Quantifying target conspicuity in contextual modulation by visual search

Contextual elements can strongly modulate visual performance. For example, performance deteriorates when a vernier is flanked by neighboring lines. On a neural level, such contextual modulation is often explained by local spatial interactions such as lateral inhibition or pooling. However, these mechanisms cannot account for a number of recent results which showed that global rather than local factors play a key role in contextual modulation. On a level of perceptual organization, we proposed that contextual modulation increases when the target groups with the flankers and decreases when the target stands out from the flankers. To quantify this "standing out" in foveal vision, here, we performed both a visual search and a vernier offset discrimination task on the same stimulus configurations. Stimulus configurations yielding short reaction times in visual search yielded good vernier discrimination performance. Stimulus configurations yielding long reaction times yielded weaker discrimination. Hence, vernier offset discrimination is superior for targets that are efficiently searched and vice versa.     

Perceptual learning in contrast discrimination: the effect of contrast uncertainty

Performance in perceptual tasks improves with repetition (perceptual learning), eventually reaching a saturation level. Typically, when perceptual learning effects are studied, stimulus parameters are kept constant throughout the training and during the pre- and post-training tests. Here we investigate whether learning by repetition transfers to testing conditions in which the practiced stimuli are randomly interleaved during the post-training session. We studied practice effects with a contrast discrimination task, employing a number of training methods: (i) practice with a single, fixed pedestal (base-contrast), (ii) practice with several pedestals, and (iii) practice with several pedestals that included a spatial context. Pre- and post-training tests were carried out with the base contrast randomized across trials, under conditions of contrast uncertainty. The results showed that learning had taken place with the fixed pedestal method (i) and with the context method (iii), but only the latter survived the uncertainty test. In addition, we were able to identify a very fast learning phase in contrast discrimination that improved performance under uncertainty. We contend that learned tasks that do not pass the uncertainty test involve modification of decision strategies that require exact knowledge of the stimulus.     

Isolation of stimulus characteristics contributing to Weber's law for position

To examine the independent contribution of various stimulus characteristics to positional judgements, we measured vernier alignment performance for three types of Gabor stimuli. In one, only the contrast envelope of the upper and lower stimulus elements was offset, with the luminance-modulated carrier grating remaining in alignment. In the second, only the carrier grating was offset. In the third, both carrier and envelope were offset together. Performance was examined over a range of element separations. When both cues are available, thresholds for small separations are dominated by carrier offset information and are inversely proportional to carrier frequency. At large separations, thresholds are governed by the spatial scale characteristics of the envelope. For broad-band stimuli such as lines, bars or dots typically used for vernier acuity, their higher frequency content can be used when separations are small, but as separation increases a smooth transition between the scales that determine threshold results in the continuum known as Weber's law for position. That is, with increasing separation, larger scales must be used, and thresholds increase in direct proportion to 1/frequency.     

Mechanisms of generalization in perceptual learning

Learning in many visual perceptual tasks has been shown to be specific to practiced stimuli, while new stimuli have to be learned from scratch. Here we demonstrate generalization using a novel paradigm in motion discrimination where learning has been previously shown to be specific. We trained subjects to discriminate directions of moving dots, and verified the previous results that learning does not transfer from a trained direction to a new one. However, by tracking the subjects' performance across time in the new direction, we found that their speed of learning doubled. Therefore, we found generalization in a task previously considered too difficult to generalize. We also replicated, in a second experiment, transfer following training with 'easy' stimuli, when the difference between motion directions is enlarged. In a third experiment we found a new mode of generalization: after mastering the task with an easy stimulus, subjects who have practiced briefly to discriminate the easy stimulus in a new direction generalize to a difficult stimulus in that direction. This generalization depends on both the mastering and the brief practice. The specificity of perceptual learning and the dichotomy between learning of 'easy' versus 'difficult' tasks have been assumed to involve different learning processes at different cortical areas. Here we show how to interpret these results in terms of signal detection theory. With the assumption of limited computational capacity, we obtain the observed phenomena--direct transfer and acceleration of learning--for increasing levels of task difficulty. Human perceptual learning and generalization, therefore, concur with a generic discrimination system.     

Hastening orientation sensitivity

Previous perceptual learning studies have shown that sensitivity to subtle orientation differences improves with practice at oblique axes but not with practice at cardinal axes. The cause of this anisotropy in angular resolution is uncertain, and it is not known whether the same anisotropy pertains to temporal resolution-the minimum stimulus duration needed to achieve a specified angular resolution. Here, we investigated the hypothesis that cardinal improvements were previously absent because long stimulus durations yielded maximal precision, even at the start of training. Accordingly, we exploited the relatively imprecise responses that occur naturally when masked stimuli are presented for extremely brief durations. After 110,000 trials were completed over seven daily sessions, temporal resolution improved by 51% at cardinal axes and by 86% at oblique axes. This hastening of the visual response was accompanied by significant improvements in angular resolution, which were specific to the trained axis. The data demonstrate plasticity in the response to cardinal orientations and indicate that sufficient initial levels of neural imprecision may be necessary for perceptual learning.     

Vernier in Motion: What Accounts for the Threshold Elevation?

Vernier acuity is susceptible to degradation by image motion. The purpose of this study was to determine to what extent vernier thresholds are elevated in the presence of image motion because of reduced stimulus visibility, due to contrast smearing, or to a shift in the spatial scale of analysis. To test the visibility hypothesis, we measured vernier thresholds as a function of stimulus velocity (0–6 deg/sec), for various levels of stimulus visibility, each normalized to the detection threshold at the respective velocity. Contrary to the prediction of the visibility hypothesis, vernier thresholds worsen as the velocity increases, even when the stimuli are equally visible. To test the shift in spatial scale hypothesis, we determined spatial frequency tuning functions for vernier discrimination and line detection tasks, using a masking paradigm. We measured vernier and line detection thresholds as a function of spatial frequency of a sine-wave mask (0.5-32 c/deg), and for stimulus and mask velocities ranging from 0 to 4 deg/sec. Peak masking for both vernier discrimination and line detection, which indicates the most sensitive band of spatial frequencies for each task, shifts systematically toward lower spatial frequencies as the velocity increases. The progressive increase in spatial scale largely accounts for the worsening of vernier thresholds for moving stimuli. Differences between peak masking for vernier discrimination and line detection were found at 0 and 1 deg/sec, suggesting that different mechanisms mediate the two tasks, at least at low velocities. The masking results are consistent with previous findings that directionally selective motion detectors mediate detection of moving stimuli, but suggest that these detectors do not analyze vernier offsets. We conclude that the elevation of vernier threshold for a moving stimulus is accounted for primarily by a shift of sensitivity to mechanisms of lower spatial frequency, and not by decreased stimulus visibility. Copyright © 1996 Elsevier Science Ltd.     

Systematic misestimation in a vernier task arising from contrast mismatch

Luminance signals mediated by the magnocellular (MC) pathway play an important role in vernier tasks. MC ganglion cells show a phase advance in their responses to sinusoidal stimuli with increasing contrast due to contrast gain control mechanisms. If the phase information in MC ganglion cell responses were utilized by central mechanisms in vernier tasks, one might expect systematic errors caused by the phase advance. This systematic error may contribute to the contrast paradox phenomenon, where vernier performance deteriorates, rather than improves, when only one of the target pair increases in contrast. Vernier psychometric functions for a pair of gratings of mismatched contrast were measured to seek such misestimation. In associated electrophysiological experiments, MC and parvocellular (PC) ganglion cells' responses to similar stimuli were measured to provide a physiological reference. The psychophysical experiments show that a high-contrast grating is perceived as phase advanced in the drift direction compared to a low-contrast grating, especially at a high drift rate (8 Hz). The size of the phase advance was comparable to that seen in MC cells under similar stimulus conditions. These results are consistent with the MC pathway supporting vernier performance with achromatic gratings. The shifts in vernier psychometric functions were negligible for pairs of chromatic gratings under the conditions tested here, consistent with the lack of phase advance both in responses of PC ganglion cells and in frequency-doubled chromatic responses of MC ganglion cells.     

Stimulus uncertainty in auditory perceptual learning

Stimulus uncertainty produced by variations in a target stimulus to be detected or discriminated, impedes perceptual learning under some, but not all experimental conditions. To account for those discrepancies, it has been proposed that uncertainty is detrimental to learning when the interleaved stimuli or tasks are similar to each other but not when they are sufficiently distinct, or when it obstructs the downstream search required to gain access to fine-grained sensory information, as suggested by the Reverse Hierarchy Theory (RHT). The focus of the current review is on the effects of uncertainty on the perceptual learning of speech and non-speech auditory signals. Taken together, the findings from the auditory modality suggest that in addition to the accounts already described, uncertainty may contribute to learning when categorization of stimuli to phonological or acoustic categories is involved. Therefore, it appears that the differences reported between the learning of non-speech and speech-related parameters are not an outcome of inherent differences between those two domains, but rather due to the nature of the tasks often associated with those different stimuli.     

Configuration specificity in bisection acuity

Crucial for the perception of form are the spatial relationships between the elements of a visual stimulus. To investigate the mechanisms involved in coding the distance between visual stimuli, thresholds for detecting whether a central marker accurately bisects a spatial interval were compared for a variety of configurations. Thresholds are best when all three members of the bisection configuration are identical. Performance is impaired, often by as much as a factor of two, when the outer delimiters of the spatial interval differ from the central marker in either length, orientation or contrast polarity. Illusory contours act poorly as borders for bisection by a central line. Disparity thresholds are not affected by orientation differences between test and flanking lines. Because in peripheral vision bisection acuity improves with practice, transfer of training between configurations can be used to gauge overlap of neural processing mechanisms. Transfer is complete only between patterns where all markers are similar, reduced when the outer markers differ by 20 degrees in orientation and absent when they are orthogonal. The dependence of bisection discrimination on similarity between the elements of the stimulus demonstrates that the encoding of spatial location and spatial extent are coupled to the coding of other stimulus properties.     

Selectivity of the evoked potential for vernier offset

Hyperacuity thresholds of a few arc seconds can be achieved psychophysically for a variety of spatial localization tasks. The present experiments show that evoked potentials can be elicited in response to the introduction of vernier offsets, but not by the introduction of other cues to hyperacuity such as bisection or relative pattern motion, although each of these cues is equally salient psychophysically. Moreover, vernier acuity measurements and the evoked potentials elicited in response to vernier offsets are strongly degraded by the introduction of flanking stimuli 2-4 min from the vernier target. This suggests that the hyperacuity VEP is a cortical correlate of a very specific type of hyperacuity, that produced by vernier offsets (colinearity failure).     

Preexposure disrupts learning of location-contingent perceptual biases for ambiguous stimuli

The perception of a bistable stimulus as one or the other interpretation can be biased by prior presentations of that stimulus. Such learning effects have been found to be long lasting even after small amounts of training. The effectiveness of training may be influenced by preexposure to the ambiguous stimulus. Here we investigate the role of preexposure for learning a position-dependent perceptual bias. We used rotating Necker Cubes as the bistable stimuli, which were presented at two locations: above or below fixation. On training trials, additional depth cues disambiguated the rotation direction contingent on the location. On test trials, the rotating cube was presented without disambiguation cues. Without preexposure to the ambiguous stimulus, subjects learned to perceive the cube to be rotating in the trained direction for both locations. However, subjects that were preexposed to the ambiguous stimulus did not learn the trained percept-location contingency, even though the preexposure was very short compared to the subsequent training. Preexposure to the disambiguated stimulus did not interfere with learning. This indicates a fundamental difference between ambiguous test and disambiguated training trials for learning a perceptual bias. In short, small variations in paradigm can have huge effects for the learning of perceptual biases for ambiguous stimuli.