Augmented Hebbian reweighting accounts for accuracy and induced bias in perceptual learning with reverse feedback

Transfer of visual perceptual learning over a task-irrelevant feature through feature-invariant representations: Behavioral experiments and model simulations

A large body of literature has examined specificity and transfer of perceptual learning, suggesting a complex picture. Here, we distinguish between transfer over variations in a "task-relevant" feature (e.g., transfer of a learned orientation task to a different reference orientation) and transfer over a "task-irrelevant" feature (e.g., transfer of a learned orientation task to a different retinal location or different spatial frequency), and we focus on the mechanism for the latter. Experimentally, we assessed whether learning a judgment of one feature (such as orientation) using one value of an irrelevant feature (e.g., spatial frequency) transfers to another value of the irrelevant feature. Experiment 1 examined whether learning in eight-alternative orientation identification with one or multiple spatial frequencies transfers to stimuli at five different spatial frequencies. Experiment 2 paralleled Experiment 1, examining whether learning in eight-alternative spatial-frequency identification at one or multiple orientations transfers to stimuli with five different orientations. Training the orientation task with a single spatial frequency transferred widely to all other spatial frequencies, with a tendency to specificity when training with the highest spatial frequency. Training the spatial frequency task fully transferred across all orientations. Computationally, we extended the identification integrated reweighting theory (I-IRT) to account for the transfer data (Dosher, Liu, & Lu, 2023; Liu, Dosher, & Lu, 2023). Just as location-invariant representations in the original IRT explain transfer over retinal locations, incorporating feature-invariant representations effectively accounted for the observed transfer. Taken together, we suggest that feature-invariant representations can account for transfer of learning over a "task-irrelevant" feature.

Asymmetric stimulus representations bias visual perceptual learning

The primate visual cortex contains various regions that exhibit specialization for different stimulus properties, such as motion, shape, and color. Within each region, there is often further specialization, such that particular stimulus features, such as horizontal and vertical orientations, are over-represented. These asymmetries are associated with well-known perceptual biases, but little is known about how they influence visual learning. Most theories would predict that learning is optimal, in the sense that it is unaffected by these asymmetries. However, other approaches to learning would result in specific patterns of perceptual biases. To distinguish between these possibilities, we trained human observers to discriminate between expanding and contracting motion patterns, which have a highly asymmetrical representation in the visual cortex. Observers exhibited biased percepts of these stimuli, and these biases were affected by training in ways that were often suboptimal. We simulated different neural network models and found that a learning rule that involved only adjustments to decision criteria, rather than connection weights, could account for our data. These results suggest that cortical asymmetries influence visual perception and that human observers often rely on suboptimal strategies for learning.

Informational feedback accelerates learning in multi-alternative perceptual judgements of orientation

Experience or training can substantially improve perceptual performance through perceptual learning, and the extent and rate of these improvements may be affected by feedback. In this paper, we first developed a neural network model based on the integrated reweighting theory (Dosher et al., 2013) to account for perceptual learning and performance in n-alternative identification tasks and the dependence of learning on different forms of feedback. We then report an experiment comparing the effectiveness of response feedback (RF) versus accuracy feedback (AF) or no feedback (NF) (full versus partial versus no supervision) in learning a challenging eight-alternative visual orientation identification (8AFC) task. Although learning sometimes occurred in the absence of feedback (NF), RF had a clear advantage above AF or NF in this task. Using hybrid supervision learning rules, a new n-alternative identification integrated reweighting theory (I-IRT) explained both the differences in learning curves given different feedback and the dynamic changes in identification confusion data. This study shows that training with more informational feedback (RF) is more effective, though not necessary, in these challenging n-alternative tasks, a result that has implications for developing training paradigms in realistic tasks.

Only cortical prediction error signals are involved in visual learning, despite availability of subcortical prediction error signals

Both the midbrain systems, encompassing the ventral striatum (VS), and the cortical systems, including the dorsal anterior cingulate cortex (dACC), play roles in reinforcing and enhancing learning. However, the specific contributions of signals from these regions in learning remains unclear. To investigate this, we examined how VS and dACC are involved in visual perceptual learning (VPL) through an orientation discrimination task. In the primary experiment, subjects fasted for 5 hours before each of 14 days of training sessions and 3 days of test sessions. Subjects were rewarded with water for accurate trial responses. During the test sessions, BOLD signals were recorded from regions including VS and dACC. Although BOLD signals in both areas were associated with positive and negative RPEs, only those in dACC associated with negative RPE showed a significant correlation with performance improvement. Additionally, no significant correlation was observed between BOLD signals associated with RPEs in VS and dACC. These results suggest that although signals associated with positive and negative RPEs from both midbrain and cortical systems are readily accessible, only RPE signals in the prefrontal system, generated without linking to RPE signals in VS, are utilized for the enhancement of VPL.

Perceptual learning across saccades: Feature but not location specific

Perceptual learning is the ability to enhance perception through practice. The hallmark of perceptual learning is its specificity for the trained location and stimulus features, such as orientation. For example, training in discriminating a grating's orientation improves performance only at the trained location but not in other untrained locations. Perceptual learning has mostly been studied using stimuli presented briefly while observers maintained gaze at one location. However, in everyday life, stimuli are actively explored through eye movements, which results in successive projections of the same stimulus at different retinal locations. Here, we studied perceptual learning of orientation discrimination across saccades. Observers were trained to saccade to a peripheral grating and to discriminate its orientation change that occurred during the saccade. The results showed that training led to transsaccadic perceptual learning (TPL) and performance improvements which did not generalize to an untrained orientation. Remarkably, however, for the trained orientation, we found a complete transfer of TPL to the untrained location in the opposite hemifield suggesting high flexibility of reference frame encoding in TPL. Three control experiments in which participants were trained without saccades did not show such transfer, confirming that the location transfer was contingent upon eye movements. Moreover, performance at the trained location, but not at the untrained location, was also improved in an untrained fixation task. Our results suggest that TPL has both, a location-specific component that occurs before the eye movement and a saccade-related component that involves location generalization.

Learning spatial frequency identification through reweighted decoding

Perceptual learning, the improvement of perceptual judgments with practice, occurs in many visual tasks. There are, however, relatively fewer studies examining perceptual learning in spatial frequency judgments. In addition, perceptual learning has generally been studied in two-alternative tasks, occasionally in n-alternative tasks, and infrequently in identification. Recently, perceptual learning was found in an orientation identification task (eight-alternatives) and was well accounted for by a new identification integrated reweighting theory (I-IRT) (Liu et al., submitted). Here, we examined perceptual learning in a similar eight-alternative spatial frequency absolute identification task in two different training protocols, finding learning in the majority but not all observers. We fit the I-IRT to the spatial frequency learning data and discuss possible model explanations for variations in learning.

Roving: The causes of interference and re-enabled learning in multi-task visual training

People routinely perform multiple visual judgments in the real world, yet, intermixing tasks or task variants during training can damage or even prevent learning. This paper explores why. We challenged theories of visual perceptual learning focused on plastic retuning of low-level retinotopic cortical representations by placing different task variants in different retinal locations, and tested theories of perceptual learning through reweighting (changes in readout) by varying task similarity. Discriminating different (but equivalent) and similar orientations in separate retinal locations interfered with learning, whereas training either with identical orientations or sufficiently different ones in different locations released rapid learning. This location crosstalk during learning renders it unlikely that the primary substrate of learning is retuning in early retinotopic visual areas; instead, learning likely involves reweighting from location-independent representations to a decision. We developed an Integrated Reweighting Theory (IRT), which has both V1-like location-specific representations and higher level (V4/IT or higher) location-invariant representations, and learns via reweighting the readout to decision, to predict the order of learning rates in different conditions. This model with suitable parameters successfully fit the behavioral data, as well as some microstructure of learning performance in a new trial-by-trial analysis.

Asymmetric perceptual confounds between canonical lightings and materials

To better understand the interactions between material perception and light perception, we further developed our material probe MatMix 1.0 into MixIM 1.0, which allows optical mixing of canonical lighting modes. We selected three canonical lighting modes (ambient, focus, and brilliance) and created scenes to represent the three illuminations. Together with four canonical material modes (matte, velvety, specular, glittery), this resulted in 12 basis images (the "bird set"). These images were optically mixed in our probing method. Three experiments were conducted with different groups of observers. In Experiment 1, observers were instructed to manipulate MixIM 1.0 and match optically mixed lighting modes while discounting the materials. In Experiment 2, observers were shown a pair of stimuli and instructed to simultaneously judge whether the materials and lightings were the same or different in a four-category discrimination task. In Experiment 3, observers performed both the matching and discrimination tasks in which only the ambient and focus light were implemented. Overall, the matching and discrimination results were comparable as (a) robust asymmetric perceptual confounds were found and confirmed in both types of tasks, (b) performances were consistent and all above chance levels, and (c) observers had higher sensitivities to our canonical materials than to our canonical lightings. The latter result may be explained in terms of a generic insensitivity for naturally occurring variations in light conditions. Our findings suggest that midlevel image features are more robust across different materials than across different lightings and, thus, more diagnostic for materials than for lightings, causing the asymmetric perceptual confounds.

Regulation of cholinergic basal forebrain development, connectivity, and function by neurotrophin receptors

Cholinergic basal forebrain (cBF) neurons are defined by their expression of the p75 neurotrophin receptor (p75NTR) and tropomyosin-related kinase (Trk) neurotrophin receptors in addition to cholinergic markers. It is known that the neurotrophins, particularly nerve growth factor (NGF), mediate cholinergic neuronal development and maintenance. However, the role of neurotrophin signalling in regulating adult cBF function is less clear, although in dementia, trophic signalling is reduced and p75NTR mediates neurodegeneration of cBF neurons. Here we review the current understanding of how cBF neurons are regulated by neurotrophins which activate p75NTR and TrkA, B or C to influence the critical role that these neurons play in normal cortical function, particularly higher order cognition. Specifically, we describe the current evidence that neurotrophins regulate the development of basal forebrain neurons and their role in maintaining and modifying mature basal forebrain synaptic and cortical microcircuit connectivity. Understanding the role neurotrophin signalling plays in regulating the precision of cholinergic connectivity will contribute to the understanding of normal cognitive processes and will likely provide additional ideas for designing improved therapies for the treatment of neurological disease in which cholinergic dysfunction has been demonstrated.

High reward enhances perceptual learning

Studies of perceptual learning have revealed a great deal of plasticity in adult humans. In this study, we systematically investigated the effects and mechanisms of several forms (trial-by-trial, block, and session rewards) and levels (no, low, high, subliminal) of monetary reward on the rate, magnitude, and generalizability of perceptual learning. We found that high monetary reward can greatly promote the rate and boost the magnitude of learning and enhance performance in untrained spatial frequencies and eye without changing interocular, interlocation, and interdirection transfer indices. High reward per se made unique contributions to the enhanced learning through improved internal noise reduction. Furthermore, the effects of high reward on perceptual learning occurred in a range of perceptual tasks. The results may have major implications for the understanding of the nature of the learning rule in perceptual learning and for the use of reward to enhance perceptual learning in practical applications.

A Computational Model of Perceptual and Mnemonic Deficits in Medial Temporal Lobe Amnesia

Damage to the medial temporal lobe (MTL) has long been known to impair declarative memory, and recent evidence suggests that it also impairs visual perception. A theory termed the representational-hierarchical account explains such impairments by assuming that MTL stores conjunctive representations of items and events, and that individuals with MTL damage must rely upon representations of simple visual features in posterior visual cortex, which are inadequate to support memory and perception under certain circumstances. One recent study of visual discrimination behavior revealed a surprising antiperceptual learning effect in MTL-damaged individuals: With exposure to a set of visual stimuli, discrimination performance worsened rather than improved [Barense, M. D., Groen, I. I. A., Lee, A. C. H., Yeung, L. K., Brady, S. M., Gregori, M., et al. Intact memory for irrelevant information impairs perception in amnesia. Neuron, 75, 157-167, 2012]. We extend the representational-hierarchical account to explain this paradox by assuming that difficult visual discriminations are performed by comparing the relative "representational tunedness"-or familiarity-of the to-be-discriminated items. Exposure to a set of highly similar stimuli entails repeated presentation of simple visual features, eventually rendering all feature representations maximally and, thus, equally familiar; hence, they are inutile for solving the task. Discrimination performance in patients with MTL lesions is therefore impaired by stimulus exposure. Because the unique conjunctions represented in MTL do not occur repeatedly, healthy individuals are shielded from this perceptual interference. We simulate this mechanism with a neural network previously used to explain recognition memory, thereby providing a model that accounts for both mnemonic and perceptual deficits caused by MTL damage with a unified architecture and mechanism.