Plasticity and Learning

Talk Session: Tuesday, May 20, 2025, 2:45 – 4:45 pm, Talk Room 1

Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions

Talk 1, 2:45 pm

Homotopic reorganization of the visual word form area following surgical resection of connected cortex

Beth Rispoli¹ (), Tina Liu¹, Kyungji Moon¹, Radhika Chatterjee², Kareem Zaghloul², Sara Inati²; ¹Georgetown University Medical Center, ²National Institute of Neurological Disorders and Stroke

The visual word form area (VWFA), typically located in the left ventral temporal cortex (VTC), is crucial for word reading due to its privileged connectivity to higher-order visual, semantic, and language areas. While acute lesions involving the VWFA in adulthood can result in reading impairments (e.g., alexia), the impact of lesions to connected cortex on VWFA function remains unclear. In this study, we ask whether a resection of anterior temporal lobe (ATL) impacts reading selectivity in the VTC and how outcomes may vary by hemisphere of resection. We analyzed a clinical dataset of adult patients (n=9) who underwent surgical resection of ATL as a treatment for drug-resistant epilepsy. Patients completed an fMRI reading task both pre-operatively and one year post-operatively. During the reading task, patients read on-screen stories and viewed a matched control condition. VTC anatomical ROIs were drawn in the native space of each patient’s brain for each hemisphere, and mean responses for the reading and control conditions were computed. Reading selectivity was calculated in each ROI as the difference in beta weights between the reading and control conditions for each session. We found that the hemisphere of resection was predictive of a change in reading selectivity after surgery in the intact, homotopic VTC. Following left ATL resection (n=5), all patients showed a consistent increase in reading selectivity in the intact right VTC. Conversely, patients with right ATL resection (n=4) did not show this pattern, likely because language and reading functions remained lateralized to the left hemisphere. Together, these findings highlight the plasticity of reading-related circuitry in adulthood, showing that disruption of left hemisphere VWFA connectivity via ATL resection can promote compensatory reliance on the intact right VWFA. Importantly, this homotopic reorganization of the VWFA appears to be asymmetric, as no comparable reorganization occurs following right hemisphere ATL resection.

Talk 2, 3:00 pm

Short-term monocular deprivation in adult humans: a meta-analysis and new perspectives

Cecilia Steinwurzel¹ (), Giacomo Pennella², Claudia Lunghi³, Paola Binda¹; ¹University of Pisa, ²University of Florence, ³Ecole Normale Supérieure & CNRS, Paris

Few hours of monocular deprivation in adult humans produces a transient shift of ocular dominance in favor of the deprived eye. This phenomenon has been investigated with several methodologies in 75 studies since 2011. We compiled a meta-analysis of these studies, structured following the PRISMA checklist and selectively including studies on healthy humans. Each study includes multiple experiments (for a total of about 180); for each, we computed a standardized effect size and, where possible, we quantified the decay rate of the effect. In the majority of studies, deprivation was achieved by covering one eye with an opaque or a translucent patch, with comparable outcomes (two-sample t(108) = -0.02, p = 0.98). Deprivation effects were mainly measured with three types of tasks: binocular cooperation, binocular competition, and monocular thresholds. Effects are larger when measured with binocular cooperation compared to binocular competition (two-sample t(100) = 4.66, p = 0.001), possibly reflecting the different time windows used to estimate effects. Longer periods of deprivation induce larger (r(107) = 0.32, p = 0.001) and longer lasting effects (r(38) = 0.38, p = 0.020). Patching either eye (dominant or non-dominant) produces a similar outcome (two-sample t(98) = -0.74, p = 0.46). Depriving one eye produces opposite effects in the two eyes; however, pooling across studies, we find that opaque patching primarily suppresses the non-deprived eye, while translucent patching primarily boosts the deprived eye (two-sample t(31) = 2.99, p = 0.005). A growing number of studies show that monocular deprivation effects can be mimicked by manipulations that do not impact the strength of the monocular signal, including degradation of image quality (pink noise, kaleidoscope), image inversion in space or time, or just monocular delay. We conclude that the observed shift in eye dominance results from a complex interplay between bottom-up visual stimulation and top-down signals

Research funded by: - grant PREDACTIVE (n. 101170249) - grant (No 948366-HOPLA) - grant (THE, CUP I53C22000780001), - grant PRIN 2022 (Project RIGHTSTRESS, Grant no. 2022CCPJ3J, CUP I53D23003960006) - grant SMILY (Italian Ministry of University and Research - FARE-2).

Talk 3, 3:15 pm

Holistic Bayesian model of perceptual adaptation that combines efficient coding and novelty detection

Jiang Mao¹, Alan Stocker¹; ¹University of Pennsylvania

Perceptual systems continually adapt to statistical changes of their sensory environment. This leads to changes in perception (i.e. aftereffects) that have been well characterized. The exact functional purpose of adaptation, however, remains unclear. The efficient coding hypothesis is one prominent account of adaptation. It proposes that neural representations adapt to changes in the input statistics in order to maximize encoding efficiency given overall resource limitations. Another account for adaptation is that it enhances the detection of novel stimuli. Neural repetition suppression as well as increased discriminability at the adaptor indicate that new stimuli that differ from the adaptor are more salient. Here we present a model of adaptation that incorporates both of these rationales within the framework of holistic Bayesian inference. The model assumes efficient sensory encoding that adapts to the distribution of future stimuli predicted by previous input statistics. Specifically, we constrain encoding accuracy with previously characterized, efficient encoding changes caused by prolonged adaptation to a single adaptor stimulus (Mao et al. 2024). The model also incorporates novelty detection by assuming that the observer performs categorical inference to determine whether the next stimulus originates from a distribution different from the distribution of the adaptor (hence is novel) or not. We show that this model can account for many of the previously reported perceptual aftereffects measured with estimation tasks (method of adjustment). It accurately explains the characteristic repulsive biases and changes in estimation variance after adaptation to a single adaptor. Furthermore, the model predicts aftereffects for arbitrary adaptors, in particular adaptors that consist of a distribution of feature values. In conclusion, we present a holistic inference model for adaptation that unites two major functional objectives of sensory adaptation, efficient coding and novelty detection.

Talk 4, 3:30 pm

Trial-level pupillary response reveals that neural gain is robustly linked to perceptual plasticity and explains the role of feedback in learning

Aaron Cochrane¹ (), Soph Bililies¹, Shreya Gulati¹, Kiley Haberkorn¹, Yuka Sasaki¹, Takeo Watanabe¹; ¹Brown University

Visual perceptual learning arises from neuroplasticity leading to improved low-level inferences about environmental signals. This experience-dependent neural change can be influenced by various factors, including explicit and informative feedback, which is thought to boost neuroplastic changes. If feedback works through such a signal boosting mechanism to improve learning, global measures of “neural gain” (e.g., pupillary dilation in response to a stimulus) should mediate the effects of feedback on perceptual learning. Here, we examined the detailed time course of perceptual improvements during five days of training on discrimination of low-coherence dot-motion under three conditions involving explicit feedback: no feedback, intermittent feedback, and first-day-only feedback. Performance improved significantly more for the first-day-only feedback group compared to the no-feedback group, even days after feedback was removed. This finding suggests a lasting effect in which initial feedback enabled participants to learn how to generate endogenous feedback signals; such signals may be revealed by measures of neural gain. To identify how the dynamics of perceptual change are linked to neural gain, we measured task-evoked pupillary response during each trial. Using local derivative analyses, we linked trial-based pupillary responses to changes in perceptual threshold (i.e., first derivative). Two striking effects emerged: First, differential pupillary responses to trials of varying difficulty (referred to as “adaptive neural gain”) were reliably associated with the trial-to-trial changes in perceptual threshold. Specifically, trials with greater adaptive neural gain occurred during periods of larger decreases in threshold, providing evidence of a single-trial mechanism for perceptual plasticity. Second, the presence of feedback enhanced this association, demonstrating that feedback strengthens the learner’s ability to dynamically regulate neural gain which leads to plasticity. These results provide evidence for the role of within-trial perceptual signal boosting and reveal a mechanism of neural gain through which feedback improves perceptual learning.

NIH R01EY019466, R01EY027841, R01EY031705, NSF-BSF BCS2241417

Talk 5, 3:45 pm

Brief memory reactivations enable generalization of perceptual learning

Taly Kondat^1,2, Yuka Sasaki³, Takeo Watanabe³, Nitzan Censor^1,2; ¹Sagol School of Neuroscience, Tel Aviv University, ²School of Psychological Sciences, Tel Aviv University, ³Brown University, Providence, USA

Perceptual learning can significantly improve human visual sensitivity. However, it requires extensive stimuli exposure, and the ability to generalize learning to untrained conditions is often limited. Traditionally, perceptual learning is attributed to practice-dependent plasticity mechanisms. Nevertheless, recent studies suggest that brief memory reactivations can efficiently improve visual perception, exhibiting comparable gains through distinct mechanisms which recruit higher-level parietal brain regions. Here, we provide evidence that such memory reactivation mechanisms promote generalization of learning. Human participants first encoded a standard texture discrimination task with the target stimulus at retinotopic location A. Brief memory reactivations of only five trials each were performed on separate days at location A. Then, generalization was tested at retinotopic location B. Results indicate remarkable enhancement of location B performance following memory reactivations, pointing to efficient offline generalization mechanisms. A control experiment with no reactivations showed minimal generalization. These findings suggest that beyond dramatically reducing stimuli exposure, reactivation-induced learning further enhances learning efficiency by promoting generalization, and may extend to additional learning domains, with potential future clinical implications.

The study was supported by the European Research Council (ERC-2019-COG 866093) and the NSF-BSF grant (National Science Foundation 2241417, Binational Science Foundation 2022666). Taly Kondat is a fellow of the Ariane de Rothschild Women Doctoral Program.

Talk 6, 4:00 pm

The rostrolateral prefrontal cortex is activated during a rule-based visual categorization tasks

Jacob Vaughn¹ (), Shinyoung Jung¹, Sean R. O'Bryan¹, Tyler Davis¹, Miranda Scolari¹; ¹Texas Tech University

Categorization is a fundamental cognitive process that allows us to make sense of the external world by extrapolating from previous visual experiences to better understand novel ones. One under-investigated aspect of categorization is how this extrapolation is neurologically supported; however, prior research indicates that humans use inferential processes, such as hypothesis testing and logical rules, when confronted with novel visual stimuli. One area that is hypothesized to instantiate these inferential processes is the rostrolateral prefrontal cortex (rlPFC). To investigate this, we conducted a visual rule-based categorization tasks while participants were in a functional magnetic resonance imaging (fMRI) machine. During the first phase of the experiment, participants (N = 50) learned the task rule via trial-by-trial feedback. During the second phase, participants were tasked with categorizing stimuli without feedback that contained novel visual features or feature combinations that were not presented during learning. This generalization phase consisted of six types of trials: three of which were ambiguous with respect to the learned rules and three were unambiguous. Our results showed that the rlPFC was significantly activated during the learning phase, and only during the ambiguous trials in the generalization phase. In contrast, primary visual cortex and MTL were activated during all trials of the generalization phase. Our results suggest that the rlPFC may be critical for early learning (times of uncertainty) and then, once a rule is learned, only activates again in situations where learned rules are inapplicable (ambiguous trials). Critically, the lack of rlPFC activation during the unambiguous trials suggests that visual feature novelty by itself is not enough to recruit the rlPFC.

This material is based upon work supported by the National Science Foundation under Grant No. 1923267 awarded to M.S.

Talk 7, 4:15 pm

Perceptual Learning Improves Pattern Classification in the Primary Visual Cortex Representation of the Trained Location

Pinar Demirayak¹, Paul Stewart¹, Elam Cutts¹, Madeline Ragland¹, Elliot Maxwell¹, Marcello Maniglia², Samyukta Jayakumar², Jaap Munneke³, Nicholas Turk-Browne⁴, Aaron Seitz³, Kristina Visscher¹; ¹Department of Neurobiology, University of Alabama at Birmingham, Birmingham, AL, USA, ²Department of Psychology, University of California at Riverside, Riverside, CA, USA, ³Department of Psychology, Northeastern University, Boston, MA, USA, ⁴Department of Psychology, Yale University, New Haven, CT, USA

Deterioration of the photoreceptors in the center of the retina can lead to loss of central vision while leaving peripheral vision relatively spared. Individuals with central vision loss often spontaneously develop a “preferred retinal locus” outside the region of vision loss which they use for daily tasks needing finely detailed vision. To better understand the functional brain plasticity associated with this compensation following central vision loss, we used a simulated scotoma paradigm in healthy individuals. Twenty-two participants were trained on basic visual tasks, including contrast sensitivity and contour integration, at a Trained peripheral Retinal Location (TRL) either to the left or right of an artificial scotoma. A control region on the other side was defined as an Unpreferred Retinal Locus (URL). We collected a total of six runs of task-based fMRI data in each pre- and post-training MRI session. Using Multi-Voxel Pattern Analysis, we calculated classification accuracy for differentiating the cortical responses to orientation discrimination stimuli (Gabors) vs contour integration stimuli (contours made up of small Gabors) at the V1 cortical area contralateral to the trained location (cTRL) or contralateral to the URL (cURL). We observed that the cTRL improved classification accuracy more with training than the cURL did (significant ROI x session interaction). Our results indicate that experience leads to more distinct neural activity in a trained location.

NEI 1R01EY031589-01

Talk 8, 4:30 pm

Filtering mechanisms in early visual cortex modulate the occurrence of visual plasticity and learning

Markus Becker¹ (), Sebastian M. Frank¹; ¹University of Regensburg

Visual perceptual learning (VPL), a type of skill learning, is an example of visual plasticity in the adult brain and can occur by mere repeated exposure to task-irrelevant visual features. However, task-irrelevant VPL poses the risk of learning features that are not necessarily adaptive and could jeopardize the stability of previously acquired VPL for task-relevant features. Therefore, one might predict that there are mechanisms that minimize the occurrence of task-irrelevant VPL. An efficient solution would be to filter out task-irrelevant features already at the earliest stage of cortical visual processing (early visual cortex) before they are further processed in higher-order visual areas. Here, we examined this possibility by exposing 30 participants to a task-irrelevant feature (coherent motion in one direction) that was either salient (easily detectable) or weak (difficult to detect) in the visual periphery while they performed a rapid-serial-visual-presentation (RSVP) task at central fixation for six sessions on separate days. Before the first and after the final exposure session functional magnetic resonance spectroscopy (fMRS) using consecutive MEGA-PRESS scans was conducted while participants were exposed to the task-irrelevant feature and performed the RSVP task. The behavioral results showed that task-irrelevant VPL occurred for the weak but not salient task-irrelevant feature. The imaging results showed that glutamate-levels in early visual cortex were significantly lower during exposure to the salient task-irrelevant feature than the weak task-irrelevant feature in each fMRS session. This difference in glutamate-levels in early visual cortex was not found in a control experiment, in which the peripherally exposed salient and weak features were rendered task-relevant. Our results suggest that filtering mechanisms exist at the earliest stage of cortical visual processing, which modulate the occurrence of visual plasticity and learning. This filtering might occur through downregulating excitatory, glutamatergic activity in early visual cortex for task-irrelevant features that are sufficiently salient.

Deutsche Forschungsgemeinschaft (DFG): Emmy Noether Grant (Project Number 491290285); Julitta und Richard Müller Stiftung