Biology subjects

Pezzulo, G.

Publications and source records attributed to Pezzulo, G..

At least 19 recordsLinked to original sources

bioRxiv2026

Expert and Novice Visuomotor Strategies in a Sequential Football Task Reflect Distinct Generative Models

In visuomotor sequential tasks, chaining actions smoothly requires tight coordination between sensory and motor systems to deal with competing demands, such as reducing uncertainty about future opportunities for action while controlling ongoing ones. In a football-based decision-making task, experts and novices used different strategies to overcome these competing demands, demonstrating a proactive and a sequential visuomotor strategy, respectively. We propose a computational explanation of expertise-related difference in policy selection, grounded in the active inference framework. We simulated the selection of sequence of actions using a partially observable Markov decision process (POMDP) model, and we manipulated three parameters: the reward structure, the prior beliefs about state transition, and the sensory precision. Simulations showed that selection of the proactive strategy relied on strong prior beliefs, high sensory precision and high valuation of fast passing. Conversely, the sequential policy was selected when rapid performance was not highly valued, or under noisy prior beliefs, or lower level of sensory precision. Decreasing sensory precision and adding noise in prior beliefs also decreased decision accuracy. These results suggest that policy selection leading to novice or expert visuomotor strategy reflects differences in the underlying generative model, particularly in how agents value different aspects of the task.

neuroscience↗

bioRxiv2026

Flexible predictive control in human interception under visual occlusion and altered gravity

Interception of moving objects requires the nervous system to compensate for sensory delays and uncertainty, yet how behavior is controlled remains debated. Key questions concern whether predictive processes play any role at all and, if so, whether they rely on simple motion extrapolation or incorporate internalized physical priors, such as gravity. Another open question is whether observers adopt a single control strategy or flexibly switch between predictive and reactive control - or between different predictive strategies - depending on task demands. To address these questions, we developed a virtual interception task in which participants intercepted moving targets under systematically varied conditions. We manipulated gravity (1g vs. 0g), visual availability (occluded vs. non-occluded), target velocity, and the initial spatial configuration of the ball and paddle (same vs. opposite side). Results indicate that interception is supported by predictive mechanisms across conditions. Behavioral patterns during occluded 0g trials suggest that participants extrapolate target motion using expectations consistent with gravity. Target velocity, visual occlusion, and task geometry modulated movement strategies, indicating that predictive control is flexibly adapted to task demands. These findings support the view that interception relies on predictive internal models incorporating structured physical priors while revealing flexible, context-dependent adaptations to sensory and task constraints.

neuroscience↗

bioRxiv2026

Generating whole-brain neural activity and behavior through unified latent dynamics

Understanding how high-dimensional neural activity and behavior emerge from shared underlying dynamics remains a fundamental challenge in neuroscience. Addressing this problem is key to enabling digital twins that can faithfully reproduce and predict the multiscale brain-behavior dynamics of living systems. Here we present NEBULA (NEural and Behavioral modeling through Unified LAtent dynamics), a generative framework that jointly models whole-brain neural activity and behavior. By applying NEBULA to brain-wide recordings from C. elegans, we identify a shared low-dimensional dynamical manifold that underlies both neural activity and behavior, enabling long-horizon generation and targeted in silico interventions. Perturbations of the learned dynamics reveal behaviorally relevant transition points, whereas steering interventions enable controlled manipulation of neural and behavioral states without retraining. These results establish a framework for linking brain dynamics to behavior in a living organism and provide a foundation for scalable virtual experimentation in neuroscience.

animal behavior and cognition↗

bioRxiv2026

Brain-wide distributed processing underlying natural vision and audition

A key challenge in neuroscience is understanding how large-scale brain dynamics give rise to perception under naturalistic conditions. Here we address this question using multivariate information decomposition of functional interactions between brain regions while typically developed, congenitally blind and congenitally deaf participants experienced audiovisual, auditory-only or visual-only versions of the same narrative. This approach allowed us to resolve information dynamics into redundant components, indexing shared information and sensory robustness, and synergistic components, indexing distributed processing that emerges only through joint interactions between brain regions. Redundant interactions were closely aligned with conventional functional connectivity and captured sensory modality- and experience-dependent differences mainly within sensory systems, whereas synergistic interactions captured these differences across high-level cortical areas. Crucially, prefrontal cortex coordinated brain dynamics through synergistic interactions independently of sensory modality and previous sensory experience, consistent with a modality-independent hub that encodes high-level semantic information and orchestrates the demands of naturalistic perception. Moreover, multimodal perception elicited the strongest high-order synergistic interactions across distributed brain subsystems, followed by unimodal visual and auditory perception, whereas sensory deprivation reduced this distributed architecture. Finally, synergistic and redundant interactions jointly supported the encoding of low-level sensory and high-level perceptual features, revealing how the trade-off between sensory robustness and distributed processing is organized across the human brain. Together, these results reveal a brain-wide information architecture for natural vision and audition, illuminating how distributed processing in the human brain supports rich perceptual experience.

neuroscience↗

bioRxiv2026

Determinants of persistence in sequential effort-based decision-making

Goal-directed behavior often requires sustained effort across a sequence of interdependent decisions, yet the determinants of persistence in such contexts remain poorly understood. Here, we investigated how individuals regulate persistence in a novel sequential effort-based task in which they controlled an avatar through successive checkpoints to reach a final goal and could make repeated attempts following failure. At each attempt, participants could choose either to persist in the same task or to disengage toward an easier but less rewarding alternative. We found that decisions to persist or disengage were jointly shaped by multiple interacting factors. Disengagement increased with task difficulty and lower skill level. It also increased with repeated attempts and time-on-task, indexing fatigue, and with accumulated errors, indexing lack of progress. Conversely, proximity to the goal promoted persistence and shaped decision dynamics by reducing choice conflict during persistence decisions and increasing hesitation during disengagement near the goal. Notably, clearing the first checkpoint produced a sharp increase in persistence, suggesting that early success plays a pivotal role. Furthermore, persistence reflected both retrospective and prospective evaluations of effort, with prior investment promoting commitment and anticipated effort reducing it. Finally, disengagement was preceded by short-term performance decline but not by gradual increases in decision conflict, suggesting relatively abrupt strategy shifts following repeated failures. Together, these findings provide a comprehensive account of persistence in sequential effortful tasks, showing that decisions to persist or disengage are jointly shaped by multiple factors related to fatigue, (lack of) progress, goal proximity, and early success.

neuroscience↗

bioRxiv2026

TRAINING SESSION INTERVALS SHAPE FUNCTIONAL CONNECTIVITY IN SPATIAL LEARNING: A BRAIN-WIDE ANALYSIS

A substantial body of research indicates that spaced training, characterized by longer inter-trial intervals between training epochs, consistently outperforms massed training in promoting durable memory. To investigate the neural mechanisms underlying this difference, we quantified c-Fos expression across 126 brain regions and mapped network activity following both spatial and cue-based learning under massed and spaced training protocols. While both training regimens (spaced and massed) and memory types (spatial and cue-based) produced small-world networks with similar overall topology, they differed in the functional organization of specific circuits. Massed spatial training preferentially activated hippocampal-thalamic and claustrum-basal ganglia-thalamic pathways. In contrast, spaced spatial training promoted stronger cortico-thalamic interactions and enhanced communication between the hippocampus and basal ganglia, indicating a shift toward a more integrated, cortically mediated network. These findings suggest that temporal spacing of training reorganizes memory-related brain networks, enhancing cortical-thalamic dynamics to support more efficient spatial memory. Finally, spaced networks were more sensitive to targeted disruptions of key connector hubs--identified through betweenness centrality analysis--than massed networks, pointing to a potential systems-level trade-off.

neuroscience↗

bioRxiv2025

Inferential planning in the frontal cortex

How the brain plans and maintains sequences of future actions remains a central question in systems neuroscience. Recent studies in the frontal cortex have revealed that multiple elements of a sequence are represented simultaneously in separable neural subspaces, challenging classical serial models of sequential planning. Here, we show that these representations emerge naturally under inferential planning in which sequential actions are inferred from sensory evidence and goals. Using a hierarchical generative model, we reproduce key neural phenomena observed in primate frontal cortex, including the simultaneous activation of multiple plan elements, the emergence of (almost) orthogonal memory subspaces, and their reuse across forward and backward sequence tasks. Our approach provides a mechanistic account of how probabilistic inference over control states gives rise to distributed and dynamic neural representations of plans. This framework not only unifies previously disparate findings on planning, working memory, and motor preparation, but also generates novel, testable predictions about the dynamics of active inference, the role of sensory subspaces, and the impact of uncertainty on sequence processing.

neuroscience↗

bioRxiv2025

Choice-induced preference change under a sequential-sampling model framework

Sequential sampling models of choice, such as the drift-diffusion model (DDM), are frequently fit to empirical data to account for a variety of effects related to accuracy/consistency, response time (RT), and sometimes confidence. However, no model in this class has been shown to account for the phenomenon known as choice-induced preference change, wherein decision makers tend to rate options higher after they choose them and lower after they reject them (and often choose the option that they had initially rated lower). Studies have reported choice-induced preference change for many decades, and the principal findings are robust. The resulting spreading of alternatives (SoA) in terms of their subjective value ratings is incompatible with traditional sequential sampling models, which consider the rated values of the options to be stationary throughout choice deliberation. Here, we propose that extending the basic DDM to incorporate independent attributes into the drift rate can allow this class of model to account for SoA. Critically, the extended model assumes that choice deliberation does not always involve the same set of attributes as is involved in individual evaluations of the same options. We show that this model can generate SoA (while simultaneously accounting for consistency and RT), as well as the relationships between SoA and choice difficulty, attribute disparity, and RT previously reported in the literature.

neuroscience↗

bioRxiv2025

Neural sampling from cognitive maps supportsgoal-directed planning and imagination

AI systems are becoming more intelligent, but at a very high cost in terms of energy consumption and training requirements. In contrast, our brains only require 20W of energy, they learn online, and they can instantly adjust to changing contingencies. This begs the question what data structures, algorithms, and learning methods enable brains to achieve that, and whether these can be ported into artificial devices. We are addressing this question for a core feature of intelligence: The capacity to plan and solve problems, including new problems that involve states which were never encountered before. We examine three tools that brains are likely to employ for achieving that: Cognitive maps, stochastic computing, and compositional coding. We integrate these tools into a transparent neural network model, and demonstrate its power for flexible planning and problem solving. Importantly, this approach is suitable for implementation by in-memory computing and other energy-efficient neuromorphic hardware. In particular, it only requires self-supervised local synaptic plasticity that is suited for on-chip learning. Hence a core feature of brain intelligence, the capacity to generate solutions to problems that were never encountered before, does not require deep neural networks or large language models, and can be implemented in energy-efficient edge devices.

neuroscience↗

bioRxiv2025

Gaze dynamics prior to navigation support hierarchical planning

The task of planning future actions in the context of an uncertain world results in massive state spaces that preclude exhaustive search and other strategies explored in the domains of both human decision-making and computational agents. One plausible solution to this dimensionality explosion is to decompose the task into subgoals that match the information geometry of the task at hand. However, how individuals identify a productive hierarchy, and perceive and select subgoals suitable to planning, is not well understood. To investigate this topic, we designed a virtual-reality based behavioral experiment which collected eye movements during a pre-navigation planning phase. By capturing gaze dynamics correlated with the simulative processes used in planning, we were able to identify the spatiotemporal evolution of visual search under uncertainty. Our results highlight gaze dynamics indicative of a search process that exhibits hierarchical structure. These include a decreasing trend seen in gaze distance from origin and a broad to narrow shift (with reducing saccade distances and longer fixation durations) as plans are established. In line with prior work, critical tiles to which landscape connectivity is most sensitive were the strongest predictors of visual attention. We also find that deeper planning was correlated with success only on the most complex maps (e.g. those with a larger number of information-nodes, higher branching factor, and more forks, according to an info-graphical map analysis). This study highlights the role of embodied visual search during planning, and the skill-dependence of the specific subgoals and hierarchical decomposition used which unlocked successful performance.

neuroscience↗

bioRxiv2025

Whole-body coarticulation reflects expertise in sport climbing

Taking sport climbing as a testbed, we explored coarticulation in naturalistic motor-behavior at the level of whole-body kinematics. Participants were instructed to execute a series of climbing routes, each composed of two initial foot-moves equal in all routes, and two subsequent hand-moves differing across routes in a set of eight possible configurations. The goal was assessing whether climbers modulate the execution of a given move depending on which moves come next in the plan. Coarticulation was assessed by training a set of classifiers and estimating how well the whole-body (or single-joint) kinematics during a given stage of the climbing execution could predict its future unfolding. Results showed that most participants engage in coarticulation, with temporal and bodily patterns that depend on expertise. Non-climbers tend to prepare the next-to-come move right before its onset and only after the end of the previous move. Rather, expert-climbers (and to a smaller extent, beginner-climbers) show early coarticulation during the execution of the previous move and engage in adjustments that involve the coordination of a larger number of joints across the body. These results demonstrate coarticulation effects in whole-body naturalistic motor behavior and as a function of expertise. Furthermore, the enhanced coarticulation found in expert-climbers provides hints for experts engaging in more refined mental processes converting abstract instructions (e.g., move the right hand to a given location) into motor simulations involving whole-body coordination. Overall, these results contribute to advancing our current knowledge of the rich interplay between cognition and motor control. NEW & NOTEWORTHYThe current study explores the way in which having formed a plan for a sequential motor task affects its execution. We showed that climbing expertise increases the extent to which participants adjust their motor execution based on the moves that follow in a planned route. These results provide evidence of coarticulation in naturalistic motor behavior and suggest enhanced skills in mentalizing forward motor control and optimal-control strategies in expert climbers.

neuroscience↗

bioRxiv2024

Separable cognitive and motor decline tracked across adult life-span for goal-directed navigation

Spatial navigation is a fundamental cognitive skill, yet our understanding of the strategies people employ in real-world navigation remains incomplete. This study investigates navigation strategies using data from the large-scale Sea Hero Quest (SHQ) game, in which participants face navigational problems consisting in controlling a boat to reach a series of goal locations, within maps having different shapes. By using a combination of behavioral and computational modeling approaches, we report three key findings: First, participants predominantly use goal-directed strategies compared to visibility-based strategies. Second, participants show signatures of sequential planning, both during memorization - as they spend more time looking at difficult levels - and during navigation - as they show primacy effects, suggesting that they remember better the first navigational goal. Third, our results show a significant age-related decline in navigation performance, which depends both on reduced motor skill - as indexed by poorer control of the boat - and decreased cognitive strategy - as indexed by a loss of goal-directedness during spatial navigation. Unexpectedly, the loss of goal-directedness with age is not associated with a shift to visibility-based strategy or to an incorrect sequencing of goals. Taken together, our findings offer a nuanced understanding of spatial navigation strategies in naturalistic conditions and their decline with age. Our computational approach permits distinguishing motoric and cognitive variables, showing that both decline with age. These results align with and extent previous research on spatial memory decline with age, including potential links to neurodegenerative diseases like Alzheimers. The computational models developed in this study provide a robust framework for analyzing navigation strategies, applicable to other virtual environments and video games.

animal behavior and cognition↗

bioRxiv2024

Planning-while-acting: addressing the continuous dynamics of planning and action in a sequential embodied task

Everyday tasks, such as selecting routes when driving or preparing meals, require making sequences of embodied decisions, in which planning and action processes are intertwined. In this study, we address how people make sequential embodied decisions, requiring balancing be-tween immediate affordances and long-term utilities of alternative action plans. We designed a novel virtually embodied task in which participants controlled an avatar tasked with "crossing rivers" by jumping across rocks. The task permitted us to assess how participants balanced between immediate jumping affordances ("safe" versus "risky" jumps) and the utility (length) of the ensuing paths to the goal. Behavioral and computational analyses revealed that partici-pants planned ahead their path to the goal rather than simply focusing on the most immediate jumping affordances. Furthermore, spatial and embodied components of the task influenced participants decision strategies, as participants current direction of movement and momen-tum influenced their choice between safe and risky jumps. Additionally, participants showed (pre)planning before making the first jump, but they continued deliberating during it, with movement speed decreasing at decision points and when approaching them. Computational modeling indicates that farsighted participants who assigned greater weight to the utility of future jumps showed a better performance, highlighting the usefulness of planning in embodied settings. Finally, analysing participants performance across trials indicates that during the experiment, they become faster in moving and deciding but they do not change their overall strategy. Our findings underscore the importance of studying decision-making and planning in ecologically valid, embodied settings, providing new insights into the interplay between action and cognition in real-world planning-while-acting scenarios. Significance statementEveryday activities, such as navigating a crowded street or assembling objects, often involve contin-uous interactions between planning and action. Yet, it is unclear how people balance opportunities for action (affordances) with long-term planning in embodied settings. In a novel virtually embod-ied task, in which people have to "cross rivers" by making a series of jumps across rocks, we found that they plan ahead their path to the goal rather than simply focusing on the immediate jumping affordances - and they continue deliberating during movement. Crucially, spatial and embodied constraints, such as movement direction and momentum, influenced decision-making, challenging traditional models that treat decision and action as distinct stages and highlighting the importance of studying decisions in ecologically valid, embodied settings.

animal behavior and cognition↗

bioRxiv2024

Eye and hand coarticulation during problem solving reveals hierarchically organized planning

During everyday activities--such as preparing a cup of coffee or traveling across cities--we often plan ahead and execute sequences of actions. However, much remains to be understood about how we plan and coordinate such sequences (e.g., eye and hand movements) to solve novel and challenging tasks, for which plans must be formed from scratch. This study investigates how participants coordinate gaze and cursor movements during problem solving tasks that involve selecting a trajectory on a grid connecting multiple targets. By focusing on the action execution phase, we aimed to probe the structure of the gaze-cursor plans that participants used to solve the tasks. Our analysis reveals three main findings. First, consistent with previous studies, participants segment the problem into sequences of gestures; within each gesture, gaze focuses on a target and remains fixed until the cursor reaches it, then shifts to the next target. Second, both gaze position--while fixating on the current target--and the kinematics of cursor movement leading up to the current target allow prediction of the next cursor movements direction, revealing coarticulation in both cursor-cursor and gaze-cursor movements. Third, and most interestingly, the position of the gaze around the current target aligns with the direction of the next saccade, revealing coarticulation between successive gaze fixations. Together, these findings show that participants break the problem into gesture sequences and plan multiple eye and cursor movements in advance to efficiently reach both the current and upcoming gesture targets. This suggests a hierarchical planning strategy, with participants planning ahead at two levels: gesture targets and cursor movements.

animal behavior and cognition↗

bioRxiv2024

Learning and embodied decisions in active inference

Biological organisms constantly face the necessity to act timely in dynamic environments and balance choice accuracy against the risk of missing valid opportunities. As formalized by embodied decision models, this might require brain architectures wherein decision-making and motor control interact reciprocally, in stark contrast to traditional models that view them as serial processes. Previous studies have assessed that embodied decision dynamics emerge naturally under active inference - a computational paradigm that considers action and perception as subject to the same imperative of free energy minimization. In particular, agents can infer their targets by using their own movements (and not only external sensations) as evidence, i.e., via self-evidencing. Such models have shown that under appropriate conditions, action-generated feedback can stabilize and improve decision processes. However, how adaptation of internal models to environmental contingencies influences embodied decisions is yet to be addressed. To shed light on this challenge, in this study we systematically investigate the learning dynamics of an embodied model of decision-making during a two-alternative forced choice task, using a hybrid (discrete and continuous) active inference framework. Our results show that active inference agents can adapt to embodied contexts by learning various statistical regularities of the task - namely, prior preferences for the correct target, cue validity, and response strategies that prioritize faster or slower (but more accurate) decisions. Crucially, these results illustrate the efficacy of learning discrete preferences and strategies using sensorimotor feedback from continuous dynamics.

neuroscience↗

bioRxiv2024

Embodied decisions as active inference

Decision-making is often conceptualized as a serial process, during which sensory evidence is accumulated for the choice alternatives until a certain threshold is reached, at which point a decision is made and an action is executed. This decide-then-act perspective has successfully explained various facets of perceptual and economic decisions in the laboratory, in which action dynamics are usually irrelevant to the choice. However, living organisms often face another class of decisions - called embodied decisions - that require selecting between potential courses of actions to be executed timely in a dynamic environment, e.g., for a lion, deciding which gazelle to chase and how fast to do so. Studies of embodied decisions reveal two aspects of goal-directed behavior in stark contrast to the serial view. First, that decision and action processes can unfold in parallel; second, that action-related components, such as the motor costs associated with selecting a particular choice alternative or required to "change mind" between choice alternatives, exert a feedback effect on the decision taken. Here, we show that these signatures of embodied decisions emerge naturally in active inference - a framework that simultaneously optimizes perception and action, according to the same (free energy minimization) imperative. We show that optimizing embodied choices requires a continuous feedback loop between motor planning (where beliefs about choice alternatives guide action dynamics) and motor inference (where action dynamics finesse beliefs about choice alternatives). Furthermore, our active inference simulations reveal the normative character of embodied decisions in ecological settings - namely, achieving an effective balance between a high accuracy and a low risk of missing valid opportunities. Author summaryIn this study, we introduce a novel modeling approach to explore embodied decision-making, where decisions and actions occur simultaneously in dynamic environments. Unlike traditional models that treat decision and action as separate, our framework, based on active inference, reveals that crucial features of embodied decisions - such as feedback loops between decision and action dynamics - emerge naturally. By simulating real-time decision-making tasks, we show how organisms continuously refine their choices by integrating sensory information and motor dynamics. This allows them to strike a balance between decision accuracy and the need for fast, adaptive actions. Our model offers a new perspective on how decisions are influenced by the actions taken, highlighting the importance of considering motor control as an integral part of decision processes. This approach broadens the scope of decision-making research and provides new insights into behavior in ecologically valid, time-sensitive contexts, with potential implications for neuroscience, cognitive science, and fields involving human and animal behavior.

neuroscience↗

bioRxiv2024

Meta-Reinforcement Learning reconciles surprise, value and control in the anterior cingulate cortex.

The role of the dorsal anterior cingulate cortex (dACC) in cognition is a frequently studied yet highly debated topic in neuroscience. Most authors agree that the dACC is involved in either cognitive control (e.g. voluntary inhibition of automatic responses) or monitoring (e.g. comparing expectations with outcomes, detecting errors, tracking surprise). A consensus on which theoretical perspective best explains dACC contribution to behaviour is still lacking, as two distinct sets of studies report dACC activation in tasks requiring surprise tracking for performance monitoring and cognitive control without involving surprise monitoring, respectively. This creates a theoretical impasse, as no single current account can reconcile these findings. Here we propose a novel hypothesis on dACC function that integrates both the monitoring and the cognitive control perspectives in a unifying, meta-Reinforcement Learning framework, in which cognitive control is optimized by meta-learning based on tracking Bayesian surprise. We tested the quantitative predictions from our theory in two different functional neuroimaging studies at the basis of the current theory crisis. We show that the meta-Reinforcement Learning perspective successfully captures all the neuroimaging results by predicting both cognitive control and monitoring functions, resolving the theoretical impasse about dACC function within an integrative framework. In sum, our results suggest that dACC function can be framed as a meta-learning optimisation of cognitive control, providing an integrative perspective on its roles in cognitive control, surprise tracking, and performance monitoring. Significance statementAn important debate in cognitive neuroscience concerns the role of the anterior cingulate cortex (ACC) in cognition. Two effective and competing frameworks suggest a role for the ACC in surprise monitoring or optimizing cognitive control, respectively. So far, none of these frameworks has succeeded as a unified theory of ACC function. In this study, we reanalyzed previous neuroimaging data on ACC activity during cognitive tasks using a novel computational perspective: meta-Reinforcement Learning. We show that this computational framework can explain a variety of data on ACC function that, globally, could not be captured by any of the previous models. We propose that meta-Reinforcement Learning offers a unified theory of ACC cognitive and computational function.

neuroscience↗

bioRxiv2023

Learning to segment self-generated from externally caused optic flow through sensorimotor mismatch circuits

Efficient sensory detection requires the capacity to ignore task-irrelevant information, for example when optic flow patterns created by egomotion need to be disentangled from object perception. To investigate how this is achieved in the visual system, predictive coding with sensorimotor mismatch detection is an attractive starting point. Indeed, experimental evidence for sensorimotor mismatch signals in early visual areas exists, but it is not understood how they are integrated into cortical networks that perform input segmentation and categorization. Our model advances a biologically plausible solution by extending predictive coding models with the ability to distinguish self-generated from externally caused optic flow. We first show that a simple three neuron circuit produces experience-dependent sensorimotor mismatch responses, in agreement with calcium imaging data from mice. This microcircuit is then integrated into a neural network with two generative streams. The motor-to-visual stream consists of parallel microcircuits between motor and visual areas and learns to spatially predict optic flow resulting from self-motion. The second stream bidirectionally connects a motion-selective higher visual area (mHVA) to V1, assigning a crucial role to the abundant feedback connections: the maintenance of a generative model of externally caused optic flow. In the model, area mHVA learns to segment moving objects from the background, and facilitates object categorization. Based on shared neurocomputational principles across species, the model also maps onto primate vision. Our work extends the Hebbian predictive coding to sensorimotor settings, in which the agent actively moves - and learns to predict the consequences of its own movements. Significance statementThis research addresses a fundamental challenge in sensory perception: how the brain distinguishes between self-generated and externally caused visual motion. Using a computational model inspired by predictive coding and sensorimotor mismatch detection, the study proposes a biologically plausible solution. The model incorporates a neural microcircuit that generates sensorimotor mismatch responses, aligning with experimental data from mice. This microcircuit is integrated into a neural network with two streams: one predicting self-motion-induced optic flow and another maintaining a generative model for externally caused optic flow. The research advances our understanding of how the brain segments visual input into object and background, shedding light on the neural mechanisms underlying perception and categorization not only in rodents, but also in primates.

neuroscience↗