Yves Boubenec
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yves Boubenec has a total of 84 co-authors. Here are the top co-authors by number of shared publications:
- Pascal Mamassian
Laboratoire des systèmes perceptifs, Département d’études Cognitives, École Normale Supérieure, PSL University, CNRS, Paris, France
11 shared publications Shihab Shamma
Laboratoire des systèmes perceptifs, Département d'études Cognitives, École Normale Supérieure, PSL University, CNRS, Paris, France
7 shared publications- Srdjan Ostojic
Laboratoire de Neurosciences Cognitives et Computationnelles, INSERM U960, Ecole Normale Superieure - PSL Research University, Paris, France
7 shared publications - Célian Bimbard
UCL Institute of Ophthalmology, University College London
6 shared publications - Agnès Landemard
Laboratoire des Systèmes Perceptifs, Département d’Études Cognitives, École Normale Supérieure, PSL University
4 shared publications - Shihab Shamma
Neural Systems Laboratory, University of Maryland College Park
4 shared publications
Latest Publications
- Open Access
Neural networks as decision trees: an analytical solution for learning and neural selectivity
Nonlinear neural networks develop structured internal representations, yet how their geometry is determined by the tasks being learned remains poorly understood. Here, we develop an analytical framework for piecewise-linear feedforward and recurrent networks that links learning, activation-region structure, and neural selectivity. We show that, at any stationary point of gradient-aligned learning, a nonlinear network decomposes into local linear regressions over its activation regions. Deviations from the corresponding least-squares solutions are jointly constrained by network weights shared across regions and vanish in the low-error regime, yielding an approximate piecewise least-squares decomposition of the task. This structure admits a decision-tree interpretation, which we recover numerically by fitting trees to predict network activation patterns from the input. We further derive how the regions in which neurons are active determine the task statistics they capture, thereby organizing neural selectivity into distinct subpopulations. The resulting predictions closely match the selectivity geometry observed in simulated networks and two empirical neural datasets. Finally, we show that neural baseline regulates activation-pattern diversity, placing networks along a continuum between coarse, generalizing representations and fine-grained, expressive representations. Together, these results establish activation regions as a unifying framework for describing how nonlinear networks decompose task structure into local computations and how those computations can be revealed through neural activity.
Open AccessNeural networks as decision trees: an analytical solution for learning and neural selectivity
Nonlinear neural networks develop structured internal representations, yet how their geometry is determined by the tasks being learned remains poorly understood. Here, we develop an analytical framework for piecewise-linear feedforward and recurrent networks that links learning, activation-region structure, and neural selectivity. We show that, at any stationary point of gradient-aligned learning, a nonlinear network decomposes into local linear regressions over its activation regions. Deviations from the corresponding least-squares solutions are jointly constrained by network weights shared across regions and vanish in the low-error regime, yielding an approximate piecewise least-squares decomposition of the task. This structure admits a decision-tree interpretation, which we recover numerically by fitting trees to predict network activation patterns from the input. We further derive how the regions in which neurons are active determine the task statistics they capture, thereby organizing neural selectivity into distinct subpopulations. The resulting predictions closely match the selectivity geometry observed in simulated networks and two empirical neural datasets. Finally, we show that neural baseline regulates activation-pattern diversity, placing networks along a continuum between coarse, generalizing representations and fine-grained, expressive representations. Together, these results establish activation regions as a unifying framework for describing how nonlinear networks decompose task structure into local computations and how those computations can be revealed through neural activity.
- Open Access
- 19 August 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Sensorimotor remapping drives task specialization in prefrontal cortex
The prefrontal cortex (PFC) plays a key role in flexible, context-dependent decision-making. Yet, the population-level mechanisms that support this flexibility remain unclear: is there a generalist neural population involved in representing useful information across tasks, or are representations distributed over multiple modules of task-specialized subpopulations? Here, we investigate whether the population structure modularity in medial prefrontal cortex (mPFC) increases with sensorimotor remapping between task rules in three different paradigms. In a paradigm where all stimuli were remapped, we found a highly modular structure with two task-specialized subpopulations and one generalist. In a different paradigm where only half of the stimuli were remapped, modularity decreased but remained significant. In a last paradigm where no stimuli were remapped, modularity disappeared. Based on predictions drawn from a reservoir neural network fitted to experimental data, we propose that a gain modulation by global inhibition of mPFC can explain how task-specialized neurons emerge to varying degrees in each paradigm.
- 19 August 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessSensorimotor remapping drives task specialization in prefrontal cortex
The prefrontal cortex (PFC) plays a key role in flexible, context-dependent decision-making. Yet, the population-level mechanisms that support this flexibility remain unclear: is there a generalist neural population involved in representing useful information across tasks, or are representations distributed over multiple modules of task-specialized subpopulations? Here, we investigate whether the population structure modularity in medial prefrontal cortex (mPFC) increases with sensorimotor remapping between task rules in three different paradigms. In a paradigm where all stimuli were remapped, we found a highly modular structure with two task-specialized subpopulations and one generalist. In a different paradigm where only half of the stimuli were remapped, modularity decreased but remained significant. In a last paradigm where no stimuli were remapped, modularity disappeared. Based on predictions drawn from a reservoir neural network fitted to experimental data, we propose that a gain modulation by global inhibition of mPFC can explain how task-specialized neurons emerge to varying degrees in each paradigm.
- Open Access
- 13 May 2026
- Published by Public Library of Science (PLoS) in PLOS Biology
- Vol. 24 (5) , e3003768
Premotor cortex hemodynamic responses primarily reflect perceptual rather than specific motor aspects of decision making
Decisions are driven by perception, but also by non-perceptual factors. It remains an open question, however, whether frontal brain regions involved in perceptual decision-making tend to uniquely reflect the perception of an animal, or the final choice of action driven by perceptual and non-perceptual factors. Using functional ultrasound imaging (fUSI), we investigated how the premotor cortex (PMC) in ferrets represents stimuli in a Go/NoGo auditory categorization task, varying the difficulty in order to manipulate the rates of perceptual errors relative to non-perceptual errors. We found that on Easy error trials, where error in perception was less likely, PMC activity was similar to correct answers for the same stimulus category. In contrast, on Difficult error trials, PMC activity more closely reflected the choice the animal made, being similar to correct answers for the opposite category. These results together are consistent with PMC activity reflecting the reward-predictive perceptual category. Perceptual errors could be refined further by assessing licking patterns, but licking patterns alone did not explain the effect. This study advances our understanding of the functional role of the frontal cortex in decision-making, suggesting that the PMC integrates sensory inputs to guide behavior based on perceptual information, rather than motivational information.
- 13 May 2026
- Published by Public Library of Science (PLoS) in PLOS Biology
- Vol. 24 (5) , e3003768
Open AccessPremotor cortex hemodynamic responses primarily reflect perceptual rather than specific motor aspects of decision making
Decisions are driven by perception, but also by non-perceptual factors. It remains an open question, however, whether frontal brain regions involved in perceptual decision-making tend to uniquely reflect the perception of an animal, or the final choice of action driven by perceptual and non-perceptual factors. Using functional ultrasound imaging (fUSI), we investigated how the premotor cortex (PMC) in ferrets represents stimuli in a Go/NoGo auditory categorization task, varying the difficulty in order to manipulate the rates of perceptual errors relative to non-perceptual errors. We found that on Easy error trials, where error in perception was less likely, PMC activity was similar to correct answers for the same stimulus category. In contrast, on Difficult error trials, PMC activity more closely reflected the choice the animal made, being similar to correct answers for the opposite category. These results together are consistent with PMC activity reflecting the reward-predictive perceptual category. Perceptual errors could be refined further by assessing licking patterns, but licking patterns alone did not explain the effect. This study advances our understanding of the functional role of the frontal cortex in decision-making, suggesting that the PMC integrates sensory inputs to guide behavior based on perceptual information, rather than motivational information.
- Open Access
Sound-evoked facial motion in ferrets: evidence for species differences in sensorimotor coupling
Sensory stimuli elicit uninstructed movements in the mouse, and cortical activity covaries with these movements, raising the question of how much apparent sensory encoding reflects behaviorally driven signals. Still, whether sounds systematically elicit uninstructed movements in non-rodent species is an open question. Here we characterize sound-evoked facial motion and pupil dynamics in head-fixed ferrets, a carnivore species, passively exposed to broadband, natural, and synthetic auditory stimuli. Ferrets reliably produced facial responses to sounds, but unlike mice, these responses were dominated by a single onset-locked component scaling with sound loudness, tracked only temporal modulations below 2 Hz, and carried no information about sound identity or category. Paradoxically, synthetic sounds, which preserved spectrotemporal statistics but lacked higher-order natural structure, elicited stronger responses than their natural counterparts for several sound categories, suggesting that ferret behavioral responses reflect selective sensitivity to acoustic novelty rather than ecological salience. Pupil dynamics mirrored facial motion, consistent with a shared arousal mechanism. Together, these results show that sound-evoked facial motion in ferrets is less stimulus-specific than in mice, and suggest that the tight sensorimotor integration observed in rodents may reflect a species-specific organization rather than a general mammalian principle.
Open AccessSound-evoked facial motion in ferrets: evidence for species differences in sensorimotor coupling
Sensory stimuli elicit uninstructed movements in the mouse, and cortical activity covaries with these movements, raising the question of how much apparent sensory encoding reflects behaviorally driven signals. Still, whether sounds systematically elicit uninstructed movements in non-rodent species is an open question. Here we characterize sound-evoked facial motion and pupil dynamics in head-fixed ferrets, a carnivore species, passively exposed to broadband, natural, and synthetic auditory stimuli. Ferrets reliably produced facial responses to sounds, but unlike mice, these responses were dominated by a single onset-locked component scaling with sound loudness, tracked only temporal modulations below 2 Hz, and carried no information about sound identity or category. Paradoxically, synthetic sounds, which preserved spectrotemporal statistics but lacked higher-order natural structure, elicited stronger responses than their natural counterparts for several sound categories, suggesting that ferret behavioral responses reflect selective sensitivity to acoustic novelty rather than ecological salience. Pupil dynamics mirrored facial motion, consistent with a shared arousal mechanism. Together, these results show that sound-evoked facial motion in ferrets is less stimulus-specific than in mice, and suggest that the tight sensorimotor integration observed in rodents may reflect a species-specific organization rather than a general mammalian principle.
- Open Access
- 4 December 2025
- Published by Elsevier in Current Biology
- Vol. 35 (24) , 6114-6125.e7
Neurons in auditory cortex integrate information within a constrained and context-invariant temporal window
- 4 December 2025
- Published by Elsevier in Current Biology
- Vol. 35 (24) , 6114-6125.e7
Open AccessNeurons in auditory cortex integrate information within a constrained and context-invariant temporal window
- Open Access
Encoding and combining naturalistic motion cues in the ferret higher visual cortex
Natural visual scenes contain rich flows of pattern motion that vary not only in orientation but also in spatial sizes and temporal rhythms. To properly interpret the motion, the brain must extract individual features and integrate them into coherent parts, and sometimes also segregate these parts from each other. Here, we performed single-neuron recordings in the motion-sensitive higher-order visual cortex (PMLS) of awake ferrets to investigate how complex motion signals are encoded and combined. We presented motion clouds—naturalistic stimuli with parametrically controlled spatiotemporal frequency content—and found that motion features were encoded in a temporally ordered sequence: orientation and spatial frequency emerged within 120 ms after stimulus onset, while temporal frequency and direction followed at later latencies. Time-resolved decoding revealed that this selectivity evolved dynamically within neurons and was distributed across the population. To probe motion integration, we introduced compound motion clouds composed of two or three localized frequency components. Neuronal responses were well explained by a linear pooling model, suggesting a simple summation mechanism of the individual components. However, a distinct subset of neurons exhibited late responses sensitive to changes in speed content despite matched marginals, consistent with receptive fields differentiating along the speed gradient. Together, we have uncovered a structured and distributed code for motion in high-level visual cortex, and provide mechanistic insights into how the brain parses complex motion in natural scenes.
Open AccessEncoding and combining naturalistic motion cues in the ferret higher visual cortex
Natural visual scenes contain rich flows of pattern motion that vary not only in orientation but also in spatial sizes and temporal rhythms. To properly interpret the motion, the brain must extract individual features and integrate them into coherent parts, and sometimes also segregate these parts from each other. Here, we performed single-neuron recordings in the motion-sensitive higher-order visual cortex (PMLS) of awake ferrets to investigate how complex motion signals are encoded and combined. We presented motion clouds—naturalistic stimuli with parametrically controlled spatiotemporal frequency content—and found that motion features were encoded in a temporally ordered sequence: orientation and spatial frequency emerged within 120 ms after stimulus onset, while temporal frequency and direction followed at later latencies. Time-resolved decoding revealed that this selectivity evolved dynamically within neurons and was distributed across the population. To probe motion integration, we introduced compound motion clouds composed of two or three localized frequency components. Neuronal responses were well explained by a linear pooling model, suggesting a simple summation mechanism of the individual components. However, a distinct subset of neurons exhibited late responses sensitive to changes in speed content despite matched marginals, consistent with receptive fields differentiating along the speed gradient. Together, we have uncovered a structured and distributed code for motion in high-level visual cortex, and provide mechanistic insights into how the brain parses complex motion in natural scenes.
- Open Access
- 23 September 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Hierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic ‘foreground’ sounds are often masked by more stationary ‘background’ sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging, which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.
- 23 September 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Open AccessHierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic ‘foreground’ sounds are often masked by more stationary ‘background’ sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging, which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.
- Open Access
- 5 September 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (9) , e1013271
The detection of algebraic auditory structures emerges with self-supervised learning
Humans can spontaneously detect complex algebraic structures. Historically, two opposing views explain this ability, at the root of language and music acquisition. Some argue for the existence of an innate and specific mechanism. Others argue that this ability emerges from experience: i.e. when generic learning principles continuously process sensory inputs. These two views, however, remain difficult to test experimentally. Here, we use deep learning models to evaluate the factors that lead to the spontaneous detection of algebraic structures in the auditory modality. Specifically, we use self-supervised learning to train multiple deep-learning models with a variable amount of either natural (environmental sounds) and/or cultural sounds (speech or music) to evaluate the impact of these stimuli. We then expose these models to the experimental paradigms classically used to evaluate the processing of algebraic structures. Like humans, these models spontaneously detect repeated sequences, probabilistic chunks, and complex algebraic structures. Also like humans, this ability diminishes with structure complexity. Importantly, this ability can emerge from experience alone: the more the models are exposed to natural sounds, the more they spontaneously detect increasingly complex structures. Finally, this ability does not emerge in models pretrained only on speech, and emerges more rapidly in models pretrained with music than environmental sounds. Overall, our study provides an operational framework to clarify sufficient built-in and acquired principles that model human’s advanced capacity to detect algebraic structures in sounds.
- 5 September 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (9) , e1013271
Open AccessThe detection of algebraic auditory structures emerges with self-supervised learning
Humans can spontaneously detect complex algebraic structures. Historically, two opposing views explain this ability, at the root of language and music acquisition. Some argue for the existence of an innate and specific mechanism. Others argue that this ability emerges from experience: i.e. when generic learning principles continuously process sensory inputs. These two views, however, remain difficult to test experimentally. Here, we use deep learning models to evaluate the factors that lead to the spontaneous detection of algebraic structures in the auditory modality. Specifically, we use self-supervised learning to train multiple deep-learning models with a variable amount of either natural (environmental sounds) and/or cultural sounds (speech or music) to evaluate the impact of these stimuli. We then expose these models to the experimental paradigms classically used to evaluate the processing of algebraic structures. Like humans, these models spontaneously detect repeated sequences, probabilistic chunks, and complex algebraic structures. Also like humans, this ability diminishes with structure complexity. Importantly, this ability can emerge from experience alone: the more the models are exposed to natural sounds, the more they spontaneously detect increasingly complex structures. Finally, this ability does not emerge in models pretrained only on speech, and emerges more rapidly in models pretrained with music than environmental sounds. Overall, our study provides an operational framework to clarify sufficient built-in and acquired principles that model human’s advanced capacity to detect algebraic structures in sounds.
- Open Access
- 21 August 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
Hierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic "foreground" sounds are often masked by more stationary "background" sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging (fUSI), which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets, this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.
- 21 August 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
Open AccessHierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic "foreground" sounds are often masked by more stationary "background" sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging (fUSI), which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets, this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.
- Open Access
Hierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic "foreground" sounds are often masked by more stationary "background" sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging (fUSI), which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets, this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.
Open AccessHierarchical encoding of natural sound mixtures in ferret auditory cortex
Extracting relevant auditory signals from complex natural scenes is a fundamental challenge for the auditory system. Sounds from multiple sources overlap in time and frequency. In particular, dynamic "foreground" sounds are often masked by more stationary "background" sounds. Human auditory cortex exhibits a hierarchical organization where background-invariant representations are progressively enhanced along the processing stream, from primary to non-primary regions. However, we do not know whether this organizational principle is conserved across species and which neural mechanisms drive this invariance. To address these questions, we investigated background invariance in ferret auditory cortex using functional ultrasound imaging (fUSI), which enables large-scale, high-resolution recordings of hemodynamic responses. We measured responses across primary, secondary, and tertiary auditory cortical regions as ferrets passively listened to mixtures of natural sounds and their components in isolation. We found a hierarchical gradient of background invariance, mirroring findings in humans: responses in primary auditory cortex reflected contributions from both foreground and background sounds, while background invariance increased in higher-order auditory regions. Using a spectrotemporal filter-bank model, we found that in ferrets, this hierarchical structure could be largely explained by tuning to low-order acoustic features. However, this model failed to fully account for background invariance in human non-primary auditory cortex, suggesting that additional, higher-order mechanisms are crucial for background segregation in humans.