LBLucas Benjamin

Lucas Benjamin

*This is an algorithmically generated author record based on publicly available information.

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


Lucas Benjamin has a total of 62 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Flexible compression of sensory evidence makes human decisions more resilient to noisy neural computations

    by Julie Drevet, Lucas Benjamin, Lindsay Rondot, Jan Drugowitsch, Valentin Wyart

    Categorizing ambiguous stimuli requires accumulating sensory evidence over time through noisy neural computations. Theories disagree on whether this evidence is accumulated in its native sensory space or in an abstract and compressed category space that retains only decision-relevant information. Here we designed a visual categorization task in which revealing decision categories before or after a stimulus sequence determines which space can be used to accumulate evidence, without changing the optimal decision. Humans were more accurate, and their inferences less noisy, when they could accumulate evidence in category space. Concurrent magnetoencephalographic (MEG) recordings revealed flexible and accuracy-predictive representations of evidence compressed onto decision categories in these conditions. Recurrent neural networks (RNNs) trained on the same task developed the same strategy, which made their decisions more resilient to noisy computations. Together, these findings suggest that the human brain mitigates the cost of noisy inferences by focusing limited computational resources on decision-relevant information.

  • Open Access

    Neural abstraction of evidence accumulation enables flexible multidimensional perceptual decisions

    by Lucas Benjamin, Julie Drevet, Valentin Wyart

    Human perceptual decisions adapt to rich and changing contexts but are typically tested under impoverished conditions. To probe this flexibility, we examined decisions requiring both the selection and the combination of multiple sensory features across sequences of objects. Here we show that human decision flexibility extends beyond feature selection to combination, even across sensory modalities, supported by the abstraction of underlying neural representations. Rapid magnetoencephalographic (MEG) responses reveal a hierarchy of object representations that become increasingly abstract and gated according to their decision relevance. A shared abstract representation of updates reflects evidence accumulation collapsed onto the relevant axis across unidimensional (selection-based) and multidimensional (combination-based) contexts. Recurrent neural networks tested under the same conditions acquire and benefit from abstract representations only when operating under human levels of computational noise. Together, these findings provide a ‘flexibility through abstraction’ account of human decisions across a broader range of everyday perceptual conditions.


  • Open Access

    Infants at high and low likelihood for autism show different EEG developmental trajectories in speech tracking and statistical learning

    by Michel Godel, Ana Fló, Lucas Benjamin, Ghislaine Dehaene-Lambertz, Marie Schaer

    Delayed onset of canonical babbling and first words is often reported in infants later diagnosed with autism spectrum disorder. Identifying the neural mechanisms underlying language acquisition in autism is therefore critical to inform early diagnosis, prognosis, and intervention strategies. In this study, we investigated two speech processing mechanisms previously identified as atypical in children and adults with autism: the neural ability to track syllables, and statistical learning, the capacity to detect speech regularities beneath surface variability. We recorded 83 longitudinal high-density electroencephalograms from 44 infants (2.5–22.6 months) at high (HL) and low (LL) likelihood for autism and assessed their verbal outcomes at 20 months. Neural entrainment was measured at syllable and word frequencies during exposure to a multi-speaker stream of concatenated tri-syllabic words, followed by a word recognition test using evoked response potential (ERP) recording. Our findings revealed reduced tracking abilities at the syllabic level in HL infants, a measure that correlated with verbal outcomes. While HL infants did not exhibit deficits in statistical learning itself, they displayed reduced novelty orientation during the word recognition test, indicated by a reduced late ERP. By contrast, multi-talker variability temporarily disrupted word segmentation around 12 months in LL infants, but not in HL infants, potentially reflecting decreased sensitivity to human voices variability in the HL group. These results emphasize the importance of longitudinal protocols employing online, implicit measures to track the hierarchical stages of speech processing in both HL and LL infants.

  • Open Access

    Long-horizon associative learning as a unifying framework for statistical learning across scales

    by Lucas Benjamin, Ana Fló, Fosca Al Roumi, Ghislaine Dehaene-Lambertz

    Sensory inputs are rich with temporal patterns that unfold across multiple timescales. Uncovering these regularities is essential for anticipating future events and navigating the environment efficiently. Numerous models have been proposed to account for learning at specific temporal scales; however, they are often designed in isolation and rely on narrowly tuned statistical measures, limiting their generalizability to other paradigms. In contrast, humans typically learn without prior knowledge of the underlying structure or the relevant timescale at which regularities occur. Here, we present a unifying account of statistical learning that spans a wide range of temporal dependencies, from adjacent and nonadjacent transitions to complex network structures. This model, long-horizon associative learning, offers a biologically grounded implementation of the successor representation, or equivalently, the free energy minimization model. Reanalyzing data from 11 previously published studies, we show that a single neural mechanism captures both local statistical regularities and higher-order structural properties. This mechanism rests on graded temporal overlap of associative traces and is governed by a single free parameter (β). This initial domain-general associative learning process, emerging from the graded structure of associations, may later scaffold to higher-level operations such as grouping, categorization, rule abstraction, and memory formation. Overall, this framework offers a conceptual synthesis that bridges disparate strands of the statistical learning literature and reframes apparent paradigm-specific effects as different expressions of a common underlying computation.


  • Open Access

    An ordinal Language of Thought supports human memory for regular sequences

    by Elyes Tabbane, Santiago Figueira, Lucas Benjamin, Stanislas Dehaene, Fosca Al Roumi

    How do humans store sequences that far exceed working memory capacity? Using visuo-spatial and binary auditory sequences, we previously showed that a Language of Thought (LoT) architecture, in which simple primitives are recursively combined into hierarchical programs, enables efficient storage of structured sequences. Here we ask whether this principle extends to purely ordinal structure: sequences defined by how items repeat and in what order, as in AABBCCAABBCC, independently of their spatial content. Across three experiments, participants reproduced 12-item sequences of spatial locations with various ordinal structures. The minimal description length derived from the LoT model predicted recall accuracy with remarkable precision (r = .96), substantially outperforming Shannon entropy, Lempel-Ziv complexity, chunking models and subjective complexity ratings. Critically, fine-grained analyses of participants' inter-click intervals during reproduction revealed systematic slowdowns at the hierarchical boundaries predicted by the LoT programs, providing a behavioral signature of the underlying mental syntax. These results identify a compact vocabulary of mental primitives, repetition, mirroring, and interleaving, whose composition accounts for the symbolic compression of ordinal structures. For ordinal regularities, human sequence memory operates as a form of program induction, leveraging a domain-general capacity for hierarchical compression to encode complex structured information.

  • Human sensitivity to statistical regularities at different scales in auditory sequences : evidence from electrophysiology and behavior in adults and neonates

    by Lucas Benjamin

    Perception humaine des régularités statistiques à différentes échelles dans les séquences auditives : études comportementales et en électrophysiologie chez l’adulte et le nouveau-né: Comment un cerveau parvient-il à extraire des informations de l'environnement et à y découvrir des structures pertinentes, a fortiori lorsqu'il est immature et peu expérimenté, comme peut l'être le cerveau d'un nouveau-né ? La question de savoir comment un nouveau-né découvre la structure de la parole, la complexité des structures qu'il peut extraire et les mécanismes cérébraux dont il dispose pour le faire, a motivé notre travail. Nous avons étudié les capacités d'apprentissage de nouveau-nés endormis et comparé leurs performances à celles d'adultes éveillés. Sur la base d'un ensemble de résultats, nous soutenons qu'un grand nombre de structures séquentielles couvrant différentes échelles de complexité pourraient être traitées par un seul calcul cérébral basé sur le principe de l’apprentissage associatif. Le cadre que nous proposons ici rassemble des résultats jusqu'à présent considérés comme des compétences distinctes : d'une part "l'apprentissage statistique" et d'autre part "l'apprentissage de réseaux". Pour tester les propriétés et les limites de ce cadre, nous présentons une série d'expériences d’apprentissage séquentiel à différentes échelles et dans différentes populations : des patients comateux aux adultes éveillés en passant par les nouveau-nés endormis. Nous avons utilisé des mesures comportementales, des approches de modélisation et des techniques de neuro-imagerie telles que l'électroencéphalographie (EEG) et la magnétoencéphalographie. Tout d'abord, en utilisant l'EEG pour évaluer l'apprentissage de populations non verbales, nous avons étudié l’aptitude de nouveau-nés endormis et de patients dans le coma à percevoir des associations simples entre des syllabes consécutives dans une séquence de parole continue. L’hypothèse a été faite que cette capacité est un mécanisme majeur permettant aux bébés d’extraire les mots d’un flot de parole continu. En effet, les adultes éveillés, les nouveau-nés endormis et même certains patients dans le coma montrent des signes comparables d'apprentissage de ces dépendances locales, ce qui plaide en faveur d'un processus hautement automatique ne nécessitant pas d'attention explicite, et qui serait fonctionnel dès la naissance. En outre, les adultes éveillés et les nouveau-nés endormis ont montré des limitations similaires quant à la longueur des mots possibles à extraire, ce qui démontre que malgré l'immaturité du cerveau du nourrisson, les propriétés et les limitations de l'apprentissage statistique dans ce contexte sont très similaires à celles d'un cerveau adulte mature. Deuxièmement, nous avons proposé un nouveau dessin expérimental, le paradigme de ‘communauté incomplète’, pour étudier simultanément les apprentissages dits statistiques et ceux de graphe dans une unique séquence. Les réponses comportementales d’adultes ont montré que les humains complétaient leur représentation des structures, malgré une présentation incomplète. Des approches de modélisation nous ont permis d'émettre l'hypothèse que ces différents ordres d’apprentissage des séquences relèveraient en fait d’un seul et même processus cognitif. L'exploration MEG sur ce paradigme a révélé des corrélats neuronaux compatibles avec cette hypothèse, montrant que l'apprentissage statistique local et l'apprentissage de graphes pouvaient être sous-tendus par un unique mécanisme d’apprentissage associatif. Enfin, nous avons passé en revue la littérature sur l'apprentissage statistique et l'apprentissage de réseaux, et proposé un nouveau cadre unifiant ces différentes échelles d'apprentissage sous un principe commun. Nous avons montré que la plupart des résultats précédemment rapportés dans la littérature, avec une variété de modèles à considérer, peuvent correctement s’expliquer par un seul modèle d'apprentissage associatif. Cette hypothèse fournit une explication plus parcimonieuse des mécanismes permettant aux humains d’extraire une structure statistique et fait de nombreuses nouvelles prédictions.

  • Open Access

    Sensory integration, temporal prediction, and rule discovery reflect interdependent inference processes

    by Lucas Benjamin, Benjamin Morillon, Valentin Wyart

    Deciphering the structure of variable sensory input is key to building an accurate model of one’s environment. Humans can accumulate evidence from sequences of stimuli to estimate their sensory statistics, predict the timing of upcoming stimuli, but also discover rules governing sequence generation. However, whether these three forms of inference operate independently or synergistically remains untested. Here, we report selective interactions between sensory integration, temporal prediction, and rule discovery in humans. Participants were exposed to rhythmic sequences of 10 stimuli governed or not by a latent rule—a predictable change in stimulus statistics after five stimuli—and then asked to predict the 10th stimulus from incomplete sequences. Individual differences in sensory integration timescale for rule-free sequences predicted efficient rule discovery. Conversely, discovering the latent rule shaped the timescale and format of sensory integration for rule-based sequences. Tampering with the rhythmicity of stimulus presentation impaired rule discovery without affecting sensory integration accuracy. Selective perturbations of recurrent neural networks trained in the same conditions confirmed these specific interactions. Together, these findings provide insights into the flexibility of human inferences based on variable yet predictable sensory input.


  • Open Access

    A geometric shape regularity effect in the human brain

    by Mathias Sablé-Meyer, Lucas Benjamin, Cassandra Potier Watkins, Chenxi He, Maxence Pajot, Théo Morfoisse, Fosca Al Roumi, Stanislas Dehaene

    The perception and production of regular geometric shapes, a characteristic trait of human cultures since prehistory, has unknown neural mechanisms. Behavioral studies suggest that humans are attuned to discrete regularities such as symmetries and parallelism and rely on their combinations to encode regular geometric shapes in a compressed form. To identify the brain systems underlying this ability, as well as their dynamics, we collected functional MRI in both adults and 6-year-olds, and magnetoencephalography data in adults, during the perception of simple shapes such as hexagons, triangles, and quadrilaterals. The results revealed that geometric shapes, relative to other visual categories, induce a hypoactivation of ventral visual areas and an overactivation of the intraparietal and inferior temporal regions also involved in mathematical processing, whose activation is modulated by geometric regularity. While convolutional neural networks captured the early visual activity evoked by geometric shapes, they failed to account for subsequent dorsal parietal and prefrontal signals, which could only be captured by discrete geometric features or by bigger deep-learning models of vision. We propose that the perception of abstract geometric regularities engages an additional symbolic mode of visual perception.


  • Open Access

    A geometric shape regularity effect in the human brain

    by Mathias Sablé-Meyer, Lucas Benjamin, Cassandra Potier Watkins, Chenxi He, Maxence Pajot, Théo Morfoisse, Fosca Al Roumi, Stanislas Dehaene

    The perception and production of regular geometric shapes, a characteristic trait of human cultures since prehistory, has unknown neural mechanisms. Behavioral studies suggest that humans are attuned to discrete regularities such as symmetries and parallelism, and rely on their combinations to encode regular geometric shapes in a compressed form. To identify the brain systems underlying this ability, as well as their dynamics, we collected functional MRI in both adults and six-year-olds, and magnetoencephalography data in adults, during the perception of simple shapes such as hexagons, triangles and quadrilaterals. The results revealed that geometric shapes, relative to other visual categories, induce a hypoactivation of ventral visual areas and an overactivation of the intraparietal and inferior temporal regions also involved in mathematical processing, whose activation is modulated by geometric regularity. While convolutional neural networks captured the early visual activity evoked by geometric shapes, they failed to account for subsequent dorsal parietal and prefrontal signals, which could only be captured by discrete geometric features or by bigger deep-learning models of vision. We propose that the perception of abstract geometric regularities engages an additional symbolic mode of visual perception.


  • Open Access

    From Syllables to Words: EEG Evidence of Different Age Trajectories in Speech Tracking and Statistical Learning in Infants at High and Low Likelihood for Autism

    by Michel Godel, Ana Fló, Lucas Benjamin, Ghislaine Dehaene-Lambertz, Marie Schaer

    Delayed onset of canonical babbling and first words is often reported in infants later diagnosed with autism spectrum disorder. Identifying the neural mechanisms underlying language acquisition in autism is therefore critical to inform early diagnosis, prognosis, and intervention strategies. In this study, we investigated two speech processing mechanisms previously identified as atypical in children and adults with autism: the neural ability to track syllables; and statistical learning, the capacity to detect speech regularities beneath surface variability. We recorded 83 longitudinal high-density EEGs from 44 infants (2.5–22.6 months) at high (HL) and low (LL) likelihood for autism and assessed their verbal outcomes at 20 months. Neural entrainment was measured at syllable and word frequencies during exposure to a multi-speaker stream of concatenated tri-syllabic words, followed by a word recognition test using ERP recording. Our findings revealed reduced tracking abilities at the syllabic level in HL infants, a measure that correlated with verbal outcomes. While HL infants did not exhibit deficits in statistical learning itself, they displayed reduced novelty orientation during the word recognition test, indicated by a reduced late ERP. By contrast, multi-talker variability temporarily disrupted word segmentation around 12 months in LL infants, but not in HL infants, potentially reflecting decreased sensitivity to human voices variability in the HL group. These results emphasize the importance of longitudinal protocols employing online, implicit measures to track the hierarchical stages of speech processing in both HL and LL infants.