Evaluating the impact of short educational videos on the cortical networks for mathematics.

Many teaching websites, such as the Khan Academy, propose vivid videos illustrating a mathematical concept. Using functional magnetic resonance imaging, we asked whether watching such a video suffices to rapidly change the brain networks for mathematical knowledge. We capitalized on the finding that, when judging the truth of short spoken statements, distinct semantic regions activate depending on whether the statements bear on mathematical knowledge or on other domains of semantic knowledge. Here, participants answered such questions before and after watching a lively 5-min video, which taught them the rudiments of a new domain. During the video, a distinct math-responsive network, comprising anterior intraparietal and inferior temporal nodes, showed intersubject synchrony when viewing mathematics course rather than control courses in biology or law. However, this experience led to minimal subsequent changes in the activity of those domain-specific areas when answering questions on the same topics a few minutes later. All taught facts, whether mathematical or not, led to domain-general repetition enhancement, particularly prominent in the cuneus, posterior cingulate, and posterior parietal cortices. We conclude that short videos do not suffice to induce a meaningful lasting change in the brain's math-responsive network, but merely engage domain-general regions possibly involved in episodic short-term memory.

Spatiotemporally distributed frontotemporal networks for sentence reading.

Reading a sentence entails integrating the meanings of individual words to infer more complex, higher-order meaning. This highly rapid and complex human behavior is known to engage the inferior frontal gyrus (IFG) and middle temporal gyrus (MTG) in the language-dominant hemisphere, yet whether there are distinct contributions of these regions to sentence reading is still unclear. To probe these neural spatiotemporal dynamics, we used direct intracranial recordings to measure neural activity while reading sentences, meaning-deficient Jabberwocky sentences, and lists of words or pseudowords. We isolated two functionally and spatiotemporally distinct frontotemporal networks, each sensitive to distinct aspects of word and sentence composition. The first distributed network engages the IFG and MTG, with IFG activity preceding MTG. Activity in this network ramps up over the duration of a sentence and is reduced or absent during Jabberwocky and word lists, implying its role in the derivation of sentence-level meaning. The second network engages the superior temporal gyrus and the IFG, with temporal responses leading those in frontal lobe, and shows greater activation for each word in a list than those in sentences, suggesting that sentential context enables greater efficiency in the lexical and/or phonological processing of individual words. These adjacent, yet spatiotemporally dissociable neural mechanisms for word- and sentence-level processes shed light on the richly layered semantic networks that enable us to fluently read. These results imply distributed, dynamic computation across the frontotemporal language network rather than a clear dichotomy between the contributions of frontal and temporal structures.

Dimensionality and Ramping: Signatures of Sentence Integration in the Dynamics of Brains and Deep Language Models.

A sentence is more than the sum of its words: its meaning depends on how they combine with one another. The brain mechanisms underlying such semantic composition remain poorly understood. To shed light on the neural vector code underlying semantic composition, we introduce two hypotheses: (1) the intrinsic dimensionality of the space of neural representations should increase as a sentence unfolds, paralleling the growing complexity of its semantic representation; and (2) this progressive integration should be reflected in ramping and sentence-final signals. To test these predictions, we designed a dataset of closely matched normal and jabberwocky sentences (composed of meaningless pseudo words) and displayed them to deep language models and to 11 human participants (5 men and 6 women) monitored with simultaneous MEG and intracranial EEG. In both deep language models and electrophysiological data, we found that representational dimensionality was higher for meaningful sentences than jabberwocky. Furthermore, multivariate decoding of normal versus jabberwocky confirmed three dynamic patterns: (1) a phasic pattern following each word, peaking in temporal and parietal areas; (2) a ramping pattern, characteristic of bilateral inferior and middle frontal gyri; and (3) a sentence-final pattern in left superior frontal gyrus and right orbitofrontal cortex. These results provide a first glimpse into the neural geometry of semantic integration and constrain the search for a neural code of linguistic composition.SIGNIFICANCE STATEMENT Starting from general linguistic concepts, we make two sets of predictions in neural signals evoked by reading multiword sentences. First, the intrinsic dimensionality of the representation should grow with additional meaningful words. Second, the neural dynamics should exhibit signatures of encoding, maintaining, and resolving semantic composition. We successfully validated these hypotheses in deep neural language models, artificial neural networks trained on text and performing very well on many natural language processing tasks. Then, using a unique combination of MEG and intracranial electrodes, we recorded high-resolution brain data from human participants while they read a controlled set of sentences. Time-resolved dimensionality analysis showed increasing dimensionality with meaning, and multivariate decoding allowed us to isolate the three dynamical patterns we had hypothesized.

Sensitivity to geometric shape regularity in humans and baboons: A putative signature of human singularity.

Among primates, humans are special in their ability to create and manipulate highly elaborate structures of language, mathematics, and music. Here we show that this sensitivity to abstract structure is already present in a much simpler domain: the visual perception of regular geometric shapes such as squares, rectangles, and parallelograms. We asked human subjects to detect an intruder shape among six quadrilaterals. Although the intruder was always defined by an identical amount of displacement of a single vertex, the results revealed a geometric regularity effect: detection was considerably easier when either the base shape or the intruder was a regular figure comprising right angles, parallelism, or symmetry rather than a more irregular shape. This effect was replicated in several tasks and in all human populations tested, including uneducated Himba adults and French kindergartners. Baboons, however, showed no such geometric regularity effect, even after extensive training. Baboon behavior was captured by convolutional neural networks (CNNs), but neither CNNs nor a variational autoencoder captured the human geometric regularity effect. However, a symbolic model, based on exact properties of Euclidean geometry, closely fitted human behavior. Our results indicate that the human propensity for symbolic abstraction permeates even elementary shape perception. They suggest a putative signature of human singularity and provide a challenge for nonsymbolic models of human shape perception.

A Spatiotemporal Map of Reading Aloud.

Reading words aloud is a fundamental aspect of literacy. The rapid rate at which multiple distributed neural substrates are engaged in this process can only be probed via techniques with high spatiotemporal resolution. We probed this with direct intracranial recordings covering most of the left hemisphere in 46 humans (26 male, 20 female) as they read aloud regular, exception and pseudo-words. We used this to create a spatiotemporal map of word processing and to derive how broadband γ activity varies with multiple word attributes critical to reading speed: lexicality, word frequency, and orthographic neighborhood. We found that lexicality is encoded earliest in mid-fusiform (mFus) cortex, and precentral sulcus, and is represented reliably enough to allow single-trial lexicality decoding. Word frequency is first represented in mFus and later in the inferior frontal gyrus (IFG) and inferior parietal sulcus (IPS), while orthographic neighborhood sensitivity resides solely in IPS. We thus isolate the neural correlates of the distributed reading network involving mFus, IFG, IPS, precentral sulcus, and motor cortex and provide direct evidence for parallel processes via the lexical route from mFus to IFG, and the sublexical route from IPS and precentral sulcus to anterior IFG.SIGNIFICANCE STATEMENT Reading aloud depends on multiple complex cerebral computations: mapping from a written letter string on a page to a sequence of spoken sound representations. Here, we used direct intracranial recordings in a large cohort while they read aloud known and novel words, to track, across space and time, the progression of neural representations of behaviorally relevant factors that govern reading speed. We find, concordant with cognitive models of reading, that known and novel words are differentially processed through a lexical route, sensitive to frequency of occurrence of known words in natural language, and a sublexical route, performing letter-by-letter construction of novel words.

Working Memory for Spatial Sequences: Developmental and Evolutionary Factors in Encoding Ordinal and Relational Structures.

Sequence learning is a ubiquitous facet of human and animal cognition. Here, using a common sequence reproduction task, we investigated whether and how the ordinal and relational structures linking consecutive elements are acquired by human adults, children, and macaque monkeys. While children and monkeys exhibited significantly lower precision than adults for spatial location and temporal order information, only monkeys appeared to exceedingly focus on the first item. Most importantly, only humans, regardless of age, spontaneously extracted the spatial relations between consecutive items and used a chunking strategy to compress sequences in working memory. Monkeys did not detect such relational structures, even after extensive training. Monkey behavior was captured by a conjunctive coding model, whereas a chunk-based conjunctive model explained more variance in humans. These age- and species-related differences are indicative of developmental and evolutionary mechanisms of sequence encoding and may provide novel insights into the uniquely human cognitive capacities.SIGNIFICANCE STATEMENT Sequence learning, the ability to encode the order of discrete elements and their relationships presented within a sequence, is a ubiquitous facet of cognition among humans and animals. By exploring sequence-processing abilities at different human developmental stages and in nonhuman primates, we found that only humans, regardless of age, spontaneously extracted the spatial relations between consecutive items and used an internal language to compress sequences in working memory. The findings provided insights into understanding the origins of sequence capabilities in humans and how they evolve through development to identify the unique aspects of human cognitive capacity, which includes the comprehension, learning, and production of sequences, and perhaps, above all, language processing.

Does word flickering improve reading? Negative evidence from four experiments using low and high frequencies.

Does word flickering facilitate reading? Despite a lack of scientific evidence, flickering glasses and lamps for dyslexia are being marketed in various countries. We conducted four experiments to assess their efficacy. Two experiments involved a computerized lexical decision task with constant display or low-frequency flickering (10 or 15 Hz). Among 375 regular adult readers, flicker noticeably slowed down word recognition, while slightly biasing the decision towards pseudowords. No significant effect was observed in 20 dyslexic children. In 22 dyslexic children, we also evaluated the impact of the Lexilight lamp and Lexilens glasses, which operate at higher frequencies, on reading fluency, letter identification and mirror letter processing. No detectable impact was observed. Lastly, in two participants who claimed to benefit from flickering glasses, we orthogonally manipulated whether the glasses were actually on, and whether the participant thought they were on. Only a small placebo effect was noted in one participant. Our findings starkly contrast with marketing claims that these tools can help 90% of dyslexics, and emphasize the role of rigorous scientific research in empowering dyslexic individuals to make informed decisions.

Characterizing different types of developmental dyslexias in French: The Malabi screener.

Reading is a complex process involving multiple stages. An impairment in any of these stages may cause distinct types of reading deficits- distinct types of dyslexia. We describe the Malabi, a screener to identify deficits in various orthographic, lexical, and sublexical components of the reading process in French. The Malabi utilizes stimuli that are sensitive to different forms of dyslexia, including "attentional dyslexia", as it is traditionally refered to, characterized by letter-to-word binding impairments leading to letter migrations between words (e.g., "bar cat" misread as "bat car"), and "letter-position dyslexia", resulting in letter transpositions within words (e.g., "destiny" misread as "density"). After collecting reading error norms from 138 French middle-school students, we analyzed error types of 16 students with developmental dyslexia. We identified three selective cases of attentional dyslexia and one case of letter-position dyslexia. Further tests confirmed our diagnosis and demonstrate, for the first time, how these dyslexias are manifested in French. These results underscore the significance of recognizing and discussing the existence of multiple dyslexias, both in research contexts when selecting participants for dyslexia studies, and in practical settings where educators and practitioners work with students to develop personalized support. The test and supporting materials are available on Open Science Framework (https://osf.io/3pgzb/).

Disruption of Conscious Access in Psychosis Is Associated with Altered Structural Brain Connectivity.

According to global neuronal workspace (GNW) theory, conscious access relies on long-distance cerebral connectivity to allow a global neuronal ignition coding for conscious content. In patients with schizophrenia and bipolar disorder, both alterations in cerebral connectivity and an increased threshold for conscious perception have been reported. The implications of abnormal structural connectivity for disrupted conscious access and the relationship between these two deficits and psychopathology remain unclear. The aim of this study was to determine the extent to which structural connectivity is correlated with consciousness threshold, particularly in psychosis. We used a visual masking paradigm to measure consciousness threshold, and diffusion MRI tractography to assess structural connectivity in 97 humans of either sex with varying degrees of psychosis: healthy control subjects (n = 46), schizophrenia patients (n = 25), and bipolar disorder patients with (n = 17) and without (n = 9) a history of psychosis. Patients with psychosis (schizophrenia and bipolar disorder with psychotic features) had an elevated masking threshold compared with control subjects and bipolar disorder patients without psychotic features. Masking threshold correlated negatively with the mean general fractional anisotropy of white matter tracts exclusively within the GNW network (inferior frontal-occipital fasciculus, cingulum, and corpus callosum). Mediation analysis demonstrated that alterations in long-distance connectivity were associated with an increased masking threshold, which in turn was linked to psychotic symptoms. Our findings support the hypothesis that long-distance structural connectivity within the GNW plays a crucial role in conscious access, and that conscious access may mediate the association between impaired structural connectivity and psychosis.

Evolution of reading and face circuits during the first three years of reading acquisition.

Although words and faces activate neighboring regions in the fusiform gyrus, we lack an understanding of how such category selectivity emerges during development. To investigate the organization of reading and face circuits at the earliest stage of reading acquisition, we measured the fMRI responses to words, faces, houses, and checkerboards in three groups of 60 French children: 6-year-old pre-readers, 6-year-old beginning readers and 9-year-old advanced readers. The results showed that specific responses to written words were absent prior to reading, but emerged in beginning readers, irrespective of age. Likewise, specific responses to faces were barely visible in pre-readers and continued to evolve in the 9-year-olds, yet primarily driven by age rather than by schooling. Crucially, the sectors of ventral visual cortex that become specialized for words and faces harbored their own functional connectivity prior to reading acquisition: the VWFA with left-hemispheric spoken language areas, and the FFA with the contralateral region and the amygdalae. The results support the view that reading acquisition occurs through the recycling of a pre-existing but plastic circuit which, in pre-readers, already connects the VWFA site to other distant language areas. We argue that reading acquisition does not compete with the face system directly, through a pruning of preexisting face responses, but indirectly, by hindering the slow growth of face responses in the left hemisphere, thus increasing a pre-existing right hemispheric bias.

A theory of memory for binary sequences: Evidence for a mental compression algorithm in humans.

Working memory capacity can be improved by recoding the memorized information in a condensed form. Here, we tested the theory that human adults encode binary sequences of stimuli in memory using an abstract internal language and a recursive compression algorithm. The theory predicts that the psychological complexity of a given sequence should be proportional to the length of its shortest description in the proposed language, which can capture any nested pattern of repetitions and alternations using a limited number of instructions. Five experiments examine the capacity of the theory to predict human adults' memory for a variety of auditory and visual sequences. We probed memory using a sequence violation paradigm in which participants attempted to detect occasional violations in an otherwise fixed sequence. Both subjective complexity ratings and objective violation detection performance were well predicted by our theoretical measure of complexity, which simply reflects a weighted sum of the number of elementary instructions and digits in the shortest formula that captures the sequence in our language. While a simpler transition probability model, when tested as a single predictor in the statistical analyses, accounted for significant variance in the data, the goodness-of-fit with the data significantly improved when the language-based complexity measure was included in the statistical model, while the variance explained by the transition probability model largely decreased. Model comparison also showed that shortest description length in a recursive language provides a better fit than six alternative previously proposed models of sequence encoding. The data support the hypothesis that, beyond the extraction of statistical knowledge, human sequence coding relies on an internal compression using language-like nested structures.

A language of thought for the mental representation of geometric shapes.

In various cultures and at all spatial scales, humans produce a rich complexity of geometric shapes such as lines, circles or spirals. Here, we propose that humans possess a language of thought for geometric shapes that can produce line drawings as recursive combinations of a minimal set of geometric primitives. We present a programming language, similar to Logo, that combines discrete numbers and continuous integration to form higher-level structures based on repetition, concatenation and embedding, and we show that the simplest programs in this language generate the fundamental geometric shapes observed in human cultures. On the perceptual side, we propose that shape perception in humans involves searching for the shortest program that correctly draws the image (program induction). A consequence of this framework is that the mental difficulty of remembering a shape should depend on its minimum description length (MDL) in the proposed language. In two experiments, we show that encoding and processing of geometric shapes is well predicted by MDL. Furthermore, our hypotheses predict additive laws for the psychological complexity of repeated, concatenated or embedded shapes, which we confirm experimentally.

A mesial-to-lateral dissociation for orthographic processing in the visual cortex.

Efficient reading requires a fast conversion of the written word to both phonological and semantic codes. We tested the hypothesis that, within the left occipitotemporal cortical regions involved in visual word recognition, distinct subregions harbor slightly different orthographic codes adapted to those 2 functions. While the lexico-semantic pathway may operate on letter or open-bigram information, the phonological pathway requires the identification of multiletter graphemes such as "ch" or "ou" in order to map them onto phonemes. To evaluate the existence of a specific stage of graphemic encoding, 20 adults performed lexical decision and naming tasks on words and pseudowords during functional MRI. Graphemic encoding was facilitated or disrupted by coloring and spacing the letters either congruently with multiletter graphemes (ch-ai-r) or incongruently with them (c-ha-ir). This manipulation affected behavior, primarily during the naming of pseudowords, and modulated brain activity in the left midfusiform sulcus, at a site medial to the classical visual word form area (VWFA). This putative grapheme-related area (GRA) differed from the VWFA in being preferentially connected functionally to dorsal parietal areas involved in letter-by-letter reading, while the VWFA showed effects of lexicality and spelling-to-sound regularity. Our results suggest a partial dissociation within left occipitotemporal cortex: the midfusiform GRA would encode orthographic information at a sublexical graphemic level, while the lateral occipitotemporal VWFA would contribute primarily to direct lexico-semantic access.

Probing the limits of activity-silent non-conscious working memory.

Two types of working memory (WM) have recently been proposed: (i) active WM, relying on sustained neural firing, and (ii) activity-silent WM, for which firing returns to baseline, yet memories may be retained by short-term synaptic changes. Activity-silent WM in particular might also underlie the recently discovered phenomenon of non-conscious WM, which permits even subliminal stimuli to be stored for several seconds. However, whether both states support identical forms of information processing is unknown. Theory predicts that activity-silent states are confined to passive storage and cannot operate on stored information. To determine whether an explicit reactivation is required before the manipulation of information in WM, we evaluated whether participants could mentally rotate brief visual stimuli of variable subjective visibility. Behaviorally, even for unseen targets, subjects reported the rotated location above chance after several seconds. As predicted, however, at the time of mental rotation, such blindsight performance was accompanied by (i) neural signatures of consciousness in the form of a sustained desynchronization in alpha/beta frequency and (ii) a reactivation of the memorized information as indicated by decodable representations of participants' guess and response. Our findings challenge the concept of genuine non-conscious "working" memory, argue that activity-silent states merely support passive short-term memory, and provide a cautionary note for purely behavioral studies of non-conscious information processing.

Subject-specific segregation of functional territories based on deep phenotyping.

Functional magnetic resonance imaging (fMRI) has opened the possibility to investigate how brain activity is modulated by behavior. Most studies so far are bound to one single task, in which functional responses to a handful of contrasts are analyzed and reported as a group average brain map. Contrariwise, recent data-collection efforts have started to target a systematic spatial representation of multiple mental functions. In this paper, we leverage the Individual Brain Charting (IBC) dataset-a high-resolution task-fMRI dataset acquired in a fixed environment-in order to study the feasibility of individual mapping. First, we verify that the IBC brain maps reproduce those obtained from previous, large-scale datasets using the same tasks. Second, we confirm that the elementary spatial components, inferred across all tasks, are consistently mapped within and, to a lesser extent, across participants. Third, we demonstrate the relevance of the topographic information of the individual contrast maps, showing that contrasts from one task can be predicted by contrasts from other tasks. At last, we showcase the benefit of contrast accumulation for the fine functional characterization of brain regions within a prespecified network. To this end, we analyze the cognitive profile of functional territories pertaining to the language network and prove that these profiles generalize across participants.

The emergence of the visual word form: Longitudinal evolution of category-specific ventral visual areas during reading acquisition.

How does education affect cortical organization? All literate adults possess a region specialized for letter strings, the visual word form area (VWFA), within the mosaic of ventral regions involved in processing other visual categories such as objects, places, faces, or body parts. Therefore, the acquisition of literacy may induce a reorientation of cortical maps towards letters at the expense of other categories such as faces. To test this cortical recycling hypothesis, we studied how the visual cortex of individual children changes during the first months of reading acquisition. Ten 6-year-old children were scanned longitudinally 6 or 7 times with functional magnetic resonance imaging (fMRI) before and throughout the first year of school. Subjects were exposed to a variety of pictures (words, numbers, tools, houses, faces, and bodies) while performing an unrelated target-detection task. Behavioral assessment indicated a sharp rise in grapheme-phoneme knowledge and reading speed in the first trimester of school. Concurrently, voxels specific to written words and digits emerged at the VWFA location. The responses to other categories remained largely stable, although right-hemispheric face-related activity increased in proportion to reading scores. Retrospective examination of the VWFA voxels prior to reading acquisition showed that reading encroaches on voxels that are initially weakly specialized for tools and close to but distinct from those responsive to faces. Remarkably, those voxels appear to keep their initial category selectivity while acquiring an additional and stronger responsivity to words. We propose a revised model of the neuronal recycling process in which new visual categories invade weakly specified cortex while leaving previously stabilized cortical responses unchanged.

Can humans perform mental regression on a graph? Accuracy and bias in the perception of scatterplots.

Despite the widespread use of graphs, little is known about how fast and how accurately we can extract information from them. Through a series of four behavioral experiments, we characterized human performance in "mental regression", i.e. the perception of statistical trends from scatterplots. When presented with a noisy scatterplot, even as briefly as 100 ms, human adults could accurately judge if it was increasing or decreasing, fit a regression line, and extrapolate outside the original data range, for both linear and non-linear functions. Performance was highly consistent across those three tasks of trend judgment, line fitting and extrapolation. Participants' linear trend judgments took into account the slope, the noise, and the number of data points, and were tightly correlated with the t-test classically used to evaluate the significance of a linear regression. However, they overestimated the absolute value of the regression slope. This bias was inconsistent with ordinary least squares (OLS) regression, which minimizes the sum of square deviations, but consistent with the use of Deming regression, which treats the x and y axes symmetrically and minimizes the Euclidean distance to the fitting line. We speculate that this fast but biased perception of scatterplots may be based on a "neuronal recycling" of the human visual capacity to identify the medial axis of a shape.

Impact of literacy on the functional connectivity of vision and language related networks.

Learning to read leads to functional and structural changes in cortical brain areas related to vision and language. Previous evidence suggests that the Visual Word Form Area (VWFA), a region devoted to the recognition of letter strings in literate persons, acts as an interface between both systems. While different studies have performed univariate analyses to study the effects of literacy on brain function, little is known about its impact on whole functional networks, especially when literacy is acquired during adulthood. We investigated functional connectivity in three groups of adults with different literacy status: illiterates, ex-illiterates (i.e., who learned to read during adulthood), and literates (i.e., who learned to read in childhood). We used a data-driven, multivariate whole brain approach (Independent Component Analysis [ICA]) combined with a region of interest (ROI) analysis in order to explore the functional connectivity of the VWFA with four ICA networks related to vision and language functions. ICA allowed for the identification of four networks of interest: left fronto-parietal, auditory, medial visual and lateral visual functional networks, plus a control right fronto-parietal network. We explored the effects literacy on the connectivity between the VWFA and these networks, trying furthermore to disentangle the roles of reading proficiency and age of acquisition (i.e., literacy status) in these changes. Results showed that functional connectivity between the VWFA and the left fronto-parietal and lateral visual networks increased and decreased, respectively, with literacy. Moreover, the functional coupling of the VWFA and the auditory network decreased with literacy. This study provides novel insights in the mechanisms of reading acquisition and brain plasticity, putting to light the emergence of the VWFA as a bridge between language and vision. Further studies are required to characterize the interplay of proficiency and age of reading acquisition, and its relevance to models of brain plasticity across lifespan.

Illiterate to literate: behavioural and cerebral changes induced by reading acquisition.

The acquisition of literacy transforms the human brain. By reviewing studies of illiterate subjects, we propose specific hypotheses on how the functions of core brain systems are partially reoriented or 'recycled' when learning to read. Literacy acquisition improves early visual processing and reorganizes the ventral occipito-temporal pathway: responses to written characters are increased in the left occipito-temporal sulcus, whereas responses to faces shift towards the right hemisphere. Literacy also modifies phonological coding and strengthens the functional and anatomical link between phonemic and graphemic representations. Literacy acquisition therefore provides a remarkable example of how the brain reorganizes to accommodate a novel cultural skill.

Brain networks for confidence weighting and hierarchical inference during probabilistic learning.

Learning is difficult when the world fluctuates randomly and ceaselessly. Classical learning algorithms, such as the delta rule with constant learning rate, are not optimal. Mathematically, the optimal learning rule requires weighting prior knowledge and incoming evidence according to their respective reliabilities. This "confidence weighting" implies the maintenance of an accurate estimate of the reliability of what has been learned. Here, using fMRI and an ideal-observer analysis, we demonstrate that the brain's learning algorithm relies on confidence weighting. While in the fMRI scanner, human adults attempted to learn the transition probabilities underlying an auditory or visual sequence, and reported their confidence in those estimates. They knew that these transition probabilities could change simultaneously at unpredicted moments, and therefore that the learning problem was inherently hierarchical. Subjective confidence reports tightly followed the predictions derived from the ideal observer. In particular, subjects managed to attach distinct levels of confidence to each learned transition probability, as required by Bayes-optimal inference. Distinct brain areas tracked the likelihood of new observations given current predictions, and the confidence in those predictions. Both signals were combined in the right inferior frontal gyrus, where they operated in agreement with the confidence-weighting model. This brain region also presented signatures of a hierarchical process that disentangles distinct sources of uncertainty. Together, our results provide evidence that the sense of confidence is an essential ingredient of probabilistic learning in the human brain, and that the right inferior frontal gyrus hosts a confidence-based statistical learning algorithm for auditory and visual sequences.