Long-Horizon Associative Learning Explains Human Sensitivity to Statistical and Network Structures in Auditory Sequences

Journal of Neuroscience, Journal Year: 2024, Volume and Issue: 44(14), P. e1369232024 - e1369232024

Published: Feb. 26, 2024

Networks are a useful mathematical tool for capturing the complexity of world. In previous behavioral study, we showed that human adults were sensitive to high-level network structure underlying auditory sequences, even when presented with incomplete information. Their performance was best explained by model compatible associative learning principles, based on integration transition probabilities between adjacent and nonadjacent elements memory decay. present explored neural correlates this hypothesis via magnetoencephalography (MEG). Participants (

Language: Английский

---

Humans parsimoniously represent auditory sequences by pruning and completing the underlying network structure

eLife, Journal Year: 2023, Volume and Issue: 12

Published: May 2, 2023

Successive auditory inputs are rarely independent, their relationships ranging from local transitions between elements to hierarchical and nested representations. In many situations, humans retrieve these dependencies even limited datasets. However, this learning at multiple scale levels is poorly understood. Here, we used the formalism proposed by network science study representation of higher-order structures interaction in sequences. We show that human adults exhibited biases perception elements, which made them sensitive high-order such as communities. This behavior consistent with creation a parsimonious simplified model evidence they receive, achieved pruning completing elements. observation suggests brain does not rely on exact memories but world. Moreover, bias can be analytically modeled memory/efficiency trade-off. correctly accounts for previous findings, including transition probabilities well structures, unifying sequence across scales. finally propose putative implementations bias.

Language: Английский

---

Hebbian learning can explain rhythmic neural entrainment to statistical regularities

Developmental Science, Journal Year: 2024, Volume and Issue: 27(4)

Published: Feb. 19, 2024

In many domains, learners extract recurring units from continuous sequences. For example, in unknown languages, fluent speech is perceived as a signal. Learners need to the underlying words this signal and then memorize them. One prominent candidate mechanism statistical learning, whereby track how predictive syllables (or other items) are of one another. Syllables within same word predict each better than straddling boundaries. But does learning lead memories words-or just pairwise associations among syllables? Electrophysiological results provide strongest evidence for memory view. responses can be time-locked boundaries (e.g., N400s) show rhythmic activity with periodicity durations. Here, I reproduce such simple Hebbian network. When exposed statistically structured syllable sequences (and when not excessively long), network activation duration maxima on word-final syllables. This because receive more excitation earlier which they associated less predictable that occur words. The also sensitive information whose electrophysiological correlates were used support encoding ordinal positions thus explain neural tasks without any representations might rely cues beyond learn their native language. RESEARCH HIGHLIGHTS: Statistical may utilized identify speech) but generate explicit Exposure leads period words). memory-less model well putative encodings observed research. Direct tests needed establish whether declarative

Language: Английский

---

Larger lexicons enable representation of fine-grained phonological similarity structure: Evidence from English L2 speakers’ sound similarity judgments of word pairs

Journal of Memory and Language, Journal Year: 2025, Volume and Issue: 142, P. 104619 - 104619

Published: Feb. 4, 2025

Language: Английский

---

The role of conscious attention in statistical learning: evidence from patients with impaired consciousness

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 8, 2024

Abstract The debate over whether conscious attention is necessary for statistical learning has produced mixed and conflicting results. Testing individuals with impaired consciousness may provide some insight, but very few studies have been conducted due to the difficulties associated testing such patients. In this study, we examined ability of patients varying levels disorders (DOC), including coma, unresponsive wakefulness syndrome, minimally patients, emergence from state extract regularities an artificial language composed four randomly concatenated pseudowords. We used a methodology based on frequency tagging in EEG, which was developed our previous speech segmentation sleeping neonates. Our study had two main objectives: firstly, assess automaticity process explore correlations between level covert abilities regularities, second, potential new diagnostic indicator aid patient management by examining correlation successful markers level. observed that were preserved suggesting inherently automatic low-level process. Due significant inter-individual variability, word not be sufficiently robust candidate clinical use, unlike temporal accuracy auditory syllable responses, correlates strongly coma severity. Therefore, propose stimulus train, simple measure, should further investigated as possible metric DOC diagnosis.

Language: Английский

---

The successor representation subserves hierarchical abstraction for goal-directed behavior

PLoS Computational Biology, Journal Year: 2024, Volume and Issue: 20(2), P. e1011312 - e1011312

Published: Feb. 20, 2024

Humans have the ability to craft abstract, temporally extended and hierarchically organized plans. For instance, when considering how make spaghetti for dinner, we typically concern ourselves with useful “subgoals” in task, such as cutting onions, boiling pasta, cooking a sauce, rather than particulars many cuts onion, or exactly which muscles contract. A core question is decomposition of more abstract task into logical subtasks happens first place. Previous research has shown that humans are sensitive form higher-order statistical learning named “community structure”. Community structure common feature tasks characterized by ordering subtasks. This can be captured model where learn predictions upcoming events multiple steps future, discounting further away time. One “successor representation”, been argued hierarchical abstraction. As yet, no study convincingly this abstraction put use goal-directed behavior. Here, investigate whether participants utilize learned community informed action plans Participants were asked search paintings virtual museum, grouped together “wings” representing museum. We find participants’ choices accord museum their response times best predicted successor representation. The degree reflect correlates several measures performance, including These results suggest representation subserves abstractions relevant

Language: Английский

---

Probing sensitivity to statistical structure in rapid sound sequences using deviant detection tasks

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 23, 2024

Abstract Statistical structures and our ability to exploit them are a ubiquitous component of daily life. Yet, we still do not fully understand how track these sophisticated statistics the role they play in sensory processing. Predictive coding frameworks hypothesize that for stimuli can be accurately anticipated based on prior experience, rely more strongly internal model world “surprised” when expectation is unmet. The current study used this phenomenon probe listeners’ sensitivity probabilistic generated using rapid 50 milli-second tone-pip sequences precluded conscious prediction upcoming stimuli. Over three experiments measured response time deviants frequency outside expected range. Predictable were either triplet-based or network-style structure deviant detection contrasted against same set tones but random, unpredictable order. All found structured enhanced relative random sequences. Additionally, Experiment 2 different instantiations community demonstrate level uncertainty modulated saliency. Finally, 3 placed within an established immediately after transition between communities, where perceptual boundary should generate momentary uncertainty. However, manipulation did impact performance. Together results contexts from statistical modulate processing ongoing auditory signal, leading improved detect unexpected stimuli, consistent with predictive framework.

Language: Английский

---

Algebraic structures emerge from the self-supervised learning of natural sounds

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: March 14, 2024

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 existence an innate specific mechanism, like “merge” (Chomsky) or “neural recursion” (Dehaene). Others that ability emerges from experience (e.g. Bates): i.e. when generic learning principles continuously process sensory inputs. These views, however, remain difficult to test experimentally. Here, we use deep models evaluate factors lead spontaneous detection structures in auditory modality. Specifically, train multiple with a variable amount natural sounds self-supervised objective. We then expose these experimental paradigms classically used processing Like humans, repeated sequences, probabilistic chunks Also diminishes structure complexity. Importantly, emerge alone: more are exposed sounds, they increasingly Finally, does not pretrained only on speech, rapidly than environmental sounds. Overall, our study provides operational framework clarify sufficient built-in acquired model human’s advanced capacity Significance Statement Experimentalists have repeatedly observed human advantage structures, notably through paradigms. This is thought be key emergence cognitive operations. Yet, it remains debated if discovered form mechanism. In article, authors show how progressively learns structure. The replicate several findings but under certain developmental conditions. Notably, exposition detection. As result, work proposes as abstract abilities.

Language: Английский

---

Can non-human primates extract the linear trend from a noisy scatterplot?

iScience, Journal Year: 2024, Volume and Issue: 28(1), P. 111661 - 111661

Published: Dec. 20, 2024

Recent studies showed that humans, regardless of age, education, and culture, can extract the linear trend a noisy scatterplot. Although this capacity looks sophisticated, it may simply reflect extraction principal graph, as if cloud dots was processed an oriented object. To test idea, we trained Guinea baboons to associate arbitrary shapes with increasing or decreasing trends noiseless scatterplots, while varying number points, noise level, regression slope. Many successfully learned conditional match-to-sample task, their accuracy varied sigmoid function t-value regression, same statistical index upon which humans also base answers. The perceptual component human graphics abilities seems thus be based on recycling phylogenetically older competence primate visual system for extracting axes displays.

Language: Английский

---

Long-horizon associative learning explains human sensitivity to statistical and network structures in auditory sequences

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 16, 2024

Abstract Networks are a useful mathematical tool for capturing the complexity of world. In previous behavioral study, we showed that human adults were sensitive to high-level network structure underlying auditory sequences, even when presented with incomplete information. Their performance was best explained by model compatible associative learning principles, based on integration transition probabilities between adjacent and non-adjacent elements memory decay. present explored neural correlates this hypothesis via magnetoencephalography (MEG). Participants passively listened sequences tones organized in sparse community comprising two communities. An early difference (~150 ms) observed brain responses tone transitions similar probability but occurring either within or This result implies rapid automatic encoding sequence structure. Using time-resolved decoding, estimated duration overlap representation each tone. The decoding exhibited exponential decay, resulting significant representations successive tones. Based extended decay profile, long-horizon novelty index found correlation measure MEG signal. Overall, our study sheds light mechanisms sensitivity structures highlights potential role Hebbian-like supporting at various temporal scales.

Language: Английский

---