scholarly journals A compositional neural architecture for language

2018 ◽  
Author(s):  
Andrea E. Martin
Keyword(s):  
Hierarchical Structures ◽  
Expressive Power ◽  
Structure Information ◽  
Systems Neuroscience ◽  
Neural Entrainment ◽  
Perceptual Representations ◽  
The Brain ◽  
Computational Properties ◽  
Perception Action ◽  
Processing Mechanism

Hierarchical structure and compositionality imbue human language with unparalleled expressive power and set it apart from other perception-action systems. However, neither formal nor neurobiological models account for how these defining computational properties might arise in a physiological system. I attempt to reconcile hierarchy and compositionality with principles from cell assembly computation in neuroscience; the result is an emerging theory of how the brain could convert distributed perceptual representations into hierarchical structures across multiple timescales while representing interpretable incremental stages of (de)compositional meaning. The model's architecture - a multidimensional coordinate system based on neurophysiological models of sensory processing - proposes that a manifold of neural trajectories encodes sensory, motor, and abstract linguistic states. Gain modulation, including inhibition, tunes the path in the manifold in accordance with behavior, and is how latent structure is inferred. As a consequence, predictive information about upcoming sensory input during production and comprehension is available without a separate operation. The proposed processing mechanism is synthesized from current models of neural entrainment to speech, concepts from systems neuroscience and category theory, and a symbolic-connectionist computational model that uses time and rhythm to structure information. I build on evidence from cognitive neuroscience and computational modeling that suggests a formal and mechanistic alignment between structure building and neural oscillations, and moves towards unifying basic insights from linguistics and psycholinguistics with the currency of neural computation.

scholarly journals A Compositional Neural Architecture for Language

2020 ◽  
Vol 32 (8) ◽  
pp. 1407-1427 ◽  
Author(s):  
Andrea E. Martin
Keyword(s):  
Hierarchical Structures ◽  
Expressive Power ◽  
Structure Information ◽  
Systems Neuroscience ◽  
Neural Entrainment ◽  
Perceptual Representations ◽  
The Brain ◽  
Computational Properties ◽  
Perception Action ◽  
Processing Mechanism

Hierarchical structure and compositionality imbue human language with unparalleled expressive power and set it apart from other perception–action systems. However, neither formal nor neurobiological models account for how these defining computational properties might arise in a physiological system. I attempt to reconcile hierarchy and compositionality with principles from cell assembly computation in neuroscience; the result is an emerging theory of how the brain could convert distributed perceptual representations into hierarchical structures across multiple timescales while representing interpretable incremental stages of (de)compositional meaning. The model's architecture—a multidimensional coordinate system based on neurophysiological models of sensory processing—proposes that a manifold of neural trajectories encodes sensory, motor, and abstract linguistic states. Gain modulation, including inhibition, tunes the path in the manifold in accordance with behavior and is how latent structure is inferred. As a consequence, predictive information about upcoming sensory input during production and comprehension is available without a separate operation. The proposed processing mechanism is synthesized from current models of neural entrainment to speech, concepts from systems neuroscience and category theory, and a symbolic-connectionist computational model that uses time and rhythm to structure information. I build on evidence from cognitive neuroscience and computational modeling that suggests a formal and mechanistic alignment between structure building and neural oscillations, and moves toward unifying basic insights from linguistics and psycholinguistics with the currency of neural computation.

Relation Between Level of Prefrontal Activity and Subject's Performance

1999 ◽  
Vol 13 (2) ◽  
pp. 117-125 ◽  
Author(s):  
Laurence Casini ◽  
Françoise Macar ◽  
Marie-Hélène Giard
Keyword(s):  
Brain Activity ◽  
Sensory Information ◽  
Practice Phase ◽  
Trained Subjects ◽  
Anagram Task ◽  
Economic Information ◽  
Long Duration ◽  
Level Of Activity ◽  
The Brain ◽  
Processing Mechanism

Abstract The experiment reported here was aimed at determining whether the level of brain activity can be related to performance in trained subjects. Two tasks were compared: a temporal and a linguistic task. An array of four letters appeared on a screen. In the temporal task, subjects had to decide whether the letters remained on the screen for a short or a long duration as learned in a practice phase. In the linguistic task, they had to determine whether the four letters could form a word or not (anagram task). These tasks allowed us to compare the level of brain activity obtained in correct and incorrect responses. The current density measures recorded over prefrontal areas showed a relationship between the performance and the level of activity in the temporal task only. The level of activity obtained with correct responses was lower than that obtained with incorrect responses. This suggests that a good temporal performance could be the result of an efficacious, but economic, information-processing mechanism in the brain. In addition, the absence of this relation in the anagram task results in the question of whether this relation is specific to the processing of sensory information only.

Intact Priming for Novel Perceptual Representations in Amnesia

1997 ◽  
Vol 9 (6) ◽  
pp. 699-713 ◽  
Author(s):  
Stephan B. Hamann ◽  
Larry R. Squire
Keyword(s):  
Recognition Memory ◽  
Declarative Memory ◽  
Brain Structures ◽  
Perceptual Identification ◽  
Single Exposure ◽  
Memory Representations ◽  
Perceptual Representations ◽  
The Creation ◽  
And Control ◽  
The Brain

Recent studies have challenged the notion that priming for ostensibly novel stimuli such as pseudowords (REAB) reflects the creation of new representations. Priming for such stimuli could instead reflect the activation of familiar memory representations that are orthographically similar (READ) and/or the activation of subparts of stimuli (RE, EX, AR), which are familar because they occur commonly in English. We addressed this issue in three experiments that assessed perceptual identification priming and recognition memory for novel and familiar letter strings in amnesic patients and control subjects. Priming for words, pseudowords, and orthographically illegal nonwords was fully intact in the amnesic patients following a single exposure, whereas recognition memory was impaired for the same items. Thus, priming can occur for stimuli that are unlikely to have preexisting representations. Words and pseudowords exhibited twice as much priming as illegal nonwords, suggesting that activation may contribute to priming for words and wordlike stimuli. Additional results showed that priming for illegal nonwords resulted from the formation of new perceptual associations among the component letters of each nonword rather than the activation of individual letter representations. In summary, the results demonstrate that priming following a single exposure can depend on the creation of new perceptual representations and that such priming is independent of the brain structures essential for declarative memory.

scholarly journals The restless brain: how intrinsic activity organizes brain function

2015 ◽  
Vol 370 (1668) ◽  
pp. 20140172 ◽  
Author(s):  
Marcus E. Raichle
Keyword(s):  
Brain Function ◽  
Functional Organization ◽  
Sensory Information ◽  
Intrinsic Activity ◽  
Systems Neuroscience ◽  
Brain Functions ◽  
Molecular Neuroscience ◽  
Constant State ◽  
Health And Disease ◽  
The Brain

Traditionally studies of brain function have focused on task-evoked responses. By their very nature such experiments tacitly encourage a reflexive view of brain function. While such an approach has been remarkably productive at all levels of neuroscience, it ignores the alternative possibility that brain functions are mainly intrinsic and ongoing, involving information processing for interpreting, responding to and predicting environmental demands. I suggest that the latter view best captures the essence of brain function, a position that accords well with the allocation of the brain's energy resources, its limited access to sensory information and a dynamic, intrinsic functional organization. The nature of this intrinsic activity, which exhibits a surprising level of organization with dimensions of both space and time, is revealed in the ongoing activity of the brain and its metabolism. As we look to the future, understanding the nature of this intrinsic activity will require integrating knowledge from cognitive and systems neuroscience with cellular and molecular neuroscience where ion channels, receptors, components of signal transduction and metabolic pathways are all in a constant state of flux. The reward for doing so will be a much better understanding of human behaviour in health and disease.

scholarly journals Backwards is the way forward: Feedback in the cortical hierarchy predicts the expected future

2013 ◽  
Vol 36 (3) ◽  
pp. 221-221 ◽  
Author(s):  
Lars Muckli ◽  
Lucy S. Petro ◽  
Fraser W. Smith
Keyword(s):  
Predictive Coding ◽  
Internal Models ◽  
Prediction Errors ◽  
Conceptual Level ◽  
Cortical Hierarchy ◽  
The Brain ◽  
Descriptive Level ◽  
Perception Action ◽  
The Way

AbstractClark offers a powerful description of the brain as a prediction machine, which offers progress on two distinct levels. First, on an abstract conceptual level, it provides a unifying framework for perception, action, and cognition (including subdivisions such as attention, expectation, and imagination). Second, hierarchical prediction offers progress on a concrete descriptive level for testing and constraining conceptual elements and mechanisms of predictive coding models (estimation of predictions, prediction errors, and internal models).

scholarly journals Reverse hierarchies and sensory learning

2008 ◽  
Vol 364 (1515) ◽  
pp. 285-299 ◽  
Author(s):  
Merav Ahissar ◽  
Mor Nahum ◽  
Israel Nelken ◽  
Shaul Hochstein
Keyword(s):  
General Population ◽  
Perceptual Learning ◽  
Special Populations ◽  
Hierarchy Theory ◽  
Training Techniques ◽  
Sensory Learning ◽  
Perceptual Representations ◽  
The Brain

Revealing the relationships between perceptual representations in the brain and mechanisms of adult perceptual learning is of great importance, potentially leading to significantly improved training techniques both for improving skills in the general population and for ameliorating deficits in special populations. In this review, we summarize the essentials of reverse hierarchy theory for perceptual learning in the visual and auditory modalities and describe the theory's implications for designing improved training procedures, for a variety of goals and populations.

Differential effects of fasting and leptin on proopiomelanocortin peptides in the arcuate nucleus and in the nucleus of the solitary tract

2007 ◽  
Vol 292 (5) ◽  
pp. E1348-E1357 ◽  
Author(s):  
Mario Perello ◽  
Ronald C. Stuart ◽  
Eduardo A. Nillni
Keyword(s):  
Energy Balance ◽  
Arcuate Nucleus ◽  
Transcriptional Level ◽  
Solitary Tract ◽  
Energy Status ◽  
Melanocyte Stimulating Hormone ◽  
Pomc Mrna ◽  
Induced Changes ◽  
The Brain ◽  
Processing Mechanism

The α-melanocyte-stimulating hormone (α-MSH), derived from proopiomelanocortin (POMC), is generated by a posttranslational processing mechanism involving the prohormone convertases (PCs) PC1/3 and PC2. In the brain, α-MSH is produced in the arcuate nucleus (ARC) of the hypothalamus and in the nucleus of the solitary tract (NTS) of the medulla. This peptide is key in controlling energy balance, as judged by changes observed at transcriptional level. However, little information is available regarding the biosynthesis of the precursor POMC and the production of its processed peptides during feeding, fasting, and fasting plus leptin in the ARC compared with the NTS in conjunction with the PC activity. In this study we found that, in the ARC, pomc mRNA, POMC-derived peptides, and PC1/3 all decreased during fasting, and administration of leptin reversed these effects. In contrast, in the NTS, where there is a large amount of a 28.1-kDa peptide similar in size to POMC, the 28.1-kDa peptide and other POMC-derived peptides, including α-MSH, were further accumulated in fasting conditions, whereas pomc mRNA decreased. These changes were not reversed by leptin. We also observed that, during fasting, PC2 levels decreased in the NTS. These data suggest that, in the NTS, fasting induced changes in POMC biosynthesis, and processing is independent of leptin. These observations indicate that changes in energy status affect POMC in the brain in a tissue-specific manner. This represents a novel aspect in the regulation of energy balance and may have implications in the pathophysiology of obesity.

scholarly journals Laterality enhances cognition in Australian parrots

2009 ◽  
Vol 276 (1676) ◽  
pp. 4155-4162 ◽  
Author(s):  
Maria Magat ◽  
Culum Brown
Keyword(s):  
Cognitive Function ◽  
Strong Support ◽  
Dual Processing ◽  
Cerebral Lateralization ◽  
Widespread Occurrence ◽  
Fitness Advantage ◽  
Parrot Species ◽  
Types Of Information ◽  
The Brain ◽  
Processing Mechanism

Cerebral lateralization refers to the division of information processing in either hemisphere of the brain and is a ubiquitous trait among vertebrates and invertebrates. Given its widespread occurrence, it is likely that cerebral lateralization confers a fitness advantage. It has been hypothesized that this advantage takes the form of enhanced cognitive function, potentially via a dual processing mechanism whereby each hemisphere can be used to process specific types of information without contralateral interference. Here, we examined the influence of lateralization on problem solving by Australian parrots. The first task, a pebble-seed discrimination test, was designed for small parrot species that feed predominately on small seeds, which do not require any significant manipulation with the foot prior to ingestion. The second task, a string-pull problem, was designed for larger bodied species that regularly use their feet to manipulate food objects. In both cases, strongly lateralized individuals (those showing significant foot and eye biases) outperformed less strongly lateralized individuals, and this relationship was substantially stronger in the more demanding task. These results suggest that cerebral lateralization is a ubiquitous trait among Australian parrots and conveys a significant foraging advantage. Our results provide strong support for the enhanced cognitive function hypothesis.

scholarly journals Higher order thoughts in action: consciousness as an unconscious re-description process

2012 ◽  
Vol 367 (1594) ◽  
pp. 1412-1423 ◽  
Author(s):  
Bert Timmermans ◽  
Leonhard Schilbach ◽  
Antoine Pasquali ◽  
Axel Cleeremans
Keyword(s):  
Social Cognition ◽  
Signal Detection Theory ◽  
Causal Chain ◽  
Type I ◽  
Self Awareness ◽  
First Order ◽  
Network Simulations ◽  
Dual Route ◽  
The Brain ◽  
Perception Action

Metacognition is usually construed as a conscious, intentional process whereby people reflect upon their own mental activity. Here, we instead suggest that metacognition is but an instance of a larger class of representational re-description processes that we assume occur unconsciously and automatically. From this perspective, the brain continuously and unconsciously learns to anticipate the consequences of action or activity on itself, on the world and on other people through three predictive loops: an inner loop, a perception–action loop and a self–other (social cognition) loop, which together form a tangled hierarchy. We ask what kinds of mechanisms may subtend this form of enactive metacognition. We extend previous neural network simulations and compare the model with signal detection theory, highlighting that while the latter approach assumes that both type I (objective) and type II (subjective, metacognition-based) decisions tap into the same signal at different hierarchical levels, our approach is closer to dual-route models in which it is assumed that the re-descriptions made possible by the emergence of meta-representations occur independently and outside of the first-order causal chain. We close by reviewing relevant neurological evidence for the idea that awareness, self-awareness and social cognition involve the same mechanisms.

scholarly journals Fractals in Action: An fMRI study on the Generation of new Hierarchical Levels in Motor Sequences

2017 ◽  
Author(s):  
Mauricio J.D. Martins ◽  
Roberta Bianco ◽  
Daniela Sammler ◽  
Arno Villringer
Keyword(s):  
Motor Planning ◽  
Hierarchical Structures ◽  
Human Cognition ◽  
Generative Process ◽  
Fmri Study ◽  
General Systems ◽  
Hierarchical Levels ◽  
The Brain ◽  
Putative Domain ◽  
Active Generation

AbstractGeneration of hierarchical structures, such as the embedding of subordinate elements into larger structures, is a core feature of human cognition. Discrimination of well-formed hierarchies is thought to rely on lateral prefrontal cortex (PFC). However, the brain bases underlying the active generation of new hierarchical levels remain poorly understood. Here, we created a new motor paradigm to isolate this active generative process. In fMRI, participants planned and performed (identical) movement sequences based on three previously learned rules: (1) a hierarchical ‘fractal’ rule that involved generation of new levels, (2) a linear ‘iterative’ rule adding items to existing hierarchical levels, and (3) simple ‘repetition’. We found that generation of new hierarchical levels (using the fractal rule) activated a bilateral motor planning-and imagery network, but did not involve lateral PFC. Conversely, adding items to existing hierarchical levels required M1 directly during execution. These results show that the generation of new hierarchical levels can be achieved without involvement of putative domain-general systems such as those ascribed to lateral PFC. We hypothesize that these systems might be important to parse hierarchical sequences in a multi-domain fashion but not necessarily to generate new hierarchical levels.

Sign in / Sign up

Export Citation Format

Share Document