Abstract:Acquisition of language semantics is widely believed to be facilitated by biological mechanisms of associative learning [1]. However, current understanding of brain mechanisms underlying learning and memory is primarily based on simple models, including animal models, with mechanisms of learning in the cognitive realm largely unknown. Semantic language learning serves as a compelling example of these cognitive processes, whose brain mechanisms pose a challenge to explanation at the level of basic neural networks. Particularly, many processes underlying such learning still lack clear understanding, including the mechanisms that enable correlational coincidence of neuronal representation required for Hebbian plasticity, the reverberatory replay of the to-be associated representations in working memory, and the involvement of consolidation, among others. We conducted a series of studies modeling the acquisition of semantic knowledge of action-related vocabulary to investigate these questions. We implemented a unique trial-and-error-based procedure for semantic acquisition, in which participants obtained associations between newly introduced pseudowords and actions through active learning. The experimental procedure comprised the delivery of acoustically equivalent pseudowords that were previously unknown to the subjects. Participants were instructed to execute movements of their hands and feet in response to specific pseudowords. Following each trial, participants received either positive or negative feedback depending on their performance. During the learning phase, MEG was recorded, as well as during passive presentations of the same pseudowords before and after learning to control for attention and motor preparation effects. Our analysis involved event-related fields and beta oscillations. Our experiments showed that cortical plasticity related to word learning could be detected immediately after learning, without the need for prolonged consolidation time [2]. We observed two effects during learning that could contribute to Hebbian plasticity by linking auditory speech representations of pseudowords with associated movements. First, we observed the reactivation of auditory speech representations during the initialization of movement. This mechanism apparently creates the conditions for temporal coincidence of activations in two representations, specifically an auditory speech representation and a representation of a motor program. During the learning process, we observed a considerable rise in beta oscillations that occurred during and after action execution. This effect comprised a typical beta-rebound in the sensorimotor regions and a significant surge in beta activity in various associative cortical regions. We postulate that these beta oscillations constitute a reverberation mechanism that reinforces the associative connection and guards it against potential interference [3]. In further experiments, we administered the learning process over two consecutive days with an overnight interval between recordings. We evaluated beta power dynamics during the acquisition of novel pseudowords on day 1 and during task rehearsal following overnight sleep on day 2. Beta event-related synchronization in the frontal regions emerged upon the attainment of the pseudoword learning rules on day 1, and remained unchanged after sleep and subsequent rehearsal on day 2. The prefrontal beta rhythm’s dynamics mirrored the corresponding behavioral changes: subjects executed the task with minimal errors on day 2 upon meeting the learning criterion on day 1. Beta event-related synchronization augmented continuously in the posterior temporal and parietal cortices on day 2 and potentially contributed towards the assimilation of newly-acquired associations in long-term memory. In summary, our research offers fresh perspectives on the intricate neural mechanisms underlying plasticity in the human brain within the cognitive domain.

The backpropagation-based recollection hypothesis: Backpropagated action potentials mediate recall, imagination, language understanding and naming

Backpropagation-Based Recollection of Memories: Biological Plausibility and Computational Efficiency

Backprojections in the cerebral cortex : implications for memory

Coherent noise enables probabilistic sequence replay in spiking neuronal networks

Neural mechanisms of associative cortical plasticity in cognitive domain: Magnetoencephalographic studies

Backward Projections in the Cerebral Cortex: Implications for Memory Storage

Different Propagation Speeds of Recalled Sequences in Plastic Spiking Neural Networks

Inferring neural activity before plasticity as a foundation for learning beyond backpropagation

Principles for generation of reverberation

Reconstructing Spatiotemporal Trajectories of Visual Object Memories in the Human Brain

Reconstructing Spatio-Temporal Trajectories of Visual Object Memories in the Human Brain

Information bottleneck-based Hebbian learning rule naturally ties working memory and synaptic updates

A model for integrating elementary neural functions into delayed-response behavior

Reversed and forward buffering of behavioral spike sequences enables retrospective and prospective retrieval in hippocampal regions CA3 and CA1

A neuroanatomically grounded Hebbian-learning model of attention-language interactions in the human brain

Replay in human visual cortex is linked to the formation of successor representations and independent of consciousness

Networks for Memory, Perception, and Decision-Making, and Beyond to How the Syntax for Language Might Be Implemented in the Brain.

Memory Recall and Spike Frequency Adaptation

A history of spike-timing-dependent plasticity

Human Cognition and Language Processing with Neural-Lexicon Hypothesis

Predictive learning rules generate a cortical-like replay of probabilistic sensory experiences