Causal influence of linguistic learning on perceptual and conceptual processing: A brain-constrained deep neural network study of proper names and category terms

Abstract

Language influences cognitive and conceptual processing, but the mechanisms through which such causal effects are realized in the human brain remain unknown. Here, we use a brain-constrained deep neural network model of category formation and symbol learning and analyze the emergent model-internal mechanisms at the neural circuit level. In one set of simulations, the network was presented with similar patterns of neural activity indexing instances of objects and actions belonging to the same categories. Biologically realistic Hebbian learning led to the formation of instance-specific neurons distributed across multiple areas of the network, and, in addition, to cell assembly circuits of ‘shared’ neurons responding to all category instances – the network correlates of conceptual categories. In two separate sets of simulations, the network learned the same patterns together with symbols for individual instances (‘proper names’) or symbols related to classes of instances sharing common features (‘category terms’). Learning category terms remarkably increased the number of shared neurons in the network, thereby making category representations more robust while reducing the number of neurons of instance-specific ones. In contrast, proper-name learning prevented substantial reduction of instance-specific neurons and blocked the overgrowth of category-general cells. Representational Similarity Analysis further confirmed that the neural activity patterns of category instances became more similar to each other after category-term learning, relative to both learning with proper names and without any symbols. These network-based mechanisms for concepts, proper names and category terms explain why and how symbol learning changes object perception and memory, as revealed by experimental studies.Significance StatementHow do verbal symbols for specific individuals (Micky Mouse) and object categories (house mouse) causally influence conceptual representation and processing? Category terms and proper names have been shown to respectively promote category formation and instance learning, potentially by respectively directing attention to category-critical and object-specific features. Yet the mechanisms underlying these observations at the neural circuit level remained unknown. Using a mathematically precise deep neural network model constrained by properties of the human brain, we show category-term learning strengthens and solidifies conceptual representations, whereas proper names support object-specific mechanisms. Based on network-internal mechanisms and unsupervised correlation-based learning, this work offers neurobiological explanations for causal effects of symbol learning on concept formation, category building and instance representation in the human brain.

Verbal Symbols Support Concrete but Enable Abstract Concept Formation: Evidence From Brain‐Constrained Deep Neural Networks

“…Such associative grounding mechanisms may appear feasible for learning aspects of the referential meaning of concrete symbols, but they cannot explain all aspects thereof. For example, for explaining the difference between proper names, which can only be used to speak about one specific entity or person, and category terms applicable to a large set of instances, more is required than simple association (Nguyen et al, 2024). An unsurmountable set of challenges is apparently posed by highly abstract concepts and meanings, which, according to standard views, have correlates not in the world, but rather in mental space, where they relate to equally abstract conceptual and semantic features.…”

“…These effects emerged from the implementation of neurobiological and neuroanatomical principles in a brain-constrained neural network model (see prediction (b)). As these are general principles of Hebbian learning, they do not apply only to category formation through semantic similarity and category terms, but might also underlie behavioral effects of category terms versus proper names on attention modulation and memory performance (see, e.g., Althaus & Mareschal, 2014;LaTourrette & Waxman, 2020;Nguyen et al, 2024).…”