Brain-constrained neural modeling explains fast mapping of words to meaning

Abstract

Although teaching animals a few meaningful signs is usually time-consuming, children acquire words easily after only a few exposures, a phenomenon termed “fast-mapping.” Meanwhile, most neural network learning algorithms fail to achieve reliable information storage quickly, raising the question of whether a mechanistic explanation of fast-mapping is possible. Here, we applied brain-constrained neural models mimicking fronto-temporal-occipital regions to simulate key features of semantic associative learning. We compared networks (i) with prior encounters with phonological and conceptual knowledge, as claimed by fast-mapping theory, and (ii) without such prior knowledge. Fast-mapping simulations showed word-specific representations to emerge quickly after 1–10 learning events, whereas direct word learning showed word-meaning mappings only after 40–100 events. Furthermore, hub regions appeared to be essential for fast-mapping, and attention facilitated it, but was not strictly necessary. These findings provide a better understanding of the critical mechanisms underlying the human brain’s unique ability to acquire new words rapidly.

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

“…Following earlier modeling work (Constant et al, 2023;Henningsen-Schomers et al, 2023;Henningsen-Schomers & Pulvermüller, 2022;R. Tomasello et al, 2018R.…”

“…Following earlier modeling work (Constant et al., 2023; Henningsen‐Schomers et al., 2023; Henningsen‐Schomers & Pulvermüller, 2022; R. Tomasello et al., 2018, 2019), we modeled neuronal learning and brain activity using brain‐constrained deep neural network models (see Pulvermüller et al., 2014, 2021; see Figure 2, Panels A and B, and Appendix S1 in the Supporting Information online for details). The model was implemented on the neural network simulation platform Felix (Wennekers, 2009).…”

“…Note that most previous neural simulation studies used networks quite distant from human brain structure and function, favoring learning efficacy or standard connectionist architectures instead of brain constraints. Previous research with brain‐constrained network models has shown them to be well suited for addressing mechanisms of word–meaning mapping (see Constant et al., 2023; Garagnani & Pulvermüller, 2016; R. Tomasello et al., 2017, 2018, 2019).…”

Grounded Cognition Entails Linguistic Relativity: Response to Commentators

“…In addition, this framework has inspired research on the intersubjective "we" representations that facilitate coordinated multiagent behavior (Sebanz & Knoblich, 2021). Furthermore, and of greatest relevance here, recent neurocomputational investigations based directly on the GCM incorporate some-certainly not all, but some-of the social aspects of learning language-specific concepts (Constant, Pulvermüller, & Tomasello, 2023). Even apart from these considerations, however, the fact that language-specific concepts are public rather than private and are acquired through social interaction rather than personal bodily experience does not really bear on my argument that the GCM entails linguistic relativity, because that argument hinges on the content of languagespecific concepts, not their origin.…”

Can Human Brain Connectivity explain Verbal Working Memory?

“…Furthermore, it is needless to say that there is a wide range of different semantic word types and learning scenarios. The modelling of 'pure' object and action words grounded directly into sensorimotor system is one way to address such differences (for other word types see [25,44,112]). The differences in the simulated semantic grounding process, either objects or actions t, led to qualitative differences of the related CAs, which extended more strongly either into the primary visual or the primary articulatory-motor cortex, respectively.…”