Concept cells in the temporal pole: neuronal basis of semantic knowledge in humans
Year: 2025
Type: conference-paper
Study design: Observational Study
Abstract: Single-neuron recordings in humans provide a unique opportunity to evaluate cognitive models at the neuronal level 1,2. Intracerebral recordings have identified neurons selective to one particular concept, termed concept cells 3, in the hippocampus and adjacent medial temporal lobe (MTL) structures as the building blocks of declarative memory 4. However, these concept cells demonstrate strong semantic properties, such as multimodal responses 5 and coding for broad semantic category 6, rather than episodic properties 7. Finding such selective neurons within the hippocampus and MTL structures presents a paradox given that it is the temporal pole (TP) that is widely regarded as the primary hub for semantic memory 8. This raises critical questions regarding the type of neuronal code underlying semantic knowledge in the TP and its functional relationship with the neuronal code observed in the hippocampus. To investigate this issue, we recorded the activity from 521 single units across 16 patients implanted with tetrodes 9 from the TP (n=129), MTL (n=261), and other brain regions (n=149), while they watched series of pictures (see 3,10 for methodological details). Among these, 80 single units in the temporal pole of five different patients exhibited increased activity to at least one concept. At the single-unit level we observed highly selective cells. These units displayed several defining characteristics: (1) invariant responses to different exemplars of the same concept, (2) response latencies shorter than those observed in the MTL (approximately 250 ms post-stimulus onset), (3) multimodal encoding of conceptual information, (4) explicit representation of concepts, (5) a less sparse code than in the hippocampus and (6) absence of responses to non-semantic stimuli. Preliminary neural manifold analyses 11revealed a neuronal code for broad semantic category with different time trajectories between semantic categories, revealing a dual-level code: at the single neuron level for specific concept and at the population level for broad semantic category. These findings provide novel insights into the neuronal coding of semantic concepts. Specifically, they suggest that semantic concepts are represented by highly specialized neurons within the TP, akin to the coding mechanisms observed in the hippocampus and MTL. A key question emerging from these results is the functional significance of such a distributed semantic coding scheme across multiple, anatomically distinct brain regions. Further investigation is required to elucidate the mechanisms underlying this distributed representation and its implications for the organization of semantic memory.
Source: HAL
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7205040725