Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Semantic representations in the prefrontal cortex: insights from single-cell recordings in humans and monkeys
Poster Session E, Sunday, September 14, 11:00 am - 12:30 pm, Field House
Benjamin Mash1, Irene Caprara2, Seunghyun Lee1, Tamsin Lambourne1, Douglas Kellar1, Raymundo Báez-Mendoza3, Mohsen Jamali1, Ziv Williams1; 1Massachusetts General Hospital / Harvard Medical School, 2KU Leuven, 3The German Primate Centre - Leibniz Institute for Primate Research
Semantic information is essential for navigating and interacting with the environment, not only in humans but also across many animal species. It is thought that key components of semantic processing systems are evolutionarily conserved, given the shared need to extract similar types of information from a certain stimulus. While prior studies have explored neuroanatomical similarities across species, semantic representation at the level of single neurons remains poorly understood. To address this gap, we conducted single-unit recordings from homologous regions of the prefrontal cortex—areas involved in semantic processing and categorization—in macaque monkeys (n = 2) and human participants (n = 3) undergoing awake neurosurgery. During the recordings, participants were shown images spanning a wide range of semantic categories, valences, and themes. To facilitate cross-species comparison, matched sets of images were used. Human participants also heard spoken sentences reflecting similar semantic content. We derived vectorial representations of each stimulus using two approaches: (1) BERT-based sentence embeddings of descriptive labels and (2) a set of 14 behaviorally relevant attributes, encompassing sensory, motor, spatial, temporal, affective, social, and cognitive dimensions, rated by participants via Amazon Mechanical Turk. These vectors were employed for clustering stimuli and modeling neuronal responses. Neurons selective to specific semantic domains were identified in both species. Decoding analyses revealed that both human and macaque neuronal populations encoded semantic information in strikingly similar ways. However, we also observed species-specific variability in how semantic domains were organized and mapped within population-level response patterns. These differences suggest evolutionary divergence layered upon a conserved hierarchical framework for semantic representation. Together, these findings offer novel insights into the single-neuron and population-level coding of semantic information, illuminating both conserved mechanisms and species-specific adaptations in the neural basis of meaning.
Topic Areas: Multisensory or Sensorimotor Integration, Meaning: Lexical Semantics