Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Taxonomic-thematic knowledge and verbal fluency modulate written word processing
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
Aino Saranpää1, Gayane Ghazaryan1, Tiina Lindh-Knuutila1, Riitta Salmelin1; 1Aalto University, Espoo, Finland
Semantic memory contains both taxonomic (category-based) and thematic (relation-based) knowledge. Previous literature has aimed to tease apart the neurobiological basis of these knowledge systems and candidate regions have been identified, but the findings are inconsistent, potentially due to individual variability. Here, we linked success of brain-level decoding of words presented in the thematic vs. taxonomic contexts with the participants’ linguistic skills, quantified by a verbal fluency task. Verbal fluency is measured as the number of words a participant can generate within a minute in a given category. This task heavily relies on taxonomic structure, and higher fluency is likely associated with more separable taxonomic representations. Consequently, we hypothesized that better semantic fluency would be linked to better decoding accuracy, especially in the taxonomic context. Participants were native Finnish speakers (n=25). Their verbal fluency skills were evaluated using both semantic (animals and store items categories) and phonemic (words starting with letters k and s) fluency tasks. Magnetoencephalography (MEG) responses were measured while the participants were reading word triplets. Two words created taxonomic (e.g., bus-ambulance) or thematic (e.g., ramp-bridge) context for the third target word (car). Participants were divided into two groups based on their fluency scores across tasks. Evoked responses triggered by the target words were analysed in three predefined time windows: 100–300, 300–500 and 500–700 ms. Zero-shot decoding of the target word was performed using word2vec embeddings. Successful decoding within a time window indicates that different words evoke distinct neural patterns. Across all participants, and separately for the higher- and lower-fluency groups, the thematic context produced greater activation than taxonomic context at 300–500 ms in the sensors over the left-temporal cortex. The lower-fluency group additionally showed this effect in the right temporal cortex. Neural decoding was successful in the higher-fluency group in both taxonomic and thematic contexts at 300–500 ms. Thematic decoding was further successful at 100–300 ms. In the taxonomic condition, decoding was successful only in the left hemisphere. For the lower-fluency group, decoding was successful especially in thematic condition at 300–500 ms, but the accuracies were lower than for the higher-fluency group. Decoding analysis indicated that for the lower-fluency group, information related to the target word meaning was present bilaterally in both thematic and taxonomic contexts. Our results suggest that both context and skill jointly shape semantic retrieval of written words. While target-word activation was stronger in the thematic than taxonomic context in the left-temporal cortex at the time when semantic retrieval is commonly considered to occur, we were able to decode words similarly across contexts. Importantly, higher fluency score was associated with more left-lateralized patterns and more accurate decoding across contexts. Efficient semantic memory retrieval may depend on well differentiated left hemisphere representations, especially of the taxonomic structure.
Topic Areas: Meaning: Lexical Semantics, Computational Approaches