Cross-category structure in semantic memory

Area of Science:

• Cognitive Psychology
• Psycholinguistics
• Computational Linguistics

Background:

• Understanding semantic relationships between concepts is crucial for cognitive science.
• The Battig and Montague (1969) norms provide a foundational dataset for category research.
• Previous research has explored category structure through various methods, but direct links between label similarity and exemplar relations require further investigation.

Purpose of the Study:

• To analyze the semantic structure of 56 category terms from the Battig and Montague norms.
• To investigate the relationship between category label similarity and the similarity of high-dominance exemplars within those categories.
• To determine if the derived semantic distances predict behavioral performance, specifically reaction times (RTs) in a same-different task.

Main Methods:

• Experiment I: Participants performed similarity judgments on 56 category terms.
• Hierarchical clustering analysis was applied to the similarity judgments to reveal semantic structures.
• Experiment II compared the structure of category labels with the structure of high-dominance exemplars.
• Experiment III used the semantic distances from clustering to predict RTs in a same-different task.

Main Results:

• Hierarchical clustering revealed distinct semantic groupings among the category terms.
• The relationships among category labels closely mirrored the relationships among their high-dominance exemplars.
• Distances derived from the hierarchical clustering analysis successfully predicted reaction times in the same-different task.

Conclusions:

• Category labels effectively represent the semantic relationships between concepts.
• The structure derived from label similarity is a robust predictor of cognitive processing, specifically reaction times.
• This research validates the use of similarity judgments and hierarchical clustering for mapping semantic knowledge and predicting cognitive performance.