Linguistic Distributional Knowledge and Sensorimotor Grounding both Contribute to Semantic Category Production

Area of Science:

• Cognitive Science
• Computational Linguistics
• Neuroscience

Background:

• Human concepts integrate sensorimotor simulation and linguistic distributional data.
• The linguistic shortcut hypothesis suggests prioritizing efficient linguistic information when sufficient.

Purpose of the Study:

• To investigate the interplay of sensorimotor and linguistic information in category production.
• To test the linguistic shortcut hypothesis in semantic processing.
• To develop and validate a computational model of conceptual activation.

Main Methods:

• A pre-registered category production study involving verbal naming of category members.
• Novel measures of sensorimotor similarity and linguistic proximity were employed.
• A follow-up study incorporated typicality ratings.
• A fully grounded computational model of conceptual activation was created.

Main Results:

• Both sensorimotor similarity and linguistic proximity predicted category production order and frequency.
• Linguistic proximity provided predictive power beyond sensorimotor similarity.
• Typicality was a strong predictor but did not fully explain sensorimotor and linguistic effects.
• The computational model, integrating both data types, closely matched human performance.

Conclusions:

• Results support the linguistic shortcut hypothesis in semantic processing.
• Evidence suggests both linguistic and grounded (sensorimotor) representations are integral to the human conceptual system.
• The developed computational model offers a robust framework for understanding conceptual activation.