Lexical interference effects in sentence processing: evidence from the visual world paradigm and self-organizing...

Area of Science:

• Psycholinguistics
• Computational Neuroscience
• Cognitive Science

Background:

• Decades of psycholinguistic research show mixed results on how context constraints are integrated during real-time sentence processing.
• Evidence exists for both rapid, anticipatory context integration and for late or non-integration of contextual information.

Purpose of the Study:

• To propose and test a self-organizing neural network framework for understanding lexical information integration with sentence context.
• To reconcile diverging findings on the timing of constraint integration in language comprehension.

Main Methods:

• Developed a self-organizing neural network model simulating the integration of lexical and contextual information.
• Conducted two simulations to test the model's ability to predict existing psycholinguistic findings.
• Performed two experiments using the visual world paradigm to validate new model predictions.

Main Results:

• The neural network framework successfully predicted both classic lexical ambiguity resolution findings (suggesting late integration) and anticipatory eye movement data (suggesting rapid integration).
• Experimental results confirmed a novel prediction: listeners fixated predictable objects and also non-predictable but contextually consistent objects more than distractors in constraining environments.
• This indicates a nuanced integration process that considers both direct predictions and broader contextual consistency.

Conclusions:

• The proposed neural network framework offers a unified account for diverse findings in sentence processing and constraint integration.
• The findings suggest that language comprehension involves both rapid and flexible integration of contextual information, accommodating predictable and plausible alternatives.
• This research advances our understanding of the dynamic interplay between lexical access and contextual constraints in real-time language processing.