Related Experiment Video
Updated: Jun 28, 2026
07:36
Eye Tracking During Visually Situated Language Comprehension: Flexibility and Limitations in Uncovering Visual Context Effects
Published on: November 30, 2018
Data from eye-tracking corpora as evidence for theories of syntactic processing complexity
1School of Informatics, University of Edinburgh, Edinburgh, UK. v.demberg@ed.ac.uk
Cognition
|October 22, 2008
Summary
Dependency Locality Theory (DLT) partially predicts reading times for nouns, but not all words. Unlexicalized surprisal effectively predicts reading times, suggesting both theories are needed for processing complexity.
Area of Science:
- Psycholinguistics
- Computational Linguistics
- Cognitive Science
Background:
- Syntactic processing complexity theories aim to explain how readers process sentence structures.
- Dependency Locality Theory (DLT) and surprisal are two prominent but distinct theoretical frameworks.
- Evaluating these theories with naturalistic reading data is crucial for advancing psycholinguistic models.
Purpose of the Study:
- To test the predictive power of DLT and surprisal against real-world reading data.
- To investigate the relationship between DLT integration cost and surprisal in predicting reading times.
- To assess the utility of eye-tracking corpora for theory development in sentence processing.
Main Methods:
- Analysis of eye-tracking data from 10 participants reading 51,000 words of newspaper text (Dundee Corpus).
- Evaluation of Dependency Locality Theory (DLT) integration cost as a predictor of word reading times.
- Assessment of an unlexicalized surprisal model's ability to predict reading times.
- Statistical comparison of DLT and surprisal predictive performance.
Main Results:
- DLT integration cost did not significantly predict reading times for arbitrary words but showed success for nouns.
- Evidence for integration cost effects was found at auxiliary verbs, a prediction not covered by DLT.
- An unlexicalized surprisal formulation effectively predicted reading times for arbitrary words.
- DLT integration cost and surprisal were found to be uncorrelated.
Conclusions:
- Neither DLT nor surprisal alone fully explains reading times in naturalistic text.
- A comprehensive theory of syntactic processing complexity likely requires integrating both locality-based and probability-based factors.
- Eye-tracking corpora offer valuable, ecologically valid data for refining and testing psycholinguistic theories.
