The article proposes Syntactic Belief Update, a model that predicts processing difficulty in garden path sentences by measuring the magnitude of syntactic belief updates via generalized Rényi divergence. This approach outperforms lexical surprisal by providing a better fit to human reading time data.
- Garden path sentences cause processing difficulty when initial interpretations are proven wrong by critical words.
- Lexical surprisal fails to accurately predict processing difficulty for these specific sentence structures.
- The proposed model actively predicts and updates a probability distribution over syntactic trees after each word.
- Processing difficulty is quantified using generalized Rényi divergence, which depends on lexical items but is independent of their probability.
- This metric provides a superior fit to human reading time data compared to existing measures.
The findings suggest a new research direction in psycholinguistics focused on examining purely non-lexical alternatives to surprisal for understanding sentence processing.