Readers make targeted regressions to plausible errors in reanalysis of "noisy-channel garden-path" sentences

2026-05-18

AI Summary

This study investigates how readers perform goal-directed reanalysis of early linguistic input when encountering “noisy-channel garden-path sentences”—structures that initially appear grammatical but later violate expectations. Combining eye-tracking experiments with a Bayesian noisy-channel computational model, the research provides the first empirical evidence that readers strategically regress to likely error sites, with their regressive eye-movement patterns closely aligning with the model’s posterior predictions. These findings support a noisy-channel framework of language comprehension, offering critical evidence for an information-theoretic account of dynamic reanalysis during reading.

Abstract

A key question in psycholinguistics is how inferences about the meaning of linguistic input unfold incrementally a comprehender's mind. In this work, we study reading dynamics for ``noisy-channel garden-path'' sentences, which temporarily appear well-formed but feature late-appearing violations of expectation that can be resolved not by inferring an alternative syntactic structure, but by inferring the presence of an error. We find evidence for targeted regressions -- eye movements towards regions that are promising loci of possible errors in light of later-arriving information, showing patterns consistent with the posterior inferences of a model of noisy-channel processing with reanalysis. We discuss the implications of these findings for theories of noisy-channel language comprehension and information-theoretic explanations of reading dynamics.

Problem

• noisy-channel
• garden-path
• reanalysis
• reading dynamics
• targeted regressions

Innovation

• noisy-channel processing
• garden-path sentences
• targeted regressions
• reanalysis
• reading dynamics