A key goal in psycholinguistics is to quantify real-time language processing effort. A notable amount of literature links the processing effort to the amount of unexpected information carried by the new input. However, such theories fail to capture the dynamic evolution of comprehenders’ interpretation which may involve revision and operate under cognitive constraints. We propose peak information flux, a metric that better captures the cost of reanalysis through a resource rational framework. We characterize language comprehension as continuous belief updating under finite information processing capacity and that the maximal rate of belief change signals language processing effort. We validate the performance of peak information flux on Syntactic Ambiguity Processing and Natural Stories corpora featuring self-paced reading time and a variety of eye-tracking metrics representing different stages of lexical processing. Peak information flux predicts reading behavior across multiple measures and complements surprisal in measures associated with reinterpretation and recovery. Additional analyses reveal that garden-path constructions exhibit delayed and asymmetric information flux dynamics, supporting the hypothesis of a sustained probabilistic revision process. Our model provides an information-theoretic metric for language comprehension under resource constraints.

Speech errors are a natural part of communication, yet they rarely lead to complete communicative failure because both speakers and comprehenders can detect and correct errors. Although prior research has examined error monitoring and correction in production and comprehension separately, integrated investigation of both systems has been impeded by the scarcity of parallel data. In this study, we present SPACER, a parallel dataset that captures how naturalistic speech errors are corrected by both speakers and comprehenders. We focus on single-word substitution errors extracted from the Switchboard speech corpus, accompanied by speaker’s self-repairs and comprehenders’ responses from an offline text-editing experiment. Our exploratory analysis suggests asymmetries in error correction strategies: speakers are more likely to repair errors that introduce greater semantic and phonemic deviations, whereas comprehenders tend to correct errors that are phonemically similar to more plausible alternatives or do not fit into prior contexts. Our dataset enables future research on the integrated approach of language production and comprehension.

Current pre-trained language models have enabled remarkable improvements in downstream tasks, but it remains difficult to distinguish effects of statistical correlation from more systematic logical reasoning grounded on the understanding of real world. We tease these factors apart by leveraging counterfactual conditionals, which force language models to predict unusual consequences based on hypothetical propositions. We introduce a set of tests from psycholinguistic experiments, as well as larger-scale controlled datasets, to probe counterfactual predictions from five pre-trained language models. We find that models are consistently able to override real-world knowledge in counterfactual scenarios, and that this effect is more robust in case of stronger baseline world knowledge—however, we also find that for most models this effect appears largely to be driven by simple lexical cues. When we mitigate effects of both world knowledge and lexical cues to test knowledge of linguistic nuances of counterfactuals, we find that only GPT-3 shows sensitivity to these nuances, though this sensitivity is also non-trivially impacted by lexical associative factors.