Subjective and objective text difficulty in L2 English reading: exploring their interplay in relation to eye-movement behavior

Text difficulty has traditionally been evaluated using objective indices such as readability measures. However, difficulty is also inherently subjective. In this study, we examined how second-language English readers’ subjective judgments of text difficulty were associated with established readability indices. We also explored their relationship with difficulty estimates derived from large language models (LLMs) as a further objective indicator. Additionally, we investigated how these subjective and objective difficulty measures related to eye-movement behavior during reading. The results indicated that subjective ratings were strongly correlated with objective difficulty measures. Bayesian multilevel models further indicated that the LLM-derived estimate and text-level mean subjective difficulty were associated with selected passage-level eye-movement measures, such as total reading time and skipping count. However, model comparisons showed that these text-level predictors provided only limited incremental improvement once participant- and text-level differences were modeled. The clearest incremental information came from reader-specific deviations in subjective difficulty. These findings suggest that objective and subjective difficulty measures are related but not interchangeable. Objective measures are useful for characterizing text-level difficulty, whereas reader-specific subjective ratings provide additional insight into how individual readers experience and process the same text.