<p>During reading, readers frequently generate causal inferences to fully comprehend a text. However, the causal inferences they initially make about the cause of an event may later turn out to be incorrect. The cognitive processes underlying the revision of causal inferences during Chinese text comprehension remain to be examined. To address this, the present study combined an eye-tracking technique with the Word-printed Visual World Paradigm (VWP) to investigate causal inference revision processes during Chinese text reading, with a special focus on the dynamics of initial and revised causal inferences within this cognitive process. Participants’ eye movements on target words representing initial and revised causal inferences were recorded while they listened to the sentences that either induced the revision of causal inferences or did not. It was found that following the onset of the discrepancy, revised causal inference received significantly more fixations in revised texts compared to non-revision texts, around 1400 ms, while the initial inference was not found to be significantly different across different types of texts. This result suggested that readers could successfully generate revised causal inferences during Chinese text reading. Moreover, it suggested that to build a coherent situation model, revised causal inference was activated while initial causal inference might not be suppressed but co-existed during situation model updating. The study contributes to the resonance, integration, and validation model (RI-Val model) by proposing a “competition phase” to elucidate the real-time processing of causal inference revision.</p>