The coherence threshold marks the point at which a reader has gained a sufficient comprehension level to move on in a text. Previous research has demonstrated that the readers' coherence threshold can be manipulated by increasing or decreasing task demands. The present experiments examined a manipulation of the coherence threshold within the text itself. Experiment 1 demonstrated that the participants' coherence threshold was raised following an in-text disruption, evidenced by slowed reading times on the semantic anomaly target sentence that followed. Experiment 2 demonstrated that this effect was not limited to contextually related information. In Experiment 3, the effect occurred even when the inconsistency and anomaly were separated further by filler text. In Experiments 4 and 5 the threshold was not reset following a pause in reading but was reset following a full passage break with a comprehension question. Results are discussed within the context of the RI-Val model of reading comprehension.