A central question in psycholinguistics concerns whether linguistic knowledge and non-linguistic (world) knowledge are processed differently during sentence comprehension. Competing accounts propose either a two-step model, in which linguistic knowledge is integrated first, and world knowledge is incorporated only afterwards, or a one-step model, in which the two knowledge types are integrated simultaneously. It remains unclear whether these knowledge sources give rise to similar or distinct processing patterns. The present study examines how comprehenders respond to temporal inconsistencies arising from linguistic constraints or world knowledge. Linguistic inconsistencies are determined by incompatible time adverbials in past-under-past constructions with a complement clause, where a forward-shifted reading is not allowed, as in #One week ago, Emma said that David smiled two days ago. World-knowledge inconsistencies arise when adverbials contradict expected event sequences, as in #During the rehearsal, Emma said that David smiled at the concert, where the typical rehearsal-before-concert relation is reversed. Two experiments contrast these sources of inconsistency and comprehenders’ sensitivity to them during temporal interpretation. An acceptability-rating study assesses whether linguistic and world-knowledge inconsistencies are recognised as similarly anomalous. A self-paced reading experiment examines the extent to which the same inconsistencies disrupt reading in similar sentence regions. Comparable disruption patterns would support the one-step model, whereas divergent patterns would support the two-step model. The findings of the present study will clarify whether it is necessary to distinguish between these knowledge sources from a language-processing perspective.