Three eye movement studies examined the role of punctuation in reading. In Experiment 1, although a comma at the end of a clause facilitated overall reading times for the sentence, first pass times were longer at the end of comma-marked clauses than clauses without a comma (or the same material in clause medial position). The data supported the dwell time hypothesis which claims that readers pause at punctuation marks. To explore the possibility that dwell time/wrap-up effects are really due to pausing at the ends of intonational phrases, Experiment 2 tested non-redundant punctuation in structures where a comma corresponds to an obligatory intonational phrase boundary in speech. The intonation hypothesis was supported for vocatives and parentheticals but not for non-restrictive relative clauses. Relative clauses were further investigated in Experiment 3. Overall, the data were consistent with the intonation explanation for wrap-up effects but only if it is assumed that non-restrictive relatives and other asides receive less attention than do other parts of a sentence.