In two experiments, we investigated whether automatic activation, integration, and validation processes apply to sequentially presented text-picture combinations. In previous experiments, we already showed that simultaneously presented text-picture combinations are automatically integrated and validated. In these experiments, we used an adapted version of the Epistemic Stroop paradigm (Richter et al., 2009): Participants saw either valid (text and picture matched) or invalid text-picture combinations (text and picture mismatched). These stimuli were followed by the probe words “right” or “wrong” to which participants reacted with different key presses. Even though participants were instructed to react independently of the stimuli to the probe words, participants showed a Stroop effect: they answered slower and made more errors in incongruent trials (validity of the stimuli and the probe words mismatched) as in congruent trials (validity of the stimuli and the probe word matched). In the current experiments, we extended these findings by presenting the text-picture combinations sequentially, that is, an unrelated task was presented between the text and the picture part of each stimulus. We found a Stroop effect in Experiment 1 (N = 251), which used one-sentence picture combinations, and in Experiment 2 (N = 266), which used multiple-sentence picture combinations. This suggests that participants not just automatically integrate and validate text-picture information but that they also automatically activate text or picture information which is no longer available in working memory. These findings have implications for text comprehension and multimedia learning theories.