Three experiments explored how pictures in award-winning children's storybooks contribute to their cohesion. In Experiment 1, one group of college students read storybooks with pictures, and another group read them with the pictures removed. Both groups answered questions inserted periodically. The source for about one half of the questions was the pictures, and the source for the remaining questions was the written text. The answers to one half of the questions from each source were explicitly contained in the immediately preceding text or pictures, and the remainder required inferences that went beyond the immediately preceding information. Not surprisingly, questions based on the pictures could not be answered without the pictures. In Experiment 2, the texts without pictures were rewritten to provide the essential storyline information that the pictures had supplied. Cohesion of both the original and rewritten texts was assessed with Coh-Metrix indexes. The changes resulted in increased cohesion for referential and semantic information, with relatively little change in syntactic cohesion indexes or readability levels. In Experiment 3, the pattern of college students' responses to questions on the revised texts was very similar to responses in Experiment 1 to the texts with pictures. (Contains 3 tables, 2 figures, and 1 footnote.)