The Illustrated Page: Analyzing Illustrations of Historical Children’s Books Using Citizen Science

This paper presents the first large-scale, systematic study of historical children’s book illustrations through a combination of citizen science and computational analysis. Using a corpus of 27,901 digitized illustrations from 2,827 books from the Internet Archive’s Children’s Library, we developed a structured annotation workflow deployed on Zooniverse to collect over 400,000 annotations from 902 volunteers. Tasks included identifying depicted characters, objects, settings, and emotional tone. We assess inter-annotator reliability across task types and derive consensus labels to explore three central questions: who and what is most commonly visualized, which entities co-occur, and how visual depictions change over time. Findings reveal dominant portrayals of patriarchal figures and animals, the centrality of nature, and gendered patterns in emotional framing. Temporal analysis shows a surprising visual stability over 140 years. This work demonstrates the value of human-in-the-loop annotation for visual cultural heritage and provides a new resource for studying the visual language of childhood in print.