What can we learn from 28,000 historical children’s book illustrations?

In our most recent project, we analyzed nearly 28,000 illustrations from historical children’s literature—spanning over 140 years—using a combination of citizen science and computational analysis.

Illustrated children’s books offer a rich archive for understanding cultural history, combining visual and textual modes of storytelling that reflect societal values across time. Scholars have long emphasized that picture books are complex narrative artifacts in which words and images interact to produce meaning. This multimodal structure makes them powerful instruments
of cultural transmission, encoding assumptions about childhood, gender, identity, morality, and
emotion within their language and visual design. Yet despite this longstanding interdisciplinary recognition, large-scale analysis of historical children’s book illustrations remains limited especially with respect to their visual content.

Turning Citizen Readers into Researchers

One of the biggest challenges in studying visual culture is scale. Unlike text, images are harder to search, categorize, and analyze—especially historical ones that are often noisy, stylized, or degraded.

So we took a different approach.

We built a citizen science project on Zooniverse and invited the public to help annotate images. Nearly 1,000 volunteers contributed over 450,000 annotations in just two weeks. The full project now contains over 1.4 million annotations based on the work of more than 11,000 volunteers.

Participants answered simple but powerful questions:

• Who is in the image?
• Where is the scene set?
• What objects are present?
• How emotionally intense is the scene?

This allowed us to transform thousands of images into structured data—something we can analyze at scale.

What Shows Up in Children’s Books?

Once we aggregated the data, some patterns became immediately clear.

1. Adults—especially men—dominate childhood worlds

Men appear far more often than women and are among the most frequently depicted figures overall. Even in books ostensibly about children, adult male presence looms large .

2. Animals are everywhere

Animals rival humans in frequency, suggesting that childhood is imagined not just socially, but ecologically—deeply embedded in the natural world.

3. Nature is the primary setting

Scenes are far more likely to take place outdoors than indoors. Childhood, at least in these books, unfolds in fields, forests, and open spaces rather than domestic interiors .

4. Fantasy is surprisingly rare

Despite our assumptions about whimsical children’s literature, mythical beings and animated objects appear infrequently. The visual world of these books is more grounded than we might expect.

5. Emotion is mostly calm

Most scenes cluster around low to moderate emotional intensity. Childhood is not depicted as chaotic or dramatic, but as measured, controlled, and often serene.

Who Appears Together—and Who Doesn’t?

Looking beyond frequency, we also examined co-occurrence patterns: which characters tend to appear together.

Here, the results are even more revealing.

• Boys and girls tend to appear together, forming a shared “child world.”
• Girls are often depicted alongside women, suggesting stronger ties to domestic or relational spaces.
• Men and animals frequently appear alone, occupying more autonomous or self-contained roles.
• High-intensity scenes cluster around men, animals, and fantasy figures, while women are associated with more moderate emotional contexts .

In other words, these images don’t just show who is present—they structure relationships, roles, and emotional expectations.

Has Anything Changed Over Time?

Given that our dataset spans from the 18th to early 20th centuries, we might expect dramatic change.

But that’s not what we found.

Across 140 years, many patterns remain strikingly stable:

• The dominance of nature persists
• The balance between boys and girls holds
• Emotional tone stays centered around neutrality

There are exceptions—most notably fluctuations in the presence of adult men—but overall, the visual language of childhood is remarkably consistent over time .

This stability suggests that children’s illustrations may function less as mirrors of rapid social change and more as carriers of enduring cultural ideals.

Why This Matters

At one level, this project is about data: building a large, structured dataset of historical illustrations.

But at a deeper level, it’s about interpretation.

Children’s books are not neutral. Their images encode ideas about:

• Who matters
• What relationships look like
• Where life happens
• How emotions should be expressed

By analyzing these patterns at scale, we can begin to see how cultural assumptions about childhood are constructed, repeated, and normalized over generations.

A New Kind of Humanities Research

This project also points to a broader shift in how we study culture.

Traditionally, literary and cultural analysis has been close, interpretive, and small-scale. Here, we combine that tradition with:

• Citizen participation (who helps generate the data)
• Computational methods (which allow scale)
• Machine learning applications (which can extend analysis further)

The result is not a replacement for human interpretation, but an expansion of it.

We can now ask questions that were previously impossible:
What does childhood look like—not in one book, but across thousands?

Looking Forward

This dataset doesn’t just tell us about the past. It also provides a foundation for:

• Training multimodal AI systems on historical data
• Improving search and accessibility of illustrated archives
• Studying long-term cultural patterns in visual storytelling

Most importantly, it reminds us that even the most familiar cultural objects—like children’s books—are deeply structured, historically situated, and worth studying anew.