Scholarly communication · May 14, 2026 · 5 min read

Lessons from the Slave Voyages database

What the Trans-Atlantic Slave Trade Database teaches about building scholarly datasets: provenance, versioning, editorial vetting, ethics and citation.

Historic bound registers open on a reading-room table beside a laptop

Few scholarly datasets have shaped a field as much as the Trans-Atlantic Slave Trade Database. It records individual slaving voyages: ships, dates, ports, captains, and the numbers of people embarked and landed. Historians, teachers, genealogists and descendants use it to study a system of forced migration that moved millions of Africans to the Americas.

The database also offers practical lessons for anyone building a research dataset. Its history shows how to handle sources, revisions, contributions and the ethics of data about people who were treated as cargo. This post looks at those lessons.

From CD-ROM to the web

The database began as scholarship on disk. The Trans-Atlantic Slave Trade Database CD-ROM, published in 1999 by David Eltis and colleagues, documented 27,233 voyages compiled from archives on several continents. It was a landmark, but a CD-ROM is a fixed edition. Corrections and new findings had to wait for another release.

In the mid-2000s a project based at Emory University set out to change that. The MetaScholar Initiative, a digital library program run from the Woodruff Library, partnered with slave trade researchers to build an expanded, online and freely available version. The project site described the goal: a revised database of more than 35,000 voyages, released in 2008 to mark the bicentennial of the abolition of the slave trade in Britain and the United States.

Funding came from the Woodruff Library and the National Endowment for the Humanities, which designated the project a "We the People" grant. Harvard's W.E.B. Du Bois Institute and an Emory history endowment also contributed. The resource went public as Voyages, and today it lives on at slavevoyages.org, extended over the years with further datasets and tools.

Provenance is the dataset

A voyage record is only as credible as its sources. The database was built from shipping registers, port records, naval reports, newspapers and other archival material. Many voyages appear in more than one source, and the sources do not always agree.

The lesson is that provenance cannot be an afterthought. A scholarly dataset should let users see where each value came from, and ideally which source was preferred when two conflict. Researchers argue about the slave trade's scale and patterns. They can only do so productively if they can trace figures back to evidence.

  • Record sources per field where it matters, not only per record. A ship's name and a count of captives may come from different documents.
  • Separate observed values from estimates. Imputed numbers are legitimate scholarship, but users must be able to tell them apart from documented ones.
  • Keep the original notation. Spellings of ships, places and names vary across languages and centuries. Normalize for search, but preserve what the source said.
record.xml — before / after enrichment
Harvested
titleLessons from the Slave Voyages dat…
date2026-05-14
creator—
subjectmisc
rightsempty
Enriched
dc:titleLessons from the Slave Voyages database
dc:date2026-05-14
dc:typeArticle
dc:subjectDigital libraries
dc:rightsRights statement attached
premisfixity sha256 ✓
Figure 1 · interactive illustration

Versioning and editorial vetting

The move online made the database a living resource. The project planned a way for researchers to submit new data to an editorial board for vetting before inclusion. That model sits between a closed edition and an open wiki. Contributions are welcome, but they pass through expert review.

Vetting works best when the rules are written down. Contributors need to know what evidence is required, how conflicts with existing values are resolved, and how long review takes. Reviewers need a way to record their decision and the reason for it. Without that record, a dataset slowly fills with changes nobody can explain. With it, every value has a trail back to a source and a person who accepted it.

A living dataset raises a hard question. If the numbers change, how does anyone reproduce a published result? The answer is explicit versioning. Every release should have a version label and a date. Earlier versions should remain available, or at least documented, so that a 2010 article can be checked against the data it used.

PracticeWhy it mattersSimple way to do it
Stable record identifiersCitations keep pointing at the same voyage after editsNever reuse or renumber an ID
Named releasesResults can be reproduced against a known stateVersion number and date on every export
Change logUsers see what moved and whyShort notes per release, linked from the data
Editorial review of contributionsGrowth without loss of trustA small board with written criteria
Documented methodsEstimates can be evaluated, not just acceptedA methods essay published with the data

None of this is exotic. It is the same discipline software projects use. Scholarly datasets benefit from it just as much. Our post on persistent identifiers covers the identifier side in more depth.

Ethics of data about enslaved people

The records behind the database were created by slave traders, merchants and officials. They describe human beings as commodities. Turning those records into tables risks repeating that view unless the presentation is handled with care.

The project's own plans pointed in a useful direction. It aimed to serve two audiences: professional researchers and K-12 and general readers. Maps, archival documents and essays were meant to give the numbers context. Later work by the wider Voyages team added material about the names and lives of individual Africans, which shifts attention from ships to people.

A few principles follow for any dataset with this kind of history:

  • Explain how and by whom the source records were made, and what they leave out.
  • Pair quantitative views with interpretation written by specialists.
  • Consult descendant communities and educators, not only researchers, about presentation.
  • Choose language carefully, for example "enslaved people" rather than terms taken from the ledgers.

Access design is part of ethics too. Some users come to the data looking for ancestors, not for statistics. Interfaces that let them search by port, ship or name, and that explain the limits of the records gently, serve them better than a raw table.

We discuss related questions in our post on the ethics of digitizing civil rights collections.

A magnifying glass over a handwritten manuscript next to its digital scan
Photo illustration

Citation and credit

A dataset that changes over time is hard to cite well. A useful citation names the resource, the version or access date, and the specific record identifiers used. Voyage identifiers make this practical. A researcher can cite a single voyage, and a reader can find it again.

Credit matters too. Datasets like this are built by many hands: original compilers, archivists who preserved the sources, developers who built the interface, and contributors who submitted corrections. Good documentation names them and describes their roles. That makes the dataset easier to trust and easier to sustain, because the people who maintain it can show the value of their work.

Small projects can borrow the same habits. Publish a short "how to cite" note beside the data. Give each release a date and a version. Offer a downloadable export so that results do not depend on a live interface that may change or disappear. And deposit at least one release in a repository with a persistent identifier, so the dataset outlives the website that first presented it.

The broader point is simple. A scholarly dataset is a publication. It needs sources, editions, review and credit, just like a book. The Trans-Atlantic Slave Trade Database showed what that looks like at scale, and its approach still holds up.

Questions

Where can I use the Trans-Atlantic Slave Trade Database today?
The database and related resources are freely available at slavevoyages.org, the continuation of the online Voyages project launched in 2008.
How did the online project differ from the 1999 CD-ROM?
The CD-ROM was a fixed edition of 27,233 voyages. The online version expanded the data to more than 35,000 voyages, added maps and documents, and planned vetted contributions from researchers.
What is the most transferable lesson for other dataset projects?
Treat provenance and versioning as core features. Users should see where values come from and be able to cite a stable record in a known release.