About SlaveryResearch.org
Slavery, war, and forced migration are bound together. The great majority of armed conflicts involved some form of enslavement, including forced labor, abduction, trafficking, and exploitation. However, the relationship between slavery, war, and migration remains poorly understood, in the present and in the past. SlaveryResearch.org approaches this nexus historically and at global scale. The aim is to reconstruct, town by town and year by year, how warfare in pre-colonial Africa produced enslavement, and where the enslaved went not only into the internal migrations that fed African states and their slave systems, but also into the external migrations of the African diaspora across the Atlantic, the Sahara, the Red Sea, and the Indian Ocean.
This temporal map helps us better understand when and where people came from within Africa, and when and where they went in diaspora. That matters because the records of slavery so often begin at the coast, reducing millions of individuals to the ports where they were sold. Restoring the inland half of their histories returns homelands, slave routes, and historical context to people whom the trade tried to erase. It connects the cultures of the diaspora to the specific wars and regions that formed them, and shows how warfare redistributed population within Africa itself. The aim is a global, cumulative framework in which the geography of African conflict can be linked to the geography of the diaspora, giving the study of slavery and migration in war the deep empirical foundation that understanding and prevention require.
Mapping Uncertainty
Pre-colonial West Africa lacks detailed, reliable historical maps of inland towns and political boundaries. The effort to remedy that began with “Redrawing Historical Maps of the Bight of Benin Hinterland, c. 1780” (Canadian Journal of African Studies, 2013), which reconstructed the region’s towns and boundaries on the eve of the upheavals that followed. “Mapping Uncertainty: The Collapse of Oyo and the Trans-Atlantic Slave Trade, 1816–1836” (Journal of Global Slavery, 2019) then set those maps in motion. The biggest challenge in mapping Africa’s pre-colonial past revolves around the sources, which are thematically, spatially, and temporally fragmented. The founding experiment produced twenty-one annual maps of the dissolution and formation of towns during the jihad and the collapse of Oyo, joined to records of nearly 75,000 people, the majority of them Yoruba speakers, who boarded slave ships between Little Popo and Lagos for Brazil, Cuba and, through British abolition, Sierra Leone.
SlaveryResearch.org turns that limitation into its core method. Uncertainty is recorded rather than hidden. Every instance of conflict enters the database as a dated range with a graded assessment of its accuracy. The maps express probability rather than false precision, with conflict shown as gradients. The outputs are probabilities, which can be tested. Wherever prediction and evidence disagree, the maps improve. An approximation that states its own uncertainty can be corrected by every new source. That is the strength on which the whole framework rests, and it is why this site claims not to know where any one person came from, but when and where enslavement was most likely, for whom, and with what probability.
Shared Vocabulary for African Geography
Scaling from one kingdom to a continent requires consistent geography. AfricanRegions.org, developed in consultation with scholars across the field to support systematic continental expansion, proposed six broad regions and thirty-four sub-regions as a controlled vocabulary for digital history (History in Africa, 2019 and 2021). Regions as geographic containers facilitate collecting, classifying, and linking records about Africa’s past while deliberately avoiding terms inherited from slave traders, colonial rule, or modern borders. That vocabulary is the scaffold on which SlaveryResearch.org is being built continent-wide. Each of the thirty-four sub-regions is a future module, and the present map, spanning the Central Savanna and the Western Bight, demonstrates the model any of them can follow, linking regional warfare to both internal enslavement and every branch of the diaspora.
A Schema for Conflict
“Mapping Conflict during the Era of the Slave Trade” (Joseph C. Miller Memorial Lecture, Bonn Center for Dependency and Slavery Studies, 2022) set out the metadata schema at the core of the database. Every settlement is a panel of dated periods of peace, slave raiding, open conflict, destruction or abandonment, and refoundation. Each period is classified on a conflict-intensity index, attributed to a controlling polity, and graded for dating accuracy. The schema converts narrative history, from chronicles and traveler itineraries to court records and two centuries of regional scholarship, into data a statistical model can use without severing it from its sources. The version displayed here holds roughly 690 towns and more than 2,000 town-periods, rendered as the flashing conflict symbols and annual heat surfaces on the map. This schema also allows for including a plethora of information connecting places and the people associated with them, including king lists, military leaders, or other enslaved people documented in the diaspora.
From Conflict to Origins
The spatial statistical keystone is “A Modelling Strategy to Estimate Conditional Probabilities of African Origins” (Wiens, Lovejoy, Mullen & Vance, Journal of the Royal Statistical Society, 2022). This model works in three movements. Because places and dates are approximate, quantities enter as Gaussian distributions, spreads of probability around known points rather than false certainties. A Markov decision process then moves captives across the trade network toward Atlantic ports and interior markets, choosing routes step by step under penalties of distance and warfare along the way. Kriging, a geostatistical technique, interpolates the town panel into annual surfaces of conflict intensity, turning scattered dated events into a continuous map of where enslavement was most likely in each year. Bayesian inversion then turns the model around to yield, for any port and year, the conditional probability that a person embarked there came from a given inland place. On this site the method is extended from the Oyo region to slave ship departures. The “Likely Inland Origins” shown when clicking a ship are its output.
Testing the Model Against Names
Ships seized by the naval blockade and condemned before Vice Admiralty Courts and Mixed Commissions involved registers of the people found aboard. For this phase, over 33,000 individuals were registered, and their documentation is assembled at LiberatedAfricans.org. Interpreting the languages of more than 20,000 of these recorded names against dictionaries of Yoruba, Gbe, Hausa, and other naming practices provides an out-of-sample test. In the peak decades covered by the registers, the model’s predicted mix of origins matches the observed language of names within a few percentage points. As a result, it is possible to get a sense of the ethnolinguistic composition of the people on board a single slave ship, and associate likeliest inland origins of those people to instances of intra-African conflict. The mismatches are findings in their own right. They indicate that the northern wars of the savanna fed interior slave systems far more than the Atlantic, a geography of retention the map makes visible.
Where the Project Stands
The map presented here is the framework’s first module: the Central Savanna and Western Bight of West Africa from 1800 to 1865, the era of the Sokoto jihad and the collapse of Oyo. It was chosen because its sources are rich enough to test every part of the method, from conflict panels to statistical models of origins to validation against the names of “Liberated Africans”. Subsequent modules will extend the same schema to other regions and periods of the continent.
A Work in Progress
Conflict chronologies, model parameters, and name interpretations remain under active revision, and the geographic scope and periodization will expand. Future work will expand on a continental and temporal scale. The model will also integrate time and distance to better align conflict with the movement of captives, alongside the itineraries of enslaved people who documented their own routes to the coast.