This essay interprets a classification of Africa's Bantu languages which used statistical tools guided by assumptions about farming and its chronology to analyze fresh vocabulary evidence. It shows a peeling movement from Cameroon's grassfields, into southern Cameroon, then along a savanna corridor through West Central Africa's rainforests, into the Savannahs, then to Southern Africa, the Great Lakes, and Indian Ocean coast. The clear sequence of movement masks methodological and historical factors. Language death, multilingualism, and the limits of vocabulary evidence restrain the classification's authority. 'Transformations' from food collecting to food producing or from no metals to full engagement with metals were mutable, unfolded at different speeds, and involved interactions with firstcomers. In Central Africa, Bantu speakers were often the first farmers and metal-users in the region but elsewhere they were commonly neither. Their arrivals did not immediately displace firstcomers. Computational methods can accommodate many of these issues.