The story of Bantu
01
From Kongo Grammars to Ur-Bantu
The first outsiders to notice the pattern were missionaries. In 1659 the Italian Capuchin Giacinto Brusciotto published in Rome a Latin grammar of Kikongo that already sorted its nouns by their prefixes. Over the next two centuries travellers and missionaries along both coasts of Africa found that languages thousands of kilometres apart shared the same machinery of prefixes and a common stock of words: -ntu 'person', -tatu 'three', -tano 'five'.
The German linguist Wilhelm Bleek, working at the Cape, set this out in A Comparative Grammar of South African Languages (1862–69) and gave the family its name from the plural ba-ntu 'people'. The decisive step came with Carl Meinhof, who around 1900 applied the methods of Indo-European sound laws to Bantu and reconstructed an ancestral 'Ur-Bantu' with regular correspondences between languages. That regularity, rather than resemblance alone, is what made the family certain.
In the mid-20th century two scholars pulled in different directions. Malcolm Guthrie, at the School of Oriental and African Studies in London, catalogued the languages into lettered zones (1948) and compiled the four-volume Comparative Bantu (1967–71). Joseph Greenberg, in the United States, argued that Bantu is not a family on its own level at all but one small branch of a much larger Niger-Congo family, its nearest relatives the Bantoid languages of Nigeria and Cameroon. Greenberg's view of Bantu's place has prevailed. The Belgian A. E. Meeussen's grammatical reconstructions (1967) and the lexical reconstructions later compiled at the Royal Museum for Central Africa in Tervuren became the standard references for Proto-Bantu.
02
Where It Began and How It Spread
Guthrie placed the Bantu centre in the south of the Congo basin, where the languages seemed to him most typical. Greenberg (1972) argued the opposite from diversity: a family's homeland is usually where its oldest splits are found, and the closest non-Bantu relatives and the most divergent Bantu languages are in the Nigeria–Cameroon borderland. Nearly all specialists now accept the northwestern homeland.
Three kinds of evidence are used to trace the spread. Reconstructed vocabulary shows what the early speakers knew: yams, oil palms, goats, boats and pottery, but not, it seems, iron or cattle. Archaeology tracks the arrival of pottery and village settlement across Central Africa from the first millennium BCE, and of Early Iron Age farming communities in eastern and southern Africa in the first centuries CE. Genetics shows that the spread involved migration, not just the passing on of a language: many Bantu speakers from Kenya to South Africa share West-Central African ancestry, mixed with that of the local peoples they met.
The details are debated. Ancient DNA from Shum Laka, a rock shelter in the Cameroon Grassfields, published by Lipson and colleagues in 2020, came from people who were not the direct ancestors of most Bantu speakers today, a reminder that the homeland region had a complex population history. The timing is uncertain: dates for the first departure range from about 5,000 to 4,000 years ago. There is also disagreement on how the migrants crossed the equatorial rainforest, and on whether a climatic crisis around 2,500 years ago, which opened savanna corridors, was the trigger.
03
Noun Classes, Long Verbs and Tone
The feature that defines the family is the noun-class system. Proto-Bantu had around twenty classes, most of them paired as singular and plural, and each noun's prefix is echoed on the words that agree with it. In Swahili kitabu kikubwa kimeanguka means 'the big book has fallen'. Make it plural and every prefix changes: vitabu vikubwa vimeanguka. The classes loosely track meaning (people in classes 1 and 2, trees in 3 and 4, languages and small things in others), so that Swahili m-tu is 'person' and ki-swahili the Swahili language.
The verb is built by stringing pieces together in a fixed order: subject, tense, object, root, and a final vowel. Swahili ni-na-ku-pend-a is 'I love you' (literally I-now-you-love). Suffixes called extensions change what a verb does to its participants: from pika 'cook' come pikia 'cook for' and pikwa 'be cooked'. A single verb can therefore carry what English expresses in a whole clause.
The basic word order is subject–verb–object. Syllables are mostly open, so words end in vowels, and nasal-plus-consonant clusters like mb, nd and ng are common at the start of words, as in Ndebele or Swahili mbwa 'dog'. Almost all Bantu languages are tonal, using pitch to tell words and tenses apart; Swahili, which has lost its tones, is the best-known exception. Most are written in the Latin alphabet introduced by 19th-century missionaries. Swahili and some of its neighbours were written in Arabic script for centuries before that.
04
Unsettled Questions
The inner shape of the family is still being worked out. Guthrie's zones are convenient but geographical, and specialists agree that they do not reflect descent. Computational phylogenetic studies since the early 2000s, notably by Rebecca Grollemund and colleagues (2015), have produced trees in which the Northwest languages form several early offshoots and Eastern Bantu emerges late from within the western languages. These trees are built from vocabulary lists, and critics point out that shared words can also spread between neighbouring languages, so that a tree may partly reflect later contact rather than descent.
Where the family ends is also debated. The label 'Narrow Bantu' covers the languages in Guthrie's zones, but several of them, such as the Mbam languages of central Cameroon, may belong outside the core group, while some Bantoid languages of the Grassfields resemble Bantu so closely that the boundary looks arbitrary.
Dating depends on assumptions. Bayesian methods give estimates of the family's age, but they rest on calibrations from archaeology, and archaeologists do not always agree on which pottery traditions correspond to which languages.
Further out, Bantu's membership of Benue-Congo and Atlantic-Congo is secure. Whether Atlantic-Congo in turn belongs with Mande, Dogon and others in a single Niger-Congo family, as Greenberg proposed, is doubted by some specialists, including Tom Güldemann, who argue that the evidence for the outer branches has never been demonstrated by the comparative method.