© 1999 Macmillan Magazines Ltd

An African Classical Age: Easternand Southern Africa in WorldHistory, 1000 BC to AD 400by Christopher EhretUniversity of Virginia Press/James Currey:1998. 341 pp. $45, £35

Jared Diamond

Between 1000 BC and AD 400, the prehistoryof sub-equatorial Africa was transformed bythe expansion, over most of the region, offarmers speaking Bantu languages. A newbook by Christopher Ehret transforms ourunderstanding of this event. Ehret gives usrich insights into big questions of world history, while also showing how prehistorycan be deciphered with a method termed linguistic archaeology.

The central problem of sub-equatorialAfrican history is as follows. Today, nativepeoples of the region speak languagesbelonging to six groups: Bantu (by far themost widespread), Central Sudanic, EastSahelian, South Nilotic, South Cushitic andKhoisan. Different traditional peoples of theregion practise different mixes of farming,herding and hunting–gathering. Farmers usedifferent mixes of West African and Indone-sian wet-climate crops and Sahel and Red Seadry-climate crops. None of the importantcrops or domestic animals are native to sub-equatorial Africa; all must have been broughtin from outside. What peoples invaded, atwhat times and from what directions, andwhich crops, livestock and other cultural baggage did each invader bring? What advan-tages enabled the Bantu to emerge from thathuman melting pot with a much wider distri-bution than the other five groups?

If only we had tape-recordings of conver-sations from African villages 3,000 years ago!Analysis of those recordings would go a longway towards answering our questions. In theabsence of recordings, one can still try toreconstruct some by using linguistic archae-ology. One thereby aims to deduce culturalattributes, population interactions andmigrations in the preliterate past, not by theusual archaeological methods of excavation,but by analysis of modern languages. Amongthe members of a modern language family,the analysis maps the distributions of wordroots that were evidently borrowed fromvarious other languages at various pasttimes.

Consider these examples from Indo-European languages that will be more familiar to most Nature readers than Bantulanguages. The word for ‘sheep’ is similar inmost Indo-European groups: ewe, ovis, avis,ois and oi in English, Latin, Sanskrit, Greekand Old Irish, respectively. The same is truefor ‘fart’: fart, pardate, perdet, perdzu and

perdo, in English, Sanskrit, Russian,Lithuanian and Greek, respectively. Theseshared words permit reconstruction of theroots ‘owis’ and ‘perd’ for sheep and fart in the ancestral proto-Indo-European lan-guage spoken at some time in the preliteratepast. But the word for ‘gun’ is entirely differ-ent in different Indo-European languages:gun in English, fusil in French, Gewehr inGerman, ruzhyo in Russian, and so on. Theinference is obvious: proto-Indo-Euro-peans, who probably expanded out of WestAsia around or before 4000 BC, already hadsheep and farts, but no guns, so that differentmodern Indo-European languages shareancient inherited roots for sheep and fart butindependently coined new words for gunsupon their recent invention.

In Ehret’s hands, linguistic archaeologyyields even clearer and richer conclusionswhen applied to sub-equatorial Africansthan to Indo-Europeans. The reasons for thisare: sub-equatorial African languages are sodiverse and distinct (six stocks and manysubstocks); the main population expansionwas more recent in sub-equatorial Africa(1000 BC–AD 400) than in West Eurasia(7000–4000 BC); and, as a result, ancient sub-equatorial language stocks are still mostlywell represented by living languages, where-as many ancient European stocks are nowextinct and can only be inferred from strangeword roots in existing languages (for exam-ple, Greek words evidently borrowed from a now-extinct pre-Indo-European languageof Greece). In broader or narrower subsets of

the now-dominant Bantu language family,Ehret recognizes words evidently borrowedfrom each of the five other stocks.

For instance, all Bantu substocks sharesimilar words for wet-forest animal speciessuch as leopard and hippopotamus, but dif-ferent Bantu substocks have different wordsfor East African dry-grassland species suchas zebra and ostrich. This suggests that theBantu who colonized sub-equatorial Africaarose from a wet-forest environment and didnot encounter East African dry grasslanduntil later in their history. While this princi-ple is familiar to linguists, I am impressed bythe wealth of detail that Ehret extracts whenhe applies it to African languages. For exam-ple, he demonstrates that, in their eastwardsexpansion, the Bantu interacted successivelywith Central Sudanics, East Sahelians, South Cushites, South Nilotes, Khoisan andIndonesians in approximately that sequence(because Bantu words for items such assorghum, iron, cow, black and white cow,click consonants and chickens are loanwords from these language stocks, respec-tively, but are confined to successively narrower subsets of Bantu languages).

As one example of Africa’s lessons forworld history, the spread of food productionaround the world from its few prehistoriccentres of origin is thought usually to have involved the spread of farmers andherders themselves, bringing their crops and livestock with them and replacing local hunter–gatherers. The best-attestedexample known to me of the alternative explanation — local hunter–gatherers

NATURE | VOL 398 | 8 APRIL 1999 | www.nature.com 477

Bantu banter and African rootsbook reviews

Speech marks: linguistic archaeology can trackhow this Khoisan woman’s language . . .

. . . was upstaged by the language spoken by thisBantu woman.

CO

RB

IS/P

ET

ER

JOH

NSO

NC

OR

BIS

/ CH

AR

LES

& JO

SET

TE

LE

NA

RS

© 1999 Macmillan Magazines Ltd

acquiring livestock or crops from food-pro-ducing neighbours — comes from SouthernAfrica, where Khoisan hunter–gatherersacquired sheep and cattle to become theancestors of modern so-called Hottentotherders. Ehret identifies loan words for ‘ram’,‘young ram’, ‘milk ewe’, ‘grain food’ and ‘por-ridge’ in a reconstructed early Khoisan lan-guage. Astonishingly, though, these loanwords have not been borrowed from Bantulanguages, which are now the sole nativefarming languages in Africa south of theZambezi River, but from East Sahelian lan-guages, now confined to regions over 1,000km north of the Zambezi. This suggests avanished former thrust of East Sahelians farsouth of their modern distribution. Here is ahypothesis awaiting testing by archaeolo-gists: will excavations yield any evidence ofEast Sahelians 2,000 years ago living as farsouth as Malawi, Zimbabwe and Botswana?

Ancient East Africa was evidently a melt-ing pot of peoples adapted to different life-styles: the Bantu with their West African wet-forest root crops and goats; Central Sudanicsand East Sahelians near the West Rift Valley,with their dry-country grain crops, sheep andcattle; South Cushites in the Eastern grass-lands and South Nilotes at high altitude, withtheir cattle; and Khoisan hunter–gatherers inareas unsuitable for agriculture.

As each group of food producers trickledinto East Africa, they gravitated towards envi-ronments similar to those they had left — justas, during the European colonial era, Spanishimmigrants preferred warmer areas of theNew World, while British immigrants pre-ferred cooler areas of North America, Aus-tralia, South Africa and the Kenya highlands.Bantu latecomers were able to gain a footholdin East Africa because their wet-forest agri-culture let them farm areas unsuitable for thedry-country grain agriculture and herding ofthe other food-producing groups.

Once the Bantu had added grain crops,cattle, sheep and iron tools to their originalinventory, they found themselves in posses-sion of an economic package that wasunstoppable in the subequatorial Africa ofthose times. Within a few centuries of reach-ing the east coast of Tanzania, Bantu food-producers advanced southwards for severalthousand kilometres, through areas onlythinly populated by Khoisan hunter–gather-ers, to reach South Africa.

Why, out of this human melting pot, wasit the Bantu and not one of the other food-producing groups that developed theunstoppable package? Ehret hints at possiblecultural factors — the wild card of environ-mental historians. Before the Bantu arrived,Khoisan hunter–gatherers had coexisted fora long time in East Africa with Sudanic, Sahe-lian, Nilotic and Cushitic farmers, becausemost of those food-producers did not hunt.The Bantu did hunt, so there was no roomleft for separate Khoisan hunter–gatherer

societies, which were speedily out-competedand eliminated. But why did not the Sudan-ics and those other peoples hunt and elimi-nate the Khoisan earlier? And why was it theBantu who added dry-country crops to theirwet-country inventory and emerged the vic-tors, instead of all those dry-country food-producers adding Bantu wet-country cropsto their own repertoire and emerging the vic-tors? Why, in East Africa’s environmentalmosaic, was there such an extensive melting-together of economically distinct societies,whereas India’s superficially similar environ-mental mosaic continued to harbour specialized groups of farmers, herders andhunter–gatherers until modern times?

Now that Ehret’s linguistic analyses havebrought these questions into focus, archae-ologists can test his hypotheses for Africaitself, comparative historians can use Africanhistory to illuminate world history, andcomparative linguists can imitate hisapproach elsewhere in the world.Jared Diamond is in the Department of Physiology,University of California Medical School, Los Angeles, California 90095-1751, USA.

For Charlie and NickThe Shape of Actions: What Humansand Machines Can Doby Harry Collins and Martin KuschMIT Press: 1998. 212 pp. $26.50, £15.95

Bruce Mazlish

As the human species enters a new millenni-um, it is engaged in a fateful intensification ofa long-standing question: is there any differ-ence between humans and machines, and, ifthere is, what is it? A few thousand years ago,Chinese automata posed the issue physically,with mechanical figures putatively lookingand functioning like their makers. How wasone to tell the difference? The Greek con-struction of automata also gave actual fea-tures to what we can call the ‘android look’.By the seventeenth century, René Descarteshad posed the question philosophically in hisclassic Discourse on Method, claiming thatthere were two certain ways to recognize thedifference. Put in the simplest terms, his twocriteria for distinguishing between man andmachine were that the latter has no feedbackmechanism (“it could never modify itsphrases”) and no generalizing reason (“rea-son is a universal instrument which can beused in all sorts of situations”).

Needless to say, the debate raged fiercelyin Descartes’ time, and has continued eversince. The issue is omnipresent even in suchan unlikely place as the childish Oz stories ofthe early twentieth-century writer FrankBaum. One need only think of the Tin Man,but the mechanical creatures of Ozma of Oz

serve as especially fascinating examples.Here we meet the Wheeler, formed like aman but moving on wheels rather than feet,and Tiktok, the machine man who “thinks,speaks, acts” and “does everything but live”.Having been fitted with a brain, he claims thepower of ratiocination. What then distin-guishes him (it?) from humans? We are,clearly, only a short step away from deviceswith artificial intelligence (AI), where, inpractice and not just fictionally, mentalactivity is added to physical activity as part ofthe repertoire of mechanical ‘beings’.

It is in a context such as the above that weneed to place The Shape of Actions. One of theauthors, Harry Collins, is a sociologist whoteaches at the University of Wales. The other,Martin Kusch, is a philosopher who lecturesat Cambridge University. Both have manybooks to their credit, and Collins is knownfor his work on the sociology of science, andespecially for his embrace of the notion ofthe social construction of science (that is,that scientists do not so much discover whatis in nature, as impose certain conventionalideas on it), and the role of tacit knowledge(knowledge acquired through a collectivesocialization process and irreducible to pre-cise formulae) in the actual doing of science.Together, they claim to be doing “philosoph-ical sociology … following the tradition ofthe later [Ludwig] Wittgenstein (1953) andof [Peter] Winch (1958)”.

Well, we may ask, what does philosophi-cal sociology have to say about thehuman–machine relation? The authors’great insight, announced with much fanfareand claim to originality, is a version of thephilosophy of action. Their theory of mor-phicity, they tell us, goes like this: “Humanscan do two kinds of actions and — likemachines — they can merely behave. Whenhumans do polymorphic actions, they drawon their understanding of their society;when they do mimeomorphic actions, theyintentionally act like machines — entitiesthat do not need to understand society …”Machines, a term they define broadly, theauthors conclude, are incapable of poly-morphic action. Basically, the rest of theirbook is a variation on this theme.

Specifically, they apply their analysis toall manner of things — letter writing, golfswings, bike balancing and riding, restau-rants (especially McDonald’s), armies, andother such examples of ‘mechanical’ possi-bilities. Their text is filled with tables andbranching trees, which we are encouraged toscramble up and down in seeking to estab-lish the differences between humans andmachines. (The boundaries, in fact, as theauthors point out, are permeable and shifting.) Everywhere, we encounter types,and types within types: we are told that, typically, there are four types of action, threeof which are polymorphic. These three are then divided into “open”, “occasional”

book reviews

478 NATURE | VOL 398 | 8 APRIL 1999 | www.nature.com