Words, trees, and the dispersal of iron working in sub-Saharan Africa: Some explorations of a computational linguistic approach to tracing the spread of words for ‘iron’ across Africa

Keywords
Loading...
Thumbnail Image
Authors
Issue Date
2016-01-29
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Problems in our knowledge of human (pre)history are always best looked at from multiple angles, studied from multiple disciplines and approached with different methods. The issue treated in the current paper, where and when iron working spread across sub-Saharan Africa, is a good example. The archaeological literature on the earliest iron technologies in Africa is plagued by a number of problems to do with radiocarbon dating. C14 dating is of limited use due to (a) the nature of the material that is typically dated (i.e. charcoal), which cannot always be trusted to be contemporary with the iron working activities themselves, and (b) the 'black hole' on the radiocarbon calibration curve between 800-400 BCE that prevents us from obtaining precise dates for the early iron age in Africa. To supplement archaeological data, historical linguists and linguistic historians have tried to derive where iron initial! ly spread by looking at the dispersal of iron-related vocabulary across the continent. In this literature, the direction of the spread of a certain word is derived, among others, from the shape of its current distribution, and where the most shattered areas are. An issue that crops up there, however, is that scholars have been interested more in those areas where they can find common vocabulary and have neglected possible data coming from areas where less such commonality is found. For this reason, linguists have tended to equate ‘absence of evidence’ with ‘evidence of absence’. The current paper reviews the previous archaeological and linguistic literature and tries to find some remedies for this ‘absence of evidence’ issue in the linguistic literature. One solution that is explored is the quantification of diversity in terms for ‘iron’ across sub-Saharan Africa. Applying automatic cognate detection to a dataset of 1,559 words for ‘iron’ gathered from various sources, we can not only get an idea of where related words for ‘iron’ are used (and hence where we can ascertain iron-related contact between communities), but also of where most diversity in ‘iron’ terms is found, and where a lot of lexical replacements have assumably happened. When applied to the data, this method allows us to say something about the likelihood of iron arriving in the Bantu-speaking area from the east vs. from the west, and gives some interesting new data on diversity across West Africa that remains to be explained. The second possible solution involves ! building phylogenetic trees from individual cognate sets, which can then be projected as areal sequences on a map. For studying the dispersal of iron working across sub-Saharan Africa, however, this second method appears to be of limited value, and does not provide any new results.
Description
Citation
Faculty
Faculteit der Letteren