Anthropology

by A. L. Kroeber

170. Limitations on the Diffusion Principle

To essay, by the mere principle of converting spatial extent into temporal duration, an accomplishment of such magnitude and ultimately of such complexity as this, may seem simplistic; and it would be. The distribution principle may be the most useful of the weapons in the ethnologist’s armory. But it requires supplement and qualification.

Fig. 33. Schematic illustration of distributions of culture traits as indicative of their history. A, distribution corresponding to one by accident, and suggesting that each occurrence is independent of the others. B, distribution by contiguous occurrences, strongly suggesting a single invention and subsequent diffusion. C, distribution interpretable as due either to independent, parallel origins; or to a single origin, diffusion over the whole area, and subsequent loss of the trait in most parts, with survival only in marginal tracts. The loss in the central area might be due to the growth of a supplanting trait, whose later diffusion had not yet penetrated to the farthest ends. D, distribution suggesting a single origin old enough for its diffusion to have become extensive, but checked in certain directions by adverse conditions in nature, communications, or cultural preoccupation. The specific demonstration of such adverse factors would substantiate the interpretation.

First of all, it is obvious that spatial extension must not be measured mechanically. To work on the assumption that a custom or art practised over a million square miles was a third as old again as one practised over seven hundred and fifty thousand, would be too often contrary to the evidence of known history as well as the dictates of reason. Culture traits do die out, from inanition, from sterility of social soil, through supplanting by more vigorous descendants. Continuity is therefore not a necessary ingredient of geographical range. An ancient trait may have been displaced in all but a few remote peripheral tracts. The areas of these may aggregate but little. Yet the distances between them are likely to remain greater than the longest range of a later trait which has replaced the earlier one over most of its original territory.

Thus, alphabetic writing is more recent than the ideographic and rebus methods, but in the year 1500 A.D. was in use over a larger area in Europe, Africa, and Asia than the surviving Chinese and Mexican systems occupied. Yet these two outlying systems enclosed between them a larger tract than those over which the alphabets had diffused.

So, at the same period, was agriculture practised by peoples holding more area than was occupied by non-agricultural ones. But the former constituted two great and continuous groups, one in each hemisphere, to which the non-agricultural peoples in the north of Asia, the south of Africa, the remote continent of Australia, the north of North America, and the south of South America were obviously peripheral. Agriculture being of necessity later than the non-agricultural state, and there being thus no doubt that the marginal hunting peoples represent the remnant of a condition that was once world-wide, it appears that there must be a presumption of validity in favor of reckoning the extent of a scattered custom by its included rather than its actual area.

Of course, the situation is not always so simple. There may exist the possibility of two or more marginal areas sharing a trait as the result of parallelism. Half-hitch basketry coiling in Tasmania and at Cape Horn might logically be the last survival of a very ancient world-wide diffusion, or the product of two thoroughly independent inventions, or of parallel processes of degeneration in isolated and culturally unstimulated nooks. The last two interpretations in fact seem more conservative than the first. If half-hitch coiling were as antecedent in its nature to other coiling and to weaving as wild foods are to cultivated ones; or if the Old Stone Age remains showed it to have been actually so; or if it were practised by a considerable number of tribes in four or five rather large marginal areas instead of two quite narrow ones, diffusion, and the consequent antiquity of the trait, could be inferred with high probability. In short, the periphery argument must not be stretched too thin.

Obviously, too, comparables must be compared: coiling with twining, hand-weaving with loom-weaving; not, however, the very special variety of half-hitch coiling with the entire array of weaving techniques. Nor would it be fair to balance the whole group of true alphabets in the year 500 B.C. against the particular rebus system of Egyptian hieroglyphs from which they were possibly derived but which they had already much exceeded in their diffusion. Yet the distribution of all alphabets as against that of all ideographic and rebus systems would lead, at that date as two thousand years later, to the same interpretation that the facts of history actually give.