¹ In the original article—Quantitative Expression of Cultural Relationships, Un. Cal. Publ. Am. Arch. Ethn. 31.211–56 (1932)—the letter values are reversed, c denoting common elements, a and b those not shared by the two populations. In his formula 6 Ellegård cites (143) a simple measure p = a/(a + b + c) which ‘may be of help to those who find it difficult to see intuitively what r_n means’, though r_n is preferable to p ‘for statistical work’. This measure was also proposed, as ‘T’, by Driver and Kroeber in 1932.

² Significant in what they show, not ‘statistically significant’, although in general it is the high coefficients and the very low ones that are statistically most significant also, i.e. farthest from randomness.

³ New problems, though somewhat analogous ones, will arise when we shall have enough information on other old IE languages like Tocharian, Illyrian, Thracian, and Phrygian to want to draw them into the comparative picture. (Albanian presents a somewhat separate problem because of the recency of its data, compared with the languages we are here dealing with.) For most of these, we may never have full data on Meillet's 74 features; and to compare incomplete with complete arrays of features for expression in coefficients seems basically a very dubious procedure. However, when the time is ripe, the incomplete languages can be tested in special, separate studies, on the basis of such an array of traits as are known from them, comparing them with the most relevant of the ‘complete’ languages; and they may then be hooked into the general framework of these where indicated as appropriate. By then, also, it is to be hoped that further analysis may have much enlarged the present Meillet list, and we may expect to see the framework somewhat differently.

Another opportunity of the future may lie in supplementing the present tendency of dealing primarily with the oldest forms of speech—our ‘Germanic’ of course is really Proto-Germanic, and so on—and attacking contemporary languages like French, English, Hindi, and Modern Persian for their likenesses and unlikenesses. The present methods would probably not be applicable to this sort of problem, because the differences in the phenomena are so great that coefficients would mostly run so uniformly low as to have little intellectual significance—let alone the disconcerting expectability of their lacking statistical significance. However, a more fully developed typological index, such as Greenberg has broken the ice for, may well be able to handle this problem with considerable adequacy.

⁴ 45 occurrences of avoidance primarily between husband and his wife's kin × 65 occurrences of uxorilocal residence ÷ 350 tribal units investigated = 8.4 chance distribution, as against 13 actual cases of association found. N = 350 probably includes some common absences (d), but also cases of lack of data, and the class of common absences is not recognized as such.

⁵ c was then used for modern a, a for a + b, and b for a + c, in computing A, G, T.

⁶ The fault mentioned above in treating doublets like Celto-Italic coordinately with units like Germanic or Greek is similar to a botanist's equating supergenera coordinately with genera.

⁷ Lg. 15.71. The idea and its expression were Chrétien's.

⁸ It is interesting that it is Italic rather than Celtic that enters so preponderantly into the very low-value Q coefficients which prove significant. With r_n = G, it was Celtic that appeared most distal. With Q₆ the four lowest nonsignificant coefficients all involve Ce: those with Ba, Sk, Ir, Gr, for values of –30, –30, –30, –25. For those four Ce coefficients the a‘s run higher than for the five It lows (5–7 as against 3–7), the d‘s also run higher (25–31 as against 18–24), and (b + c) runs lower (37–41 as against 43–52). But in both groups the a values are so insignificant compared with the d values, that on suppression of the latter in converting to r_n, the differences b + c overweight the unsupported a agreements even more in Celtic than in Italic, mostly.

⁹ Thus also, Ge-Sl 110 c, 76 m; but Ge-Ir 122 c, 133 m is an exception. The lines for these two values have been omitted so as not to overcrowd the diagram.

¹⁰ A. S. C. Ross, on p. 29 of his paper cited above in §2.2, says: ‘To the philologist the presence of a root’ (discussed in comparison with the philological features of Kroeber-Chrétien) ‘in two languages only (X and Y) would seem to be worth more as evidence for their close relationship than would its presence in X, Y, and one or more other languages; the method takes no account of such factors.‘ This is true if the point is to show the closeness of relationship of X and Y. Our problem however was to investigate the relative closeness of all IE branches—the taxonomy of the family as a whole. My point above is that I do not know in which type of question Meillet's interest was strongest.

¹¹ D. G. Macrae, in Darwinism and the Social Sciences, A century of Darwin 311 (S. A. Barnett, ed.; 1958), says: ‘What, in all probability, sociology most needs at the moment is not either a Newton or a Darwin, but a Linnaeus to elaborate a really workable classification of social structures and of the range and variety of institutional patterns and sequences.‘

¹² This view of the situation during the last century and a half sheds considerable light on the anomaly of linguistics having successfully operated on a genetic basis without having concerned itself seriously with a general taxonomy. It suggests that the assumption of genetic unity in linguistics was partly parallel to the positing of hypothetical origins in culture—although better evidenced. The whole tangle of the complexity of actual history was thereby avoided in both cases. I am not implying that the positive results of the comparative study of Indo-European, Semitic, Sino-Tibetan, Malayo-Polynesian, or Athabascan are about to be overthrown or discarded. They are founded and they will stand. But they may be considerably modified by the recognition of other processes which comparativists have mostly ignored; and the hitherto accepted results will at any rate have to be viewed as forming only part of a larger web of causes and effects.

In twentieth-century study of culture, solid and positive results of assumptions of unitary origin and unfolding development have been very much fewer than in linguistics, and the impending revision in understanding may accordingly be smaller, or even in reverse direction. This would be so because, first, the cultural findings of development from a single origin have usually been world-wide and universal, but linguistic ones have been limited to one family of speech at a time out of many; and second because the cultural findings have been supported by less systematic evidence. Still, it does seem possible that some portions of Tylor's, Morgan's, and Frazer's findings may ultimately be revindicated, now that there is no longer danger that the massive effects of contact and diffusion will be underweighted as they used to be on the ground of being perhaps undeniable but also accidental, random, of little significance, and discardable.

¹³ Thus, Simple Mixed-Relational French, which ‘might nearly as well have come under D’, Complex Mixed-Relational, is also Fusional and mildly synthetically Analytical; Complex Mixed-Relational Nootka is ‘very nearly Complex Pure-Relational’ and is also Agglutinative with a symbolic tinge and Polysynthetic; Chinook is Complex Mixed-Relational, Fusional-Agglutinative, and mildly Polysynthetic; Polynesian and Haida are both Complex Pure-Relational and Agglutinative-Isolating but differ in being respectively Analytic and Polysynthetic. (Sapir, Language 150–1 [1921]).

¹⁴ It might be suggested, if comfort is needed, that even in biology, where phylogenies are simple in that mainly they only diversify, and we do have extinctions but no positive complications through convergences and recombinations—even in biology the natural genetic system of life branches or classes works clearly for the Vertebrates, and nearly as well for the Arthropods, but breaks down if we try to trace the connection of the Chordate and Arthropod phyla.