Amerind personal pronouns: A reply to Campbell

DISCUSSION NOTE Amerind personal pronouns: A reply to Campbell Johanna Nichols and David A. Peterson University of California, Berkeley Introduction. LyIe Campbell's comments (Campbell 1997a) on Nichols & Peterson 1996 appear to be based on misunderstandings. Our paper surveyed pronominal systems with first person ? and second person m, a mini-paradigm well attested in western America, and showed that (a) the ? : m system is not sufficient to prove, or even to strongly suggest, genetic relatedness among the languages exhibiting it and (b) nonetheless its distribution can hardly be due to universale or chance. We argued that, though neither genetic marker nor universal, the ? : m system can nonetheless be given a historical interpretation. It is one of several features with nearly identical Pacific Rim distributions (including tones, numeral classifiers, and others we did not discuss), and the set ofcongruent distributions strongly suggests that language populations all around the Pacific Rim, and only there, include a good many lineages ultimately dispersed from the same founding population. For the dispersal a chronology (both absolute and relative), a center, and a trajectory can be reconstructed, although, given the great time depth involved, neither a mechanism of dispersal (migration? diffusion?) nor a descent (are all n.m language families sisters? some of them? none?) can be securely established . Our point was that, despite this indeterminacy, a fairly determinate origin and history can be posited. We posited a late phase in the linguistic colonization of the Americas, in which a few successive entrants—not necessarily sister languages—all emanated from the same coastal Asian population, brought markers of that population (such as n:m pronouns, numeral classifiers, etc.) into America, and retained their coastal orientation in a southward spread from the Beringian point of entry well into South America. The set of entries began at about the end of glaciation and had ended by about 6000 bp (the chronology follows from the geography: the markers of the Pacific Rim population are nearly continuous around the Pacific, but are stopped in Australasia at the point where postglacial sea-level rise separated Australia from New Guinea, and are cut off in the far north by the more recent Chukchi-Kamchatkan and Eskimo-Aleut expansions). Our reconstruction is consistent with the linguistic facts and with what can plausibly be assumed about the colonization of the Americas. We would welcome other interpretations of the evidence; the main point of our paper was that the distribution of n:m pronoun systems cannot be dismissed as due to random chance or to universale. We have organized Campbell's remarks into two groups: those having to do broadly with method (sampling in general, construction of our sample, how we surveyed pronouns ) and those having to do with our interpretation of the results. We quote or summarize each of his major criticisms in bulleted paragraphs, with our response in the following paragraph.1 1. Method. Our paper used a genetically based worldwide sample of 173 languages, broken down into 10 geographically defined large areas. We gave very strict classifica1 Campbell has recently published the same critique (repeating parts of it verbatim) in Campbell 1997b: 247ff. 605 606LANGUAGE, VOLUME 74, NUMBER 3, (1998) tions of pronoun categories, allomorphs, and phonotactic positions, and used them to determine the frequencies of pronominal patterns such as 'initial « in a basic allomorph of the independent first-person singular pronoun and initial m in a basic allomorph of the independent second-person pronoun', 'n as first consonant in any first person singular form and m as first consonant in any second person singular form', etc. We compared the proportions ofthe sample languages in our ten areas that displayed various pronominal consonant patterns, and used standard statistical techniques to determine whether the differences in these proportions were statistically significant; they were, so we said (again using standard procedure) that the high frequency of ? : m systems in western America cannot be dismissed as due to accident or universals. Several ofCampbell' s remarks are not principled objections but seem to us to indicate unfamiliarity or unease with the idea of sampling and the fundamentals of survey design and statistical analysis. We begin with some of these. • Over 109 (over 63%) of the languages...