In lieu of an abstract, here is a brief excerpt of the content:

REVIEWS929 Historical linguistics. Edited by Barron Brainerd. (Quantitative linguistics, 18.) Bochum: Brockmeyer, 1983. Pp. 236. DM 29.80. Reviewed by Anthony J. Naro, Universidade Federal do Rio de Janeiro The title of this book is somewhat misleading, even given the series name. The papers are devoted almost entirely to problems involving type change in the lexicon—or, in some cases, change over time in the frequency of use of lexical items. Most of the authors look to Morris Swadesh as the founder of this field of inquiry; they attempt to develop, modify, and apply his methodology in the light of facts or problems that arise in various language families. Nonetheless, the book presents some innovative techniques, and even some rare non-lexical, non-quantitative material. In the first article, 'Incorporating borrowing rates in lexicostatistical tree reconstruction' (1-24), Sheila M. Embleton provides a good (although obviously biased) survey of developments to date in the field. This part would have been much more useful if E had provided a balanced discussion of the many controversies and confusions of the 60's and early 70's, which still leave most linguists with severe doubts as to the validity of any Swadesh-like methodology . In any event, she reviews a series of possible revisions to Swadesh's original assumption that the rate of change is a constant multiplied by the number of remaining cognates; she also discusses David Sankoffs method, originally intended for use in biology, of reconstructing an evolutionary tree based on a similarity measure that considers only proportions of cognates, and postulates constant probabilities of internal word replacement and external word borrowing. In a computer simulation of Sankoffs method, E finds that satisfactory reconstructions can be achieved with a 200-word list—which, however , need no longer be constrained to rule out borrowing, since this factor is now explicitly considered. She goes on to propose modifying Sankoffs method to allow replacement and borrowing rates to differ for different languages and language pairs. Finally, she proposes to test the resulting method on real data from the Germanic languages; but this has not yet been carried through. Abstract, intuition-oriented linguists, on the one hand, will of course reject E's paper (as well as most of the others in the volume), since they would never countenance the idea of measuring similarity between languages, or any other significant property, on the basis of proportions of cognates. Real-world, data-oriented linguists, on the other hand, will be severely disappointed by the lack of any empirical proof of goodness of fit. E (p. 3) approvingly quotes Sankoff 1972 as having stated, correctly, that in models like that under consideration, 'any degree of deviation is predicted to occur in a certain proportion of cases'. However, it is wrong to infer from this, as E does in her next sentence, that empirical evidence is therefore in effect irrelevant. If the model is good, the proportion of cases in which significant deviation occurs will be low. In fact, the only reason Sankoffs models in the area of linguistic variation are now so widely accepted is precisely the fact of their good fit to real data in a large number of cases. Until such time as parallel results can be presented for lexicostatistic models, linguists are fully justified in remaining skeptical. ? stochastic model for language change' (25-49), by Brainerd, uses the same basic Swadesh-like framework, but different mathematical techniques. B first discusses, unfortunately without explicit reference, Whinnom's relexification hypothesis (1956, 1965), which postulates that pidgins arise from previous pid- 930LANGUAGE, VOLUME 60, NUMBER 4 (1984) gins as a result of shift in lexical items from one language to another, while retaining essentially the same reduced grammar. This would seem to be the ideal linguistic site for application of lexical methods, since the syntax is assumed to be constant, but one critical point must not be forgotten: the theory of relexification is intended to explain only the derivation of one pidgin from another—it is irrelevant to the problem of the origin of pidginization. Thus relexification would be applicable to the transition from the original Romance (Portuguese) pidgin to English-based pidgins, but not to the...

pdf

Share