BALEE & MOORE: SIMILARITY AND VARIATION IN PLANT NAMES 211

five languages: Arawetd (Ar) pinuwa-'i, Asurini (As) pinuwa-'iwa, Ka'apor (K)
pinuwa-'4, Temb6 (T) pinuwa-'iw, Wayapi (W) pino. By contrast, the words
for 'moela de mutum' are bewilderingly different: Ar iwa-pedi, As iwa-kaw-
'iwa, K kupapa-'iran-I', T iwa-zu-'iw-ran, W mitil-'ay.
 This paper attempts to answer this question, at least partially, for the
semantic domain of ethnobotany, by investigating similarity and variation
among words for a given corpus of plant species in five different languages of
Tupi-Guarani. These five languages are spoken in a broad arc in lower
Amazonia. The possible factors that may a priori help explain why words for
some plant species are similar while words for others vary across languages of
the same family include (1) cultural ones, such as plant utility and/or
management; (2) geographical ones, such as proximity and similarity of
environments; (3) diachronic linguistic ones, such as borrowing and degree of
genetic relatedness, as well as (4) the linguistic properties of the words used to
designate plants, including their morphological or semantic structure. An
investigation of such factors, to our knowledge, has never before been carried
out with regard to South American languages. We have collected data on
similarity and variation of words for 167 botanical species native to the
neotropics in the Arawet6, Asurini do Xingu, Ka'apor, Temb6, and Wayapi
languages of eastern Amazonia (Tables 1, 2, and 3). Although these data were
collected initially for non-linguistic purposes, they are highly appropriate for
the investigation of factors involved in similarity and variation in plant words
among different languages of the same family.
 First, the five languages are dispersed in four linguistic sub-groupings of
Tupi-Guarani (A.R. Rodrigues 1984/85; A.R. Rodrigues, pers. comm. 1988),
with only Wayipi and Ka'apor being classified in the same sub-grouping.
Second, these five languages are spoken in three ecologically diverse regions:
the Xingu River basin of north-central Brazil for Arawet6 amd Asurini, the
Gurupi/Turiaqu River basins of extreme eastern Amazonia for Ka'apor and
Temb6, and the Oiapoque River basin of northern Amazonia for WayIpi.
Third, although all five groups are horticultural, they exhibit notable
differences in crop staples and patterns of utilization of non-domesticated
species. For example, the Arawet6 rely heavily on maize, in contrast to the
other groups who are more dependent on tubers; the Asurini traditionally
eschewed hog plum (Spondias mombin L.), which is an esteemed edible fruit
of the other groups. Fourth, collections and determinations of voucher
specimens for most of the 167 species in our sample have been obtained for the
five languages. Finally, the corpus of data is large enough to test statistically
propositions regarding similarity and variation in plant words across the five
languages.
 Some years ago, Brent Berlin and his colleagues (Berlin et al. 1973)
published a pioneering paper on the retention of plant words in two Mayan
languages. They proposed that such retention reflected the cultural