BALEE & MOORE: SIMILARITY AND VARIATION IN PLANT NAMES 217

criteria exist for distinguishing between categories more or less encompassing
than the species (see Gould 1980:206) with the obvious exception of the
individual plant. For logistical reasons, it was not feasible to obtain responses
in the five languages for individual plant specimens.
 One of the criteria for including species in our list was that they be of
neotropical origin. This is because names for non-neotropical species, clearly,
would be most likely introduced and hence of non-Tupi-Guarani origin as well.
There is some doubt, nevertheless, about the origins of a few species included
in our analysis, all of which are domesticates. These include papaya (Carica
papaya L.), bananas and plantains (Musa spp.), and bottle gourds (Lagenaria
siceraria Mol.). As for papaya, it has been most recently argued on botanical
grounds that it is a New World cultigen (Storey 1976:23); moreover, Sousa
(1974: 99) refers to its introduction in 16th century Bahia, indicating that it
came from Pernambuco to the north. Although the genetic evidence indicates
a Southeast Asian origin for bananas, Smole (1980) argued that Musa spp.
existed in the neotropics in pre-Columbian times. Early 16th century explorers
noted that the Tupinamba cultivated bananas and called these pakoBa (L6ry
1960:157; Lisboa 1967:122; Sousa 1974:98; Vasconcellos 1865:136), a
reconstructable term in Proto-Tupi-Guarani (A.D. Rodrigues, pers. comm.
1988). Bottle gourds were also cultivated by the aboriginal TupinambB (Sousa
1974:95). The bottle-gourd probably probably arrived in South America via
Africa in remote pre-Columbian times (Heiser 1979:114-116). Although it is
probably not, therefore, a true native to the neotropics, it seems unlikely that it
was introduced by human beings (but see Lathrap 1977). This means that
there is no a prior reason to assume that the name for it in modern Tupi-
Guarani languages was introduced. Our exclusion of "borrowed" plants is, first,
an attempt to exclude borrowed words. All domesticates here included are of
sufficient antiquity in the neotropics (i.e., probably older than the five
languages in our analysis) that they can be considered for historical linguistic
purposes to be neotropical.
 One other requirement for species inclusion in our comparison concerns
the number of responses. Only species for which names in three or more of
the five languages were obtained are included. This is to guarantee that each
species occurs in at least two of the three ecological regions. If species were
included where there were only two or more responses, then the ecological
region occupied by proximate groups (the K/T and the As/Ar) perhaps would
be overrepresented in the lists of species. It is plausible, moreover, that
linguistic borrowings are more likely to exist between neighbors. The three-or-
more rule, then, is one more means of controlling the possible occurrence of
borrowed words between the five languages in the sample. As will be seen, we
are able to draw statistically significant conclusions about similarity and
variation in plant words between these languages on the basis of the data.