BULLETIN FLORIDA MUSEUM NATURAL HISTORY VOL. 35(4)


If the range of meaning were not restricted and "looking up and down the list"
were permitted, then from the point of view of Temb6, both words would have
a similar Ka'apor counterpart, on the basis of the head terms, iwiri. But
paradoxically, from the point of view of Ka'apor, only one of the words has a
similar counterpart in Temb6. The situation would become yet more confusing
when considering five instead of two languages. Further, a skewing would
result in that a pair of languages both having relatively complete collections
would offer a greater possibility of finding similar words than would a pair of
languages both having relatively incomplete collections.
 So it is necessary to restrict the range of meaning of the referent, then
examine whether the words for it are similar or different. We opted to restrict
referents to the taxonomic rank of botanical species. This is because one may
argue that the botanical species is the most objective level of abstraction for
distinguishing between individual plants. The species is more objective than
higher order units, such as tribes, genera, and families, since "rank is not
inherent in supraspecific groups" (Cronquist 1968:31) [2]. Species are natural
units, not products of mind (see Gould 1980:204-213).
 Species, for our purposes, are also more suitable referents for comparing
indigenous plant names than taxa of infraspecific ranks. The classification of
many neotropical cultivars (i.e., varieties) of a single botanical species is far
from possessing taxonomic exactitude. In their taxonomic revision of the genus
Manihot (which includes cultivated manioc, Manihot esculenta Crantz), for
example, Rogers and Appan (1973:34) observed that "It is impossible to apply
formal subspecific taxon epithets to fleeting variants which are not related to
some precise geography or ecological region." In an exhaustive study,
Albuquerque and Cardoso (1980) discussed several possible means of
classifying manioc cultivars. One classification was based on color of the tuber,
yielding only three basic types: white, yellow, and cream. Each of these types
had sweet, bitter, and sweet/watery sub-types. Each sub-type was further sub-
divided into cultivars that had erect stem habits vs. ones that showed branching
stem habits. The total number of phenotypically distinct cultivars based on this
classification, then, would be only 18 (cf. Albuquerque and Cardoso 1980:138-
139). Another proposed classification scheme, based on floral parts, yielded
only 6 possible cultivars. The number of phenotypically distinct manioc
cultivars named and recognized by the Ka'apor, however, is at least 19 (Bal6e
and G61y 1989:138); the WayApi name as many as 29 (Grenand 1980:310). The
lack of correspondence between folk species and botanical infraspecific taxa is
not limited to manioc. With respect to neotropical cotton (Gossypium
barbadense L.), Fryxell (n.d.) wrote: "There are many difficulties in
determining the identity of individual plants among the tetraploids from the
Greater and Lesser Antilles (and elsewhere), where spontaneous and artificial
hybridizations among different taxa have blurred the distinctions between them
and made a rational classification difficult at best." No "correct" taxonomic