504 ROBERT D. VAN VALIN, JR. & DAVID P. WILKINS
cates. The Dowty system as employed in RRG is clearly a minimalist
approach. It succeeds in capturing those aspects of the meaning of a predi
cate which are definitive of the larger semantic class to which it belongs,
and as such it can capture a lot of valuable generalizations about higher
level verb classes. This approach has been shown to be widely applicable
cross-linguistically, e.g. Tagalog, Lakhota (FVV), Sama (Philippines; Wal
ton 1986), Georgian (Holisky 1979), Italian (Centineo 1986), and Tepehua
(Totonacan; Waiters 1986,1988). Its major shortcoming is that it provides
only enough semantic content to allow general syntactic predictions for
whole verb classes but not enough for subtler syntactic predictions to be
made for individual predicates.
An extreme example of the rich semantic approach to decomposition is
found in Wierzbicka's (1972, 1980a, 1988) program of semantic explication.
The main feature of her approach is that she uses a minimal (primitive) voc
abulary to construct paraphrases in natural language which must be sub-
stitutable for the analysandum without change of meaning. The primitive
vocabulary precludes circularity in definitions; the use of natural language
permits it to be intuitively comprehended by laypersons and hence gives it a
wide range of applicability; and the requirement for substitutibility salva
veritae provides a test for verifying the accuracy of the definition. While
Wierzbicka (p.c.) currently claims that the minimal primitive vocabulary
only contains between 20 and 30 items of natural language, she tends to
work with a much larger vocabulary containing words which she has already
defined in terms of the minimal primitive vocabulary. The approach has
been shown to have cross-linguistic validity, e.g. Russian (Wierzbicka
1980b), Polish (Wierzbicka 1976), Japanese (Wierzbicka 1979), Ewe
(Ameka 1986), and Mparntwe Arrernte (Wilkins 1986a; 1990), and it
clearly allows for rich semantic insights. It does, however, have serious
shortcomings. Both McCawley (1983) and Nichols (1982), while praising
the insights which the Wierzbickan program gives, have cogent criticisms.
McCawley points out that she does not make fully clear what formal nature
she ascribes to her analyses: are they strings of words, trees or dependency
structures? This confusion arises from the fact that although Wierzbicka
provides a vocabulary for her natural language metalanguage, she does not
describe the combinatorial properties of her items, nor does she discuss the
syntax and discourse structure of her paraphrases. Nichols comments that
Wierzbicka is a "maximally content-oriented linguist" and that "[her] exclu
sive orientation to content leaves her study unequipped to recognize the