36 R. M. W. Dixonadjectives, but then developed dependent markers. Syntactic function is, as in Jap-
anese, shown by particles following an NP, and some of these are thought to have
developed recently. (For example, Sohn 1999: 30 mentions that the subject particle
ka—which is now a conditioned allomorph of the earlier subject marker i—first
appeared in the literature in 1572 CE and may possibly have been a borrowing from
the Japanese subject particle ga.)
I feel that extending the Japanese scenario to also apply to Korean maybe tran-
scending speculation in the direction of fantasy. Like other results in linguistic typ-
ology, (33) is a statistical correlation, not a hard-and-fast rule. There are exceptions
to it. Besides Korean, these include Southern Paiute (Uto-Aztecan; Sapir 1930-1),
which is also dependent-marking and shows 'verb-like' adjectives. It is, of course,
worthwhile according detailed examination to these and other exceptions, to see
whether there is an explanation (along historical or other lines). But it is unlikely
that every exception will be provided with an explanation, and it is an error to try
forcibly to provide one. Some languages just do have a typologically unusual com-
bination of properties in some area of the grammar.10. Semantic overlapping between word classes
I mentioned in §1.1 that while each word class has a similar semantic core between
languages, there are a number of non-core concepts whose word class membership
varies between languages; for example 'hunger/hungry/be hungry' can be a noun,
an adjective, or a verb.
As discussed at several places above, some languages have small adjective class-
es whose members are typically taken from the core adjectival types, DIMENSION,
AGE, VALUE, and COLOUR; languages that have an adjective class with more than
about ten members tend to include in it some PHYSICAL PROPERTY terms. How-
ever, most languages have a large, open adjective class; it is interesting to compare
the semantic membership of these classes.
In §6.1.2, I outlined the membership of the open adjective class in Fijian—it
includes all DIMENSION, AGE, COLOUR, PHYSICAL PROPERTY, and SPEED terms. But
in Fijian HUMAN PROPENSITY terms are verbs. This is explainable in terms of the
grammatical organization of Fijian; adjectives may function just as intransitive
predicates, while verbs may be either intransitive or transitive. Coding HUMAN
PROPENSITY terms as verbs allows them to take an object argument, correspond-
ing to the optional prepositional phrase in English, as in happy (about), clever (at),
jealous (about).
During the remainder of this section, I want to focus on the adjective class in
the Australian language Dyirbal, and compare it with English. The adjective class
in Dyirbal is large and open, and it is like the noun class in its grammatical proper-
ties (very similar to European languages). The semantic contents of the adjective
classes in Dyirbal and English are similar; that is, most adjectives in Dyirbal cor-
respond to adjectives in English, and vice versa.