This paper focuses on nominal classification in Bantu, Romance and Chinese.
The relation between numeral classifiers, noun classes and grammatical gender has often been noted in the typological literature (e.g. Craig 1986, Senft 2000), based, on the one hand, on the (perceived) comparable semantic parameters involved in the classification underlying all three types of system (shape, size, animacy, and others), and, on the other hand, on the frequent relation between classification and agreement, both within the nominal phrase and with constituents outside it.
However, despite these apparent similarities, the establishment of a common core of these three systems (if there is one) (represented here by Cantonese classifiers, Italian gender and Swahili noun classes) remains problematic, as well as the question why we have classification at all.
In this paper, we give a fine-grained description of the three systems, showing that Bantu and Romance are very similar, not only for the common pervasive agreement facts, but also because both families allow for derivational uses of noun classes and gender seem to involve similar semantic domains. These two characteristics set Chinese apart.
We propose that a common rationale underlying these classification systems can still be discerned, and is suggested by the fact that the expression of classification and number often, though not always, go hand in hand: in our view, class and number are both part of the individuating process that turns a descriptive “predicate” NP into a referential expression.
