One of the nice things about writing a grammar is that you have to make decisions about things you don’t know how to analyse, which means working out how to analyse them! One of these things for Bardi has been the Noun Phrase structure. I knew that adjectives and other modifiers could either precede or follow the noun, but I didn’t know what conditioned it — whether it was simply related to the focus marking system, or prosodic, or something else. Since Australian languages tend to have fairly free constituent order, I think I had led myself to belief that it was going to be related to information structure.
However, when I started looking into it in more detail, several things become apparent immediately. One was that the Noun Adjective order was pretty rare compared to the Adj Noun one. Another was that although Noun Adj order was deemed ungrammatical in elicitation, it did occur in texts. Lastly, Noun Adj order tended to occur in noun phrases which were not initial in the clause; that is, clause initial NPs had Adj Noun, but ones elsewhere in the clause (particularly finally) had both orders.
It turns out that there is a distinction in semantics, as far as I can tell, though I need to confirm this for certain with Bardi speakers. The difference seems to be very similar to Spanish adjective placement. In Spanish, the default order for most adjectives is Noun Adj, and in such an order, the adjective is classificatory (broadly, it describes a Noun with property X, or the restricts the reference of the Noun to Nouns with that particular property). The order Adj Noun is unrestricted. That is, it refers to a Noun (already established in the discourse), which happens to have the property described by the adj. The difference is parallel to restricted and unrestricted relative clauses.
In Bardi, it’s just the same, except that the order is reversed. That is, Adj Noun orders are restricted (and the default), while Noun Adj are unrestricted. For example, if we say boordiji aamba ‘big man’, it refers in theory to all men who are big (that is, boordiji ‘big’ classifies or delimits the reference of aamba ‘man’). On the other hand, aamba boordiji ‘man big’ refers to a man, who happens to be big (that is, it describes a property but does not function to restrict the reference of the noun).
The semantics of the Noun Adj placement also explains why such structures do not occur in initial noun phrases. Initial NPs are focused, and contain new information. This is a position where one is very unlikely to find an unrestricted relative clause, since they tend to occur when one is providing more information about established referents.
This work was motivated in part by Erich Round’s Kayardild dissertation, where he looks in some detail at Noun Phrase structure and the conditions under which apparent Noun Modifier constructions occur. He found that there is not a true position of post-nominal modification in Kayardild, but rather that such structures are two noun phrases, the second headless.
I have not seen this restricted/non-restricted difference reported for other Australian languages, and I wonder how common it is. It’s also made me wonder how much structure we would uncover in other languages which are otherwise non-configurational, but that will be a topic for another post.