I recently read Foley’s grammar of Yimas, and I wanted to share one particularly interesting aspect of it. I’ve got a pretty worn copy of Foley’s The Papuan Languages of New Guinea, which I’d strongly recommend for an overview of the region’s languages, and I’ve always found him to be very good at simple, readable, but nevertheless comprehensive explanations.

In any case, one interesting thing about Yimas is that it has two kinds of noun phrase structure, a tightly integrated but non-recursive structure, and a much looser, more appositional structure. Starting with the non-recursive structure, he documents the following properties:

• A single modifier (true adjective, possessor, …) immediately precedes the head noun
• Nothing can be placed between the two of them
• The modifier is bare and unmarked for normal agreement categories (especially noun class)

The following examples demonstrate:

Not all modifiers can occur in this construction; for example, Yimas has verbal adjectives as well as more nouny adjectives, and they cannot act as an integrated modifier in the same way mama “bad” can.

When a modifier cannot be placed in its bare form immediately before the noun, either because it is not compatible with the construction or because multiple modifiers are present, the alternative is to add a noun class agreement suffix on to it. In this form, almost all constraints are relaxed:

• The modifier can head up an NP on its own, without an overt noun to modify
• The modifier can occur either before or after the noun
• The modifier does not have to be adjacent to the “head” noun (i.e. it can be separated from it by other material that is not even part of the same “noun phrase”)
• There is no limit on the number of modifiers that can included. The structure is recursive or infinitely expandable.

Some examples:

The conclusion that Foley draws from this is that the integrated but non-recursive structure without agreement is a noun phrase, and the scrambleable versions with agreement are multiple, apposed noun phrases. While this analysis has a certain theoretical elegance, I can’t say if it’s correct or not… another obvious possibility is that the rigidly ordered but limited structures are compounds, in the same way that Yimas has e.g. noun-noun compounds, and the expandable structure is appropriately called a “noun phrase”, in the same way that Latin or many other languages with case and/or gender agreement and relatively free word order are analysed as having noun phrases.

All I can say is that I find an attested language with a distinction between compact, productive, morphologically minimal but structurally limited “noun phrases” and morphologically complex, agreeing, loose, structurally complicated “noun phrases” very interesting. Especially since the language is otherwise somewhat enthusiastic about gender / noun classes and concord.