Since the beginning of the Karyoł project the language has not had a sense of nominal number. I originally had an idea that verbal number, signaled by an affix or by suppletion, would indirectly signal nominal number, but I’m beginning to feel that this not in keeping with the design goals of the language.

I already have a well-grammaticalized and not really productive collective in –hoal, which is just delightful in that it may be so easily confused with –hoał ‘seat, basis, nest, throne’. I imagine –hoal could be to an extent combined with new vocabulary, but I feel a native speaker would reject it, l for the most part, as being awkward.

Lehmann (p 60) writes, ‘[V]erbs may acquire the category of number by the agglutination of a personal pronoun. This is also a possible origin of nominal number’. – As a matter of fact, toward the beginning of the Karyoł project, when I was reading a grammar of Great Andamanese, I had an idea that a word ɲIb would be used to coordinate both NP’s and VP’s along the lines of ɲIb=P₁ P₂=CONJ, with CONJ most likely being =wa ‘and’. I scrapped the idea, though, when I grammaticalized =wa as a conditional marked, and I turned ɲIb into a pronoun, specifically a third person plural pronoun.

Let’s say that ɲIb is in fact a third person plural pronoun, but that it specifically refers to animate arguments, or even just human arguments. I already have Wackernagel auxiliaries agreeing with A’s in [±human], and although I really enjoyed this being the only corner of the grammar where the distinction mattered, I’m comfortable with spreading the agreement around a bit. ɲIb could then be procliticized onto a human NP to signal number, or perhaps onto a VP if the S or O was human and not explicitly stated. Coordination between human NP’s would be handled with the pattern ɲIb=P₁=LINK P₂, where LINK is the adjectival morphology I talked about the day before yesterday. As the –(a)t linker would no longer apply simply to adjectives, I’ll need some other test for adjective-ness (or just retire the word class entirely). Perhaps an adjective is a para-nominal that cannot exist by itself in the predicate, but requires another NP to support it. (Along the lines of ‘He is good’ being illegal but ‘He is a good boy’ being well-formed.)

If –(a)t is available as a NP to NP marker, it’ll be neither head- nor dependent-marking; it’ll attach to the first word in a phrase, with order in part determined by pragmatics. This (mild) non-configurationality is actually what I had pictured to begin with – and I’ve since moved away from it. I’m interested in what it’ll do to my system of possession – which is broken down into alienable and inalienable categories. It’s also a classifying system, and each alienable noun requires the support of an inalienable noun – so ‘my spoon’ would be rendered omagkwebe loabał ‘my tongue spoon’ and ‘mother’s spoon’ would be ōmace omagkwe loabał. Now that the linker is available to nouns, I’m thinking I’m going to have it join a possessor to an inalienable possessum – the linker wouldn’t attach the omagkwe to loabał, for example, as these are two separate appositive NP’s.