It has been claimed that the conlang Kēlen, winner of the 2009 Smiley award, is supposed to challenge Greenberg's universal of always having a noun-verb distinction by eliminating verbs. In light of the fact that Kēlen has a special word-class called relationals, which are four predicating words that take noun phrase arguments, is it not more factual to analyse the relationals of Kēlen as a very limited and closed class of verbs? They do seem to fulfill the same function words called "verbs" do in other languages, and are necessary for predication — another domain typically reserved for verbs (even if we assume zero-copula).
Using the Natural Semantic Metalanguage as a good baseline of what a human language can communicate, there are several core verbs:
- Mental predicates: THINK, KNOW, WANT, FEEL, SEE, HEAR
- Speech: SAY
- Actions, events, movement: DO, HAPPEN, MOVE
- Location, existence, specification: BE (SOMEWHERE), THERE IS, BE (SOMEONE/SOMETHING)
LA is clearly polysemous for THERE IS, BE (SOMEWHERE), BE (SOMEONE/SOMETHING), and probably also marks PART in reverse.
NI probably covers DO and MOVE.
SE is even higher level, expressing the verb of give, as well as primes like FEEL and SAY.
So the claim that Kēlen does not have verbs is patently false.
There are in fact natural languages with even more restricted verb inventories than Kēlen, such as Jingulu, which has been analysed as only have verbs for go, come, and do. (I haven't seen an NSM style analysis of Jingulu, I wonder how they would see the verbs like THINK etc being expressed in Jingulu?)