On 9 Jan 2016 16:57, "Jeffrey Brown" <[log in to unmask]> wrote: > > > > > > > > > > On Thu, 7 Jan 2016 18:02:22 +0000 > > > And Rosta <[log in to unmask]> wrote: > > > > > > > Genders are classes (of noun) that it is necessary for the grammar to > > recognize. Grammatical behaviour that is fully predictable from semantics > > is not necessary for the grammar to recognize. The rules of grammar need > > specify only what is not fully predictable. > > This does not seem correct to me. For example, in Arabic, verbal agreement > in some circumstances depends on the animacy of the subject (human vs > non-human). The animacy of the subject is semantic (doctors are human; > trees are not). Yet, I still have to know the agreement rule to form a > grammatical sentence. That sure sounds like "grammar" to me (and my > professor thought so, too, since he marked me down for my error in the > exam). You need look no further than me, for evidence, to be assured of the fallibility of professors. I don't know Arabic (of course), but I'll suppose the facts are as you describe. In that case, you have to have a grammatical distinction between two versions of the verb, Versions A & B, and semantic rules stating that Version A verbs' subjects refer to animates and Version B verbs' subjects refer to inanimates, but, crucially, nouns don't have to be marked in the grammar as animate or inanimate. This is just like the situation for Sylvia's conlang, which I had provisionally concluded has no noun classes (for all that the grammar is sensitive to animacy, motility, sessility). --And.