About the tripartition of 3rd declension adjectives.

An interesting topic, old boy. Thanks for the tag.

A general statement about (Indo-European) adjectives is that they are not infrequently morphologically indistinguishable from nouns of the same class. The older & therefore 'purer' the language in question, this is increasingly true (that is, Hittite & Vedic Sanscrit are more likely to prove this than, say, Modern Lithuanian). And so our belief is that, at one very ancient time, they were 'one & the same thing', whatever that means.

In Latin (& Italic generally), the i-stem adjective predominates considerably with respect to other IE families because it has: (1) in effect absorbed the u-stems in their entirety (cf. Latin suāvis against G ἡδύς & S स्वादु svādu), a feature it has more or less in common with Lithuanian & Latvian; (2) extended the old so-called dēvī-type noun, which is itself a reduced i-stem + fem. suffix 2nd laryngeal (yielding G words like πότνια with short alpha, cf. L genetrīx against S patnī, janatrī); (3) reassigned most of the consonant-stems to itself; (4) even conditioned the reassignment of certain inherited o-stems (cf. similis against G ὁμαλός).

Regarding the tripartite split (incidentally, I was recently looking over a section of a book on Arabic grammar when I came across the pertinent term 'triptote', a noun having three cases), I can do no better than Sihler's comparative grammar, which I quote:
'The type with different masc. and fem. endings in the nom.sg., like acer m. 'sharp', but acris f., is a development from the normal type, as follows. A regular sound law (74.3) would give masc. AND fem. acer from *ākris via *ākers. Beside this there was a restored ācris, likewise epicene †[meaning grammatically applicable to both genders] in OL; it was reconstituted on the pattern of other i-stems. In the o-stems, however, syncopated forms like ruber 'red' < *-ros (74.4), are masc.-only (cf. rubra f., rubrum n.), and this provided the basis for the specialization of i-stems in -er as masculine-only.'

†My own note. I wasn't familiar with this term till recently, which maybe comes as a surprise.

And so, the answer, as always, is analogy. The tripartite endings operate analogically on the fact that 2nd decl. nouns ending in -er are always masculine (from an older -os, e.g. OL puerus > CL puer), such that any -er ending in i-stems must also be masc. Meanwhile, the 2nd decl. has 3 distinct endings, so the 3rd 'should also have them', a Roman might say.

I never knew that about vetus, but I just supposed it was an old s-stem adj., of the type which Sanscrit has in abundance. If I had to guess, I'd say dives is related to G δαίμων, from an extended IE root (*dH3-y-) for 'give, apportion', which we all know forward & back from L dare. I also don't think it impossible that noun uber rather comes from the adjective, much in the way the IE word for 'honey' is rather a substantivized neuter u-stem for 'sweet (thing)'.

Perhaps it even echoes in their terminology... I mean, what we call nouns and adjectives were all lumped together by ancient grammarians under the single category nomen.

Then at some point (I don't know when) someone felt the need to distinguish the two and labeled adjectives as a special kind of nomen—nomen adjectivum, that is a nomen that's typically added to another nomen. Still further down the road nomen was dropped from the phrase so that, in English etc., the word adjective is a noun.

The translations for "adjective" suggested by my English-Latin dictionary are:

- adjectivum nomen (cited as appearing in Priscian),
- adjectivum alone (Macrobius),
- appositum (Quintilian).

So the distinction may have been made as early as Quintilian (but one would have to see the context; whether appositum was really meant as a category of words or if it was just like "in this particular case this word happens to be added to that one").

So sequens may be another word for it, hm.

Well, yes, I'm aware that a lot of our terminology came from Quintilian. However, an adposition is just a subcategory of an attribute.

That's exactly right. It seems that even in many non-IE langs, the class of 'adjective' which is signaled by some morphological innovation has its origins in words that emphasize how someone/something is, as opposed to what he/it 'does'. In your example of 'victor', the semantic emphasis is not on the 'beating (that took place in the past)' but on the present status of being one who has beaten.

Your example also shows the special relationship between agentive nouns & (derivative) participles, which we commonly call 'verbal adjectives' but we could just as rightly call 'adjectival verbs'. In fact, before the Germanic langs had the suffix -er < L -ārius as in 'singer', they chiefly used the inherited present participle for agent nouns, like friend & fiend (< PG *frijōnd(i)z and fijand(i)z, respectively: [one who loves] & [one who hates]).