-eris, -oris, -uris?

Question

Much to students' annoyance, nouns ending in -us can belong to either the second (servus), third (tempus), or fourth (circus) declensions. I understand the origin of the second and fourth: Proto-Indo-European stems in -o- vs -u-. But the third-declension -us nouns are less predictable:

• tempus → temporis (also corpus, decus, nemus)
• latus → lateris (also genus, funus, scelus)
• jūs → jūris (also crūs, pūs, rūs)

The last of these makes sense to me: the root was originally jūs, and the intervocalic s became r in pre-classical Latin (cf flōs, flōris).

But for the other two patterns, why sometimes e and sometimes o? The two patterns seem very distinct: even nouns of uncertain declension, like penus, -ī/-ūs/-oris, never interchange -eris and -oris forms. What led to this distinction?

Answer 1

The usual explanation given in historical grammars, e.g. those of Weiss, Sihler, and Buck, is that the -er- stems result from regular sound change, while the -or- stems result from analogical remodeling on the basis of the nominative/accusative.

A well-known Latin sound change turned all short vowels in word-medial open syllables to i. Since short unstressed ir is a disfavored sequence in Latin, though, before r such vowels became e instead. (Compare second-person passive forms like cap-e-ris, vs. cap-i-tur.) This is how you get the medial vowel of the lateris type.

Apparently, though, in some cases the vowel of the nominative/accusative -- which would still have been o, e.g. tempos, corpos -- was restored throughout the declension, giving the temporis type. Why this happened in some nouns and not others is unexplained: it's often the case that analogical changes hit some words but miss others for unclear reasons.

This, as I said, is the standard explanation. It is interesting, however, that in three of the four -oris examples you list, the consonant before the o is a labial. This could be coincidence, of course, but maybe part of the explanation is phonetic, since the presence of a labial could plausibly favor the rounded vowel o. It would be interesting to look at the full set of such neuter s-stems to see if there are any phonetic patterns of this kind.