Remnants of the dual number

Question

To my understanding, Latin doesn't have a dual number at all. The adjectives duo "two" and ambō "both" have some special forms derived from the PIE dual (-ō, -ābus, -ōbus), but are considered irregular rather than part of any real pattern.

However, from an answer to an unrelated question:

By the way, another heterogeneous noun with a similar story behind it is frēnum "rein", pl. frēnī. In this case the Latin masculine plural ending continues what was in PIE terms not a plural at all, but a neuter dual.

This points to another trace of the dual number which I hadn't heard about before.

Are there any other irregularities or fossilized forms in Latin which derive from the dual number in PIE or Proto-Italic?

Answer 1

The -ī of vīgintī "20" is originally a dual ending, the same one as in frēnī (PIE *-ih₁). This is why the ending of vīgintī is different from that of the other tens (trīgintā etc.)

Answer 2

According to W.M. Lindsay The Latin Language: An Historical Account of Latin Sounds, Stems, and Flexions (p. 253), octo could also be a dual in form, with the sense of 'two sets of four'. He sees a comparison with ambo and duo because of the ending ō-sound.

I don't get why it would mean 'two sets of four' though.

Answer 3

It looks like a couple of neuter nouns of the fourth declension might have nominative singular forms derived from PIE duals. (Note that the nominative neuter ending -u might have been pronounced either as -ŭ or -ū; it seems that we don't have any clear evidence either way for the quality of the vowel in this context.) A form -ū with a long vowel could be explained as deriving from PIE dual *-uh₁; the dual form is mentioned as a possible source of a Latin form in De Vaan's entries for genu "knee" ("PIE *ǵen-u(-h₁ dual), *ǵen-u- (obl.)") and cornu "horn" ("If ū is original, cornu may reflect the nom.acc.dual ending *-uh₁; especially for cornū *'pair of horns' > 'horn'").