On 2006-01-31 14:16, tgpedersen wrote:
> But the suffix(es) is/are *-(t)er-o-. You might as well demand for
> there to be a /t/ in Lat. levir.
Analogy doesn't _have to_ apply in such cases, but it often does apply.
Cf. Skt. ja:ma:tar-, Av. za:ma:tar- 'son-in-law', with exactly this kind
of secondary /t/. Or Slavic *jeNtry, gen. *-trUve 'husband's brother's
wife' (originally one of the *-h2ter- family terms) on the analogy of
*svekry, -krUve 'husband's mother'. In the case of the cardinal
directions, a set like *aus-t(e)ra- : *wesp-(e)ra- :: *sun-þ(e)ra- :
*nur-þ(e)ra-, involving a recognisable suffix of contrast (as in
"other", "whether", "further", etc.) in three out of the four terms,
would have been crying out for analogical regularisation.
BTW, le:vir probably owes its /i/ to some sort of folk-etymological
association with <vir>, so it's also analogical in its own way.
Piotr