Re: Mille (thousand)

Piotr said:

>Well, most etymologists do support a connection between Skt.
>sahásra-/Av. hazaNra- and Lat. mi:lle, though it's obvious that the
>relationship is indirect. The Indo-Iranian numeral is a true compound
>(*sm.-g^Heslo-), whereas the Latin one seems to be a univerbated phrase
>(*smih2 g^H(e)slih2 > *(s)mi:ksli > mi:lle; the geminate is regular
>after a long vowel). The main objection is the non-attestation of any
>reflexes of *smih2 in Italic, but as the form must have existed in
>pre-Italic IE, its loss as a free form is quite irrelevant. The
>prototype of <mi:lle> was probably already opaque to speakers of
>Proto-Italic. *g^Heslo- etc. may be related to *g^Héso:r/*g^Hesr- 'hand'
>("a large handful"?)

Sihler has a good discussion of this (New Comparative Grammar section 396).
He says there are two theories, with and without the (e) in Piotr's version.
Without the (e) he suggests we would get *smi:ksli as above, but then a
single l as in te:la from *teksla (note the short e, whereas (sm)i:ksli has
a long i, so you may be right, Piotr)
With the (e), he suggests we would get *mi: he:li, which gives mi:le, from
which we get the plural mi:lia. The double -ll- in the singular is a
problem for this theory.

A problem for both theories is the plural form, which is neuter. The
starting phrase for both forms involves a feminine noun.

Peter