geminus

Latin <geminus> 'twofold; twin' is usually explained by anlaut-contamination of <genus> 'offspring' or the like with inherited *jeminus, from the Indo-European root *jemh{x}- 'to pair up' reflected in Sanskrit <yamá-> 'twin', Old Irish <emon> 'id.', Latvian <jùmis> 'pair', etc. Apart from the singularity of such a contamination, this ad-hoc explanation is weakened by <trigeminus> 'threefold; triplet' and <quadrigeminus> 'fourfold; quadruplet'. If the original sense of <geminus> was indeed 'paired, twinned', it is very odd that words literally meaning 'thrice twinned, sixfold' and 'four times twinned, eightfold' would be used in the senses attested. It seems more likely that <geminus> earlier meant something like 'brought together, born(e) together', hence 'litter-mate', and no duality was implied. With lambs, *digeminus would be appropriate to distinguish double from triple or quadruple births, but with humans, multiple births are so uncommon that <geminus> alone would normally serve to designate a twin.

Greek <gaméo:> and the like are generally referred to a root *g^em(h{x})- 'to marry'. Skt. <ja:rá-> 'suitor, lover' implies a set.-root to yield a long syllabic nasal, *g^m:-ró-, but Grk. <gambrós> 'son-in-law' requires an anit.-root with excrescent -b-. I think the original sense of this root was 'to bring or bear together', with 'to marry' as a specialized technical development. Then Lat. <geminus> can be referred to a normal-grade *-nó-participle *g^emh{x}-nó- 'brought or born(e) together; litter-mate'; this is morphologically parallel to Grk. <semnós> 'revered' from *tjegW-nó-.

In the Nordwestblock we have at least eight examples of the place-name *Gamapia or *Gamapium, all found at the confluence of two rivers. I agree with Dittmaier's suggestion that the first element is identical to the root of <gaméo:>. But again I assume the basic sense 'to bring together', not 'to marry', so that *Gamapia is the Belgo-Latin equivalent of Gallo-Latin <Conda:te> (now mostly <Condé>), and no idea of marriage should be inferred. Indeed, most river-names in this area are feminine, such as the Haine and the Trouille, which unite at Jemappes (Gamappium 1065, Gamapium 1122, Gamapia 1150), part of the Belgian comm. Mons, prov. Hainaut. Given the overall conservatism of ancient society, it is unlikely that same-gender river-marriage would have been approved.

I regard *Gamapia as parallel in formation to <Messa:pia>, the latter adapted from Messapic or Epirote Illyrian *Mez^z^a:pja(:), a fem. adj. 'amidst waters'. This compound was probably formed within Illyrian as *mez^z^a-apja- with crasis *-a-a- > *-a:-, though of course one could posit a purely formal PIE cpd. adj. *medHjo-h2ap-jo-. In my view Belgic was closely related to Messapic and a similar cpd. adj. was formed from *gama- 'union' (formally identical to Grk. <gámos> 'marriage', but lacking the specialized sense) and *apa:- 'water', namely *gama:pja- 'pertaining to the union of waters, confluent'. I posit a Late Belgic shortening of long vowels in this position, yielding *gamapja-; this is parallel to the short-vowel syncope which I have postulated to get Latin <Venta> from Late Belgic *Wentja (perh. [wenc^a]), earlier *Wenetja:, name of three presumed Venetic settlements in England taken over by Belgae (and in turn by Celts). The short middle vowel in *Gamapia is necessary to explain Dutch <Gempe>. Post-Belgic syncope in the Dutch/Low German area (but not in High German or Walloon/Picard/Norman) is also found with simple apa-Namen. Thus we have LG <Elspa>, <Elspe> against HG <Elsaff>, <Elsoff> from *Alisa(:)pa 'Alder-Water'.

It should be clear that I do not regard Belgic as a form of Celtic. However, I think that Celtic reflexes of *g^em(h{x})- may be continued by some Ibero-Romance words. Viewed from the outside in, a branch is a confluence, and we might tentatively ascribe Asturian <gamacha> 'great branch', West Ast. <gamaya> 'id.' (Asturo-Latin *gama:cula) to a Celtic base *gamo- from *g^m.h{x}ó-. Antlers branch, so the semantic shift 'branch' > 'antler' (> 'horn') is straightforward, and we might also assign Santandrian <gama> 'antler' here, with the more widely distributed <gamo> 'antlered animal, fallow deer' as a back-formation. Anlaut-contamination would then explain why Spanish <gamuza> 'chamois' has initial /g/, when most dialects north of the Pyrenees reflect /k/ (first attested as Gallo-Latin <camox>, Pol. Silv.). And Sp. <gamón> 'Asphodelus ramosus', a plant noteworthy for branching roots, could be placed in this group as well.

DGK