Radius Solis wrote:Anybody is free to invent words in their source language whenever the author of that language didn't provide something you want to have. Publishment of such words encouraged but not required.
Wait, seriously? I certainly see where you're coming from, and indeed, in the
talideon wiki's take on this experiment, I did a fair amount of this. But then, that was quite necessary, 'cause our starting lexicon there was the items of the Babel text and nothing else; here we already have plenty of material.
If you're inventing a word out of whole cloth, it certainly makes sense to do so by projecting it back to the source lang, to get something with the general phonological feel of an evolved word, and to pick up whatever morphological alternations your sound changes bring about, and whatever. But positing this word to have been part of the source language all along? That feels, well, somewhat violating. Certainly I've already made a number of conscious decisions about the AhH lexis: semantic space will split this way, so that these semantic distinctions will be made but not those; these other concepts I don't want lexicalized at all -- it's not always just a matter of non-provision. And I wouldn't especially like having that ridden over roughshod.
And (on a theme I've touched on before) don't forget that there are plenty of ways aside from complete invention to get a word for X that don't necessitate your source language having had a word for X. Semantic shifts especially; but also borrowings, and fossilization of compounds or derived forms, &c. These have the further benefit (beyond parsimony) of being much more interesting than plain invention, from the etymological perspective.
----
[...] then /m/ must have been unconditionally lost in at least some languages. The question is, lost in what way? I don't know, but I have a hunch that the number of such instances must be too large to explain entirely with cataclysmic losses of labials, which are quite rare.
True. But my own hunch is that in these cases it'd more likely be something like /m/ > /b/ or /m/ > /w/, or indeed /m/ > /0/.