I have just discovered that entirely coincidentally the word for “barbarian” in one of my constructed languages is homophonous with a Celtic word meaning “Irishman”. As you might remember, I don’t simply make the words in my constructed languages up, instead I attempt to model naturalistic sound-change through several millenia in order to gain a measure of verisimilitude. So it’s not through any unconscious antigoidelism on my part, but a pure case of happy homophony.
For a while now, the name given to a branch of the Kalpo-Lacaran family has been bugging me. That is the name of the Thærskan branch, which is roughly equivalent to Indo-European’s Germanic family. I came up with the name before actually devising the soundchanges which give the branch its shape, with the vague intention of deriving the name from the established Kalpo-Lacaran root √tʰeres-, meaning “tribe”, from something like an adjectival derivative *tʰeres-ka- “tribal, common to us”- which is basically a calque on the etymology of Deutsch and Dutch. Unfortunately, however, the reflex of *tʰereska- in Proto-Thærskan would be *terska-, which I don’t particularly like.
So, unwilling to tweak the soundchanges so that *þǣrskaz- would be an acceptable outcome, I thought about what proto-form could produce the form þǣrsk-. It turns out to be *dēreska-, which looks like an adjectival derivative of a stem √deres-, which in turn could be an extension of a more “basic” root *der-. On checking my lexicon of Kalpo-Lacaran, I see that there’s no existing roots of that shape. However, there is the already well-established root *del- “to speak”. Kalpo-Lacaran roots occur in two types: biconsonantal roots and triconsonantal roots, the latter often being derivatives or modifications of the former. For example, there’s √bʰar- “to shape, mould” and the extension √bʰaragʰ “to knead dough”. (There also exists a large number of triconsonantal roots with no such correspondence: √dur- “circular” and √durutʰ- “strong”, for example.) A plausible semantic extension of “to speak” would be “to speak intelligibly”: thus we have *þǣlskaz “understandable”, which is a good way of referring to your own language in contrast to a foreigner’s. So the Thærskan languages have become the Thælskan languages.
Satisfied with resolving that little niggle, I started playing around with this new root, tracing its development in the other branches of the family. Adding the pejorative prefix *wai- to a nominalisation of the root gives *waidelso- “someone who cannot be understood”, which is a good term for a barbarian. In Classical Tailancan, this becomes aidelsos by regular sound-change. In the most significant of Tailancan’s daughter languages, Carastan (whose soundchanges are broadly modelled after those of Breton), this becomes gouezel “savage, barbarian”.
And the actual Breton term for “Gael, Goidel” is also gouezel.