Improbable regular cognates

In Zenaga (the Berber language of Mauritania), the word for "slave" is oʔḅḅäy.

In the "Shelha" Berber spoken near Touggourt, the word for "black" is aɣəggal. (In Tamasheq - Malian Tuareg - ɣǎggal means "to be brown".)

As you've probably guessed from the title, these are originally the same word. The semantic shift is sadly predictable, given Saharan history, but how can the consonants be related? Well:

Zenaga ʔ regularly corresponds to pan-Berber ɣ, eg iʔf "head" = iɣəf, iʔy "arm" = iɣil, iʔssi "bone" = iɣəs.

Proto-Berber *ww becomes bb in Zenaga and gg(ʷ) almost everywhere else in Berber, eg "year": Zenaga äššäbbaš = pan-Berber asəgg(ʷ)as.

Pan-Berber l becomes Zenaga y word-finally, eg ađ̣abbäy "male in-law" = pan-Berber aḍəgg(ʷ)al. But if you add the feminine ending -t, the resulting cluster lt becomes L. Sure enough, "slave (f.)" in Zenaga is toʔḅḅäL.

So if you're tired of repeating Armenian "erku" = English "two" every time you need an example of a non-trivial sound change, consider opting for a Berber example.

(All Zenaga data from Taine-Cheikh 2010; Tamasheq data from Heath 2006; Touggourt (specifically Tala n Aʕməṛ) data courtesy of a friend. The correspondences in question are discussed in more detail in Kossmann 1999.)