My pet “phon-prob” concerns velar nasals: what makes them so special (and rare)? Why is it they are far more constrained than n and m, at least in most languages I know of? E.g., why doesn’t English have words like *ngail, *ngight, *ngine, *ngoon, etc. There must be a simple answer to this obvious question, but I can't understand why the combination of "velarity" and "nasality" is somehow exceptional, whereas velar stops are unproblematic.
I don’t know whether anyone has an answer that would satisfy Robert. I haven’t got one myself. The usual explanations in terms of markedness theory seem to me to be post hoc and circular (because X is rare, we say it’s marked; then it is the markedness of X that’s responsible for its rarity).
There are of course plenty of languages in the world that do have initial velar nasals. One must remember that the languages familiar to Europeans are by no means representative of the world's languages. Even in Europe, an initial velar nasal is quite common in Welsh, though only through mutation of an initial velar plosive: e.g. fy nghi ‘my dog’ və ŋhiː, from ci ‘dog’ kiː; and fy ngardd ‘my garden’ və ŋarð, from gardd ‘garden’ ɡarð. In colloquial spoken Welsh the fy və bit can be dropped, leaving the velar nasal phrase-initial.
See the map and discussion in the World Atlas of Language Structures. Part of the map is reproduced in miniature here:You can see that initial velar nasals are found particularly in central Africa and in the southeast Asia – Pacific area.
Of those languages in the database that have a velar nasal as an independent phoneme (234), there are more that allow it in initial position (146) than that disallow it (88), which you may find surprising.