By Anatoly Liberman
The number of people in the English speaking world who distinguish in pronunciation between witch and which, wine and whine, wen and when is relatively small, and those who make this distinction do not say w-hitch, w-hine, and w-hen, but rather hwitch, hwine, and hwen. What follows is a breathtaking story of the hw-sound and the wh-spelling.
All the Old Germanic languages had words beginning with hl-, hn-, hr-, and hw-. At first, their initial h- was pronounced like ch in Scots loch or in German ach. Very early, thirteen or fourteen centuries ago, the fricatives, that is, f, s, th, and ch (the latter as in loch) began to weaken, and ch was hit the hardest: it turned into a sound we now hear in hope and help (kindly note the alliteration in the preceding part of the sentence). As time went on, h tended to disappear altogether. In English it no longer occurs between vowels. Word initially people drop their h’s “like a house on fire.” Before l, m, and n it was also lost, and, thank heavens, spelling retains no trace of it. As a result, modern speakers do not realize that listen, neck, and roof go back to hlystan, hnecca, and hrof (nor do they need to be informed about the previous stages of their language). By contrast, wrong, wreak, and their likes—at one time they began with “real” w—have preserved this venerable ruin in spelling. The same holds for knock and gnaw, in which k- and g- have been mute for centuries.
In the Germanic group of languages, only Icelandic still has a semblance of hl-, hn-, and hr- in pronunciation. Why fricatives began to weaken, why they were sometimes dropped and sometimes hung on, and why the position before l, n, and r turned out to be more vulnerable than before vowels (compare house and help with listen, neck, and roof) are important and interesting questions studied by historical phonetics, but discussing them here would take us too far afield. One of the mysteries of this process is that h- managed to stay before w long after it was shed before l, n, and r. In Icelandic, it was even reinforced: in one of the two main dialects of that language, hw- (by that time hv-) became kv-. A similar process took place in northern Middle English, as evidenced by the use of qu-, quh-, and cu- for hw- in multiple manuscripts. In southern English, including London, hw- and w- merged eventually, but in the north and in some central regions words like which and witch remained distinct. English words with the historical initial hw- are not too many; yet the highly frequent which, when, what, why, and where (who stands somewhat apart from this group) do not allow us to forget them. Why then do we write when rather than hwen?
Above I said that in southern English hw and w merged. This is true, but it should not be understood as meaning that one day h was lost before w and that where (to give a rand