Ed’s comment to my last post allows me to widen the discussion from the specific question of th > f in modern English, and its history, to that of how much linguists/dialectologists/orthoepists – call them/us what you will – have or have not registered change in the language.
I’ve been reading Alexander J Ellis’s monumental “Early English Pronunciation”, jointly published by the Early English Text Society, the Chaucer Society and the Philological Society between 1869 and 1889, in five parts, and well over 2000 pages. Despite its title, Part V is an analysis of the English dialect area of Great Britain (i.e. of that part of the island that has been English-speaking for many centuries) and the Isle of Man (which Ellis admits is less interesting as English is mainly a “learned” language for the Celtic-speaking natives) during the 19th century. Many of the dialects include th > f as a feature, at least in some words.
The nearest modern equivalent to Ellis is John Wells’ “Accents of English”, and I was surprised when I looked in the bibliography to Wells’ work that Ellis is not mentioned. While Wells gives a set of key words to identify particular vowels, Ellis goes back to the Old English basis as his starting point, and shows the way in which each OE sound has developed in the various dialects. This accounts for the way in which some sounds have developed differently in the modern forms. As an example, some dialects of the north west and north west midlands of England have, as a reflex of OE /ū/, the phoneme /ai/, so that house is pronounced /(h)ais/, and mouse /mais/. So what happens to OE /ī/, which gives standard modern English /aɪ/? It becomes /ɛi/, so that mouse and mice are still distinguished. Wells does not mention this, but then he is talking about accents, not dialects. This raises the question, which I shall not attempt to address here: where is the boundary between an accent and a dialect?
Ellis was not only concerned with vowels, as is shown by the th > f example. He also recognized /tl/ for /kl/ in many dialects, and the rhotacization of /t/ in some contexts and in some dialects. John Maidment wrote about this here and the comments to his post show that /t/ to /r/ is quite widespread. Ellis expressed no surprise at any of these features, and was at pains to point out several times in the work that it was a linguist’s job to describe what he heard, not to pass judgment on the correctness or otherwise of his informants’ speech. Perhaps he has been neglected because, writing before the development of the IPA, he used a phonetic script called palaeotype, which is not always easy to interpret nowadays.