Are there impossible changes? θ > f but f ≯ θ

One question that historical phonology should reasonably seek to answer is: are there impossible changes? That is: are there plausible changes that we could reasonably expect to occur in the diachrony of languages’ phonologies, but which nonetheless do not ever occur? In this paper I seek to spell out what it really means to consider this question and what we need to do in order to answer it for any specific case. This will require a consideration of some fundamental issues in historical phonology, including the distinction between exceptionless and lexically-specific/sporadic changes (which I call ‘N-changes’ and ‘A-changes’), and the connection between that distinction and the ‘misperception’ model of phonological change. It will involve an analysis of aspects of the phonological history of Pulo Annian, Arabic, Italic, Spanish and several varieties of English. I argue that the current state of evidence indicates that there are indeed impossible changes (which I symbolise using ‘x ≯ y’ to represent that ‘x cannot change into y’) in a very specific but phonologically real way, and that f ≯ θ is one