Sound change has no memory Sound change does not discriminate between the sources of a sound. If a sound change causes X>Y, it cannot affect only original X's. If it helps, think of a stampede of animals, each erasing its predecessor's footprints.
Sound change ignores grammar A sound change can only have phonological restraints, like X>Z in unstressed syllables. It cannot drop final W, except on adjectives, or the like.
Sound change is exceptionless If a sound can happen at a place, it will. It affects all sounds that meet the criteria for change. Exceptions are possible, due to either analogy and other regularization processes, or another sound change.
Sound change is unstoppable Nobody knows why, but all languages invariably vary from place to place and time to time. Writing does not keep languages from changing. This would be true if we learnt languages from reading books. We don't. We learn our native tongue by imitating the speakers in our environment. Only dead languages, artificially resurrected and kept alive languages like Latin, and international languages like Esperanto are immune to sound change. (The only thing that has stopped Esperanto from having dialects is the fact that it is almost entirely second-language speakers. If they speak it to their children, and they do the same and so on, sound change WILL happen.)The formal notation of sound change:
For example,
Unless a change operates unconditionally (in all positions), we have to specify the context in which it applies:
If the symbol "#" stands for a word boundary (initial or final), the notation "/__#" = "word-finally", and "/#__" = "word-initially". For example:Rules of Sound Change