Sound changes originate in the varying speech habits of speakers of a given language. I would like to propose that there are two kinds of sound change:
(1) Those that are typologically triggered or motivated, and
(2) those that do not carry the additional typological motivation.
The shared features of (1) in a given family are: (a) Such sound changes are actuated and implemented in different languages and subgroups independently, producing an identical output and eventually giving the impression that they have resulted from a shared innovation; (b) they have different time profiles for different members of the same family, (c) they are more regular than (2) and could even be exceptionless. Supporting data are given from the Dravidian family, e.g. the merger of highly marked segments with unmarked ones, leading to the ultimate elimination of the former, e.g. alveolar *_t [= t <sub-bar>] with dental t/d or retroflex .t /.d [= t/d with <subdot>]; *.z [= z with <subdot>] (retroflex frictionless continuant) with a host of different segments.d, .r, .l,.n [=
d, l, r, n <subdot>], y, w, zero. Proto-Oceanic had words ending in vowels and consonants, but some of the derived subfamilies/languages have become totally vowel-ending without exception. The deaspiration of voiced aspirated stops of Proto-Indo-European in most of the descendant subgroups is also accountable by this phenomenon. Data from other language families will also be presented. What is proposed is that certain sound changes, which are supported by system-internal pressures, tend to be highly regular compared to those, which lack such typological backing.