Univerbation -- merging of several morphemes into one, or at least into one word -- seems to be pretty common. How commonly does the opposite happen, and what sorts of environments does it happen in?

I can thing of various English examples, most of them fairly colloquial, like _(al)coholic_, _(ra)dar_, _(veg)etarian_, etc. Is this process more or less common in English than other languages? Do the resulting morphemes in some other languages end up *not* colloquial (well, I suppose given enough time, the English ones will cease to be colloquial too)?

ObConlang: I've had this idea for a while that I want to evolve from a system where:

- there are certain forms which can work as pronouns or as conjugated copulas, depending on context
- those forms don't really share any morphemes with each other; they probably derive from older pronouns

to one where:

- the original pronoun/copula forms have been reanalyzed as consisting of two morphemes
- the first morpheme of each new two-morpheme word still has basically the same sense as the original one-part word (viz., that of a pronominal/copular base)
- the second morpheme of each new two-morpheme word is analyzed as a person/number/gender agreement marker on the copula, and by analogy, on verbs in general
- the new pronominal/copular forms (the first morpheme of the new complex form) may or may not be able to stand on its own as a finite copula (I haven't decided); but it will be the base for further operations, e.g. attaching a subjunctive marker to


original conjugated pronoun form:
/?atam/ "I, I am"
/ka:n/ "You, you are"

leading to:
/?at/ pronominal/copular form: "I, I am"
/am/ 1st-person singular agreement marker

/k/ pron./cop. form: "You, you are"
/a:n/ 2nd-p. sg. agreement marker

So, what does everyone think? Some thoughts I have:
- Do we know how Athabaskan multipart verbs came about? From sets of two original verbs which became closely associated and then lost their independent meaning?
- Maybe if my system developed infixation, the infixes could come to be seen as non-infixal affixes, meaning the stuff on each side of them would cease to be seen as one morpheme.