Alec's response to Jim's email

 HI Jim,

Good questions!

On (1), one needs to think through the relationship between selection and allosemy.  In my 2013 paper, I wanted it to be the case that, in your example, ORGAN would need to be interpreted in the context of IZE before ATION was considered, although it could be ATION that sent ORGAN-IZE off for interpretation,  What you're correctly saying is that, in the "Innocuous piece," I haven't worked through the implications of my phase-al identifications made from the point of view of phonological selection and the phase by phase semantic interpretation that would need to parallel phase by phase phonological interpretation.  I will try to spend a few minutes thinking this through and get back to you.

On (2), yes, this is a good question -- are root-root Mergers unexceptional?  For the specific example you give, it's not clear to me whether the bracketing [ [ [ CURI OUS ] a ] ity ] and the bracketing [ [  CURI [ OUS a] ] ity ] are distinct (make distinct predictions) on the approach I was outlining.  Worth thinking about.  My theoretical intuition is that we want to say that OUS projects a, but of course one could imagine it doing so even if CURI has Merged to it.  The theory needs tightening, one way or the other.

On (3), yes, I'm imagining a close connection between derivational morphology and compounding -- perhaps to the extent that the syntax of derivationally derived words and compounds is the same.  This connection shows up in the literature in places like the analysis of Spanish -mente, whose phonology, I believe, is that of compounding (that is, X-mente words have compound phonology) and of English -ish, where -ish can attach to everything including phrases and can be used by itself.  Seems to me that the two major historical sources of derivational morphemes are (a) from participle morphemes in the extended projection of the verb and (b) from compounds.  Consider -wala in Indian English (as in tiffenwala) and -man (the -man in Walkman) in American English.  I guess the suggestion one might derive from "Innocuous" is that root-root compounds don't necessarily need to involve a bare root as either member of the compound -- the first, modifier, member would just need to Merge syntactically with the head of the compound the way that roots merge (i.e., not in the extended projection of the head or as a complement of the head).  The internal structure of the first member wouldn't matter.  (Works for me,)

--Alec