tree, drive: affrication

"start with the realization of /t/ in initial /tr/ (in _tree_ etc.) as
a retroflex affricate. this is anticipatory retroflexion of the /t/,
plus a retroflex fricative version of the aspiration on the /t/. this
realization of /t/ is very common in english, certainly in american
english; i have it myself, and it shows up in the spellings of
children who have devised their spellings on the basis of the names of
the letters in english (so that _tree_ is spelled CRE).

i believe that anticipatory retroflexion can also affect /pr/ and /
kr/, but there the retroflexion would be just a gesture accompanying
labial p and velar k; the gesture could be detected by phonetic
studies, but it would be hard to detect just by listening. for /tr/,
however, retroflexion is not just an accompanying gesture, but a shift
in point of articulation. then the aspiration on the t is itself
realized as retroflex -- that is, as a retroflex fricative, and we get
a retroflex affricate for the whole business (something very close to
the palato-alveolar affricate C).

for me, the retroflex affricate appears for /t/ in /tr/ only in
contexts where the /t/ would be aspirated. so, it's there in _tree_,
but not in _street_ .

but it's open for others to take the _tree_ realization of /t/ as
evidence for a retroflex affricate as the allophone of /t/ before /r/
in general, including in _street_. then, i think, the /s/ would be
necessarily retroflex as well. so the retroflexion spreads to /s/
through the /t/, rather than the reverse (which is what geoff seems to
be suggesting).

retroflexion could also spread to /s/ through /k/ and/or /p/. (it
might well be that spread through /k/ is more likely than spread
through /p/. in any case, there's room for lots of variation here.)

one more step: some speakers might have interpreted (some or all of)
these retroflex obstruents as palato-alveolars, on the basis of their
acoustic similarity. but i simply don't know what the facts are.

a side point: wilson writes: "the act of retroflexion appears to
require special
effort". well, you make the effort for the /r/; the only question is
whether you make the effort *only* for the /r/ or make it for some
larger stretch.

arnold"