Max,
    Thanks for your detailed reply. I am curious about a lot of it, but
in particular I am struck by the fact that fragments were counted as
part of the T-uinit to which they naturally belong. Did Kelly ever
discuss, or do you have any ideas about, the relationship between the
underlying cognitive (or psycholinguistic) functions and the production
of sentences? In my work, I'm counting fragments as separate units
because I think I am measuring the writer's ability to chunk linguistic
units in STM. (See my psycholinguistic model at
http://www2.pct.edu/courses/evavra/ENL111/Syntax/PLModel/Int000.htm.) My
hypothesis is that fragments result from an overloading of STM. The
overloading causes a crash -- the student puts down a period and writes
the rest of the sentence starting with a capital letter. This is, by the
way, related to the discussion of the definition of a clause -- in my
view a main clause is not a philosophical unit -- it is a
psycholinguistic one. The constraints of STM (now usually called
"working memory") limit the number of words (and hence ideas
(propositions)) that can be conveyed in any specific main clause. Do you
know what Kelly would have said about this? What do you think?
Thanks,
Ed