Word count

[Brian MacWhinney]

Dear Anne, Ann, Carson, and Info-CHILDES,

   Carson's message shows clearly how much linguistic theory gets
involved immediately in the morpheme count issue. And thanks to Ann
Peters for explaining the additional issues about syntactic combination
and productivity.
   When I referred to "initially" and "forever", I meant that a child
may begin with a monomorphemic analysis initially and remain with that
analysis throughout subsequent language learning.  I was not referring
to the idea that linguistic theory would ever stick with something
"forever."
    The idea that a word has "one bit of meaning" is a rather quaint one
that never occurred to me.  It is interesting to think that back in
1967 Ursula Bellugi may have thought of "don't" as a single meaning
primitive. Personally, I think of morphological analysis as involving
the division of perhaps (just for the sake of illustration) 80
dimensions of meaning in a word into two subsets, one of which may
contain perhaps just a dozen "features."
   Apart from all the various theoretical perspectives one can adopt
here, I have always liked the notion proposed by Roger Brown that
semantic analysis typically precedes formal morphological analysis.
The idea was that children talk about "two shoe" or "many shoe" even
before producing "shoes".  I don't imagine that the data for this are
always so clear, but the idea that children develop a concept of
plurality and perhaps also some phonological awareness that there are
sounds like /s/ on the ends of words, well before hooking up these two
ideas has always seemed reasonable to me.  In fact, I wonder how it
could happen any other way.
   I realize that one can easily postulate stem-modifying rules for the
analysis of forms like "don't" and "wont".  My only point here was to
make sure that Anne realized how much this issue takes us right to the
heart of linguistic theory and child language theory.  I think Carson's
message makes this quite clear.

--Brian MacWhinney