<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
Indeed this is a crucial point that requires fine analytical tools to be
answered.
<br>And to my mind, depending on developmental level, the answer might
be only probabilistic rather than definite
<br>(rules of the thumb like those mentioned by Yonata help to go on but
do not solve the basic issues).
<br>One of the things that I think should be avoided is to consider productions
in isolation, something that unfortunately may occur with online ways of
coding.
<p>One bit of evidence that might be used more widely consists in looking
for morphologically relevant variation in <b>one and the same</b> <b>type</b>
of word, although also this is not bullet-proof because each one of the
items could be learned as a separate entry. However, considerations of
the overall trend in the child's production might provide a further crucial
hint.
<br>A more encompassing system approach to the analysis of children's productions
might be a revealing useful tool in this thorny domain.
<p>Edy
<br>
<br>
<p>Annette Karmiloff-Smith wrote:
<blockquote TYPE=CITE>The problem that remains for me mentioned by Ann
Peters "at least for adults".
<br>Surely MLU calculations are an attempt to get at what the linguistic
<br>representations
<br>are "for the child". So, it could be that raggazzi is a plural
for
<br>the child of raggazzo
<br>and that "for the child" there is a representational relationship
<br>between the two,
<br>but it could be that the child has learnt "raggazzi" as an unanalyzed
<br>whole linked to a
<br>specific context. In other words, the child could be matching
the
<br>phonology rather than
<br>the morphology. I have no solution except there may be clauses
in
<br>intonation or the
<br>rest of each child's particular productions, etc., but the problem
is
<br>real in my view,
<br>and may be more so for highly agglutinating languages.
<br>Annette</blockquote>
</html>