MLU and comparison criteria

[Ann Peters]

While I agree with the points made by Dan and Yonata, I'd like to add a
couple more thoughts. One way to look at MLU is to consider how many *bits
of meaning* are being strung together. This is the issue Dan was
addressing.
Another way is to consider how many *segmentable pieces* of language are
being strung together. My feeling is that we need to think about *both*
kinds of measures.
>From this point of view, to take Dan's examples:
	ragazz-i probably contains 2 'pieces'
	parl-o probably contains 2 as well - at least for adults.
Although I'd guess that the Hebrew restriction on at most 2
"morphemes" for any word probably works fine for Hebrew and Italian, it
wouldn't do for Turkish or Finnish. In other words, the degree of
'agglutinativity' of a language is also important.
I agree with Dan's and Yonata's points about productivity.
ann


****************************
Dr. Ann M. Peters, Professor
Department of Linguistics
University of Hawai`i               email: ann at hawaii.edu
1890 East West Road, Rm 569         phone: 808 956-3241
Honolulu, HI  96822                 fax:   808 956-9166
http://www2.hawaii.edu/~ann/