MLU counts

[Yonata Levy]

Dear Brian,
What you say is exactly what we have been trying to accomplish - there are
details that pertain to the language which guide one in the decision of
which morphemes should be counted - gender, root, case - however, the
ultimate goal of MLU is comparability both within a language and
cross-linguistically. To achieve that, or come close to it, one has to set
some ad hoc rules like the ones I suggested - never count a single lexical
item as longer than 2 or a single utterance as longer than 9. I am not sure
about the 9 though - it may be too high.
Why do you think that there might be a way to count MLU that will work for
diverse languages with rich morphologies? My understanding of comparative
grammar is that there is no such dividing line between syntax and
morphology. therefore a system should be constructed that will enable
comparative studies regardless of the typology of the language. So, here we
are again - we need to propose language specific+language general ways of
reducing the count to the levels which have been set mostly by developmental
studies of English . Alternatively, we may try to construct a generalized
scale that will be based on a sample of languages for the sake of cross
linguistic research.
Yonata

________________________

Prof. Yonata Levy
Psychology Department
The Hebrew University
Jerusalem, Israel 91905

Phone: 972-2-5883408 (o)      Fax:    972-2-5881159
            972-2-6424957 (h)      e-mail: msyonata at mscc.huji.ac.il


----- Original Message -----
From: "Brian MacWhinney" <macw at cmu.edu>
To: "Yonata Levy" <msyonata at mscc.huji.ac.il>; "Hyams, Nina"
<hyams at humnet.ucla.edu>; <info-childes at mail.talkbank.org>
Sent: Wednesday, February 18, 2004 5:51 PM
Subject: Re: MLU counts


> Dear Nina, Yonata, and Info-CHILDES,
>   The discussion of how to handle MLU in morphologically rich languages
has
> been on the table since the 1960s without any real resolution.  Yonata's
> guidelines seem to reflect current state of the art and seem just about
> right.  However, it seems to me that people working with morphologically
> rich languages should really compute two indices.  The first would not be
> cross-linguistically meaningful, but would be maximally meaningful within
> the language.  That index would count morphemes in terms of what they
> express.  Here, you still may wish to be a bit conservative.  For example,
> do you really want to count the German article "die" as four morphemes
> (definite, case, number, gender)?  I would say not.  Maybe two morphemes
> would be about right (definiteness and case-number-gender).
>   The second MLU for morphologically rich languages should be constructed
on
> the basis of a real comparative program of research.  Comparing normal
> children of similar ages in similar urban (or rural) environments and
> similar (mutually culturally relevant) activities, can you come up with a
> method of scoring that yields parallel counts across morphologically rich
> language (Hebrew, Inuktitut, Hungarian) and an analytic one (English,
> Chinese).  As far as I can tell no one has yet attempted this obviously
> important but perfectly feasible study.
>
> --Brian MacWhinney
>