MLU for languages with complex morphology
Shanley Allen
shanley at bu.edu
Mon May 20 15:18:57 UTC 2013
Hi Katie,
Another relevant article is this:
Hickey, T. (1991). Mean length of utterance and the acquisition of Irish. Journal of Child Language 18: 553-569.
Tina calculates both MLUm and MLUw for Irish, and again shows that they are highly correlated. She also discussed problems in distinguishing individual morphemes in Irish, especially in child language.
As was pointed out earlier, some of the major problems with MLUm are:
1. It is hard to determine the productivity of a morpheme, especially at the early stages.
2. In languages with a lot of phonological processes at morpheme boundaries (e.g. Irish, many North American Indian languages), it is hard to determine where morpheme boundaries are or even what the morphemes are.
3. In languages with portmanteau morphemes or synthetic morphology (e.g. Inuktitut or English), a morpheme may well carry more than one grammatical function - it is not clear how these should be counted.
4. Some morphemes are only one phoneme long - the recording may not be good enough, or the transcription careful enough, to determine if the morpheme is really there or not.
Also, as Yonata hinted, neither MLUm nor MLUw can accurately be compared across languages because of the very different morphological and syntactic systems across languages.
Isabelle mentioned that MLUw would not be a good measure for Inuktitut because of its high morphological complexity. Indeed, utterances in Inuktitut can easily be only one word long but contain 5 or more morphemes (10+ for adults). So MLUw in Inuktitut would work as a measure of development but it would be quite a gross measure compared to MLUm. I have used mean length of word in morphemes in one publication:
Allen, S.E.M, Crago, M.B. & Pesco, D. (2006). The effect of majority language exposure on minority language skills: The case of Inuktitut. International Journal of Bilingual Education and Bilingualism 9(5), 578-596.
Again, though, the best method of measuring increasing linguistic complexity in children's utterances likely differs depending on the typology of the language.
Best,
Shanley.
On May 20, 2013, at 4:49 PM, Gordon, Peter wrote:
> It seems to me that one of the points that Roger Brown made about use of morphology and away from telegraphic speech was that this was indicated by a separation of the measures for MLUw and MLUm. In other words, when the morpheme count began to branch away from the word count, then this indicated the incorporation of morphology into the grammar, and a trend away from telegraphic speech. In the Parker & Brorson paper, they show that MLUw and MLUm are almost exactly correlated. The problem is that a correlation does not reveal the shapes of the two curves being correlated -- only whether one goes up when the other goes up. Generally words and morphemes will both increase in utterances over age, but this doesn't mean that the rates are the same. Presumably the morpheme count accelerates at some point that the word count does not, yet the correlation would not be affected by this unless there were a reverse trend, which is unlikely. So, the morpheme count can be more informative than the word count, but perhaps comparing the two is the most informative!
>
> Peter Gordon
>
>
> On Mon, May 20, 2013 at 10:28 AM, lofa <lofa4 at hotmail.com> wrote:
> hi Katie,
> I only know of an article written in 2005 regarding this matter.
> Kind regards,
> Véronique Devianne
> Speech therapist
> Doctorante Sciences du Langage, Université Sorbonne Nouvelle Paris 3
>
> From: Yonata Levy
> Sent: Monday, May 20, 2013 3:30 PM
> To: CHILDES
> Subject: Re: MLU for languages with complex morphology
>
> Since MLU is a heuristic measure, seems to me the solution to this problem cannot be universal. Rather, it is necessary to establish a method of counting which will be verified against theoretically-motivated developmental trajectories to achieve comparative standards. In our studies of Hebrew speaking children (Hebrew being a root-based, rich morphological language) we are using a method of counting MLU that takes into consideration the structure of Hebrew along with the need to guard against an inflated MLU. Our way of calculating MLU could perhaps inform other root-based languages such as Arabic but probably not languages with a different typology.
> Yonata.
>
>
> On Mon, May 20, 2013 at 2:47 PM, marilyn vihman <marilyn.vihman at york.ac.uk> wrote:
> I think one of the key problems with calculating MLU in morphemes - although it seems the right way to compare across languages, in principle - is the productivity problem: How do we know which morphemes the child really has any kind of mastery over? Is it sensible or appropriate to count all morphemes as if the child were using them productively, without any test to see if that is true or not? This may give an artificially high count to languages with synthetic morphology, lots of morphemes packed into a single affix...or is a morpheme only counted based on form? Even then, if the presence or absence of a single consonant means having one morpheme or two...less of an issue for English, say, than for many other Indo-European languages...then we probably need much more careful transcription of the phonetics of the child's speech than is usually thought necessary in morphosyntactic studies...
>
> -marilyn
>
>
> On 20 maj 2013, at 13.40, Isa Barriere wrote:
>
>> I meant to add:
>>
>> In contrast with a mophologically impoverished/poor language, the number of morphemes will tend to be positively correlated with the number of words, which is why it is not as crucial to calculate MLU in morphemes.
>>
>> Good morning,
>> Isn't there a negative correlation between morphological complexity of a given language and the average number of words that an utterance contains in the same language?
>>
>> If we take the examples in Inuktitut (given Shanley Allen's book based on her dissertation- Allen 1989 or 1990, in the appendix), looking at the utterances taking into account the # of words, many of them would only 1 word. However this does not capture the complexity of the meaning and the structure- that MLU does.
>>
>> So for different stages of devlpt and in the adult language the number of word per utterance may be limited and pretty stable and the progression is therefore best assessed taking into account number of morphemes (less stable across ages and stages).
>>
>> Isabelle Barriere, PhD
>>
>>
>> On Mon, May 20, 2013 at 7:37 AM, Isa Barriere <barriere.isa at gmail.com> wrote:
>> Good morning,
>> Isn't there a negative correlation between morphological complexity of a given language and the average number of words that an utterance contains in the same language?
>>
>> If we take the examples in Inuktitut (given Shanley Allen's book based on her dissertation- Allen 1989 or 1990, in the appendix), looking at the utterances taking into account the # of words, many of them would only 1 word. However this does not capture the complexity of the meaning and the structure- that MLU does.
>>
>> So for different stages of devlpt and in the adult language the number of word per utterance may be limited and pretty stable and the progression is therefore best assessed taking into account number of morphemes (less stable across ages and stages).
>>
>> Isabelle Barriere, PhD
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
>> To post to this group, send email to info-childes at googlegroups.com.
>> To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/CANNGd2b6ELzcqj3umutoAa1GBL1RyqzMBs4s75cxuBGgc12NBw%40mail.gmail.com?hl=en-US.
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>>
>
>
> --
> You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to mailto:info-childes%2Bunsubscribe at googlegroups.com.
>
> To post to this group, send email to info-childes at googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/E552879A-092E-41ED-BFCB-515BF8782633%40york.ac.uk?hl=en-US.
>
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>
>
>
> --
> Prof. Yonata Levy
> Psychology Department
> and Hadassah-Hebrew University Medical School
> Mount Scopus
> Jerusalem 91905, ISRAEL
>
> tel:972-2-5883408 (o)
> 972-547905997 (c)
> --
> You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
> To post to this group, send email to info-childes at googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>
> --
> You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
> To post to this group, send email to info-childes at googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>
>
>
> --
> Peter Gordon, Associate Professor
> 1155 Thorndike Hall
> Teachers College, Columbia University, Box 180
> 525 W120th St.
> New York, NY 10027
> Phone: 212 678-8162
> Fax: 212 678-8233
> E-mail: pgordon at tc.edu
> Web Page:http://www.tc.columbia.edu/faculty/index.htm?facid=pg328
>
>
> --
> You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
> To post to this group, send email to info-childes at googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/CAJE3P%2B8sAqBP8FqMYKpyi4_C1Yvi%3D_tnJANOAC1ECXQhE-a2mw%40mail.gmail.com?hl=en-US.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
****************************************************************************
UNTIL MAY 31, 2010:
Shanley Allen, PhD
Associate Professor and Chair
Department of Literacy and Language,
Counseling and Development
School of Education, Boston University
2 Silber Way, Boston, MA, 02215, USA
e-mail: shanley at bu.edu
phone: +1-617-358-0354
office: SED 331
web: http://efolio.bu.edu/portfolio/showPublicPortfolio.do?shareId=127
****************************************************************************
AS OF JUNE 1, 2010:
Shanley Allen, PhD
Associate Professor
Psycholinguistics and Language Development Group
Faculty of Social Sciences, University of Kaiserslautern
Erwin-Schrödinger-Strasse, Geb. 57
67663 Kaiserslautern, Germany
e-mail: allen at sowi.uni-kl.de
phone: +49-631-205-4136
office: Building 57, Room 409
web: http://www.sowi.uni-kl.de/english-linguistics/home/
****************************************************************************
--
You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.
To unsubscribe from this group and stop receiving emails from it, send an email to info-childes+unsubscribe at googlegroups.com.
To post to this group, send email to info-childes at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/info-childes/6229E208-0109-4E73-8BA5-A92754B8F734%40bu.edu?hl=en-US.
For more options, visit https://groups.google.com/groups/opt_out.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/info-childes/attachments/20130520/b99c88e0/attachment.htm>
More information about the Info-childes
mailing list