CHAT/CLAN discussion

[Bracha Nir-Sagiv]

Dear Yahya,
I am a research assistant at Ruth Berman's lab at Tel Aviv university,
where I have extensively used the CHILDES system in the framework of
several studies on lexicon and morpho-syntax in both early and later
language development as well as on discourse development. The system has
proved itself an invaluable resource for multiple analyses, as it is so
well adapted to the needs of linguists and language researchers.
Although the system requires you to "stick" to specific conventions
(which is an advantage in itself, as it enables you to compare across
databases), I have found it quite flexible, and was able to find ways to
implement whatever analysis needed. As to your statements, I can say that:
- Working first with MS Word and then converting to CHILDES is possible
(and rather easy), although you have to be careful when converting the
files, especially if you are going to use special fonts. But working
within CHAT mode is very easy and very efficient - not only can you
insert dependent tiers as you go along, you can insert main tier codes
(no need to copy-paste), work directly with sound files, and constantly
check your work and learn from your mistakes. Also, since CHAT files are
TXT based, I don't see a reason for having difficulties in exporting to
other programs - on the contrary, I have found it much less complicated
than exporting WORD documents, and there are even several CLAN utilities
that can help you perform whatever adjustments you might need.
- CLAN allows you to do much more than count codes or compute ratios,
for example, I have been able to use it for complicated automatic
analyses, automatically code extensive corpora at different levels,
create concordances, compute frequencies, analyze texts from specific
words to entire units, compare parent input with child output and so on.
If you prepare the files according to instructions (for example, use the
ID tier), you can run a statistics utility (STATFREQ) which enables you
to import data directly into Excel (and from there very easily into
SPSS). No need to type by hand.
- It is also possible to overcome the problem of nested structures, it
is all a function of how you transcribe the main speaker tier and of the
codes you choose to use (and this is also a very good point - you can
use whichever codes you want, all you have to do is define them). Even
if there are things that can't be visually represented within the file,
you can always use a special code.

I hope my answer has been of help, and can only sum up by saying that I
highly recommend using the CHILDES system.
Good luck with your work,
Bracha Nir-Sagiv
Department of Linguistics
Tel-Aviv University



Yahya E-rramdani wrote:

> Dear Colleagues,
> In preparation of the transcription and the analysis of data of a
> research on child language development (3-6 years), focussing on:
> lexicon, morpho-syntax and discourse, we have been discussing the
> (dis)-advantages of using CHAT/CLAN programme. I am listing below a
> number of statements on the disadvantages of using CHAT/CLAN in data
> analysis. I would like from to hear from you whether these statements
> are true or.
>
>    *
>       We should reconsider whether to make transcripts directly in
>       CHAT or first in MS Word (and to export the transcript as .rtf
>       to CHAT or to another processor). Arguments contra working
>       directly in CHAT: less flexibility, more difficulties with
>       exporting to other programs.
>     * There is discussion and careful consideration needed how to do
>       the actual coding. The classical CHAT-approach is adding (and
>       filling out) code-lines below each utterance. With the analysis
>       program CLAN a number of operations are possible, but as far as
>       we know it all boils down to counting codes per transcript or
>       computing a ratio. For further analysis, over sessions and
>       subjects, a form of exportation to SPSS (or EXCEL) is needed,
>       but CHAT/CLAN seems to lack direct options for that. Exportation
>       will mean entering data (that is, printed results from many
>       CLAN-runs) in SPSS by hand.
>     * Another disadvantage is that CHAT/CLAN seems to have
>       difficulties with dealing with different levels of aggregation,
>       that is, provides no explicit and easy ways of coding of the
>       'multilevel/nested' structure of the speech data. SPSS (or
>       EXCEL) is much more transparent in this respect and has
>       procedures for computing aggregate scores (and the reverse, for
>       de-aggregating scores).
>
>
> We welcome all other comments on the use of CHAT/CLAN in general.
> Thanks in advance.
>
> Yours,
>
> Yahya E-Rramdani
> Tilburg University
> Babylon: Centre for Studies of the Multicultural Society
> Tilburg University
> PB 90153
> 5000 LE Tilburg
> The Netherlands
> Tel: +31 (0)13-466.20.27
> FAx: +31-(0)13-466.31.10
> http://let.uvt.nl/GENERAL/PEOPLE/YE-RRAMD/
>
> +++++++++++++++++++++++++++++++++++++++++++
> This Mail Was Scanned By Mail-seCure System
> at the Tel-Aviv University CC.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/info-childes/attachments/20050617/bb1abd20/attachment.htm>