Using CHILDES: Replies

[Carol Slater]

Dear Colleagues:
Thanks to all of the list members who replied to my query about using
CHILDES in an undergraduate course in language acquisition. It looks as
though there are lots of ways to use it, varying widely in technical
complexity and intensity.

A number of people use CHILDES to create handouts for class  discussion
and analysis. Erika Hoff says, ” I have [only] used CHILDES 
 to get the
original Adam, Eve,  and Sarah transcripts. I hand them out to the class
and use them as examples of various phenomena and the students get to
see the real words of famous subjects. I have also used those
transcripts as the basis of take home exams with a question something
like ‘what does Eve know and what about language does she not know at
this point in her development?’  Students have thought that was
interesting and valuable.”

Lynn Santelmann makes more extensive use of CHILDES handouts. She says,
” I use the transcripts in two ways: First, I have created a packet of
transcripts for the students to analyze, one set for child-directed
speech, one for phonology, one for morphosyntax, and one for
discourse/conversation. (I also have a set for narratives, but they did
not come from CHILDES). Unfortunately, I haven't been able to get a good
data set for word learning yet. I "clean up" these transcripts a bit to
remove some of the analysis tiers, and I give them transcripts of
different ages so they can see change over time.
I give the students very specific questions or features to analyze, and
then they work either in small groups or at home. We discuss the results
in class. This gives them not only a chance to see some of the features
that we've talked or read about first hand, but gives them a chance to
see how hard it is to analyze things sometimes (e.g., is a morpheme
missing because the child doesn't produce it  or because the context
does not provide an opportunity for the child to use it?).”

People also involve students in analyzing CHILDES datasets with and
without reliance on CLAN programs. Catherine Snow says, ”I have used
CHILDES quite extensively in my course on child language 
to the extent
of teaching the class while logged on to CHILDES so we could pursue
particular issues 
(when does past tense first show up? what gets added
when MLU goes from 1 to 2? what are the first words that kids say and to
what extent are they the same across kids?) by downloading the relevant
files preparatory to ... doing analysis right there and displaying the
results. 
I also give analysis exercises as homework that students can
do pretty efficiently using CLAN, or less efficiently without it  (since
some don’t want to really learn to use the system), and I provide CHAT
formatted files as a basis for the longer analyses I assign for
take-home essays. Again, the students can analyse the files using a word
processor, or they can avail themselves of the CLAN options
I also
strongly encourage students to used archived data for their research
projects, because they can then do something much bigger and more
sophisticated.

Santelmann prefers that students  analyze their own data but  “a few
students can’t do this, or want to analyze a language other than
English, so I let them use CHILDES data. They’re able to do some nice
analyses because they have had some practice in class.” Michelle Barton
systematically helps students develop skill using CLAN. Her experience
has been that they like using the CHILDES system and  “[In] several
cases, having the skills has been a real plus for grad school
applications and research assistant positions.”


Some people also teach students about CHILDES transcription. Catherine
Snow says, “I 
 use the projection system to display transcripts linked
to videos
so that students get a sense of how one translates interaction
into analyzable text.” Margaret Friend has had students carry out their
own transcriptions. “My approach was to have students practice using the
transcription system and ... complete two transcripts: one standard
transcript which could be corrected for errors and on which they could
obtain assistance from other students and one transcript that they had
collected and recorded themselves. Students were assigned to groups of
four and each group recorded narrative data from children of different
ages. At the end of the semester they compared their transcriptions, did
a count of open and closed-class words and presented an in-class
developmental analysis based on the data 
. I was impressed with
students’ insights at the end of the course 
.”

In addition, there were pointers to relevant resources. Barton uses the
CLAN exercises included in Brian MacWhinney, The CHILDES Project, 3rd
ed. Volume I. (There is an introductory Tutorial at the beginning of
Part 2: The Programs, and a  set of Exercises at the end; Barton notes
that longer versions of the latter appeared in Sokolov and Snow (1994)
Handbook of Research on Language Development Using CHILDES.) Judith
Becker Bryant mentions that her chapter in Jean Berko Gleason (ed.), The
Development of Language, is followed by several suggestions for projects
using CHILDES.

Finally, people gave helpful counsel. Barton cautioned about glitches
in  CHILDES manual exercises: “[B]e sure to run them yourself and find
the errors, before assigning them to the students. It’s frustrating for
all if it doesn’t work right.” Santelmann says, “[I]t’s sometimes
difficult to find a set of transcripts that illustrate a particular
phenomenon, so students may need some direction in where to look.
(That’s particularly true for phonology kinds of things.) [T]he less
computer savvy of my students have had trouble down-loading
[transcripts].” Snow suggested that “[I]t is slightly more efficient to
have likely corpora pre-downloaded” if you plan to utilize them "on
line" in class discussion and notes that encouraging sophisticated
projects using archived data "only works, I find, if the class
activities have demonstrated how to get into CHILDES, what data are
available, and how to use CLAN." . Friend advises that the data
transcription projects she assigned  “turned out to be a considerable
amount of work for the students so you will need to think carefully
about grading.”

I am indebted to Info-childes list members for their generous support
and look forward to next steps made possible by their help..

Carol Slater
Department of Psychology &
Cognitive Science Group
Alma College
Alma MI 48801