🚀 go-pugleaf

Greg Thomson wrote:
>But if you added a picture selection task, even though they repeated, "My
>pencil ate a teacher", you might be surprised to find that they chose the
>picture which portrayed a teacher eating a pencil.

Back in the early 1980s, someone here in Japan was advocating a dictation
test which utilized progressively longer sentences. In a presentation at a
JALT Conference, they said that their research had shown that at some point
in the sequence a student will be no longer able to remember the entire
sentence long enough to write it down accurately. They also said that there
were few if any errors before the cutoff and few if any correct answers
after it.

At the time I was in need of a simple test to separate large groups of EFL
students in roughly accurate general levels, The previous test had been
grammar based and was totally inappropriate as a placement test for our
program.

I worked with a variety of tests but soon found that Rasch analysis of one
our standard in house tests allowed me to get sufficient information for our
purposes. Therefore, I did not finish the work on the graduated dictation
test (GDT). I did find, however, something similar to what Greg discusses.
Without taking meaning into account, the GDT results did not seem to relate
to the students' ability level. We asked the students to translate the
sentence into Japanese after they wrote it and found that the correctness of
the translations did not seem to relate to the correctness of the
transcription. In other words there seemed to be two, or more, effects
taking place. One set of results even indicated that there might be a
difference between presenting the sentences in increasing order of length
and randomly. One other interesting point appeared and that was that if a
student made a mistake in transcribing a sentence before the cutoff or got
one correct after the cutoff, it was very likely that many other students
would also have the same results. I did not investigate far enough to make
any guess as to why this happened. Two possibilities which I started to
consider at the time were: 1. this was something that all the students had
been taught in school, 2. this represented some general characteristic of
the students' interlanguage.

Also I have found that I can increase the chunk length that students can
remember through a simple exercise. I put students in pairs. One reads a few
words from a text and the other then repeats without looking at the text.
The students are instructed to adjust the chunk size to what their partner
can remember and they are to gradually make them longer by just enough to
challenge but still be mostly doable. About 2 hours of such practice spread
out over a semester allows most students to reach a maximum which seems to
be fairly fixed in that more practice does not seem to change it much. This
limit appears to be somewhere in the vicinity of the 18 syllables that
someone mentioned recently.

Charles Adamson
adamson@mxu.meshnet.or.jp