Article
Greg Thomson
June, when I'll have ready access to a cheap two-sided copier in the dorm
where I'll be living? That is an extremely helpful piece of literature,
particularly (of course) the peer commentary. The material in the Epstein
et al. portion isn't entirely new to me. If I may make a couple of comments
on your portion... The linguist's ability to talk about language may be
deceptively simple, relying tremendously on the existence of language
processes. Chomsky talked somewhere of how Martians would find Swahili and
English to be very similar. The problem is (with some nods to people like
Ania) Martians, if they don't themselves have a human language, will
probably be unable to make any sense out of Swahili or English. They'll be
behaviourists in the most extreme possible sense. In the balance, it could
be that linguistic ability does _most_ or at least _much_ of the work that
is done in metalinguistic thinking. We just aren't aware of it. Second, a
real speech production system (taking Levelt's model as still the industry
standard, as far as I know) really doesn't have much use for ther more
"general cognitive aspects" of metalinguistic knowledge. Message level
representations stream by, triggering chains of very rapid, modular (in the
Fodorean sense) events that end in a stream of crummy L2 speech. How I wish
my metalinguistic knowledge of Russian would do more for my output. Problem
is, it often takes a half hour before I realize what I said.
Another more minor point has to do with the elicited imitation task. I
personally think it is a wonderful task, and there is a body of literature
in support of it, which I can't readily cite (and Epstein et al. by all
means should have). As I understand it, there is a limit on the ability to
retain an utterance in an auditory loop (or in verbal form in short term
memory), with a parrot-like response, and I think that for English it is
something like 15 syllables. After that, listeners are forced to go ahead
and convert the form of what they heard into a non-linguistic conceptual
representation, i.e., the form in which text is most commonly and best
remembered, and flush the formal details down the toilet. They then must
use the memory of "meaning" in order to produce a replica of the sentence
by normal production mechanisms. Thus elicited imitation seems to be
tapping a wide range of comprehension and production processes as long as
the sentences are fairly long. Epstein et al. didn't invent this idea.
I really like your suggestions of using elicited imitation with
semantically bizarre and ungrammatical sentences. I would suggest that
these would need to be short and simple enough that native speakers would
achieve near 100% verbatim accuracy. I bet that NNSs would have enormous
difficulty with semantically bizarre sentences, and would often readily
repeat syntactic errors in short, easily repeatable sentences. Native
speakers are supposed to have a tendency to unconsciously correct
grammatical errors if they are performing under time pressure.
Anyway, I look forward to many enjoyable hours in that article and
commentary (especially the latter). Thanks again. And if it's OK, I'll
return it in early June. I can't imagine you have a huge supply of
offprints like that.
Greg
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
Morning by morning new mercies I see
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
Greg Thomson, Ph.D. Candidate (gtho...@gpu.srv.ualberta.ca)
SIL/Thomson, Westpost P.O. Box 109, FIN 53101,
Lappeenranta, FINLAND
Phone: 7-812-246-35-48 (in St. Petersburg, Russia)
Kevin R. Gregg
>Thanks for the article. If it's OK, I'll keep it until some time in early
>June,
****It's yours; enjoy. In haste, Kevin
Kevin R. Gregg
Momoyama Gakuin University
(St. Andrew's University)
1-1 Manabino, Izumi
Osaka 594-1198 Japan
tel.no. 0725-54-3131 (ext. 3622)
fax. 0725-54-3202
Kevin R. Gregg
>my metalinguistic knowledge of Russian would do more for my output. Problem
>is, it often takes a half hour before I realize what I said.
***You're lucky it's only a half hour. A couple of years ago, I was
reporting to the faculty on our decision to reject a recommended candidate
for a job, on the grounds that his only qualification seemed to be that he
was a devout Christian. The next day, I suddenly realized--I'd like to
know how and why--that in fact I had said he was a new car. (*sinsha*
instead of *sinja*; by rights it *should* be *sinsha*, but that's little
comfort.)
>Another more minor point has to do with the elicited imitation task. I
>personally think it is a wonderful task, and there is a body of literature
>in support of it, which I can't readily cite (and Epstein et al. by all
>means should have). As I understand it, there is a limit on the ability to
>retain an utterance in an auditory loop (or in verbal form in short term
>memory), with a parrot-like response, and I think that for English it is
>something like 15 syllables. After that, listeners are forced to go ahead
>and convert the form of what they heard into a non-linguistic conceptual
>representation, i.e., the form in which text is most commonly and best
>remembered, and flush the formal details down the toilet. They then must
>use the memory of "meaning" in order to produce a replica of the sentence
>by normal production mechanisms. Thus elicited imitation seems to be
>tapping a wide range of comprehension and production processes as long as
>the sentences are fairly long. Epstein et al. didn't invent this idea.
****Yeah, that's the philosophy of elicited im, and I have no objection to
it (and a good thing, too, given that I have no specialist knowledge of the
method). But deciding whether to use it certainly depends on the kind of
knowledge you're testing for, and using it correctly certainly requires
avoiding the kind of shortcircuiting possibility I mentioned, where the
mere collocation of a verb and its arguments suffices to produce
comprehension, without implicating the IL grammar. Chaudron et al, or
maybe Bley-Vroman et al--you know, I could get off my tush and walk the 5
feet to the bookshelf; hold on a mo--I'm back, did you miss me?
Bley-Vroman & Chaudron, no al, 'Elicited Imitation as a measure of
second-language competence', in Tarone, Gass, & Cohen, Research Methodology
in Second-Language Acquisition, Erlbaum 1994, (ISBN 0-8058-1424-8)--I can
copy it if you want--is relevant. Needless to say Flynn ignores it.
>I really like your suggestions of using elicited imitation with
>semantically bizarre and ungrammatical sentences. I would suggest that
>these would need to be short and simple enough that native speakers would
>achieve near 100% verbatim accuracy. I bet that NNSs would have enormous
>difficulty with semantically bizarre sentences, and would often readily
>repeat syntactic errors in short, easily repeatable sentences. Native
>speakers are supposed to have a tendency to unconsciously correct
>grammatical errors if they are performing under time pressure.
****I'd think you'd want long enough sentences so that verbatim accuracy
would be effortful, if even possible, in the absence of sufficient
grammatical knowledge. I mean, even my students could probably repeat 'My
pencil ate a teacher' accurately. Later, Kevin
Greg Thomson
reading discussions of SLA on SLART...
At 13:08 +0900 04-22-1999, Kevin R. Gregg wrote:
...
>****Yeah, that's the philosophy of elicited im, and I have no objection to
>it (and a good thing, too, given that I have no specialist knowledge of the
>method).
No doubt Bley-Vroman & Chaudron point this out, but the method has a long
history in psychology, and I imagine a search in PychLit will turn up more
hits than a search in LLBA or the MLA database. I first learned about it in
a small, rustic hotel in the foothills of the Himalayas when somebody
shared her research on the validity elicited imitation task for measuring
""proficiency"". Some of her research was later published her book
_Sentence Repetition Testing_, by Karla Radloff (Dallas: SIL). SIL (my one
time employer) uses this technique all around the world for testing levels
of bilingualism. They find it correlates well with more elaborate
techniques for estimating levels of bilingual proficiency. But for getting
at details of knowledge of ""structure""...
>But deciding whether to use it certainly depends on the kind of
>knowledge you're testing for, and using it correctly certainly requires
>avoiding the kind of shortcircuiting possibility I mentioned, where the
>mere collocation of a verb and its arguments suffices to produce
>comprehension, without implicating the IL grammar.
....
>****I'd think you'd want long enough sentences so that verbatim accuracy
>would be effortful, if even possible, in the absence of sufficient
>grammatical knowledge. I mean, even my students could probably repeat 'My
>pencil ate a teacher' accurately.
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. Epstein et al. said
they wanted to detect effects "due to the structure of the sentence". This
sort of thing always concerns me, of course. There example of present tense
is:
(1) (=61) The nervous professor inspects the broken television.
What is that supposed to mean? "Present tense" has various _functions_ in
English, and in L1 acquisition they aren't all aquired at the same time.
Thus it is possible that an L2 learner would have acquired the "present
tense" function in
(2) The nervous professor teaches several difficult courses.
before acquiring the function of "present tense" seen in (1).
Likewise, past tense with event verbs might be acquired earlier than past
tense with stative verbs, (their choice was _wanted_). So the idea of
studying the acquisition of forms without functions is suspicious, anyway.
Maybe this comes up in the peer reviews. This is not to say there won't be
a relationship between things that can't be reliably repeated, and things
that have not yet been acquired. The problem, as I see it, is arguing from
what _can_ be repeated to the conclusion that something specific _has_ been
acquired. This is because over time many aspects of language use will
become easier, increasing the possibility of repeating something even
though you haven't acquired it.
Greg
Epstein, D.S., Flynn, S., & Martohardjono, G. Second Language Acquisition:
Theoretical and Experimental Issues in Contemporary Research. _Behavioral
and Brain Sciences. 19:677-758, with peer commentaries).
mary gillespie
I'm wondering if any of you out there have done research or are interested
in depression and language acquisition. I know that the cognitive
suppression typical of some depressive episodes can mimic learning
disabilities. Obviously the affective factors influence learning, and
retention et al. Would be interested in hearing what you all might have to
say.
Thanks,
Mary Gillespie
Kwantlen University College
British Columbia, Canada
Adamson
>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
ada...@mxu.meshnet.or.jp
Greg Thomson
off-line messages to two SLARTers at the same time, and one I did right,
and the other I sent to the list by accident. It was late.)
At 20:07 +0800 04-22-1999, Greg Matheson wrote:
>I agree that formal details can be flushed down the toilet. I
>wonder if it is a function of the complexity of the sentence or
>of its length. Thinking about it, it is probably a combination of
>both which determines when the details start going down the
>toilet.
Now I went overboard in saying the formal details get flushed down the
toilet once the sentence is understood, and knew it, but that was when I
thought only Kevin would be reading it (and it was late). In fact, lots of
formal details can be and are remembered. But there appears to be less
memory for the formal details than for the meaning. So there is evidence
that memory for the verbatim form of an utterance decreases after a phrase
boundary, decreases even more after a clause boundary, and decreases a lot
after a sentence boundary. There is other evidence (don't ask for it all
right now) of some kind of special wrap-up activity at the end of
sentences, and probably at that point the meaning is stored in longer-term
storage, and the formal details are abandoned, although there is nothing to
prevent a lot of them from still being recalled. It's not that there is
immediate amnesia wrt the form of the utterance, but recall for exact form
after a long and complex utterance will be poorer than recall for the
"content" of the utterance.
>Grammar 'smoothing,' like recorrecting typos
>unconsciously, are the things which are likely to be done in the
>shorter sentences. Meaning changes are what is likely to be
>focused on in longer sentences.
This seems reasonable to me. I would say especially if grammar problems
involved long distance dependencies the situation will often be too
complicated to go back and try to straighten things out when the listener
already got the point. (This applies to native language speech. In a
nonnative language some of us suffer from an obsessive compulsion to
correct any error we notice or even suspect.) In any, I think "local"
grammar violations will stand out better than long distance ones, where,
for example speakers commonly "solve" a subjacency violation with a
resumptive pronoun and listeners don't bat an eye. But I'm not sure what
this has to do with elicited imitation (though no doubt it has something).
>I'm wondering if this means there are two different processes
>taking place here, a grammar one at shorter sentence lengths and
>a meaning one at longer sentence lengths
The difference that was of interest was between verbatim recall of short
strings, possibily in the presence of a less-than-total grammatical
analysis, versus the impossibility of such string repetition following
longer sentences, and the resultant need to reproduce the longer sentences
from scratch based on the better-retained memory of its meaning.
Gregards,
Greg
p.s. to Greg Mattheson. I personally wouldn't turn down a free psychoanalysis.
Adamson
>Now I went overboard in saying the formal details get flushed down the
>toilet once the sentence is understood, and knew it, but that was when I
>thought only Kevin would be reading it (and it was late). In fact, lots of
>formal details can be and are remembered. But there appears to be less
>memory for the formal details than for the meaning.
I think this needs to be softened even more. Greg's "formal details can be
and are remember" should possibly be changed to something like "some formal
details are remembered by some people".
The reason for saying this is that I know a weak bilingual (one language
much stronger than the other) who not only remembers no formal details, she
does not even remember the language that was used - either for things she
hears or things she says herself. And of course, there are some people with
edetic memories who remember every word.
Charles Adamson
ada...@mxu.meshnet.or.jp
Jonathan Centner
I would be careful about drawing exuberant analogies among brain systems,
but this sounds an awful lot like how the thing upstairs fills in the
details of visual representations where it has been determined there is a
superposed object and the underlying object is correctly, sometimes,
guessed at.
The formal information in this case would seem also to be abandoned once
the mind locks onto a conclusion. Ramachandran writes about this in
_Phantoms in the Brain_.
This might be the pulse point of formal acquisition, and could clear up
some confusion ( for me I hope) on how formal processes could be possibly
extrapolated from input, the projection thing, in which one might hope
eventually to get that penny to drop.
I am in the problematic position of agreeing more with your hyberbolic,
according to you, description, than with your emendation of it.
Pretty good stuff, I hope you and Kevin will walk down this path a bit, in view.
Jon Centner
>Now I went overboard in saying the formal details get flushed down the
>toilet once the sentence is understood, and knew it, but that was when I
>thought only Kevin would be reading it (and it was late). In fact, lots of
>formal details can be and are remembered. But there appears to be less
Jonathan Centner
Well yes, some bilinguals apparently don't even know they are speaking
different languages, and apparently they know when to switch between them
without ever having to think about it. (none of that _oh there's
grandmaman I have to start thinking in French_). A very difficult to
explain thing, I think, not quite in sync with Greg's latest, but maybe
yes.
Jon Centner
Adamson
>I would be careful about drawing exuberant analogies among brain systems,
>but this sounds an awful lot like how the thing upstairs fills in the
>details of visual representations where it has been determined there is a
>superposed object and the underlying object is correctly, sometimes,
>guessed at.
>
I have written about it before here, but Deane argues that we use the same
brain systems for understanding the structure of visual images and for the
structure of language.
Charles Adamson
ada...@mxu.meshnet.or.jp
Jean-Jacques d'Aquin
> Well yes, some bilinguals apparently don't even know they are speaking
> different languages, and apparently they know when to switch between them
> without ever having to think about it. (none of that _oh there's
> grandmaman I have to start thinking in French_). A very difficult to
> explain thing, I think, not quite in sync with Greg's latest, but maybe
> yes.
** I remember that as a child our immediate family would constantly be
changing among three languages, very often during the same communication,
or at least the same conversation. The reason for (code) switching almost
always was that a semantic term or an idiomatic expression in one of the
other languages "fit" our communicative intent better, and we continued on
in the new language until the same thing happened again. It wasn't a
deliberate decision, but a reflex action. Once the switch was made we
were semi-conscious that it had happened, but attached no conscious
importance to it. There usually was no conscious decision of which
language to start off with either. That was usually prompted by the
immersion environment of the moment, or what was on the radio or record
player, or whatever we were reading or overhearing at the moment, or even
thinking. My fascination with languages started somewhere in the pre-teens
when I realized that I actually thought within a "sonic" and/or "graphic"
paradigm (didn't know that word yet, though!) which influenced which
language my next communicative effort would be in.
When the extended family gathered, up to six languages would be
mix-mastered around in the same way, necessitating appropriate occasional
interpreting by a sensitive adult or remote cousin.
And yes, there was the equivalent of " _oh there's
> grandmaman I have to start thinking in French_ " in the sense that the
"extended" members didn't all speak all the same languages, and an
executive decision had to be made regarding the start-off language, a
decision that was totally unconscious among the immediate family members.
Sorry, if this anecdotal stuff is too boring ...
Cheers!
Jean-Jacques d'Aquin, Language Lab Director
University of South Alabama, HUMB-322, Mobile, AL 36688
VOX 334-460-6291 FAX 334-460-7123