14 April 2012

Ooooh... it looks like I was wrong....

For years, I've been toeing the line that non-native languages never become like native languages, and that there's two language processing mechanisms in the brain: one for the infant learner, and one for the adult-learner.  This was science's best guess based on the data they had to hand: victims of "selective aphasia".  These people had suffered brain damage that had affected their abilities in languages, but not equally across languages.  The observed pattern was that all native languages would be affected more or less equally, and all non-native languages would be affected more or less equally, but the two would often be differently affected.
Well, via an article in the New York Times, I came across this paper at PLoS ONE that says otherwise.  Apparently they've been able to track the brainwaves of learners and find that there are similarities between the brainwaves of a proficient learner and patterns typical of the native.  This I find really cool.  (I wish I could afford a brainscanner so that I could measure my proficiency in terms of native-like brainwaves!!)

I'm concerned, though, that one thing they say will be overinterpreted:
Interestingly, both before and after the delay the implicitly trained group showed more native-like processing than the explicitly trained group, indicating that type of training also affects the attainment of native-like processing in the brain.
There are many people who advocate immersion from day one, and this would appear to be proof of the efficacy of the approach.  However, we are dealing with a very simple language here -- 13 words and a handful of extra inflectional morphemes, leading to a sum total of 1404 possible sentences in the language.  Even if the grammar's very different from English, there's still a very small number of "decision points" to be considered, so it is much easier to devine the meaning from the context -- this is something that just doesn't match the experience with real language.

They do, of course, admit as much in the body of the paper:
it may be that the results reported here are due to the limited size of the artificial language
Furthermore, the study says:
In the explicit training condition, participants were provided with 13.5 minutes of input of a type similar to that found in traditional grammar-focused classroom settings. Auditorily-presented metalinguistic explanations structured around word categories (e.g., nouns, verbs) were presented along with meaningful Brocanto2 phrases and sentences (which were also auditorily-presented, together with visually-presented corresponding game board configurations).
It's a shame the paper doesn't include the script.  My concern here is with the mention of "metalinguistic explanations".  Explicit instruction does not necessarily mean a lot of "metalinguistic" explanations -- yes, it will always require some, but often what starts out as an explanation overcomplicates things with unnecessary jargon and overly conscious processing.  Finely tuned explicit input can be very clear to the learner even if it looks messy on paper, and poorly tuned explicit input can be nigh-on impenetrable to the learner even if it looks neat on paper.

Which is why I object when they claim that:
A second possibility is that at end of training the explicit group's dependence on explicit, declarative memory-based knowledge resulted in the inhibition of the learning or use of procedural knowledge (see above), thus precluding anterior negativities. On this view, explicit training actually prevents, or at least slows, the development of native-like processing.
It is impossible for them to state that explicit training in particular prevents anything -- but that the training that they provided did.  (Of course, if I was to see their scripts, I'd definitely find fault with them, but that could always be me rationalising away anything that disagrees with me -- classic confirmation bias.  We're all guilty of it at times.)

Anyway, despite their statements about the difference between the explicit and implicit instruction, the  study does state towards the end:
the implication of declarative and procedural memory in this study is consistent with the predictions made by the declarative/procedural model for second language acquisition. This neurocognitive model posits that during L2 learning grammar initially depends largely on declarative memory, but that gradually aspects of grammar are increasingly learned and processed in procedural memory.
I am heartened to see that they haven't out-and-out supported the hardline distinction between "learning" and "acquisition" that Krashen and his ilk propose.

So it looks like I may have been wrong about the neurological differences between native and adult-learned languages, and I might still be proven wrong about the benefits of explicit vs implicit instruction, but this is not the study that proves that one.  If and when one comes along, I'll rethink my position.  Until then, I'll stick with what I currently believe.


gbarto said...

Here's a hypothesis for you:

1) Native-like proficiency requires that language be understood implicitly.

2) Implicit understanding means gathering meaning without conscious translation or parsing.

3) Developing implicit understanding requires prolonged exposure to comprehensible input.

4) Comprehensible input means that the input can be understood without outside guidance or helps to understanding the phrase.

Leaving to the side your thoughts about the use of cues in the classroom as an artificial sign language support, the question is: Where do you get the most comprehensible input over the long term? In the immersive environment, of course. This isn't because the immersive environment is a panacea. It's because believers in it will go to great lengths to develop strategies that provide comprehensible, if overdone and overwrought, input. By contrast, those who favor explicit teaching will fall too easily into thinking that a rule understood is a rule learned.

I would propose, however, that if the key is how much comprehensible input you get, then what is needed is a balance between explicit teaching to make input comprehensible faster and implicit teaching where the student gets comfortable with the language involved.

I would hope that with all the grad students out there, someone has investigated where explicit teaching makes input comprehensible faster and where it causes excess analysis on the part of the learner, but maybe it hasn't been done. Certainly your note about confirmation biases leaves open the possibilities that everybody is either so strongly in the explicit or the implicit camp that nobody wants to test out and refine a hybrid approach. Then it would turn out there's no panacea, just a requirement that the teacher be sensitive to when particular students are starting to get something presented implicitly and when their time is being wasted for the sake of pedagogical ideology.

When I was in grad school, in a CLT program, our pedagogy director always told us beginning instruction should be between 95% and 98% target language. She never explicitly said it shouldn't be 100% or that it wouldn't be dandy if it were, but she seemed to take it for granted that it wouldn't be, so better to restrict it than outlaw it. And so, when I taught, I'd offer a short explanation of tricky grammar points or a quick translation of a difficult word rather than, say, trying to pantomime the imperfect tense. My presentation was largely implicit, but with explicit clues about what needed to be watched for. It seemed to work well for certain types of students, especially the people in engineering and pre-med who really didn't want to play charades or 20-questions-in-a-language-I-don't-know-yet. I think this approach made things comprehensible to them faster.

So I think the answer is in here somewhere: Implicit learning may be better than explicit learning, but that doesn't mean that implicit teaching is the most effective way to effect implicit learning.

(Forgive any incorrect or outdated jargon; I've been out of grad school over a decade now.)

Nìall Beag said...

I can agree with points 1 and 2, but I just don't buy the idea that "comprehensible input" is the be-all and end-all of language "aquisition".

I've seen innumerable people who have spent decades in Scotland, and while they have learnt a lot from input, it's still notable that people have achieved better native-like structure in a few years with conscious instruction than many immigrants do through immersion.

Core to the problem is that comprehension does not require a complete model of grammar. Nature has built a great deal of redundancy into every language, so that we can understand each other even if we don't catch every word.

So the comprehensible input hypothesis fails because input does not provide any intrinsic motivation to grammatical accuracy.

To give a concrete example (AKA an anecdote!) I was teaching a couple last year, and I asked them to transcribe a short audio recording including the phrase "house prices". Every time I played it, they said afterwards "prices of houses". They had 100% comprehension of the phrase, but their internal grammar was altering their perception of the received input.

I fully believe that the given instruction is part of the learning process, and my perception* from my own learning is that the controlled production exercises in Michel Thomas's courses (the real ones, not the courses under the brand by other teachers) were more effective to me in building an accurate internal model of the languages than any other instruction I have received, and more effective than many hours of viewing and reading.

I don't accept that the goal of initial instruction is to get people to CI quicker, I believe instead that it is a vital part of developing an accurate internal model.

*Caveat emptor: no-one can give a truly unbiased account of their own experience.