A somewhat left-field theory on the discrepancy between learner performance in the written and spoken modes

In the post 4 skills safe, I argued that writing could be carried out using declarative memory, but that procedural memory was required during speech, but there may still be more to it than that, and I have a theory.  Feel free to tell me I'm crazy - it doesn't deny what I said about declarative vs procedural memory in the first place.

I have been told when you are reading, your eye scans each word on average three times.  This is because the written sentence is missing many important cues we would have in the spoken form, and it needs information from the context to reconstruct the full meaning.  And this is in your own language, so what must it be like in a foreign language you're not fluent in yet?  Your eyes dart backwards and forwards across the page as you try to decode the meaning, and in the end, without realising it, you develop the habit of reading in the wrong order.  You could be faced with a French sentence like:

Je le lui ai dit

and your brain might decide to jiggle the order round until it's reading the same as English:

*Je ai dit lui le

Why would the brain do that?  Because it already knows English, so it's easier that way. The thing is, you won't necessarily be consciously aware you're doing it, and the only way to ever find out that you are might be to head to your local uni's language science department for an eye-tracking study.

Well actually, maybe not, because if you're reading in the wrong order, you're probably going to... (drum roll please)... try to speak in the wrong order, because you end up creating a procedural knowledge of grammar based on your reading style.  And guess what?  Yup, lots of learners do indeed try to speak in the wrong order.

So what appears superficially to be a good "reading skill" is actually flawed reading, bad reading, disordered reading.  We celebrate a student's success in reading as motivation when they're not doing well in speaking, but in isolating and rewarding reading as a single "skill", we may actually be encouraging and reinforcing the very behaviour that is limiting their spoken fluency.

That can't be right, though, because they're still writing in the correct order!

Well yes, but the brain is subtle, and writing is a very slow activity compared to speaking or signing, so it has a hell of a lot more time and freedom.  The thing is that whatever language you're writing in, native or foreign, your brain is likely to be several words ahead of your hand.  This is where it gets twisted.  In theory, the brain has enough time to recall the words in the wrong order and then shuffle them about spacially to write them down.  As a skill, this would be good enough and fast enough for writing, but would not transfer into speaking; it may even prejudice against proper speaking.  By isolating and rewarding writing as a single "skill", we may again be encouraging and reinforcing a problematic behaviour.  I may be wrong, but without testing it, is this a risk we want to take?

And this time we can't even use eye-tracking software to detect the problem, because everything goes on inside the brain.

Except that there is one very subtle clue that comes along a little down the road: some people's grammar is great in a short sentences, but even simple grammar is beyond them when the sentence grows in length and complexity.  Traditional thinking puts this down as simply being "a difficult sentence", but really, it's just a combination of language points* that we have already taught and tested to our satisfaction.  If the students know the rules, why do they fail to combine them.

What if what we're really seeing is the writer running out of working memory or time?  If learners do indeed recall the structure out-of-order and reconstruct it on the fly, then it stands to reason that they will quite quickly fill up their working memory once they have to hold something in it while constructing a complex phrase, or even an embedded clause.

I think a good example is the difference in how German and English handle defining clauses (and you'll have to forgive me if this isn't quite right as I've not learned German properly yet.  Corrections gratefully received.)

I would like to buy the book you like.
Ich möchte das Buch kaufen, das Sie mögen.

Here we have a slight crossover as "the book" and "to buy" switch places.  But (as I understand it) it's actually that book in German, and the that is repeated after "to buy".  This means that "the book you like" is split up, and if you're trying to hold the whole structure in working memory, you'll be taxing your memory.

And it gets worse as you add in more information, as German lets you insert things in a multitude of ways that I'm personally not comfortable with yet.  And if its "the book you told me about yesterday", it gets even messier...
And thus the "out-of-order recall" strategy that was initally the simplest strategy for the brain to follow becomes unworkable and a barrier to further learning.

Consequences for teaching

Now first of all, I'll stress that it's just a theory and so any change of teaching practices should balance "what if he's right" with "what if he's wrong".  Furthermore, I'm not claiming that this is an inevitable consequence of certain teaching methods, but that certain teaching methods open the possibily that a student develops these flawed strategies.

What I want, therefore, is for teachers to work to reduce the possibility for students to develop suboptimal or counter-productive strategies.  I suggest this can be done by adopting two simple principles:

1. Students should be made to produce spoken language of equal or greater complexity to their written language from the beginning.  This way the student is forced to adopt a strategy suited to spoken language.
2. Language should be integrated with previously-taught language points early and often.

This second principle I cannot stress enough.  I was once made to teach children from a book in which each unit consisted of two 1-hour lessons.  The first lesson taught a verb structure in the positive declarative (=statement) form, and the second lesson introduced the negative declarative and the positive interrogative (=question) forms.  But the structures taught included such things as "I used to ", and the negative and interrogative forms taught were fully regular ("I didn't use to..." & "Did you use to...?") so could have been dealt with from the word go.  Neither lesson integrated with the present tense or the past simple to produce sentences such as "I used to play football, but I don't anymore" or "I used to play tennis but I stopped a year ago".  These are sentences that any learner at that level should be able to produce, yet we often delay them, and students are left without the confidence or competence required to use these straightforward conversational devices.

Footnote: Why did I come up with this crackpot theory?

When I started doing written grammar drills in Spanish, I found myself frequently missing the object pronoun then writing it in afterwards (object pronouns appear before the verb in Spanish, like in French).  I got better at doing this, until I was thinking a few words ahead of my pen by a word or two.  So I was still thinking of the verb before I had thought of the pronoun, and in the end I made a conscious effort to stop doing this and I refuse to put pen to paper for as long as my brain tried to put the verb first.

That's a sample size of one, so doesn't really prove anything.  But it does give a plausible mechanism for observed data, and one of my big problems with much of the writing on language that I've read is that in general, mechanisms are rather vague and hand-wavy.  Empirical data is all well and good, but all too often what is recorded is merely the tasks given to the students and the end result -- the process followed by the student is rarely tracked.

If anyone knows of an eyetracking experiment that has explored this, I'd be interested to know.  And if anyone fancies studying it as a masters thesis, let me know how you get on!


* "Language point" is a catch-all term for vocabulary items, fixed phrases, grammatical rules, etc.

Speed reading's great - unless you want to learn a language.

Every now and then, I read someone on the internet quoting the remarkable reading speed of trained speed readers, and then asking whether this would help a language learner.

In principle, it sounds like a good idea.  After all, if you can read four times faster, you can read four times as much, right?  It seems like you can therefore expose yourself to more foreign language in a shorter time.

Unfortunately, the mechanics of speed reading -- and indeed reading in general -- do not allow this.

When we read, we take advantage of known, familiar patterns in the input.  This means that we don't actually need to process everything that is printed in order to understand it.  Tak# th{s sent#nce, f*r ex*m*pl*.  As a reader, your knowledge of English is probably enough to fill in the gaps.  We do this all the time.  If we didn't, reading speeds wouldn't improve so dramatically just through frequent reading, but they do.  Basically, our brains get in the habit of knowing the bits they need to read.  "The book is ## the table": the eye knows those two ## shapes are "on" as there's nothing else they could be.  "The book is ##### the table": well, that's just got to be under.  So the brain recognises the rough shape and/or size and matches it with the known patterns -- it knows what it must be without actually reading it.

Speed-reading training isn't anything special.  Although some speed-reading schools will tell you to do this, that or the other, in the end, all speed-reading classes do is encourage you to take this natural process to its logical conclusion.

So where does this leave the language learner?
Nowhere.

When we're learning a new language, we do not have deep knowledge of the language we are studying, so our brain cannot take the shortcuts required for speed-reading.  Any assumptions it makes will come from outside the target language, and therefore will result in errors.

So, nice thought, but no.