07 June 2013

The filter of perception

I've been a bit preoccupied with exam season and have been putting off many things, including blogging.  But I'm going to try to get back into it, and I'm going to try to get this blog back on the language track, and I'll mostly be posting my thoughts on MOOCs elsewhere.

So here for your perusal is some recent thoughts on the filter of perception, and how it affects notions of learning by absorption.

During a recent break, while I was holed up at my parents' house due to a foot injury (a pratfall in a doorway), I started studying German with one of the free interactive web courses.  I'm relatively happy with the course itself, although there are (as with any course) some rather stupid things in it.

There are various types of questions in the course, including translation, multiple-choice and taking dictations.

So there I was, and I heard the computer say to me:
"Liest du Büchen?"
which I started typing.  But then I stopped, because I'd made a mistake with this before.  The form I should have been typing was "Bücher".  I knew this.  But I'm telling you, I heard Büchen.  Yes, the computer said Bücher, but I heard Büchen, clear as day.

The brain is a remarkable thing, and what we perceive is not always what hits our eyes or ears.

There are rules to the universe, and once our brain knows the rules, it filters what we receive to produce a perception that matches our expectation of the universe.  If we see a man standing half-hidden behind a wall, we don't perceive "half a man", we perceive "a man that we can see half of".  We make a rough guess at the hidden bits based on proportions relative to what we can see and the many hundreds of humans we've seen in our lives.  If you see a man with a hand over one (presumed) eye, you assume there's an eye underneath, and you would only be surprised when he moved his hand if there wasn't an eye there.

In language, this is particularly useful as it lets us understand dialectal variation without too much effort, and crucially without ever having to truly "learn" the dialect we're trying to listen to. 
If an Irishman said to me "ten times tree is tirty", I might well perceive "ten times three is thirty", and if the conversation was quick enough, I might not even be consciously aware that he had said T sounds instead of TH.  And if I said to him that "ten times three is thirty", he wouldn't have any problems understanding me, just because I used the "extra" TH sound that isn't in his inventory.

But while that's good for the fluent speaker, it's a potential pitfall for the language learner.  In the case of my German lesson, I had an unconscious rule in my head that said "-en is the German plural suffix" and that filtered the received "Bücher" and gave me the perceived "Büchen".  Now before anyone blames "rules" for my error, let me make it clear that this was an internalised, procedural rule rather than a conscious, declarative one.

Had I never been punished for perceiving it wrong, my ear would probably never have been learned to perceive the difference, because there would have been no impetus to do so.  (Say, for example, I was only asked to translate from German to English.)

And so it is for anyone living through a foreign language.  I recall one interesting experience when doing a listening lesson with two private students (I thought I'd mentioned this here before, but I can't find it in my posting archive).  There was a gist-listening exercise with comprehension questions, and then there was a series of close-listening tasks consisting of a sentence or two of audio and a fill-in-the-gaps version of the sentence on the worksheet.  As they whittled away the gaps word by word, they were left with two gaps, but that wasn't enough, because every time they listening to the recording, they heard three words: "prices of houses".  I replayed the file several times, watching them in fascination: "prices of houses", "prices of houses", "prices of houses".  How was it that even when they were listening very, very closely, they couldn't perceive the simple phrase "house prices"?

As far as I can see, it comes down to this:
that structure wasn't part of their language model, and continued exposure to the language only trained their ability to filter the input to adapt it to their structure, rather than adapting their structure to match the input.

If we comprehend input by mangling it to match our internal model, then accurate acquisition by comprehensible input alone must be an impossible dream.

14 comments:

acutia said...

I had a very similar experience this week practicing dictation from a short Euronews video in German.

After working through the sentences, I checked my dictation with the transcript on the site. I found that I'd 'heard' in certain sentences words I'd recently studied or checked the meaning of but which weren't in the video or text. This fits with what you've described. But the other aspect in my case was that for several words I had got wrong I intropectively remembered my actual first barely conscious guess was more correct, but I thought better of it when I listened for the second or third time and wrote down another word.

I see this as a clash between my sound perception (which is generally very good) and my internal German language model (lexis, word forms, sound to orthography) is only at a beginner level.

My takeaway for next time will be to try to listen for and trust my pre-conscious guesses.

Titch said...

I think that's a very different issue. If your first reaction was correct, you perceived it correctly. If your changed your mind, that's a conscious or semi-conscious thing, and that's not about your internal language model, because the internal model is automatic.

random review said...

[QUOTE] As far as I can see, it comes down to this:
that structure wasn't part of their language model, and continued exposure to the language only trained their ability to filter the input to adapt it to their structure, rather than adapting their structure to match the input.[/QUOTE]

I think this sentence sums up one of the most important things in learning (especially language learning). How do we train ourselves to gradually refine our model rather than our ability to filter the input? It seems to me that most of the ingenious techniques people have come up with are ways to "force feed" yourself corrective feedback of some sort or another. What do you think? Glad to hear you're finally on the German!

Titch said...

Hi RR,

Well, German's been a sort of "now and then" dabbly thing for years. It was actually the first language I ever started attempting to learn, sitting outside the caravan in Kärnten when I was about 8 years old. "Herr Ratte ist ein*** Grau Rat***, Frau Ratte ist ein*** Wasserratte" or something like that.

I did about half of the first Michel Thomas course one day in 2006, and I've revisited it a couple of times, almost completing it, but at that point I was still studying for my French/Spanish degree and working on my Gaelic, so it never got very far.

In this case, though, I wasn't studying German for German's sake -- I wanted to check out the software, and the only language they offer that I know less about is Portuguese, but given that I know several Romance languages, it wouldn't have been an authentic beginner experience.

If I was wanting to learn German properly, I'd redo the end of the MT Foundation, then do the Advanced and plug your declension rules from HTLAL into a computer and practice that way!

Titch said...

As for building a better language model, the best way is to get it right from the word go, like MT tries. I'm not of the school of thought that says mistakes are necessary.

But once there is a mistake, yes, corrective feedback is required, and unlike what the communicative approach preaches, it has to be immediate. (For reasons discuss in this post.

But it also needs to be systematic errors, and they need to be systematically corrected. There's no point correcting a complicated structure if it's built up of 3 or 4 components that a student gets wrong. Correct the component errors and the larger combined error necessarily disappears. This was one of Thomas's strengths. A mere "correct answer" is not a "correction", because it doesn't necessarily resolve the underlying problem, and without an expert teacher like Thomas to "rewind" us to the basic components, it can be difficult to self-teach away these errors, but it can be done....

random review said...

Yes, I agree, with language feedback needs to be immediate. Most of the more useful techniques like shadowing give you that instant feedback. I also agree with your second comment, but that is MUCH more difficult to deal with as a self-learner (as you point out). An FSI (or possibly DLI) course I once read insisted that to be successful, learners had to be proactive about analysing their errors and trying to figure out the underlying reason. In the end, though, I agree: prevention is better than cure.

How do you try to deal with this?

I knew you'd dabbled in German for a while and thought you'd finally decided to get serious with it. One day you will ha ha.

random revuew said...

Er, should be a comma after "with language" in the above. Sorry if that makes the first sentence hard to read.

Titch said...

I think the best way to avoid making mistakes is again something Thomas did.

Thomas never divorced the structure from a meaningful context, compared with a great many older books that focus on structure and come up with meaningless sentences just to demonstrate the sentence.

On the other hand, Thomas's meaningful sentences were not as specific as many of the newer courses that focus on situational dialogues to give context, and were far more varied.

Old style books cause you to overgeneralise, by suggesting the target language structure is directly equivalent to the native language structure, new style books cause you to undergeneralise by not giving enough context.

For example, when I started his Spanish course, I didn't have any experience of a ser/estar style distinction (I knew that you say "sto bene" in Italian, rather than "sono bene", but that was as close as I'd got). Thomas introducing "es" without explaining the difference is one of the biggest complaints I hear from Spanish teachers, but every time he uses it, it's something that is unambiguously an intrinsic feature. I don't remember being at all confused or surprised when estar was later introduced.

So I suppose the trick is finding a "safe zone" to practise in where the structure and meaningmap well between languages, and only to step out of that safe zone when you've got a way to verify you're saying it a right. (IE with a teacher or a native speaker.)

Titch said...

And German was the last "big" language which I could use to evaluate new learning materials, so I was saving it for when I saw something I really wanted to give a proper try. I have now cashed it in, so I can give it a proper try some time in the near future!

random review said...

Safe zone, hmmmm. As a learner the only way I can think to be sure that I'm in a safe zone is NOT to create my own sentences unless there is immediate feedback (e.g. in Thomas' courses or from a native). When learning Spanish, I often used to ask myself, "how would I say X in Spanish?" It was something I started doing myself, but I know a few people even recommend this on HTLAL, which I think is unfortunate advice for any beginner that takes it. With German I'm forcing myself NOT to do this yet. Sometimes I will spontaneously think (and want to say) a phrase I heard in a TV program or Assimil dialogue in an appropriate context, but that's something quite different and much safer of course.

Thanks for your replies.

Titch said...

You've just made me think there...

Time was I'd look in a nice big dictionary if I wanted to find out a new word. I'd be looking at all the examples and looking for some kind of "centre" to the meaning -- the same sort of idea as the "safe zone". I would try to learn/internalise the meaning as per the centre, and let the "edges" remain fuzzy and uncertain.

So I would learn the "centre" (or multiple centres, for words with multiple meanings), but I'd guess at using the edges when necessary.

But these days I spend far less time looking at dictionaries, as I'm more likely to just look up the bare minimum of information required to understand or write the message in question.

I don't know whether I'm being lazy in a bad way or an efficient one....

random review said...

Depends how much reading (etc) you're doing in that particular language IMO. If you are doing a lot, you are likely to come across the word or idiom again in different contests (unless it's a rare word) until that centre (sematic core?) starts to become clear- and that seems efficiently lazy to my mind. If it's one of your languages you don't use that much, then I would say you need to make the most of your previous time with the language and you're being lazy in a bad way. Just my two cents FWIW. Hope I'm not being too presumptuous.

random review said...

Damn spellchecks! Typos and dpelkcheck software are a terrible combination. I meant context and precious; not contest and previous. Sorry, man.

Titch said...

Presumptuous? No. Right? Probably.

After 8 years of formal language study, with all the structure that brings, I suppose I've formed habits based on regular study and graded materials, something which I haven't had for Corsican (or Polish, which I never ever learned properly).

I suppose I have to reassess my ways of studying.

(And on a sidenote, now you've got me wondering what MS Office would do to "correct" some of Gollum's lines from the Hobbit and the Lord of the Rings...!)