kitchen table math, the sequel: second language learning - the "less is more" hypothesis

Wednesday, July 20, 2011

second language learning - the "less is more" hypothesis

I'm reading Choke by Sian Beilock, a terrific book. Mostly it's about why and how people choke under pressure, but I've been surprised at the number of ancillary topics that turn out to be related to choking -- including foreign language learning by grown-ups.

Beilock says that the reason children learn language better than adults is that children have less working memory (pdf file). Less is more.

She has fascinating things to say about math and problem solving, too.
Statement of Research Interests (pdf file)
Alan W. Kersten
This research has been testing one hypothesis for why adults have so much difficulty successfully acquiring a second language, namely the “Less is More” hypothesis of Elissa Newport (1990). According to this hypothesis, the reduced working memory capacity of children relative to adults actually results in better language learningby forcing children to focus on small chunks of language. Adults, on the other hand, can remember larger chunks of language, allowing them to memorize useful expressions in a foreign language (e.g., “Where is the bathroom?”), but making it difficult for them to extract the lower-level meaning elements from which those expressions are constructed. Adults are thus limited to the set of phrases that they have acquired, and are unable to recombine the lower level elements from which those phrases are constructed to express novel meanings. If this hypothesis is correct, one may predict that adults will learn a language better if they are forced to focus on small chunks of language rather than being allowed to learn entire phrases. We have tested this prediction using a miniature artificial language learning paradigm (see Kersten & Earles, 2001). One group of adults was presented immediately with complete “sentences” from this language, whereas a second group was presented initially only with individual words from the language. This second group was subsequently presented with incrementally longer chunks of language until ultimately they were hearing the same sentences that the other group heard all along. The group that was initially forced to focus on small chunks of language showed better ultimate learning of the word meanings and morphology of that language, consistent with the “Less is More” hypothesis. We are currently investigating whether starting small benefits the acquisition of a natural language with more complex grammar, namely French (Chin & Kersten, in press).

update: Less Is Less in Language Acquisition

Choke: What the Secrets of the Brain Reveal About Getting It Right When You Have To


Catherine Johnson said...

Less Is Less in Language Acquisition

Anonymous said...

I'm sorry, but their test doesn't confirm their hypothesis.

The folks who got only individual words, then longer chunks, then sentences were experiencing distributed practice of the vocabulary.

If this were truly the best approach, then experts wouldn't be so critical of Rosetta Stone.

Glen said...

As language learners, adults are better in some ways and children are better in others. In general, adults learn more quickly, while children learn more accurately.

Programmers tend to reuse working code when they can and build custom code only when they must. Our brains are like that. If your brain can find an existing pattern that seems to work well enough, it will use it. Only if it can't will it go to the effort of learning a new pattern.

Adults, with a vast storehouse of semantic and syntactic patterns already in their heads, tend to build a new language like a pre-fab house, quickly combining existing components with some new, custom parts to put up a workable, but clunky structure ASAP. Once it's good enough, resources are diverted elsewhere so, while it may be improved here or there, it's never rebuilt from scratch. It remains clunky but functional.

Kids, with fewer existing patterns to repurpose (as well as extraordinary plasticity for precise custom adaptations), are forced to build much more from scratch. It takes longer, but the ultimate results are superior.

Different social circumstances between adults and children (immigrant families, for example) often mask the adult-type superiority, because adults often find ways to reduce their exposure to the local culture, while their kids end up fully immersed. Kids will, of course, learn more quickly if they get more exposure. So will anyone. Many adults will also resist the pressure to conform, while their children want to fit in. Skilled adult language learners, like skilled actors, are able to relax into a new role and become chameleons, like kids.

Given equal desire to conform to the new language and equal exposure, an adult can come up to speed a lot more quickly than a kid. If the immersion continues for several years, though, the kid will inevitably surpass the adult.

Amy P said...

That is very interesting, Glen.

"Given equal desire to conform to the new language and equal exposure, an adult can come up to speed a lot more quickly than a kid."

It drives me nuts when I hear the radio ad for Rosetta Stone where they promise listeners that if you buy their product, you'll learn your new language just like you learned your first language! I don't know about everybody else, but one of my kids barely said anything until he was 2. As an adult learner, I wouldn't be very happy with that rate of progress.

Glen said...

Right. To get the same language exposure that a typical suburban kid gets before kindergarten, an adult would have to study an hour a day, 365 days a year, for roughly 30-40 years! An hour a day every day with no days off is a solid commitment, and adults working that hard on a language simply don't take anything close to forty years to reach the communicative effectiveness of a kindergarten kid.

The kid will be obviously native sounding, while the adult may still be obviously non-native, but the adult will be able to exploit a vast storehouse of background knowledge, both linguistic and non-linguistic, in the service of communication. Even among adults, the Defense Language Institute discovered that superior knowledge of one's native language and superior general knowledge of the world both seemed to contribute to one's ability to learn a new language. Consider how big those advantages are when comparing an adult to a young child.

And yet, given enough time, a child's greater plasticity is an insuperable advantage.

Anonymous said...

How is this even a statement about working memory? Why do we believe that children have less working memory than adults?

Isn't a more obvious explanation that children haven't chunked as much knowledge as adults, and so they don't have convenient handles for whole sets of structures the way adults do, because they haven't yet had enough exposure to the structures to create those abstractions?

Adults are experts at many things; being an expert means large amounts of knowledge are structured in a specific way, and then those structures are stored into compact subroutines. That's chunking. You can have small working memory, but if you've chunked nearly all of the necessary procedures, it won't be an impediment. Think about driving a car: if you've become an expert at driving a car, you don't need working memory at all to do it. Your brain also knows how to shift gears with the hand while working the clutch and accelerator with feet, or how to check the blindspot while turning on the turn signal.

wrt to multiple language acquisition, I'm not sure I believe as Glen does that "Adults, with a vast storehouse of semantic and syntactic patterns already in their heads, tend to build a new language like a pre-fab house, quickly combining existing components with some new, custom parts to put up a workable, but clunky structure ASAP." Clearly adults as experts have these vast storehouses of patterns in their heads for lots of tasks, but for language? Naah. Now, SOFTWARE ENGINEERS do, sure, but they are experts in a way that generic adults are not. "The first ten languages are the hardest" my old linguistics prof would say, and assuredly a software engineer whose seen ten has hierarchical knowledge that a student on their second programming language does not.

But I agree with Glen's general point that experts can just graft pointers on top of already existing structures, while children still need to organize the structures.

The Kersten piece confusing: adults recalling larger chunks of language is means they ARE NOT USING more working memory; the lack of distinction between recalling and newly "remembering" here makes their experiment mostly useless.

Catherine Johnson said...

Why do we believe that children have less working memory than adults?

I'll check, but I don't think there's any real dispute about this.

Lower working memory is one of the things that makes kids kids as far as I know.

Catherine Johnson said...

Wait 'til you see the math problem people with brain damage do better than people without.

There is also a set of problems that children solve better than adults. (Not math problems --- )

Catherine Johnson said...

I posted a study of working memory in children.

Working memory is a large part of executive function; almost by definition working memory is lower in kids.

Catherine Johnson said...

I would have to read this again to confirm what I'm going to say, which I'm not going to do now, but I don't think we're talking about 'chunking' here. Not in the way we normally talk about chunking.

I think the authors are talking about adults losing the 'ability' to perceive tiny little parts of a stimulus.

That was what Animals in Translation was about; Temple and I argued that animals and autistic people share an ability to see tiny parts instead of wholes.

Animals are terrified by details that normal humans simply can't 'see.'

It's extremely difficult to 'see' parts instead of a whole. That's why the SAT geometry questions are so difficult, I believe. (Even the best scoring kids & tutors tend to miss the geometry questions disproportionately, in my experience.)

Betty Edwards upside-down Picasso exercise is a terrific illustration of the extreme difficulty normal adults have in breaking a visual image into component parts.

Catherine Johnson said...

In other words, it's not that grownups chunk brand-new language into sentences in order to remember it better.

It's that grownups can't help but perceive big chunks of a foreign language instead of the small bits & phonemes making up those sentences & words.

Glen said...

Allison: Clearly adults as experts have these vast storehouses of patterns in their heads for lots of tasks, but for language? Naah.

Mon dieu, but of course we do. My native English, like that of other native-speaking adults, is almost entirely a vast storehouse of patterns of pronunciation, prosody, morphology, syntax, idiomatic usage, vocabulary usage, and so on, that I have accumulated and practiced to fluency over many years. I can't help carrying many of these patterns over into other languages, usually without realizing it, just as native speakers of other languages unwittingly bring their non-native patterns with them when they try to speak English: Throw Momma from the train a kiss!

But you know this, so maybe you thought I was saying that adults somehow become experts in "language" in general, as opposed to any specific language. Definitely not. I meant they accumulate a vast storehouse of patterns in their own native dialect. They know how to express a huge number of things in one language. When they try to learn another language, they don't have that database, so they fall back on patterns from their native language for many things. Sometimes this fails, but often it works. People often understand you, even if your way of saying things isn't the native way, and you may be able to quickly put together a collection of "ways to say things" that meets most of your needs. When you want to use an "Even though..." sentence, or an "If I hadn't..." sentence, or whatever, you have a stock way of doing it that native speakers will understand, even if they aren't native sounding. This is the clunky, pre-fab construction I was talking about. Much of it is not learned from the new language but brought over from the old.

Catherine Johnson said...

I found the term that distinguishes 'chunking' in the sense we're talking about it here from 'chunking' in the sense of chunking knowledge in order to hold more content in working memory -- will get that posted, too.