Slashdot Mirror
Localizing Language In the Brain
RogerRoast writes "A new study by MIT scientists pinpoints areas of the brain used exclusively for language (PDF), providing a partial answer to a longstanding debate in cognitive science. According to the study, there are parts of our brain dedicated to language and only language. After having their subjects perform the initial language task, which they call a 'functional localizer,' they had each one do a subset of seven other experiments: one on exact arithmetic, two on working memory, three on cognitive control, and one on music; since these are the functions 'most commonly argued to share neural machinery with language.' The authors say the results don't imply that every cognitive function has its own dedicated piece of cortex; after all, we're able to learn new skills, so there must be some parts of the brain that are both high-level and functionally flexible."
For those who need a bit of background about what this is all about, and why this study is so important to the study of cognitive linguistics, we turn to a bit of history.
Linguistics has always been closely intertwined with psychology. So much so, in fact, that both modern cognitive linguistics and psychology approaches stem from reactions to an idea called behaviorism. Everyone's 'favorite' linguist, Noam Chomsky, was one of the first to try and go beyond behaviorism's explanations. Much has been said and written, and I won't go into that whole mess, but suffice to say after the dust settled Chomsky had decided that the human acquisition of language is very much an innate property of our species, something inherent in our brains, which he would come to refer to as the 'black box' that just acquires language like a sponge that dries up after a certain age in childhood. Once again, the whole debate around this topic is what cognitive linguistics is currently bent on figuring out--a question that has existed since man first wondered "why language?"
Anyway, before this becomes a true wall of text, I'll come down to what this study means to cognitive science: the two camps directly affected by this study are named 'nativists,' who believe that the human brain has structures specifically designed for the acquisition, processing, and production of language, and the other side are called 'structuralists,' who believe that the natural human proclivity for pattern recognition is naturally reinforced during language acquisition, bootstrapping its own language recognition abilities by simply recognizing patterns. Pinpointing specific, exclusive areas for language supports the nativist conclusion, dealing a blow to the structuralist theory. Evolution at work, perhaps?
Quote: "Researchers in Israel, Canada and France used brain imaging to observe the neural activity of eight blind subjects as they read Braille. They found that although the blind subjects were using their sense of touch, their brains showed activity in the same so-called visual region that sighted people use when they read."
More at http://www.scientificamerican.com/article.cfm?id=the-reading-region
- "Every demand is a prison, and wisdom is only free when it asks nothing." Sir Betrand Russell
As someone who does neuroimaging research, what appears to be exciting about this approach to fMRI is that it is on an individual-by-individual basis, not at a group level (which is mentioned in the MITNews article). Almost all fMRI work is at a group level. While I perform some group analyses, most of my work is on an individual basis (but I do structural imaging, not functional). Group analyses can have severe limitations that are not always discussed by the researchers and are almost never understood by people outside the field of neuroimaging.
From the article: "It’s the same way for brains. 'Brains are different in their folding patterns, and where exactly the different functional areas fall relative to these patterns,' Fedorenko says. 'The general layout is similar, but there isn’t fine-grained matching.' So, she says, analyzing data by 'aligning brains in some common space is just never going to be quite right. Ideally, then, data would be analyzed for each subject individually; that is, patterns of activity in one brain would only ever be compared to patterns of activity from that same brain."
This process of aligning brains is called registration. Even if you are just working within one subject, there is registration involved (between the functional scan, in this case, and the structural - so you know what part of the brain is being activated). I spend about 25% of my imaging work dealing with checking registrations or trying to improve registrations. It's really a key step in neuroimaging work, one that not enough researchers consider seriously enough. So that's why this approach to fMRI is interesting - the researchers are trying to minimize the effects of poor registration, which can lead to completely invalid results.
I just watched this @Google talk yesterday, which finds wide variation in the way people think about various tasks. It doesn't contradict the findings of this MIT team, it just shows how variable and "plastic" these functions can be. One example that comes to mind is students from one country (France, IIRC) showed a lot of activation in the hearing areas of the brain when doing simple arithmetic tasks. They said this was because they learned arithmetic through rote repetition of tables, and thus used those aural regions when doing the tasks. (They also said they preferred doing math problems in a quiet environment to avoid distraction.)
IANA neuro-scientist, I just enjoy learning about this stuff. For any other armchair brain enthusiasts out there, you might also enjoy this lecture series on Human Behavioral Biology by Robert Sapolsky at Stanford.
XML is like violence. If it doesn't solve your problem, you're not using enough of it. --AC
After having their subjects perform the initial language task, which they call a 'functional localizer,' they had each one do a subset of seven other experiments:
They could have just monitored a Geek trying to talk to a sexy girl and look for the part of the brain that shut down.
Having to work for a living is the root of all evil.
I've got an undergraduate degree in linguistics - which, granted, isn't much, but i did spend some time learning about language acquisition. The general consensus within linguistics is that there exists both a language acquisition device (LAD) and a critical period for language learning. Language learning is a biological process on par with learning how to process visual data that (in neurotypical individuals) unfolds regularly given adequate input. After somewhere between 11-13 years the LAD switches off, and languages that are learned after this critical period are typically learned imperfectly.
I've seen side-by-side fMRI scans of people speaking two languages they learned before the critical period and of people speaking a language learned before and a language learned after. In the true bilingual speakers, both languages lit up the same area of the brain, and in the speakers who learned a new language after the critical period, the post-critical language lit up a different area of the brain from the native language.
As far as post-critical-period second language acquisition goes, there is some indication that the LAD is involved in the process - there is a specific order in which English speakers will learn grammatical features of German regardless of who taught them or what method they used to learn the language. There are actually some language acquisition theorists (Krashen in particular) who think that language processing and production (at a grammatical level) is all done at an automatic level, and that all our conscious brains do is monitor what comes out.
The environment you learn the language in and your degree of identification with the target language's culture do play a pretty big role in how accurately you'll be able to reproduce the target language, though.
Also, programming "languages" aren't capital-L languages and are (presumably) not handled by the part of the brain that handles language.
I can speak a little from practical experience, even if I'm not an authority. My wife has aphasia as a result of a stroke in her left temporal lobe. Immediately after her stroke she struggled to remember her own name and language-based communication of virtually any kind was almost impossible (written, verbal). However, she had no trouble understanding pictures or drawing them, and heavy use of a smartphone with google image search was able to get us through the early problems.
Since then she has recovered quite a bit, though she struggles especially with proper names, and her vocabulary is nowhere near what it used to be.
She never had any problems with movement (but some with vision - the stroke carried over into the occipital lobe). She could go through the operational aspects of maintaining her checkbook though she often got the math wrong (I suspect largely because she couldn't recognize the numbers - not because she didn't know how to add). From the moment she was home to this day if there were any question about where she needed to be I'd just get in the car and have her give me turn-by-turn navigation instructions and we'd end up exactly where she wanted to be.
Before this whole episode I would have assumed that the brain just worked like some like of abstract neural network where data goes in and comes out and how it gets from one to the other is just the result of training and varies person by person. Since then I've learned quite a bit and varies lines of evidence exist that suggest that many areas of the brain are highly specialized. Sure, within those areas neural networks may be what cause learning and adaptation, but if you stick a blood clot in the left temporal lobe, or the right temporal lobe, you'll wipe out a person's ability to use language in two completely different ways.
And I'm talking about language here - which encompasses a lot (listening, speaking, reading, writing, and likely more). Reading might be relatively new, but verbal language is likely to be MUCH older. And, since people can do it an most animals can't, it stands to reason that there is some biological reason for this.
Oh, aphasia can impact lots of other things as well - like short-term memory. The thinking (as I've heard), is that our short term memory often is augmented by repeating things to ourselves, and aphasia apparently inhibits your brain's ability to even talk to itself inside your own head.
I think these kinds of findings might have profound impacts on the pursuit of AI. It isn't enough to have a big network and good training method. You might need to pre-wire the network in some way to get something that resembles a human intelligence and not just the neural net you might find in a jellyfish or something.
Good points. And there's another dimension, that hasn't been picked up in this discussion yet - modality of learning. Second-language learners may be visual, kinetic or audile types (or combinations of these, though there's usually a dominant mode), and successful learning depends on a teacher recognizing a mode and adopting an appropriate methodology. I've also observed that some learners have an affinity for one language (or language group), and experience difficulty getting to grips with other language groups. Another thing I've noticed is that, until the age of nine or ten, youngsters may have a separate 'language' for each of the people close to them. This is often seen in marriages where the parents have different mother-tongues and a child attends a playgroup or nursery school in a third language. Acquiring a language by analysis and conscious learning strategies usually kicks in around eleven or twelve.
... After somewhere between 11-13 years the LAD switches off, and languages that are learned after this critical period are typically learned imperfectly.
...
So your saying the Spanish classes I flunked in high school wasn't my fault?
Be seeing you...