Last week the linguistics department here at Michigan hosted the 2012 “Marshall M. Weinberg Symposium”:http://www.lsa.umich.edu/linguistics/events/marshallmweinbergsymposium. The theme for this year’s symposium was _bilingualism_. I learned a ton from the various speakers, much of it about how hard it was to learn a second language after very early childhood.

Even people who appear, to naive judges, to be fluent in a second language they learned after childhood, perform “well below native speakers at cognitively demanding linguistic tasks”:http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9922.2009.00507.x/abstract, such as understanding speech in noisy environments, or explaining proverbs. I don’t have the citation link for this, but “Jürgen Meisel”:http://www1.uni-hamburg.de/romanistik/personal/w_meis.html reported that German students learning French by immersion did much better if the immersion started between 32 and 42 months than they did if they started after 42 months. The errors that he reported were common among the older learners after several months of immersion, like not getting the genders of articles right even for words like maman where you would think it was obvious, were really striking. “Karen Emmorey”:http://emmoreylab.sdsu.edu/director.php reported that the same thing was true for learners of ASL; late learners can become fluent enough for practical purposes, but are never as good as people who learn ASL in early childhood.

The striking contrast to all this is how successful first language acquisition is. To a first approximation, 100% of people successfully learn the syntax of their first language, and do so at a staggeringly young age.

I realised a few days after the symposium that there was a huge question I wish I’d asked. _Why_ are we so good at learning a first language, and so poor at learning a second language. What cognitive system would have such a feature(/bug), and what evolutionary advantage could there be to having such a system?

Here’s one possible answer that I think is simple, explanatory, and sadly not consistent with the data. As Gilbert Harman noted in his talk, philosophers have long argued about the question of whether humans think in language. (Being philosophers, they’ve also argued about what the question even means, and that’s not a trivial issue.) Let’s adopt the following working hypothesis: humans who have learned a first language think in it, those who haven’t, don’t. This transition, from not thinking in language to thinking in it, runs very deep in the system. Once you have learned a language, it is impossible to not think in it. Compare the striking fact that once you learn a language, it is impossible to not interpret sounds you hear that are communications in that language. Once this transition is made, learning a language goes from being an instinctive task to a cognitive task. In “Kahneman’s terms”:http://www.powells.com/biblio/18-9780374275631-0, it goes from being a system 1 task to a system 2 task. And learning a language is just too hard a task for system 2; it is literally harder than rocket science or brain surgery.

This obviously can’t be the complete story; a full explanation would need to fill in a lot of gaps. And the analogy I appeal to with Kahneman’s system 1/system 2 can’t be completely right. In Kahneman’s examples, system 2 is supposed to be more accurate than system 1, but self-consciously learning a language ends up being less accurate. The bigger problem, however, is that the hypothesis gets the timing all wrong.

The Meisel studies I mentioned do say that learning a language gets harder after about 42 months. They don’t say it gets harder after 24 months, let alone after 36 months. In fact, the students who start second language immersion then seem to do pretty well. But my little hypothesis would predict they’ll do badly, since by those ages they do speak their first language.

There’s also a problem at the other end. Profoundly deaf children with hearing parents often don’t learn a sign language until very late. And since they are deaf, they don’t learn a spoken language either. But that doesn’t mean their minds retain the plasticity to adopt a new language as a native speaker. Instead (at least according to results Karen Emmorey mentioned in Q&A), they do worse than even second language learners of ASL.

So I don’t have much of a theory as to why we should be wired this way. I assume that something in the ballpark of my hypothesis is right. Our ability to learn language isn’t switched off, it is inhibited by some other abilities we acquire. But how that inhibition works, I don’t really know.

It’s possible of course that there is a widely known and well supported explanation for this phenomenon. If so, I’ll be a little disappointed that it didn’t come up last week. But I rather doubt it does exist. One of the themes of the talks was that there was much less research on bilingualism than you’d expect, given how big a feature of the world it is. And most of the relevant data being discussed seemed to be from very recent studies. So I think until recently we didn’t know many of the facts to be explained, let alone have an explanation of them. Still, this is not at all an area I’m an expert in, and I’m sure many TAR readers will be able to point to more informed speculation than mine.

The talks at the Weinberg Symposium were videotaped, and I believe they’ll be posted to the web shortly. I’ll update this post when that happens.

Thanks again to Marshall Weinberg for sponsoring this event. Marshall sponsors many, many things at the University of Michigan (including my job!), and the intellectual environment here is much richer for it.