Coolly rational in a second language
« previous post | next post »
Boaz Keysar, Sayuri Hayakawa and Sun Gyu An published an intriguing paper last month in Psychological Science in which they found that several different groups of bilinguals were more immune to common cognitive biases when making decisions in their second languages than in their native tongues. The paper has received a fair bit of attention in blogs and the media. I've added my own commentary in this post over at Discover Magazine, expanding on two plausible explanations for the effect that are alluded to in the original paper. Feel free to toss your comments in the hat over there, but I'll keep the comments open here as well for those who are in the mood for a more Language Loggy discussion.
Bob Couttie said,
May 31, 2012 @ 9:48 pm
Carrying out a similar study in the Philippines might be interesting. Many, if not most, people are multilingual in Tagalog/Filipino, a regional language such as Chabachanco, Visayan, Ilocano or Buhinon for example, and in English. In addition, most also speak a mixture known as Taglish and/or Engalog.
michael ramscar said,
June 1, 2012 @ 2:03 am
Another view of these results that they reflect a depressingly superficial understanding of how language works.
The rational answer to these questions is surely, "it depends." If the 600,000 people represent the total population of a unique people with their own, long history, as opposed to being a projected casualty rate in a population of billions, one's attitude to risk may well change. However, given the contexts in which these decisions were made — the rational, deliberative atmosphere that is promoted when students are asked to scrawl down answers on questionnaires for course credit — our decision makers opportunities to seek further informational in this way were limited.
Given this, what is a rational decision maker to do? If a decision maker is uninformed about how language works, or not particularly sensitive to the full range of information conveyed in a given linguistic message, he might decide to just flip a coin. Which is what it looks like the bilinguals did in these experiments.
But given that natural language is the medium used to pose these questions, is it really rational to make decisions about them from a position of linguistic ignorance, or in a language in which one is far from expert? (After all, in theory, people's — and peoples' — lives could be at stake.)
Given the way these questions were posed, surely the more rational thing for a decision maker to do would be to best inform herself about how natural languages work, and then bone up on the particular language in which the questions were posed.
A linguistically informed decision maker might reason that, given the constrained circumstances in which she is forced to choose between options a and b, by far the most rational of the options open to her would be to do a little Gricean reasoning. She might reason that given the way natural languages communicate meaning, there will be useful information about the questioner's communicative intentions in the actual framing of the question (given that we know questions might have been framed otherwise, and given that we know that she has no access to further information, this seems like a very rational thing to do). Indeed, if our decision maker was very skilled at using the language in which the questions are framed, she might even do this automatically. Which seems to be what the monolinguals in this experiment did.
Or, to put it another way, Wittgenstein once observed:
“One keeps hearing the remark that philosophy really makes no progress, that the same philosophical problems that had occupied the Greeks are still occupying us. … The reason [why this is true] is that our language has remained the same and seduces us into asking the same questions over and over. As long as there is a verb ‘to be’ which seems to function like ‘to eat’ and ‘to drink,’ as long as there are adjectives like ‘identical,’ ‘true,’ ‘false,’ [and] ‘possible', as long as one talks about a flow of time and an expanse of space, etc., etc., humans will continue to bump up against the same mysterious difficulties, and stare at something that no explanation seems able to remove."
Do these results show anything more than he really should have added 'rational' to that list.
Bump, bump…
GeorgeW said,
June 1, 2012 @ 5:32 am
My guess would be that this is because we don't have the same degree of pragmatic knowledge in our L2 as our L1. Therefore, we have a more literal interpretation of speech.
Amanda said,
June 1, 2012 @ 6:03 am
Has anyone considered that the more slow and thoughtful processing of bilingual speakers in their non-native language might have something to do, not with how they use their own fluency, but how they learned the language to begin with? First language acquisition is itself a more instinctive, quick, gut instinct type of learning. You don’t intend to learn it, you aren’t thinking about the grammar or semantics consciously, you are simply grasping at whatever will allow you to communicate and you aren’t interested in analyzing that form of communication. Second languages, however, are almost always learned by deeply studying the structure of the language. Examine the paradigms, learn the grammars, memorize the vocabulary. It’s a slow and labored process which generally has nothing to do with communicating (at least on an immediate level) and more to do with scientific exploration. It’s not necessary for survival, but the pursuit of knowledge. And the way we use our language can be influenced by the way we think about the language we’re using. So we very well might be using a second language for analytical thoughts because those are the methods by which we learned to think in it.
I’d be really curious about what processing techniques are found in the various languages of children from bilingual homes. I’d bet it’s the more instinctive type.
AlexB said,
June 1, 2012 @ 7:20 am
I can attest to the effect on swearing in a second language. I am a lot more vulgar in French than in Dutch.
Ruben Polo-Sherk said,
June 1, 2012 @ 8:32 am
One possible mechanism is that when processing the question posed in a foreign language, they translate it into their native language; in that process, it is possible that the framing effect is removed by abstracting the information to "facts" in order to be able to do translation.
J.W. Brewer said,
June 1, 2012 @ 9:17 am
Yes, further to what Ruben said, it's not at all clear to me that it's accurate to say (any/most/all of) these test subjects are "making decisions" in the L2 as opposed to hearing/reading the question in L2 and internally translating it into L1 and then "making decisions" in L1.
Circe said,
June 1, 2012 @ 12:24 pm
I can say from anecdotal evidence that the swearing-is-less-inhibited-in-second-language effect is rather strong in India (where many people are bilingual in their local language and English). People who would never utter a single swear word in their native language often have no difficulty going all out swearing in English.
LDavidH said,
June 1, 2012 @ 12:31 pm
@Ruben and J.W. Brewer: As a non-native English-speaker who uses English for most every-day situations, including decision-making, let me emphatically state that I do *not* translate anything into my mother tongue! On the contrary, I sometimes find it very difficult to do so when required. I also did not plan this comment in Swedish and then translate it into English; my decision to write it was based on reading an English blog with English comments, and deciding (although whether that involved using language or not is a different question) to respond in English.
Most people who regularly speak a second or third language do not translate between it and their mother tongue; we simply function in several different languages, with no translation involved.
Glenn Bingham said,
June 1, 2012 @ 1:48 pm
@AlexB
That is why there is a saying in English used before a sentence spiced with tabooisms: "Pardon my French, but…"
Ran Ari-Gur said,
June 1, 2012 @ 2:56 pm
@Amanda:
> Has anyone considered that the more slow and thoughtful processing of bilingual speakers in their non-native language might have something to do, not with how they use their own fluency, but how they learned the language to begin with?
Yes, people have considered that. Dr. Sedivy, in the commentary she links to, wrote that according to one possible explanation, "Words in a foreign language simply don’t set off emotional resonances as they do in a native tongue because the associative links are fewer and weaker (and perhaps connected to the more cerebral classroom settings in which they were first learned)—meaning that we quite literally feel less when using a second language." And while the original paper is a bit less explicit about this, it does make a point of stating that "Participants had acquired the foreign language mainly in a classroom setting and did not have a parent who spoke it as a native tongue", and it alludes repeatedly to this fact, which suggests that the study authors considered it to be relevant.
john said,
June 1, 2012 @ 3:23 pm
Yes, it's very important to qualify how the foreign language was acquired.
For example, to relate to LDavidH's comment, in my experience Swedes coming to London for the first time initially sound like Americans and speak in a very "cerebral classroom-like" manner, but then become more English in their speech (especially swearing!) and mannerisms over time.
Unfortunately there aren't too many opportunities to speak Chinese in the UK, but I agree that certainly in everyday situations, I would never translate on-the-fly. However, I do find myself using the language in which I was taught a subject. As I learned maths predominantly using Chinese, I often find it easier to perform mental arithmetic in that language, but ask me to talk about economics in Chinese and I do need to translate from my English knowledge base.
quixote said,
June 1, 2012 @ 3:58 pm
I grew up bilingual (Russian first, English second) in English-speaking countries. We spoke Russian at home, I learned English in day care before being aware of doing so, and the rest of my family spoke English fluently. I was always aware my English speech was less emotion-connected than my Russian. That's still true, decades later, when I've spent my whole life speaking English and my Russian is very rusty.
Now, whether that makes me more rational in English, I'm not sure.
I also second the commenters who point out that multilinguals don't translate. I learned French, German, and Dutch later in life. Once I was fluent in those languages, it took effort if someone using another language asked me what had just been said. Then I translated. That's a very different sense of thought processes from the plain understanding-response in a given language, whether it's one you learned at three or at thirty.
Dan Hemmens said,
June 1, 2012 @ 4:06 pm
Having skimmed the paper, I'm a bit concerned about some of the decisions they give people (particularly in experiment two).
My understanding of rational choice theory is limited, but from my very basic understanding it is not "irrational" to refuse a bet with a positive expected value, because while the nominal value of the bet may be positive, the actual consequences of losing may be far greater than they first appear – I would not bet a hundred thousand pounds against an even chance of winning a hundred and fifty thousand, because I do not *have* a hundred thousand pounds, and so losing the bet would completely remove my ability to support my lifestyle, while winning it would give me a lot of money for which I have no particular use.
So perhaps all that is happening here is that people answering questions framed in a second language understand context *less clearly* (looking at the assessments people made of their language proficiency – something I note that nobody actually tested) and as a result made less considered decisions. Notably, in every case the most "rational" option was always the one which required the subject to gamble the most (or in the case of the first test, to be indifferent between two options).
By way of anecdotal evidence, I work in an international school, and I do sometimes find that my students (for whom English is a second language) avoid the "standard" mistakes that native speakers make in their initial response to a problem, only to fall *into* the standard mistake once they have understood the question clearly enough to be misled by it (I've found this happens a lot with the "disease strikes one in ten thousand / test is 99% accurate" problem).
I also wonder if working in a second language (and, more specifically, a second language in which one is not fully proficient and which one perhaps associates only with an artificial, classroom environment) makes one less likely to take things "seriously". Being asked to work in a foreign language might make people more inclined to treat the whole thing as a game, which might make them more likely to take the gambling/risk-taking options.
To put it another way, I rather suspect that *drunk* people are more likely to be indifferent to the context surrounding a question, and to gamble if given the opportunity. I don't think that constitutes evidence that drunk people are less prone to cognitive biases.
Coby Lubliner said,
June 1, 2012 @ 5:56 pm
It is not clear what the study means by "bilinguals." Are they (1) people who just learned an additional language as a "foreign language" in school, or (2) people who are fully at home in two languages, or (3) diglossics who use one language at home and another at school or work? It would seem that, even if language did influence thought, the effect would be quite different.
Ruben Polo-Sherk said,
June 2, 2012 @ 8:56 am
LDavidH, quixote: I'm just like you–no matter what language I'm using, there's no translating going on. But from my experience talking to other people about using second languages, I gather that its quite common even among people with reasonable proficiency to do this sort of translation (or, at least, that they perceive that they do).
J.W. Brewer said,
June 3, 2012 @ 8:00 am
If you click through from the original post to the paper itself they describe their various groups of test subjects, e.g. Anglophone U.S. college students who have taken several years of instruction in Japanese, or Anglophone college students studying in Paris after however many years of French instruction, or Korean college students that have had however many years of ESL instruction in the Korean school system. The test subjects certainly do not self-identify as equally fluent in the L2 as the L1. How many of them are across the threshold where they are "thinking in" the L2 (when primed by being asked a question in L2) rather than internally translating is not at all clear from the description, but I see no reason to suppose that all of them are.
J. Goard said,
June 5, 2012 @ 1:00 am
I find that many interesting facts become more understandable when the body of evidence that is normally summarized as
"Adult language learners are worse than early childhood learners."
is reconceptualized as
"Adult humans are more highly resistant than young children to chronic infection by linguistic replicators."
A.M. said,
June 6, 2012 @ 5:00 am
Thanks for the link.
I found out that, if I ask a question at a Russian (my 1st language) web site and then go to a similar English (2nd language) resource, half of the time I don't actually post the question, because, after I put it into English, it makes very little sense. Either a false dichotomy becomes apparent, or one of the alternatives suddenly shines in a new light, or something, but the result is that I don't ask anything on Russian web sites anymore, if the question can be answered by English-speaking persons and do so very seldom.
srina said,
June 7, 2012 @ 8:58 am
Speaking my native language needs no efforts. I believe that language learning process has been completed during the childhood. I do not think what to say and how to say things. I just know them. That is not the case with English, my second language, even though I have learned it since I was ten. I still think what to say and how to say things.