Does my Chinese language subconsciously influence me to save more, avoid smoking, exercise more, and eat more healthily?
Prof. Chen’s argument essentially runs as follows: certain languages, such as English and Greek, oblige speakers to separate out present and future tense (“I will go to the store tomorrow”), while other languages, such as Chinese and German, prevent speakers from doing so (“我明天去商店”, or, “I go to the store tomorrow”). Could this linguistic separation of present and future in certain languages lead speakers of that language to discount the future, thus reducing their propensity to save and prepare for the future? Since this is certainly not a question that can be answered using a randomized controlled trial, Prof. Chen goes the big data route; pulling data from the largest data sets in the world and running regressions to parse out the effects of numerous possible confounding variables, so that he can compare the savings behavior of people born in the same country, living in the same country, of the same sex, age, income level, educational achievement, family structure, religion, etc. and differing only in the language they speak. What he found was a surprisingly rigorous correlation despite controlling for all of these other factors: “Futureless” language speakers are 30% more likely to report having saved in any given year, 20-24% less likely to smoke, 13-17% less likely to be obese by retirement, and 21% more likely to report having used a condom in their last sexual encounter.
Two Linguists Offer Their Criticisms
I listened to a recording of this TED talk while practicing my (apparently linguistically-driven) morning run yesterday, and slowly shifted from skepticism to mind-blown acceptance as Prof. Chen covered his statistical bases. But it’s easy to become a believer when you’re neither a professional linguist or economist. Upon arriving home, I did a quick search and sure enough, uncovered well-written criticisms from two linguistics professors:
Mark Liberman of the University of Pennsylvania highlights the dangers of running statistics on factors that may very well not be completely independent. He uses graphical depictions to show that as soon as you introduce the assumption that certain factors are correlated in their geographic diffusion (e.g. language and culture), it becomes quite easy to find highly statistically significant correlations between non-causally related phenomenon.
“If you torture the data long enough, it will confess.” -Economist Ronald Coase
Geoffrey K. Pullum of the University of Edinburgh has a slightly more piquant response (including taking a jab at Prof. Chen for being a “non-linguist”), challenging Prof. Chen’s categorization of languages into strong and weak “future time reference” languages. This underlied some of my initial skepticism as well (after all, I can easily say, “我明天会去商店”, or “I will go to the store tomorrow”). Perhaps most alarmingly, Osten Dahl–the same scholar whose work Prof. Chen relies upon to categorize his languages–shows up in the comments section to emphasize that he never created hard categorizations in the first place.
These posts touched off a series of additional conversations, which you can find outlined here. Not being an expert in either field, I’ll leave it to you if you want to continue down the rabbit hole of this very interesting debate. However, I did want to highlight one remark from Prof. Chen that I think we can all learn from.
Stepping Out of the Echo Chamber
Earlier in the week, I had listened to a TED Talk (maybe I’ve overdosing a bit here) by Margaret Heffernan, who recounted the story of Alice Stewart. Alice Stewart was an English physician and epidemiologist who in 1956 published a paper showing for the first time that children of women who received X-rays while pregnant were significantly more likely to die from childhood cancer. What’s most striking about this story is the fact that Alice had to fight entrenched interests for 25 years before the British and American medical authorities changed their practices, and how her working relationship with a statistician named George Kneale helped her persevere through the fight. As Ms. Heffernan explains, George saw it as his job to prove Dr. Stewart wrong. He crunched the numbers in different ways, explored possible confounding factors, and tried everything within his statistical power to poke holes in her conclusion.
“Because it was only by not being able to prove that she was wrong, that George could give Alice the confidence she needed to know that she was right.” -Margaret Heffernan
This brings me to a statement that Prof. Chen included in his response to Profs. Liberman and Pullum that jumped out at me:
“…all analyses I conduct are on publicly available data, and I’d love to talk more with readers interested in replicating, extending, or testing these results in ways I haven’t thought of.”
Too often we enclose ourselves in an echo chamber of like-minded thinkers, looking for the next killer argument to throw at “the other side”. Regardless of whether Prof. Chen’s hypothesis turns out to be valid or not, his openness to engaging in constructive conflict is something I think we should all strive to emulate.