3 ms·
As a reader of the article kindly submitted here who speaks and reads multiple human languages from more than one language family, I'm not convinced that this a
by tokenadult 11y ago
As a reader of the article kindly submitted here who speaks and reads multiple human languages from more than one language family, I'm not convinced that this article passes the linguist's or polyglot's smell test. The sample size reported in the article is wholly inadequate for showing that there is a real effect here. That the underlying study was reported in a rather minor journal is a further indication that there isn't any real-world significance behind the statistical significance claimed by the researchers. For good background reading on why you shouldn't look merely to statistical significance claims for evaluating research on human subjects like this, see the articles by Uri Simonsohn[1] and his statistically astute co-authors, who have identified phenomena called "researcher degrees of freedom" (there are a lot of researcher degrees of freedom here) and "p-hacking" (a news release that mentions "more accurately than by random chance" is almost always a sign of p-hacking) that allow many weak, underpowered studies of human behavior to be published in journals with poor statistical skills among the editorial staff.
[1] http://opim.wharton.upenn.edu/~uws/ http://opim.wharton.upenn.edu/~uws/
http://www.p-curve.com/ http://www.p-curve.com/
- canjobear 11y ago> That the underlying study was reported in a rather minor journal Cognition is one of the top journals in cognitive science. Definitely not a minor journal for this kind of work, and plenty of their reviewers are statistically competent. So my prior would not be that the stats here are bogus. I share your skepticism of their result about more synesthetic people being more able to guess the meanings of foreign words. The effect size there looks tiny, the motivation is a bunch of hand-waving, and I wouldn't be surprised if it doesn't replicate. On the other hand I'm inclined to believe their result that people can guess the meanings of certain foreign certain words with above-chance accuracy (about 60% for big vs. small). It's just a version of the well-replicated bouba/kiki effect[1], plus the idea that this can influence words in a language, which doesn't seem far-fetched at all. What I'm most concerned about is that people might be picking up on similarities between words beyond what the authors identify as "cognates". The sample languages are Albanian, Dutch, Gujarati, Indonesian, Korean, Mandarin, Romanian, Tamil, Turkish, and Yoruba. 4 of those are Indo-European and might have obscure etymological relations that show up as a few overlapping letters. [1] https://en.wikipedia.org/wiki/Bouba/kiki_effect https://en.wikipedia.org/wiki/Bouba/kiki_effect