The story of research on linguistic relativity can be summarized thusly: early cognitive scientists, inspired by the work of Edward Sapir and Benjamin Whorf, were all-too eager to find that thought is influenced, if not determined, by language (either by its grammatical categories, ala Whorf, or by the words we use). Their enthusiasm caused them to get greedy, and instead of starting simple, they went straight to perceptual properties that are sorted out very early in visual perception. If the perception of such properties could be shown to have top-down influences (like linguistic influences), then linguistic relativity/determinism would be on very solid empirical ground. Thus, much if not most of the early research (in the 1950s and 60s) focused on color categories, with researchers arguing that differences in color terms across languages led to differences in the perception of color among speakers of different languages. In the late 1960s and early 1970s, this enthusiasm was severely dampened, particularly by the work of E.R. Heider, who presented studies purporting to demonstrate that two very different languages -- that of the Dugum Dani in New Guinea, with 2 basic color terms, and English, with 11 -- show roughly identical patterns of color memory, and therefore (or so the reasoning goes) color perception1. If such extreme differences in language (2 vs 11 basic color categories) didn't produce differences in perception, it's unlikely any linguistic differences would.
With the rug taken from under it by Heider, linguistic relativity remained virtually dormant until the mid-1990s, when some cognitive psychologists realized that starting with perceptual categories probably wasn't a good idea. They decided it would have been wiser to start with more cognitive categories, like time and space, number, gender, and other minds. Since these sorts of things are already processed at the top (i.e., conceptually rather than perceptually), top-down influences like language shouldn't have to work very hard to affect them. And sure enough, it was pretty easy to demonstrate the influence of language on such categories. Studies were conducted, books were published, and linguistic relativity was back, baby. With linguistic relativity's newfound respectability, it was only a matter of time until someone got greedy and went back to color perception. But this time, armed with more sophisticated research techniques, and relatively cooler heads, researchers like Debi Roberson2 were able to find fairly strong evidence for differences in color perception and memory (contra Heider!) related to differences in color terms.
There's a problem inherent in comparing very different languages and cultures like those of Western Europe and the U.S., and the Dugum Dani, Berinmo (also of New Guinea), and Himba (of Namibia) tribes (all three were studied by Roberson). Since both the languages and the cultures are very different, it's very difficult to sort out the influence of the languages from those of the cultures. And since the Dani, Berinmo, and Himba's cultures are so different, it's also very difficult to use complicated research methods in studying them. So, while we were left with tantalizing evidence of the influence of color terms on color perception, we weren't able to say for sure that it was the language, and not the culture, that was really doing the influencing.
I bring all this up because a paper in last week's issue of the Proceedings of the National Academy of Sciences takes the first steps in addressing this culture v. language problem, and thus the next steps in linguistic relativity research. The paper, by Jonathan Winawer and 62,000 other people3, starts by using languages and cultures -- Russian and English -- that while pretty different, aren't quite as radically different as, say, Berinmo and English. There aren't many differences between English and Russian basic color terms, but there is one important one. In English, we refer to pretty much any shade of blue, be it light or dark, as "blue." In Russian, however, there is an "obligatory distinction" between light blues, which are called "goluboy"(Голубой), and darker blues, which are called "siniy" (синий). Utilizing this small difference, Winawer et al. are able to test for differences in color perception between Russian and English speakers.
The first part of their experiment is simple. Using a range of blues from light to dark (the goluboy to siniy range; see the figure below, Winawer et al.'s Figure 1, p. 7781 ), they presented each participant with three color squares, one at the top and two beneath it (as in the figure).
Participants were tasked with deciding which of the two bottom squares was the same color as the top square. On some trials, the two squares on the bottom would both be either from the goluboy or siniy side of the range (within-category trials) of colors, while on others, they would be from different sides (between-category trials). For Russian speakers, this difference should affect performance: since their language makes a distinction between the two types of blue, they should be better at discriminating shades if they come from different categories (goluboy and siniy). This should lead to better matching performance in the between-category trials, for Russian speakers. English speakers, on the other hand, don't make a category distinction between the two types of blue, and therefore should perform about equally well on within and between-category trials, controlling for color distance.
If that was all there was to the experiment, it would be interesting, but it wouldn't tell us anything more than the studies with the Berinmo, Himba, etc. already have: that differences in color perception are related to (but not necessarily caused by!) differences in color terminology. But Winawer et al. added a twist to their experiment, to make it really interesting. They hypothesized that linguistic influences on perception may occur online (that is, while people are processing the information). If so, then interfering with that influence while people are processing the color information should reduce or eliminate the effect of color terms on color perception. To test this possibility, they divided their participants into three conditions. In the first condition, participants just completed the task as described above: they were given three squares, and told to match one from the bottom pair to the target square on top (control condition). In the second condition, participants were asked to keep a spatial figure in memory while completing the task (spatial-interference condition). Since spatial and verbal working memory have been shown to be mostly independent, the spatial memory task should not interfere with the influence of language on color perception. Finally, in the verbal-interference condition, participants silently recited digit strings while performing the color task. Since reciting digit strings takes up verbal working memory capacity, this task should interfere with any linguistic influence on color perception if that influence occurs online.
Just to be clear, the predictions for this experiment are as follows: Russian speakers should do better (as measured by reaction times) when the two comparison colors (the ones on the bottom) come from different linguistic categories (goluboy and siniy) than when they come from the same category (goluboy or siniy). English speakers' performance, on the other hand, should simply be a function of color distance: the more the two shades differ from each other, the better English speakers should do in the matching task. Furthermore, the verbal-interference task (but not the spatial-interference task) should reduce this pattern for Russian speakers, but not for English speakers, because the latter aren't influenced by linguistic categories (all of the shades are just "blue"). Here are the results (Figure 2, p. 7782):
As you can see, Russians were able to give the correct matches faster in the between-category trials than the within-category trials in the control and spatial-interference conditions, but not in the verbal-interference condition. In fact, in the verbal-interference condition, they were actually faster in the within-category trials than the between-category trials, though this difference only approached statistical significance. English-speakers, on the other hand, performed the same on within and between-category trials in all three conditions.
What makes this experiment so interesting, then, is that it provides the first evidence that language, independent of other aspects of culture, is directly influencing color perception. It does so by showing that that influence is occurring online, and can thus be reduced or eliminated (reversed, even) by interfering with that influence using a verbal working memory manipulation. Obviously, this is just the first step in conclusively showing that language, independent of other cultural influences, can affect perception, and how it does so, but it's a much needed first step. For anyone who's interested (or mildly obsessed, as I am) in linguistic relativity, this is therefore a very exciting study.
1Heider, E.R. (1972). Universals in color naming and memory. Journal of Experimental Psychology, 93, 10-20.
2E.g., Roberson, D., Davidoff, J., Davies, I.R.L. & Shapiro, L. R. (2004) The Development of Color Categories in Two languages: a longitudinal study. Journal of Experimental Psychology: General, 133, 554-571.
3Winawer, J., Witthoft, N., Frank, M., Wu, L., Wade, A., and Boroditsky, L. (2007). Russian blues reveal effects of language on color discrimination. Proceedings of the National Academy of Sciences, 104(19), 7780-7785.
Nice study. Your readers may be interested to know that E. R. Heider is better known as Eleanor Rosch.
Count me in as another soul who is mildly obsessed with linguistic relativity. Thanks for sharing this.
Color naming reflects optimal partitions of
Terry Regiera,b, Paul Kayb,c,d, and Naveen Khetarpala
Thanks Abel. I read that a while back, and thought about posting on it. It's a bit complicated, though, so I passed.
It doesn't appear that any of these experiments take pains to distinguish between "noumenal" and "phenomenal" events when talking about perception, memory, and cognition generally. That is, at some point between the reception of sense data* and the recognition or categorization, a change occurs in cognition from passive reception to active conceptualization. On which side of this line does "perception" occur? Where does pure perception become cognition?
I wonder if what is being measured is actually much further downstream than anything that can rightly be called "perception."
(* even the idea of passively receiving sense data is debateable, not least because we choose what to experience, and, if Schroedinger and Heisenberg are correct, perhaps even partially determine what we experience).
"I read that a while back, and thought about posting on it. It's a bit complicated, though, so I passed."
Damn. I was hoping you could explain it to me.
Chris S., in cognitive psychology, "perception" is generally used to refer to what occurs during and after the processing of sense data. The reception of sense data is called sensation. Also, cognitive psychologists tend to refer to the representation of perceptual properties (e.g., color) as perception, and the classification of those properties as perceptual categorization. I'm not sure any of this has anything to do with a distinction between the noumenal and phenomenal, of course, but perhaps you're using those in a way that I'm not familiar with.
Oh, and Abel, maybe I'll give it a go in a week or to. By then I should have a little extra time to blog, and I can sit down and figure out how best to describe that paper.
If I may throw in my 2c. There are three major issue with this study which is actually about 5 years old at this point. These issues are the main reasons why it took so long to push it through to publish in PNAS.
1) there is NO "obligatory difference" between light- and deep-blue in Russian. Any proper Russian linguist will tell you that.
2) the first part of this study has not been replicated in a different lab. In fact, John Trueswell's lab at UPenn failed to replicate the results in something like 16 different versions of the expt...
3) I'm not sure that the effect actually statistically reverses for Russians in the case of verbal interference. And if it doesn't, the whole thing should be attributed to general processing difficulty (Russian being a "harder" language than English) rather than to linguistic basis - in the case of verbal interference all reaction is slower resulting in ceiling effects...
So, don't believe everything you see or read. ;-)