Color categories make the world easier to live in. Granny Smith (green) and Red Delicious (red) apples belong in different bins; so do violets (blue) and roses (red).

To most of us, those categories seem natural, but in many other languages the categories differ. Some African languages have five primary color words or fewer; Russian has as many as English, plus an additional kind of blue. Often the boundaries between two colors shift as one moves from one language community to another.

Now, a new study by researchers at the University of Surrey suggests that the process of learning new color categories produces subtle but significant changes in how people actually perceive those colors.

The findings, published in the Journal of Experimental Psychology: General (Vol. 131, No. 4), support the linguistic relativity hypothesis--the idea that the language one speaks can affect the way one thinks about and perceives the world.

"The main conclusions of the study are basic: that color perception is not as rigid and inflexible as was thought before," says the study's lead author, Emre Ozgen, PhD. "This is the first time that it's been shown that a new perceptual color category boundary can actually be induced through laboratory training."

The experiment

Previous studies have shown that people find it easier to distinguish between similar hues that belong to different color categories than between hues that fall within a single color category. A bluish green and a greenish blue, for example, are easier to tell apart than a bluish green and a yellowish green.

The central question of the current study was whether these improvements in performance at the boundaries of color categories--an effect known as "categorical perception"--are fixed or changeable. Can training enhance the effect, making people more sensitive to color differences across boundaries? Can new boundaries be created, even ones that lie right in the middle of conventional color categories?

In their first experiment, Ozgen and Ian Davies, PhD, sought to answer the more basic question: whether training could improve participants' ability to distinguish between similar hues of a single color. The answer was yes: Participants became increasingly accurate over the course of three days of training.

Ozgen and Davies then moved to the second, critical question: whether novel categorical perception effects could be acquired in the laboratory. Participants were trained to divide a basic color category, blue or green, into two new categories. The boundaries of the new categories lay at the focal points of the old categories--the greenest greens, for example, now lay at the boundary between a category of yellowish greens and a category of bluish greens.

After three days of training, participants were better able to distinguish between hues that fell on either side of the novel color boundaries than between hues within a single category, even when the absolute difference between the two hues was the same--a classic categorical perception effect. At the same time, participants effectively unlearned their pre-existing color categories: They stopped showing categorical perception effects at "natural" boundaries that lay within the range of hues on which they had trained.

A follow-up experiment showed that the change in categorical perception could be produced after a single training session of 500 trials, though the improved performance was not evident until the next day--perhaps, the researchers speculate, because improvement at the end of the first session was masked by fatigue.

In their final experiment, Ozgen and Davies explored whether participants would learn new categories based on differences in lightness, just as participants in the previous experiments had learned categories based on differences in hue. They found that lightness training produced the same kinds of categorical perception effects that hue training did, but they also found an interesting asymmetry: Participants who trained on hue-based categories showed no new categorical perception effects when later tested on lightness, whereas participants trained on lightness-based categories showed categorical perception effects for hue. The results suggest that lightness can be ignored when it is irrelevant to a task, but hue is processed automatically.

"People have shown categorical perception effects for color before, but there's been a large number of people who've argued that these are innate color categories," says Robert Goldstone, PhD, a professor of psychology at Indiana University. "What the current study shows is that you can get acquired categorical perception."

Linguistic relativity

The study also offers a new spin on the work of University of California, Berkeley, psychologist Eleanor Rosch, PhD, who suggested in the 1970s that linguistic differences have little effect on how people actually perceive colors. She reported that speakers of a New Guinean language with only two color words--one for cold, dark colors and one for warm, light colors--learned and remembered colors more easily when they were prototypical examples of colors identified in English, such as red, green and blue, than when they were not. The results were a blow to the linguistic relativity hypothesis, also known as the Sapir-Whorf hypothesis, which suggests that cognition is shaped by the languages people speak.

In recent years, however, University of Essex psychologist Debi Roberson, PhD, and others have tried to replicate Rosch's work among other tribes with few color words, and have found results that appear to contradict hers. Their findings suggest that there are differences--small but nonetheless significant--in the color perception of speakers of different languages.

"These kinds of categorical perception effects seem to be language-dependent," says Davies, who has collaborated with Roberson on some of those studies. "If an African language doesn't mark a blue-green boundary, then adult speakers don't seem to show categorical perception across that boundary, whereas English speakers do."

The current study is partly motivated by the cross-cultural research. But as MIT psychologist Lera Boroditsky, PhD, points out, unlike the cross-cultural studies, it does not directly address the linguistic relativity hypothesis. It does not, for instance, provide evidence that learning a new linguistic distinction can produce a new categorical perception effect.

What the study does provide, says Ozgen, is evidence that categorical perception can change quickly as well as a plausible mechanism for how it changes. It is a small leap from there to being able to show how a distinction that starts off as merely linguistic--this sort of color goes in category A; that sort goes in category B--can become deeply ingrained in perception.

"Linguistic relativity may work with similar principles, in that as a child grows up, he or she will have to continually learn a category boundary just as our subjects learn in the lab," says Ozgen. But he cautions, "neither the linguistic relativity nor the universal hypotheses would hold if we were to take an extreme position. To say that language completely shapes thought would be as extreme as saying that thought is entirely hardwired."