Does our language affect our thoughts?

One of the oldest questions in the study of language involves how it influences our thought. One of the most controversial answers comes from Benjamin Whorf, the student of renowned anthropologist Edward Sapir: language not only influences thought; language determines thought—thought cannot exist without language. The Sapir-Whorf hypothesis, at least in its strongest form, has been discarded by mainstream psychologists. After all, it's not difficult to come up with many examples of thought that do not involve language, such as mentally rotating an object or learning how to juggle (think about it: by the time you verbalized the tiny adjustments necessary to juggle successfully, the floor would be littered with juggling balls). But a weaker form of the hypothesis has yet to be disproved: the idea that the available linguistic expression does to a certain extent constrain our thoughts.

Hanako Yoshida and Linda Smith of Indiana University devised a clever experiment to see if they could identify a concrete instance of language influencing thought ("Linguistic Cues Enhance Learning of Perceptual Cues," Psychological Science, 2005). The experiment relied on a linguistic feature of the Japanese language: the fact that articles (like the word a in "a dog") don't necessarily give any cues as to the type of noun they are modifying. For example, in English, we don't ever say "a sand." We might say "the sand" or "some sand." These words offer us a clue as to the type of thing we're talking about. By saying "some sand," we let readers know that this is an object that we classify based on the material it's made of. We never say "some table," because tables are classified by shape. When we encounter a new word ("I have some sploot"; "this is a bive"), the article a or some instantly lets us know whether to think of it as a material or a shape. In Japanese, by contrast, it's possible to refer to "sand" without indicating that it's a thing we typically categorize based on the material it's made out of.

There are Japanese words that make this distinction (hitotsu and sukoshi, corresponding to a and some respectively); it's just that they are not required for proper communication—and indeed, they are rarely used in everyday conversation. Yoshida and Smith studied a group of 2-year-old Japanese children who had not yet learned the words hitotsu and sukoshi. The experimenters played with the children for ten 30-minute sessions where they introduced several new objects with invented names. Some of these objects were solid and so would in English normally be accompanied by the article a. Others were non-solid and would normally be accompanied by some. The experimenters took care to name each object at least 20 times per session, but with half the children, no accompanying article was used. With the other half, the solid objects were consistently referred to with the article hitotsu no, and the non-solid objects were consistently referred to with the article sukoshi no.

After the "training" sessions, the children were tested in two different ways. In one test, they were shown a new object (with a new, invented name) that either fit into the "solid" or "non-solid" category, and asked to choose another one:

The diagram is a schematic representation of the test: in the real test, actual physical objects were used. Remember, in Japanese, the lack of an article would in no way prejudice the response.

Next, they were given the same test, but the articles—a linguistic cue to the type of object they were being shown—were now included:

For solid objects, both sets of children performed equally well on the test. This corresponds to a large body of research indicating 2-year olds can classify objects by shape. However, for nonsolid objects, the kids who had been trained using linguistic cues performed nearly three times better on the test when the linguistic cues were given. The children who had never learned the word sukoshi to indicate that all objects with this article should be classified by the material they were made of, seemed simply unable to classify objects by material.

Yoshido and Smith have demonstrated that in this instance, language does appear to influence thought. I should emphasize, however, that this is still a very limited example of the Sapir-Whorf hypothesis. All Japanese children do eventually learn how to classify objects based on the materials they are made of, even though this classification is not required by the Japanese language. And think about it—wouldn't the world be a bizarre place if Japanese people were never able to learn to classify objects by material? This thought experiment alone should be enough to convince most skeptics that a strict interpretation of the Sapir-Whorf hypothesis would result in a society scarcely able to function. Worst of all, it would be a world without juggling, which would be a very sad place indeed.