Wednesday, August 12, 2009

How does language shape the way we think?

A more hypothesis-free question might be "Does language shape the way we think?"

This question, in fact, goes back a long way. It can be traced to the European romantic period of the early 19th century, with antecedents that are much earlier.

But a more modern form of the question is well-known to many as the Whorfian Hypothesis or the Sapir-Whorf Hypothesis. In even more modern terminology, this is the question of linguistic relativity. (That Wikipedia article gives a fairly detailed summary.)

There's a recent essay at Edge by neuroscientist Lera Boroditsky that outlines a number of the issues and gives an entertaining account of simple experiments that have been performed to test various linguistic relativity hypotheses.

How does language shape the way we think?
Humans communicate with one another using a dazzling array of languages, each differing from the next in innumerable ways. Do the languages we speak shape the way we see the world, the way we think, and the way we live our lives? Do people who speak different languages think differently simply because they speak different languages? Does learning new languages change the way you think? Do polyglots think differently when speaking different languages?

These questions touch on nearly all of the major controversies in the study of mind. They have engaged scores of philosophers, anthropologists, linguists, and psychologists, and they have important implications for politics, law, and religion. Yet despite nearly constant attention and debate, very little empirical work was done on these questions until recently. For a long time, the idea that language might shape thought was considered at best untestable and more often simply wrong. Research in my labs at Stanford University and at MIT has helped reopen this question. We have collected data around the world: from China, Greece, Chile, Indonesia, Russia, and Aboriginal Australia. What we have learned is that people who speak different languages do indeed think differently and that even flukes of grammar can profoundly affect how we see the world.

Let's first review the history of linguistic relativity. In romantic thought during the early 19th century in Germany language was seen as expressing the "spirit" of a nation. Implicitly, then, there was a close connection between language and the social and political attitudes of a place. However, the direction of a causal link, if any, was not clearly articulated. If anything, one could infer that society and culture affect language more than the reverse.

This might have been true before the invention of writing, but since then, language generally changes much more slowly than social attitudes. Instead of following culture, one might expect language to be one of the ways that attitudes are transmitted from generation to generation. So the natural expectation now is for language to causally affect society.

Some of the assertions in Ludwig Wittgenstein's Tractatus Logico-Philosophicus (TLP) tend to support the notion that language strongly affects thought. TLP was published in 1921 and belongs to the "early period" of Wittgenstein's thought. In his later period, represented in the posthumously published Philosophical Investigations (1953), he reversed course on a number of issues. Yet many people still regard much of TLP as powerfully insightful.

The basic thrust is that one's language places limits on what can usefully be thought, especially: "The limits of my language mean the limits of my world." (TLP, 5.6) And the concluding assertion of the book: "Whereof one cannot speak, thereof one must be silent." (TLP, 7) One interpretation is that although one might have certain thoughts one can't express in language, such thoughts would be largely futile, since they could not be communicated to others (at least not verbally) or even to oneself at a later time – perhaps only hours or even minutes later. At all events, limitations on the expressive power of a particular language place constraints on useful thinking. (To a large extent, this should be an empirical question about how much the brain uses a natural language to help with long-term storage of abstract thoughts. It seems very plausible – we need only recall a short aphorism like Wittgenstein's to access the more complex ideas. In fact, there's now a common term for this: "sound bites".)

A century after the German romantics, Edward Sapir (1884-1939), an American anthropologist and linguist, began to put the study of the relationship between language and thought on a more systematic and scientific basis. Sapir's linguistic studies adopted an anthropological approach (under the influence of Franz Boas), including field work exploring native American languages.

Although Sapir's connection with the question of a relationship between language and thought becomes more explicit in the work and writings of his student, Benjamin Lee Whorf, Sapir himself did much to encourage examination of the relationship. His introductory book on linguistics, Language: An Introduction to the Study of Speech is still in print. In Chapter X of it he wrote "Nor can I believe that culture and language are in any true sense causally related." But nevertheless, "Language and our thought-grooves are inextricably interrelated, are, in a sense, one and the same."

Benjamin Lee Whorf (1897-1941) was much more closely identified with the question, so much so that the idea of an influence of language on thought is often called simply the Whorfian Hypothesis. His collection of writings, still in print, Language, Thought, and Reality, is the work for which he is best known. Representative quote:
We dissect nature along lines laid down by our native language. The categories and types that we isolate from the world of phenomena we do not find there because they stare every observer in the face; on the contrary, the world is presented in a kaleidoscope flux of impressions which has to be organized by our minds—and this means largely by the linguistic systems of our minds.

For some time more recently, especially in the 1980s, the Whorfian Hypothesis (which wasn't actually stated explicitly by Whorf) and the idea of linguistic relativity in general was not well regarded by cognitive scientists, because it lacked precision, testable hypotheses, and (consequently) experimental evidence. It was, after all, largely speculations based on anthropological observations. Lack of experimental evidence made a skeptical view of linguistic relativity easier to hold. But since the 1990s many attempts have been made to study the question scientifically and experimentally.

Although Boroditsky's essay could be construed as implying that most of the research has been done in her laboratory, in fact it has been undertaken rather more widely. However, we won't attempt to survey the other work right now. Let's just recap some of the research findings she discusses.

One interesting group of studies is anthropological in origin, like the original investigations of Sapir and Whorf. It involves how the languages of certain Australian aborigines, including but not limited to a group known as the Kuuk Thaayorre, describe 2-dimensional space. In English, and indeed in most languages of which English speakers have ever heard, positions in 2-dimensional space, in which the observer is situated, are usually described relative to the observer: left vs. right, in front of vs. behind. But for the aborigines, 2-dimensional space in which they are located is described in terms of absolute directions: north, south, east, west. Such a system has obvious advantages for navigation in possibly unfamiliar territory, by forcing the observer to remain oriented to fixed directions. It's also natural for people who spend a lot of time outdoors. This anthropological fact has been known for some time.

Many languages can apply spatial concepts to non-spatial information as well, particularly in the special-case of 1-dimensional data of many types. Examples include size, weight, age, time ("earlier" vs. "later"), moral values ("better" vs. "worse"), kinship ("near" vs. "far"), musical pitch ("high" vs. "low"). (It's now understood, of course, that pitch can be quantified as vibrational frequency, but this is a modern development.)

Each of these examples either involves things that can be quantified using numbers (size, weight, age), or if that's not possible there are terms specialized for the particular concept. Quantifiable concepts generally also have their own special terms too, which indicate magnitude in a vaguer way, when numeric values aren't convenient ("big" or "small", "heavy" or "light", "old" or "young").

In the case of quantifiable concepts it's not unusual to think of attribute values spatially laid out on a line from smaller to bigger (though the actual direction in space may vary – left to right or right to left, for example). Some concepts that aren't obviously quantifiable also customarily involve spatial metaphors (kinship, pitch).

Other concepts more often are described with non-spatial terms, but sometimes admit spatial metaphors. For instance, a particular time may be "near" or "distant", like kinship. Moral values may be "higher" or "lower", like pitch.

Obviously, languages other than English may customarily use different terms and metaphors to describe the same concepts. Every language seems to have its own way of conceptualizing and describing certain things, even with quantifiable concepts. Consider time, for example. Sometimes it's represented in non-spatial terms ("earlier" vs. "later"), but spatial metaphors are also common ("before" vs. "after"). When thinking spatially of time, English speakers tend to think of it in a horizontal line, perpendicular to the observer's line of sight, from earlier to later (and usually left to right). Occasionally time is described, like kinship, in a line that runs outward from the observer. However, Mandarin arranges time in a vertical direction – the way English arranges musical pitch.

(There are obviously even more complexities in all this if similar concepts customarily used in differing languages aren't quite congruent or coextensive to begin with. For instance, the concept "moral values" is one thing if such values are customarily believed in one culture to be supernaturally ordained, but something rather different if they are thought to be products of a rational ethics.)

In mathematics there is a uniform way of dealing with such 1-dimensional concepts. This is especially true when the concepts are quantifiable, but it also true even when there is only a greater-than-less-than relationship between specific attributes of different individuals. This kind of relationship of attribute values is called a linear ordering. However, most people, and most languages, don't express things in abstract mathematical terms. And so the actual language used for different types of linearly orderable attribute values can vary dramatically from one language to another.

With all these ways of using metaphors and terminology, which vary from language to language, for describing concepts, one can't help but wonder whether the choice of terminology or metaphor in a particular language affects how a speaker of the language thinks about a concept. For instance, does use of a spatial metaphor for moral values ("higher" or "lower") predispose a speaker to think differently about the concept than a different metaphor ("conservative" or "liberal", say)? If we are comparing different languages, do the different customary metaphors employed lead to observable differences between cultures where one or the other language predominates?

So the interesting general scientific question that arises from this situation is whether the peculiarities of how a particular language describes certain types of attributes have an effect on how the speakers of the language think about the attributes. More specifically, is it possible to set up experiments that demonstrate consistently different behaviors by speakers of different languages, where the behaviors make sense in terms of how each language typically represents the attributes?

When, as is quite common, a single language offers choices of metaphors or terminology for describing a given concept, it should be easy to set up experiments to detect different ways people think about a concept (as determined by observable behavior) according to selections of metaphor or terminology presented to an experimental subject. Or better yet, can we measure how much or how likely use of a specific metaphor affects specific behavior with respect to a given concept under given conditions?

This kind of experiment, that uses only speakers of a single language, can separate out effects due to language (as expressed in metaphor or terminology) from effects due to underlying culture, as long as experimental subjects are drawn from the same culture or classified by culture (or specific demographic or ethnic groups, for example). This is a way to deal with the questions that always arise about whether it's the language or the culture in which the language is spoken that accounts for observed differences. A similar option, which Boroditsky mentions, is to deliberately train experimental subjects to use unfamiliar metaphors or terminology drawn from languages other than the subject's native language.

Boroditsky cites a few examples of experiments that have been performed to study such questions. In addition to considering languages that use different spatial metaphors (either 1- or 2-dimentsional), she mentions experiments involving other types of language differences. One example is whether and how "gender" is encoded in the grammar of a language. English is a little unusual in not having much grammatical role for gender (except for pronouns). In Russian, however, most nouns have one of three genders (masculine, feminine, neuter). In a Russian sentence, adjectives, pronouns, and even verbs that refer to a noun must agree in gender.

As Boroditsky notes, "some Australian Aboriginal languages have up to sixteen genders, including classes of hunting weapons, canines, things that are shiny." Although English and other Indo-european languages seem to rely on sex (i. e. genitals – same linguistic root as "gender") as the metaphor for gender, in other languages gender is simply about whatever the language, or the underlying culture, considers to be the most important categories for partitioning the world of discourse. Obviously, this can lead to huge differences in how people who use different languages think about the world.

Or consider the way languages differ in how they divide up a space of perceptual qualia – color in particular. Russian, for instance, recognizes two different types of blue (varying in the brightness dimension), using different words. This can lead to observable differences in behavior between Russian and English speakers. (Of course, English also recognizes multiple types of blue, such as azure and blue-green. It may simply use hyphenated and compound words to denote many of the different types. There may be thousands of English terms. What's distinctive about Russian is that it doesn't have a single word that is applicable to something that's normally just "blue" in English.)

It would be very interesting to study linguistic effects with respect to other types of qualia, such as taste. Taste is especially interesting, since there are physiological reasons that, cross-culturally, five different "flavors" are commonly recognized: "sweet", "salty", "sour", "bitter", and "umami". The result is a 5-dimensional perceptual space, instead of the 1-dimensional space of loudness or hue. For that matter, color space is actually 3-dimensional (hue, saturation, brightness), and sound has various dimensions too (loudness, pitch, timbre, etc.)

All this raises many interesting questions. How much do languages vary in the way they conceptualize perceptual spaces? Even if spatial dimensions are the same between two languages, are the spatial axes oriented differently? And what effects do such differences have on how people think about the perceptual spaces?

Interestingly, even in the case of color, the strength of a linguistic effect on perception of hue (e. g. 1 vs. 2 types of blue) seems to depend on how much the experiment calls for cognitive processing in the left vs. the right brain hemisphere, presumably because of the special role of the left hemisphere in language processing.

And then, supposing some correlation exists between how languages represent certain attributes and how people behave in thinking about the attributes, how can it be decided whether the language has affected the thinking, as opposed to both language and thought being affected by cultural history and circumstances (such as a lifestyle that places high importance in being oriented to directions of the compass). Boroditsky suggests experiments that can be done, such as deliberately teaching people different ways of describing things, to find out whether and how this affects thought and behavior.

But let's begin to wrap this up by returning briefly to Boroditsky's observations of the Kuuk Thaayorre. There was a specific reason she wanted to study these people. As noted, it was known that they tended to think of (2-dimensional) spatial information using coordinates independent of the observer. That's also true in thinking about 1-dimensional data, which speakers of English (and most other language) think of as left vs. right, relative to the observer.

Boroditsky wanted to know how the Kuuk Thaayorre conceptualized and described other 1-dimensional linear orderings, such as time, where spatial metaphors are less likely to be used in other languages. The answer is that the Kuuk Thaayorre continued to use the same absolute coordinate system for such things. For example, given pictures illustrating temporal progressions, they would tend to arrange them in a line in the east-west direction. So not only did they stick to absolute coordinates, but they chose a direction that just happens to be the same as the Sun's motion through the sky – as opposed to north-south, say. Of course, that might be due to dealing with temporal information.

One wonders whether they would do the same with other linearly ordered data, such as taste of food (from "unpleasant" to "delicious"). One also wonders what the outcome of such experiments would be if they were done with Kuuk Thaayorre people indoors, under conditions where they would be unaware of actual compass directions.

Interestingly, speakers of English and other languages that are written in a left-right direction, tend to arrange pictures illustrating temporal progression in the same left-right direction, while speakers of languages written from right to left tend to use this right-left ordering for 1-dimensional data. Right there you have an apparent effect of language on thought patterns and behavior.

Concluding thoughts

This is a vast subject. It's not at all easy to summarize even the discussion presented here. So I won't try. I'll just add some general remarks.

It has generally been assumed that if there is any causal relationship between language and thought it should be in the direction of language affecting thought instead of vice versa. This is probably because language is assumed to have arisen without conscious design, and because it's difficult to imagine what relatively complex thought would be like if it came before language. We assume that non-human animals, which lack language, can't have complex thoughts. But we know almost nothing about the origins of human language, at least tens of thousands of years ago, or what human thinking might have been like at the time. Our assumptions could be wrong. Language and complex thought probably co-evolved to some extent that we can now only guess about.

For another thing, it should be obvious that the effect of language doesn't have to be all or nothing. The Whorfian Hypothesis isn't necessarily generally true or generally false. Instead, what needs to be investigated are tendencies and correlations under many different special conditions, such as the kind of cognitive activity one suspects might be affected. One should look for the strength (i. e. probability) of an effect, and its pervasiveness (i. e. range of circumstances in which it can be observed).

Another important consideration is that language has a variety of parts, with vocabulary and grammar being the major divisions. The vocabulary and grammar of a language probably differ a lot in whatever way they affect thinking.

As far as vocabulary is concerned, if a language lacks a word for a specific concept, it's considerably more difficult to think about the concept, though generally not impossible. Without a word for a concept, it's harder to keep it in mind in order to examine it logically in order to understand it, or to use it for classifying or organizing things. For example, a study of an Amazonian tribe, the Piraha, who don't have words for numbers, suggests that these people have difficulty with numerical concepts in tasks where memory is involved, even though they presumably can recognize differences in size of small collections of things. (See below for more about this study.)

Likewise, suppose we didn't have words for many types of mammals, how easily could we distinguish between a fox and a coyote, say? Whether we regard two things as the "same" or "different" depends a lot on whether we have separate, common words for the two things. (For example: different shades of the same color.) Even if we could visually perceive differences between the species, would we remember the differences long enough to keep track of other ways the species differed?

So languages with sparse vocabularies make many kinds of thinking difficult. In language without words for abstract concepts like "liberty", "freedom", "justice", etc. it would be difficult to develop much of a political philosophy. Indeed, the introduction of such terms is often the starting point for new directions in philosophy. All this argues for a significant effect of language on thought, mediated by the language's vocabulary.

But vocabulary is a rather different aspect of language from grammar, and it is the grammar of a language that is often the focus of questions about how language affects thought. For instance, there are all the questions about the role of grammatical gender. Other important grammatical issues have to do with verbs – the way in which they involve time ("tense"), attitude towards the action ("mood"), whether action is ongoing or completed ("aspect"), issues of effect of an action ("transitivity") and so forth. A language can potentially pack quite a variety of disparate kinds of information in its verbs. Each of these aspects of grammar may affect thinking in different ways and different degrees. Research should be looking for whether certain aspects tend to be more influential than others.

Another interesting difference between vocabulary and grammar is that the former probably changes more quickly than the latter. Indeed, a language's grammar tends to change pretty slowly, and in small increments, if at all, over hundreds of years. But vocabulary can change in a matter of decades, or less. Think of the difference in English vocabulary between Shakespeare's time and ours. (With Shakespeare himself responsible for quite a few innovations.) Yet the grammar of Shakespearean English is hardly different from that of contemporary English.

What this may mean is that the grammar of a language has a more profound effect than vocabulary on thinking because of its relative constancy. Consequently, perhaps, speakers of the language can hardly imagine that distinctions imposed by grammar might admit any possible alternative kinds of distinctions. Just as speakers of English have a difficult time imagining any relevance of gender to non-living things.

Lastly (for now), I should point out the relevance of this discussion to the even more general topic of "social construction of reality", as briefly touched on here.

In that regard, Edward Sapir wrote: "No two languages are ever sufficiently similar to be considered as representing the same social reality."

A few recent relevant research reports

Language Without Numbers: Amazonian Tribe Has No Word To Express 'One,' Other Numbers (7/15/08)
An Amazonian tribe, the Piraha, with only 300 speakers of its language, apparently has no words for any natural number, even "one", though it has terms for comparing quantities ("some", "more"). Consequently the size of a very small collection (1 to 4 things) would be described with a word meaning "few", while larger collections would be "many".

Aboriginal Kids Can Count Without Numbers (8/18/08)
In contrast to the preceding report, a study of children in two Australian aboriginal communities, whose language does not contain words for numbers, showed that the children were able to copy and perform number-related tasks.

Pre-verbal Number Sense Common To Monkeys, Babies, College Kids (2/15/09)
As in a number of other studies, this research shows that human infants, as well as macaque monkeys, show an awareness of the different size of very small sets (2 or 3 objects), even though they do not have language. In fact, even dogs have a similar awareness, using similar testing protocols. (See here.) So certainly the Piraha also have such awareness. But as is pointed out in the Piraha study, lack of words for numbers seems to significantly affect the ability to perform tasks where memory is involved.

Scientists show that language shapes perception (2/26/09)
Modern Greek, like Russian, has different words for light and dark blue, unlike English. Greek speakers automatically use different words to describe the color of the sky or the color of the ink of a (dark) blue pen. Using a measuring technique called "event related brain potentials", differences in brain activity caused by seeing different shades of blue can be detected before the appropriate color word enters consciousness. However, this doesn't necessarily mean that the brains of Greek speakers and English speakers (say) are genetically different – only, perhaps, that the cultural habit of sharply distinguishing two kinds of blue has been learned in non-verbal, as well as verbal, parts of the brain.

Expressing comparisons is possible even without language, researchers find (6/30/09)
Deaf children who had not learned a spoken language were compared with hearing children who had learned some language. Deaf children are able to use gestures to indicate a recognition that (for example) a house cat is different from a tiger. However, "as the children grew older, the comparisons diverged — deaf children's comparisons remained broader in nature, while hearing children's comparisons became more complex and focused, such as saying that the hair was brown like a brown crayon, after hearing children started using the word 'like'."


Update, 8/24/09: Here's a relevant new article I haven't had a chance to evaluate yet: Does Language Shape What We Think?

Update, 9/4/09: And here's an interesting quote from Steven Pinker (via Amira):
Human beings do not live in the objective world alone, nor alone in the world of social activity as ordinarily understood, but are very much at the mercy of the particular language which has become the medium of expression for their society. It is quite an illusion to imagine that one adjusts to reality essentially without the use of language and that language is merely an incidental means of solving specific problems of communication or reflection.

The fact of the matter is that the ‘real world’ is to a large extent unconsciously built up on the language habits of the group … We see and hear and otherwise experience very largely as we do because the language habits of our community predispose certain choices of interpretation.


Tags: , , social construction of reality

Labels: , , ,

0 Comments:

Post a Comment

<< Home