Thursday, August 19, 2004

Psychometrics and the Sapir-Whorf Hypothesis

Nature is reporting on the existence of a South American tribe whose rudimentary counting system seems to have placed limits on the members' ability to deal with even moderately large numbers.

A study of an Amazonian tribe is stoking fierce debate about whether people can count without numbers.
Psychologists, anthropologists and linguists have long wondered whether animals, young children or certain cultures can conceptualize numbers without the language to describe them.
To tackle the issue, behavioural researcher Peter Gordon of Columbia University in New York journeyed into the Amazon. He carried out studies with the Pirahã tribe, a hunter-gatherer group of about 200 people, whose counting system consists of words which mean, approximately, 'one', 'two' and 'many'.
Gordon designed a series of tasks to examine whether tribe members could precisely count and conceive of numbers beyond one or two, even if they lacked the words. For example, he asked them to look at a group of batteries and line up a matching amount.
The tribe members struggled to perform these tasks accurately after the numbers were greater than three, Gordon reports in Science1; and their performance got worse the higher the numbers climbed. "They couldn't keep track at all," he says.

Opposing views

Other researchers in the field have welcomed the study. But they disagree about what it means. Psychologist Charles Gallistel, at Rutgers University in Piscataway, New Jersey, says that the Pirahã simply may not recognize when one quantity of items exactly equals another, so they have trouble with matching tasks. He argues that people do possess an innate, non-verbal ability to conceive of all numbers, and that language simply helps them to refine it.
Psychologist Susan Carey of Harvard University in Massachusetts argues the opposite: she says we lack an innate ability to count beyond very small numbers, and that the Pirahã difficulty with numbers proves it. "It's a spectacular finding," she says.

[............]

On a broader level, the study also addresses a long standing and controversial hypothesis developed by Benjamin Lee Whorf in the late 1930s: that language can determine the way we think or what we are able to think.
But Gordon's study is one of the best examples in which language allows people to think something completely new, says cognitive psychologist Lisa Feigenson of Johns Hopkins University in Baltimore, Maryland. "This is by far the strongest piece of evidence," she says. In this case, a lack of language seems to prevent the Pirahã from thinking about larger numbers, she says.
I don't know if this study really is as paradigm-shaking as all that, as to a certain extent, it actually is obvious that the Sapir-Whorf hypothesis must be true. To see what I'm getting at, consider two languages, one spoken by, say, mathematics professors in the United States, and another spoken by an isolated pre-literate tribe; the latter language will lack words for concepts like "manifold" and "commutativity", and it is hard to imagine how any member of the group could even begin to grapple with such concepts without clear and rigorous definitions for them, while to the former group these ideas will be old hat, so much so that they'll be able to see applications of these ideas all over the place in the world around them. Much the same argument could be made for any jargon used by specialists in a given field, as the entire rationale for specialist terminology is the utility it provides as a mental shorthand. What is really interesting here is that this new research seems to indicate that the Sapir-Whorf hypothesis touches upon aspects of daily life one would have imagined to be so basic as to be invariant amongst human groups everywhere.

An interesting issue raised by this paper is the extent to which it is at all meaningful to speak of such things as "culture-independent" or "culture-free" intelligence tests. If abilities as rudimentary as a facility with numbers above 3 are contingent on one's language providing for them, how reasonable is it to expect that one can take a supposedly "culture free" test like the Ravens Progressive Matrices and expect illiterate villagers to know which answer is the "right" one to pick, when their languages* might not even possess the terms needed to describe what you require of your subjects? And this is before one even begins to ask awkward questions such as how one determines what the "right" answer ought to be on tests of this sort, even within a given culture; as the mathematically savvy will know, even a supposedly straightforward series like
1, 2, 4, 8, ...
isn't necessarily constrained to have "16" as the next term in the sequence, even if we apply some plausible-seeming "simplicity" principle (and why the "simplest" answer ought to be the "right" one would also need further explaining ...)

If the idea behind applying IQ-tests ("culture-independent" or otherwise) to illiterates in the lower Congo basin is to discern whether they'd able to thrive in a Western academic environment given their upbringings - the very thing IQ tests were designed for and are best at - one needs no psychometric expertise to see that the answer will be "no", any more than one needs a fancy test to determine that teenagers from Scarsdale, NY will make absolutely lousy trackers in the Kalahari. The very notion that "intelligence" can be assessed independently of culture is a dubious one, and I expect that as more results of the nature of the one above come in, its plausibility will continue to dwindle. Simply asserting that a test is "culture independent" or "culture free" doesn't make it so.

*This on the assumption that the test administrator even knows the native language to begin with, an exceedingly dubious assumption given the pre-existing belief in the intellectual inferiority of the testees displayed by most of those who go in for such surveys.