A study just published in Science (scroll down to “Numerical Cognition Without Words: Evidence from Amazonia; Peter Gordon; Published online August 19 2004; 10.1126/science.1094492″—linked article and abstract only available to subscribers; brief Scientific American story, longer Science Daily piece) attempts to test the Sapir-Whorf hypothesis in terms of number:
During the late 1930s, amateur linguist Benjamin Lee Whorf posed the theory that language can determine the nature and content of thought. But are there concepts in one culture that people of another culture simply cannot understand because their language has no words for it?
No one has ever definitively answered that question, but new findings by Dr. Peter Gordon, a bio-behavioral scientist at Teachers College, Columbia University, strongly support a “yes” answer. Gordon has spent the past several years studying the Pirahã, an isolated Amazon tribe of fewer than 200 people, whose language contains no words for numbers beyond “one,” “two” and “many.” Even the Piraha word for “one” appears to refer to “roughly one” or a small quantity, as opposed to the exact connotation of singleness in other languages.
What these experiments show, according to Gordon, is how having the right linguistic resources can carve out one’s reality. “Whorf says that language divides the world into different categories,” Gordon said. “Whether one language chooses to distinguish one thing versus another affects how an individual perceives reality.”
When given numerical tasks by Gordon in which they were asked to match small sets of objects in varying configurations, adult members of the tribe responded accurately with up to two or three items, but their performance declined when challenged with eight to 10 items, and dropped to zero with larger sets of objects. The only exception to this performance was with tasks involving unevenly spaced objects. Here, the performance of participants deteriorated as the number of items increased to 6 items. Yet for sets of 7 to 10 objects, performance was near perfect. Though these tasks were designed to be more difficult, Gordon hypothesizes that the uneven spacing allowed subjects to perceive the items as smaller “chunks” of 2 or 3 items that they could then match to corresponding groups.
According to the study, performance by the Piraha was poor for set sizes above 2 or 3, but it was not random. “Pirahã participants were actually trying very hard to get the answers correct, and they clearly understood the tasks,” Gordon said. Participants showed evidence of using methods of estimation and chunking to guess at quantities in larger set sizes. On average, they performed about as well as college students engaged in more complex numerical estimation tasks. Their skill levels were similar to those in pre-linguistic infants, monkeys, birds and rodents, and appeared to correlate to recent brain imaging studies indicating a different sort of numerical competence that seems to be immune to numerical language deprivation. Interestingly, Gordon noted, while Pirahã adults had difficulty learning larger numbers, Piraha children did not.
There’s more at the Science Daily story linked above; thanks to Mike for the tip! [And see Mark Liberman's Language Log post for more details.]
By the way, there’s more interesting stuff on Piraha here in relation to their neighbors the Wari and “a heretofore undocumented grammatical sound… rendered as tp~ and pronounced as the t consonant sound followed immediately by what linguists call a ‘bilabial trill,’ which sounds like a person releasing air between vibrating lips in imitation of a snorting horse — or flatulence.” And here is a brief description of the language, which “is phonologically the simplest language known, having just ten phonemes, one fewer than in Rotokas.”