Word-detecting baboons are a tough read
Studies differ on whether monkeys learn to recognize letter combinations
MINNEAPOLIS — Baboons use the order of regularly appearing letter pairs to tell words from nonwords, new evidence suggests.
Psychologist Jonathan Grainger of the University of Aix-Marseille reported earlier this year that baboons can learn to tell real four-letter words from nonsense words (SN: 5/5/12, p. 5). But whether these animals detect signature letter combinations that enable their impressive word feats has been tough to demonstrate.
Monkeys that previously learned to excel on this task are more likely to mistake nonwords created by reversing two letters of a word they already recognize as real, much as literate people do, Grainger reported November 16 at the Psychonomics Society annual meeting.
“Letters played a role in baboons’ word knowledge,” Grainger concluded. “This is a starting point for determining how they discriminate words from nonwords.”
Science News headlines, in your inbox
Headlines and summaries of the latest Science News articles, delivered to your email inbox every Thursday.
Thank you for signing up!
There was a problem signing you up.
Grainger’s team tested the six baboons in their original investigation. Some of the monkeys had previously learned to recognize many more words than others. In new trials, the best word identifiers made more errors than their less successful peers when shown nonwords that differed from known words by a reversed letter combination, such as WSAP instead of WASP and KTIE instead of KITE.
Grainger’s team fed the same series of words and nonwords into a computer simulation of the experiment. The computer model best reproduced the animals’ learning curves when endowed with a capacity for tracking letter combinations.
Psychologist Emmanuel Keuleers of Ghent University in Belgium is skeptical of the findings, he said. His own computer simulations mimicked baboons’ word learning in Grainger’s experiments without being able to track specific letter combinations.
Keuleers’ program calculated the similarity of new words to previously learned four-letter strings based on the number of insertions and deletions of single letters. As with the baboons, this computer system easily learned new nonwords but often stumbled when supplied with new words. That’s because, using this single-letter strategy, new words presented to baboons in Grainger’s experiments were typically most similar to previously learned nonwords, Keuleers suggested.“Baboons’ processing strategy for discriminating words from non-words is inconclusive,” he said.