NEW YORK (Reuters) – No one is exactly using the words “reading” and “baboons” in the same sentence, but a study published Thursday comes close.
Researchers report in the journal Science that they trained six Guinea baboons (Papio papio) to distinguish real, four-letter English words such as “done” and “vast” from non-words such as “dran” and “lons.” After six weeks, the baboons learned to pick out dozens of words – as many as 308 in the case of the clever Dan, and 81 for Violette – from a sea of 7,832 non-words.
Each of the monkeys performed significantly better than 50 per cent, which they would have scored by randomly guessing which letters formed words or non-words. They averaged almost 75 per cent right, with some scoring 90.
The study is “extraordinarily exciting,” said cognitive psychologist Stanislas Dehaene of the College de France in Paris, an expert on the neural basis of reading who was not involved in the research. “For the first time, we have an animal model of a key component of literacy, the recognition of the visual word form.”
Ancient brain circuits
The study was intended less to probe animal intelligence than to explore how a brain might learn to read. It suggests that, contrary to prevailing theory, a brain can take the first steps toward reading without having language, since baboons don’t.
“Their results suggest that the basic biological mechanisms required for reading have deeper evolutionary roots than anyone thought,” said neuroscientist Michael Platt of Duke University, who co-authored an analysis of the study. “That suggests that reading draws on much older neurological mechanisms” and that apes or monkeys are the place to look for them.
Reading has long puzzled neuroscientists. Once some humans started doing it (about 5,000 years ago in the Middle East), reading spread across the ancient world so quickly that it cannot have required genetic changes and entirely new brain circuitry. Those don’t evolve quickly enough. Instead, its rapid spread suggests that reading co-opted existing neural structures.
For the new study, scientists at Aix-Marseille University in France trained the six baboons by setting up nine booths, equipped with computers and touch screens, in two trailers in the monkeys’ 80-by-100-foot (25- by-30 metre) enclosure. The animals wandered in when they felt like it.
Once a baboon was at the screen, it was shown a string of four letters that was either a real English word (“them”) or a non-word (“telk”). The baboon learned to touch an oval on the screen when it saw a word, and a plus (+) for a non-word. Every correct response brought a reward of dry wheat.
Once they got the hang of things, it was test time.
Again strings of four letters appeared on the screen, but now they were never-before-seen strings. If a baboon decided the letters formed a word, it pressed the oval sign; for non-words (always three consonants and a vowel), it pressed a plus. In a video released by the scientists, the baboons swat the screen (“itcs”: plus!; “kite”: oval!) with the alacrity and assuredness of a 12-year-old acing a video game.
The baboons were engaging in “orthographic processing,” said Aix-Marseille’s Jonathan Grainger, who led the study. That term means recognizing letters and their positions. It has nothing to do with sounding out a word, let alone understanding it. Still, it is a necessary early step in reading
Statistical properties of words
The scientists’ best guess as to how the animals were able to identify real words is that they learned “the statistical properties that distinguish words from non-words,” said Grainger. That is, they learned that “sl” and “dr,” say, are found in many words, but “dl” and “sr” are not.
“When people learn words, and when we learn to distinguish words from non-words, we’re doing statistical pattern recognition, too,” said Duke’s Platt. “The critical finding is that the baboons were able to transfer this learning to words they had never seen before.”
Grainger called the baboons’ accuracy remarkable, since the words were so superficially similar to the non-words (not a “sxjayud” in the bunch).
To be sure, other animals have learned to recognize letters. In a 1982 experiment, for instance, pigeons were able to identify all 26 letters of the English alphabet.
But the baboons were not simply memorizing which strings of letters were words, said Grainger. When shown a word for the first time, they identified it correctly about 70 per cent of the time, suggesting the animals were applying the statistical rules they had inferred.
The word-savvy baboons may be drawing on “more generalized learning mechanisms and visual processing abilities rather than specialized mechanisms unique to humans,” said Diana Reiss of Hunter College in New York, who has done pioneering work in animal intelligence.
A prime candidate for those processing abilities lies in a region of the brain that becomes active when people read. Discovered by Dehaene, it is called the “visual word form area” and is located behind the left ear. It recognizes strings of letters, and the more active it is in 7- to-18-year-olds, studies show, the better readers they tend to be.
“Neuroimaging shows that this region is specific for words and not meaningless strings of letters,” said Duke’s Platt.
Since reading arrived on the scene a mere blink of the eye ago, evolutionarily speaking, the visual word form area cannot have developed in order to support reading. If baboons or human ancestors also had this structure, the question becomes what they used it for. Best guess: recognizing objects by visually assembling their parts, such as tall cylinder + bushy top = tree.
Among the many surprises in the study is that it involved baboons rather than a primate known for braininess.
“Guinea baboons have a lot of social savvy, since they have to learn about complex male-male and male-female interactions in their troop,” said primate curator Craig Demitros of the Brookfield Zoo outside Chicago. “They’re smart, but not at the level of chimps.” Apart from the glimpses it provides into the evolution of the brain’s ability to read, the study has implications for education. “You might conclude that phonics doesn’t work” as well as teaching children to read by recognizing the entire word, said Platt. “This study suggests that reading is all about pattern recognition and not working out phonemes.”