ADVERTISEMENT

Parsing Primates: Baboons Learn to Recognize the Difference between Real and Fake English Words

New research suggests that some aspects of reading are not dependent on previous experience with language or speech
baboon-close-up



Andre Karwath, Wikimedia Commons

I bet you can raed waht tihs syas, eevn tohguh smoe wrods are sracbmled.

In his PhD thesis, Graham Rawlinson of the University of Nottingham showed that if one jumbles a word's interior letters, but preserves the positions of its boundary letters, people can still read the word with more ease than one might expect. Language standardizes the position of letters in words—what linguists call orthography and what is more colloquially known as spelling—but clearly the human brain can handle a little disarray.

But what if you have no idea how to spell? What if you have never seen written words before? What if you are a baboon? For the first time, scientists have taught baboons to tell the difference between English words and strings of letters that do not spell real words. The researchers conclude that—without any prior knowledge of written language—the primates learned to identify individual English letters and what kinds of letter combinations differentiated true English from nonsense words. The new findings suggest that some of the ways human brains understand written language do not depend on previous experience with written words or spoken language. Instead, when learning to recognize letters and words, our brains may co-opt circuits that evolved to recognize objects in our environment.

Jonathan Grainger of Aix–Marseille University in France and his colleagues trained six baboons to distinguish four-letter English words from nonsense strings of three consonants and one vowel. The baboons sat in front of a touch screen on which either a real or nonsense word appeared followed by an oval and a cross. Using wheat pellets as rewards, the researchers trained the baboons to touch the oval when a real word appeared and the cross when the letters formed only gibberish. One and a half months later, the baboons had learned to distinguish as many as 308 true English words from nonsense strings with nearly 75 percent accuracy. The study is published in the April 13 issue of Science.

Remember, the baboons have no idea what any of the words mean. As far as anyone can tell, all the baboons know is that there are two categories of patterns on the screen: pressing the oval in response to one category and the cross for the other earns a wheat pellet. But Grainger and his colleagues also see evidence that the monkeys are not simply memorizing which patterns correspond with either the oval or cross. Over time, the baboons got better at distinguishing the true words from nonsense. At first, hundreds of trials were required for the baboons to learn that, say, "wasp" and "feet" belong to one category and that "tokl" and "tezp" belong to another, but the more time they spent in training, the faster they differentiated real and made-up words, making fewer mistakes.

Grainger thinks that the baboons gradually learned which combinations of syllables distinguish real and fake words. Further evidence for this conclusion comes from the fact that the monkeys' mistakes increased with the degree of similarity between a nonsense word and one that was correctly spelled. Grainger showed this with a standard measure of "orthographic similarity" called OLD20, which quantifies how similar the order of letters in one word is to the order of letters in another word by counting the number of substitutions, deletions, insertions and swaps required to transform the first word into the second. The lower the OLD20 score, the more similar two words are. Baboons made the most mistakes with nonsense words that had low OLD20 scores.

Because the baboons had never learned to read by pairing written syllables with different speech sounds—which is how most children acquire the skill—Grainger suspects that the monkeys treated syllables like components of any other visual object, such as the shelves of a bookcase or the handle of a mug. However, he stresses that this is only one explanation and that his new findings are preliminary. "Most people would agree that skilled human readers depend on previous knowledge and experience to read written words," he says, "and although we think this is true, we think it is only part of the story. Letters are parts of words the same way that legs are part of a table. When we code information about letters and their approximate position, we may be mimicking what we are doing with ordinary object recognition."

Colin Davis, a psychologist who studies language processing at Royal Holloway, University of London, says the baboons' success is "truly remarkable." He adds: "There's an ongoing debate about the extent to which the ability to identify words is largely visual or the extent to which it relies on phonology—getting sounds out of letters. This study makes a big contribution to solving that debate. Although being able to utilize phonological code is helpful, it's not essential."

Rights & Permissions
Share this Article:

Comments

You must sign in or register as a ScientificAmerican.com member to submit a comment.
Scientific American Dinosaurs

Get Total Access to our Digital Anthology

1,200 Articles

Order Now - Just $39! >

X

Email this Article

X