A while back now, I wrote a brief piece commenting on the ability of at least some chimpanzees to recognise numbers. So it didn’t come as a huge surprise to hear that members of a baboon troop could distinguish between ‘real’ words and random strings of letters. Yes, really.

A group of psychologists led by Jonathan Grainger (Grainger, Dufaur, Montant, ZIegler & Fagot, 2012) have just published a paper in Science entitled “Orthographic Processing in Baboons (Papio papio)”, where ‘orthography’ is a standardised system for using a particular writing system (script) to write a particular language. The team noted that most research on visual word recognition hasn’t treated words as ‘visual objects’, instead dealing with the relationship between information at the letter level and ‘higher-level linguistic properties including semantics & syntax. But it seems that the ability to recognise words as entities resides in a part of the brain that’s also involved in recognition of objects & faces, and primates are pretty good at faces, so Grainger & his colleagues decided to investigate whether baboons could extend their facial recognition skills to identifying words.

More specifically:

The computation of letter identities and their relative positions is referred to as orthographic processing, and there is a large consensus today that such processing represents the first “language-specific” stage of the reading process that follows the operations involved in the control of eye movements (bringing words into the focus of central vision) and early visual processing (enabling visual feature extraction). In the present study, we examined whether the ability to efficiently process orthographic information can operate in the absence of prior linguistic knowledge.

Hence the decision to work with a non-human primate species: baboons don’t use any phonological equivalents of English words (or, most likely, words in any other human language), & so can’t be said to have any prior knowledge of a human linguistic system .

So, what did the researchers do? They worked with a captive social group of baboons that were living in a large enclosure with various climbing structures & sleeping areas, & set things up so that the animals also had free access to a set of test computers that used touch-screen technology & provided operant conditioning: the animals would get a food reward for correctly recognising an English word (as opposed to a string of random letters). The ‘free access’ thing is important – the baboons could get involved, or not, as they chose.

Using that operant conditioning, the baboons learned

to recognize four-letter English words and distinguish them from strings of letters that are not English words.

Each time a letter string (word or non-word) showed on the screen before it, a baboon could touch either a blue cross (for a random set of letters such as DRAN, LONS, TELK, or VIRT) or a green oval (for a four-letter word such as such as DONE, LAND, THEM, or VAST). A correct response was rewarded with a blank computer screen & some food (dry wheat), while an incorrect choice got a green screen for 3 seconds. They began with a single genuine word option & worked up from there (my emphasis):

Words and nonwords were presented randomly in blocks of 100 trials. The 100-trial sessions were composed of 25 presentations of a novel word to learn, 25 presentations of words randomly selected from already learned words, and 50 nonword trials. Each new word was added to the ever-increasing pool of already learned words, once responses to that word exceeded 80% correct within the preceding session. Thus, in terms of explicit information available to the baboons, a word was defined as a string of letters that was repeatedly presented, whereas a nonword was rarely repeated.

During the course of the experiment, individual animals learned to recognise a surprising number of 4-letter English words (from 81 for ‘VIO’ to 308 for ‘DAN’) – correctly distinguishing the words they recognised from a total of 7832 ‘non-word’ combinations.

Obviously the baboons were simply making random choices at the start of the experiment, and in fact they did this in quite a biased way, with each individual tending to go repeatedly for either the green or the blue button. But – after 2000 trials, they became a lot more accurate, correctly identifying words around 75% of the time. And they were doing this on the basis of different frequencies of letter combinations, rather than ‘just’ memorising the real words (although that would be a rather amazing feat in itself). What’s more,

words that were seen for the first time triggered significantly fewer “nonword” responses than did the nonword stimuli. This implies that the baboons had extracted knowledge about what statistical properties characterize words and nonwords and used this information to make their word versus nonword decision without having seen the specific examples before.

And:

The more similar a nonword was to a known word, the more false positive responses it produced.

The researchers noted that this mirrors responses in skilled human readers in the same situation – a rather unexpected outcome.

So, are we looking at some feature(s) of the way the primate brain is wired, that could be regarded as exaptations when it comes to processing visual symbols? Grainger & his colleagues certainly think so:

The primate brain might therefore be better prepared than previously thought to process printed words, hence facilitating the initial steps toward mastering one of the most complex of human skills: reading.

Grainger J, Dufau S, Montant M, Ziegler JC, & Fagot J (2012). Orthographic processing in baboons (Papio papio). Science (New York, N.Y.), 336 (6078), 245-8 PMID: 22499949