What Can We Learn From Monkeys About Orthographic Processing in Humans? A Reply to Ziegler et al.

In two provocative articles (Grainger, Dufau, Montant, Ziegler, & Fagot, 2012; Ziegler et al., 2013), Ziegler and his colleagues have recently made 6 baboons unexpected contributors to reading research. The baboons were trained using operant conditioning to differentiate between repeated four-letter word stimuli consisting of high-frequency English bigrams, such as DONE, and four-letter nonword stimuli consisting of low-frequency bigrams, such as VIRT. The articles report that the baboons learned to discriminate the words from the nonwords with relatively high accuracy, and like humans, they also showed transposed-letter effects—the baboons tended to confuse nonwords as belonging to the word category they had been trained on if the nonwords involved letter transpositions (e.g., DONE → DNOE). The authors argue that because baboons do not have a linguistic system but nevertheless perform like humans do, the neural mechanisms underlying orthographic processing in the two species must be similar and therefore nonlinguistic.

We argue that these conclusions are logically fallacious and do not withstand empirical scrutiny. If the performance of baboons with printed material is at all similar to that of humans, it does not follow that the neural mechanisms underlying orthographic processing in humans is similar to that in baboons. Similarly, the presence of transposed-letter effects in the absence of a linguistic system does not imply the absence of linguistic modulation of transposed-letter effects. More important, however, close inspection reveals that the baboons’ behavior, as reported by Ziegler and his colleagues, is critically different from that of humans.

The issue at stake is the extent to which humans and baboons respond similarly to misspellings of words that contain transposed letters. Researchers who have examined the impact of manipulating letter order on reading performance in humans have shown a small cost of letter-transpositions in terms of reading time, along with robust masked priming effects when primes and targets share all of their letters but in a different order (e.g., Perea & Lupker, 2003). Ziegler and his colleagues have shown that the 6 baboons in their studies classified both words (e.g., DONE) and their transposed-letter version (e.g., DNOE) as “words.” Although, at first glance, this finding may seem to be similar to that of humans, Ziegler and his colleagues seem to forget that the transposed-letter phenomenon presupposes a substantial ability to differentiate words from their transposed-letter versions in the first place. Considering the baboons’ absolute level of accuracy, they seem to consistently perceive the transposed-letter versions of words as “words,” making as many positive responses to trained words as they make false-positive responses to transposed-letter nonwords. This stands in sharp contrast to the performance of humans, who correctly reject transposed-letter nonwords in a lexical decision task (albeit more slowly and slightly less accurately than nonwords with substituted letters; e.g., Chambers, 1979). Thus, humans have a genuine flexibility in coding letter position in spite of their explicit knowledge that letter order matters in constructing words.

For now, the only thing that Ziegler and his colleagues have shown is that baboons learned that the presence of certain shapes or symbols in a series has a relation to a particular response category and that the order in which they are presented does not matter. This does not mean that they demonstrated “flexible” letter coding. Moreover, as Ziegler and his colleagues report, the baboons could not discriminate between words and nonwords that were different by a single letter, again in sharp contrast to humans. This suggests that a critical proportion of mismatches is required for baboons to consider two series of shapes to be different, which demonstrates severe limitations on how far “orthographic processing” can develop nonlinguistically via the object-recognition system. Ziegler and his colleagues also seem to forget that transposed-letter effects for humans are not always present and are modulated by the linguistic properties of the stimuli (e.g., Duñabeitia, Dimitropoulou, Grainger, Hernández, & Carreiras, 2012; see Frost, 2012a, 2012b, for a review). Baboons by definition are blind to such linguistic factors.

Comparing the abilities of humans and nonhumans allows researchers to trace the demarcation line between processing mechanisms shared with other species and those that are specifically human. However, such investigations should also seek the point at which the performance of species diverges rather than halt at an apparent convergence. For example, finding that both humans and rats can segment a stream of continuous speech (e.g., Toro & Trobalon, 2005) does not imply that the cues that govern speech segmentation are identical for humans and rats. Indeed, Toro and Trobalon demonstrated that rats are sensitive to simple frequency of co-occurrence, whereas humans rely on transitional probabilities. Returning to “reading” performance of baboons, the results simply show that baboons can learn probabilistic conjunctions of three to four individual shapes and that this learning does not extend to the order of the shapes. Whether this form of statistical learning should be labeled “orthographic processing” seems very doubtful. More important, the data regarding how letter transpositions affect baboons versus how they affect humans certainly does not suggest that the processing mechanisms for orthographic information are the same in the two species. The inevitable conclusion is, therefore, that the recent findings with baboons reveal something about their statistical learning abilities but have no important implications for theories of human visual word recognition.