Ciccione, L.; Dighiero Brecht, T.; Claidiere, N.; Fagot, J.; Dehaene, S.

Recent studies showed that humans, regardless of age, education, and culture, can extract the linear trend of a noisy graph. Here, we examined whether such skills for intuitive statistics are confined to humans or may also exist in non-human primates. We trained Guinea baboons (Papio papio) to associate arbitrary geometrical shapes with the increasing or decreasing trends of noiseless and noisy scatterplots, while varying the number of points, the noise level, and the regression slope. Many baboons successfully learned this conditional match-to-sample task for both noiseless and noisy plots. Crucially, for successful baboons, accuracy varied as a sigmoid function of the t-value of the regression, the same statistical index upon which humans also base their answers, even after controlling for other variables. These results are compatible with the hypothesis that the human perception of data graphics is based on the pre-emption and recycling of a phylogenetically older competence of the primate visual system for extracting the principal axes of visual displays.