| dc.description.abstracteng | Intuitive statistical reasoning is the capacity to draw intuitive probabilistic inferences based on an understanding of the relations between populations, sampling processes and resulting samples. This capacity is fundamental to our daily lives and one of the hallmarks of human thinking. We constantly use sample observations to draw general conclusions about the world, use these generalizations to predict what will happen next and to make rational decisions under uncertainty. Historically, statistical reasoning was thought to develop late in ontogeny, to be biased by general-purpose heuristics throughout adulthood, and to be restricted to certain situations and specific types of information. In the last decade, however, evidence has accumulated from developmental research showing that even pre-verbal infants can reason from populations of items to randomly drawn samples and vice versa. Moreover, infants can flexibly integrate knowledge from different cognitive domains (such as physical or psychological knowledge) into their statistical inferences. This indicates that neither language nor mathematical education are prerequisites for intuitive statistical abilities. Beyond that, recent comparative research suggests that basic forms of such capacities are not uniquely human: Rakoczy et al. (2014) presented nonhuman great apes with two populations with different proportions of preferred to non-preferred food items. Apes were able to infer which population was more likely to lead to a preferred food item as randomly drawn sample. Hence, just like human infants, great apes can reason from population to sample, giving a first hint that human statistical abilities may be based on an evolutionary ancient capacity.
The aim of the present dissertation is to explore the evolutionary roots of intuitive statistics more systematically and comprehensively by expanding on the initial findings of Rakoczy et al. (2014). I examined three questions regarding the i) generality and flexibility of nonhuman great apes´ statistical capacities, ii) their cognitive structures and limits, as well as iii) their interaction with knowledge from other cognitive domains. To address these questions, I conducted three studies applying variants of the paradigm established by Rakoczy et al. (2014).
In the first study, zoo-living great apes (Pan troglodytes, Pan paniscus, Pongo abelii, Gorilla gorilla) were required to infer from samples to populations of food items: Apes were presented with two covered populations and witnessed representative multi-item samples being drawn from these populations. Subsequently, subjects could choose which population they wanted to receive as a reward. I found that apes´ statistical abilities in this direction were more restricted than vice versa. However, these limitations were potentially due to accessory task demands rather than limitations in statistical reasoning. The second study was designed to gain deeper insights into the cognitive structure of intuitive statistics inchimpanzees and humans. More specifically, I tested sanctuary-living chimpanzees and human adults in a task requiring inferences from population to sample and I systematically varied the magnitude of difference between the populations´ ratios (the ratio of ratios, ROR). I discovered that the statistical abilities of both chimpanzees and human adults varied as a function of the ROR and thus followed Weber´s law. This suggests that intuitive statistics are based on the analogue magnitude system, an evolutionary ancient cognitive mechanism common to many aspects of quantitative cognition. The third study investigated whether chimpanzees consider knowledge about others´ mental states when drawing statistical inferences. I tested sanctuary-living chimpanzees in a task that required subjects to infer which of two populations was more likely to lead to a desired outcome for the subject. I manipulated whether the experimenters had preferences to draw certain food types or acted neutrally and whether they had visual access to the populations while sampling or drew blindly. Chimpanzees chose based on proportional information alone when they had no information about experimenters’ preferences and (to a lesser extent) when experimenters had preferences for certain food types but drew blindly. By contrast, when biased experimenters had visual access, subjects ignored statistical information and instead chose based on experimenters’ preferences. Consistent with recent findings on pre-verbal infants, apes seem to have a random sampling assumption that can be overridden under the appropriate circumstances and they are able to use information about others´ mental states to judge whether this is necessary.
Taken together, the findings of the present dissertation indicate that nonhuman great apes possess intuitive statistical capacities on a par with those of human infants. Therefore, intuitive statistics antedate language and mathematical thinking not only ontogenetically, but also phylogenetically. This suggests that humans´ statistical abilities are founded on an evolutionary ancient capacity shared with our closest living relatives. | de |