The use of abstract higher-level knowledge (overhypotheses) allows humans to learn quickly from sparse data, and make predictions in new situations. Previous research has suggested that humans may be the only species capable of abstract knowledge formation, but this remains controversial, and there is also mixed evidence for when this ability emerges over human development. Kemp et al. (2007) proposed a computational model of overhypothesis formation from sparse data. We provide the first direct test of this model: an ecologically valid paradigm for testing two species, capuchin monkeys (Sapajus spp.) and 4-5-year-old human children. We compared performance to predictions made by models with and without the capacity to learn overhypotheses. Children’s choices were consistent with the overhypothesis model predictions, whereas monkeys performed at chance level.

The relational matching-to-sample task (RMTS) is a gold standard in measuring abstract concepts. Most preschoolers and non-human animals do not spontaneously succeed in the classic, two-item version of the task. It is debated whether this failure indicates a lack of abstract reasoning ability, perhaps linked to limited language capabilities, or rather stems from learned biases for other bases of matching. We developed a physical, causal RMTS task for 4- to 5-year-old children based on matching the weight relations within object pairs by asking them to align two balance scale apparatuses. We presented conflicting object matches in half of the trials and a transfer phase with a new set of stimuli. By age five, children benefitted from the causal context of the task, suggesting that not solely abstract reasoning abilities but other factors, like biases to match individual object features, influence their performance in classic arbitrary RMTS tasks.