Abstract
Evidence suggesting non-human primates infer the hidden functional properties of tools is equivocal, possibly because subjects had to reason about two factors: the tools’ functional properties and the out-of-reach reward. We designed a novel inference task with a simplified set-up in which subjects only had to reason about one factor. Subjects could choose between a breadstick that was intact or broken. If subjects chose the intact breadstick they could eat all of it, whereas choosing the broken breadstick would mean they could only eat less than half or, in some cases, none at all. Covers were placed over the breadsticks’ middle sections that hid their broken or intact properties but allowed their ends to protrude. The breadsticks now looked identical but it was inferable which one was intact by watching the experimenter move one end of each breadstick, which moved both ends of the intact breadstick but not the broken breadstick’s unattached end. We found that only one of the 12 subjects chose the intact breadstick significantly above chance. In a series of follow-up experiments, we investigated if procedural modifications would improve the subjects’ performance. Although some experiments resulted in more subjects being successful, learning during the test procedures could not be excluded. Therefore, we tested subjects with novel types of breadstick configurations and most subjects passed these transfer tests suggesting that orangutans and chimpanzees show some evidence of inferring the intact and broken properties of breadsticks in this task. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1038/s41598-026-38796-x.