Demonstrating inferential reasoning through behavioural choices: an experimental analysis

We critically examine theoretical and experimental evidence that animals’ reason by exclusion, focusing on a widespread protocol known as the 2-Cups Task. We test the hypothesis that subjects solve inference by exclusion tasks by responding independently to two options, without implementing complex reasoning, as posited by the Sequential Choice Model. In the critical 2-Cups Task subjects are signalled that a bait is hidden in one of two cups and then shown that one is empty while the other’s content is not revealed. Subjects are then allowed to choose and are considered successful if they choose the unrevealed cup. We emulate the procedure using arbitrary stimuli and starlings (Sturnus unicolor) as subjects. Birds learned in encounters with single stimuli that one coloured pecking key never yielded rewards ("empty cup") and another yielded rewards on half the trials ("resting cup"). When facing only one of the two stimuli, birds displayed differential behaviour, pecking at the resting cup and not at the empty one. From the first time they faced the stimuli together they all chose appropriately. We argue that primates in the 2-Cups Task can behave similarly, with correct choices not being evidence for inference by exclusion. Studies of inference by exclusion should report data on behaviour towards single cups, to consider the possibility that simple mechanisms explain subjects’ choices. This contribution highlights the contrast between simple, generalisable, and falsifiable models of comparative cognition and alternatives that rely on high-level operations such as ’causal understanding’, but without controls to exclude the former.