An insect’s view of a numerosity illusion: a simple strategy may explain complex numerical performance in bumblebees

Summary Despite some similarities in numerical abilities among vertebrates, animals seem to differ with respect to the perceptual factors underlying numerical estimation. This is particularly evident in the case of numerosity illusions, a type of illusory phenomena elicited by the spatial arrangement of the elements that compose it. We investigated whether an insect experiences the most famous numerosity illusion, the Solitaire Illusion. Bumblebees ( Bombus terrestris audax ) presented with two mixed arrays containing magenta and yellow dots were trained to select the array containing the larger number of a target colour (either magenta or yellow). In the test phases, bumblebees were presented with sets of stimuli differing in the numerical ratio, including the illusory arrays: in one array, magenta dots were centrally located (yellow dots in the perimeter), whereas in the other one, magenta dots were located in the perimeter (yellow dots in the centre). Bees discriminated up to a ratio of 0.78 (14 vs. 18). Video analyses showed that bumblebees sequentially scanned the stimuli before making a choice. Bumblebees required longer scanning time for patterns with smaller numbers of reinforced colour dots, suggesting that bumblebees spent longer time on searching for more information or finding a reinforced arrangement of spatial dots before making an accurate decision. In the presence of the “illusion arrangement”, bees behaved in a manner consistent with misperception of numerosity, overestimating the number of centrally located dots. The spatial configuration of the stimuli influenced how the patterns were assessed: bees examined the clustered reinforced colour dots longer. Flying paths analysis suggested that bumblebees followed the paths formed by similar colour dots rather than counting each dot to make a numerical discrimination. These results challenge the notion that visual illusions can be inferred simply from evaluating choices of different visual patterns and reinforce the idea that multiple strategies can potentially be used to achieve seemingly similar cognitive outcomes. Detailed analyses of decision-making processes are essential to deduce the cognitive strategies underpinning discrimination tasks, whether they are numerical or other visual tasks.