The performance of cleaner wrasse, Labroides dimidiatus, in a reversal learning task varies across experimental paradigms

Testing cognitive performance in controlled laboratory experiments is a powerful tool for understanding the extent and evolution of cognitive abilities in non-human animals. However, cognitive testing is prone to a number of potential biases, which, if unnoticed or unaccounted for, may affect the conclusions drawn. We examined whether slight modifications to the experimental procedure and apparatus used in a spatial discrimination and reversal learning task affected performance outcomes in the bluestreak cleaner wrasse, Labroides dimidiatus (hereafter ‘cleaners’). Using two-alternative forced choice tests, fish had to learn to associate a food reward with a side (left or right) in their holding aquarium. Individuals were tested in one of four experimental treatments that differed slightly in procedure and/or physical set-up. Cleaners from all four treatment groups were equally able to solve the initial spatial discrimination test. However, groups differed in their ability to solve the reversal learning task: no individuals solved the reversal task when tested in small tanks with a transparent partition separating the two options, whereas over 50% of individuals solved the task when performed in a larger tank, or with an opaque partition. These results clearly show that seemingly insignificant details to the experimental set-up matter when testing cognitive performance, and might significantly influence the outcome of experiments. When designing the methodology for comparative cognitive tests, care should be taken to ensure that all groups understand and can respond to the relevant cue to avoid misinterpretations.