Exploring visual categorical learning in teleost fish Danio rerio and Xenotoca eiseni

Categorization skills enable animal species to sort entities based on their similarities, providing significant advantages in terms of learning efficiency. The extent of these skills and the mechanisms involved vary across taxa, making it critical to explore categorization in as many lineages as possible. From an evolutionary standpoint, fish are valuable models for studying categorization processes; however, until recently, this group of vertebrates has been neglected in visual categorization research. To address this gap, we conducted three experiments to investigate, for the first time, the visual categorical learning skills of the redtail splitfin (Xenotoca eiseni) and the zebrafish (Danio rerio), two teleost species studied in other contexts of animal cognition. Using a two-alternative forced-choice task, we first trained the fish to discriminate between several pairs of stimuli. Then, we tested their ability to generalize to pairs of new stimuli of the same categories. The stimulus categories varied across the experiments (Experiments 1 and 2: alphanumeric characters; Experiment 3: fish silhouettes) and were consistent with those used in past studies involving different taxa. Both fish species met the learning criterion in the training phase across all three experiments, but their ability to generalize to novel stimuli differed based on stimulus features. Moreover, both species presented interindividual differences in their ability to generalize what they learnt. Our results validate the use of X. eiseni and D. rerio as models for research on categorical learning. They also encourage further investigation into how experimental methods and stimuli could affect learning-based generalization abilities in these species.