Questioning robustness of a human proto-symbolic capacity: associative symmetry

Several proto-symbolic cognitive capacities have been documented in mammals and birds, at least to some degree. But one capacity, routinely reported in humans and called associative symmetry – being able to reverse an association learned in a forward manner, has largely resisted demonstration in non-human animals. It is thus generally considered to represent a building block for human symbolic cognition and mark a cognitive discontinuity with nonhumans. Yet, recently revealed methodological approximations in human data weaken such claim. Here, we tested 165 human adults in three experiments, in order to 1) provide a much-needed replication establishing symmetry, 2) and 3) assess its robustness with new tests. We obtained positive results, but showed that minor changes to the standard, well-established protocol produce major disruptions of evidence in favor of symmetry. This study suggests a prominent role of task-related inductive biases and underscores the benefit of minimal verbal instrutions to participants. Our results downplay associative symmetry as a defining feature of human cognition and question the reality of an interspecific cognitive gap.