Abstract
Like humans, several mammalian and avian species prefer foretold over unsignalled future events, even if the information is costly and confers no direct benefit. It is unclear whether this is an epiphenomenon of basic associative learning mechanisms, or whether these preferences reflect a derived form of information-seeking that is reminiscent of human curiosity. We investigate whether a fish that shares basic reinforcement learning mechanisms with birds and mammals also shows such a preference, with the aim of elucidating whether widely shared conditioning processes are sufficient to explain paradoxical preferences resulting in unusable information. Goldfish (Carassius auratus) chose between two alternatives, both resulting in a 5 s delay and 50% reward chance. The 'informative' option immediately produced a stimulus correlated with the trial's forthcoming outcome (reward/no reward). Choosing the 'non-informative' option instead triggered an uncorrelated stimulus. Goldfish discriminated between the different contingencies but did not develop a preference for the informative option, suggesting that in goldfish associative learning mechanisms are not sufficient to generate preferences between alternatives differing only in outcome predictability. These results challenge the notion that informative preferences are a by-product of ubiquitous associative processes, and are consistent with the possibility that derived information-seeking mechanisms have evolved in some vertebrate species.