Task-dependent reversal learning dynamics challenge the reversal paradigm of measuring cognitive flexibility

Abstract

Cognitive flexibility may be necessary for animals facing changing conditions and has long been tested by the reversal learning paradigm. However, while this paradigm is typically based on training animals to discriminate between a rewarding and nonrewarding stimulus and then reversing their roles, under natural conditions animals usually face more than one set of two stimuli. Here, we addressed these potential intricacies by studying the reversal learning of house sparrows, Passer domesticus , in a two-set task. Sparrows previously trained as either colour or shape specialists exhibited different reversal dynamics: While colour discrimination was acquired faster than shape discrimination, shape specialists, including those reaching perfect preference of the rewarding shape, reversed faster than colour specialists. The reversal success of shape specialists was also less variable than that of colour specialists. Additionally, despite being slower, during the reversal, colour specialists sampled shapes more than shape specialists sampled colours. These results suggest that (1) when comparing different reversal tests, significant differences in reversal behaviour may be explained by the type of learned stimulus (e.g. shape versus colour) rather than by differences in cognitive flexibility and (2) under realistic conditions of multiple foraging options, preferences that are difficult to reverse (and typically viewed as indicating low cognitive flexibility) may nevertheless facilitate an ecologically flexible shift to novel food types. Finally, our results also demonstrate that a two-set (multiple-cue) experimental design may help to tease apart some of the different processes underlying reversal behaviour and cognitive flexibility. • The results of reversal learning tests differed depending on stimulus type. • Differences were independent of the discrimination level reached before reversal. • Associations that were difficult to reverse induced flexible shift to novel cues. • Reversal scores may not reflect ecological flexibility in multiple-cue environments. • A multiple-cue set-up helps tease apart the subcomponents of the reversal process.