Abstract
Human observers take longer to re-direct gaze to a location they previously fixated (inhibition of saccadic return: ISR). Arguments have been made for the adaptive value of this phenomenon in the context of search and foraging. However, ISR is only adaptive if previously fixated locations are unlikely to become behaviourally relevant soon after the original fixation.
In this study we examined whether ISR reflects a “hard-wired” assumption about the probabilistic structure of the world or whether it is sensitive to variations in probabilistic structure across different contexts. Observers made sequences of 2 saccades triggered by central, symbolic cues in a gaze-contingent manner. The probability that the initial starting position would become the target of the 2nd saccade was systematically manipulated between three groups of observers (low, equal and high return probability).
All three groups initially showed a strong inhibitory effect, with return saccades having much longer latencies than non-return saccades. ISR was strongest in the group for which return probability was low. Over time, ISR was completely abolished in the group for which return probability was high.
Individual observers' latency distributions were fit with an evidence accumulation model. Replicating our previous work we show that, under equal return probability conditions, ISR is best modelled as a reduction in the rate at which evidence is accumulated to a response threshold. This rate reduction was independent of the probabilistic context. However, the magnitude of behaviourally observed ISR depended strongly on the probabilistic context. This effect was mediated by adaptive changes in the response boundary. We conclude that a flexible response boundary allows observers to overcome strong in-built assumptions about the statistical structure of the world, and ensures that behaviour is more optimally adapted to the local context.
Supported by The Wellcome Trust (079473/Z/06/Z) and EPSRC (EP/E054323/1).