Fiammetta Marini¹ (), Meike Ramon^2,3, Julia Föcker^4,5, Mauro Manassi¹; ¹School of Psychology, University of Aberdeen, King’s College, Aberdeen, UK, ²Applied Face Cognition Lab, Business School, Bern University of Applied Sciences, Bern, Switzerland, ³AIR – Association for Independent Research, Zürich, Switzerland, ⁴Danish School of Education, Aarhus University, Aarhus, Denmark, ⁵College of Health and Science, School of Psychology Sport Science and Wellbeing, University of Lincoln, Lincoln, UK

In serial dependence, our actions, perception, decisions, and memories are pulled towards recently seen stimuli. This bias has been proposed as a purposeful mechanism, enhancing efficiency by merging past and present visual input. In fact, initial evidence in orientation perception showed that serial dependence between consecutive similar orientations leads to higher precision (i.e., reduced variability) in current perceptual judgements. Recently, this functional advantage of serial dependence has been questioned, as the same increase in precision has been observed even when successive stimuli were largely different and serial dependence was absent. To unravel these seemingly contrasting findings, we analysed multiple datasets from adjustment tasks related to orientation (N=45), position (N=72), shape (N=42), facial expression (N=41), and face identity (N=45) stimuli to test the consistency of these results. First, we found evidence for serial dependence across all the tested stimuli. Second, we observed evidence for improved efficiency in perceptual judgements at small stimulus differences across all tested datasets. Crucially, only datasets with orientation and position stimuli revealed a precision increase at larger stimulus differences, whereas no efficiency improvement was found for shapes and faces. Based on these findings, we propose that the efficiency advantage at larger stimulus differences is strictly related to the observer’s knowledge of the stimulus space and not to serial dependence itself. When observers have a clear representation of the stimulus space (orientation or position stimuli), they exhibit enhanced efficiency in recognising orthogonal stimuli. Conversely, when this clear representation of the stimulus space is absent (shape or face stimuli), this specific efficiency advantage does not occur. Overall, these results confirm that serial dependence improves efficiency across multiple levels of visual processing independently of concomitant stimulus-related biases, further reinforcing the idea of serial dependence as a purposeful, beneficial mechanism that enhances the efficiency of vision.

Acknowledgements: This research was supported by the UKRI ESRC Research Grant ES/Z503381/1.