Long-term representations systematically bias ongoing perception

The perception of a stimulus can be systematically biased by the immediately preceding stimulus, an effect known as serial dependence (Fischer and Whitney, 2014). This has been typically demonstrated by asking participants to report a feature (i.e., orientation) of an object (i.e., grating) and measuring to what extent the report of that feature is biased by the previous trial. However, it is unclear whether similar biases can emerge from task-relevant representations over longer time scales. In the present online study, participants were asked to report the orientation of gratings with different background colours. Participants were informed that gratings of a specific colour (e.g., blue; cue gratings) would always be followed by a target grating of another specific colour (e.g., green) and tilt (i.e., 60 degrees). All other gratings, including cue gratings, had random orientations. Replicating previous findings of serial dependence, we found that orientation reports of random gratings were systematically repelled from the orientation of the immediately preceding grating. Interestingly, we found that participants were systematically repelled from the orientation of the target grating in their orientation reports of all randomly tilted gratings. Furthermore, the bias exerted by the target grating did not differ immediately before it was expected (when reporting cue gratings) than when it was unexpected (when reporting random gratings). Finally, we found that the general biasing effect of the target grating on all other gratings was additive with the classic serial-dependence effect, suggesting that they reflect distinct phenomena. From this, we concluded that long-term representations can additionally bias ongoing perception.