Abstract
Studies of serial dependence have shown that the perception and memory of the current-trial stimulus are systematically biased by stimuli seen on previous trials. Is this effect driven by the perception of the previous stimuli or by the use of these stimuli to achieve given task goals? To answer this question, we designed a two-feature discrimination task in which random dot arrays moved in one of several directions and were drawn in one of several colors. At the end of the stimulus period, a feature cue was presented to indicate which of the two features (direction of motion or color of the dots) should be reported. Observers either reported the perceived motion direction by adjusting a response line or reported the color of the moving dots by selecting a color from two alternative color patches. The two colors could be from the same category (e.g., two pinks) so that mere categorical color memory was not sufficient to perform the task. Crucially, the feature cue was completely unpredictable (p = 0.5), so observers had to perceive and remember both the color and the direction of motion. We assessed serial dependence for motion perception as a function of whether the observer had been cued to report motion direction or color on the previous trial. We found a typical serial dependence effect in motion perception only when observers reported motion direction in the previous trial. To ensure that this effect was not driven by the difficulty of the color task, we replicated the main results in Experiment 2, in which dots were always either pink or green. Together, these findings suggest that the serial dependence effect is not merely driven by perceiving a given feature but also requires the report/retrieval of the relevant feature on the previous trial.
Acknowledgement: This work was supported by grant R01MH076226 to SJL.