Retention interval modulates motion-history biases in visual motion perception

Abstract

Serial dependence, the tendency for recent perceptual history to bias current judgments, has been proposed as a mechanism supporting perceptual stability, although its cognitive origin remains debated. We investigated serial dependence in motion direction estimation using random-dot kinematograms (RDKs). Participants reported perceived motion direction after delays of 0, 1, 3, or 6 s following stimulus offset. Trial-level linear mixed-effects modelling revealed a significant serial dependence effect and a significant interaction with delay, indicating that the magnitude of the DoG-based history effect varied across retention intervals. Contrary to accounts predicting immediate repulsion and delayed attraction, the parametric DoG analysis yielded positive peak estimates at all delays, including the nominal 0-s condition, in which no additional retention interval followed stimulus offset. However, a complementary folded-bias analysis indicated that this parametric attractive component was not uniformly expressed across angular distances but was strongest at small-to-intermediate current-minus-previous angular differences and coexisted with local repulsive deflections. These findings do not support a simple dissociation between immediate sensory repulsion and delayed mnemonic attraction. Instead, they suggest that recent motion history already influences current reports in the immediate-report condition, although this condition should not be interpreted as isolating a purely perceptual stage. Retention interval then modulated the strength and expression of this influence. Exploratory analyses further suggested that motion judgments may be influenced not only by the exact previous direction but also by the broader motion axis, whereas response-time analyses did not reveal a simple overall dependence on trial-to-trial angular similarity.

The raw data, analysis code, and MATLAB/Psychtoolbox experiment task code for this study are publicly available on the Open Science Framework at https://osf.io/9ea8p/. The study was not preregistered.