Regularities in temporal context influence the window of temporal integration

Perceptual content is integrated as it unfolds in real time within temporal windows. Understanding the dynamics of temporal integration requires understanding the influence of an ongoing temporal context. Toward this goal, we developed a novel paradigm of temporal integration where two halves of a Kanizsa square are integrated together (seen as a full square) or segregated (not seen as a full square) as a function of a temporal gap between them. The two halves are embedded within a preceding and succeeding context of rotating Kanizsa discs leading up to and trailing away from them. We wanted to investigate whether integration is subject to frames of interest or whether the dynamics of the context influence form-part integration. Through three experiments, we manipulated the nature of the rotation: (a) the context being temporally correlated or random in Experiment 1, (b) the variance of the temporally correlated context in Experiment 2, and (c) the mean of the temporally correlated context in Experiment 3. Our results showed that the sensitivity of seeing an integrated square was better when the context aided anticipation of the target frames, the effect being selective at different inter-stimulus intervals depending on the properties of the context. We interpret these results in light of a probabilistic view of temporal windows of integration. Our results provide evidence for an inherent adaptiveness in the extent or window of temporal integration. The results also reinforce the general account of extended temporal processing and iterative updating and inform rough timescales of context effects on this process.