Cross-dimension serial dependence effects between numerosity and duration

Abstract Background Our perception and decisions are not only driven by present information, but are also influenced by past information. For instance, previous stimuli can affect the judgement of current ones in an attractive way—a phenomenon known as “serial dependence.” Serial dependence has been shown to occur even across stimuli with different low-level features, suggesting the involvement of high-level computations. Here we further address the level of abstraction at which serial dependence originates by investigating biases across different perceptual dimensions, that is, numerosity and duration—two “magnitude” dimensions that usually bias each other when modulated together. Participants were shown sequences of briefly presented dot-arrays, and asked to reproduce either their duration or the average numerosity computed across the sequence, in separate conditions. We then assessed the influence of duration on numerical estimates, and vice versa, both within the same stimulus (“magnitude integration”) and across successive stimuli (“serial dependence”). Results Our results show significant influences across the two dimensions, occurring both within the same stimulus and across successive stimuli. Moreover, we show that the strength of serial dependence can be predicted based on the strength of magnitude integration, suggesting a relationship between the two effects. Conclusions Our findings demonstrate that serial dependence can occur between two different perceptual dimensions when they provide compatible information (i.e., “more” vs. “less”), and thus that serial dependence originates from computations shared across different magnitude dimensions. The relationship between serial dependence and magnitude integration additionally suggests a reliance on similar mechanisms, integrating both past and present magnitude information to build a generalized estimate of magnitude.