Considering the present results, the central question thus is why serial dependence is limited to a single stimulus dimension, whereas adaptation can transfer across dimensions (at least from duration to numerosity). We thought about two possible accounts of this result. The first possibility is that serial dependence occurs, in this context, at a more limited level of abstraction compared to adaptation. Intuitively, this might seem surprising, because adaptation is usually considered a low-level physiological process (e.g.,
Kohn, 2007), whereas serial dependence very often shows the hallmarks of a much higher-level effect (e.g.,
Fornaciai & Park, 2019b;
Pascucci et al., 2019). However, adaptation effects in magnitude perception often show high-level properties. For instance, duration adaptation is not selective for the spatial position of the stimuli, highlighting a neural substrate beyond early, topographically-organized, visual areas (
Maarseveen, Hogendoorn, Verstraten, & Paffen, 2017). Numerosity adaptation generalizes not only across different stimulus formats, but also across different sensory modalities (
Arrighi, Togoli, & Burr, 2014) and across the perceptual and the motor systems (
Anobile, Arrighi, Togoli, & Burr, 2016;
Anobile, Domenici, Togoli, Burr, & Arrighi, 2020;
Togoli, Crollen, Arrighi, & Collignon, 2020). On the other hand, the effect of serial dependence in numerosity perception appears to be much more limited compared to adaptation, because it does not transfer across different sensory modalities (i.e., vision and audition;
Fornaciai & Park, 2019b). In line with these last empirical observations, the absence of cross-dimensional attractive biases in our study might be explained by assuming that serial dependence occurs within a dimension-specific processing pathway, upstream to the putative generalized magnitude system. The second possibility instead concerns the role of task set in limiting serial dependence to the task-relevant dimension only. Indeed, in our paradigm we probed duration and numerosity in two separate tasks, where each of them was the only dimension relevant for the task. The role of task-set in serial dependence has been shown by previous studies. For instance, Van der
Burg and colleagues (2019), in the context of face perception, showed that when participants were required to judge different aspects of a face in successive trials (i.e., gender in one trial and attractiveness in the next) serial dependence effects disappear. In a different study,
Fritsche and de Lange (2019) showed that feature-based attention driven by a task strongly affects the magnitude of the serial dependence effect. Namely, when participants judged the size, instead of the orientation, of a grating stimulus, the serial dependence effect on the orientation of the next stimulus was reduced. Differently from these previous studies, however, our task did not require an active judgement of the “past” stimulus inducing serial dependence (i.e., the inducer). Our results would thus further suggest that an explicit decision is not necessarily needed to influence the pattern of serial dependence, but the “task set” itself would be sufficient to modulate the effect. However, this only concerns the attractive effect between two successive stimuli, and not the repulsive effects such as adaptation (which may even be facilitated by the suppression of serial dependence; see also
Fornaciai & Park, 2019a).