Xuelian Zang1,2, Xiuna Zhu2, Fredrik Allenmark2, Jiao Wu1, Stefan Glasauer3, Hermann J. Müller2, Zhuanghua Shi2
- Center for Cognition and Brain Disorders, Affiliated Hospital of Hangzhou Normal University, 310015, China.
- General and Experimental Psychology, Department of Psychology, LMU Munich, 80802, Germany
- Institut für Medizintechnologie, Brandenburgische Technische Universität Cottbus-Senftenberg, 03046, Cottbus
Duration estimates are often biased by the sampled statistical context, yielding the classical central-tendency effect, i.e., short durations are over- and long duration underestimated. Most studies of the central-tendency bias have primarily focused on the integration of the sensory measure and the prior information, without considering any cognitive limits. Here, we investigated the impact of cognitive (visual working-memory) load on duration estimation in the duration encoding and reproduction stages. In four experiments, observers had to perform a dual, attention-sharing task: reproducing a given duration (primary) and memorizing a variable set of color patches (secondary). We found an increase in memory load (i.e., set size) during the duration-encoding stage to increase the central-tendency bias, while shortening the reproduced duration in general; in contrast, increasing the load during the reproduction stage prolonged the reproduced duration, without influencing the central tendency. By integrating an attentional-sharing account into a hierarchical Bayesian model, we were able to predict both the general over- and underestimation and the central-tendency effects observed in all four experiments. The model suggests that memory pressure during the encoding stage increases the sensory noise, which elevates the central-tendency effect. In contrast, memory pressure during the reproduction stage only influences the monitoring of elapsed time, leading to a general duration over-reproduction without impacting the central tendency.
Keywords: time perception, dual-task performance, attention-sharing, cognitive/memory load, Bayesian integration, central-tendency effect