In dual-tasking, the cognitive capacity for task processing is limited, resulting in the well-researched psychological refractory period (PRP) effect, in which the the response to the second of two tasks presented in quick succession is delayed more the less time there was between the stimulus presentations. This effect is often interpreted as evidence for serial processing where the processing of the second task can only start once the first task was successfully processed.

Two types of theories make different claims about whether it is possible to process the second task while the first task is processed. The structural bottleneck theory states that cognitive capacity is distributed in an all-or-nothing manner while the capacity sharing theory states that cognitive capacity can be be distributed gradually.

Here we investigate these two hypotheses in three experiments (N=155, N=183, in data acquisition) that aim to manipulate the sharing factor, using spatially distinct congruency rates and different stimulus onset asynchrony frequencies. Because the capacity sharing theory allows for a strategic bottleneck to form dynamically, changes might only occur in parts of the population.

We use a Bayesian multilevel modeling approach to distinguish interindividual variability in the 'true' effect sizes from measurement noise in order to identify whether there are individuals with an RT1 PRP effect. We derive immediate theoretical constraints from the model comparisons and discuss further implications.