Response selection in dual task paradigms: observations from random generation tasks.

Performance of attention-demanding tasks is worse if two tasks are carried out simultaneously than if each of the tasks is performed alone. Our aim was to determine whether these 'dual task costs' can be attributed to mechanisms on a supra-trial level such as switching of limited resources between trials or concurrent breakdown of supervisory functions, or to mechanisms effective within each trial such as demands of response selection. Twenty healthy volunteers performed verbal random number generation (RNG) and random movement generation (RMG) at three different rates. For each rate, both tasks were examined once in a single task condition and once in a dual task condition. Results showed that performance (quality of randomness) in each random generation task (RNG/RMG) was reduced at faster rates and impaired by concurrent performance of a secondary random generation task. In the dual task condition, transient increase or decrease of bias in one random generation task during any short interval was not associated with concurrent increase or decrease of bias in the other task. In conclusion, the fact that during dual task performance transient bias in one task was not associated with concurrent improvement of performance in the other task indicates that alternation of supervisory control or attentional resources from one to the other task does not mediate the observed dual task costs. Resources of the central executive are not re-allocated or 'switched' from one to the other task. Dual task costs may result from mechanisms effective within each trial such as the demands of response selection.