In the research, a hypothesis that intelligent subjects are characterized by a more efficient attentional mechanism was proved. A new experimental task was exploited, which measures the ability to control two concurrent mental activities, but imposes minimal motor demands. Significant number of errors during the dual-task condition, and fast automatization were observed, which confirm theoretical aptness of the task as a measure of attention efficiency. An attempt was also made at avoiding limits of correlational methods based on tasks designed within attentional resources paradigm. Causal relationships were examined with a model of attentional mechanism based on ACT-R computational theory of mind. The model implements both parallel processes of selective attention and serial control on reaction selection, and explains 87% of the error variance observed in the experiment. Additional manipulation of a model's parameter, which reflects available attentional resource, enabled to explain 74% of error variance for 4 groups of subjects with different intelligence levels. The results of simulation suggest two-level causal relationship between intelligence and attention: intelligence as a trait influences attentional and memory processes, which then influence the efficiency of reasoning processes determining the level of psychometric intelligence.