Abstract Cortical excitability is assumed to depend on cortical arousal level in an inverted U-shaped fashion: Largest (optimal) excitability is usually associated with medium levels of arousal. It has been proposed that under conditions of low arousal, compensatory effort is exerted if attentional demands persist. People tend to avoid this resource-consuming top–down compensation by creating or selecting environmental conditions that provide sufficient bottom–up stimulation. These assumptions were tested in an attention-demanding dual-task situation: We combined a simulated driving task to induce three different arousal levels by varying stimulation (high vs. low vs. self-chosen) with a visual two-stimulus paradigm to assess cortical excitability by the initial contingent negative variation (iCNV) component of the event-related potential. Additionally, we analyzed the oscillatory power of the beta2 band of the electroencephalogram at anterior frontal sites, which is assumed to reflect low-arousal compensatory activity. The iCNV amplitude differed in all three arousal conditions as expected: It was highest in the condition of self-chosen stimulation and lowest in the low- and high-arousal conditions. Additionally, in the low-arousal condition, anterior frontal beta2 power was found to be significantly higher than in the other two conditions and correlated positively with subjective strain. This pattern of results suggests that subjects select medium levels of stimulation which optimize cortical excitability under attentional demand conditions. The elevated fronto-central beta2 power in the low-stimulation condition may indicate the involvement of the anterior cingulate cortex in compensating for reduced arousal by top–down stimulation of the noradrenergic arousal system.