This paper proposes a new theory to account for the effects of underload on performance.Malleable attentional resources theory posits that attentional capacity can change size in response to changes in task demands. As such, the performance decrements associated with mental underload can be explained by a lack of appropriate attentional resources.These proposals were explored in a driving simulator experiment. Vehicle automation was manipulated at four levels, and mental workload was assessed with a secondary task.Eye movements were also recorded to determine whether attentional capacity varied with mental workload. The results showed a clear decrease in mental workload associated with some levels of automation. Most striking, though, were the results derived from the eye movement recordings, which demonstrated that attentional capacity varies directly with level of mental workload. These data fully supported the predictions of the new malleable attentional resources theory. The implications of this theory are discussed with regard to capacity models of attention, as well as to the design of future vehicle systems.