A task requiring the mental rotation of letters was performed either alone or during the retention interval of two short-term memory tasks. The retention of eight digits or random configurations of eight dots slowed overall reaction time on the mental-rotation task, but did not significantly alter the estimated rate of mental rotation. Conversely, performance on the memory tasks did not vary with the angle of mental rotation. These results support the view that attentional control is required to set up the mental structures required in mental rotation, but that the actual execution of the mental-rotation component can be relegated to subordinate systems that do not compete for attentional resources.