BACKGROUND: Limited information processing capacity in the brain necessitates task prioritisation and subsequent adaptive behavioural strategies for the dual-task coordination of locomotion with severe concurrent cognitive loading. Commonly observed strategies include prioritisation of gait at the cost of reduced performance in the cognitive task. Alternatively alterations of gait parameters such as gait velocity have been reported presumably to free processing capacity for the benefit of performance in the cognitive task. The aim of this study was to describe the neuroanatomical correlates of adaptive behavioural strategies in cognitive-motor dual-tasking when the competition for information processing capacity is severe and may exceed individuals' capacity limitations. METHODS: During an fMRI experiment, 12 young adults performed slow continuous, auditorily paced bilateral anti-phase ankle dorsi-plantarflexion movements as an element of normal gait at .5 Hz in single and dual task modes. The secondary task involved a visual, alphabetic N-back task with presentation rate jittered around .7 Hz. The N-back task, which randomly occurred in 0-back or 2-back form, was modified into a silent counting task to avoid confounding motor responses at the cost of slightly increasing the task's general coordinative complexity. Participants' ankle movements were recorded using an optoelectronic motion capture system to derive kinematic parameters representing the stability of the movement timing and synchronization. Participants were instructed to perform both tasks as accurately as possible. RESULTS: Increased processing complexity in the dual-task 2-back condition led to significant changes in movement parameters such as the average inter-response interval, the coefficient of variation of absolute asynchrony and the standard deviation of peak angular velocity. A regions-of-interest analysis indicated correlations between these parameters and local activations within the left inferior frontal gyrus (IFG) such that lower IFG activations coincided with performance decrements. CONCLUSIONS: Dual-task interference effects show that the production of periodically timed ankle movements, taken as modelling elements of the normal gait cycle, draws on higher-level cognitive resources involved in working memory. The interference effect predominantly concerns the timing accuracy of the ankle movements. Reduced activations within regions of the left IFG, and in some respect also within the superior parietal lobule, were identified as one factor affecting the timing of periodic ankle movements resulting in involuntary 'hastening' during severe dual-task working memory load. This 'hastening' phenomenon may be an expression of re-automated locomotor control when higher-order cognitive processing capacity can no longer be allocated to the movements due to the demands of the cognitive task. The results of our study also propose the left IFG as a target region to improve performance during dual-task walking by techniques for non-invasive brain stimulation.