Fractals - signals that display scale-invariant behaviour - are ubiquitous in nature including a wide variety of physiological processes. Fractal analysis of blood oxygen level dependent (BOLD) time-series of fMRI acquisitions from the brain can be achieved by decomposing the data Into a hierarchy of temporal scales so that although the signal may well be irregular and contain singularities, the properties of these singularities are constant in time and the entire series can be characterised by a single scaling exponent: the Hurst exponent, H. The observation that a signal has a noninteger fractal dimension suggests that the generating system is complex and has the potential to adapt to a wide variety of challenges. In contrast, the emergence of white noise or, alternatively, signal periodicity can be seen as degradation of fractal complexity and hence, maladaptivity. We tested the hypothesis that exogenous stimuli affects fractal signal properties In the context of brain function by manipulating the cognitive demand of a working memory task and using H as a summary measure of signal complexity. We show that this stimulus has a significant effect on H estimated from resting data acquired immediately before and after the task, and that the degree of change is related to cognitive load.