Human behaviour is exquisitely complex, because it is staggeringly multi-facetted and subtly varies across individuals. Characterising its yet incompletely understood underlying biological mechanisms has profound clinical implications. Here, our goal was thus to study inter-individual behavioural variability. To accomplish this feat, we monitored the dynamics of brain activity through functional magnetic resonance imaging. The acquired signals only indirectly reflect neuronal mechanisms, and can be modulated by several confounding factors. A notorious one is head motion of the scanned volunteers. As an initial step, we addressed to what extent such movement would relate to inter-individual behavioural variability. We revealed strong, yet overlooked ties between a broad range of anthropometric, cognitive, sensory and clinical measures, and spatio-temporal head motion specificities. Next, as behaviour is an intrinsically dynamic process, we scrutinised the literature to pinpoint tailored time-resolved approaches for the analysis of brain activity. In a first application, we opted to track the dynamic reconfigurations of functional interactions€ (or dynamic functional connectivity€”) between brain regions over time in the context of an audiovisual stimulation paradigm. To mitigate the deleterious influences of head motion, computations were performed across subjects (a process known as inter-subject functional correlation). We discovered that some cues from the movie would be perceived homogeneously by a healthy population, while others would not. In addition, typically developing individuals exhibited more homogeneous brain activity than subjects diagnosed with autism spectrum disorder. Furthermore, the latter differed in their individual response to the movie as a function of the extent and balance of their socio-communicative and repetitive behaviour symptoms. In a second step, we attempted to link behaviour and brain activity in the absence of an overt task. We decomposed the interactions of a dorsolateral prefrontal region of interest with the rest of the brain into a set of co-activation patterns, and observed that their expression in the resting-state could predict attentional performance. As the spatio-temporal complexity of functional brain dynamics transpired from this analysis, we then set to develop a novel sparse coupled hidden Markov model framework that enables to quantify the dynamic interplays that take place between brain networks. To try and minimise spurious effects driven by head motion, we also considered an alternative methodological avenue, called graph signal processing, which enables to inject knowledge regarding brain structure (known to constrain neuronal activity) into the analyses. By studying functional brain dynamics from this viewpoint, we revealed an extensive array of links to behaviour. Furthermore, we also extended the originally static approach to the dynamic setting, and by this mean, further broadened the set of unraveled brain/behaviour relationships. Finally, we looked back at all the methods that we applied in theoretical terms and in their ability to predict inter-individual differences in behaviour. We concluded that they render complementary aspects of functional brain dynamics, and that in future years, the neuronal significance of brain/behaviour relationships should be further strengthened by pursuing more extended multimodal analyses.