Head motion during resting-state functional magnetic resonance imaging acquisitions is an infamous confound requiring dedicated preprocessing steps. Here, we probed the level of complexity at which motion should be characterised. To do so, we designed a clustering-based approach that determines the spatio-temporal motion patterns that can fingerprint an individual. We found that each subject's motion could be fingerprinted at a different space/time granularity, some more easily than others. Our results call for refined motion descriptions in which space and time should not be over-simplified, and point towards the relevance of motion-related fingerprints as an individual's functional trait.