To investigate local neuronal activity changes and functional connectivity during cognitive and sensory stimuli, functional magnetic resonance imaging (fMRI) relies mainly on the haemodynamic response to brain activation and on measuring small fluctuations in the blood oxygenation levels. Since the basal amount of glucose and oxygen consumptions of the brain are much higher than the activity evoked by cognitive stimuli, the effect of spontaneous cerebral activity must be deduced to evaluate the fMRI responses to given cognitive tasks. We report that in our experiments with water flowing in pipes at an oblique angle with respect to the scan slice a diffuse response can be observed besides other flow artefacts, which can undermine fMRI interpretations. We point out that those artefacts are considered as such because hydrodynamic models disregard the appearance of flow in the images, and more generally the differences between mechanical representations and images.