Changes in the activity of sensory neurones carry information about a given stimulus. However, neuronal activity changes may also arise from noise sources within or outside the nervous system. Here, the reliability of encoding of visual motion information is analysed in the visual motion pathway of the fly and compared to the findings obtained in other animal species. Several constraints determine and limit the reliability of encoding of visual motion information: (i) the biophysical mechanisms underlying the generation of action potentials; (ii) the computations performed in the motion vision pathway; and (iii) the dynamical properties of motion stimuli an animal encounters when moving around in its natural environment. The responses of fly motion-sensitive neurones are coupled to visual motion on a timescale of milliseconds up to several tens of milliseconds, depending on the dynamics of the motion stimuli. Only rapid velocity changes lead to a precise time-locking of spikes to the motion stimuli on a millisecond scale. Otherwise, the exact timing of spikes is mainly determined by fast stochastic membrane-potential fluctuations. It is discussed on what timescale behaviourally relevant motion information may be encoded.