Mass responses that are obtained using electrophysiological or psychophysical techniques are inadequate to characterize motion detectors at the single-unit level. Therefore, we have modelled a population of motion detectors and fitted their mass response to motion-onset EEG data. By examining a single unit of the modelled population we could assess the range of directions and the level of adaptation an individual motion detector responded to. Stimuli were patterns of randomly distributed dots. After subjects had adapted to either a non-moving pattern or to motion in one fixed direction, the pattern moved in one of seven possible test directions. The population model assumed elementary motion detectors with a Gaussian angle-selectivity profile. The population response was calculated as the sum of all adaptation-weighted individual responses to the test direction. In all subjects, we observed sizeable effects of adaptation on test directions close to the adapted direction and small amounts of direction-independent adaptation for all test directions. The model fit explained 75% (Oz) and 94% (lateral derivation) of the total variance and revealed a motion detector bandwidth (full width at half maximum) of 72 degrees (Oz) and 62 degrees (lateral derivation). The maximal adaptation depth was estimated as 72% (Oz) and 66% (lateral derivation). Estimates inferred from the population model representing human motion detectors were found to be close to those obtained by single-cell data from non-human primates.