Identical stimuli receive higher category ratings when presented in a positively skewed distribution (smaller stimuli occur more often than larger) and lower ratings in a negatively-skewed distribution (larger stimuli occur more often). This frequency effect has been explained by a tendency to assign the same categories to the same stimuli (‘consistency model’ by Haubensak, 1992 Journal of Experimental Psychology: Human Perception and Performance18 303) rather than by a tendency to assign the same number of stimuli to each category. The consistency model postulates the prime importance of stimuli that occur early in a sequence, with the underlying assumption that more frequent stimuli have higher probability of occurring earlier. Thus, presentation order rather than frequency of stimuli might be decisive. We asked whether a ‘frequency’ effect would still obtain when stimuli are in fact presented with equal frequency, but in a sequence derived from positively or negatively skewed source samples. Ratings of visual velocity were obtained for a dot that started to move at 0.5 deg below fixation in leftward or rightward direction for 2, 2.5, or 3 deg. Five velocities (3.0, 4.5, 6.0, 7.5, 9.0 deg s−1) were presented 10 times each (in a balanced combination of stimulus duration and displacement) and were rated by 32 subjects using three categories (slow, medium, fast). Our results clearly support Haubensak's model: identical velocities were rated higher/lower in presentation sequences that mimicked the order effects of positive/negative skewing, respectively. Moreover, computer simulation of the Haubensak model reveals good agreement between predicted and observed results.