Surprisingly, perceived speed increases as luminance decreases (Vaziri-Pashkam & Cavanagh, VSS 2006; Thompson et al, Vision Research 2006). Here we propose that the visible trail that follows a moving object is the cue to speed that causes the overestimation at low luminance. Motion smear lengthens at higher speeds, hence its value as a cue to speed, but smear must also lengthen with increased persistence – as is found at low luminance (Di Lollo & Bischof, Psychological Bulletin 1995). To demonstrate this with a moving stimulus, we presented a dot on a rotating disc viewed in continuous light (incandescent DC source) at two difference luminance levels. Subjects increased the rotation speed of the disc until the dot appeared to trace out a continuous circle (Ptolemy, 150). As expected, the criterion speed was slower (indicating greater persistence) at the lower luminance. In the second experiment, the subjects matched the perceived speed of two computer generated, rotating dot patterns, one at high luminance (1000 cd/m2) and one at low (1 cd/m2), while fixating between the two. When the rotating dots had long lifetimes, the low luminance pattern appeared to move faster than the high luminance pattern. However, when the dots had short life times, the speed difference was greatly reduced. Decreasing the dot lifetime decreases the possible length of the motion smear (the persisting trace cannot be longer than the dot’s path), removing this cue from the display. The results of these two experiments suggest a direct link between the amount of speed overestimation in dim light and the perceived motion smear.