The magnocellular deficit theory is one of the prominent hypotheses in dyslexia research. However, recent studies have produced conflicting results. Ten dyslexic children and 12 controls were examined with visual evoked potentials elicited by random dot kinematogram. The experiment comprises two sequences, one with randomly moving dots (control condition) and a second sequence where a fraction of the dots were moved coherently at the left or right side (depending on the level of coherence, 10%, 20%, and 40% of the dots). Randomly moving dots elicited two components, a P100 and P200, which were not different between the groups. Coherently moving dots elicited a late positivity between 300 and 800 ms, which was significantly attenuated in dyslexic children. The area of this component becomes larger at a higher level of coherence. This study supports the hypothesis of an impairment of a specific magnocellular function in dyslexia.