The regression to right angles explanation of the Poggendorff illusion was tested by varying the distance apart of the parallels. The varying amount of illusion this produced was shown to be consistent with a constant angular distortion. When, however, the apparent angle of the transversal was also measured, this was not found to be sufficient to account for as much as half the illusion with a 30° transversal. It was shown also that the extra amount of illusion could not be accounted for by the Müller‐Lyer component of the Poggendorff illusion nor by any tendency to see the distance between the parallels as smaller than it was.