The Ebbinghaus illusion, in which a central target surrounded by larger context figures looks smaller than when surrounded by smaller context figures, is usually classified as a size contrast illusion. Thus "size contrast" is the dominant account of this effect. However, according to an alternative "contour interaction" account this phenomenon has little to do with size contrast but is rather caused by distance-dependent attractive and repulsive interactions between neural representation of contours. Here evidence is presented against the size contrast account and consistent with the contour interaction account. Experiment 1 was a control study confirming that the illusion can be obtained using displays consisting only of squares, which are more convenient to manipulate than the standardly used circles. In Experiment 2, the standard configuration involving small context figures surrounding the target was compared to a novel configuration, which involved many "spread" small context figures. The illusory effect of the standard context was stronger than the illusory effect of the spread context, in accord with the prediction of the contour interaction account, and contrary to the prediction of the size contrast account. In Experiment 3 two novel configurations were used, based on standard and spread contexts. The results were in accord with the prediction of the contour interaction account, whereas the size contrast account had no prediction because the stimuli did not involve conventional size contrast. Additional aspects of the stimuli and an account of the illusion based on a perspective interpretation are also discussed.