The method of average error was used with a mixed design to measure the horizontal-vertical illusion (HVI) for 40 Ss. Six stimulus configurations (L, J , 1-, -I, 1, +) were combined with seven angular orientations of the upright standard, and on each trial the variable horizontal was adjusted to appear equal to the standard in length. Results showed that for no stimulus configuration did the vertical orientation of the standard yield the greatest illusion. The magnitude of the HVI was dependent upon the stimulus configuration, upon the orientation of the standard, and upon an interaction between these variables. For the 1 and +, equal inclinations of the standard to either side of the vertical yielded equal effects; for the other figures, asymmetrical effects were produced. The results are discussed in relation to the perspective theory of visual illusions.