Analysis of individual variations in the classical horizontal-vertical illusion.

In the horizontal-vertical illusion (HVI), the length of the vertical line is overestimated, whereas in the bisection illusion (BI), the horizontal bisecting line is expected to be overestimated. Here, only half of our 22 observers showed the expected BI, whereas the other half underestimated the bisecting line. Observers also differed in their judgments of the strength of the HVI: The HVI was stronger for observers showing the classical bisection effect, and weaker or absent for those underestimating the bisecting line. To account for these results, we used a linear model to individually estimate the strength of two putative factors underlying both illusions. Whereas the strength of the HVI and BI were highly correlated, the estimated factors were uncorrelated. Therefore, in two control experiments, we then measured the pure horizontal-vertical (pHVI) and bisection (pBI) illusions. A significant correlation between the estimated factors and the measured illusion variants was found. Results were robust against variations of contrast, repetitive presentations, and choice of adjusted line. Thus, the classical HVI as an additive combination of two independent factors was confirmed, but we found considerable interindividual variations in the strength of the illusions. The results stress the importance of analyzing individual data rather than taking sample means for understanding these illusions.