Abstract
In the VSS ’Best Illusion ‘ contest 2010, Thompson demonstrated that a face presented upside-down looked markedly thinner than when upright. We now present experiments to investigate this effect. Experiment 1 measured the size of the illusion in 14 subjects with a method of constant stimuli. Comparisons with two upright faces (UU), two inverted faces (II) and one upright and one inverted UI) were made so that psychometric functions could be determined and points of subjective equality (PSE) calculated. This experiment established that upright and inverted faces are not significantly different in their width discriminability, and that an inverted face looks significantly narrower than its upright counterpart. In Experiment 2 we determined that the internal features of the face are responsible for the effect. Again, psychometric functions and PSEs were determined for UU, II and UI conditions, in 13 subjects. The results showed that removing the internal features abolished the illusion. Experiment 3 tested the hypothesis that inverting faces disables the ‘holistic’ encoding we use for upright faces and hence inverted faces revert to some mean size. We predicted therefore that whereas fat faces look thinner upside-down, thin faces should look fatter upside-down. Nineteen subjects adjusted the width of 12 inverted faces until they appeared to match their upright version. The faces selected varied from those with the widest internal features to those with the narrowest. The results showed that only the two narrowest faces appear fatter when inverted; the remaining 10 faces looked significantly thinner inverted. This suggests that inverted faces do not regress to some ‘average’. We shall propose a model that interprets the internal features of the face as a version of the horizontal-vertical illusion, for which it has been reported that a great illusion is seen in an inverted-T than an upright-T (Lee & Freire 1999).
Meeting abstract presented at VSS 2012