Experiments and Computational Models for the Ames Window Illusion

T V Papathomas1, M Karakatsani2, S M Silverstein3, N Baker4, M de Heer5

1Center for Cogn. Science/ Lab Vision Research, Rutgers University, NJ, United States
2Dept Biomedical Engineering, Rutgers University, NJ, United States
3Division of Schizophrenia Research, University of Medicine and Dentistry of NJ, NJ, United States
4Department of Cognitive Science, Johns Hopkins University, NJ, United States
5retired, Nieuwezijds Voorburgwal 173 1012 RK, Amsterdam, Netherlands
Contact: papathom@rci.rutgers.edu

Purpose: To examine systematically factors affecting the Ames Window illusion, toward a future study comparing schizophrenia patients and controls; to produce stimuli that span the range from extremely weak to extremely strong illusions; to develop models that predict performance by assigning appropriate weights to the examined factors. Methods: Factors examined: (1) Long-to-Short base ratio “LS”, (2) Height-to-Short base ratio “HS”, (3) presence of Shadows “SH”. These factors were varied systematically to produce nine rotating stimuli; these were used in two experiments to assess illusion strength using two measures: Asking observers to report (A) which side was in front at selected instances; (B) reversals in rotation direction. The data were fed to an algorithm to determine optimal weights. Results: The two measures produced results that agreed closely, thus confirming the validity of the methods. The optimization algorithm yielded weights that produced significant correlations with the experimental data. Conclusions: Illusory strength increased primarily with growing LS ratio, followed by the presence of shadows, then with decreasing HS ratio. The results have set the stage for the next step: testing schizophrenia patients and controls to test for potential differences in the top-down bias for perceiving frontoparallel trapezoids as rectangles slanted in depth.

Up Home