Skip to main content
. Author manuscript; available in PMC: 2011 Jan 1.
Published in final edited form as: J Vis. 2010 Sep 1;10(11):8. doi: 10.1167/10.11.8

Table 8.

Confusion matrix, derived from the empirical data in Table 3, but corrected for the lapse rate and picture-height confusions using the methods described in Appendix 2. The rows are stimuli (Target Presented) and the columns are responses (Subject Response). The cells show percentages of response X given stimulus Y. The upper values in each cell are the corrected values. The values in parentheses are the model fits. The correlations between corrected values and model values are 0.996 (computed across all 25 cells), 0.911 (computed across the five cells on the diagonal, representing the accuracies for the five targets), and 0.896 (computed across the 20 off-diagonal cells). All three correlations are statistically significant (P < .05).

Subject Response (%)
Step Up Step Dn Ramp Up Ramp Dn Flat
Target Presented (%) Step Up 92.9 (93.1) 0.48 (0) 4.83 (4.30) 0.00 (0) 1.81 (2.58)
Step Dn 0.36 (0) 69.6 (73.0) 3.01 (0) 14.1 (16.9) 12.9 (10.1)
Ramp Up 1.08 (0) 1.93 (0) 80.1 (81.9) 3.38 (0) 13.5 (18.1)
Ramp Dn 0.84 (0) 8.20 (0) 3.26 (0) 74.3 (81.9) 13.4 (18.13)
Flat 1.69 (0) 4.35 (0) 4.71 (0) 0.00 (0) 89.1 (100)