We were interested in the magnitude of the estimation error when only reliant on the VE rather than the direction of the respective errors. The percent error of the target location was thus calculated from the absolute bias measures of the psychometric functions for each subject and summarized in
Figure 5. In
Figure 5A, the VE + CD percent error in target location is plotted as a function of the corresponding VE-based percent error. First, when considering the condition in which trans-saccadic shifts occurred at the fovea (left panel,
Figure 5A), VE-based percent errors were significantly greater than VE + CD based percent errors across both directions and amplitudes (three-way ANOVA, main effect of CD benefit, F
1,11 = 55.45,
p < 0.001, η
2p = 0.83) (Note that the majority of the data and the respective mean values (represented by the crosses) across subjects are below the unity line). Overall, when collapsing across amplitude and direction, this perceptual estimate at the fovea was approximately 53% more accurate (VE + CD mean = 5.80 vs. VE-only mean = 12.43) and 30% less variable (VE + CD SD = 4.29 vs. VE-only SD = 6.16) than estimates based solely on visual information. The magnitude of percent error was lower for 8° saccadic eye-movement targets than for saccades to the 4° saccadic eye-movement targets as evidenced by a main effect of amplitude (F
1,11 = 4.91,
p = 0.049, η
2p = 0.32), but percent error was similar across horizontal and vertical conditions (main effect of saccade direction, F
1,11 = 0.37,
p = 0.55, η
2p = 0.03). None of the two- or three-way interactions were statistically significant (
p > 0.1 for all cases). Similarly, when considering peripheral shifts, VE-based percent errors were significantly greater than VE + CD based percent errors across both directions and amplitudes (three-way ANOVA, main effect of CD benefit (F
1,11 = 122.93,
p < 0.001, η
2p = 0.92) (As in the left panel, here again the majority of the data and the respective mean values across subjects are below the unity line). For peripheral shifts, when collapsing across amplitude and direction, the perceptual estimate was approximately 34% more accurate (VE + CD Mean = 7.65 vs. VE-only mean = 11.62) and 9.5% less variable (VE + CD SD = 3.98 vs. VE-only SD = 4.41) than estimates based solely on visual information. The magnitude of percent error was significantly lower for 8° saccadic eye-movement targets than for saccades to the 4° saccadic eye-movement targets (main effect of amplitude, F
1,11 = 8.76,
p = 0.013, η
2p = 0.44), and slightly greater for vertical saccadic targets, but this main effect of direction did not reach significance (F
1,11 = 3.51,
p = 0.09, η
2p = 0.24). Here, like for the foveal shifts, none of the two- or three-way interactions were significant (
p > 0.15 for all cases)