We analyzed the mean heading bias with respect to the independent variables using a three-way repeated measures ANOVA (6 switch angles
\(\times\) 3 preswitch headings
\(\times\) 2 postswitch durations). The analysis revealed a significant main effect of preswitch heading,
\(F\)(2, 22) = 5.746,
\(p\) = 0.010,
\(\eta _p^2 = 0.343\), a significant main effect of switch angle,
\(F\)(1.3, 14.34) = 14.525,
\(p\) = 0.001, Greenhouse-Geisser corrected,
\(\eta _p^2 = 0.569\), and a significant interaction between switch angle and postswitch duration,
\(F\)(5, 55) = 8.677,
\(p\) \(\lt \) 0.001,
\(\eta _p^2 = 0.441\). The significant interaction prompted us to follow up with separate one-way repeated measures ANOVAs for the two postswitch durations (
Figure 3b). We found a significant effect of switch angle for both 250-ms,
\(F\)(1.39, 15.26) = 19.284,
\(p\) \(\lt \) 0.001, Greenhouse-Geisser corrected,
\(\eta _p^2 = 0.637\), and 500-ms,
\(F\)(1.58, 17.34) = 7.807,
\(p\) = 0.006, Greenhouse-Geisser corrected,
\(\eta _p^2 = 0.415\), postswitch durations. To examine whether the mean bias increases with switch angle, we collapsed across switch angle sign and performed Tukey honest significant difference (HSD) post hoc tests on the bias within each postswitch duration group. For the 250-ms postswitch duration, there was a significant difference between the 12
\(^\circ\) absolute switch angle and the other absolute switch angles ( 3
\(^\circ\) vs. 12
\(^\circ\):
\(p\) \(\lt \) 0.001, 6
\(^\circ\) vs. 12
\(^\circ\):
\(p\) = 0.003). There was no significant difference between the 3
\(^\circ\) and 6
\(^\circ\) conditions (
\(p\) = 0.120). For the 500-ms postswitch duration, there was only a significant difference between the 3
\(^\circ\) versus 12
\(^\circ\) absolute switch angles (3
\(^\circ\) vs. 12
\(^\circ\):
\(p\) \(\lt \) 0.025, 6
\(^\circ\) vs. 12
\(^\circ\):
\(p\) = 0.208, 3
\(^\circ\) vs. 6
\(^\circ\):
\(p\) = 0.632). This analysis indicates an increase in bias with absolute switch angle for both postswitch durations, at least between the smallest and largest absolute switch angles considered here.