Indeed, when we considered only the subset of static trials preceded by another static trial, the mean timing error of the animate sessions was still significantly smaller than the mean timing error of the inanimate sessions (paired t-test, P<0.05).
Figure 4 plots the average difference between the value of the mean timing error in dynamic trials and the corresponding value in static trials.
This was shown by comparing the mean timing error over the first 10 repetitions starting from session onset with the mean timing error over the last 10 repetitions (two-tailed t-test, P>0.31), and by performing a linear regression of timing error as a function of all repetitions (P>0.23).
In contrast with the mean value of the timing errors, the variance across trials did not show any systematic trend as a function of experimental condition (Bartlett's χ2 test, P>0.18 in all but one subject, in whom the interaction between character type and figure type was significant, P<0.005).
(B) CMRglc was sensitive to timing errors.
First, diverter timing errors were estimated by linear regression.
Essentially, there are many sources of potential timing errors.
A two-way repeated-measures ANOVA ("outcome timing variability" × "trial type") revealed a significant interaction effect (F1,15 = 6.84, p = .019) and a main effect of "trial type" (F1,15 = 8.5, p = .011) on mean time judgement errors across subjects.
Following previous studies on binding of subjective action time (Haggard et al., 2002), we subtracted the mean time judgement error in "baseline" blocks from mean time judgement errors in "variable outcome timing" and "fixed outcome timing" blocks for each participant.
V. Dixon: Origin of timing error?
