Data were processed and analyzed using Analysis of Functional NeuroImages (AFNI; afni.nimh.nih.gov; Cox, 1996). Motion in the time series data was corrected by registering each acquisition to a minimally deviant repetition with an iterated least squares algorithm (Cox and Jesmanowicz, 1999); 3 rotational and 3 displacement parameters were output for each repetition of each participant. Trained raters examined the corrected time series data and removed repetitions containing residual visually discernible head motions. If >22% of repetitions were removed, the participant was not included (n = 5 not described in this paper). On average, 6% of repetitions were removed due to movement. Time series data were deconvolved with a reference vector coding the alternating task conditions while covarying for linear trends and the motion correction applied to control for spin history effects (Bandettini et al., 1993), and modeling typical delays in the hemodynamic response (Bandettini et al., 1993; Boynton et al., 1996). Functional data were transformed into standardized space (Lancaster et al., 2000; Talairach and Tournoux, 1988) and resampled into 3.5 mm3 voxels, and a spatial smoothing Gaussian filter (full-width half
