, 2004). The remaining functional volumes were spatially realigned to the first image of the series, and distortion corrections were applied based on the field maps using the Unwarp routines in SPM (Andersson et al., 2001; Hutton et al., 2002). Each participant’s structural scan was then co-registered to a mean image of their realigned, distortion-corrected functional scans. The structural images were segmented into grey matter (GM), white matter (WM), and cerebral spinal fluid using the New Segment tool within SPM8. The

DARTEL normalization process was then applied to the GM and WM segmented images, which iteratively warped the images into a common space using nonlinear registration (Ashburner, 2007). Using the output of this nonlinear warping process, all functional SB431542 and structural images were normalised to MNI space using DARTEL’s ‘Normalise to MNI’ tool. The functional images were smoothed using a Gaussian kernel with full-width at half maximum of 8 mm. Structural MRI scans were analysed using voxel-based morphometry (VBM; Ashburner and Friston, 2000, 2005) implemented in SPM8, employing a smoothing kernel of 8 mm full-width at half maximum. For a priori ROIs (HC, PHC and RSC – see Section 2.7), we applied a statistical threshold of p < .001 uncorrected

for multiple comparisons. For the rest of the brain, we employed a family-wise error (FWE)-corrected threshold of p < .05. We searched for structural correlates of individual differences in BE, and found no significant selleck compound effects in the MTLs, or elsewhere in the brain. Statistical analysis of the fMRI data was applied to the

pre-processed data using a general linear model. The primary analysis involved a comparison of activity elicited by the first scene presentation on trials where BE occurred and those first presentation trials where it did not. To do this, we used each participant’s behavioural data in order to divide the trials into those where BE occurred (all trials where the second scene was judged to be closer than the first – the BE condition), and those where it did not occur (the Null condition). The Null 4��8C condition consisted of trials where the second scene was judged to be the same or further away than the first, as in both cases BE did not occur. By pooling across both types of Null trial in this way, we increased the power of the analysis. We used a stick function to model the onset of each first scene presentation, dividing the trials into two conditions based on the subsequent behavioural choice data, thus creating a BE regressor and a Null regressor. These stick functions were convolved with the canonical haemodynamic response function and its temporal derivative to create the two regressors of interest. We also used a stick function to model the second scene presentations, dividing them into BE and Null conditions, which were included as regressors of no interest.