Task-specific pre-processing included removal of initial volumes (Siemens 8TRs; Philips 8TRs; GE DV25 5 TRs; GE DV26 16TRs), normalization and demeaning. Estimation of task-related activation strength used a general linear model (GLM) using AFNIs 3dDeconvolve (Cox 1996), with nuisance regressors to model baseline, quadratic trend and motion (motion estimates, derivatives, and squared estimates and derivatives included); time points with frame-wise displacement (FD > 0.9mm) were censored (Siegel et al., 2014); the hemodynamic response function was modelled as a gamma function with temporal derivatives using AFNIs SPMG model. Events were modeled as instantaneous. The resulting GLM coefficients and t-statistics were sampled on to FreeSurfer-generated cortical surface projected 1mm into cortical grey matter.