The problem of low variance voxels in statistical parametric mapping; a new hat avoids a 'haircut'

► Statistical parametric mapping judges significance with a signal-to-noise ratio. ► Low noise, e.g. outside the brain, can yield artefactually high statistical values. ► Spatial smoothing can shift peaks substantially towards regions of low variance. ► Source-reconstructed M/EEG data exhibits the problem particularly severely. ► The problem can be addressed by modifying the noise variance estimate.