R. Henson , J. Mattout , K. Friston , S. Hassel , A. Hillebrand, G.R. Barnes & K.D. Singh 1. MRC Cognition & Brain Sciences Unit, Cambridge, England, 2. Functional Imaging Laboratory, University College London, England, 3. Wellcome Trust Laboratory for MEG Studies, Aston University, England Introduction We tested recent methods for distributed source localisation using Restricted Maximum Likelihood (ReML) [1] on an MEG dataset that isolated the M170 component believed specific to faces. Unlike previous methods, multiple constraints on the covariance of the sources and/or sensor noise can be applied, and their weightings (hyperparameters) estimated automatically from the data.