Neuroscientific studies often involve some form of group analysis over multiple participants. This requires alignment of recordings across participants. A naïve and commonly used solution is to assume that participants' recordings can be aligned anatomically in sensor space. However, this assumption is likely violated due to functional-anatomical organization differences between individual brains. In this work, multiple methods were compared to find a common representation of magnetoencephalographic (MEG) activations recorded from different participants performing a grasping task with three different grasps. Two most methods are based on anatomical alignment, namely alignment in sensor space and signal space separation (SSS, Taulu 2005). Three other methods (multiset canonical correlation analysis (MCCA), Kettenring 1971; group ICA, Calhoun 2009; Hyperalignment, Xu 2012) tested here seek to find a functional alignment by maximizing similarity of activation time-series between individuals in a latent space under different constraints. For the comparison of inter-individual information transfer across alignment methods, data were split between training and test sets of participants in a leave-one-subject-out fashion before finding the common space. Then, three class, single-trial inter-individual decoders were trained on the aligned data representation from the training set and the transfer of the decoders to data from the left-out participant was tested in the aligned space. The results show that functional inter-individual alignment (accuracy: MCCA, 0.65; ICA, 0.64; Hyperalignment, 0.60; chance level 0.33) outperform anatomical inter-subject alignment methods (sensor space, 0.45; SSS, 0.45). This indicates that the important features for across subject generalization lie within the latent functional spaces, while anatomical-functional representations can be idiosyncratic. Therefore, an approach to combine MEG data over individual brains should relax the assumption that brain anatomy and function are tightly linked. Potential applications are in the combination of large amounts of datasets and warm-starting brain-computer interfaces.