Cerebral currents responsible for the extra-cranially recorded magnetoencephalography (MEG) data can be estimated by applying a suitable source model. A popular choice is the distributed minimum-norm estimate (MNE) which minimizes the l2-norm of the estimated current. Under the l2-norm constraint, the current estimate is related to the measurements by a linear inverse operator. However, the MNE has a bias towards superficial sources, which can be reduced by applying depth weighting. We studied the effect of depth weighting in MNE using a shift metric. We assessed the localization performance of the depth-weighted MNE as well as depth-weighted noise-normalized MNE solutions under different cortical orientation constraints, source space densities, and signal-to-noise ratios (SNRs) in multiple subjects. We found that MNE with depth weighting parameter between 0.6 and 0.8 showed improved localization accuracy, reducing the mean displacement error from 12 mm to 7 mm. The noise-normalized MNE was insensitive to depth weighting. A similar investigation of EEG data indicated that depth weighting parameter between 2.0 and 5.0 resulted in an improved localization accuracy. The application of depth weighting to auditory and somatosensory experimental data illustrated the beneficial effect of depth weighting on the accuracy of spatiotemporal mapping of neuronal sources.