This paper brings together results obtained by applying Magnetic Field Tomography (MFT) to the analysis of Magnetoencephalography (MEG) data over the last decade. It emphasizes the most recent developments where the availability of helmet-like MEG probes with well over 100 sensing coils provide a full coverage of the head. The paper shows that it is possible to extract tomographic information from single trial, millisecond by millisecond MEG signal, and demonstrates two ways that this capability can be exploited. First, the single trial reconstructions are used to obtain robust statistical measures of changes of activity over small latency windows. Second, the interaction between areas is studied by computing the mutual information between short, time-lagged sections of the single trial time-courses. The usefulness of the computationally demanding approach is demonstrated by analyzing experiments using two widely used protocols, one for face and affect recognition and the other for Contingent Magnetic Variation (CMV). The results show foci of significant changes of activity, which are consistent with what is reported in the literature and provide a deeper understanding of their significance. Some new, but not all that unexpected, findings also emerge from the analysis.