In a complex system such as the brain, the coordination of multiple areas as quantified, e.g., by measures of coupling, phase-locking, synchronization, etc., is fundamental to the system’s function. Unfortunately, the properties of the EEG signal (electrical field and volume conduction) create spatial correlations which introduce severe biases in coordination measures.. When the signal is dominated by a single active source, spatial correlation creates spurious inphase and antiphase synchronization. In addition, when other sources are recruited without being synchronized (co-active sources) the dynamics of phase correlation gives rise to spurious inphase and antiphase tendencies, with noticeable impact on quantitative measures of coupling. Finally, when multiple sources are genuinely synchronized, their real amplitude and phase properties are masked, and a strong tendency for zero-lag synchronization is observed. Such zero delay coordination may not be genuinely present between the sources. As a result of these biases, studying coherence on the raw EEG signal is untenable, especially if real synchrony is a transient phenomenon in brain dynamics. To remedy the problem of spatial correlation, two routes are possible: assessing synchronization in mathematical estimates of source activity (inverse problem) or learning to distinguish real and false synchronization in raw EEG data (predictive forward approach). We developed a framework along the latter route. Forward models provide the distinct signatures of real and false synchronization in the dynamics of amplitude, phase and frequency. Colorimetric mapping of continuous EEG is used to identify episodes of apparent locking. Properties of these episodes are then studied to identify real synchrony. This framework opens up the possibility to test theoretical models of brain coordination dynamics in particular the metastable régime which is a subtle blend of integration and segregation. [PPT].