The aim of this chapter is to show how the validity of neuropsychological models of attention can be explicitly tested using advanced techniques for the analysis of event-related activity in electroencephalography (EEG) and magnetoencephalog-raphy (MEG). These techniques are based on biophysical models of EEG/MEG which afford a neurobiological perspective on event-related-potential (ERP) research and on cognitive neuroscience in general. In particular, they allow one to reinterpret neurodynamical effects of attention in terms of context-dependent changes in neuro-nal couplings between remote regions embedded in a global network of attention. First, we present the nodes of the network hypothesized and what the relationships are between attention, synchronization, neural coupling, and ERPs. Second, we describe briefly the mathematical details of a recent modeling approach (dynamic causal modeling) to estimate neural couplings from ERPs. Finally, we will show how dynamic causal modeling for ERPs can be used to compare different neuropsycho-logical models using two examples: the mismatch negativity in auditory oddball paradigms and the activation of the ventral visual pathway by emotional attention.