Despite a voluminous literature on the physiology of seizures, there remains no dynamical understanding of what a seizure is. Long before the extraction of neuronal interactions became technically feasible in experiments, Penfield and Jasper (1954) described seizures dynamics as “hypersynchronous”, and the monolithic description of seizures as "synchronous" neuronal discharge has entered our canon. However, we pose the hypothesis that, as in any description of the physics of ensembles, the macroscopic phenomena of a seizure needs to be characterized in terms of the interactions between the relevant neuronal subtypes taking part in this process. To dissect the specific roles of distinct cellular subtypes in seizures, I perform simultaneous multiple whole-cell recordings in inhibitory (I) and excitatory (E) neurons in the cortical rat brain slices during spontaneous in vitro seizures generated by the potassium channel blocker 4-aminopyridine. Two main findings emerged from these studies. First, using current clamp recordings, I observed a novel pattern of EI cell spiking activity interplay (Ziburkus et al., 2006). The second set of experiments using voltage clamp recordings allowed us to measure dynamic inhibitory (Gi) and excitatory (Ge) conductances in E and I cells during these seizures. We confirmed that, during the initialization of the seizures, changes in inhibitory conductance not only scales in amplitude with excitatory conductance, but also leads in time. During the persistent stage of the seizure, Gi transiently diminishes, and then increases with Ge during the post-seizure bursting. These findings support a natural partitioning of these in vitro seizures into at least three stages and underscore the complexity of cell-type specific mechanisms in the formation of seizure patterns.