How the brain transitions into and out of a seizure is mysterious. Using the intact mouse hippocampus preparation, recurrent seizure-like events (SLEs) in low Mg2+/high K+ perfusate were measured in the CA3 region. The SLE was characterized by a "preictal phase", which abruptly turns into a higher frequency "ictal" phase. Blockade of GABAA receptors shortened the preictal phase, abolished interictal bursts, attenuated the slow preictal depolarization, but with no effect on the ictal duration. On the other hand, SLEs were blocked by glutamate receptor blockade, whereas the interictal activity remained intact. In CA3 pyramidal cells and stratum oriens non-fast and fast spiking interneurons, recurrent GABAergic inhibitory postsynaptic currents (IPSCs) predominated interictally and during the early preictal phase, synchronous with extracellularly measured recurrent field potentials (FPs). These IPSCs then decreased to zero or reversed polarity by the onset of the higher frequency ictal phase. However, postsynaptic muscimol-evoked GABAA responses remained intact. Simultaneously, excitatory postsynaptic currents (EPSCs) synchronous with the FPs, markedly increased to a maximum at the ictal onset. The reversal potential of the compound postsynaptic currents (PSCs: combined simultaneous EPSCs and IPSCs) became markedly depolarized during the preictal phase, but the muscimol-evoked GABAA reversal potential remained unchanged, implying increasing glutamatergic input during this phase and not a further depolarization of the GABAA reversal potential. During the late preictal phase, interneuronal excitability was high, but IPSCs, evoked by local stimulation, or osmotically by hypertonic sucrose application, were diminished, disappearing at the ictal onset. EPSCs evoked by hypertonic sucrose application, were maximal at ictal onset, disappearing at the the end of the ictus. We conclude that the interictal and early preictal states are dominated by GABAergic activity, with the onset of the ictus heralded by exhaustion of presynaptic release of GABA, and unopposed increased glutamatergic responses. The ictus stops when presynaptic release of glutamate is exhausted. Supported by the CIHR.