Early seizure-prediction paradigms were based on detecting electroencephalographic (EEG) features, but recent approaches are based on dynamic systems theory. Methods that attempted to detect predictive features during the preictal period proved difficult to validate in practice. Brain systems can display bistability (both normal and epileptic states can coexist), and the transitions between states may be initiated by external or internal dynamic factors. In the former case prediction is impossible, but in the latter case prediction is conceivable, leading to the hypothesis that as seizure onset approaches, the excitability of the underlying neuronal networks tends to increase. This assumption is being explored using not only the ongoing EEG but also active probes, applying appropriate stimuli to brain areas to estimate the excitability of the neuronal populations. Experimental results support this assumption, suggesting that it may be possible to develop paradigms to estimate the risk of an impending transition to an epileptic state.