Epileptic seizures are generally considered to result from excess and synchronized neural activity. Additionally, changes in amplitude and frequency are often seen in local field potential or electroencephalogram recordings during a seizure event. To investigate how seizures initiate, and how dynamical changes occur during seizure progression, we develop a neocortical network model based on a model suggested by Wilson [J. Theor. Biol. 200, 375 (1999)]. We propose a possible mechanism for seizure initiation as a bifurcation, and suggest that experimentally observed changes in field potential amplitude and frequency during the course of a seizure may be explained by noise-induced transitions among multistable states.