One-channel routine recordings of the scalp electroencephalogram (EEG) from unmedicated children strictly classified as unprovoked typical (3 c/s) absence seizures were selected. The dynamics of spike-and-wave discharges (SWD) were then examined by means of autocorrelation, correlation dimension, averaged pointwise dimension and largest Lyapunov exponent. For one EEG signal with pronounced spike-and-wave (SW) patterns, these measures were used complementary to a surrogate data method, a nonlinear (SETAR) modeling approach, and a SW simulation procedure providing five types of SW test signals. The SETAR model exhibited stationary SW dynamics, visually very similar to the EEG target signal, and with clear nonlinear structure. According to the results, the EEG episodes investigated represent low-dimensional dynamics, possibly recorded during nonstationary periods. Arguments that justify the assumption of deterministic chaos in our EEG signals were not obtained with the current methods. From the results one may conclude that two global oscillatory modes are present for the model, and three modes are active during the EEG recording period.