Epileptic seizures are one of the most well-known dysfunctions of the nervous system. During a seizure, a highly synchronized behavior of neural activity is observed that can cause symptoms ranging from mild sensual malfunctions to the complete loss of body control. Epileptic seizures and their prediction have been studied theoretically and experimentally, mostly based on using electroencephalography (EEG) data. However, the dynamical mechanisms that cause seizures are far from being understood. In this paper, we try to contribute towards the understanding by viewing the prediction and dynamical analysis as a multiscale problem involving multiple time as well as multiple spatial scales. On the smallest spatial scale we consider single neurons and investigate predictability of spiking. For clusters of neurons (or neuronal regions) we use patient data near the onset of epileptic seizures and find oscillatory behavior and scaling laws near the seizure onset. On the largest spatial scale we introduce a measure based on phase-locking intervals and wavelets and use it to resolve synchronization between different regions in the brain. We also compare our wavelet-based multiscale approach with the classical technique of maximum linear cross-correlation. At each level of the analysis we find interesting effects that show the multiscale nature of the problem and which could be used to test dynamical models or to improve prediction algorithms.