Spike EEG source localization results are influenced by different errors and approximations, e.g., head-model complexity, EEG signal noise, electrode misplacements, tissue anisotropy, tissue conductivity noise as well as numerical errors. For accurate source localization, understanding the affects of these errors on the source localization is very crucial. Six finite element head models are selected for a head-model complexity study. A reference head model is used to create the synthetic EEG signals by placing a dipole inside the model to mimic the epileptic spike activity. To understand the influence of EEG signal noise, tissue conductivity noise and electrode misplacements on the EEG source localization, different level of noises are added to EEG signals, tissue conductivities and electrode positions, independently. To investigate the influence of white matter anisotropy, a realistic head model generated from T1-weighted MRI is used and the conductivity anisotropy for the white matter is calculated from diffusion tensor imaging (DTI). Major findings of the study include (1) the CSF layer plays an important role to achieve an accurate source localization result, (2) the source localization is very sensitive to the tissue conductivity noises, (3) one centimeter electrode misplacement cause approximately 8 mm localization error, (4) the source localization is robust with respect to the EEG signal noise and (5) the model with white matter anisotropy has small source localization error but large amplitude and orientation errors compared to the isotropic head model.