In the last ten years many papers have been devoted to a description of the electroencephalographic (EEG) activity in terms of chaos, namely in terms of a hypothetical underlying low dimensional nonlinear dynamics (attributed to neuron synchronization). However, the imperfect scaling of correlation sums to a power law and the incomplete saturation of the same for increasing embedding dimension have strongly reduced the expectation of an exhaustive EEG description in terms of low dimensional deterministic chaos. Further evaluations of embedding dimension, like the false nearest neighbors method or the singular value decomposition, confirm the limits of this approach. However, such a result would not look surprising. In fact, the EEG activity is only approximately stationary, as required by dimension evaluation; furthermore, the EEG time series are contaminated by intrinsic dynamical noise (sensory stimulation and membrane fluctuations) and by minor electrical noise of measurement apparatus. These characteristics of the EEG activity have been pointed out by the analysis, performed by linear and nonlinear methods, in some experiments of periodic photo-stimulation. This is a significant result since measurement of noise can put severe restrictions on the classification of the tracings and, on the other side, the noise reduction methods may severely disturb the structure of the signal in the time and phase-like space domain. [ABSTRACT FROM AUTHOR]