Electroencephalography (EEG) is a non-invasive neuroimaging tool which can be used to measure brain activity with excellent temporal resolution. By solving the so-called ‘inverse problem’, one can not only study the time-course of activation during a particular task, but also identify the location of the underlying neural sources in the brain. Usually, these methods are applied to averaged data obtained from EEG recordings during which volunteers perform a given, usually brief (0.1-5s) task over many dozens repetitions. This averaging across many brief trials results in clearer responses, as artifactual or randomly occurring events (noise, eye blinks, eye wanderings, etc) will be reduced. Here, we take a radically new approach, and ask whether one can obtain reliable source localization associated to a particular stimulus feature (such as visual contrast) when using complex natural stimuli presented over long periods of time. We made EEG recordings (64 channels) in 7 subjects who passively watched 2 minute long segments from different commercially available movies (the segments were repeated 20 times). We then developed a method based on independent component an