The P300 wave is a positive wave induced when an infrequent (target) stimulus is randomly mixed in series of repetitive frequent (non target) stimuli. The P300 auditory oddball design is defined by numerous components (N1-N2-P2-P3a, P3b and closing processing (CP)) that are believed to represent the adaptation of the working memory to the new stimuli (context switching) or the closing of the usual context (context closure). Since 15 years, we have collected P300 data in normal subjects (from 10 to 90 years old) and in pathological states by using this kind of auditory oddball paradigm. While we were able to show that P300 latency increases with aging, we were not able to discriminate between pathological states, probably because we mixed (by using grand averages) a lot of different neural processes. In this study, we have developed and analysed EPs by dividing the presented auditive stimulations in subsequent blocks of 5 endogenous (infrequent) CEPs compared with 20 exogenous (frequent) CEPs. We have computed the “time-evolution of the neural network variability” (TINNV) on the basis of the P300 parameters (amplitude and latency) between these different blocks. More precisely, we compare the TINNV during a weak task (count target) and a strong constraining task (updating a date at each target), in normal and in 2 pathological groups (showing pure memory defect or memory defect associated with depression (DMS4)). By using this new index (TINNV) we can statistically discriminate between these two pathological populations.