We proposed a Bayesian model for the detection of asynchronous EEG patterns. Based on a skew normal model of the pattern of interest in the time-domain and on the assumption that background activity can be modeled as colored noise, we estimate both the pattern of interest and the time onset in each trial from the data using a Monte Carlo Markov Chain algorithm. Initial tests on synthetic data showed that the methods estimated correctly the pattern and the time onsets in all trials.