Saccadic eye movements provide important information about the neuron system at several levels. In recent years computer analysis of saccades has been adapted for use in clinical work. The most common detection methods do not always function without the user's control and aid. In the present paper a digital filter is described for the detection of saccades. This non-recursive filter unscrambles saccade data which has been collected during the execution of an algorithm. The method is suitable for use with microcomputers. The filter is adaptive. Two concise experiments using the method are described.