We present a method for discriminating between healthy subjects and Alzheimer’s diseases patients from on-line handwriting. Departing from the current state of the art methods, that adopts machine learning methods and tools for building the classifier, we propose to apply the Negative Selection Algorithm. The major advantage of the proposed method in comparison with others machine learning techniques is that it requires only data by healthy subjects to build the classifier, thus avoiding to collect patient data, as requested by competing techniques. Experiments results involving data produced by 175 subjects show that the proposed method achieves state-of-the-art performance.