Automatic classification methods are frequently used in early diagnosis of different diseases that affect speech production. These methods can also be applied to identify speech samples from patients affected by Parkinson's disease (PD) or depressive disorder (DD). This paper is interested in applying automatic stress detection and prosodic phrasing approaches on pathological speech samples in order to assess to what extent these tools can be useful either in characterizing in an unsupervised manner the prosodic attributes of pathological samples from individuals affected by PD and DD, or classifying samples as belonging to healthy or non-healthy individuals. We formulated hypotheses in connection with the duration of phonological phrases and the number of words grouped by them. We also briefly analyzed the phrase distributions. Our results show that healthy and pathological samples can be separated from each other by means of these prosodic analysers, and deep neural network or support vector machine based classifiers built on top of them.