The paper examies various modes of HMM segmentation on two stylistically different samples of speech.