Perception of intonational phrase integrity by hearing-impaired (HI) listeners was investigated in three experiments. Listeners compared fluently spoken sentences and sentences assembled by concatenating words. Signals consisted of speech and noise that was amplitude modulated by speech envelopes. Listeners were able to perceive prosodic phrasing in speech, but were significantly less accurate with noise signals in spite of their success in matching speech utterances to their noise replicas. Neither speech style (clear vs. conversational) nor the limited practice showed significant effect on performance accuracy. Individual differences could not be attributed either to the listenersÂ’ preferred communication mode or the degree of hearing loss.