The question is raised in how far hearing impaired subjects process the relatively robust 'voicing* feature differently from normal listeners. Starting from the German words "Deich", "Teich" and "Eich", twelve different stimuli were produced by using a splicing technique in which the cues 'burst', 'aspiration', 'VOT' and 'vowel-onset' were interchanged, deleted or added. For both groups of listeners, the VOT only played a minor role in the discrimination of the voicing feature. A difference between the two groups was found when the burst of the /d/ was substituted by the burst of the /t/. In this case, a clear majority of the normal listeners heard a /t/, while the hard of hearing still recognized a /d/. Starting from the stimulus "Teich", both the burst and the aspiration were sufficient to evoke the judgement 'voiceless stop'. But again the burst was less important for the hearing impaired subjects. Altogether, the aspiration seemed to play the dominant role in all listeners'judgements. Keywords: speech perception, recognition of stops, comparison of normal and hearing impaired listeners, auditory feature analysis, acoustic analysis of speech signals, auditory training, perception strategies.