In this paper a model of vowel reduction is introduced that emphasizes the role of the listener in a natural speech situation. A listener has such powerful strategies at his disposal at phonological, grammatical, semantic and pragmatic levels to restore missing acoustic Information, that a talker is not obliged to articulate every phoneme at its target position. This will particularly be the case in a spontaneous speech situation; in a more formal speech style the talker will increase his articulatory effort which will result in less reduced vowels. Since the listener has more need for content words than for function words, which do not contribute much to the meaning of the message, a talker will consequently put more articulatory effort in content words resulting in less reduced vowels. These views were confirmed by the present investigation in which speech characteristics of free conversation were compared with those of the same sentences but now read. The clear relationship between vowel duration and spectral undershoot suggested by Lindblom [152] is not supported in this investigation.