A good reproduction of speech should allow the listener to suppress noise and reverberation as if the sounds were heard in real life. An experiment was designed where room properties and reproduction techniques were varied in a way that allowed evaluation of noise and reverberation suppression based on speech intelligibility measurements. Speech intelligibility was considerably better in real life compared to artificial head recordings presented through headphones. The headphone reproductions did not provide enough information to allow the listener to suppress noise and reverberation as well as in real life. It was found that subjects were very consistent. This makes the method precise and it should be useable for making comparisons of different reproduction techniques.