Optical cues for visual and auditory-visual (A-V) perception of speech were varied by placing a sheet of rough-surfaced Plexiglas between talker and lipreader and systematically changing the distance between Plexiglas and talker. This distorts the optical environment in a way that is analogous to masking or filtering in the acoustic domain. In studies with normal-hearing adults and with hearing-impaired children, speech (words, sentences) was presented live under different degrees of optical distortion, and observers attempted to identify the stimuli. Visual-along (lipreading) scores dropped abruptly to the chance level as Plexiglas distance (blurring) was increased. A-V scores were relatively high for clear conditions but diminished gradually as Plexiglas distance (blurring) was increased. Under extremely poor optical conditions, A-V scores reached a plateau. This represents essentially auditory perception without meaningful optical cues for speech. Results parallel those of previous acoustic studies that compared auditory with A-V perception of speech as a function of S/N ratio or sensation level, demonstrating a reciprocal aspect of optical and acoustic cues for speech perception. Optical distortion seems to have potential as an auditory training technique to shift attention of hearing-impaired observers to non-dominant acoustic cues during A-V perception of speech.