In recent studies, it has been suggested that dyslectic symptoms result from an impaired ability to distinguish irrelevant from relevant sensory information. To attest this perceptual noise exclusion hypothesis, we performed an experiment that consisted of two sentence repetition tasks, one in the auditory modality using spoken stimuli, and the other in the visual modality registrating eye movements while reading. Both versions were conducted by the same groups of adult dyslectics and typical language users. Materials consisted of Dutch sentences with either a semantically congruent or incongruent word of a minimal pair like mouse-house as the final word, as in The cheese was eaten by the mouse/*house. The sentences were presented in three noise levels, ( no, mild (-5 dB), and heavy (-10 dB) fluctuating noise for the auditory version, and no, 30%, and 60% text masking blocked patterns for the visual version). Dyslexics made more mistakes in incongruent than congruent conditions in both the auditory and visual modality. Interestingly, they frequently replaced the incongruent final word by the congruent counterpart of the minimal pair. Especially in the visual modality, this pattern became more prominent with increasing noise levels. Typical language users showed no difference between congruent and incongruent conditions in no, or mild noise conditions. Finding effects of noise in both modalities indicates the involvement of a general deficit in processing noisy sensory information in dyslexia.