Associating letters with speech sounds is essential when learning to read. Recent studies on audio-visual integration suggested that 1) the integration of letter and speech sound takes place automatically and 2) that temporal synchrony is critical for the integration. In the current study, we wished to determine the neural networks associated with an automatic integration of written and heard syllables in literate adults. The mismatch negativity (MMN), an index of automatic change detection in the brain, was recorded. Subjects were presented with auditory stimuli (consonant-vowel syllables) together with visual stimuli, which differed between conditions. The auditory stimuli included vowel or consonant changes, or changes in syllable intensity, frequency, or vowel length. In the experimental condition the visual stimuli were written syllables, whereas in the control condition, they were scrambled images of the written syllables. The visual stimuli were presented in half of the blocks synchronously with the auditory stimuli and in the other half with a time delay. During the stimulation, the subjects pressed a button to designated visual targets. Changes in auditory stimuli elicited an MMN/N2 response, in all conditions. The type of visual material (syllables vs. scrambled images) as well as the timing of presentation (delay vs. no delay) influenced the topography of this response. The results indicated that different neural networks are activated when speech stimuli are presented together with linguistic than with non-linguistic visual stimuli and that this effect is modulated by temporal synchrony. Implications for dyslexia are discussed.