The integration of environmental multisensory information is a process that allows us to perceive stimuli coherently. While there are several studies investigating this process in the adult, the extent to which this ability occurs in children, as well as the role of congruent vs incongruent information on the identification process of multisensory stimuli, are not well investigated. In this study we explored the effects of audio-visual semantic congruency on identification of visually-presented stimuli. Eight groups of children aged from 6 to 13 years performed a speeded object identification task of visually-presented pictures. Results showed that children under 12 years of age do not benefit from multisensory presentation in speeding up the identification, suggesting that facilitation from semantic congruency only emerges late in childhood. Furthermore, the incoherent audio-visual condition exerted an interfering effect, regardless of age. From present data it seems that the system underlying multisensory intergration process of complex and meaningful stimuli reaches maturity later than the system underlying multisensory intergration process of simple stimuli.