Four-, 6-, and 8-month-old infants’ perception of the multimodal features of the human face was investigated. First, infants were habituated to a visible and audible face of a person reciting a prepared script. Then they were tested by changing various features of just the audible, just the visible, or both components of the face. When features were changed, such as the lexical-syntactic content, the speaker’s gender, or the synchrony relation between the audible and visible components, the infants discriminated their multimodal and visible representation but not the audible one. When the manner of speech was changed from adult- to infant-directed, the 2 older groups discriminated all 3 types of changes but the 4-month-old infants only discriminated its visible and multimodal representation. Results show that speech-related exaggerated prosody cues facilitate detection of the audible features of multimodally represented faces but not until 6 months of age. The perceptual world of the infant is multimodal in nature. Objects and events often can be seen, heard, felt, smelled, or tasted. The features that simultaneously specify objects and events in different sensory modalities may b e amodal (i.e., equivalent across the different sensory modalities) or modalityspecific. For example, the human face can be specified by a host of amodal features (i.e., tempo, duration, and synchrony) because of correspondences created by the way the mouth moves while it produces sounds. It also can be specified concurrently by a host of modality-specific features such as size and color of eyebrows, hair color, lip size, overall configuration of the various facial features, skin color, and pitch and timbre of the voice. How infants respond to these kinds of amodal and modalityspecific features of multimodally specified objects and events has become a question of considerable research interest ( Lewkowicz & Lickliter, 1994). Most of the research designed to answer this question has focused on infants’ ability to integrate the multimodal attributes of events and objects on the basis o f amodal information. As a whole, this research has shown that many basic intersensory integrative abilities emerge during the first year of life (Lewkowicz & Lickliter, 1994). In a typical study o f intersensory integration, infants are presented with two, side-by-side, dynamic visual displays and a sound that corresponds to one of