Four experiments are described which investigated the role of the mother's voice in facilitating recognition of the mother's face at birth. Experiment 1 replicated our previous findings (Br. J. Dev. Psychol. 1989; 7: 3-15; The origins of human face perception by very young infants. Ph.D. Thesis, University of Glasgow, Scotland, UK, 1990) indicating a preference for the mother's face when a control for the mother's voice and odours was used only during the testing. A second experiment adopted the same procedures, but controlled for the mother's voice from birth through testing. The neonates were at no time exposed to their mother's voice. Under these conditions, no preference was found. Further, neonates showed only few head turns towards both the mother and the stranger during the testing. Experiment 3 looked at the number of head turns under conditions where the newborn infants were exposed to both the mother's voice and face from birth to 5 to 15 min prior to testing. Again, a strong preference for the mother's face was demonstrated. Such preference, however, vanished in Experiment 4, when neonates had no previous exposure to the mother's voice-face combination. The conclusion drawn is that a prior experience with both the mother's voice and face is necessary for the development of face recognition, and that intermodal perception is evident at birth. The neonates' ability to recognize the face of the mother is most likely to be rooted in prenatal learning of the mother's voice.