The purpose of this study was to investigate stimulus factors, i.e.what kind of facial features elicit imitative responses. The experiments of tongue-protruding imitation were carried out with several patterns of the face(FIG.1). FIRST EXPERIMENT; 1-, 3-, 7-, 9- month old infants(respectively N=20, 20, 10, 5)were presented tongue protrusion in the experimenter's real face, the mirror image of the face, the schematic face, the mouth pattern and the rectangle protrusion. The result showed that the complete face, whether real, schematic or mirror image, was efficient for 1-month old infants; in 2-month old infants, the movements of tongue protrusion itself was as efficient as human face patterns, face from or its parts for all 3-month old infants except for mirror image. It was, however, not later than 7 months that few imitative responses were elicited except by real face(FIG.2). In general, it was suggested that facial imitation did not occur in responses to physical movements per se of stimuli; still in the context of human face, the primitive form of imitation at 1-month reached its peak just before 3 months, and then started declining to show the first proper imitation emerging. SECOND EXPERIMENT; 1-to 12-month old infants(N=60)were presented tongue protrusions, in a schematic normal face, an eyes-and-mouth pattern without contour, a scrambled face, and a contour pattern without eyes. The result showed that more facial imitations were produced with human face-like stimuli(a schematic face or an eyes-and-mouth pattern)than complex stimuli(a scrambled face)or a contour pattern before 5-month old; still in 2 month-old a scrambled face was effective too. But for 5-month old, patterns of face elicited few imitative responses. These results showed almost the same developmental changes as in EXP.1. An analysis of other responses showed that, since facial imitation accompanied smile responses at 3 months, it was believed to be associated with positive emotion; and there occurred many reaching responses inhibiting facial imitation after 5 months. In short, it may be suggested that facial imitation rquires the cognition of human face-likeness which is considered to be based on primary socio-emotional communication.