The Perceived Self in Infancy

Discussions of the have frequently generated distinctions between ways of viewing the self: for example, James's distinction between an existential and a categorical referred to by Butterworth; and more recently, Neisser's (1988) thoughtful article distinguishing five kinds of selfknowledge-the ecological, interpersonal, extended, private, and conceptual selves. We intend to make only one distinction in our comments-between a presumed conceptual and a perceptually grounded self-a distinction recognized by Butterworth. Most people who write about the have in mind a representational construct composed of images (e.g., body images, physiognomic features), attitudes, childhood memories, beliefs, and the like. Discussion of the as perceived by way of direct detection of information is rare indeed. Neisser referred to a perceived as an ecological self and Butterworth in his target article discusses information for a perceived in visual proprioception located in optic flow patterns generated during locomotion and postural control. J. J. Gibson (1979) said, One perceives the environment and coperceives (p. 126). Optic flow patterns that define one's direction of movement and provide information for control of posture are indeed examples of information specifying the as distinct from the world. But there are other sources of information that specify a self, enough to make the construction of a representation a luxury, if not superfluous. We describe these sources, showing their ubiquity and their adaptive value, leaving formulations of a conceptual to others. Flow patterns created by self-movement, as Butterworth indicates, do indeed provide information for oneself. But how is it that as we move through the layout, shifting position with respect to the room and the objects around us, objects do not appear to move with us, confusing any potential usefulness of optic flow for specifying ourselves as distinct from the world? Analysis of patterns of flow in the optic array caused by one's own movement shows them to be global, producing perspective transformations of all the surfaces in the array so that motion vectors vary with distance of the surface and its position relative to the observer. Object motion, on the other hand, produces local motion in the optic array, giving rise to shearing, occlusion and disocclusion, and patterns of accretion and deletion (wiping of texture) in relation to other objects of the layout. Adult observers have no problem distinguishing between these two sources of motion information, even when both occur at once. These two types of change are reliable information for a distinction between and environment, and the information is detected and used for control of locomotion and other actions. Now we have evidence that it is detected by infants as young as 4 months. Kellman, Gleitman, and Spelke (1987) put infants in a special apparatus, where object motion and motion could be manipulated experimentally. Infants were seated in a chair mounted on a frame, connected with a rod, in front of the baby, that moved behind a screen. Either the frame supporting the baby's seat or the rod could be moved independently, in synchrony, or not at all. Habituation studies compared four conditions: (a) both rod and baby motionless; (b) rod moving, baby motionless; (c) baby moving, rod motionless; and (d) both rod and baby moving in synchrony. Results showed that infants distinguished visually between object movement (of the rod) and passive movement of themselves. Infants attend to visual events in the world very early and do not confuse information for such events with information specifying their own movement. When self-movement is active and spontaneous, it should be even more informative. Babies, like adults, have sources of information about their own movement in addition to visual proprioception. They not only see themselves moving in the world, they feel themselves via somatosensory and vestibular systems. How do these sources of information mesh? Do they specify the same thing, an objective self? Older studies of toddlers observing themselves in a mirror suggested that at some point in development, children identified the mirror image of their actions with somesthetic-vestibular information for them (Loveland, 1986). Recently, Bahrick and Watson (1985) showed that 4month-olds perceive optical and somesthetic-vestibular information resulting from their own movements as identical, emanating from the same invariant source, their own bodies. Infants watched spontaneous kicking movements of their own legs on a video monitor, with direct view of the legs hidden by a screen. On another monitor placed alongside was an analogous display of another infant's legs (similarly dressed) or of their own leg movements videotaped at a different time. Preferential-looking measures provided strong evidence that infants distinguished between the contingent live display of their own leg m-. on and a noncontingent display of themselves or another infant. They perceived the invariant intermodal relationship between visualproprioceptive and somatic-proprioceptive information. Identification of this invariant makes possible still another remarkable source of information for oneself as an actor upon the world. Though young infants have a limited repertoire of actions, numerous experiments have shown that they can use the movements available to them (leg kicks, arm waves, head turns, nonnutritive sucking, etc.) to produce a specific desired effect in the environment. For example, when spontaneous leg kicks were linked to the movement of a mobile strung overhead, infants discovered their own causal efficacy, first by kicking their legs to explore the experimental contingency, then by executing the movements with increased frequency and vigor to produce the desired effect (e.g., Rovee Thelen & Fisher, 1983). Yoked