The distinction drawn between cognitive, affective and social actions is historical, dating back at least to Aristotle. It is a distinction which has proven useful; however, this distinction often serves to obscure. Consider a situation where an eightmonth-old infant is approached by a stranger. At first, as the stranger enters the room, the child turns its head and stares at the intruder, but after the stranger approaches, the infant breacks into tears and crawls to its mother. How should we understand this situation? Are the child's behaviors cognitive, emotional or social actions, or are they, in fact, all three? Clearly, the attention given to the stranger has important cognitive elements. The sounds of someone entering produced interest and attention. The domains interface; interest is an emotion (Izard, 1977), attention, a cognitive act (Lewis, Goldberg & Campbell, 1969). The child compares the image of the stranger to the image of known people and there is a discrepancy between them. This appraisal involves memory capacities along with other cognitive skills (see Bruyer, 1986). In this particular child, discrepancy causes alarm and fear. Here we see cognition leading to an emotion. On the other hand emotion leads to social action and cognition the child moves to its mother and holds her for protection. The connection between emotional, social and cognitive aspects of the child's life is complex. Most investigators have offered linear models of explanation, either emotion leads to cognition (see, for example, Zajonc, 1980) or cognition leads to emotion (Lazarus, 1982). Clearly, the entry point of observation determines the causal chain in these linear models (Lewis, Sullivan & Michalson, 1984). We have suggested elsewhere (Lewis & Michalson, 1983) that linear models will not serve us well in this regard and that it is better to consider that social/emotional and cognitive life is better considered as a fugue or strange loop (see also Hofstadter, 1980) rather than a simple linear chain of events. Such an analysis requires that we consider theories of social development which are complex mixtures of all these elements. Professor Frye (1989, this issue) has discussed some of the more important issues of social and cognitive development in infancy, however he has neglected for the most part to pay attention to the complexity of their interaction. For example, consider relationships. If we mean by relationship the features mentioned by Hinde (1976, 1979), they must include (1) goal structures, (2) diversity of interactions, (3) degree of reciprocity, (4) meshing of interactions, (5) frequency, (6) patterning and multidimensional qualities. Elsewhere I have suggested that these six features characterize interactions of both human and non-human child-parent dyads (Lewis, 1987). For example, a pup rat and its mother can match, align or attune their behavior to each other, but these interactions are hardly relationships; especially if we consider relationship from the mature human perspective (Lewis, 1989). Hinde suggests two further features of relationship: (7) cognitive factors or those mental processes that allow