RECENT years have witnessed an upsurge in interest in the change of attitudes and behaviors in children following being imitated. In general, experimental findings have revealed that, when given an ambiguous task, children tend to be attracted to, give rewards to, and imitate an observer who imitated them rather than an observer who did not imitate them (Fouts, Waldner and Watson, 1976;Hallahan et aL, 1977;Kauffman, Snell and Hallahan, 1976;Kauffman et al., 1977;Kauffman Gordon and Baker, 1978;Parton and Priefert, 1975;Thelen, Dollinger and Roberts 1975;Thelen and Kirkland, 1976; Thelen et aL, 1977). Thelen and his associates have mterpreted these findings through a modified reinforcement-affect model of attraction (Byrne and Clore, 1970;Byrne, 1971).This model implies that, for the tasks under consideration, the observer's imitations provide consensual validation and a reduction of uncertainty for the model. Consensual vaUdation and reduction of uncertainty increase attraction and reciprocal imitation ofthe observer by the model.A logical extension of this research area would be the exploration of possible generalization effects. The present study was designed to assess whether induced reciprocal imitation would generalize to an observer other than the one who had imitated the child and to a task different from the one on which the child had been imitated. Experiment 1 was designed to assess generalized imitation across observers. The children in this experiment interacted with an adult who imitated or did not imitate them and with another adult having no scheduled opportunities to respond (i.e. to imitate or not to imitate). It was reasoned that if behavioral similarity induced by the imitating adult acquires reinforcement properties over trials (Parton and Priefert, 1975), the child might, in the absence of any other cues mdicating no reinforcement (SA) strive to induce maximal behavioral similarity by copying the response of both adults. Such a generalization effect could be facilitated by using adults who share salient characteristics such as age, height, body build, sex and the degree of (un) familiarity, and by randomly alternating their modeling trials (Peterson, 1968). Experiment 2 was designed to demonstrate generalization effects across different (ambiguous) tasks. Findings related to this effect have been observed by Thelen and Kirkland (1976). In contrast to their studies, these authors used previously acquainted children both as confederates (i.e. one confederate older than and one •Requests for reprints to: Dr. P. M. Smeets,