Although most of the bystander-intervention studies found that bystanders were prone to help when they were alone, one study (Latané, 1967) found that virtually all undergraduates failed to intervene in behalf of an assaulted child, even when they were alone. Similar results were obtained in a study by Kaufmann (1968). Only 11%0 of subjects who were asked to observe a “teacher” deliver what seemed to be increasingly severe and dangerous shocks to a “learner’ (cf. Milgram, 1963) responded to the learner’s pleas for help. No differences were found among conditions in which the bystander was given a position of high status, was led to believe he would later be the “teacher,” or had reason to doubt the legitimacy of the teacher’s authority. People fail to help in some situations, it would appear, on the slightest excuse.
A developmental study by Staub and Feagans (1969) found that children also failed to help in emergency situations. Unlike their adult counterparts, though, they were less prone to help when they were alone than when they were with another child. Nursery school, first- and second-grade children tended to help another child who appeared to fall and hurt himself more when they were in pairs than when they were alone. At the fourth and sixth grade the trend tended to reverse itself and the children behaved more like their adult counterparts. It is possible that the presence of a partner reduced the fear associated with doing something wrong in the younger children, but served as a source of potential negative evaluation for the older children.
Whatever the precise reasons, it seems likely that the behavior of bystanders is predicted on their ignorance concerning the consequences of helping. If they were assured that no harm would come to them, their apparently heartless inaction would seem less likely. And if they were assured that they would be rewarded, helping behavior would seem virtually certain. It is surprising, therefore, that none of the studies which manipulated perception of consequences has achieved a clear effect.
Modeling as a Function of Perception of Consequences
Models who supply information about the consequences of altruistic behavior almost always supply other information. Information about the consequences of particular choices of action usually supplies an indication of its appropriateness. The observation of the act also, of course, increases the salience of related behavioral alternatives.
Of the experiments which tested for the influence of perception of consequences, one used adult subjects and two used children. In the study on adults, Hornstein et al., ( 1968) presented passersby with a wallet that had apparently been found by a model and then relost. Passersby who were led to believe they were similar to the model were more likely to return the wallet when an attached note indicated that the model felt good or neutral about returning the wallet than when the note indicated he felt bad.
In the first of the studies on children, Midlarsky and Bryan (1967) failed to find any difference between the number of self-sacrificial and charitable responses of elementary school girls who observed a model who emitted expressive signs of joy after making self-sacrificial choices, and models who did not. M. Harris (1968), using an age group similar to that of Midlarsky and Bryan (1967), failed to find a difference between the altruistic behavior of children who were exposed to models who were praised for their altruism, and those who were not. Because the praising agent was absent when the children performed, however, they had reason not to expect similar consequences.
None of the studies that examined the effect of perception of consequences found a significant difference between positive and neutral consequence conditions. The Hornstein et al. (1968) study, though, found that the perception of negative consequences had an inhibiting effect on altruism. It is possible that the lack of similarity between observer and model in the studies which used children made the consequences of the model’s actions irrelevant.
Although situations in which consequences can be observed were analyzed according to their effect on the performance of altruistic behavior, some investigators (e.g., M. Harris, 1968; Midlarsky & Bryan, 1967) analyzed them in relation to the internalization of relatively permanent behavioral dispositions. None of the studies, however, met the two criteria of internalization—situational generality and longitudinal stability.
Several studies, though, have set out to examine the antecedents of modeling effects. Instead of explaining imitative behavior with what is essentially a descriptive rubric, they have attempted to find out what it is about observing a model that results in the acquisition of congruent behavioral dispositionse Three general approaches have been taken. The first examined modeling effects in terms of identification, the second examined them in terms of secondary reinforcement, and the third examined them in terms of empathy.
Modeling and the Internalization of
Altruistic Dispositions
Although several investigators have viewed modeling in terms of identification (see Bandura & Walters, 1963), there has been only one study which examined the relationship between identification and altruism. Rutherford and Mussen (1968) found that generous nursery school boys saw their fathers as more warm and nurturant than less generous boys. These findings were interpreted as support for the hypothesis that generosity results from identification with warm and nurturant samesexed parents. Although the findings indicated that the generous boys saw their fathers as warm and nurturant, no information was supplied about the generosity of the fathers, nor was the assumption that the boys were identified with their fathers substantiated. Moreover, the hypothesis that identification caused the boy’s generosity was supported by only correlational data.
Mowrer (1960) suggested that rewarding models are imitated because stimuli that are associated with the performance of imitative responses, especially proprioceptive feedback, are secondarily reinforcing. Hartup and Coates (1967) tested Mowrer’s suggestion by exposing nursery children who had histories of frequent and infrequent peer reinforcement to rewarding and nonrewarding peer models. Both history of reinforcement and rewardingness of model were determined by time-sampled observations of a nursery school class. The altruism of rewarding peer models was imitated more by children with histories of frequent peer reinforcement, and the altruism of nonrewarding peer models was imitated more by children with histories of infrequent peer reinforcement. No differences in attractiveness of model or social acceptance of subjects were found among groups. The results suggested that Mowrer’s (1960) theory applies only to subjects with histories of frequent reinforcement. However, on the basis of the positive correlation between giving and receiving reinforcement from peers (Charlesworth & Hartup, 1967) the results can also be interpreted as support for the notion that subjects modeled the responses of those who were similar to them. The latter interpretation is consistent with the Hornstein et al. (1968) findings. However, several other studies, which did not control for history of reinforcement (e.g., Grusec & Skubiski, in press; Rosenhan & White, 1967), failed to support the notion that rewarding models are imitated more than nonrewarding models.
The results of a study by Aronfreed and Paskel (1968) on first- to third-grade girls can be interpreted in accord with a secondary reinforcement position. Aronfreed and Paskel ( 1968) found that children who were exposed to a female model, who emitted both expressive signs of joy and hugs (EH) after she made self-sacrificial responses, evidenced more imitation than children who were exposed to expressive cues (E) or hugs (H) alone. If the model who emitted expressive cues and affection responses is seen as the most rewarding of the three models (and the fact that she was the only one who displayed pretask nurturance makes this likely) , then the modeling effect can be interpreted as a function of the secondary reinforcement effects of imitation.
Aronfreed and Paskel (1968) interpreted their results differently—as support for the notion that the self-sacrificial responses of the girls were reinforced by empathically experienced positive affect. Because the girls who were exposed to both affective responses and expressive cues were exposed to the most warmth, and because they were given the most indication of what was desired of them, interpretation of the results is difficult.
A later study by Midlarsky and Bryan ( 1967) controlled for pretask warmth and added a condition in which the model failed to emit expressive cues in the test situation. It also took a second measure of altruism in a different situation. Although the investigators did not interpret the results as support for the position of Aronfreed and Paskel, their results were very similar. Both studies found that exposure to the EH condition resulted in more self-sacrificial responses than exposure to E or H alone. Moreover, as Aronfreed and Paskel would have predicted, more self-sacrificial responses were made when the model emitted expressive cues during testing than when she did not. The findings, though, that self-sacrificial responses occurred in conditions other than the HE and EH conditions, and that most of the girls who sacrificed candy also donated to charity in the absence of the model, suggest that more was involved than the empathic transmission of positive affect.
There has been no concrete support for positions that view modeling as a function of identification, secondary reinforcement, or empathy. Studies which appear to support each position have alternative interpretations. Moreover, although all of the studies were concerned with the internalization of altruistic dispositions, none established the situational generality and longitudinal stability of the effects. Some researchers would argue that the two criteria are beyond the range of laboratory studies. What is needed, perhaps, is supplementary evidence from naturalistic studies. If, for example, correlates of altruism could be found in the behavior of parents or friends, relevant experimental analogues would receive support. Such correlates have been reported in three studies. Rosenhan (1967) found that a group of active civil rights workers had a close relationship with at least one altruistic parent; Tomkins (1965) reported that prominent abolitionists were influenced by altruistic friends; and Rettig (1956) found that altruism in parents was positively correlated with scores made by college students on an altruism scale. Some evidence was found in the Rettig (1956) study for the notion that the effects of direct reinforcement of altruistic behavior by parents dissipates over time, whereas modeling effects are more enduring.
Although more precise studies are needed, the results of studies which have found a relationship between attributes of parents and altruism in children are encouraging, especially in view of the failure of most studies to find a relationship between moral behavior in parents and personality correlates in children (cf. Kohlberg, in press).
Experimenter Effects
An appropriate way of ending the review of modeling and altruism is to draw attention to the relationship between modeling and experimenter effects. R. Rosenthal (1966) and others have demonstrated that experimenters unintentionally effect the responses of their subjects, even when they do not perform the tasks in question. Modeling can be considered a type of experimenter effect—it causes subjects to act in ways congruent with the behavior (and, usually, expectations) of models. The similarity between modeling and experimenter effects has three implications. The first emphasizes the suggestion that modeling effects are temporary and situationspecific. The second emphasizes the subtlety of modeling effects—a decade of research has been unable to isolate the means by which expectations are communicated in experiments. And finally, the necessity for methodological caution in modeling studies becomes obvious. Most studies attribute modeling effects to more than experimenter bias. In spite of this, few studies have controlled for expectations of the model.
