1 Journal of Personality and Social Psychology Copyright 1997 by the American Psychological Association, Inc. 1997, Vol. 73, No. 3, /97/$3.00 Does Empathy Lead to Anything More Than Superficial Helping? Comment on Batson etal. (1997) Steven L. Neuberg, Robert B. Cialdini, Stephanie L. Brown, Carol Luce, and Brad J. Sagarin Arizona State University Brian E Lewis University of California, Los Angeles To properly test the hypothesis that empathy-associated helping is altruistic, one needs to (a) consider plausible nonaltruistic alternatives for the observed empathy-helping effects, (b) validly and reliably measure these nonaltruistic alternatives, and (c) examine whether the empathy-belping relationship remains after removing the effects of the full complement of reasonable nonaltruistic alternatives. C. D. Batson, K. Sager, E. Garst, M. Kang, K. Rubchinsky, and K. Dawson (1997) failed to meet these criteria. New data, and reanalyses of existing data, bolster the case that self-other overlap-- a nonaltruistic motivator--underlies the association between empathy and costly helping. At best, empathy per se leads to superficial helping. In a postscript, the authors comment briefly on C. D. Batson's (1997) reply to this comment and, given his remarks, speculate as to whether the empathyaltruism formulation is even relevant to understanding meaningful forms of help. We have always found back-to-back (perhaps better termed eye-to-eye) scholarly exchanges fascinating reading. Although, for observers, such exchanges can be highly engaging--even entertaining--pieces of academic theater, they typically carry an associated burden. Rarely can they be fairly judged through a surface reading of article abstracts or summaries. A careful assessment of the sometimes dense particulars of methods, measures, and analyses is required. The present exchange seems no exception. However, we have tried to lay out our position in as reader-friendly a manner as possible, beginning with a brief background summary, then moving to a consideration of the strengths and weaknesses of the work of Batson, Sager, Garst, Kang, Rubchinsky, and Dawson (1997) as well as our own, and ending with the presentation of new data that reconcile the two seemingly divergent data sets. Background For more than 20 years, Daniel Batson and his colleagues have argued that altruistic helping exists and that it occurs when the potential helper experiences for the person in need. According to this empathy-altruism hypothesis, empathy-based helping is motivated not by self-interest but by a Steven L. Neuberg, Robert B. Cialdini, Stephanie L. Brown, Carol Luce, and Brad J. Sagarin, Department of Psychology, Arizona State University; Brian P. Lewis, Department of Psychology, University of California, Los Angeles. We thank C. Daniel Batson for sharing his data with us. Correspondence should be addressed to either Steven L. Neuberg or Robert B. Cialdini, Department of Psychology, Arizona State University, Tempe, Arizona Electronic mail may be sent via the Internet to either or concern for the other's welfare. Indeed, an impressive number of research studies appear to support this view (Batson & Shaw, 1991 ). We recently suggested, however, that empathy-helping findings are open to an alternative interpretation (Cialdini, Brown, Lewis, Luce, & Neuberg, 1997). In modem theories of the self, as well as in current evolutionary thought, important features of the self can be located outside of the person and inside others. In addition, those conditions that typically lead to (e.g., kinship, familiarity, perspective taking) also lead people to see parts of their selves in others. The possibility exists, then, that empathy-associated helping is not selfless but is rooted in the (usually implicit) desire to help that part of the self that is located in the other. Our experiments supported this possibility: The powerful impact of on helping was consistently eliminated when oneness--a measure of perceived self-other overlap--wa s also considered. We concluded that empathy-associated helping can no longer be presumed to be altruistic because as empathy increases, so does the presence of the self in the other. Batson et al. ( 1997 ) responded to our findings with two studies of their own, which they interpreted as incompatible with our claims and consistent with the empathy-altruism hypothesis. Although many issues could be raised in reaction, space constraints force us to focus instead on just three main points: First, as with previous studies from their lab, Batson et al.'s procedures failed to rule out fully the action of egoistic motives for helping. Second, a new experiment puts to rest several of Batson et al.'s concerns about our studies. Third, reanalyses of the Cialdini et al. (1997) experiments revealed that empathic concern seems to orient people toward help and to stimulate superficial helping, but it does not appear to predict the levels of meaningful help people provide. In contrast, it is in the arena of meaningful (nontrivial) helping that oneness and egoistic motivators seem to have their greatest influence. These findings 510
2 EMPATHY AND SUPERFICIAL HELPING: COMMENT 511 not only explain the apparently conflicting outcomes of our work versus those of Batson et al. but also suggest a conceptual integration of existing theory and empirical findings in the larger literature. Properly Assessing Altruism According to the empathy-altruism hypothesis, altruistic behavior is motivated by the desire to benefit another. Should any nonaltruistic concern lead people to help, the helping cannot be viewed as purely altruistic. If we accept this premise, rigorous explorations of the empathy-altruism hypothesis must meet, at minimum, three criteria: First, in any investigation, the researcher must assess each of the nonaltruistic variables that may be reasonably considered to motivate aid in that situation (e.g., those shown by prior research to spur helping). Second, the researcher must measure these nonaltruistic alternatives validly and reliably. Third, the researcher must examine whether the observed empathy influence on helping still remains after removing the effects on helping of the full complement of relevant nonaltruistic alternatives; it does not suffice to test the empathyaltruism hypothesis against just one nonaltruistic alternative at a time (Cialdini, 1991; Sorrentino, 1991 ). Although stringent, these hurdles are conceptually necessary because the altruistic explanation posits that no egoistic variable can account for helping if the help is truly altruistic. Unfortunately, Batson and his colleagues have yet to design an experiment that meets these criteria. In the typical study, they assess at most a subset of reasonable nonaltruistic alternatives; they do not always measure the nonaltruistic alternatives with the validity and reliability they afford measures of empathic concern; and they never remove the effects on helping of a full complement of reasonable nonaltruistic alternatives before concluding that empathy-associated helping is altruistic. Instead, they adopt the strategy of dispatching, in one study at a time, compelling nonaltruistic alternatives. For instance, they concerned themselves in one investigation with social approval; in a separate investigation, with guilt; in a third, with sadness reduction; and in a fourth, with empathic joy. Although this strategy can create the appearance that the compelling nonaltruistic alternatives have been adequately disconfirmed in aggregate, these studies--by their design--cannot provide strong evidence for altruistic aid. The two studies reported by Batson et al. (1997) neatly illustrate these problems. First, plausible nonaltruistic motivators (e.g., social desirability) were not measured or taken into account. Consequently, it could not be determined whether the observed empathy-helping effects remained after partialing out the influences on helping of the full complement of nonaltruistic factors. Finally, the focal nonaltruistic motivator du jour--selfother overlap--was measured unreliably, especially as compared with the assessment of? Burdened by such problems, the Batson et al. studies do not provide compelling evidence for altruistic helping. Addressing Batson et al.'s Critique Batson et al. (1997) criticized our studies on four grounds: (a) We did not directly manipulate or perspec- tive taking; (b) our experimental paradigm captured only "imagined needs and self-reports of imagined willingness to help"; (c) self-other distinctiveness, and not overlap, contributes to empathy-associated helping; and (d) we measured empathic concern and the egoistic motivators only after the helping opportunity. We address these criticisms in turn. First, although we did not manipulate using perspective-taking instructions, we did directly manipulate empathic concern (without affecting oneness) through our severity of need variable. It is not clear why we would need to manipulate through perspective-taking instructions. If per se influences helping, then it should do so whether it arises from perspective-taking instructions, from severity of need, from perceptions of kinship or friendship, or from other factors. Our reading of the empathy-altruism hypothesis provided no suggestion that the hypothesis should be 1 Space limitations preclude a thorough treatment of this issue here, so we must be brief. First, Batson et al. (1997) used a truncated version of our oneness index in their operationalization of self-other overlap. Our index included two components: the Inclusion of Other in the Self (IOS) scale (Akron, Aron, & S mollan 1992) and a measure of willingness to use the term "we" to describe the relationship with the needy other. Batson et al. used only the los scale, reducing the reliability of their measure of self-other overlap to that of a one-item scale. Second, they changed the wording of the IOS scale instructions from that of Aron et al. (1992) in a way that we believe channeled participants into a focus on their separateness from the person in need. That is, prior to Aron et ai.'s standard instructions, Batson et al. made salient participants' nonoverlapping identity with the other through the orienting clause "Considering yourself as Se/f and the person whose pilot interview you heard as Other...." Perhaps this initial orientation toward separateness accounted for the relative failure of the los scale to predict helping in the Batson et al. studies (rs =.22 and.29, as compared with a range from.43 to.70 in our studies). In any event, if the purpose of research is to test the applicability of a rival construct such as self-other overlap to a theoretical account, it seems unwise to change the measurement of that construct in the process. Third, Batson et al. used similarity as an indicator of self-other overlap, assessing it after the helping opportunity, unlike the other putative mediators, which were measured beforehand. Whether this order made a difference we cannot know, but we note that the obtained weak similarity-helping correlations (rs =. 17 and.07) run counter to much research demonstrating powerful effects of similarity on helping (see Dovidio, 1984, for a review). Finally, in an attempt to recreate the measure of self-other overlap put forth by Davis, Conklin, Smith, and Luce (1996), Batson et al. had participants rate both themselves and the person in need on a series of attributes, defining selfother overlap in terms of the absolute difference between the ratings of self and other. There are several issues of note. First, the Davis et al. measure was composed of many more attributes, likely increasing not only its psychometric reliability but also its validity as an indicator of self-concept. Second, in the hope of reducing reactivity biases, Davis et al. assessed self-attributes weeks before and in a different context than other-atla'ibutes. In contrast, Batson et al. assessed self- and other-attributes back-to-back in the lab. We suspect that this methodological change encouraged participants to seek not similarities but differences between themselves and the person in need and thereby contributed to Batson et al.'s failure to replicate the Davis et al. findings. In sum, it appears that Batson et al. assessed self-other overlap with less than desirable reliability and validity, making it difficult for them to explore rigorously the nonaltruistic hypothesis.
3 512 NEUBERG ETAL. supported only when had been generated by perspective-taking instructions. Batson et al. (1997) also suggested that because our experimental paradigm did not use currently active helping situations, it may have been unable to elicit real from participants or to capture anything beyond their willingness to play act "socially normative scripts for how one ought to act and feel." We have two responses. First, as we noted (Cialdini et al., 1997), our participants were asked not to speculate on what they might be feeling but to report what they were presently feeling. In this way, we inquired into the genuine responses of our participants to our manipulations. Second, our findings replicated quite nicely the previous work of Batson and his colleagues in multiple ways: The levels of felt by our participants were sizable and well within the range recorded in earlier studies using active need situations. Empathic concern was correlated highly with helping. Empathic concern was correlated with interpersonal closeness, sadness, and distress. Moreover, continued to predict helping after controlling for the effects of sadness and distress. Given the extent of replication, should Batson et al. now wonder whether their own studies capture only the power of "socially normative scripts"? In sum, because our methods so closely replicated results supporting the empathy-altruism hypothesis, it seems unconvincing to argue that those same methods were to blame when the empathy-helping relationship disappeared after self-other overlap had been partialed out. Third, in contrast to our position that self-other overlap underlies the influences of empathy on helping, Batson et al. (1997) argued the reverse: that self-other distinctiveness is necessary if one is to appreciate the circumstances of others and feel empathy for them. Although we have no doubt that some minimal level of self-other distinctiveness is necessary for empathy-associated helping to take place--else we place the bandage on our arm instead of on the wounded victim's-- it is not clear to us that differentiating between the self- and other-concepts increases such helping. Indeed, the suspect nature of their self-other overlap measure aside (see Footnote 1 ), a reanalysis of their Experiment 1 data revealed that perspectivetaking participants showed greater self-other differentiation than control participants on only two items--overburdened (p <.09) and pressured (p <.ll)--both of which imply not a characteristic feature of one's self-concept but rather a temporary (and objective) feature of the victim's highly stressful circumstance. Furthermore, Batson et al. found no evidence for their distinctiveness hypothesis in Study 2. For these reasons, we are not convinced that differentiating between the self-concept and the other-concept enables empathy-associated helping] Fourth, Batson et al. (1997) were concerned that we measured empathy after helping, contrary to the proposed natural sequence of these variables. To allay their concern, we ran a new study that replicated the eviction setting presented in Studies 1 and 3 of Cialdini et al. (1997) but moved the measure of empathic concern before the helping opportunity. In brief, participants were asked to describe in writing either a near stranger, an acquaintance, a good friend, or a family member (preferably a sibling). They were then asked to consider that this person had just been evicted from his or her apartment and was in need of help. We then assessed their, feelings of sadness and distress, and perceptions of oneness with the person. Finally, participants were asked to choose among seven helping options, ranging from providing no help to offering to let the person live with them rent-free. Our results replicated those of the prior studies perfectly, demonstrating that our earlier effects were not an artifact of measuring empathy after helping. First, relationship closeness led to increased willingness to help, F(3, 76) = 61.51, p = Second, relationship closeness led to both increased empathic concern, F(3, 76) = 10.82, p =.0001, and oneness, F(3, 76) = 56.14, p < Third, both and oneness were highly correlated with helping, rs =.62 and.80, respectively. Finally, we ran a hierarchical regression to determine whether the significant zero-order empathy-helping relationship would hold after we partialed out the nonaltruistic motivators (e.g., sadness, distress, oneness). These data are presented in Table 1 and reveal that empathy failed to predict helping after the nonaltruistic motivators had been considered, although oneness remained a significant predictor. In sum, these data provide the fifth replication of our findings, lending further doubt that empathy-associated helping is truly altruistic. They also reveal that our previous results were not an artifact of the order in which we measured, as Batson et al. (1997) had feared. In seeking ways to reconcile these data with those of Batson et al. (1997), we noticed that although the participants in the Batson et al. studies were provided with a four-level measure of helping--no help, 2-4 hr, 5-7 hr, or 8-10 hr of offered help--virtually all who offered to help did so at the lowest level available to them. In Experiment 1, although 25 of 40 participants offered some degree of help, only 3 of them offered to help beyond the minimum. Similarly, in Experiment 2, although 24 of 60 participants offered some degree of help, only 2 of them offered to help beyond the minimum. As a result, Batson et al.'s purportedly continuous measure of offered help was effectively a dichotomous measure. Thus, their data did not allow a determination of which levels of help may have been associated with. We used our new data set to explore the question of the kind of helping that might generate. First, after transforming participants' helping responses into dichotomous scores (would not help vs. would help), we repeated our standard analyses. The results were striking and counter to what we found when using a continuous measure of helping: Consistent with the empathy-altruism hypothesis, remained a significant predictor of willingness to help even after all the measured nonaltruistic motivators were accounted for, b =.11, F(1, 69) = 10.91, p <.001. We then analyzed the data of all participants who reported a willingness to help (74 of 79 participants were willing to help in some form)--that is, nonzero helping--and our previous findings emerged: Empathic concern lost its ability to predict amount of nonzero help after 2 As an aside, Batson et al. (1997) frequently characterized the selfother overlap hypothesis as implying psychological indistinguishability. Although such a depiction might be useful as a straw man, no theorist with an affinity for this hypothesis of whom we are aware would make such an extreme claim, nor would we. To "see" part of oneself in another is not to believe that one and other are the same.
4 EMPATHY AND SUPERFICIAL HELPING: COMMENT 513 Table 1 Summary of Hierarchical Regression Analyses for Variables Predicting Intended Helping for Evicted Person Variable B SE B Step 1: Participant gender and Empathic concern *** Step 2: Sadness and distress Empathic concern t Empathic concern Oneness *** Note. Given our theoretical focus on and oneness as predictors of helping, and for presentational clarity, we do not report here the findings for participant gender or for the sadness and distress items. ***p < t p < 13 (marginally significant). the nonaltruistic variables were accounted for, b =.068, F(1, 64) = 0.73, ns, and the variable of oneness remained a unique predictor of helping, b =.94, F(1, 64) = 44.74, p < Thus, whereas was able to 15redict whether participants expressed a willingness to help, oneness (and not ) predicted how much help participants were willing to provide. We use the remainder of this paper to explore more fully this effect and its theoretical implications. Before moving on, however, we briefly note that the findings from this study also bear on the first two concerns voiced by Batson et al. (1997). Specifically, our finding that empathy was able to predict participants' decisions about whether to help, even after we controlled for self-other overlap, further indicates that one does not need a perspective-taking manipulation to create adequate levels of ; our relationship-closeness manipulation created sufficient to replicate the Batson et al. effects. Second, because we were able to perfectly replicate the Batson et al. findings by using the dichotomous helping measure, their critique of our paradigm as unable to detect effects obtained in active helping situations becomes even less compelling. In sum, we are confident that the Cialdini et al. (1997) findings are not attributable to weak empathy effects, "artificial" elements of our paradigm, or the order in which we measured empathy. Rather, we think they say something important about the roles of and oneness in shaping helping behavior. We turn, then, to explore the intriguing results from our last analysis, proposing that the function of may be essentially preparatory, serving to orient people to opportunities for helping and acting to spur relatively superficial assistance. Empathy and Superficial Helping We believe that empathy can have two quite important influences on helping, stemming from empathy's status as an emotion. Emotions serve two functions central to human survival: They alert us to important, personally relevant features of the environment, and they provide us with the energy to respond to these features (e.g., Carver & Scheier, 1990; Frijda, 1986; Higgins, 1987; Mandler, 1984; Simon, 1967; Tomkins, 1970). For instance, when a dangerous-looking thug approaches us on a dimly lit street, the emotional system both interrupts our ongoing mental activities so that we might take notice of the stranger and releases adrenaline-like biochemicals so that we might take quick and robust physical action. Although they alert us to the need for action and facilitate that action, emotions do not by themselves determine the specific course of action taken. Although fear of the approaching stranger may prepare us for action, we still must decide whether to slip into a doorway, run across the street, clench a fist, pull out the pepper spray, or shout for help. It is at this point that the costs and benefits of various options are considered and emotional arousal provides sufficient energy to support the selected tactic. 3 Applying this functional perspective on emotions to empathy, one should expect empathy to play a central role in both alerting us to another's need and energizing our attempts to help. One might conceive, then, of empathy as a first step to providing aid--a step that turns an individual into a potential helper-- and one that may even facilitate superficial helping, if such an opportunity is readily available. Conditions that independently lead us to focus on the potentially needy person, like close relationships in the real world and the perspective-taking instructions of Batson and his colleagues, should amplify these prosocial functions of empathy. We should not, however, expect empathy to play a unique role in determining helping decisions when meaningful costs are involved. Such circumstances require potential helpers to consider the price of providing aid (see Footnote 3). Therefore, fluctuations in are likely to be derivative, and thus secondary, to those variables such as perceived self-other overlap that apply directly to self-benefit and, hence, to the outcomes of cost-benefit analyses. To test deductions from our argument, we reanalyzed all our data. Several predictions emerged from our framework. First, there should be significant zero-order relationships between empathy and the amount of aid our participants provide once they decide to help, reflecting the energizing and support functions that empathy serves. Second, because these functions are secondary within the context of cost-benefit analyses, the empathy-helping relationship should become nonsignificant once the effects of cost-relevant variables such as oneness are considered. Third, these cost-relevant variables--in our data, especially oneness--should uniquely predict the amount of nonzero help. These hypotheses are consistent with those proposed in Cialdini et al. (1997). Finally, and independent of the Cialdini et al. (1997) hypotheses, we expected empathy to be a unique predictor of the decision to help, but only when the easiest helping options were relatively costless. In the eviction need setting used in Studies 1 and 3 of Cialdini et al. (1997), the easiest helping option was handing an apartment guide to the person in need; this was rated 3 We make no claim that such cost-benefit analyses are always thoughtful or performed "on-line." In many cases, selected tactics are simply activated from memory on the basis of their success in similar circumstances, real or imagined.
5 514 NEUBERG ET AL. Table 2 Summary of Hierarchical Regression Analyses for Variables Predicting Intended Amount of Nonzero Help for Person in Need for the Five Helping Replications in Cialdini et al. (1997) Variable B SE B Study 1: Eviction Empathic concern * Empathic concern Empathic concern Oneness "** Study 2: Orphaned children Empathic concern ** Empathic concern * Empathic concern Oneness *** Study 3: Phone call Step I: Participant gender and Empathic concern ** Empathic concern * Empathic concern Oneness ** Study 3: Eviction Empathic concern ** Empathic concern Empathic concern Oneness *** Study 3: Orphaned children Empathic concern t Empathic concern Empathic concern Oneness * Note. Given our theoretical focus on and oneness as predictors of helping, and for presentational clarity, we do not report here findings for participant gender or for the sadness, distress, and other egoistic items. *p <.05. **p <.001. ***p < tp <.15 (marginally significant). by a different set of participants as having a cost level of 0.6 on a 10-point scale. We considered this level of help superficial and thus expected empathy to be a unique predictor of the decision to help. In the orphan need setting used in Studies 2 and 3, the easiest helping option was donating $10 to a fund for the children; this was rated as having a cost level of 3.0 on a 10-point scale. We considered this level of help costly and thus did not expect empathy to be a unique predictor of the decision to help. Finally, in the phone call need setting of Study 3, the easiest helping option was pointing out the nearest pay phone; this was rated as having a cost level of 0.3 on a 10- point scale. Because this level of help is essentially costless, we expected empathy to be a unique predictor of the decision to help. In sum, we expected that empathy would be a unique predictor of the decision to help in the eviction and phone call settings but not in the orphaned children setting. To test our first three hypotheses, we replicated the hierarchical regression analyses from Cialdini et al. (1997), but this time we used the amount of nonzero help offered as the dependent variable. As Table 2 reveals, these predictions were supported across all five replications: (a) Empathic concern predicted amount of nonzero helping, (b) it did so until the nonaltruistic variables were entered, and (c) oneness was a significant predictor above and beyond all other predictors. Table 3 reveals generally strong evidence in support of the fourth hypothesis. When the helping measure was dichotomized into no help versus help, we expected that would uniquely predict dichotomous helping in the two eviction replications. This was indeed the case (ps <.12 and.07). Although these effects reach only marginal levels of significance, they do replicate the more powerful effects (p <.001) discovered in the new data set reported in this article. Empathic concern was not expected to remain a predictor of dichotomous helping in the two orphan replications, as the easiest helping opportunity available to pa4"t, icipants was a relatively costly one. Indeed, empathy's unique effects did not approach significance (ps >.50 and.75). Finally, in the phone call setting, predicted marginally (at best) a unique proportion of the decision to help (p <.25), despite the superficial nature of the easiest helping opportunity. We speculate that the relative lack of an empathy effect here may stem from the relatively low level of need inherent in the circumstance (Batson & Shaw, 1991 ). In any event, the evidence for even superficial empathy-based helping is quite weak in the phone call setting. 4 In sum, although not able to provide an experimental test of our new hypothesis, reanalyses of the Cialdini et al. (1997) data reveal that the ability of to predict helping is limited to deciding between providing either relatively costless help or no help at all. These findings have several important 4 The above analyses included all four levels of relationship closeness (i.e., near stranger, acquaintance, good friend, and family member). Because the research by Batson and his colleagues has focused on aid to strangers, we combined--in order to have a sufficient sample size-- the participant populations from the three replications of the eviction setting (two from Cialdini et al., t997, and the one reported in this article), and repeated the above analyses on only the near stranger condition. The findings replicated perfectly: (a) Empathic concern uniquely enhanced the decision to help (p <.001 ); (b) did not uniquely enhance the amount of nonzero helping, and indeed, its association with amount of nonzero help was negative (p <.08); and (c) oneness uniquely enhanced the amount of nonzero help (p <.001 ).
6 EMPATHY AND SUPERFICIAL HELPING: COMMENT 515 Table 3 Summary of Hierarchical Regression Analyses for Variables Predicting Whether Participants Decided to Help Person in Need for the Five Helping Replications in Cialdini et al. (1997) Variable B SE B Study 1: Eviction Empathic concern *** Empathic concern " Empathic concern Oneness ** Study 2: Orphaned children Empathic concern Empathic concern Empathic concern Oneness Study 3: Phone call Empathic concern * Empathic concern Empathic concern Oneness Study 3: Eviction Empathic concern *** Empathic concern * Empathic concern I" Oneness I" Study 3: Orphaned children Empathic concern "~ Empathic concern Empathic concern Oneness * Note. Given our theoretical focus on and oneness as predictors of helping, and for presentational clarity, we do not report here findings for participant gender or for the sadness, distress, and other egoistic items. *p <.05. **p <.001. ***p < "p <.15. implications. First, they are consistent with the finding of Batson, O'Quin, Fultz, Vanderplas, and Isen (1983) that under conditions of substantial cost to the helper, does not facilitate helping--an outcome that led these authors to characterize empathy-based altruism as a "fragile flower easily crushed by self-concern" (p. 718). Indeed, our results reveal that the flower of empathy-motivated helping is more fragile yet, blooming only in the garden of superficial assistance. These findings present an interpretational challenge to the empathyaltruism hypothesis. Helping that occurs without genuine cost to the helper (e.g., handing an evicted person an apartment guide) invites the possibility that it may be undertaken for selfinterested reasons, such as avoiding the perception of being especially unhelpful, reducing immediate hostility, or enhancing the prospect of receiving aid in return--all of which would be made more likely by taking the other's perspective. Second, these analyses tentatively point to the promise of a modified framework of empathy influences on helping. Grounding our conceptualizations in the theories and evidence of the functions of emotion, we posit that empathy both orients people toward potential helping circumstances and energizes whatever (if any) helping attempts ensue. It may also enable superficial helping. Empathy does not, however, uniquely influence how much meaningful help people decide to give. Such helping is influenced instead by factors such as oneness, which contribute in a straightforward way to assessments of costs and benefits. We are currently designing experiments to explore this framework more directly. In Closing Should we conclude from the foregoing analysis that the empathy-associated helping effects obtained in the Batson et al. (1997) studies occurred because the helping option chosen by most participants represented essentially costless aid? We think not. In our view, those effects were more likely attributable to the combined action of a set of non-empathy-based (and nonaltruistic) factors that were not appropriately measured or not removed from causal consideration. For example, in contrast to our studies, Batson et al. did not control for social desirability concerns as participants' helping choices were not anonymous, they did not include the egoistic factors of sadness and distress in the analyses, and they did not appropriately assess the construct of self-other overlap (see Footnote 1 ). In conclusion, we believe that perceived oneness remains a viable nonaltruistic motivator of helping effects commonly attributed to altruistic motivation. We also believe that evidence relegating the unique impact of empathy on helping to the arena of superficial assistance changes fundamentally the likely interpretation of that assistance. Of course, in light of the longstanding debate generated by this topic, we have little doubt that observers have yet to see its last round. References Aron, A., Aron, E. N., & SmoUan, D. (1992). Inclusion of Other in the Self Scale and the structure of interpersonal closeness. Journal of Personality and Social Psychology, 63, Batson, C.D. (1997). Self-other merging and the empathy-altruism hypothesis: Reply to Neuberg et al. (1997). Journal of Personality and Social Psychology, 73, Batson, C. D., O'Quin, K., Fultz, J., Vanderplas, M., & Isen, A. (1983). Influence of self-reported distress and empathy on egoistic versus
7 516 NEUBERG ET AL. altruistic motivation to help. Journal of Personality and Social Psychology, 45, Batson, C. D., Sager, K., Garst, E., Kang, M., Rubchinsky, K., & Dawson, K. (1997). Is empathy-induced helping due to self-other merging? Journal of Personality and Social Psychology, 73, Batson, C. D., & Shaw, L. L. (1991). Evidence for altruism: Toward a pluralism of prosocial motives. Psychological Inquiry, 2, Carver, C. S., & Scheier, M. E (1990). Origins and functions of positive and negative affect: A control-process view. Psychological Review, 97, Cialdini, R. B. (199l). A/truism or egoism? That is (still) the question. Psychological Inquiry, 2, Cialdini, R. B., Brown, S. L., Lewis, B. P., Luce, C., & Neuberg, S. L. (1997). Reinterpreting the empathy-altruism relationship: When one into one equals oneness. Journal of Personality and Social Psychology, 73, Davis, M. H., Conklin, L., Smith, A., & Luce, C. (1996). The effect of perspective taking on the cognitive representation of persons: A merging of self and other. Journal of Personality and Social Psychology, 70, Dovidio, J. E (1984). Helping behavior and altruism: An empirical and conceptual overview. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 17, pp ). New York: Academic Press. Frijda, N. H. (1986). The emotions. Cambridge, England: Cambridge University Press. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, Mandler, G. (1984). Mind and body: Psychology of emotion and stress. New York: Norton. Simon, H. A. (1967). Motivational and emotional controls of cognition. Psychological Review, 74, Sorrentino, R. M. (1991). Evidence for altruism: The lady is still waiting. Psychological Inquiry, 2, Tomkins, S. S. (1970). Affect as the primary motivational system. In M. Arnold (Ed.), Feelings and emotions (pp ). New York: Academic Press. Received March 18, 1997 Revision received April 22, 1997 Accepted April 22, 1997 Postscript: W(h)ither Empathy-Altruism? In this postscript, we find ourselves delivering a retort to a rebuttal of a response! We are not complaining about the opportunity, though, as the Batson article (1997) that follows contains several assertions that warrant comment. Space constraints require that we limit ourselves to commenting on only a few of them, however. First, Batson (1997) suggests that if self-other merging is not synonymous with self-other indistinguishability (as it certainly is not in our thinking), our position ceases to pose a serious problem for the empathy-altruism model. We disagree. To the extent that the conditions said to produce altruistic motivation also produce at least some degree of self-other overlap--as when one recognizes aspects of oneself (e.g., one's genes) in the other--there exists a serious problem in interpreting aid for the other as truly selflessly motivated. This overlap does not require the perception that the self and other are indistinguishable but requires only that there is some degree of self-other merging. Second, Batson (1997) argues that our manipulation (relationship closeness) and our measure (perceived oneness) of self-other overlap are confounded in that they can reflect differing social categories (e.g., stran~l ers, friends, kin) that are associated with differing helping norms. He argues further that the norms associated with these differing social categories may therefore account for our findings. However, this criticism ignores our multiple demonstrations that our findings appeared even when we examined our data within these social categories. For example, in both the original data analyses of Cialdini et al. (1997) and the reanalyses of Neuberg et al. reported in this article, our typical data pattern replicated perfectly when we analyzed the helping data of only those participants in the near stranger conditions--the conditions most analogous to Batson's methodology. Thus, even when we analyzed helping within the same social category, thereby eliminating the potential confound due to category-specific norms, our standard findings have emerged. Third, Batson (1997) contends that, contrary to our argument, empathy can indeed lead to more than superficial forms of aid, as evidenced by prior work that supported the empathy-altruism hypothesis while incorporating costly helping action (volunteering to take electric shock). In response, we reassert our position that although such work included costly help, it cannot be viewed as good support for the hypothesis because it did not adequately extract the influence of the full set of nonaltruistic motivators (especially oneness) on that costly aid. Finally, Batson (1997) maintains that we have mischaracterized the empathy-altruism model, which, he states, does not claim that empathy leads to helping, as we imply. Rather, the model only posits that empathic emotion evokes altruistic motivation, which may or may not spur helping, depending on the relative strengths of the various conflicting motives in / the situation. If this is how Batson wishes to characterize the model-- as implicating motivation and not behavior--we will certainly accede. But we feel that our data continue to be highly relevant to an evaluation of the model. That is, the empathy-altruism model may remain intact in light of our data because it deals with altruistic motivation rather than helping. However, if, as our findings suggest, altruistic motivation does not uniquely affect meaningful levels of aid, one must question the utility of the model. The effect of our data, then, may be to place the mode/ in a different sort of jeopardy. The empathy-altruism model may not be wrong, but because the motivation it concerns has no impact on meaningful forms of help, it may simply be irrelevant.