Reinterpreting the Empathy- Altruism Relationship: When One Into One Equals Oneness Robert B. Cialdini, Stephanie L. Brown, Brian .P Lewis, Carol Luce, and Steven L. Neuberg Arizona State University features Important of eht can be outside located self-concept of eht individual dna or close inside related .srehto ehT said previously data reinterpret to insight this use authors ot support eht empathy- model altruism of which asserts helping, that for results concern another empathic in ssensselfles dna That true altruism. ,si they argue that eht conditions ltehaadt ot concern empathic also lead ot a sense greater of self-other raising overlap, eht possibility that helping under is conditions these not but selfless is also directed toward eht self. In 3 studies, eht impact of on concern empathic to willingness help saw eliminated when oneness--a measure of self-other perceived overlap-- sawonly helping increased concern empathic that further revealed analyses Path considered. hguorht sti relation ot throwing thereby oneness, perceived eht empathy-altruism question. into model ehT srohtua suggest that primarily helping affects concern empathic sa signal an emotional of .sseneno Although social psychologists regularly inquire into the form et al., 1989; Batson, Duncan, Ackerman, Buckley, & Birch, and intent of behavior, rarely do they address questions designed 1981; Batson et al., 1988; Batson & Shaw, 1991; Batson, Turk, to reflect directly on the basic nature of humanity. That has not Shaw, & Klein, 1995; Coke, Batson, & McDavis, 1978; Toi & been the case, however, in one notable research domain: the Batson, 1982). According to Batson's empathy-altruism hy- study of true altruism. Not content to leave the issue to the pothesis, purely altruistic action can occur reliably, provided philosophers who have pondered it through the ages without that it is preceded by a specific psychological state: empathic clear resolution (Bentham, 1789/1879; Comte, 1851/1875; concern for another. Empathic concern is defined as an emo- Hume, 1740/1896; Nagel, 1970), various social psychologists tional reaction characterized by such feelings as compassion, have examined the possibility that wholly altruistic motivation tenderness, softheartedness, and sympathy. It is brought about falls within the range of human functioning (e.g., Campbell, by the act of perspective taking, wherein one person (e.g., the 1975; McDougall, 1908; Rushton, 1989; Staub, 1978), The potential altruist) takes the point of view of another (e.g., a question of whether we are ever genuinely selfless--that is, suffering victim). Perspective taking, in turn, is brought about motivated solely or principally to enhance the welfare of an- by a perception of attachment (kinship, friendship, familiarity, other--has been advanced significantly by theorists who have similarity) to the other or by instructions to take the other's introduced the concept of empathy to the debate (Batson, 1987; perspective (Batson & Shaw, 1991 ). Hoffman, 1984; Krebs, 1975). In a remarkably extensive program of investigation, Batson Currently, the most prominent and easily the most research and his associates (see Batson, 1991; Batson, in press, for re- productive of these empathy-based formulations is that of Bat- views) as well as researchers workingi ndependently (e.g., Dov- son and his associates (Batson, 1991; Batson, in press; Batson idio, Allen, & Schroeder, 1990; Schroeder, Dovidio, Sibicky, Matthews, & Allen, 1988) have generally supported the empa- thy-altruism hypothesis. Dozens of experiments have demon- treboR .B Stephanie Cialdini, .L Luce, E Carol Brian Lewis, Brown, strated that, first, the circumstances hypothesized to lead to Steven and .L ,grebueN Department of ,ygolohcysP State Arizona -revinU perspective taking do increase empathic concern, and second, .ytis E Brian is Lewis won ta eht of Department ,ygolohcysP ytisrevinU under conditions of empathic concern for another, individuals of California, soL .selegnA help more frequently in what appears to be an altruistically eW era grateful for eht comments helpful of Nancy Bill Eisenberg, motivated attempt to improve the other's well being rather than Graziano, Brad Sagarin, dna on Mark Schaller earlier versions of this an egoistically motivated attempt to improve their own. Espe- article. eW era especially grateful ot Art Aron for eht value of his cially impressive is that the empathy-altruism hypothesis has thinking dna insights from eht outset of this project. been repeatedly confirmed in response to challenges from a ecnednopserroC be should article this concerning desserdda to treboR .B Department Cialdini, of ,ygolohcysP State Arizona ,ytisrevinU ,epmeT variety of egoistically based alternative accounts of the basic Arizona .4011-78258 mail Electronic yam be sent via eht Internet ot effect. rebor @ t.cialdini .ude.usa For instance, Batson and colleagues have produced data sug- Journal of Personality and Social Psychology, Vol. 1997, ,37 .oN .3 481-494 Copyright by 1997 the American Psychological Association, Inc. 0022-3514/97/53.00 184 482 CIALDINI, BROWN, LEWIS, LUCE, AND NEUBERG gesting that the increased helping accompanying perspective conceptual identities that are merged, not physical identities or taking, is due to a selfless regard for the other rather than a situational circumstances. selfish desire to escape aversive arousal (Batson et al., 1981) The merging of self and other identity has been suggested or social disapproval (Fultz, Batson, Fortenbach, McCarthy, & before as an explanation for helpfulness (Aron & Aron, 1986; Varney, 1986) or guilt (Batson et al., 1988) or shame (Batson Hornstein, 1982; Lerner, 1982; Piliavin, Dovidio, Gaertner, & et al., 1988) or sadness (Batson et al., 1989) or to increase Clark, 1981 ). Batson and his colleagues have examined this vicarious joy (Batson et al., 1991). Although some are not possibility within the context of the empathy-altruism model convinced that the empathy-altruism position has won every and responded in two ways. The first was to acknowledge that battle against the egoistically based alternatives (e.g., Davis, the model depends critically on the separateness of the self and 1994; Schaller & Cialdini, 1988), that position does appear to the other; without a distinct self and other and without distinct have won the war in important respects. Even champions of motivations to aid the self or the other, it is not possible to egoistic accounts of helping have conceded that Batson and his detach altruism from egoism (Batson, 1987; Batson, in press; associates have provided credible experimental evidence for the Batson & Shaw, 1991). The second response has been to doubt existence of true selflessness in the human character (Archer, the likelihood that a merging of self and other ever genuinely 1984; Cialdini et al., 1987; Piliavin & Charng, 1990), and the occurs, "except perhaps in some mystical states" (Batson & empathy-altruism hypothesis is described as generally sup- Shaw, 1991, p. 161 ), and to argue that, in any case, feelings of ported in most contemporary texts in the field (e.g., Brehm & attachment and altruistic motivation do not blur the self-other Kassin, 1996; Brewer & Crano, 1994; Franzoi, 1996; Schroeder, distinction but may even intensify it (Batson, in press; Batson & Penner, Dovidio, & Piliavin, 1995; Smith & Mackie, 1995). Shaw, 1991; Schoenrode, Batson, Brandt, & Loud, 1986). Rooted as they are, though, in centuries of philosophical and However, accumulating evidence suggests that a merging of scientific thought, nonaltruistic explanations of human conduct self and other identity can occur and that it is most likely under are not easily dismissed. Indeed, the purpose of this article is conditions linked by the empathy-altruism model to feelings to propose a nonaltruistic reinterpretation of the data supporting of attachment and altruistic motivation: relationship closeness the empathy-altruism hypothesis. To do so properly, it is neces- and perspective taking. For instance, Davis, Conklin, Smith, sary to consider two relatively recent theoretical developments. and Luce (1996) found that individuals instructed to take the The first concerns the contemporary view of the self as dynamic perspective of another subsequently saw more of themselves as and malleable, especially in the moment (Higgins, 1996; Kihl- residing in the other. In addition, research by Aron and Aron strom & Cantor, 1984; Markus & Wurf, 1987; McGuire & Mc- has demonstrated that as relationship closeness escalates, so Guire, 1988). That is, although our self-conceptions are fairly does blurring of the self-other distinction, as measured by more stable over long periods of time, they can be made to shift similar self-other word associations, the selection of self-other temporarily, flowing across established boundaries with changes representations with greater boundary intersections, and longer in various factors, such as whom we are with, what immediate me/not me reaction time latencies on nonshared traits (Aron, goals we have, which aspects of the self are currently prominent, Aron, & Smollan, 1992; Aron, Aron, Tudor, & Nelson, 1991 ). and which roles we are instructed to play (Leary, 1995; Mar- kus & Nurius, 1986; Rhodewalt & Agustsdottir, 1986; Schlenker, The Self in the Other 1985). A second relevant theoretical development is the growing application of evolutionary principles to much of human social The Other as the Self behavior (Barkow, Cosmides, & Tooby, 1992; Buss & Kenrick, in press), including helping behavior (Burnstein, Crandall, & The notion of a responsive and fluid sense of self offers the Kitayama, 1994; Cunningham, 1986). Perhaps the evolutionary provocative possibility that when one takes the perspective of principle most pertinent to the empathy-altruism hypothesis is another (either through instructions or a feeling of attachment) Hamilton's (1964) concept of inclusive fitness, because it, too, and vicariously experiences what the other is experiencing, one undermines the distinction between self and other. Hamilton's comes to incorporate the self within the boundaries of the other. position, which is now widely accepted among evolutionary If true, such a process would seriously undermine the logic theorists (Alcock, 1993; Dawkins, 1989), is that individuals do of the empathy-altruism hypothesis. That is, if the distinction not so much attempt to ensure their own welfare and survival between the self and other is compromised by perspective taking, as to ensure the welfare and survival of their genes. This is a then so is the distinction between selflessness and selfishness. crucial point for conceptualizations of true altruism in that it It is important to recognize the nature and limits of this implies that the self in self-interest can lie outside of one's body proposed merger of self and other. What is merged is conceptual, and inside the skin of another--in the form of genes that may not physical. We are not suggesting that individuals with over- be shared with the other. By this logic, it can be selfish to give lapping identities confuse their physical beings or situations away resources in a helpful act, provided that the recipient gives with those of the other. If one were to stub a toe, we would not evidence of greater than average genetic overlap with the helper. expect the other to experience the same kind of localized, sharp Of course, it is not possible to detect one's degree of genetic pain--although, tellingly, the other might wince in a different overlap with another directly; instead, one must rely on discern- sort of pain. That is, the other might well be pained emotionally ible cues that are associated with a relatively high level of ge- that someone with whom he or she shares an identity has been netic commonality (Krebs, 1991 ). It is telling that these cues-- hurt. Thus, according to contemporary views of the self, it is kinship, friendship, similarity, and familiarity (Cunningham, GNITERPRETNIER THE EMPATHY-ALTRUISM PIHSNOITALER 483 1986; Rushton, Russell, & Wells, 1984; Well s, 1987)--are pre- that as relationship closeness intensified, so would feelings of cisely those that Batson and his associates contend produce true oneness (merged identity) with the other. Third, we expected altruism through their effect on attachment (Batson & Shaw, that extent of empathic concern would predict level of helping, 1991; Batson et al., 1995). as has been found in numerous prior studies (see Batson, 1991; Batson & Shaw, 1991, for reviews). Fourth, we expected that extent of felt oneness would also predict level of helping. Fifth, Oneness we expected that empathic concern would remain predictive of The upshot of the foregoing analysis is that close attachments helping after the influence of the egoistic factors of personal may elevate benevolence not because individuals feel more em- distress and sadness was removed, as has been shown previously pathic concern for the close other but because they feel more (e.g., Batson, O'Quin, Fultz, Vanderplas, & Isen, 1983; Fultz et at one with the other--that is, because they perceive more of al., 1986). Finally and most crucially, we expected that empathic themselves in the other. This experience of oneness--a sense concern would no longer be predictive of helping after the influ- of shared, merged, or interconnected personal identities--could ence of oneness was extracted but that oneness would be sig- come about in one or both of two ways? First, consistent with nificantly predictive when empathic concern was partialed out. contemporary theories of the self as situationally malleable This last prediction was based on our view that empathic con- (Kihlstrom & Cantor, 1984; Markus & Wurf, 1987; McGuire & cern is not the functional cause of increased helping under condi- McGuire, 1988), enhanced oneness could stem directly from tions of increased attachment but a concomitant of the functional the act of perspective taking that accompanies attachment (Bat- cause: perceived oneness. son & Shaw, 1991; Eisenberg, Fabes, & Miller, 1990) and that facilitates the symbolic merging or expansion of the self into A Methodological Rationale the other (Aron & Aron, 1986; Aron et al., 1992; Aron et al. 1991; Davis et al., 1996). Second, consistent with evolutionary In testing these expectations, we sought to avoid a criticism theory, felt oneness could arise as a consequence of exposure that had been leveled against the data gathering approach of to attachment-related cues (e.g., kinship, friendship, and famil- Batson and his colleagues (Cialdini, 1991; Sorrentino, 1991 ). iarity) that signal relatively high genetic commonality (Cun- The criticism is that, although the procedural rigor of individual ningham, 1986; Rushton, Russell, & Wells, 1984; Wells, 1987). experiments supporting the model is by and large quite good, In either event, perceived oneness provides a nonaltruistic alter- there is a meta-methodological weakness in the overall ap- native account of the findings that Batson and colleagues have proach. Consider the following sequence of experimentation. To attributed to altruistic motivation. If people locate more of them- examine whether a particular egoistic factor, concern about so- selves in the others to whom they are closely attached, then the cial approval, could account for seemingly selfless action, re- helping that takes place among such individuals may not be searchers conducted a set of studies that removed or controlled selfless. for the role of such concern and that still found evidence of To test this alternative to the empathy-altruism hypothesis, selfless responding (Fultz et al., 1986). With that egoistic factor we conducted an initial experiment that operationalized degree apparently dispatched, the focus shifted to another egoistic con- of attachment in terms of the closeness of the relationship be- tender, guilt, and another set of procedures and situations was tween two individuals. Recall that according to the empathy- devised to rule out the role of guilt (Batson et al., 1988). That altruism model (Batson & Shaw, 1991; Batson et al., 1995), done, the focus shifted to yet another egoistic factor, sadness, feelings of attachment stem from such factors as kinship, friend- and its mediational influence was tested and rejected in still ship, and familiarity. Consequently, participants in our study another set of experimental situations (Batson et al., 1989). focused on an individual who was a near stranger, an acquain- Finally, using a different experimental situation again, research- tance, a good friend, or a family member. They were then asked ers assessed the possibility that empathic joy could account for to indicate their willingness to provide various levels of aid obtained altruistic patterns and found it wanting (Batson et al., upon learning that this individual had been recently evicted from 1991). his or her apartment. Participants also rated the extent of oneness Despite the success of this sequential strategy in supporting they felt with the evicted person, as well as the amount of the empathy-altruism hypothesis, there is a problem with its empathic concern, personal distress (aversive arousal), and sad- one-at-a-time approach to dealing with egoistic alternatives. ness. Thus, our participants were not asked to speculate on what After all, the fundamental dispute the empathy-altruism model they might be feeling should this set of events occur. Rather, they rated what they were currently feeling, after having spent time focused on the need situation. In this fashion, we sought As ew have noted, eht sense of shared personal identities sah been to inquire into the genuine emotional responses of our partici- detseggus by earlier workers sa importantly implicated in eht helping pants in the same way that one might inquire into the genuine decision (Aron & Aron, 1986; Hornstein, 1982; Lemer, 1982; Piliavin emotional responses of individuals viewing a movie or other et al., 1891 ). These workers have tended ot esu eht term we-ness ot refer ot this shared sense of identity (see especially Piliavin et al., 1891 ). engaging depiction. Since those earlier statements, ,revewoh we-ness sah frequently been We had several expectations as to the experimental outcomes. used ot denote a feeling of group membership or social identity First, in keeping with the empathy-altruism model (Batson & (Brewer & Gardner, 1996; ,renruT 1987). Thus, ew have opted for eht Shaw, 1991 ), we expected that as relationship closeness (attach- term oneness to etaitnereffid it from a sense of regrem with a ytivitcelloc ment) intensified, so would feelings of empathic concern. Sec- dna to reserve it for that which occurs in a dyadic relationship with a ond, in keeping with our oneness-based alternative, we expected deificeps .rehto 484 CIALDINI, BROWN, LEWIS, LUCE, AND NEUBERG attempts to resolve is not between altruism, considered as a if you passed each other on campus." Participants in the acquaintance concept, and various egoistic concepts, considered indepen- condition described "a man/woman who you don't know really well, dently; rather, it is between altruism as a whole and egoism as but you would stop and chat with him/her for a few minutes if you passed each other on campus." Participants in the good friend condition a whole, which has not been tested. Ruling out the influence of described "a man/woman who is a friend of yours, who you sometimes one egoistic motive in a specific situation does not rule out go out with outside of school." Participants in the family member condi- egoism due to other factors there. Moreover, the one-egoistic- tion described "your closest male/female family member, a sibling if factor-at-a-time approach ignores the plausible possibility that possible." different egoistic motives may be active (and inactive) in differ- ent situations. The desire to reduce sadness, for instance, may Dependent Variables not be strong or even applicable in one setting but may be quite prominent in another. If so, then demonstrating that a single Helping measure. All participants were asked to consider that the egoistic motivator, such as sadness, fails to account for helping described person "was just evicted from his/her apartment." Partici- in a particular experimental situation does not eliminate its pos- pants were then asked to indicate the level of help (if any) they would sible influence on helping in a different experimental situation be willing to give the evicted person by choosing one of seven increas- ingly costly helping options: nothing, give him or her an apartment guide, where it has not been removed or held constant. The implication help him or her find a new place to live by driving him or her around is that investigations of the empathy-altruism hypothesis should for a few hours, offer to have him or her come stay with you for a properly include procedures that allow for the simultaneous couple of days (provided you had space), offer to have him or her come elimination of multiple nonaltruistic motives. To that end, the stay with you for a week (provided you had space), offer to have him present research incorporated design and analysis procedures or her come stay with you until he or she found a new place (provided that allowed the simultaneous consideration of the impact on you had space), and offer to let him or her come live with you rent- helping of an altruistic motivator--empathic concern--and free (provided you had space). In all analyses, the level of cost associ- three nonaltruistic motivators--personal distress, sadness, and ated with each of these seven options was determined in pilot work in oneness. which 21 introductory psychology students rated (on a 0 to 100 scale) each of the options in terms of its cost to the helper. The resultant ratings ydutS (divided by 10) formed weights that were applied to each option. For 1 the seven options, the weights were, respectively, 0.0, 0.6, 2.9, 3.6, 4.2, 5.9, and 8.3. Method Mediational measures. After responding to the helping measure, Participants participants rated the extent of empathic concern they felt for the evicted person, as measured by the four empathic adjectives used by Batson Forty-four male and 46 female introductory psychology students at et al. (1995) to constitute their empathic concern index sympathetic, Arizona State University participated in the experiment in return for compassionate, softhearted, and tender. Participants also rated the course credit. We assured participants that their responses would remain amount of personal distress (aversive arousal) and sadness they felt, as anonymous in order to discourage stated helping on the basis of social measured by emotion adjectives suggested by Fultz, Schaller, and Cial- approval pressures. dini (1988) to reflect personal distress (alarmed, worried, uneasy) and sadness (sad, low-spirited, and heavy-hearted). We administered the Procedure personal distress and sadness items to assess egoistic affect. The em- pathic concern, personal distress, and sadness adjectives were intermixed In a study said to be investigating impression formation, participants on the questionnaire and were rated along 7-point scales with anchors were asked to focus on a particular individual who, depending on condi- at not at all (1) and extremely (7). tion, was a near stranger, an acquaintance, a good friend, or a family Additionally, participants rated the extent of oneness they felt with member (preferably a sibling). oT instantiate their focus, participants the evicted person by responding to two items that were combined in described in writing, as best they could, the individual's physical charac- all analyses to form a oneness index. The first item incorporated the teristics, personality traits, interests, values, and attitudes. Next, they Inclusion of Other in the Self (lOS) Scale used by Aron et al. (1992) were asked to consider a need situation in which the described individual to measure perceived self-other boundary ovedap. It consisted of a set had recently been evicted, to concentrate on that situation, and to indicate of seven pairs of increasingly overlapping circles. Participants selected the level of aid they would be willing to provide him or her in that the pair of circles that they believed best characterized their relationship situation. They also rated the amount of sadness, personal distress, em- with the evicted person. The second item asked participants to indicate pathic concern, and oneness they were feeling toward the described on a 7-point scale the extent to which they would use the term we individual. Each of these ratings followed the willingness-to-help assess- to describe their relationship with the evicted person. For purposes of ment, with the exception of the oneness measure, which was counterbal- counterbalancing, the oneness index items appeared either immediately ~" anced so that it occurred either before or after all other measures. The after participants engaged in their description of the target person or timing of the oneness measure had no significant and systematic impact after all other measures were taken. in this or any of the subsequent studies; therefore, this order factor is not considered further. Results Independent Variable Before examining our specific predictions, we performed an analysis of variance (ANOVA) testing our general expectation We manipulated relationship closeness by instructing participants to that relationship closeness would lead to greater willingness to think about and describe one of four same-sex persons: a near stranger, an acquaintance, a good friend, or a family member. Participants in help. That analysis produced a supportive and highly significant the near stranger condition described "a man/woman you don't really effect, F(3, 82) = 33.28, p < .001. The helping means associ- know...someone you would recognize from class, but not say 'hello' to ated with each of the levels of relationship closeness are pre- REINTERPRETING THE EMPATHY-ALTRUISM RELATIONSHIP 485 Table 1 Means of Helping, Empathic ,nrecnoC and Oneness Scores as a Function of Level of Relationship ssenesolC and Need Situation Level of relationship closeness Near Good Close family Need situation stranger Acquaintance friend member Study :1 Eviction Helping 1.20, b31.4 ~36.6 c98.6 Empathic concern 3.04, 063.4 b12.4 b05.4 Oneness 1.52, o61.3 c25.4 c75.4 n 22 22 20 22 Study 2: Orphaned children Helping 4.13, 6.11 ~ b69.7 9.01 b Empathic concern ~24.4 b25.5 b58.5 b28.5 Oneness ~09.1 b81.3 c42.5 c61.5 n 51 71 71 91 Study 3 Phone call Helping 0.80, b~89.0 b45.1 b55.1 Empathic concern a78.2 b~94.3 b55.4 b66.4 Oneness ~71.2 3.16, b34.4 066.4 n 33 81 20 91 Eviction Helping a77.1 b36.3 c88.5 ¢59.6 Empathic concern 3.56, bL43.4 c.b09.4 ~66.5 Oneness a61.2 b65.3 ~00.5 466.5 n 27 91 20 61 Orphaned children Helping ~51.4 ~63.5 b32.8 b38.8 Empathic concern ~35.4 4.51, b.,14.5 b12.6 Oneness 2.40, ~20.3 b84.4 b08.4 n 20 23 91 20 .etoN Within each row, means sharing the same subscript are not significantly different by Tukey test. sented in Table 1. Next, we examined the set of hypotheses istic affect (personal distress and sadness) had been removed designed to provide information about the mediation of this but would no longer be predictive after the influence of one- basic effect. ness had been extracted. A test of those hypotheses required a hierarchical regression analysis in which helping served as Effect of Relationship Closeness on Empathic Concern the criterion variable and empathic concern, personal distress, and Oneness sadness, and oneness served as predictors. In that analysis, participant gender was entered in a first step and had no sig- Our first and second hypotheses, that relationship closeness nificant impact, b = .01, F(1, 82) < 1. When entered on would lead to greater empathic concern and to greater oneness, the second step, empathic concern was a highly significant were supported by ANOVAs demonstrating a significant impact predictor of helping, b = 0.90, F(1, 81) = 19.52, p < .001. of relationship closeness on empathic concern, F(3, 82) = 5.37, Furthermore, consistent with our fourth hypothesis, when the p < .01, and on oneness, F(3, 86) = 30.58, p < .001. The egoistic affect adjectives were entered on the third step, em- means for these variables are presented in Table 1. pathic concern remained significantly related to helping, though less so, b = .62, F(1, 75) = 4.09, p = .05. Finally, Relation of Empathic Concern and Oneness to Helping supportive of our sixth hypothesis, when oneness was entered on the last step of the analysis, it both predicted helping pow- Our third and fourth hypotheses, that empathic concern and erfully, b = 1.10, F(1, 74) = 48.27, p < .001, and rendered oneness would predict helping, were confirmed by correlational the impact of empathic concern nonsignificant, b = .11, F(1, analyses showing significant relationships between helping 74) < 1. In addition, the unique proportion of variance ac- scores and scores on both empathic concern (r = .45, p < .01) counted for by oneness was many times that for empathic and oneness (r = .76, p < .01). concern (30% vs. 0.36%). This last step in the regression procedure highlights the fundamental asymmetry between Mediation of Helping oneness and empathic concern as mediators of help. That is, controlling for oneness eliminated the influence of empathic Our fifth and sixth hypotheses were that empathic concern concern, whereas controlling for empathic concern left one- would remain predictive of helping after the influence of ego- ness a powerful predictor of willingness to help. 486 CIALDINI, BROWN, LEWIS, LUCE, AND NEUBERG Discussion deeN .noitautis After participants focused on and described a near stranger, an acquaintance, a good friend, or a close family member In several ways, the outcomes of Study 1 are congruent with (preferably a sibling), they were asked to consider that the described what has been found by Batson and his associates and with individual "died in an accident leaving his/her two children without a what would be predicted according to the empathy-altruism home.' ' model. First, in keeping with statements of that model (Batson, gnipleH .snoitpo Participants indicated the amount of help they 1987, 1991), relationship closeness led to greater feelings of would be willing to give by choosing one of seven helping options: empathic concern for a needy other. Second, as has occurred in nothing, donate $10 toward a fund for the kids, donate $25 toward a much prior research (see Batson & Shaw, 1991, for a review), fund for the kids, donate $50 toward a fund for the kids, start a fund- raising campaign for the kids' welfare, have the kids come live with levels of empathic concern predicted willingness to help. More- you until a permanent home was found, and have the kids come live over, once again consistent with previous work (e.g., Batson et with you and raise them as you would your own. As in Study ,1 each al., 1983; Fultz et al., 1986), when the influence of the egoistic option was weighted according to pilot work, which for this study pro- factors of personal distress and sadness were extracted, a sig- duced helping weights of 0.0, 3.0, 4.2, 5.7, 6.0, 8.0, and 9.2, respectively. nificant relationship between empathic concern and helping re- mained. These parallels between the present findings and those Results of past studies offer a necessary degree of reassurance that our procedures and measures adequately manipulate and capture the Our first analysis was an ANOVA designed to examine the variables under consideration. influence of relationship closeness on willingness to help. That Two additional findings of Study 1 suggest a nonaltruistic analysis demonstrated a powerful impact of relationship close- alternative to the empathy-altruism model, however. First, in- ness on helping responses, F(3, 70) = 17.43, p < .001. Table creasingly close relationship attachment to another person pro- 1 provides the relevant means. duced an increasingly elevated perception of self-other oneness, raising the possibility that the enhanced helping that regularly Effect of Relationship Closeness on Empathic Concern takes place in attached relationships does not occur altruis- and Oneness tically-without consideration of its impact on the self. Support for this possibility can be seen in the second additional finding: As in Study 1, our first and second hypotheses, that relation- When the effect of oneness was extracted from the analysis, the ship closeness would lead to greater empathic concern and to empathy-altruism model's proposed source of altruistic motiva- greater oneness, were tested with ANOVAs that showed a sig- tion (empathic concern) no longer had any impact on helping. nificant effect of relationship closeness on empathic concern, One implication of these findings is that empathic concern had F(3, 70) = 6.73, p < .001, and on oneness, F(3, 64) = 27.75, appeared to motivate helping in past work only because it is a p < .001. Relevant means appear in Table .1 concomitant of oneness, which had not been measured or taken into account. Before assigning much confidence to such contentions, how- Relation of Empathic Concern and Oneness to Helping ever, we felt it necessary to replicate the basic findings of Study We examined our third and fourth hypotheses, that empathic 1 and to do so with a different need situation. It is conceivable, concern and oneness would predict helping, with correlational for example, that the eviction predicament that served as the analyses that, as in Study ,1 showed significant relationships helping context in Study 1 somehow obscured the impact of between helping scores and both empathic concern (r = .33, p altruistic motivation on our helping measures. Therefore, to es- < .01), and oneness (r = .53, p < .01). tablish the generality of our results, we decided to conduct a follow-up study in which the needy person could not be deemed responsible for his or her plight. To assure that the empathy- Mediation of Helping altruism hypothesis would have a good opportunity for support, As in Study ,1 we tested our mediational hypotheses with a we chose a type of victim that had been used frequently in hierarchical regression analysis in which helping served as the studies confirming that hypothesis (Batson et al., 1991; Coke criterion variable and empathic concern, personal distress, sad- et al., 1978) and that was likely to evoke a high degree of ness, and oneness served as predictors. The effect of participant empathic concern: recently orphaned children. gender was partialed out in step 1 of the analysis and proved Study 2 nonsignificant, b = -.21, F( ,1 65) < .1 When empathic concern was entered on step 2, it was a significant predictor of helping, Method b = .84, F(1, 64) = 15.39, p < .01. Although adding the Participants egoistic affect (personal distress and sadness) adjectives at step 3 reduced the influence of empathic concern, it remained a Thirty-eight female and 36 male introductory psychology students at Arizona State University participated in return for course credit. Again, significant predictor of helping, b = .72, F( 1, 58) = 5.14, p < participants were assured that their responses would be anonymous. .03. Just as had occurred in Study 1, however, when oneness was introduced to the analysis on the final step, it not only Procedure significantly influenced helping, b = .57, F(1, 57) = 10.39, The procedures and measures of Study 2 were identical to those of p < .01, but it reduced the impact of empathic concern to Study 1 with the exceptions of the need situation and the consequent nonsignificance, b = .37, F(1, 57) = 1.37, .sn Furthermore, helping options. the unique proportion of variance accounted for by oneness REINTERPRETING THE EMPATHY-ALTRUISM RELATIONSHIP 487 was again several times greater than that for empathic concern variable was identical to that of Studies 1 and 2. The severity of need (10.4% vs. 1.4%). variable was composed of three levels of need, the lowest of which was introduced for the first time. It consisted of a need situation in which the target person required aid in making a phone call. The next highest Discussion level of need consisted of the eviction situation used in Study .1 The highest need level consisted of the orphaned children situation used in Study 2 replicated exactly the findings of Study 1. It is note- Study 2. worthy that this was so despite substantial differences between In order to create a continuous variable for situation severity as well the studies in the victims and need situations considered by as validate our severity levels, a weighting scheme we developed from participants. Although consistency of this sort enhanced our pilot work in which 25 participants were asked to compare the three confidence that our model might apply across a wide range of situations and assign a value between 0 and 10 that would indicate "how helping situations, we recognized that additional work would important it is that this person receive help." The means of these values be necessary to establish such generality. To that end, we decided were calculated so that they could be used as a weighting factor for to test our model in a need setting quite different from those of situation severity in our analyses. Consistent with our operationaliza- our initial two studies. That is, Studies 1 and 2 involved rather tions, the phone call scenario was weighted 3.3, the eviction scenario severe and unusual forms of need. We wondered whether one- was weighted 6.6, and the orphaned children scenario was weighted 8.8. ness would have a similar mediational influence on minor and Dependent Variables mundane types of aid. One small kind of everyday help, aid in making a phone call, seemed a desirable choice in that it has gnipleH .serusaem The helping measure used in the phone call need been used frequently in helping research and has been shown situation asked participants to indicate the level of help they would be to be responsive to an array of social psychological factors willing to provide by choosing one of seven helping options: nothing, (Cialdini, Darby, & Vincent, 1973; Gaertner & Bickman, 1971; stop to tell him or her where the nearest pay phone is located, help him or her find a phone, drive him or her to a phone that is 5 min away Isen, Clark, & Schwartz, 1976; Manucia, Baumann, & Cialdini, (assume you have a car), drive him or her to a phone that is 15 min 1984). Therefore, in a third study we asked participants to away (assume you have a car), cut class to drive him or her to a phone decide how much help they would be willing to provide to an (assume you have a car), and cut class on the day of the exam to drive individual who needed aid in making a phone call. To contrast him or her to a phone (assume you have a car). As in the earlier studies, this level of need severity against much higher levels, we repli- each helping option was weighted according to pilot work, which in cated the need situations of Studies 1 and 2 as well. this instance produced respective weights of 0.0, 0.3, 1.0, 1.4, 1.9, 2.9, More than providing information about the generality of our and 7.7. The helping options for the eviction and the orphaned children model, varying severity of need within our design provided the need situations were identical to those of Studies 1 and 2, respectively. opportunity to test a hypothesis that comes from the evolution- lanoitaideM .serusaem The mediational measures designed to assess ary perspective on helping which is gaining prominence within empathic concern, sadness, personal distress, and oneness were identical to those of Studies 1 and 2. On an exploratory basis, for each need social psychology (Burnstein et al., 1994; Buss & Kenrick, in situation, we also included several items inquiring into egoistic motives press; Cunningham, 1986). A fundamental assumption within that would be specific to that need situation (situational egoistic mo- this perspective is that as indications of genetic commonality tives). For example, those participants exposed to the phone call need between individuals increase, so will willingness to offer assis- situation were asked how much they would want to help someone find tance-an assumption that was supported in our initial two a phone in order to enhance a friendship. Those exposed to the eviction studies by the significant effect of relationship closeness on need situation were asked how much they would want a roommate in helping. However, evolutionary theory makes an additional, order to enhance the security of their home. Those exposed to the or- qualifying assumption: This effect should occur primarily in phaned children need situation were asked how much they would want situations of severe need where survival of the needy other (and to raise children to gain love. These items, considered individually, did of his or her genes) could be at risk (Burnstein et al., 1994). not prove instructive in our analyses; consequently, we do not detail them further. Thus, a second purpose of Study 3 was to test the prediction that the influence of relationship closeness on helping should Results be strongest in the relatively severe need contexts involving eviction and orphaned children. A Relationship Closeness × Severity of Need ANOVA was conducted on the helping measure. Replicating the pattern of Study 3 Studies 1 and 2, there was a significant main effect for relation- ship closeness, F(3, 243 ) = 62.35, p < .001. A significant main Method effect of severity of need also appeared, indicating that helping increased as need increased, F(2, 243) = 217.62, p < .001. Participants More important, a significant interaction emerged, supporting Eighty-two male and 181 female introductory psychology students at the evolutionary-theory-based prediction that the impact of rela- Arizona State University responded to an experimental questionnaire in tionship closeness on helping would be more pronounced in the return for course credit. As in Studies 1 and 2, all participants were higher need situations, F(6, 243) = 10.33, p < .001. Figure 1 made aware that their responses would be anonymous. presents the pattern of these effects. Independent Variables Mediation of Helping: Within Levels of Need Two independent variables, relationship closeness and severity of Phone call situation. Our mediational hypotheses were need, were crossed in a 4 × 3 factorial design. The relationship closeness tested with a hierarchical regression analysis in which helping 488 CIALDINI, BROWN, LEWIS, LUCE, AND NEUBERG influence, b = .87, F(I, 50) = 27.15, p < .001, that rendered Family Member nonsignificant empathic concern's role in helping, b =. 18, F( ,1 50) = 2.15, ns. Again, oneness accounted for much more of the unique variance than empathic concern (14% vs. 0.7%). Orphaned children situation. At step 1 of a similar regres- Close Friead sion procedure done on the data from the orphaned children need situation of Study 3, participant gender was not significant, b = -.92, F(1, 73) = 1.66, ns. At step 2, empathic concern was entered and was predictive of helping, b = .55, F(1.73) = Aequaiatanee Amount 5.07, p < .05. Entering the items constituting personal distress, of sadness, and situational egoistic motives at step 3 rendered em- gnipleH pathic concern somewhat less predictive, b = .65, F( ,1 51 ) = 2.95, p < .09. At step 4, oneness was entered, again proving significant, b = .79, F(1, 50) = 16.13, p < .001, and again removing the significant impact of empathic concern on helping, b = .36, F(1, 50) = 1.12, ns. As before, the unique proportion of variance accounted for by oneness was many times that for empathic concern (13.5% vs. 0.9%). Thus, the mediational pattern obtained in Studies 1 and 2 reappeared in virtually identical form at each level of need in Study 3. I I I Mediation of Helping Across Levels of Need enohP Call Eviction Orphaned nerdlihC We used the EQS program (Bentler & Wu, 1995) to explore the mediation of helping as it occurred across the various need Severity of Need levels explored in Study 3. Relationship closeness, severity of need, and their interaction (calculated after we centered the two erugiF .1 Amount of helping in each relationship type as a function variables) served as exogenous variables. We created compos- of severity of need. ites of oneness, empathic concern, and personal distress on the basis of the simple averaging of their constituent items; these composites, along with the sadness item, were assessed as poten- served as the criterion variable and empathic concern, personal tial mediating predictors. 2 Finally, the helping measure served distress, sadness, situational egoistic motive items (as a group), as the outcome variable. and oneness served as predictors. Participant gender was entered Supporting our hypotheses, oneness was the only explored at step 1 of the analysis and proved marginally significant, indi- predictor to mediate the effects of the manipulated variables on cating a tendency for women to be more helpful, b = -.38, helping. Although relationship closeness and severity of need F(1, 84) = 3.31, p = .07. At step 2, empathic concern was did indeed increase levels of empathic concern, personal dis- significantly related to helping, b = .26, F( ,1 83) = 26.06, p tress, and sadness, these latter constructs did not influence help- < .0001. At step 3, the items related to personal distress, sad- ing when oneness was in the model. Figure 2 presents the best ness, and the situational egoistic motives were entered, reducing fitting model. This model captures the observed data quite well, but not eliminating the relationship of empathic concern to help- X2( 12, N = 236) = 16.70,p > .15, CFI = .995, and all depicted ing, b = .18, F( ,1 66) = 6.84, p < .05. However, when oneness paths are significant atp < .01 or less, two-tailed. No other paths was introduced at step 4, not only did it prove a significant reached conventional levels of significance, nor does adding or predictor of helping, b = .21, F(1, 65) = 13.39, p < .001, it eliminating paths significantly improve the model's fit. 3 More- rendered nonsignificant the effect of emphatic concern on help- ing, b = .08, F(1, 65) = 1.46, ns, and it accounted for much 2 Because the three sadness items revealed somewhat different patterns more unique variance than did empathic concern (7.7% vs. of influence in the studies reported here, we felt it inappropriate to 0.8%). Thus, even within this minor form of need, the pattern aggregate them into a composite. Findings from the following analyses of our earlier studies was replicated exactly. remain the same when we use the low-spirited or heavy-hearted items Eviction situation. The same type of regression analysis was in our models instead. conducted on the data from the eviction need situation of Study 3 Readers should note two features of the path model. First, for pur- poses of visual clarity, we do not depict the covariances among our 3. It produced comparable results. At step 1, the influence of manipulated variables and their interaction. However, adding these three gender was not significant, b = .57, F(1, 73) < .1 At step 2, paths does not improve the fit of the model, and none of these paths entering empathic concem revealed its significant relation to approaches statistical significance (all p s > .35). Second, note that the helping, b = 1.09, F(1, 72) = 38.32, p < .001. At step 3, model allows for covarying, as opposed to causal, relationships among the introduction of personal distress, sadness, and situational the psychological variables. (Covariances were estimated on these items' egoistic motive items reduced but did not eliminate the pre- error terms, which are not depicted in Figure 2 for reasons of visual dictiveness of empathic concern, b = .56, F( ,1 51 ) = 4.03, p clarity.) This decision was predicated on technical issues and not a < .06. At step 4, oneness added an independent significant theoretical stance. In the present model, the covariance arrows could REINTERPRETING THE EMPATHY-ALTRUISM RELATIONSHIP 489 I I 656. pihsnoitaleR ssenesolC 242. ssenesolC X ~ deeN ssertsiD 394.1 456. 422, .3. / 1 ytireveS ~ 3 1 8 ~ Sadness of deeN 515. erugiF .2 A path model representing the effects of relationship closeness and severity of need on helping, as potentially mediated by oneness, empathic concern, personal distress, and sadness. The model fits the data well and illustrates the sole mediational effect of oneness. over, alternative models--for instance, one in which the putative empathic concern had no such influence once perceptions of psychological mediators are viewed merely as consequences, oneness were taken into account. Extending these results, the rather than causes, of helping--resulted in a significantly de- unique mediating influence of oneness revealed itself as well graded fit to the data. In sum, consistent with the regression when participants in Study 3 confronted a lower need helping analyses conducted within each level of need, path analyses decision, whether to aid a person wishing to make a phone call. conducted across all three need levels revealed a pattern whereby And a path analysis conducted across the three need settings of perceptions of oneness (a) had an appreciable mediating influ- Study 3 further revealed that decisions to help were influenced ence on helping and (b) reduced the mediating role of empathic by participants' perceptions of oneness but not by their levels of concern to nonsignificance. empathic concern. That path analysis suggested a bidirectional causal relationship between the more cognitive variable of one- Discussion ness and the more emotional variable of empathic concern. Such bidirectionality is consistent with our view that the perception Study 3 replicated and added to .the findings of Studies 1 and of oneness with a needy other generates empathic concern and 2. As before, perceptions of oneness accounted for helping in the that the experience of empathic concern generates the perception relatively high need eviction and orphan contexts; furthermore, of oneness. However, it appears to be oneness and not empathic concern that mediates help. Although the results of Study 3 answer the question of be replaced with directional arrows--pointing in either direction--with whether oneness mediates the impact of empathic concern on no appreciable loss of fit. Indeed, consistent with theoretical expecta- helping even at low levels of need, they do not address whether tions, further analyses suggested the likelihood of bidirectional causal such mediation occurs even at low levels of relationship close- influences among these variables. Unfortunately, technical concerns-- ness. The latter question is worthy of attention, as support for related to the need to independently "anchor" all variables in bidirec- the empathy-altruism model has come almost exclusively from tionally causal relationships--precluded a formal test of this possibility. studies of near strangers. Participants in such studies typically 490 CIALDINI, BROWN, LEWIS, LUCE, AND NEUBERG rate their feelings of empathic concern toward a stranger or near ness increased, so did empathic concern for a needy other. In stranger after being exposed to that person's suffering. 4 The each, empathic concern significantly predicted willingness to standard experimental outcome is that higher helping scores are help. And in each, it did so even after the influence of the egoistic linked to higher empathic concern (Batson & Shaw, 1991 ). factors of personal distress and sadness had been removed. More To examine the effects of empathic concern and oneness on telling for the purposes of this research, however, was a fourth willingness to help another with whom one has only a minimal type of consistent internal replication: Invariably, when a nonal- relationship, we analyzed the data of just those individuals in truistic factor that covaries with empathic concern was intro- our near stranger condition (across all levels of need). The duced to the analyses, it reduced the impact of empathic concern results were identical in form to those we had obtained when to nonsignificance. That nonaltruistic factor, oneness, reflects a analyzing across relationship types: When entered at step 1 of sense of interpersonal unity, wherein the conceptions of self and a regression procedure, participant gender had no significant other are not distinct but are merged to some degree. influence on helping (b = -.53, F < 1 ); when empathic concern The implications of these results for the empathy-altruism was entered at step 2, it proved significantly predictive, b = .66 model are considerable. If the circumstances specified in the F(1,72) = 14.57, p <. 001; entering items constituting personal model as leading to truly altruistic acts (interpersonal closeness distress, sadness, and situational egoistic motives at step 3 ren- and perspective taking instructions) are the same circumstances dered empathic concern somewhat less predictive, b = .31, F( 1, that enhance the merging of self and other, as has been shown 55) = 1.40, p < .25; finally, at step 4, oneness was entered, in the present studies as well as earlier studies (Aron et al., proving significant, b = .39, F(1, 54) = 10.26, p <.01, and 1991; Aron et al., 1992; Davis et al., 1996), then one can doubt completely eliminating any influence of empathic concern on whether those helpful acts reflect the selflessness required of helping (b = .08, F < 1 ). Clearly, then, the basic form of our true altruism. As even the proponents of the model admit, if self findings is not limited to close relationships. 5 and other are not sharply distinct in a helper's mind, it is not Two other findings are of interest in Study 3. First, the path possible to separate egoism from altruism in a helper's motive analysis demonstrated that relationship closeness, severity of (Batson, 1987; Batson, in press; Batson & Shaw, 1991). After need, and their interaction influenced helping in an additional all, as the self and other increasingly merge, helping the other manner not captured by the mediating variables explored here, increasingly helps the self. making it likely that other psychological constructs account for Moreover, one can doubt the empathy-altruism hypothesis these effects. This result underscores the multiplicity of causes even further when, as we have demonstrated, oneness both cova- for help that typically apply in any given situation. Even though ries with empathic concern and is the functional mediator of oneness stood above the other mediating factors that we investi- helping when the two factors are considered simultaneously. gated, we do not wish to be misinterpreted as suggesting that That is, although relationship closeness elevated the levels of it is the only influence to be seriously weighed. Second, the both factors, only one (perceived oneness) predicted helping Relationship Closeness × Severity of Need interaction revealed when the influence of the other factor was controlled. Overall, a pattern wherein the impact of relationship closeness on helping then, our findings suggest that empathic concern may have only increased as severity of need increased: In low-need circum- appeared to mediate aid in much prior research because it is stances, relationship closeness had only a small positive effect a concomitant of perceived oneness, a construct that offers a on helping; in higher need circumstances, however, relationship nonaltruistic path to such aid. closeness had a profound influence. This pattern is compatible An attentive reader may have noticed that throughout this with an evolutionary perspective in which concern for a close article we have avoided characterizing oneness as an egoistic other (and thus a likely carrier of common genes) should be motivator of helping, preferring instead the label nonaltruistic. especially potent when that other's survival is at risk (Burnstein Such usage reflects an important feature of our argument: When et al., 1994). The pattern, which may also be compatible with the distinction between self and other is undermined, the tradi- formulations based on social or cognitive factors such as the social expectations or the need salience associated with different 4 Sometimes empathic concern is measured after a perspective-taking levels of severity, is not completely consistent with a strict evolu- manipulation has focused participants on the other's plight; other times, tionary model, however. That is, as Table 1 shows, although as in our studies, it is simply measured as a naturally occurring response help was always greater in the close family member condition to suffering. However, these two approaches, the manipulation of em- than in the good friend condition, it was never significantly so. pathic concern or its simple assessment, have produced comparable A proponent of a strict evolutionary model may wish to explain demonstrations of the facilitative effect of empathic concern on helping this in terms of the documented tendency of friends to perceive (Batson et al., 1989; Batson et al., 1983; Fultz et al., 1986). an extremely high degree of phenotypic similarity with one 5 Indeed, the form of these findings does not seem dependent on level another (Cunningham, 1986). Our own view, however, is that of relationship at all, as comparable analyses done within each of the other three levels of relationship closeness in Study 3 produced findings a variety of factors besides genetic commonality predict helping comparable to those done within the near stranger condition. In fact, in in high-need situations, including those factors associated with no instance within any of the 21 relationship conditions across our the social aspects of friendship--a view we think most evolu- three studies was empathic concern still predictive of helping after the tionary theorists would not find objectionable. influence of oneness was extracted. These results offer assurance that our data patterns were not caused by a stronger correlation between General Discussion rated oneness and our manipulated variable of relationship closeness than The data patterns of the three studies of this investigation are between empathic concern and relationship closeness. Similar patterns compelling in their consistency. In each, as relationship close- emerged within and across relationship levels.