Paying It Forward: Generalized Reciprocity and the Limits of Generosity

When people are the victims of greed or recipients of generosity, their first impulse is often to pay back that behavior in kind. What happens when people cannot reciprocate, but instead have the chance to be cruel or kind to someone entirely different—to pay it forward ? In five experiments, participants received greedy, equal, or generous divisions of money or labor from an anonymous person, and then divided additional resources with a new anonymous person. While equal treatment was paid forward in kind, greed was paid forward more than generosity. This asymmetry was driven by negative affect, such that a positive affect intervention disrupted the tendency to pay greed forward. Implications for models of generalized reciprocity are discussed.

2010; Dufwenberg, Gneezy, Güth & van Damme;Fowler & Christakis, 2010;Hamilton & Taborsky, 2005;Nowak & Sigmund, 2005;Stanca, 2009); however these paradigms often give the appearance of interdependence because they involve both extensive interpersonal interaction and/or small group settings-factors which are generally missing from naturally occurring (or "well-mixed") social groups (Rankin & Taborsky, 2009). In contrast to these paradigms, realworld-and media accounts-of paying it forward often involve brief, one-time encounters with strangers (e.g., choosing to pay a stranger's layaway bill; Beck, 2011). The first goal of the current experiments is to examine whether behavior is paid forward in such anonymous situations.
The second goal is to examine which kinds of behavior are paid forward. Past research has investigated only one class of behavior-minimally altruistic behavior (i.e., small acts of help)-but here we distinguish between equality, generosity, and greed. Research suggests that equality is a deep-seated behavioral norm: at default, people split outcomes equally (i.e., 50/50; Messick & Schell, 1992), and even children and dogs show inequality aversion (Fehr, Bernhard, & Rockenbach, 2008;Range, Horn, Viranyi, & Huber, 2009). As a result, we predict that when people receive equality, they will pay forward this treatment in kind. Equality can be violated in both the positive (more-than-fairness) and negative (less-than-fairness) directions, and it is unknown whether these behaviors are differentially paid forward 1 . Generosity may propagate more than greed because generosity seems relatively rarer-at least to the news-watching layperson-and the amplified associated affective reactions engendered by rare and novel events may compel people to pay forward generosity (Ohman,Eriksson,Fredrikson,Hugdahl,& 1 We note that economists and evolutionary biologists consider any behavior above pure self-interest as generous. Here we define generosity as treating others more than equally (i.e., more than fair), so that we can tease apart the relative effects of different kinds of positive behaviors. Olofsson, 1974). Arguing for the relative power of greed over generosity, however, is an extensive body of research showing that negative stimuli evoke stronger responses and exert a greater influence on subsequent human and animal behavior than do positive stimuli (Baumeister, Bratslavsky, Finkenauer & Vohs, 2001;Fiske, 1980;Ito, Larsen, Smith & Cacioppo, 1998;Keysar, Converse, Wang & Epley, 2008;Rutte et al, 2006;Taylor, 1991; but see Wang, Galinsky, & Murnighan, 2009). As a result, we predict that greed would be paid forward more than generosity.
Our third goal is to examine the underlying affective drivers of paying behavior forward.
While positive emotions (e.g., gratitude) lead people to pay forward generosity (Bartlett & DeSteno, 2006;DeSteno et al., 2010), these investigations do not include conditions under which people received equal or greedy treatment. Receiving greed likely engenders negative affect (Sanfey, Rilling, Aronson, Nystrom, & Cohen, 2003), which in turn typically exerts more influence than positive affect; thus, we predicted that negative affect will drive paying it forward behavior when considering the full suite of behaviors: equality, generosity, and greed.

Overview of the Experiments
We explore paying forward greed, equality and generosity in the form of both money (Experiments 1 and 2) and labor (Experiments 3, 4a and 4b). Experiment 2 explores the potential effect of overall resource distribution on paying forward greed versus generosity by manipulating recipients' initial financial endowment. Experiments 4a and 4b investigated the role of affect in paying it forward via both mediation and moderation.

Experiment 1: Paying Forward Money
One-hundred participants (51 female, M age = 23.4) were recruited from subway stations and high traffic tourist areas in a Northeastern city. After giving consent, individual participants were led to an isolated bench and told they would play an anonymous economic game in which one person splits money between themselves and another person (i.e., a dictator game).
Participants were assigned to one of four conditions. In the control condition (the giveonly condition), participants assumed the role of the "splitter" in a standard dictator game, and received $6 to split between themselves and an anonymous future receiver. They kept as much as they wished, then sealed the rest in an envelope labeled "future receiver" and returned it to the experimenters.
In the three other conditions (the greedy, equality, and generous pay-it-forward conditions), each participant acted as both the receiver and the splitter. First, these participants received an envelope with money ostensibly left to them by a previous splitter. Upon opening the envelope, they saw that the previous splitter had given them a greedy ($0 of $6), equal ($3 of $6), or generous ($6 of $6) split. Participants rated the fairness of this split on a 5-point scale (1: much less than fair, 3: fair, 5: much more than fair). Participants then acted as the splitter in a second dictator game, splitting an additional $6 between themselves and a different future receiver. Thus, participants in the three pay-it-forward conditions played the central link in a three-person chain-first receiving then splitting. The amount given to the future receiver was our measure of paying-it-forward.

Results
The pay-it-forward conditions differed in perceived fairness, The amount paid forward varied by condition, F(3, 96) = 9.09, p < .001. As seen in Figure 1, participants in the greedy condition gave the least, followed by those in the give-only Paying it Forward 7 condition, followed by the equality and generous conditions. All conditions differed from one another, all ts > 2.14, all ps < .05, except for the equality and the generous groups, t(48) = .61, p = .55. 2 Relative to the give-only condition, receiving greed led participants to be more selfish and receiving equal or generous outcomes led people to be more generous. However, receiving generosity did not prompt any greater generosity than receiving equal treatment, despite the fact that these treatments differed in terms of perceived fairness. We also compared the difference between the amount received and the amount given in each condition-the closer this value is to zero, the more closely people paid their treatment forward in kind. The one-way ANOVA was significant, F(2, 72) = 7.15, p = .001. Equality (M difference = .38) was paid forward more than greed (M difference = 1.36), p < .05, which was paid forward marginally more than generosity (M difference = 2.29), p = .07. Thus, behavior appears to be paid forward in anonymous interactions, but the extent depends on the type of behavior: equality propagates most, and greed propagates more than generosity.
Experiment 2: Manipulating Endowment Experiment 1 does not differentiate the impact of social phenomena-receiving a greedy or generous split from another person-from the effects of simply having a larger or smaller endowment. In Experiment 2, we randomly endowed people with either $0 or $6 before they received greedy or generous splits, and predicted that greed and generosity would influence paying-it-forward more than mere endowment.

Method
One hundred participants (50 female, M age = 24.3) were recruited as in Experiment 1.
This experiment followed a 2 (initial random endowment: lucky or unlucky) × 2 (social treatment: greedy or generous) design. Participants first picked three numbers then rolled a die.
If one of their lucky numbers came up, they were given $6 (lucky condition), if not, they received $0 (unlucky condition). Following this endowment, the procedure was identical to the greedy and generous conditions of Experiment 1.

Results
A 2 (initial random endowment: lucky or unlucky) × 2 (social treatment: greedy or generous) ANOVA revealed main effect of social treatment on paying it forward, F(3, 95) = 8.19, p < .01, but no main effect for initial random endowment, F(3, 95) = 1.42, p > .23, and no interaction, F(3, 95) = 0.46, p > .49. Those who received a greedy split paid forward less than Paying it Forward 9 those who received a generous split (Figure 2); those who had received $0 or $6 due to luck were less likely to pay this behavior forward, suggesting that social treatment plays a role in paying behavior forward over and above one's endowment.
When both social treatment and the initial random endowment were entered as simultaneous predictors for paying it forward in a regression, social treatment significantly predicted paying it forward behavior, β = .30, t(97) = 3.15, p < .005 , but not initial endowment, β = .15, t(97) = 1.50, p = .13. More descriptively, the fact that those who received a total of $12 (who were lucky and received generosity) paid forward only 21¢ more than those who received $6 (who were unlucky and received generosity) further suggests that endowment is not driving our effects. 3 As before, the relative difference between money received and money paid forward revealed that greed (M difference = 1.81) was paid forward more than generosity (M difference = 3.02), 3 Perhaps the most compelling argument against the role of endowment in paying it forward is that, in Experiment 1, equality was paid forward most of all despite having an endowment in between greed and generosity. In other words, if endowment was driving the effect, one would expect that generosity (total endowment: $12) should be paid forward most, followed by equality (total endowment: $9), followed by greed (total endowment: $6)-which is not the observed pattern of data. Figure 2: Money paid forward (out of a possible $6.00), after receiving both a random initial endowment and a social treatment of either greed or generosity (Experiment 2).

Experiment 3: Paying Forward Labor
Experiments 3, 4a and 4b explore paying it forward using a different currency: labor. In life, there are both enjoyable and onerous tasks, with the deeds of others leaving people with more or less of each; we explored whether people who received greedy and generous divisions of tasks paid forward that treatment to others. In addition, we test paying-it-forward in an inherently anonymous environment: an internet-based labor market.

Method
Participants (N = 60, 12 female; M age = 26.9 years) were recruited through Amazon's Mechanical Turk, an online labor recruitment tool (Buhrmester, Kwang, & Gosling, 2011); most importantly, the tool allowed us to recruit participants who would engage in truly anonymous, fleeting interactions. As in Experiments 1 and 2, participants were told that they were the middle link in an anonymous chain (in this case, a chain of labor): they would first play the role of the receiver, then that of the splitter. The experiment consisted of both good tasks (rating humorous stimuli) and bad tasks (circling vowels in dense foreign text); each split consisted of a total of four tasks (two good, two bad).
Participants first received either a greedy split (the previous participant had completed both good tasks, leaving only bad tasks for the current participant) or a generous split (the previous participant had completed both bad tasks, leaving only good ones). Participants then acted as the splitter, deciding how to split an additional four tasks between themselves and an anonymous future participant. They could pay forward either greed (leaving both bad tasks), Paying it Forward 11 equality (leaving one good and one bad task), or generosity (leaving both good tasks). Finally, participants completed their four tasks (two given from the previous participant, and the two not given to the future participant). The number of good tasks left for the future anonymous participant (0, 1, or 2) was our measure of paying it forward.

Results
The number of good tasks participants assigned to the future participant varied by condition, F(1, 58) = 5.69, p = .02. Participants in the greedy condition (M = 0.50, SD = .57) gave significantly fewer good tasks than participants in the generous condition (M = 0.86, SD = .59). We again assessed the extent to which behaviors were paid forward by comparing the difference between good tasks received and given. A one-way ANOVA revealed significant differences in paying it forward, F(1, 58) = 18.42, p < .001; as before, greed (M difference = 0.50) was paid forward more than generosity (M difference = 1.14).

Experiment 4a: Negative Affect and Paying it Forward
In Experiments 4a and 4b, we explore whether the asymmetry in paying forward greed versus generosity is driven by negative affect, and provide converging evidence by using both mediation (Experiment 4a) and moderation (Experiment 4b) approaches (Spencer, Zanna & Fong, 2005).

Method
Experiment 4a explicitly tested for potential mediation by both general positive affect and general negative affect, as well as a commonly studied discrete negative emotion: anger (Allred, Mallozzi, Matsui & Raia, 1997). In contrast to previous studies that investigated only good deeds (e.g., Bartlett & Desteno, 2006), we predicted that the inclusion of greed would lead negative affect to be the best predictor of overall paying it forward behavior; moreover, we Paying it Forward 12 predicted that generalized negative affect would be more powerful than the specific negative emotion of anger in predicting pay it forward behavior (Pillutla & Murnighan, 1996).
Ninety-six participants (63 female, M age = 30.1) were recruited as in Experiment 1. Similar to Experiment 3, participants completed a series of tasks: good tasks consisted of making free associations to words, and bad tasks consisted of circling the vowels in dense Italian prose ( Figure 3). Each participant was given eight tasks-four good and four bad-to split between themselves and a future participant; participants completed half of the eight tasks in any combination before passing on the remaining tasks by placing them in an envelope addressed to a "Future Participant." In the give-only condition, participants completed four of the eight tasks and passed on the remainder. In the three pay-it-forward conditions, participants first learned that a previous participant (the splitter) had been asked to split eight tasks and had left them either a generous (completing all four bad and leaving them all four good tasks), greedy (completing all four good and leaving them all four bad tasks), or equal division of labor (completing two tasks of each kind).
Paying it Forward 13 Figure 3: Examples of a "good task" (left) and a "bad task" (right) given to participants in Experiment 4.

Immediately after receiving this split, participants completed an abbreviated Positive and
Negative Affect Schedule (PANAS; Watson, Clark & Tellegen, 1988), indicating the extent to which they were feeling a variety of positive (happy, enthusiastic, excited, alert, inspired) and negative (distressed, upset, afraid, nervous) affective states on 5-point scales (1: not at all, 5: extremely). We also assessed feelings of anger on the same scale. Participants then rated the fairness of the split as in Experiment 1, completed the tasks left for them by the previous player, and then acted as a splitter by dividing an additional eight tasks (four good, four bad) between themselves and a future anonymous recipient. The division of tasks passed to the anonymous recipient (good v. bad) was our measure of paying it forward.

Paying it Forward
The number of good tasks participants assigned to the future participant varied by condition, F(3, 87) = 4.23, p < .01. Participants in the greedy condition gave significantly fewer good tasks than participants in all other conditions (all ts > 2.50, all ps < .03; Figure 3).
Participants in the give-only condition gave the next least, though not significantly less than those in the equality or generous conditions, all ts < 1.25, all ps > .22. We again compared the difference between amount received and amount given. The one-way ANOVA was significant, F(2, 69) = 24.45, p < .001. As in Experiment 1, equality (M difference = .04) was paid forward more Paying it Forward 14 than greed (M difference = 1.04), p = .001, which was paid forward more than generosity (M difference = 2.08), all ts > 3.45, p < .001.
Interestingly, when positive and negative affect and anger were regressed on paying it forward in the generous condition, only positive affect did predict paying forward, β = .41, p = .06 (negative affect: β = .07, p = .78; anger: β = .14, p = .51), consistent with previous research solely examining paying forward generosity (Bartlett & Desteno, 2006). Our results suggest, however, that when accounting for the full range of deeds-not just generosity, but also greed and equality-negative affect best predicts paying it forward.

Experiment 4b
If negative affect drives paying greed forward, then reducing negative affect should have the parallel effect of reducing the tendency to pay greed forward. Experiment 4b introduced a filler task between receiving and splitting, to test whether a positive mood intervention would moderate pay-it-forward behavior. Relying on previous research demonstrating that viewing cartoons repairs a negative mood (Göritz, 2006), we predicted that relative to a neutral task (viewing at abstract art), reading cartoons would mitigate paying greed forward. This experiment followed a 2 (receive: greedy, generous) × 2 (filler task: neutral, positive) design. We used the same "good" and "bad" tasks as in Experiment 4. Participants first received either a greedy split (the previous participant had completed both good tasks, leaving only bad tasks for the current participant) or a generous split (the previous participant had completed both bad tasks, leaving only good ones). Participants then rated their liking for either three pieces of abstract art (neutral) or three cartoons (positive), before next splitting an additional four tasks (two good, two bad) between themselves and an anonymous future participant. We predicted that a positive affect intervention would reduce participants' tendency to pay greed forward.

Results
Results from a 2 (receive: greedy, generous) × 2 (filler task: neutral, positive) ANOVA revealed a main effect of receiving greed or generosity, F(1, 161) = 3.74, p = .05, no effect of filler task, p > .14, and the predicted interaction effect, F(1, 161) = 7.00, p < .01, on the number of good tasks given. In the greedy condition, those who completed the neutral filler task (M = 0.37, SD = .54) paid greed forward more than those in who completed the positive filler task (M = 0.78, SD = .63) F(1,161) = 7.94, p < .01. There was no effect of filler task for those who received a generous split (M positive = 0.71; M neutral = 0.83); p = .39, suggesting that while reducing negative affect reduces paying greed forward, increasing positive affect fails to increase paying generosity forward.
Taken together, Experiments 4a and 4b provide evidence that the tendency to asymmetrically pay forward greed stems from negative affect. In Experiment 4a, negative affect mediated the link between tasks received and tasks paid forward; in Experiment 4b, a mood improving filler eliminated the paying forward of greed.

General Discussion
Five experiments demonstrate that people pay forward behavior in the sorts of fleeting, anonymous situations that increasingly typify people's day-to-day interactions. These data reveal that-in contrast to the focus of media, laypeople, and prior research-true generosity is paid forward less than both greed and equality. Equality leads to equality and greed leads to greed, but true generosity results only in a return to equality-an asymmetry driven by the greater power of negative affect. These results are both encouraging and dispiriting. When people receive equality, they rise above their baseline selfishness and pay forward this equality.
Furthermore, the stability of equality in pay-it-forward situations suggests that this basic level of pro-social behavior can propagate even in fully anonymous interactions. Deviations from equality, however, are paid forward only in the negative direction: true generosity prompts only equality, whereas greed prompts people to be greedier than they would be at baseline.
By examining paying it forward in the most parsimonious of environments with a novel paradigm, we show that generalized reciprocity can develop without the existence or appearance of interdependence-which in turn has implications for biological theories of cooperation. By examining paying it forward generosity, equality and greed of both money and labor we extend the study of paying behavior forward to everyday experiences. Finally, by documenting the asymmetry between greed and generosity and the underlying affective mechanism, we highlight a more sinister side of paying it forward-and hence human nature-that previous research and media attention concerning this phenomenon has largely ignored.

Paying it Forward 18
We note that our people are not often asked to fill envelopes with small amounts of money in their everyday lives, and so the meaningfulness of the task and the generalizability of our findings are important questions. The fact that participants responded consistently to even small amounts of money-both behaviorally and affectively-suggests that they did find the tasks meaningful and emotionally evocative. Moreover, Experiment 2 suggests that people respond less to monetary value per se and more to social aspect of greed and generosity.
Similarly, our experiments were small stakes, and future research should generalize the effects to large amounts of money and labor; we found that receiving generosity did not differ from receiving equal treatment in terms of behavior or affect, but such a difference may emerge for extremely generous acts 4 . Finally, while our studies all used similar minimal paradigms, they still capture paying it forward in the real world, especially as interactions increasingly take place via anonymous internet channels.
The research presented here provides the first systematic investigation into paying forward greed, generosity and equality. In so doing, it suggests a nuanced view of human nature.
Equality is paid forward in anonymous situations, suggesting that people will act fairly even when no-one is watching-if they have first been treated fairly. Although the recipients of generosity act positively-equitably-towards others, our results suggest that the magnitude of the positive behavior passed forward is less than what is received, whereas greed is paid forward in more equal measure. This asymmetry complements findings in both psychology and biology on the effects of socio-negative versus socio-positive outcomes (Rutte, Taborsky & Brinkhof, 2006), and provides new theoretical directions for evolutionary models of generalized reciprocity, which thus far have examined only primarily generous behaviors (Barta,McNamara,4 However, it is important to note that generous and equal acts are distinguishable by ratings of fairness: generous acts are rated as "more than fair" and equal acts as only "fair." See Experiment 1.

Paying it Forward 19
Huszár, & Taborsky, 2011). By looking at these three types of behavior, these data also extend and quantify past social psychological work on equity with the world (Austin & Walster, 1974) and the licensing effects of victimization (Zitek, Jordan, Monin & Leach, 2010), both of which suggest that people will act anti-socially if they have been treated less than fairly. The question is under what conditions cause people to be genuinely kind to others (Inbar, Zitek, Jordan, Fleuren & Breij, 2012)?
From the perspective of the person who is paying-it-forward, the asymmetry between greed and generosity may stem from a misconception of the threshold required for an act to truly reflect paying it forward. The person who awakes to gratefully find his long driveway cleared of snow may feel that he has "paid forward" the generous act by brushing off a bit of snow from a nearby car, but this discount rate is sufficiently high that the perpetuation of good will likely ends there. On the other hand, the person who awakes to find his driveway completely blocked from an errant snowplow may pile all that extra snow onto another car, thereby creating a significantly longer chain of ill will. This asymmetry suggests that to create chains of positive behavior, people should focus less on performing random acts of generosity, and more on treating others equally-while refraining from random acts of greed.