Abstract Darkness can conceal identity and encourage moral transgressions; it may also induce a psychological feeling of illusory anonymity that disinhibits dishonest and self-interested behavior regardless of actual anonymity. Three experiments provided empirical evidence supporting this prediction. In Experiment I, participants in a room with slightly dimmed lighting cheated more and thus earned more undeserved money than those in a well-lit room. In Experiment 2, participants wearing sunglasses behaved more selfishly than those wearing clear glasses. Finally, in Experiment 3, an illusory sense of anonymity mediated the relationship between darkness and self-interested behaviors. Across all three experiments, darkness had no bearing on actual anonymity, yet it still increased morally questionable behaviors. We suggest that the experience of darkness, even when subtle, may induce a sense of anonymity that is not proportionate to actual anonymity in a given situation. Keywords darkness, illusory anonymity, dishonesty, self-interested behavior Received 5/4/09; Revision accepted 8/21/09 In "Worship," Ralph Waldo Emerson (1860/1888) wrote, "As Departing from this body of work, we suggest that darkness gaslight is the best nocturnal police, so the universe protects does more than simply produce conditions of actual anonym- itself by pitiless publicity" (p. 214). This quote expresses an ity. We contend that darkness may create a sense of illusory inherent aspect of darkness: Darkness conceals identity and anonymity that disinhibits self-interested and unethical behav- decreases inhibitions. Indeed, criminal assaults are most fre- iors. Individuals in a room with slightly dimmed lighting or quent during hours of darkness (Hartley, 1974; Karnes, 1960), people who have donned a pair of sunglasses may feel anony- and dark rooms promote aggressive behavior (Page & Moss, mous not because the associated darkness significantly reduces 1976). This licensing effect of darkness might have contrib others' ability to see or identify them, but because they are uted to the popularization of streetlights in urban landscapes anchored on their own phenomenological experience of dark- during the 19th century (Bouman, 1987). ness. When individuals in such circumstances experience Darkness can disinhibit criminal acts and moral trans- darkness and, consequently, impaired vision, they generalize gressions by producing anonymity. Unethical acts are more that experience to others, expecting that others will conversely likely when transgressors cannot be identified. In The have difficulty perceiving or seeing them. Piaget (1936) Republic, Plato told the story of the ring of Gyges, which described this kind of egocentrism among young children, In granted its owner the power of invisibility and eventually one study (Piaget & Inhelder, 1956), children were presented led to the owner's corruption. Similarly, Zimbardo (1969) with a three-dimensional model of a scene with a small doll showed that participants dressed in concealing hoods and sitting opposite them and were asked to describe what the doll baggy clothing delivered longer electric shocks to strangers saw. Children between the ages of 4 and 7 tended to identify an than did participants wearing regular clothing. The same image that showed what they saw despite the difference in the effect has also been found for unrestrained, impulsive, and uncontrolled behavior when individuals experience ano- nymity or deindividuation through their association with a Corresponding Author: Chen-Bo Zhong, University of Toronto, OBHRM, 105 St. George St.. group (e.g., Festinger, Pepitone, & Newcomb, 1952; Singer, Toronto, Ontario, Canada MSS 3E6 Brush, & Lublin, 1965). E-mail: chenbo.zhong@rotman.utoronto.caZhong et al. scene that would be viewed from the doll's position. Even Participants were told that they would have 5 min to find though adults are better able to take others" perspectives, they two numbers in each matrix that added up to 10. For each pair never completely grow out of egocentrism (e.g., Epley, More- of numbers correctly identified, they would keep $0.50 from wedge, & Keysar, 2004; Tversky & Kahneman, 1974). For their supply of money; they were also asked to transfer the example, adults tend to overestimate the extent to which their remaining amount to the white envelope and drop it in a desig- thoughts, feelings, and sensations are accessible to others nated box along with the collection slip. Note that 5 min is not because they are anchored on their own experience, using it as enough time to solve all 20 matrices, In previous studies (Gino, a starting point to predict others' experiences (Gilovich, Sav- Ayal, & Ariely, 2009; Mazar et al., 2008), people were able to itsky, & Medvec, 1998). Thus, just as children playing hide- find 7 of the 20 pairs on average during this amount of time. In and-seek will close their eyes and believe that others cannot addition, there was no apparent identifying information any- see them, the experience of darkness may lead adults to feel where on the two sheets, so results seemed anonymous. Thus, they are hidden from others regardless of whether that is actu- participants had both an incentive and the opportunity to over- ally true. We propose that this illusory anonymity can conse- report their performance to earn more money. quently license unethical behaviors. All participants received the same matrices to solve, except Three experiments tested whether darkness can license dis- that a single number was unique for each participant. One of honesty and self-interested behaviors. In Experiment 1, we the three-digit numbers in the matrix used as an example on manipulated environmental dimness and examined whether par- the back of each collection slip matched the unique number on licipants cheated to earn undeserved money, In Experiment 2, we the corresponding test sheet. This allowed us to match the examined the extent to which people would act selfishly in a dic- worksheet with the collection slip of each participant and com- tator game while wearing sunglasses versus clear glasses. Finally, pute the difference between self-reported performance and Experiment 3 examined whether subjective perceptions of ano- actual performance. This difference score was our main depen- nymity mediated the licensing effect of wearing sunglasses. dent variable. Positive difference scores indicate that partici- pants overreported their performance and cheated on the task." Experiment I: Cheating in a Dim Room After the 5-min task, participants in both conditions wrote down on the collection slip the number of matrices they had A central aspect of our prediction is that darkness can license solved correctly. They dropped the collection slip and the self-interested and unethical behaviors regardless of actual remaining money in one box and the matrices sheet in another anonymity. In Experiment 1, we controlled for actual anonym- box located in a different corner of the room. ity by having participants engage in an individual task in A f test revealed that there were no significant differences which no identifying information was revealed and partici- in actual performance between the two conditions (M = 7.26 pants' choices could not be traced. We manipulated the light- correct, SD = 2.27, vs. M = 6.95 correct, SD= 2.49), 1(82) .99. Eighty-four college students at the University of North Participants in the control room reported a lower number of Carolina at Chapel Hill (40 female, 44 male; average age = correctly solved matrices (M = 7.78, SD = 3.09) than did par- 20.54 years) participated in the study for a maximum payment ticipants in the dim room (M = 11.47, SD = 4.32). This resulted of $12. Participants received a $2 fee for showing up and had in a difference of $1.85 in actual payout. We found that the opportunity to earn an extra $10. They were randomly cheating was greater in the dim room than in the control room assigned to one of two conditions (dim room vs. control room) both when we examined the average number of matrices by upon arrival. The well-lit, or control, room (15 ft x 14 fl) was which participants overstated their performance (M = 4.21, illuminated by 12 fluorescent lights mounted to the ceiling. SD = 4.12, vs. M = 0.83, SD = 1.58), ((82) = 4.92, p .99, and when we examined the percentage of partici- lights; the participants could see each other and the experi- pants who overstated their performance (M = 60.5%, SD = mental material, but this room was visibly dimmer than the 50%, vs. M - 24.4%, SD - 44%). x (1, N= 84) = 11.15, p = well-lit room. Participants in the dim-room condition were .001, Prep -.99. simply told that some of the lights were out. These results provide strong support for the predicted rela- For the task, participants received a brown envelope that tionship between darkness and cheating. Although early stud- contained $10 (nine $1 bills and four quarters) and an empty ies such as the one by Prentice-Dunn and Rogers (1980) white envelope, along with two sheets of paper. The first paper manipulated dimness and measured its effect on aggressive was a worksheet with 20 matrices, each consisting of 12 three- behaviors, in these studies room dimness was manipulated digit numbers (e.g., 4.78; Mazar, Amir, & Ariely, 2008). The along with other factors, such as white noise and confidential- second paper was a collection slip on which participants were ity of personal information; thus, the causal relationship to report their performance and answer demographic ques- between darkness and cheating has not previously been estab- tions. On the back of the collection slip we included instruc- lished. In our first experiment, the task was completely anony- tions for the task and a different matrix as an example. mous, and the only difference across conditions was roomdimness. We found that a slightly dim room increased cheating division, 1(23) = 0.78, p = 44, pre = .54. These results are above and beyond the effect of guaranteed anonymity. consistent with those of Experiment I and provide even stron- ger evidence that darkness can license dishonest and self- Experiment 2: "Shades" and interested behaviors through illusory anonymity: Wearing a Self-Interested Behavior pair of sunglasses should have no bearing on anonymity in an on-line task without face-to-face interaction. As we mentioned earlier, a useful metaphor for the illusory anonymity induced by darkness is the example of children playing hide-and-seek, who close their eyes and believe that Experiment 3: "Shades" and others cannot see them. In Experiment 2, we tested this idea by Perceived Anonymity having some participants wear a pair of sunglasses (and others In Experiment 3, we directly examined perceived anonymity as wear clear glasses") while engaging in an on-line task without a mediator of the licensing effect of darkness on self-interested expectation of face-to-face interaction. Clearly, the fact that behaviors. Experiment 3 employed the same design and proce- one is wearing a pair of sunglasses should not impair other dure as Experiment 2 except that we included a five-item mea- people's sight, especially when there is no face-to-face inter- sure of perceived anonymity (see Table 1) after the dictator action. Nevertheless, because darkness induces illusory ano- game. These items captured the extent to which participants nymity, we expected that participants wearing sunglasses felt anonymous and thought that others were not paying atten- would be more likely than those wearing clear glasses to tion to them and their choices during the dictator game behave selfishly in an anonymous dictator game. (a = .93). Participants indicated their agreement with each item Fifty students at the University of Toronto volunteered (31 on a 7-point Likert scale (1 = strongly disagree, 7 = strongly female, 19 male; average age = 21.36 years) for a maximum agree). payment of $1 1. Participants received a $5 fee for showing up Eighty-three students at the University of North Carolina at and had the opportunity to earn up to $6 during the study. The Chapel Hill participated (39 female, 44 male; average age = experiment had a one-factor (sunglasses vs. clear glasses) 20.71 years) for a maximum payment of $11 ($5 fee for show- between-subjects design. Participants were randomly assigned ing up and potential earnings up to $6). On average, partici- to "test" either a pair of sunglasses or a pair of clear glasses pants offered $2.35 (SD = $1.43). As expected, those who while completing some "unrelated" tasks. We purposely wore sunglasses gave significantly less (M = $1.93, SD = selected oversized glasses so that participants who came in $1.27) than did those who wore clear glasses (M = $2.76, SD= with their own glasses could wear the experimental glasses on $1.46), ((81) = -2.77, p .95. Also, participants in top of their own. the sunglasses condition gave significantly less than the fair The supposedly unrelated task involved an ostensibly inter- division (i.e., $3), /(40) = -5.40, p <.001 prep>.99, whereas personal interaction with a stranger in a different room. The the amount given by those in the clear-glasses condition was interaction was a typical one-shot dictator game that included not significantly different from the fair division, ((41) =-1.06, two roles, initiator and recipient. The initiator had $6 to allo- p= .30, Prep =.65. These results fully replicated the findings of cate between him- or herself and the recipient. Initiators kept Experiment 2. whatever they did not offer, recipients could choose to accept Further, participants who wore sunglasses reported feeling or reject the offer, but their choices did not affect initiators' more anonymous during the study (M = 4.73, SD = 1.10) than outcomes, Although participants were told they had been ran- did those who wore clear glasses (M = 4.01, SD= 1.17), ((81) = domly assigned to a role, they all played the initiator against 2.87, p .95. We examined whether this perceived the experimenter. We emphasized that participants would not anonymity mediated the effects of darkness on the amount see or talk to their counterparts during or after the experi- participants offered in the dictator game (Baron & Kenny, ment-all the interactions would be mediated by a computer 1986). The effect of wearing sunglasses was reduced to non- program. This ensured that the sunglasses did not affect actual significance (from B- -0.29, p <.01.p.> .95, to B = =0.09. anonymity or visibility of facial expressions. The experiment p= .28, Prep "-66) when perceived anonymity was included in ended after participants made their choice; they then answered a few demographic questions and were paid $5 plus the amount they kept for themselves in the dictator game. Participants offered between $0 and $6 (M = $2.24, SD - Table I. Items Used to Measure Perceived Anonymity and $1.62). As expected, those who wore sunglasses gave signifi- Concealed Identity in Experiment 3 cantly less (A = $1.81, SD = $1.30) than those who wore clear I. I was watched during the study.* glasses (A = $2.71, SD = $1.83), /(48) - 2.02, p= .049, P. = 2. I was anonymous during the study. 88. Also, participants in the sunglasses condition gave signifi- 3. My choice went unnoticed during the study. cantly less than the fair division (i.c., $3), /(25) = -4.688, p.95, whereas the amount given by those in the con- 5. Others were paying attention to my behavior during the study.* trol condition was not significantly different from the fair Note: Reverse-scored items are indicated by asterisks,the equation, and perceived anonymity was a significant pre- Notes dictor of the offered amount (B = -0.67, p <.001>.99). A bootstrap analysis showed that the 99% bias-corrected confi- 1. No participant underreported his or her performance. dence interval for the size of the indirect effect excluded zero 2. The glasses did not have prescription lenses. ([-0.77, -0.750), suggesting a significant indirect effect (Mackinnon, Fairchild, & Fritz, 2007). These results show References that perceived anonymity mediated the effect of darkness on Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator vari- selfish behavior. able distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality General Discussion and Social Psychology, 51, 1173-1182. Bouman, M.J. (1987). Luxury and control: The urbanity of street light- Imagine that a person who is alone in a closed room is decid- ing in nineteenth-century cities. Journal of Urban History, 14, 7-37. ing whether to lie to a total stranger in an e-mail. Clearly, Emerson, R. W. (1888). Worship. In E. W. Emerson (Ed.), The conduct whether the room is well lit would not affect the person's of life (pp. 191-230). Cambridge, MA: Riverside Press. (Original actual level of anonymity. Nevertheless, darkness may license work published 1860) unethical behavior in such situations. Across three studies, we Epley, N., Morewedge, C.K., & Keysar, B. (2004). Perspective taking found that darkness, induced by room dimness (Experiment 1) in children and adults: Equivalent egocentrism but differential cor- or wearing sunglasses (Experiments 2 and 3), licensed self- rection, Journal of Experimental Social Psychology, 40, 760-768. interested and cheating behavior. In addition, an illusory sense Festinger, L., Pepitone, A., & Newcomb, T. (1952). Some conse- of anonymity seems to have mediated this licensing effect of quences of de-individuation in a group. Journal of Abnormal and darkness (Experiment 3). Darkness appears to induce a false Social Psychology, 47, 382-389. sense of concealment, leading people to feel that their identi- Gilovich, T., Savitsky, K., & Medvec, V.H. (1998). The illusion of trans- ties are hidden. parency: Biased assessments of others' ability to read our emotional It is important to note that across all three experiments, our states. Journal of Personality and Social Psychology, 75, 332-346. darkness manipulations did not have any bearing on actual Gino, F., Ayal, S., & Ariely, D. (2009). Contagion and differentiation anonymity. In Experiment 1, we manipulated darkness by in unethical behavior: The effect of one bad apple on the barrel. dimming the lights. Although the room in the experimental Psychological Science, 20, 393-398. condition was darker than the one in the control condition, par- Hartley, JE. (1974). Lighting reinforces crime fight. Piusfield, MA: ticipants had no trouble seeing and identifying each other, In Buttenheim, Experiments 2 and 3, we manipulated darkness simply by ask- Karnes, E.B. (1960, April). Well planned lighting is city progress. ing participants to wear a pair of sunglasses. The task used in American City Magazine, 75, 104-105, these latter two experiments was fully mediated by computers, Mackinnon, D.P., Fairchild, A.J., & Fritz, MS. (2007), Mediation and participants did not expect to see or talk to the recipient of analysis, Annual Review of Psychology, 58, 593-614. their offer during or after the experiment. Further, the task was Mazar, N., Amir, O., & Ariely, D. (2008). The dishonesty of honest designed so that it promised complete anonymity. Neverthe- people: A theory of self-concept maintenance, Journal of Market- less, in each of these studies, darkness increased dishonesty ing Research, 45, 633-644. and self-interested behaviors. Page, R.A., & Moss, M.K. (1976), Environmental influences on Previous studies have treated darkness as just one of many aggression: The effects of darkness and proximity of victim. factors that induce a state of deindividuation (e.g., Zimbardo, Journal of Applied Social Psychology, 6, 126-133. 1969), but out studies suggest that the experience of darkness, Piaget, J. (1936). The origins of intelligence in children. New York; combined with the difficulty of transcending one's own phe- W.W. Norton & Co. nomenological experience, triggers a fundamental psychologi- Piaget, J., & Inhelder, B. (1956). The child's conception of space. cal belief that one is protected from others' attention and London: Routledge. inspections. Our results suggest that darkness, even experi- Prentice-Dunn, S., & Rogers, R.W. (1980). Effects of deindividuating enced one-sidedly through the act of wearing sunglasses, can situational cues and aggressive models on subjective deindividu- have potentially harmful consequences. Thus, Emerson may ation and aggression. Journal of Personality and Social Psychol have been correct when he stated that good lamps are the best ogy, 39, 104-113, police. Singer, J.E, Brush, C.E., & Lublin, S.C. (1965). Some aspects of deindividuation: Identification and conformity. Journal of Exper- Acknowledgments imental Social Psychology, 1, 356-378. The authors thank the Center for Decision Research at the University Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: of North Carolina at Chapel Hill for its support. Heuristics and biases. Science, 185, 1124-1131. Zimbardo, P. (1969). The human choice: Individuation, reason, and Declaration of Conflicting Interests order vs. deindividuation, impulse, and chaos. In W.J. Arnold & The authors declared that they had no conflicts of interests with D. Levine (Eds.), Nebraska Symposium on Motivation (Vol. 17. respect to their authorship andfor the publication of this article. pp. 237-307). Lincoln: University of Nebraska Press