Perceptions of the Self and most People’s Reactions towards Innocent and Noninnocent Victims

Abstract Research has shown that: individuals positively distinguish themselves from most other people; being consistent is positively valued; injunctive and descriptive norms are perceived to protect victims. Joining these findings, we argue that individuals present themselves as following injunctive and descriptive norms towards victims to a higher extent and more consistently than most people. In an experimental study 273 university students of both sexes indicated what they and most other people would approve of (injunctive norm) or typically do (descriptive norm) regarding various reactions towards either an innocent or a noninnocent victim. The reactions involved secondary victimization (devaluation/derogation, avoidance, suffering minimization, blaming the victim) and non secondary victimization (valuation, contact, suffering acknowledgment, not blaming the victim). Participants perceived themselves and most people as approving of more non secondary than secondary victimization reactions, except for blaming the noninnocent victim. Participants indicated they approved of most of the normative reactions to a higher extent than most people, which is interpreted as a new instance of the Primus Inter Pares effect. Participants also indicated they would show more consistency between their injunctive and descriptive norms, especially towards the innocent victim. Results suggest that individuals perceive themselves as more immune to perverse norms than most people.

A wide body of research and theorizing has repeatedly found and predicted various counterintuitive reactions towards innocent victims, especially towards those whose suffering cannot be relieved (Correia, Vala, & Aguiar, 2001;Correia & Vala, 2003;Hafer, 2000;Hafer & Bègue, 2005;Lerner, 1980). Specifically, this literature has ascertained that individuals unconsciously blame and derogate these victims in emotionally-laden situations. These surprising reactions play an important role in the lives of individuals who witness undeserved suffering: they protect their "fundamental delusion" that the world is just (Lerner, 1980). Without questioning these findings, our approach aims to complement them by focusing on the normative processes guiding reactions to victims, an issue which the literature has generally neglected (but see Lerner, 2003; see also van den Bos & Maas, 2009, for an empirical approach). Specifically, our main goal was to investigate individuals' perceptions of what they and most people approve of and how they typically react towards victims presented as either contributing or not contributing to their deep and permanent suffering. In other words, we aimed to investigate individuals' perceptions of personal and social injunctive and descriptive norms, respectively (Cialdini, Kallgren, & Reno, 1991;Cialdini & Trost, 1998) of reactions towards innocent and noninnocent victims whose suffering cannot be relieved.
The relevance of our research was twofold: theoretical and practical. Theoretically it broadens our understanding of the mechanisms involved in such reactions. In fact, by focusing on the normative aspects of reactions towards victims, it is possible to ascertain that individuals may use their discourses on what they and most people approve of and do towards victims as a device to positively distinguish themselves from other people. Specifically, they are expected to display the Primus Inter Pares effect (Codol, 1975), that is, to self-present as more normative than most people. This self-others distinction would be accomplished in two ways. First, participants could indicate that they approve of and engage in the normative reactions to a higher extent than most people. Second, participants could present themselves as more consistent than most people regarding what they approve of and typically do. For instance, given that consistency is socially approved of (Channouf & Mangard, 1997), participants were expected to indicate that they, but not most of the people, would typically contact victims to the same extent as they approve of it. As a result, individuals fall prey of a kind of a normative delusion which is likely to have practical consequences. Specifically, it may prevent individuals from acknowledging that they are likely to display the unsupportive reactions towards victims (secondary victimization, Brickman et al., 1982) as much as other people do and thus fail to correct these reactions. Although norms seemingly protect victims (Alves & Correia, 2009), the biased perception that one follows them may be a second layer of causes adding to emotional arousal that lead to unsupportive reactions. Hence, at a practical level, our research was relevant because the knowledge derived from it can be used in campaigns to make people aware of this normative delusion. In so doing, we could promote actual justice towards those people who especially need it. Lerner (1980; see also Hafer & Bègue, 2005) created what possibly is one of the most complete taxonomies of reactions towards victims. Lerner divided these reactions into two main categories: rational and nonrational. Through individuals' rational reactions, not only acknowledge that victims exist, but also engage in behaviors which are likely to objectively improve the victim situation (e.g., helping behavior). On the contrary, through nonrational reactions individuals may fail to acknowledge that injustices exist, for instance undeserved suffering. Among these reactions, Lerner (1980) identified those generally referred to as secondary victimization (Brickman et al., 1982), that is, unsupportive reactions which reflect negatively on the victim situation.

The various reactions towards victims
This failure to acknowledge injustices can be accomplished in various ways which may coexist (Correia et al., 2001). First, individuals may engage in denialwithdrawal. In the case of denial, individuals may minimize the victim suffering (Sonne & Pope, 1991), sometimes to the extent of denying it (Cohen, 2001). In the case of withdrawal, individuals may avoid physical and/or psychological contact with those who are suffering undeservedly (Hafer, 2000). Second, individuals may reinterpret the character of the victim and perceive him/her as the kind of person who deserves to suffer (victim devaluation/derogation). Furthermore, individuals may even reinterpret the objective cause(s) of victim suffering and/or blame/ hold the innocent victim responsible for his/her suffering (for other strategies, see Hafer & Bègue, 2005;Lerner, 1980). The literature has collected much empirical evidence that secondary victimization is a nonconscious, possibly automatic, and common phenomenon in emotionallyladen situations (Lerner, 2003). These reactions are common when individuals perceive that they cannot do much to relieve or end the innocent victim suffering (for reviews, see Hafer & Bègue, 2005;Lerner, 1980; for empirical evidence see Correia & Vala, 2003;Mohiyeddini & Montada, 1998). For instance, Robbenholt (2000) found that the assignment of more responsibility/ blame to victims of more serious outcomes (not only suffering) is a robust finding. However, according to Lerner (2003) when individuals are able to engage in controlled, conscious and relatively emotionallyneutral information processing, these secondary victimization reactions are not only relatively rare but also disapproved of. In other words, when individuals are able to engage in more thoughtful processing, secondary victimization reactions are counternormative, especially when innocent victims are involved. In these situations non secondary victimization reactions are more likely and approved of (but see van den Bos & Maas, 2009).

Social norms and the (counter-) normativity of reactions towards victims
Social norms may refer to what is prescribed and approved of in a given collective (prescriptive/injunctive norms) or to what members of the collective actually do (descriptive norms) (Cialdini & Trost, 1998;Cialdini et al., 1991; see also the akin distinction between desirable and factual norms, Codol, 1975). Research has used the concept of social norms in such diverse domains as pro-environmental behaviors (Félloneau & Becker, 2008;Göckeritz et al., 2010), conformity (Jacobson & Mortensen, 2011), judgments of (non) deviant in-group or out-group members (Pinto, Marques, Levine, & Abrams, 2010), the social value of certain discourses, such as internality, self-sufficiency and individual anchoring (e.g., Dubois & Beauvois, 2005) or the belief in a just world (e.g., Alves & Correia, 2008. Research has found that social norms can powerfully influence people's behaviors (Shultz, Nolan, Cialdini, Goldstein, & Griskevicius, 2007), even though individuals may fail to notice such influence (Nolan, Shultz, Cialdini, Goldstein, & Griskevicius, 2008). To our knowledge, however, only Alves and Correia (2009) have directly applied social norms, specifically the conceptual distinction between injunctive and descriptive social norms, to the study of various secondary victimization and non secondary victimization reactions towards innocent and noninnocent victims (but see Simons & Piliavin, 1972;Snyder, Kleck, Strenta, & Mentzer, 1980, for indirect evidence on derogation and avoidance, respectively).
In their study, Alves and Correia (2009) first measured participants' general belief in a just world (Dalbert, Montada, & Schmitt, 1987), that is, the degree to which Perceptions of the Self 3 they thought that people get what they deserve (Lerner, 1980). Then participants read about a situation in which a male or female innocent (someone crashed into his/her car) or noninnocent victim (he/she crashed into someone's car) suffered severely and permanently (e.g., being disfigured and in pain for life). Participants were then asked about four secondary victimization reactions (devaluation, avoidance, suffering minimization and blame) and their opposites or non secondary victimization reactions (valuation, contact, suffering acknowledgment and not blaming). Specifically, they were asked to indicate what they thought most people would approve of (perception of injunctive normativity) or would do (perception of descriptive normativity).
Results indicated that most people are perceived to approve of and to engage in valuation, contact and suffering acknowledgment, but not in devaluation, avoidance and suffering minimization towards both the innocent and the noninnocent victim. Victim innocence and the type of norm (but not participants' reported general belief in a just world nor the sex of the victim), however, moderated this general pattern. Regarding the effect of victim innocence, participants indicated that most people considered valuation and suffering acknowledgement towards the innocent victim as more normative than towards the noninnocent victim. The effect of victim innocence was even stronger on blame ratings. Specifically, results indicated that most people would approve of blaming the noninnocent victim and not blaming the innocent victim. As far as the type of norm is concerned, participants indicated that valuation and contact were more injunctively than descriptively normative. In other words, according to participants, to most people -words are stronger than actions-the extent to which they are perceived to approve of valuation and contact is higher than the extent to which they are perceived to typically value and contact victims. Alves and Correia's (2009) study, however, had two major limitations which we address here. First, participants read that the victim had bumped his/her car into another. Although they also read that the other driver had suffered only minor injury and that he/she had already fully recovered, this operationalization of the noninnocent victim left room for participants to perceive him/her not only as a victim but also as a harm-doer. This fact may have been the cause that led participants indicate that most people would approve of blaming him/her or to indicate that most people would approve of contacting and valuating him/her less (as a sort of punishment) than the innocent victim. Thus, in the current study there is no third-party involved. Second, Alves and Correia only asked about the participants' perception regarding most people's norms. As a consequence, we do not yet know the extent to which these reactions are also perceived, or at least performed as personally normative. Thus, in this study we also asked participants to indicate the extent to which they approved of and thought these reactions would be typical of them.
By asking participants about their personal stand it is possible to compare how they present their and most people's norms for reactions towards victims. We expected that participants took the opportunity to engage in normative self-other distinction and in so doing they will show the Primus Inter Pares effect (Codol, 1975; see also Hornsey & Jetten, 2004). According to the Primus Inter Pares effect, individuals tend to see themselves as more normative than most people on relevant comparative dimensions. The qualification "most other people" is important to note. In fact, although "primus inter pares" literally means "first among equals", Codol (1975) cautioned that this tendency does not necessarily imply presenting oneself as being the first of all others (i.e., the most normative). Furthermore, judging oneself as more normative than most others does not imply that the latter are judged as counternormative individuals; only that they are (slightly) less normative than the self.
In our study, participants' ratings reflected the Primus Inter Pares effect in two ways. First, they indicated that they approve of and typically put into practice the normative reactions (e.g., valuation) to a higher extent than most people. Second, participants positively distinguished themselves from most people by presenting themselves as more consistent, which was positively valued. In fact, even though it was common for individuals to be inconsistent (Critcher, Huber, Ho, & Koleva, 2009), displaying consistency was prescriptively normative at least in Western societies (Channouf & Mangard, 1997). Thus, participants indicated that their typical reactions towards victims (i.e., their indicated descriptive normativity) reflect to a higher extent than most people the reactions that they approve of (i.e., their indicated injunctive normativity). In other words, there was higher consistency between the participants' stated norms than between those attributed to most people.

Goals and hypotheses
This study was centered on individuals' perceptions of their and most people's approved of (perception of injunctive normativity) and typical reactions (perception of descriptive normativity) towards innocent and noninnocent victims who suffer deeply and permanently. We aimed at: (a) showing that participants will present themselves as prescriptively and descriptively normative individuals; (b) showing that individuals use the proposed reactions in order to positively distinguish themselves from others by stating: (b1) their higher injunctive and descriptive normativity regarding various reactions towards innocent and noninnocent victims when compared to most people, and (b2) their higher consistency between these two types of norms also when compared to most people; (c) replicating Alves and Correia's (2009) results of assessments of most people's normativity with a new operationalization of victim innocence.
Based on Alves and Correia (2009), we expected that: (a) participants would indicate that both the self and most people would approve of and typically engage in valuation, contact and suffering acknowledgement towards both the innocent victim and the noninnocent victim; (b) the effect of victim innocence would be especially strong in ratings of (not) blaming the victim. Specifically, participants would indicate that both the self and most people would approve of and typically blame a noninnocent victim and not blame an innocent victim; (c) reflecting the Primus Inter Pares effect, participants would positively distinguish themselves from most people in two ways: (c.1) by indicating that they would both approve of and typically engage in the normative reactions to a higher extent than most people, and (c.2) by presenting themselves as more consistent than most people. Specifically, ratings of descriptive norms (i.e., how typical) involving the self would not differ from ratings of their injunctive norms (i.e., how approved of); on the contrary, ratings of most people's descriptive norms would be lower than ratings of their injunctive norms.

Participants
Our convenience sample comprises 273 undergraduates of both sexes (136 males and 133 females, four unreported) studying in Lisbon (Universidade Lusófona and Faculdade de Letras de Lisboa) and Setúbal (Instituto Politécnico de Setúbal). The participants were taking various degrees, such as geography, marketing or human resources. Their ages varied between 18 and 57 (M = 23.86, SD = 6.18). We had access to these participants through acquaintances who taught them one subject.

Procedure
The 30-minute sessions occurred during class time. The teacher present in the classes introduced the experimenter (the first author), explained his presence there and indicated that participation was fully voluntary. Then the experimenter stressed the voluntary aspect of their participation and indicated that he would perfectly understand if they refused to take part (no one refused to take part in the study). Afterwards, the experimenter delivered a sheet of paper with an overview of the study which was introduced as an impression formation and behavioral judgment study. Participants read a text presented as the summary of a case taken from a study about road accidents. The word "victim" was never used in the study and he/she was referred to as X both in the text and in the measures. The text manipulated the victim innocence by presenting him/her as either responsible (noninnocent) or not responsible (innocent) for the onset of his/her situation. The text also described his/her suffering which was always presented as very severe and permanent. Next, participants responded to the dependent measures. Finally, they were probed for suspicion, thanked and debriefed.

Independent Variables
This experimental study consisted of a 2 (innocence of the victim: innocent/ noninnocent) X 2 (type of norm: injunctive/descriptive) X 2 (referent: the self/ most people) mixed-subjects design, with the two former factors between-subjects and the latter factor withinsubjects. Suffering was a controlled variable: the victim suffered severely (e.g., 85% of the body burnt, disfigurement, no eyelids, need of constant eye lubrication) and permanently (need of permanent medical care, with life expectation being equivalent to the population average, according to doctors). The various versions of our instruments were randomly distributed among participants.
The victim innocence was manipulated in the text. In the innocent condition participants read that the victim had been driving carefully and at low speed (no more than 50 Km/h -about 30 miles an hour) and that oil on the road caused the victim to lose control of the car and to fall from a cliff. In the noninnocent condition, the victim was described as having driven drunk, recklessly and at high speed (between 150-170 Km/h -about 95-105 miles an hour) resulting in falling from a cliff after having swerved too much to his/her right. The various speeds were presented as estimations by experts and blood alcohol levels were said to be obtained through blood analyses. These operationalizations were pretested among 15 participants who read one of the two versions of the text (noninnocent or innocent victim) and answered to two questions on 7-point Likert-type scales from 1 (not at all) to 7 ( very much): "To what extent does X suffer?"; "To what extent is X responsible for his/her situation?" The effect of the victim innocence on the perceived seriousness of suffering was nonsignificant, M = 6.67, SD = 0.94, but pretesters clearly distinguished between the responsibility of the innocent, M = 1.38, SD = 1.15, from Perceptions of the Self 5 perceived approval or typicality for each pair of opposing reactions, according to the participants' view and the one they attributed to most people. Since the participants rated their judgments on 7-point scales (from 1 to 7), the resulting difference could range between -6 (total agreement that a secondary victimization reaction is approved of/typical) and +6 (total agreement that a non secondary victimization reaction is approved of/typical). A rating equivalent to zero was interpreted as indicating that both the secondary victimization and the non-secondary victimization reactions were equally approved of or typical.

Results
We conducted a 2 (victim innocence) X 2 (type of norm) X 2 (referent) X 4 (reactions) mixed-subjects ANOVA, with the two former factors between-subjects and the latter two factors within-subjects. Due to the violation of the sphericity assumption we used the Huynh-Feldt estimates to correct for degrees of freedom (ε = .93, for the interaction effect between reactions and type of norm). There were several significant main effects, two-way and three-way interactions 2 . Since the fourway interaction was significant, F(2.85, 766.38) = 3.44, p = .02, η p 2 = .013, and it qualifies the previous effects, we will focus on it. One-tailed values are presented. Mean and standard deviation values are reported on Table 1.
Hypothesis a predicted that participants would indicate that both they and most people would approve of and typically engage in valuation, contact and suffering acknowledgement towards the innocent and the noninnocent victim. As far as the participants themselves were concerned, results were fully in line with our expectations. In fact, they indicated that they themselves would approve of valuation, contact and suffering acknowledgment towards both the innocent victim, Ms = 3.13, 3.26, 2.37, and the noninnocent victim, Ms = 2.69, 3.25, 2.44. Also as expected, they indicated that they would typically engage in valuation, contact and suffering acknowledgment towards both the innocent victim Ms = 2.76, 2.76, 2.15, and the noninnocent victim, Ms = 1.86, 2.16, 1.10.
As far as most people are concerned, our predictions received full support for their approved of reactions but only partial support regarding their typical reactions. In fact, participants indicated, as expected, that most people would approve of valuation, contact and suffering towards both the innocent victim, Ms = 2.36, that of the noninnocent victim, M = 6.43, SD = 0.85, t(13) = -18.58, p < .001.
The type of norm and the referent were manipulated through the instruction "Taking into account X's situation, indicate for each of the following sentences, the extent to which you/ most people approve of [injunctive norm]; it is typical of you/ most people" [descriptive norm]. Each participant responded to one type of norm from the point of view of two referents (the self and most people). The order of the referents was counterbalanced across participants.

Dependent measures
The participants were asked about both their and most people's perceived approval (injunctive normativity) or about their and most people's perceived typicality (descriptive normativity) regarding eight reactions. Each reaction was operationalized through one sentence. The sentences which operationalized the reactions were the same as in Alves and Correia (2009) and can be found in the Appendix. The Portuguese sentences were equivalent in length for the various reactions (23-25 words long) and were pretested for their perceived opposition and meanings (e.g., valuation vs. devaluation; see Alves & Correia, 2009). The order of the sentences was randomly established with the criterion that one sentence could not be immediately followed by its opposite. The sentences comprised four secondary victimization reactions (devaluation, avoidance, suffering minimization and blaming the victim) and four non secondary victimization reactions (valuation, contact, suffering acknowledgement and not blaming the victim). Participants rated the reactions on seven-point Likert type scales: approval, from 1 (I/most other people do not approve of at all) to 7 (I/most other people approve of very much); typicality, from 1(not typical of me at all / most other people) to 7 (very much typical of me/ most other people).
We calculated eight indices comprising the difference of the ratings for each "non secondary victimizationsecondary victimization" pair (e.g., "contact" minus "avoidance") 1 . These indices were our measure of 1 In order to avoid awkward expressions, such as "contact-avoidance", from now onwards we will only refer to the reaction reflected in the mean values. As we will see next, because most values go in the direction of non secondary victimization, we will refer to these indices mostly as "valuation", "contact" and "suffering acknowledgement", and will use the verbs "value", "contact" and "acknowledge", respectively. Nevertheless, because results indicate that the noninnocent victim should be and is blamed and the innocent victim should be and is not blamed, we will have to use the expressions blam(e)ing/not blam(e)ing. To note that given our operationalization of the latter dependent variables, a more precise label would have been hold(ing)/not hold(ing) the victim responsible (see Shaver & Drown, 1986). Again for the sake of simplicity we opted for blam(e)ing/not blam(e)ing.

H. Alves and I. Correia
1.90, 1.09, and the noninnocent victim, M s = 2.19, 1.75, 1.14. Also as expected, participants indicated that most people would typically engage in valuation of the innocent victim, M = 0.87, and suffering acknowledgement of both the innocent, M = 0.92, and the noninnocent victims M = 0.79. All these values were significantly different from zero, ts ≥ 2.89, ps ≤ .005, but none reached its highest possible value, that is, +6 or -6, ts ≥ 9.25, ps < .001. Contrarily to expected, however, participants indicated that most people would typically engage in a blend of secondary victimization and non secondary victimization in some cases. Specifically, they indicated that most people would engage in contact-avoidance of the innocent, M = .011, and the noninnocent victim, M = .0.26, and in valuation-derogation of the noninnocent victim, M = 0.40. The latter values were the only ones in this study to be statistically equivalent to zero, ts ≤ 1.30, ps ≥ .20.
Hypothesis b concerned the remaining pair of reactions, that is, (not) blaming the victim for his/her suffering. Specifically regarding the innocent victim, it predicted that participants would indicate that they themselves and most people would approve of and typically not blame him/her. On the contrary, as concerns the noninnocent victim, it predicted that both they and most people would approve of and typically blame him/her for his/her suffering. Consistent with this hypothesis, participants indicated that both they and most people would approve of not blaming the innocent victim, M self = 3.24; M most people = 2.74, and blaming the noninnocent victim for his/her suffering, M self = -3.59; M most people = -2.09. Similarly, participants also indicated that both they and most people would typically not blame the innocent victim for his/her suffering, M self = 2.44; M most people = 2.41, but would blame the noninnocent victim, M self = -2.24; M most people = -1.64.
According to Hypothesis c our results would reflect the Primus Inter Pares effect in two ways. First, participants would indicate that they approved of (injunctive norms) and would typically engage (descriptive norms) in the reactions to a higher extent than most people (Hypothesis c.1); second, they would indicate higher consistency between their injunctive and descriptive norms than between those of most people (Hypothesis c.2).

Note:
Scores could vary between -6 (complete agreement with devaluation, avoidance, suffering minimization and blaming) and + 6 (complete agreement with valuation, contact, suffering acknowledgement and not blaming).
Regarding ratings of engaged in reactions, Hypothesis c.1 was also supported for its most part. In fact, participants indicated that that they would engage in the majority of the normative reactions to a higher extent than most people. This pattern was obtained towards the innocent victim (comparisons of cells 5 vs. 6 of Table 1): more valuation, F(1, 269) = 47.18, p < .001, η p 2 = .15; more contact, F(1, 269) = 55.09, p < .001, η p 2 = .17; and more suffering acknowledgement, F(1, 269) = 17.59, p < .001, η p 2 = .06. Contrarily to expected, however, as concerns not blaming this victim, participants did not make such a positive distinction, F(1, 269) = 0.01, p = .46.
Finally, as far as Hypothesis c.2 is concerned, which predicted higher consistency between the participants' injunctive and descriptive norms than between those of most people, it was partially supported. As concerns participants' own norms towards the innocent victim (cells 1 vs. 5 of Table 1), there were no significant differences between what they indicated to approve of and to typically engage in the cases of valuation, contact, and suffering acknowledgement, all Fs ≤ 1.49, ps ≥ .11. Regarding the noninnocent victim, however, participants did not show such consistency (cells 3 vs. 7 of Table 1). In fact, they indicated to approve of more than to typically engage in valuation, F(1, 269) = 3.74, p = .027, η p 2 = .01; contact, F(1, 269) = 5.70, p = .01, η p 2 = .02; and suffering acknowledgement, F(1, 269) = 8.66, p = .001, η p 2 = .03. This unexpected discrepancy was also present in the (not) blaming the victim. In other words, participants indicated that they would approve of more than actually not blame the innocent victim, F(1, 269) = 12.01, p < .001, η p 2 = .016, and blame the noninnocent victim, F(1, 269) = 5.17, p = .005, η p 2 = .04.
Given that the patterns obtained for (not) blaming the victims were completely unsupported (i.e., consistency for most people and inconsistency for the self), we compared the perception of the participants' own descriptive norms with that of most people's injunctive norms. Duncan post-hoc tests showed that the differences were nonsignificant, ps = .50 and .71, which indicates that what participants say they themselves do is equivalent to what they say that most people approve of.

Discussion
With this study we intended to investigate perceptions of injunctive and descriptive norms of reactions towards innocent and noninnocent victims who suffer deeply and permanently. Specifically, our main goals were to show that individuals perceive that social norms governing most people's reactions (Alves & Correia, 2009) also apply to individuals themselves. Furthermore, we also aimed to show that individuals engage in self-others distinction by stating their superior conformity to and their higher consistency between their injunctive and descriptive norms when they refer to reactions towards a serious victimization case. As expected, most ratings were in the non secondary victimization direction, except for the case of blaming the noninnocent victim, which replicates Alves and Correia (2009). More importantly, our study not only extends these findings to self ratings but it also shows that individuals indicate that these norms apply more to them than to most people.
Our results indicate that participants distinguished themselves normatively from most people (Codol, 1975) in various ways. First, they indicate that, compared to most people, they would approve of valuation and contact with both the innocent and the noninnocent victim, and suffering acknowledgement of an innocent victim to a higher extent than most people (i.e., presenting themselves as more injunctively normative, 8 H. Alves and I. Correia Cialdini et al., 1991). Second, participants indicate that they would typically engage in these reactions to a higher extent than most people (i.e., presenting themselves as more descriptively normative, Cialdini et al., 1991). This positive distinction of the self is even more clear regarding descriptive than injunctive normativity. In fact, as far as injunctive norms are concerned, participants acknowledge that most people are clearly normative, albeit to a lesser extent than them, which is a typical Primus Inter Pares effect pattern (Codol, 1975). As regards descriptive normativity participants indicate, however, that most people engage in a blend of contact with and avoidance of both the innocent and the noninnocent victims, and a blend of valuation and devaluation of the noninnocent victim. Victims who suffer deeply and permanently, especially if they are innocent, tend to be secondarily victimized when help or support is not possible or it is too costly (Correia et al., 2001). Thus, the pattern obtained concerning most people's typical reactions is somewhat ironically closer to reality than that concerning the self. Third, individuals distinguish themselves normatively from most people by presenting themselves as more consistent than most people regarding approved and typical contact and valuation towards the innocent victim. This pattern, however, is not present in the case of the noninnocent victim. In fact, individuals indicate that both they and most people would engage in valuation and contact with the victim to a lesser extent than they approved of these reactions. Thus, it seems that it is more important to show consistency towards an innocent victim than towards a noninnocent victim, at least as far as contact and valuation are concerned. This pattern seems to reflect the fact that innocent victims are normatively more valued than noninnocent ones (Lerner, 2003).
We would like to stress that in the case of suffering acknowledgement, however, participants did not distinguish themselves from most people through higher consistency. In fact, in the case of the innocent victim, participants indicated that they and most people would descriptively follow the injunctive norm. Surprisingly, in the case of the noninnocent victim they indicated that their actual suffering acknowledgement would not be as strong as the extent to which they approved of it. This pattern again suggests that innocent victims are normatively more important than noninnocent ones. Nevertheless, participants indicated that most people would show consistency between their injunctive and descriptive norms regarding suffering acknowledgement. Although it seems that participants are stating that most people are more normative (i.e., consistent) than them, a glance at the mean values shows that according to participants it is "easier" for most people to be consistent in this case. In fact, mean values of injunctive normativity for most people are already relatively low, in fact as low as participants' ratings of their own descriptive normativity.
As expected, the victim innocence has the strongest effect on ratings of (not) blaming the victim, as was the case in Alves and Correia (2009). On the one hand, ratings of blaming the noninnocent victim for his/her situation are the only ones to be clearly of the secondary victimization type both for injunctive and descriptive normativity ratings. On the other hand, ratings involving the innocent victim are clearly of the non secondary victimization type (i.e., not blaming) also for both kinds of norms. As expected, individuals indicate that they approve of blaming the noninnocent victim and not blaming the innocent victim to a higher extent than most people. Also as expected, they also indicate that they would actually blame the noninnocent victim to a higher extent than most people. Unexpectedly, however, no difference was obtained regarding not blaming the innocent victim. The latter result seems to indicate that it is more important to find responsibility than innocence in others. Another unexpected pattern worth mentioning is that in this case participants indicated less consistency between their injunctive and descriptive norms than between most people's for both the innocent and the noninnocent victims. Yet, it seems that individuals opted to distinguish themselves from most people with another strategy because their descriptive norms are as strong as most people's injunctive norms. Thus, participants indicate that what they actually do is equivalent to what most people approve of, which contrasts with what they indicate regarding valuation and contact.
Possibly the different patterns of self-others distinction between suffering acknowledgement and (not) blaming the victim on the one hand, and valuation and contact on the other hand, may be explained by the fact that regarding the latter reactions participants perceive themselves, but not most people, as relatively immune to perverse norms. Perverse norms exist in situations in which individuals are expected to follow injunctive norms although they are impossible or very difficult to be followed due to the characteristics of the situations themselves (Fernández-Dols, 1992;Oceja & Fernandez-Dols, 2001). Although the theorizing of perverse norms has focused on formal norms, such as laws, it seems a fruitful approach to extend it to informal norms. In our case, it is likely that emotions are the mechanism that makes the fulfillment of approved contact and valuation difficult. Specifically, the perceived decrease in most people's descriptive normativity may be explained by the higher emotional involvement required than in the case of the remaining reactions (the victim was disabled and disfigured). If our reasoning is true, it Perceptions of the Self 9 seems that participants actually believe that they, but not most people, can overcome the emotional distress resulting from being with such a victim. They may also actually believe that they would typically engage in contact with such a victim to the same extent as they approve of these reactions. We put forward this is an illusion because it is not possible for most individuals to be more consistent than most other individuals. We also think that this may be a perverse mechanism because if individuals are illusioned about their ability to overcome emotional stress, they are not aware of their most probable reactions which the literature has shown to be secondary victimization (e.g., Correia et al., 2001;Hafer & Bègue, 2005). This unawareness may contribute to the perpetuation of these automatic reactions because individuals will not try to correct them. This, however, is a post-hoc explanation which was not tested in the current study. Future studies should ask participants' perceptions of their and most others' emotional reactions and test whether they mediate perceptions of injunctive and descriptive normativity.
We would like to stress that none of the reactions was rated at its highest or lowest possible point. Next, we present some explanations for this pattern which also hint at limitations to our study. A possible reason may be connected to the operationalizations of reactions which seem "extreme" at least at face value. This may have caused participants to refrain from more fully agreeing or disagreeing with them. Another reason for this pattern may be that participants rated opposing reactions, the operationalizations of which were somewhat symmetrical. This fact may have unwittingly reminded participants that there are two sides to the matter and, as a consequence, they may have moderated their ratings. Finally, there could be several norms exerting their influence simultaneously. In the case of blaming the victim, participants may have indicated that they do not completely approve of nor would completely engage in it because a norm of compassion for those who suffer deeply moderated the impact of the norm of responsibility. These limitations could be addressed in future studies by the use of less "extreme" and asymmetrical operationalizations of reactions, and by normatively focusing participants (Cialdini et al., 1991) on either a norm of compassion or a norm of responsibility before they rate reactions. Also, it could be fruitful to turn this research to an intergroup level of analysis. In fact, although we have stressed the interpersonal level of analysis because neither the victim nor most other people were categorized, our explanation for the normative distinction between selfmost others is consistent with other theoretical interpretations (Social Identity Theory - Tajfel & Turner, 1986; Social Categorization Theory -e.g., Turner, Oakes, Haslam, & McGarty, 1994). Given that these theories address intergroup relations, it would be very interesting to manipulate the social category of "others" as in-group or out-group members in order to test the hypothesis that in-group members will be judged to adhere more strongly to norms towards victims than outgroup members.
Despite the above mentioned limitations we believe that our study has indicated new research directions in this domain. Individuals engage in positive self-other distinction, as our results show, by indicating that they themselves adopt those norms to an even higher extent than most other people. Given the strong influence of norms on behaviors (Nolan et al., 2008) and that injunctive norms protect victims (Alves & Correia, 2009), it is imperative that individuals are made aware of the discrepancy between what they think they themselves would typically do (i.e., non secondary victimization) and what individuals actually do (i.e., secondary victimization) even when they were supposed to engage in non secondary victimization (van den Bos & Maas, 2009). Our work may be used to guide campaigns aiming to inform individuals about this counterintuitive discrepancy which they are most probably unaware of. As in other domains, these campaigns may not prevent automatic reactions from happening. Without such campaigns, however, individuals are likely to remain in their normative delusion that victims can count more on them than on most people. As a result, they do not exert the required effort to "correct" their automatic reactions, thus failing to provide these people with the just and humane treatment that they are entitled to. Moreover, our results may guide campaigns about victimization which involve abuse/violence (such as domestic abuse). By stressing that society at large does not accept victimization, campaigns could encourage bystanders to act (e.g., by complaining to authorities) instead of remaining silent.