What Makes Characters’ Bad Behaviors Acceptable? The Effects of Character Motivation and Outcome on Perceptions, Character Liking, and Moral Disengagement

Many characters in entertainment content behave in morally questionable ways at least some of the time. However, the negative effects of those behaviors on individuals’ judgments of the character may be diminished in some instances. This study examined the effects of character motivation and outcome in a written narrative on character perceptions and moral disengagement. The findings of a 2 (motivation: altruistic, selfish) × 2 (outcome: positive, negative) experiment (N = 123) revealed that both motivation and outcome affect perceptions of the character's positive and negative attributes, character liking, and moral disengagement. Specifically, altruistic motivations and positive outcomes led to more favorable perceptions of a character's attributes, greater character liking, and more justification of the character's actions. Furthermore, moral disengagement mediated the effects of motivation and outcome on perceptions of characters' positive and negative attributes, and in turn positive attributes mediated the effect of moral disengagement on character liking. Implications for future entertainment research are discussed.

experiment (N ¼ 123) revealed that both motivation and outcome affect perceptions of the character's positive and negative attributes, character liking, and moral disengagement. Specifically, altruistic motivations and positive outcomes led to more favorable perceptions of a character's attributes, greater character liking, and more justification of the character's actions. Furthermore, moral disengagement mediated the effects of motivation and outcome on perceptions of characters' positive and negative attributes, and in turn positive attributes mediated the effect of moral disengagement on character liking. Implications for future entertainment research are discussed.

INTRODUCTION
Main characters or protagonists are generally perceived to be the ''good guys,'' but many of these characters do bad things at least some of the time. Nonetheless, these morally ambiguous characters are often well liked by audience members. Past research has examined the effects of characters' good and bad actions on audience responses (e.g., Krakowiak & Oliver, 2012;Tamborini, Weber, Eden, Bowman, & Grizzard, 2010;. Findings suggest that typically, characters who do good things are judged to be good and are thus well liked, whereas characters who do bad things are judged to be evil and are disliked (Zillmann, 2000). However, often times even good characters perform immoral or evil acts. For example, popular television characters such as Gregory House, Walter White, Dexter Morgan, Don Draper, and others behave badly much of the time, yet many individuals like these characters and root for them to succeed. One explanation for this may be that individuals are able to excuse these actions through the process of moral disengagement (see Raney, 2004). Furthermore, this process may be facilitated if certain cues are present in the narrative. Specifically, a character's motivation for an immoral act and the outcome that the act produces may affect the level of moral disengagement that occurs. For example, it may be easier to excuse an immoral action if the character's motivation is altruistic rather than selfish. Likewise, moral disengagement may be encouraged if the immoral act produces a positive rather than negative outcome. Moral disengagement, in turn, may affect other audience responses, such as perceptions of the character and character liking (Hartmann & Vorderer, 2010;Klimmt, Schmid, Nosper, Hartmann, & Vorderer, 2006;Raney, 2004;Shafer, 2009;. This research thus empirically tests the effects of both a main character's motive for a morally questionable action and the outcome of that action for a secondary character on audience responses to narrative content. Employing an experimental design that manipulates the type of character motivation (altruistic or selfish) and character outcome (positive or negative), this study examines their impact on moral disengagement, perceptions of the main character's positive and negative attributes, and character liking.

MORAL DISENGAGEMENT
When individuals see another person perform an immoral act, it is typical for them to condemn this action. However, at times, such as when the individual commits the immoral act him-or herself or when the perpetrator of the act is well liked, individuals may be motivated to excuse or justify the bad behavior in order to maintain cognitive consistency (cf. Festinger, 1957). Raney (2004) expanded on this idea and suggested that because individuals are motivated to enjoy entertainment content, they will justify the bad behaviors of protagonists through a process of moral disengagement. Recent research has supported this proposition by showing that individuals justify the immoral actions of characters in various entertainment contexts, such as video games (e.g., Hartmann & Vorderer, 2010;Shafer, 2009), and written narratives .
Moral disengagement may occur via one of several routes, focusing on the action itself, the perpetrator of the action, or the outcomes of the action (Bandura, 1986(Bandura, , 1999(Bandura, , 2002. For example, lying can be referred to as ''telling white lies'' (euphemistic labeling) and be justified by suggesting that doing so spares others' feelings (moral justification). Implying that everyone lies can minimize the liar's role (diffusion of responsibility), and proposing that lying doesn't really hurt anyone can minimize its outcomes (distortion of consequences). Immoral actions can also be justified by blaming someone of authority for the action (displacement of responsibility), by comparing the action to something worse (advantageous comparison), by blaming the victim for the conduct (attribution of blame), or by perceiving the victim to be less than human (dehumanization). Because moral disengagement relies on specific mechanisms, it is possible that certain factors may facilitate or hinder this process. For example, Tsay and Krakowiak (2011) found that perceiving a character as being similar to oneself and identifying with a character encourages justification of the character's immoral behaviors through moral disengagement. Moreover, research has shown that when violence in a video game was presented as being justified, game players felt less guilty and had less negative affect than when no justification was given (Hartmann, Toz, & Brandon, 2010;Hartmann & Vorderer, 2010). It is thus plausible that other narrative factors, specifically, character motivations and outcomes, may likewise affect moral disengagement.

CHARACTER MOTIVATIONS
Individuals make judgments about characters based on a variety of intrinsic and extrinsic factors. For example, physical attractiveness of characters has been found to significantly predict favorability of character outcomes and perceptions of friendliness, goodness, and intelligence (Bazzini, Curtin, Joslin, Regan, & Martz, 2010). Furthermore, some personal values of protagonists that are associated with desirable behaviors include ambition, courage, responsibility, and love (Beckwith, 2009). Other factors such as a character's personality, ideals, speech characteristics, and behavioral intentions are also related to the ways we respond to media characters (see Hoffner & Cantor, 1991, for a review). However, according to Zillmann (2000) individuals mainly form affective dispositions toward characters based on the characters' actions and motivations. Furthermore, affective disposition theory proposes that if characters behave morally or with good intentions, they will be liked, whereas if they behave immorally or with bad intentions, they will be disliked (Zillmann, 2000;Zillmann & Cantor, 1977). A character's motivations for an immoral act can thus influence how the character is perceived.
According to theories of moral development, as children develop (around the age of 7), they begin to focus more on a person's motivation for an action when determining whether the action was morally righteous rather than on other factors (see Kohlberg, 1984;Piaget, 1981). Empirical evidence shows that older children are more likely to consider the intention of an action and perceive immoral behaviors performed for ''good'' reasons fairly positively (Berndt & Berndt, 1975;Leifer & Roberts, 1972), suggesting that the ability to accurately understand characters' motives and intentionality strengthens over time. In particular, Krcmar and Cooke (2001) found that younger children are more likely to view unpunished violence on television as justified, whereas older children are more likely to pardon provoked violence. With respect to specific motivations, individuals justify aggressive actions by characters in films more readily and have fewer negative dispositions toward these characters when self-defense is the reason for violence, as compared to vengeance (Hoyt, 1970). Scholars have also argued that people may enjoy violent acts when these acts restore their sense of justice (Raney, 2002(Raney, , 2005Raney & Bryant, 2002). In other words, violence is perhaps more acceptable when it comes in response to another violent act. Therefore, liking of a superhero who beats up a criminal in order to restore justice may not diminish as a result of the violent act. In fact, Sanders (2005) found that individuals may even give villains the benefit of the doubt based on the villains' motivations for their evil acts. The aforementioned findings suggest that when a character's immoral actions are performed with the proper motivation, individuals' perceptions and liking of the character may not suffer. Therefore the following hypotheses are proposed: H1: A character who does something morally questionable for an altruistic reason will be perceived to have (a) more positive attributes and (b) fewer negative attributes than a character who does the same thing for a selfish reason. H2: A character who does something morally questionable for an altruistic reason will be liked more than a character who does the same thing for a selfish reason.
One reason that individuals may like and have positive perceptions of characters who have altruistic motivations for performing immoral actions is that it may be easier for individuals to morally disengage in this case. For example, Dr. House on the television show House routinely lied to his patients, but he did so in order to save their lives. Although most individuals would agree that lying to patients is immoral, they may be more willing to forgive this action if it is done with a positive or altruistic intent. On the other hand, if Dr. House lied to his patients for selfish reasons, it may be more difficult for audience members to justify his behavior. Therefore, the following hypothesis is proposed: H3: Individuals will morally disengage more when a character has an altruistic rather than selfish motivation for performing a morally questionable act.

OUTCOMES
The outcome that a character's immoral action creates may be another factor that influences individuals' levels of moral disengagement and perceptions of the character. According to affective disposition theory, individuals derive enjoyment from watching liked characters succeed and disliked characters fail (Zillmann, 2000;Zillmann & Cantor, 1977). However, the outcomes of characters' actions may also affect individuals' perceptions of these characters. Krcmar and Cooke (2001) found that children perceived a violent action to be more justified if the protagonist was not punished, regardless of the character's motivation for the action. Similarly, Collins and Zimmermann (1975) found that children exhibited more aggression when they watched a video in which the outcomes of a character's violent action were incongruent with his motivations. The authors explained that children might evaluate the violent act as being less severe when the character is not punished. Similarly, studies have shown that a successful outcome for a main character tends to increase his or her status (Albert, 1957;Bandura, 1977). Furthermore, the outcomes of a main character's actions on other characters may also have an effect on perceptions of the main character (Collins, Berndt, & Hess, 1974). For example, Gunter (1985) found that violent actions that produced negative outcomes for a victim were judged to be more violent than similar actions that produced no observable harm. Therefore, if a main character does something immoral that results in a positive outcome for another character, this may affect perceptions of the main character. Furthermore, this may occur regardless of the character's intent or motivation. To return to the previously used example of Dr. House, he may have run a dangerous test on a patient because he wanted to experiment with a new piece of equipment, but the test may have (inadvertently) saved the patient's life. It is possible that regardless, or perhaps, in spite of the doctor's motivation, this result will positively affect individuals' perceptions of him. Conversely, an immoral action that is performed for altruistic reasons but that results in a negative outcome for a neutral other may negatively affect perceptions of the character who performed the action. This is likely to occur as long as the secondary character is not disliked or seen as deserving a negative outcome. Therefore, the following hypotheses are proposed: H4: A character who does something morally questionable will be perceived to have (a) more positive attributes and (b) fewer negative attributes when the action results in a positive rather than a negative outcome for a neutral secondary character. H5: A character who does something morally questionable will be liked more when the action results in a positive rather than a negative outcome for a neutral secondary character.
Outcomes of character actions may also affect the degree to which individuals morally disengage. Several moral disengagement mechanisms rely on the distortion or minimization of consequences (Bandura, 1986(Bandura, , 1999(Bandura, , 2002. As a result, a positive outcome of an immoral act may make it easier to justify and accept the action; conversely, a negative outcome may make moral disengagement more difficult. Specifically, the following hypothesis is proposed: H6: Individuals will morally disengage more when a morally questionable act results in a positive outcome for a neutral secondary character than when it results in a negative outcome for a neutral secondary character. In addition to the aforementioned six hypotheses, we propose a model that further delineates the relationships among the variables of interest (see Figure 1). Specifically, this model proposes that altruistic-as compared to selfish-character motivations will result in higher levels of moral disengagement (Path a). Likewise, positive-as compared to negative-outcomes will result in higher levels of moral disengagement (Path b). Higher levels of moral disengagement will then result in perceptions of more positive (Path c) and fewer negative (Path d) character attributes. Perceptions of a character's attributes will in turn affect character liking such that positive character attributes will lead to increased liking of the character (Path e), whereas negative character attributes will lead to decreased liking of the character (Path f). This model also implies some indirect effects. Specifically, motivation and outcome will lead to positive and negative character attributes via moral disengagement. Furthermore, moral disengagement will lead to greater character liking through positive and negative character attributes.

METHOD Participants and Procedure
A total of 123 students (88.6% female, 11.4% male) participated in an online experiment for extra credit. Participants were recruited from communication courses at two large universities, one in the West and the other in the Northeast. Participants ranged in age from 19 to 30 years (M ¼ 20.24, SD ¼ 1.83), and 76.4% reported having some White ancestry, 14.6% had Asian ancestry, 9.8% had Hispanic ancestry, and 6.5% had African American ancestry.
Participants were randomly assigned to one of four conditions in a 2 (motivation: altruistic, selfish) Â 2 (outcome: positive, negative) betweensubjects factorial design. Each participant answered some demographic questions before being instructed to read a short narrative (approximately FIGURE 1 Proposed model. CHARACTER MOTIVATIONS AND OUTCOMES 700 words). After reading the story, participants answered a questionnaire measuring manipulation check items and questions assessing the dependent variables. These items were counterbalanced to avoid order and demand characteristic effects.

Stimulus
A story about two mountain climbers was the basis for the four condition manipulations. The story is largely written from the perspective of the main character, Craig, and begins as the two climbers are a few hours from summiting Mount Everest. However, about halfway into the story, the secondary character, Jon, begins to feel weak, and the two characters receive a transmission from some climbers on the summit that suggests treacherous weather ahead. At this point in the story, three sentences were edited to create the two motivation conditions. Specifically, in the altruistic motivation condition, Craig decides to leave Jon behind in order to try to help the climbers from whom they received the transmission. Contrastingly, in the selfish motivation condition, Craig leaves Jon behind because he wants to make it to the summit. The story continues as Jon attempts to descend the mountain himself. The last paragraph of the story was edited to create the two outcome conditions. In the positive outcome condition, Jon makes it to a camp where he is safe; in the negative outcome condition, Jon falls asleep in the snow and succumbs to hypothermia.
A manipulation check was conducted to ensure that the manipulations were successful. After reading the narrative, participants answered five questions on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Three questions measured the main character's motivation for leaving the secondary character behind (i.e., Craig's motivation for leaving Jon behind was altruistic, Craig left Jon behind because he wanted to help the stranded climbers, and Craig's motivation for leaving Jon behind was selfish [reverse coded]; a ¼ .79, M ¼ 2.78, SD ¼ 1.08), and two questions measured the type of outcome (positive vs. negative) the secondary character experienced (i.e., The ending was positive for Jon and The story did not end well for Jon (reverse coded); a ¼ .83, M ¼ 2.57, SD ¼ 1.14). Two 2 (motivation: altruistic, selfish) Â 2 (outcome: positive, negative) analysis of variance (ANOVA) tests were conducted to examine whether the manipulations were successful. For the Motivation scale, the analysis revealed a main effect for motivation conditions, F(1, 119) ¼ 187.28, p < .001, g 2 p ¼ :61, but not for outcome conditions, F(1, 119) ¼ .105, p ¼ .75, g 2 p ¼ :001. Conversely, for the Outcome scale, the analysis revealed a main effect for outcome conditions, F(1, 119) ¼ 128.77, p < .001, g 2 p ¼ :52, but not for motivation conditions, F(1, 119) ¼ 2.21, p ¼ .14, g 2 p ¼ :02. The results were in the expected directions, such that the main character's motivation was perceived to be significantly more altruistic in the altruistic condition (M ¼ 3.63, SD ¼ .64) than in the selfish condition (M ¼ 1.95, SD ¼ .71). Likewise, the outcome for the secondary character in the story was perceived to be significantly more positive in the positive (M ¼ 3.39, SD ¼ .76) than in the negative condition (M ¼ 1.76, SD ¼ .83). Therefore, it was concluded that the manipulations were successful in creating the four conditions.

Dependent Measures
All items measuring the dependent variables used 7-point Likert-type scales ranging from 1 (strongly disagree) to 7 (strongly agree).
Positive and negative attributes. Nine items adapted from person perception and impression formation literature (Hoffner, 1996;Pfau & Mullen, 1995) measured perceptions of the main character's positive and negative traits. Character attributes were measured on two distinct scales to account for the findings that-as opposed to clearly good or bad charactersmorally ambiguous characters have both positive and negative traits (see Krakowiak & Oliver, 2012). Five items measured positive attributes (e.g., Craig does some good things, Craig has some positive attributes, etc.; a ¼ .89, M ¼ 4.25, SD ¼ 1.17) and four items measured negative attributes (e.g., Craig does some immoral things, Craig has some negative attributes, etc.; a ¼ .84, M ¼ 4.76, SD ¼ 1.19).
Character liking. Character liking is typically measured with two items (e.g., I like the main character and I dislike the main character [reverse coded]; Zillmann & Bryant, 1975;Zillmann & Cantor, 1977). Three additional items were used to measure liking of the main character (e.g., I would like to be friends with someone like Craig, I admire Craig, etc.; a ¼ .89, M ¼ 3.53, SD ¼ 1.36).
Moral disengagement. Moral disengagement has been addressed as a theoretical concept in entertainment content (Raney, 2004); however, no general scale that can be adapted to a variety of entertainment contexts exists. Krakowiak and Tsay (2011) developed a 15-item scale using Bandura's (2002) descriptions of specific moral disengagement mechanisms. This scale was designed for a specific narrative, rather than for moral disengagement in general, in order to better assess the specific moral disengagement mechanisms that were used by participants while reading the narratives. Because the narrative in the Krakowiak and Tsay (2011) study resembled the one used in this study, nine items from the scale were adapted for use in the current study. Specifically, the moral disengagement mechanism of moral justification was measured with the item, It was all right for Craig to leave Jon behind because Craig thought Jon would be safe; attribution of blame was measured with the item, It was Jon's fault that he was left behind; distortion of consequences was measured with the item, Leaving a climber behind doesn't necessary hurt him=her; diffusion of responsibility was measured with the item, Craig should not be blamed for leaving Jon behind because other climbers would have done the same thing; and advantageous comparison was measured with the item, Leaving a fellow climber behind is not as bad as killing someone. No items were used to measure euphemistic labeling, dehumanization, or displacement of responsibility because the content of the story did not lend itself to the use of those mechanisms for moral disengagement. However, four additional items were used to measure overall moral disengagement (Craig was morally justified in his actions, I consider Craig's actions as unethical [reverse coded], I believe that in general Craig is an ethical person, and It was all right for Craig to leave Jon behind). One item, measuring advantageous comparison, showed low levels of item-to-total correlations and was subsequently dropped from the analysis. The remaining eight items were averaged to create a Moral Disengagement scale, in which higher values indicate higher levels of disengagement (a ¼ .88, M ¼ 3.28, SD ¼ 1.19).

RESULTS
To test the proposed hypotheses, a series of factorial ANOVAs were conducted. H1 predicted that a main character with an altruistic motivation would be perceived to have more positive and fewer negative traits than a main character with a selfish motivation for a morally questionable act. Similarly, H4 predicted that a main character whose actions produced a positive outcome for a neutral secondary character would be perceived to have more positive and fewer negative traits than a main character whose actions produced a negative outcome. To test these hypotheses, two 2 (motivation: altruistic, selfish) Â 2 (outcome: positive, negative) ANOVA tests were conducted. For the Positive Attribute scale, the analysis revealed a main effect for the motivation conditions, such that a main character in the altruistic conditions was perceived to have significantly more positive traits than the main character in the selfish conditions, F(1, 118) ¼ 28.16, p < .001, g 2 p ¼ :19 (see Table 1 for means). Likewise, for the Negative Attribute scale, the analysis revealed a main effect for the motivation conditions, such that the main character in the altruistic conditions was perceived to have significantly fewer negative traits than the main character in the selfish conditions, F(1, 118) ¼ 35.52, p < .001, g 2 p ¼ :23. Therefore, H1 was supported.
Testing for the effect of the outcome conditions on the Positive Attributes scale revealed no significant main effect for outcome conditions, indicating that the main characters in the positive and negative outcome conditions were perceived to have equally positive traits, F(1, 118) ¼ 1.51, p ¼ .22, g 2 p ¼ :01 (see Table 2 for means). However, for the Negative Attributes scale, the analysis revealed a main effect for the outcome conditions, such that the main character in the negative outcome conditions was perceived to have significantly more negative traits than the main character in the positive outcome conditions, F(1, 118) ¼ 4.43, p < .05, g 2 p ¼ :04. Therefore, H4 was partially supported.
H2 predicted that a main character with an altruistic motivation would be liked more than a main character with a selfish motivation. Similarly, H5 predicted that the main character in the positive outcome conditions would be liked more than the main character in the negative outcome conditions. To test these hypotheses, a 2 (motivation: altruistic, selfish) Â 2 (outcome: positive, negative) ANOVA test was conducted. The results revealed a main effect for motivation, indicating that the main character in the altruistic conditions was liked significantly more than the main character in the selfish conditions, F(1, 118) ¼ 44.34, p < .001, g 2 p ¼ :27 (see Table 3 for means). Likewise, a main effect for outcome conditions was revealed, indicating that the main character in the positive outcome conditions Note. Using Holm's sequential bonferroni post hoc comparisons, means across rows with no subscript in common differ at p < .05.

CHARACTER MOTIVATIONS AND OUTCOMES
was liked significantly more than the main character in the negative outcome conditions, F(1, 118) ¼ 11.01, p < .001, g 2 p ¼ :09. Therefore, H2 and H5 were both supported in that main characters in the altruistic motivation and positive outcome conditions were liked more than main characters in the selfish motivation and negative outcome conditions, respectively.
H3 predicted that individuals would be more likely to morally disengage if the character had altruistic rather than selfish motives, and H6 predicted that individuals would be more likely to morally disengage if the outcome for a neutral other was positive rather than negative. To test these hypotheses, a 2 (motivation: altruistic, selfish) Â 2 (outcome: positive, negative) ANOVA test was conducted. The results revealed a main effect for motivation, indicating that individuals were significantly more likely to Note. Using Holm's sequential bonferroni post hoc comparisons, means across rows with no subscript in common differ at p < .05. Note. Using Holm's sequential bonferroni post hoc comparisons, means across rows no subscript in common differ at p < .05. morally disengage in the altruistic conditions than in the selfish conditions, F(1, 118) ¼ 53.84, p < .001, g 2 p ¼ :31 (see Table 4 for means). Likewise, a main effect for outcome conditions was revealed, indicating that individuals were significantly more likely to morally disengage in the positive outcome conditions than in the negative outcome conditions, F(1, 118) ¼ 3.62, p < .05, g 2 p ¼ :03. Therefore, H3 and H6 were supported in that individuals were most likely to morally disengage in the altruistic motivation and positive outcome conditions.

Supplemental Analysis
No interaction effects were found in any of the analyses. Furthermore, the effect sizes for the main effects of the motivation conditions were larger than for the outcome conditions, indicating that a character's motivation likely had a greater influence on the dependent variables than did outcomes. To test whether these differences were statistically significant, Fisher's r-to-z transformations were conducted (see Cohen & Cohen, 1983). The analyses revealed that the correlations between motivation and positive attributes, and outcome and positive attributes differed significantly, t(120) ¼ 2.84, p < .01, indicating that the main character's motivation had a greater effect on perceptions of the main character's positive attributes than did outcomes. Likewise, motivation had a greater effect than outcomes on perceptions of the main character's negative attributes, t(120) ¼ À 2.57, p < .01; character liking, t(120) ¼ 2.33, p < .01; and moral disengagement, t(120) ¼ 3.85, p < .001. Therefore, although outcome for a neutral secondary Note. Using Holm's sequential bonferroni post hoc comparisons, means across rows with no subscript in common differ at p < .05.

CHARACTER MOTIVATIONS AND OUTCOMES
character affected perceptions of the main character's negative attributes, character liking, and moral disengagement, the main character's motivation had a stronger effect on these variables.
The hypothesized model was tested using AMOS with the motivation condition coded as 0 ¼ selfish, 1 ¼ altruistic, and the outcome condition coded as 0 ¼ negative, 1 ¼ positive. The initial run of the model showed that all of the predicted paths were statistically significant; however, the overall fit of the model could be improved, v 2 (8) ¼ 52.80, p < .001, comparative fit index ¼ .88, root mean square error of approximation ¼ .21. The modification indices indicated the utility of adding two direct paths, one between moral disengagement and character liking and the other between outcome condition and character liking. Given that we believe the presence of these paths has theoretical implications (discussed in the next section), we decided to include the two direct paths in our final model. The resultant model had acceptable fit, v 2 (6) ¼ 7.23, p ¼ .30, comparative fit index ¼ 1, root mean square error of approximation ¼ .04. Figure 2 shows the paths associated with this revised model. The two additional paths were statistically significant, and all but one of the hypothesized paths were statistically significant.
Overall, the model supported our hypotheses. It revealed that an altruistic motivation (b ¼ .56, p < .001) and positive outcome (b ¼ .15, p < .05) led to greater moral disengagement. Furthermore, higher levels of moral disengagement led to perceptions of more positive (b ¼ .69, p < .001), and fewer negative (b ¼ À.69, p < .001) character attributes, and to greater character liking (b ¼ .59, p < .001). Perceptions of positive character attributes also resulted in greater character liking (b ¼ .20, p < .01). Surprisingly, perceptions of negative character attributes did not decrease character liking (b ¼ À.12, p ¼ .08), but outcome did have a direct effect on character liking (b ¼ .13, p < .01). In addition, significant indirect paths supported our prediction that outcome and motivation affect character liking via moral disengagement. Bootstrapping procedures using 2,000 bootstrap samples and bias-corrected confidence intervals revealed that moral disengagement mediated the indirect paths between motivation and positive (b ¼ .38, p < .001) and negative (b ¼ À.39, p < .001) character attributes. Likewise, moral disengagement mediated the indirect paths between outcome and positive (b ¼ .10, p < .05) and negative (b ¼ À.10, p < .05) character attributes. Furthermore, the indirect path from moral disengagement to character liking was mediated by positive character attributes (b ¼ .14, p < .01) but not by negative character attributes (b ¼ .08, p ¼ .10).

DISCUSSION
This research theoretically contributes to entertainment literature by showing that a character's motivation for an immoral action and the outcome of the action impact a host of audience responses. Specifically, altruistic (as compared to selfish) character motivations and positive (as compared to negative) outcomes facilitated moral disengagement, resulting in more positive and fewer negative attributes given to a morally ambiguous character. Although this study garners support for the role of morality in affecting character judgments (see Hartmann & Vorderer, 2010;Klimmt et al., 2006;Raney, 2004;Shafer, 2009;, it also uniquely demonstrates that appraisals of morally ambiguous characters are multifaceted and more complex when factors such as altruism and the nature of behavioral outcomes are considered in a narrative context. Results from this study clearly indicate that character assessments (e.g., attributes, liking and moral nature) are more driven by behavioral intentions as compared to action outcomes. These findings are consistent with prior research supporting the significance of character motivations when forming evaluations of characters (Berndt & Berndt, 1975;Hoyt, 1970;Krcmar & Cooke, 2001, Leifer & Roberts, 1972. In fact, research on attribution of responsibility suggests that adults tend to rely more heavily on motivational cues than on consequences when evaluating negative behaviors of individuals (Fincham & Jaspers, 1980); these patterns were also found with media characters (Collins et al., 1974). Such reasoning takes into account theories of moral development, which posits that with age, logical analysis and the consideration of goals and intentions determine the moral righteousness of actions (see Kohlberg, 1984;Piaget, 1981).
In addition to character motivation, the outcome of an immoral action had a bearing on the degree to which individuals assessed character traits when reading a narrative. Although a character was perceived to have fewer negative traits if the immoral action resulted in a positive outcome, there was no difference in perceptions of a character's positive traits. One explanation for this finding is that evaluations of a character's positive qualities may be more rigid than that of a character's negative qualities, with the latter being a more salient and malleable assessment of morally ambiguous characters. It is also possible that individuals have a threshold for rating the goodness of characters who do immoral deeds; they perhaps will not exceed this threshold regardless of the outcome of such actions. Furthermore, taking into consideration that viewers are motivated to enjoy media and will take measures to remain liking their favorite characters for the sake of pleasure and excitement (Raney, 2004), it is understandable why people may assess characters as ''less bad'' if the outcomes of their immoral actions are positive. This result lends support to the idea that moral evaluations of characters are multidimensional, such that positive and negative attributes are interpreted independently. In particular, a morally ambiguous character may be perceived to have both positive and negative traits, whereas less ambiguous characters are rated high on one type of trait and low on the other (see Krakowiak & Oliver, 2012).
This research also revealed that the nature of the outcome produced by an immoral act and moral disengagement had an effect on character liking. This finding is somewhat surprising considering that characters cannot be sure of the outcome that their actions will produce; that is, the outcomes are not always entirely a result of the characters' actions, as was true in the stories used in this study. In both the positive and negative outcome conditions, the outcome for the secondary character is not guaranteed. Nevertheless, even though the main character has minimal or no control over the outcome, individuals appear to base their judgments and liking of the character on it. These findings support the attribution of responsibility theory, which suggests that the valence of outcomes influences the amount of blame or accountability placed on those who purposefully or inadvertently caused them (see Fincham & Jaspers, 1980). Such responsibility attributed to characters based strictly on the ending of a story or aftermath of immoral actions without consideration of external factors (e.g., intention or reasoning) theoretically explains the inclusion of the direct path between positive outcome and character liking in our final model.
Although motivation and outcome affected perceived character attributes and likability, the extent to which immoral behaviors were justified or perceived as less severe also differed. The findings suggest that it may be easier to rationalize bad behaviors when the intentions underlying them are selfless and when no other characters suffer as a result of such actions. This may be especially true for the moral disengagement mechanisms that focus on moral justification and the minimization or distortion of consequences (see Bandura, 1986Bandura, , 1999Bandura, , 2002. Hence, the present study supports the significant impact of narrative content (e.g., character intentions and outcomes) on the extent to which readers defend or excuse the immoral actions of characters.
Theorizing about the role of morality in entertainment media is greatly enhanced by not only the antecedents found in this study but also its effects on the attribution of positive and negative character traits and character liking. Although moral disengagement led individuals to assign more positive attributes and fewer negative attributes to a morally ambiguous character, only positive attributes predicted character liking. This result is perhaps explained by the spillover effects of moral disengagement. Furthermore, it may be the case that moral disengagement suppresses the effects that negative attributes have on character liking. In other words, individuals who morally disengage may continue to like a character regardless of the negative traits that the character possesses. In addition, the path between moral disengagement and character liking was included in the final model because theoretically, excusing and justifying immoral character behaviors could directly impact character liking without being contingent on the attribution of character traits. Instead, it could be the case that moral disengagement fosters overall enjoyment of a narrative, and such pleasure and excitement transfers to the disposition toward the character. Taking into account the theory of attribution of responsibility (Fincham & Jaspers, 1980), it is possible that the sheer recognition of an immoral act as less severe or excusable enhances the likability of a character, especially if the motivation is selfless and the outcome is positive.

LIMITATIONS AND DIRECTIONS FOR FUTURE RESEARCH
This study has a number of limitations and therefore proposes some avenues for future research. In this study, a short written narrative was used to create different conditions; as a result, individuals were given a small amount of information about the main character on which to base their evaluations. Therefore, the effects of the manipulations may be stronger than in other entertainment contexts that provide more information about characters. For example, in films and television programs, viewers may learn more contextual and visual information, such as a character's appearance, background, environment, and nonverbal gestures or cues, which has a bearing on their interpretations of his or her motives or actions. It would be important for future research to investigate the ways in which variables, including but not limited to physical attractiveness (Bazzini et al., 2010), personal values (Beckwith, 2009), and personality and speech characteristics (Hoffner & Cantor, 1991), influence the moral disengagement process and, in turn, character attribution and liking. Moreover, these factors may hold a greater role in the context of more visually stimulating and interactive media. In light of more enhanced interactive experiences with media content, such as video games, narrative elements become even more important in how they may facilitate the justification of immoral or violent behaviors of characters (see Hartmann & Vorderer, 2010;Shafer, 2009).
In addition, the present study used a written narrative that was edited to create a positive and negative outcome. Although participants rated the positive outcome to be fairly neutral (M ¼ 3.39, SD ¼ .76) on a 7-point scale, they did perceive the positive outcome to be significantly more positive than the negative outcome. Furthermore, the sample in this study consisted primarily of women (88.6%). Taking into consideration gender differences in moral reasoning, empirical support has shown that women are more likely to use a care orientation when assessing moral conflict, whereas men rely on a more justice-focused approach (see Ford & Lowery, 1986). Future studies may consider investigating how the gender of the viewer perhaps influences the assessment of morally questionable behaviors or conflicts based on potentially different moral criteria.
Moreover, in the present study, moral disengagement measures were adapted from those used in previous research . Although the measures were applicable to the short stories used as stimuli for the experiment, the items did not tap into the use of every type of moral disengagement mechanism, and no comparisons among the specific moral disengagement mechanisms could be made. Future research may consider creating and refining a moral disengagement scale that can be used across various entertainment contexts and that measures the use of particular moral disengagement mechanisms in media content. This would allow researchers to examine the effects of different story components on the use of certain moral disengagement mechanisms. For example, positive motivations could facilitate the use of moral justification mechanisms, whereas certain outcomes could encourage the use of the distortion of consequences mechanism.
Last, the results of this study indicate that character motivation greatly affects the ways in which characters are judged. In particular, individuals are willing to accept or justify bad behaviors when they are performed with good intentions. This may be one reason why entertainment content featuring antiheroes and morally ambiguous characters are deemed appealing to audiences. Perhaps these characters are liked because their motivations are deemed to be sufficiently positive. However, it is unclear if all bad actions can be redeemed with good motivations. Future research could thus examine whether the positive influence of good intentions diminishes as the severity of a bad action increases. For example, an altruistic motivation may not affect people's judgments of a character who kills an innocent victim. Therefore, it could be the case that the effects of selfless and noble intentions are inconsequential when the immoral actions are inexcusable and morally heinous. In addition, the perceptions of the victim of the immoral act may be altered in order to restore individuals' just-world beliefs (see Haynes & Olson, 2006).
Overall, the present research confirms the importance of character motivation and action outcomes in affecting character attributions and liking via the process of moral disengagement. Elucidating our understanding of the antecedents and outcomes of moral judgment, this study theoretically provides media scholars with deeper insight to how exposure to immoral actions with positive intentions and outcomes leads to more relaxed moral reasoning. In particular, findings contribute to entertainment literature by showing that the justification or rationalization of morally questionable behaviors have strong implications for character impressions. Furthermore, this research provides practical recommendations for content creators when constructing effective narratives. It is imperative that writers consider incorporating rich character developments into their stories, taking into account altruistic goals and intentions of the protagonist and ensuring that if immoral or antisocial behaviors were committed, they result in positive or happy endings. Such narrative features will indeed foster more positive dispositions toward characters and may consequently enhance overall program enjoyment. While character impressions may be short term, prolonged exposure to entertainment media featuring morally ambiguous characters with selfless or justifiable intentions for their actions may likely influence a viewer's own moral compass and reinforce character liking for the sake of continuing to enjoy his or her media experience.