Excusing murder? Conservative Jurors’ Acceptance of the Gay Excusing murder? Conservative Jurors’ Acceptance of the Gay Panic Defense Panic Defense

We conducted a simulated trial study to investigate the effectiveness of a “gay panic” provocation defense as a function of jurors’ political orientation. Mock jurors read about a murder case in which a male defendant claimed a victim provoked the killing by starting a fight, which either included or did not include the male victim making an unwanted sexual advance that triggered a state of panic in the defendant. Conservative jurors were significantly less punitive when the defendant claimed to have acted out of gay panic as compared to when this element was not part of the defense. In contrast, liberal jurors were unaffected by the gay panic manipulation. The effect of the gay-panic defense on punitiveness was mediated by conservatives’ decreased moral outrage toward the defendant. Implications for psychological theory and the legal system are discussed.


Excusing Murder? Conservative Jurors' Acceptance of the Gay-panic defense
In 2010, Vincent McGee was charged with capital murder and arson for beating and stabbing Richard Barrett and then setting fire to Barrett's house. Barrett was found with blunt trauma and severe burns on his head as well as 16 stab wounds to his neck and body. McGee confessed, but defended his actions by claiming that he had been provoked into killing Barrett because Barrett had made an unwanted sexual advance on McGee. McGee ultimately pled guilty to lesser charges of manslaughter, robbery, and arson. McGee's defense is an example of a "gay panic" defense (Chen, 2000), whereby a defendant argues that he or she is less culpable for a crime such as murder because the victim made an unwanted same-gender sexual advance toward the defendant, causing the defendant to lose control and act in the heat of passion.
Provocation defenses are interesting psychologically because jurors are asked to make a very subjective judgment about whether they believe lethal action is reasonable in light of the facts as the defendant believed them to be. In determining whether an action was reasonable, jurors might envision whether they could have experienced a similar emotional reaction and loss of control had they been in the same situation as the defendant. In cases involving the gay-panic defense, jurors might rely on their subjective ideological biases when deciding whether the defendant's reactions were reasonable. Jurors' political ideology could be particularly influential, given that conservatives tend to have more negative reactions to same-gender sexual behavior than do liberals, because it violates their values (e.g., Skitka & Tetlock, 1993). This might lead conservative jurors to be less morally outraged toward a defendant who killed a victim who allegedly made a same-gender sexual advance, relative to a case with an otherwise identical provocation defense that does not include the advance. Thus, our goal was to investigate whether a victim's male-male sexual advance would reduce conservatives' moral outrage toward the defendant and in turn result in a reduction of the defendant's charge from murder to manslaughter. In contrast, we predicted that the gay-panic defense would not reduce liberal jurors' moral outrage toward the defendant nor lead liberals to endorse more lenient sentencing because the victim's behavior does not violate liberal jurors' values. Next, we justify these hypotheses.

Gay Panic: A Form of Provocation Defense
Provocation defenses are offered in an effort to persuade the trier of fact to reduce the criminal charge against the defendant (e.g., from murder to manslaughter) in cases where there is no doubt that the defendant committed the act. Rather than determine guilt, jurors must determine whether the crime (usually homicide) was "committed under the influence of extreme mental or emotional disturbance for which there is a reasonable explanation or excuse" (Dubber, 2002, p. 376). There is no uniformity in provocation law across the United States, despite attempts to encourage uniformity by reforming the Model Penal Code's language regarding Extreme Emotional Disturbance (Dubber, 2002). Most jurisdictions rely on some variant of the provocation defense (Robinson, 2011) based in English common law (Horder, 1992). In his analysis of current provocation defense statutes, Robinson (2011) noted that only 11 states have formally adopted the Model Penal Code's Extreme Emotional Disturbance formulation of provocation defense. Further, even among those jurisdictions that did adopt Model Penal Code reform in other areas of their criminal codes, 23 of these reform jurisdictions retain various types of the common law variant of the provocation defense. Among jurisdictions that do not use the Model Penal Code formulation, provocation statutes contain a variety of elements related to acting in the "heat of passion" with sufficient provocation and the lack of a cooling-off period (Nourse, 1997;Robinson, 2011).
Thus, provocation defenses are successful if jurors perceive the provocation as sufficient to cause a person to lose self-control and commit the offense, and, as a result, decide the defendant is less culpable than he or she would be otherwise (Kirschner, Litwack, & Galperin, 2004;Mison, 1992). The gay-panic defense is a specific type of provocation defense in which the defendant claims that the crime in question was the result of a sudden and intense passion provoked by the victim's unwanted same-gender sexual advance. It is primarily used by straight men claiming that they found the experience of an unwanted same-gender sexual advance so upsetting that they temporarily became enraged and lost control of their own behavior (Lee, 2008). Chen (2000) argues that the acceptance of a gay-panic defense implies acceptance of a nonviolent same-gender sexual advance as an adequate trigger to cause a person to fall into an uncontrollable state of panic. If jurors collectively agree that the reaction was reasonable, they can find the defendant guilty of a lesser offense, which often results in a lesser sentence (Lee, 2008). We propose that acceptance of the gay panic defense might represent another form of anti-gay prejudice, alongside a few previous experimental demonstrations of discrimination against gay, compared to straight, defendants in child sexual abuse (Stawiski, Dykema-Engblade, & Tindale, 2012;Wiley & Bottoms, 2009) and murder cases (Ragatz & Russell, 2010).
To be clear, we are not concerned with whether "gay panic" is a real psychological phenomenon. Instead, our goal is to investigate jurors' reactions to this defense, given that it is used and sometimes successful in court. How prevalent is the gay-panic defense? A few courts have ruled that a nonviolent same-gender sexual advance may not be used as part of a provocation defense strategy (see Davis v. State, 2005;Janofsky, 1999), but most courts permit it (Lee, 2008). Gay panic strategies such as these have been offered in 45 trials that we know of since 2002 (Nichols, 2013;e.g., Van Hook v. Bobby, 2011). In addition, Harrington (2009) located 189 appellate cases, dating from 1952 to 2005, in LexisNexis in which defense arguments referenced unwanted gay advances, either as a formal part of trial strategy or in their descriptions of the facts of the cases. The gay panic defense appeared to be successful in gaining a reduced sentence in 13% of 189 gay panic cases he identified, either at the original trial (20 cases), or on appeal (5 cases). Given that this analysis utilized the appellate record, these statistics might underrepresent the true number of jury trials in which the gay-panic defense is invoked successfully, because defendants who are successful at trial will not have convictions to appeal.
One noteworthy recent (2011) case in which a murder defendant successfully argued provocation based on the victim's same-gender sexual advance is that of 14-year-old Brandon McInerney. McInerney shot a male classmate twice in the back of his head after McInerney stole his grandfather's gun and told a friend of his intentions the day before. At trial, the defense argued that the victim's sexual advances had driven McInerney into a dissociative state that led him to kill the victim in the heat of passion. The prosecutor attempted to neutralize this strategy in her closing statement by making the gay panic defense explicit in hopes of discrediting it, arguing "Let's just say it, this defense is gay panic. For the past six weeks, there's been this giant smoke screen" (Broverman, 2011, as cited in Perkiss, 2013. Despite evidence of premeditation, the jury hung, with 7 of the 12 jurors favoring a lesser sentence of manslaughter. (For an indepth analysis of this case as it pertains to the gay-panic defense, see Perkiss, 2013). In other cases, the gay panic defense has resulted in not only hung juries, but also defendants being found guilty of lesser charges, such as voluntary manslaughter (e.g., Thomas v. State, 1985;People v. Chavez, 2002) and heat-of-passion manslaughter (e.g., State v. Kulseth, 1983). There are also cases in which the gay-panic defense was initiated by a defendant without tangible evidence of the victim's alleged same-gender sexual advance. As such, defendants on trial for murder, who have few alternative strategies, might latch on to the gay-panic defense in an effort at making a case for mitigation.
Although we do not have complete statistics regarding the frequency of successful cases of gay-panic defenses, the American Bar Association (ABA) was concerned enough about the misuse of this variation of the provocation defense that it passed a motion to encourage all state legislatures to alter their criminal codes to eliminate the gay-panic defense (American Bar Association, 2013). Thus, the gay-panic defense is of serious concern for the legal community.
Our study is thus a timely empirical investigation of whether adding a same-gender sexual advance to a provocation defense will make jurors less punitive. It is particularly timely given the current limited evidence regarding actual jurors' reactions to gay-panic defenses.
There have been two published studies using a mock trial simulation to investigate jurors' acceptance of the gay-panic defense, but they do not speak to our question directly. Plumm, Terrance, Henderson, and Ellingson (2010) asked undergraduate participants to read a hate-crime scenario describing a defendant who assaulted a victim either in a gay bar or a local bar after the victim either put his arm around the defendant and asked him to dance or provided no provocation. Mock jurors' verdicts and guilt judgments were not affected by the manipulationsalthough jurors did blame the victim more when he had made a same-gender sexual advance relative to when he had not. Mock jurors in a study by Kraus and Ragatz (2011) were relatively more punitive in a man-on-man gay panic defense relative to a female-female gay panic defense-but this difference was eliminated when jurors were instructed to avoid letting their bias influence their judgments (as would all actual jurors via jury instructions). Neither of these studies, however, provided a comparison of the gay panic defense to an otherwise identical provocation defense to isolate the effect of the same-gender sexual advance portion of the provocation defense. That is, it is not clear whether the results from these two studies were a result of a same-gender sexual advance, or would occur in response to any form of victim provocation. To address this specifically, we conducted a mock trial study investigating the effectiveness of a gay-panic defense relative to a similar provocation defense that did not include an unwanted same-gender sexual advance, holding all other case facts constant.

Do Provocation Defenses Invite Juror Bias?
The Model Penal Code of 1962 requires jurors to determine whether a provocation could reasonably explain or excuse a defendant's homicidal behavior (Dubber, 2002). The law further asserts that determinations regarding the acceptance of provocation defenses must be made "from the viewpoint of a person in the actor's situation under the circumstances as he believed them to be" (sic, Dubber,p. 376). Thus, the law requires jurors to make a subjective judgment that the defendant's emotional reaction was real and reasonable in light of the facts as they were perceived by the defendant (e.g., People v. Casassa, 1980). Asking jurors to rely on their own beliefs about whether responding to an unwanted same-gender sexual advance by killing the provocateur is a reasonable reaction might open the door for jurors' subjective biases to influence their perceptions of the gay-panic defense and, in turn, case judgments.
The provocation defense might be similar to other case factors that exacerbate the effect of victim factors on jurors' judgments. For example, Greene and colleagues (Greene, 1999;Greene et al., 1998) found that when a victim impact statement described a victim as more versus less respectable, mock jurors valued the victim more and judged the murder as more serious (the researchers did not include measures of verdict or sentencing). Similarly, provocation defenses-particularly the gay-panic defense-might also exacerbate jurors' leniency for defendants based on the victim's behavior (e.g., victims who made a same-gender sexual advance on the defendant).
Political orientation. An additional reason that Plumm and colleagues' (2010) study did not find an effect for the gay-panic defense might be because they expected these effects to manifest for all mock jurors. Provocation defenses introduce the opportunity for jurors' subjective biases to affect judgments, but not all jurors bring the same biases to the courtroom. A gay-panic defense, for example, should be compelling only to those for whom the victim's provocation (i.e., a same-gender sexual advance) justifies retaliation-for example, jurors who have negative reactions to same-gender sexual behavior. We explored whether jurors' political orientation might influence acceptance of the gay-panic defense, because conservatives typically also have negative reactions to same-gender sexual behavior (Haider-Markel & Joslyn, 2008;Skitka & Tetlock, 1993).
Unfortunately-and somewhat surprisingly-investigations of juror political orientation and legal judgments are somewhat rare. One exception is Clark and Wink's (2012) study of individuals who had reported to jury duty, which revealed that political orientation was unrelated to guilt judgments or sentence length. Yet, political orientation might have a more nuanced effect on legal decision making. Political orientation changes the way jurors perceive crimes (Forgas, 1980). Further, political orientation can affect the extent to which jurors rely on different types of evidence in reaching death penalty decisions (e.g., defendant remorsefulness and features of defendant's psychopathy, Cox et al., 2013;defendant's alcohol abuse, Stevenson, Bottoms, & Diamond, 2010). Finally, conservatives and liberals often have different perspectives regarding the causes of misconduct (Weiner, 1993). Thus, although political orientation might not directly affect case judgments overall, it might influence how mock jurors evaluate evidence, victims, and offenders, which might result in different case judgments for conservatives as compared to liberals. For example, in the context of hate crimes, liberals are more certain in White perpetrators' guilt relative to Black perpetrators' guilt-an effect that is not significant for conservatives (Marcus-Newhall et al., 2002). Thus, political orientation might moderate the effects of other case factors on legal judgments.
Only one study that we are aware of has tested whether the effect of a victim factor was moderated by juror political orientation. Marcus-Newhall and colleagues (2002) found that political orientation did not moderate the effect of victim race on guilt certainty in a hate-crime case. We, however, expected political orientation to moderate the effect of a gay panic provocation defense, such that this defense would be effective for conservative, but not liberal, jurors. Why? Although social conservatives tend to have negative views of African Americans (Kinder & Sears, 1981), a victim's racial status does not necessarily violate conservatives' moral values-this might explain why political orientation did not moderate the effect of victim race.
In contrast, a victim's same-gender sexual advance does violate conservatives' values.
Political conservatives are less accepting of same-gender sexual behavior than are liberals (Haider-Markel & Joslyn, 2008;Skitka & Tetlock, 1993), less likely to support gay rights, and more likely to endorse gay stereotypes (e.g., beliefs that gay individuals are sexually promiscuous, perverted, or dangerous) and endorse homophobic attitudes (Wood & Bartkowski, 2004). Further, compared to liberals, conservatives dislike, fear, and react more defensively toward individuals who violate traditional norms of sexuality (Skitka & Tetlock, 1993). Finally, punishing a same-gender sexual advance is in line with conservatives' values (i.e., punishing violations of traditional sexuality, Skitka & Tetlock, 1993). Thus, because same-gender sexual behavior violates conservatives'-but not liberals'-values, conservative jurors might think that reacting violently to this behavior is more reasonable than do liberal jurors.

Mediating Mechanism: Moral Outrage
Finally, we investigated a potential underlying psychological mechanism for the effect of the gay-panic defense and political orientation on punitiveness. Theoretically, we expected the gay-panic defense to be effective for conservatives because the victim's same-gender sexual advance would violate their values and result in feelings of moral outrage. In contrast, we expected the gay-panic defense would not be effective for liberals because the victim's samegender sexual advance would not violate their values and elicit no moral outrage. Thus, we predicted a moderated mediation model in which the effect of the gay-panic defense on conservative-but not liberal-jurors' punitiveness would be explained by moral outrage toward the victim and defendant. Moral outrage is a constellation of cognitive, affective, and behavioral reactions to acts that people perceive as moral transgressions (Skitka, Bauman, & Mullen, 2004;Tetlock, Kristel, Elson, Green, & Lerner, 2000). The extent to which one experiences moral outrage predicts a variety of outcomes relevant to legal decision making, such as retribution and compensation (Carlsmith, Darley, & Robinson, 2002;Tetlock et al., 2000), verdicts (Salerno & Peter-Hagene, 2013) including verdicts in cases involving alleged sexual abuse perpetrated by gay men (Wiley & Bottoms, 2009), and punishment for crimes (Salerno et al., 2010). Because same-gender sexual behavior violates conservatives' values, we expected them to react to the gay panic provocation with less moral outrage toward the defendant and more moral outrage toward the defendant, both of which would, in turn, make them more lenient. In contrast, because samegender sexual behavior does not violate liberals' moral values, we predicted that liberals' feelings of moral outrage toward the victim and defendant and punitiveness would be unaffected by whether the provocation defense included an unwanted same-gender sexual advance.

Hypotheses
We predicted that conservative-but not liberal-jurors would be more likely to downgrade a defendant's charge from murder to manslaughter when the defendant's provocation defense included a same-gender sexual advance relative to when it did not (i.e., provocation type). We predicted that this interaction effect would be mediated by moral outrage toward the defendant and victim. Specifically, we predicted that conservatives who read about a gay panic versus non-gay panic provocation defense would experience more moral outrage toward the victim and less moral outrage toward the defendant, which, in turn, would make them less punitive. In contrast, we predicted that liberals' moral outrage toward the defendant and victim would not be affected by whether the provocation defense included a same-gender sexual advance, and as a result provocation type would not affect their punitiveness.
We also included measures to rule alternative mediators for the effect of provocation type on conservatives jurors' punitiveness--specifically, participants' perceptions of the crime. First, one could argue that the same-gender sexual advance represents an additional provocation (i.e., insult, yelling, and same-gender sexual advance) relative to the non-gay-panic provocation condition (i.e., insult and yelling) and perhaps conservatives are more sensitive to additional provocation than liberals. Second, one could also argue that the defendant might seem to be in more danger in the gay-panic (versus non-gay-panic) provocation condition because it might seem like more of a physical assault and perhaps conservatives might be more sensitive to heightened danger than liberals. Although our manipulation included an innocuous same-gender sexual advance that we do not believe could be construed as aggressive, we did include measures of both perceived provocation by the victim and perceived danger to the defendant. Because we hypothesized the driving force of the gay panic defense on conservatives to be moral outrage resulting from the violation of values, we predicted that the gay panic manipulation would not affect the extent to which mock jurors' perceived that the victim provoked the defendant or the degree to which the defendant was in danger.

Participants and Procedure
Seventy-seven men and women agreed to complete a survey after being approached in public places around a large Midwestern city (e.g., trains, train and bus stations, outdoor eating areas around lunchtime). We excluded three participants for failing a manipulation check. The remaining 74 participants (58% men) were ethnically diverse (69% Caucasian/White, 10% Asian, 7% African American, 7% Hispanic, and 7% Other) and jury-eligible United States citizens over 18 years old (M = 35, range = 19-67 years). Participants were randomly assigned to either the gay-panic provocation or the non-gay-panic provocation condition (i.e., the provocation type manipulation). Participants read experimental instructions, the case scenario that included the provocation type manipulation, and jury instructions. Next, they completed measures of verdict, verdict confidence, moral outrage toward the defendant and victim, perceptions of the crime, their own political orientation, and manipulation checks. Participants completed the survey voluntarily-they were not compensated.

Materials and Measures
Case scenario. Participants read the following instructions: You are going to read a summary of a legal case. There is no question that the defendant, Robert Coleman, killed the victim, John Kaslov. Please put yourself into the role of a juror, read the case, and decide which of two different verdicts you would recommend if you were a juror: (a) the more serious charge of murder, or (b) the less serious charge of manslaughter.
The participants next read the following vignette that included the provocation defense.
The vignette either did or did not include details about the victim making an unwanted samegender sexual advance toward the defendant. The additional gay-panic information, given only to participants in the gay-panic provocation condition, is shown in italics: The defendant, 35-year-old Robert Coleman, and the victim, 32-year-old John Kaslov, were acquaintances who met at a bar one evening by chance. They began drinking together and talking. As the bar was closing, the two men drove to a store for a bottle of bourbon. After parking the car on a side road, the two continued to drink. According to the defendant, at approximately 3:00AM, the victim insulted the defendant's wife. Jury instructions. Participants were given a set of Heat-of-Passion jury instructions that were adapted from those used in several jurisdictions (e.g., Illinois, Connecticut). The instructions were crafted so they would be easy for participants to understand and generally representative of the instructions used in the majority of states (see Appendix A). The instructions explained that whether the defendant committed the crime was not in question, but instead that jurors needed to determine whether the crime constituted murder or manslaughter.
The instructions also highlighted the difference between the two charges.
Verdict confidence. Participants chose a verdict of either "murder" or the lesser charge of "manslaughter." Participants also indicated their level of confidence in their verdict ranging from 0% to 100% confident, which we combined with their verdicts to create a 22-point verdictconfidence scale, ranging from 1 (manslaughter, 100% confident) to 22 (murder, 100% confident). Thus, lower scores on this scale indicate less punitiveness in that they reflect greater confidence in the desire to convict the defendant of a less serious crime that would result in less punishment. Participants reported their agreement with the two items on 6-point scales ranging from -3 (Strongly Disagree) to +3 (Strongly Agree).
Political orientation. Political orientation was assessed using the item, "When it comes to politics, how liberal or conservative are you?" Responses were given on a 7-point scale ranging from 1 (Extremely Liberal) to 7 (Extremely Conservative). This measure has been used in many previous studies to assess political orientation (e.g., Clark & Wink, 2012;Marcus-Newhall et al., 2002;Stevenson et al., 2011). The sample yielded a good range of political orientation scores, as reflected by the mean score on this item for our sample being roughly at the

Results
To test our hypotheses, we conducted a hierarchical logistic regression analysis for dichotomous verdicts and a series of hierarchical linear regression analyses for all other continuous dependent measures. Each regression included a dummy code representing provocation type condition (0 = non gay panic, 1 = gay panic) and jurors' centered political orientation scores entered as predictors in Step 1, and the interaction term between the dummy code and centered political orientation score entered in Step 2. Supplementary analyses of the simple slopes were used to follow up significant interactions revealed by the primary regression tests. All simple slopes analyses tested the effect of provocation type for conservative jurors (operationalized as 1 SD above the mean on the political orientation item) and then for liberal jurors (operationalized as 1 SD below the mean). Means and standard deviations of all continuous variables are reported in Table 1. Finally, we present results from moderatedmediation analyses exploring whether moral outrage mediated the predicted interaction effect.
Step 1 of the hierarchical logistic regression analysis (i.e., the main-effects-only model) revealed that the provocation type was not a significant predictor of verdicts, Odds Ratio (OR) = .60, Wald = 1.01, p = .31. In contrast, political orientation was a marginally significant predictor of verdicts, such that increased conservatism was associated with a marginal increase in the likelihood of choosing the murder verdict, OR = 1.37, Wald = 3.13, p = .08.
Step 2 of the analysis revealed that the main effect of political orientation was qualified, however, by the predicted significant interaction effect, OR = .48, Wald = 3.82, p = .05. 1 Simple slopes analyses revealed that provocation type did not significantly affect liberal jurors' likelihood of choosing the murder verdict, OR = 1.81, Wald = .62, p = .43. In contrast, as predicted, conservative jurors were significantly less likely to choose a murder verdict in the gay-panic provocation condition relative to the non-gay-panic provocation condition, OR = .21, Wald = 4.21, p = .04.

Verdict Confidence
The hierarchical linear regression analysis on verdict confidence revealed the same pattern of effects as for dichotomous verdicts.

Moral Outrage toward the Victim
An identical analysis of jurors' moral outrage toward the victim revealed no significant effects of provocation type, juror political orientation, nor the interaction term, Bs  .10, s  .12, ts  .59, ps  .33.

Perceptions of the Crime
To strengthen our argument that the gay panic effect is driven by conservatives' decreased moral outrage toward a defendant who killed a victim who violated their values, we tested two alternative explanations regarding jurors' perceptions of the crime itself. We ran the same hierarchical linear regression model including the dummy code for provocation type, centered political orientation, and the interaction term on mock jurors' perceptions of the extent to which (a) the victim provoked the murder, and (b) the defendant was in danger. First, neither provocation type, nor the interaction between provocation type and political orientation significantly affected the extent to which jurors thought the victim provoked the killing, Bs  |.25|, s  |.16|, ts  |.96|, ps  .34. Political orientation was marginally related to perceptions of victim provocation, however, such that the more conservative jurors were, the less they agreed that the victim provoked the defendant to kill the victim, B = -.24,  = -.22, t(68) = -1.82, p = .07. This does not provide an alternative explanation for why conservatives (but not liberals) were more lenient in the gay-panic condition because (a) the effect of political orientation on perceived victim provocation was the same regardless of provocation type and (b) conservatism decreased agreement that the victim provoked the killing, which should make them more rather than less punitive toward the defendant.
Second, neither provocation type, political orientation, nor the provocation type X political orientation interaction significantly affected the extent to which mock jurors thought the defendant was in danger, Bs  |.41|, s  |.15|, ts  1.23, ps  .20. Thus, the interaction effects on dichotomous verdicts and the verdict-confidence scale are not due to the same-gender sexual advance increasing jurors' perceptions of the degree to which the defendant was in danger.

Mediation Analyses
We conducted a test of moderated mediation to directly test our hypothesis that moral outrage explains why conservative jurors-but not liberal jurors-were less punitive in the gaypanic provocation condition versus non-gay-panic-provocation condition. Because provocation type did not affect perceptions of the crime, we did not include these measures in the mediation analysis. We used Hayes' (2013) PROCESS macro for SPSS, which is a computation tool that uses bootstrapping techniques to complete this type of analysis. Specifically, we tested whether the conditional indirect effects of provocation type on jurors' verdict confidence through moral outrage toward the defendant and victim were significant for conservative jurors but not liberal jurors. Given the significant interactive effects of provocation type and political orientation on both moral outrage and verdict confidence, we tested a model in which political orientation moderated the path from the predictor (i.e., provocation type) to both mediators (i.e., moral outrage toward the defendant and victim) and the dependent variable (i.e., verdict confidence) (i.e., Model 8 in Hayes, 2013). Significant indirect effects are indicated by confidence intervals (CIs) that do not include zero (Hayes, 2013).
Contrary to our prediction, the conditional indirect effect of provocation type on verdict confidence through moral outrage toward the victim was not significant among either conservative jurors, coefficient = -.12, 95% CI: lower = -1.21, upper = .88, or liberal jurors, coefficient = -.04, 95% CI: lower = -1.84, upper = .76. 2 Thus, among conservative jurors-but not liberal jurors-the effect of provocation type on verdict confidence was mediated by moral outrage toward the defendant but not the victim. That is, among conservative jurors, those exposed to a gay panic provocation defense were less morally outraged toward the defendant and, in turn, more confident in a less punitive charge as compared to those exposed to a typical provocation defense.

Discussion
To our knowledge, the current research is the first direct test of the effectiveness of the gay-panic defense over an otherwise identical provocation defense. Mock jurors who believed that the defendant acted "reasonably" when he lost control of his emotions and killed the victim downgraded the charge from murder to manslaughter. Overall, the gay-panic defense did not affect jurors' punitiveness --had we not taken into account jurors' political orientation, it would have appeared that the gay panic aspect of the provocation defense had no impact. Yet, when we factored in potential biases stemming from jurors' political orientation, we found that the defense was effective for conservative jurors but ineffective for liberal jurors. Conservative jurors were less morally outraged toward a defendant who killed in response to a same-gender sexual advance rather than an otherwise identical provocation, which in turn, made them more lenient.
The type of provocation defense, in contrast, did not affect liberal jurors, because the samegender sexual advance did not reduce their moral outrage toward the defendant. Next, we discuss these results' implications for legal and psychological theory, and then for law and policy.

Theoretical Implications
Subjectivity surrounding provocation defenses. First, we demonstrated that provocation defenses such as the gay-panic defense might encourage jurors to rely on their subjective biases when rendering their judgments. Specifically, asking jurors to judge whether a given victim provocation was sufficient to judge the defendant's lethal reaction "reasonable" provided a context in which jurors' value-based biases influenced their judgments. Provocation defenses might have effects similar to victim impact statements that lead jurors to value respectable victims more than less respectable victims (Greene, 1999;Greene et al., 1998). Both victim impact statements and provocation defenses emphasize victim characteristics and behavior-focusing jurors on the victim's same-gender sexual advance might exacerbate antigay biases in jurors' judgments, relative to cases that do not include this kind of provocation.
Moral outrage and legal judgments. Second, we identified the underlying psychological mechanism that, at least in part, explained the effect of the gay panic provocation defense on conservatives' punitiveness: decreased moral outrage toward the defendant. Moral outrage is considered a key factor that underlies retributive motives of punishment (Carlsmith & Darley, 2008). Our results are in line with other study findings highlighting the importance of moral outrage in driving punishment decisions (e.g., Carlsmith et al., 2002;Darley et al., 2000;Salerno & Peter-Hagene, 2013;Salerno et al., 2010;Wiley & Bottoms, 2009). Our data contributes to this literature by identifying a novel factor that can drive moral outrage and, in turn, punishment: the extent to which victims' actions violate the legal decision makers' politically motivated values. Victims' actions can indirectly influence moral outrage toward the defendant by making the defendant's criminal actions appear more reasonable, ultimately resulting in leniency toward the defendant. We also ruled out alternative explanations for conservatives' leniency in response to the gay panic provocation defense by demonstrating this effect was not explained by jurors' perceptions of the extent to which the victim provoked the killing or that the defendant was in danger.
Political orientation and legal judgments. Third, we identified a (surprisingly) rare effect of juror political orientation on mock juror judgments, and highlighted the importance of investigating how political orientation might moderate the effect of case evidence factors. Our findings also contribute to basic research regarding political orientation, attribution processes, and punishment. Despite political conservatism being negatively associated with punitiveness overall in our sample, conservatives were more lenient in the gay panic provocation versus typical provocation condition, which represents a rare reversal of the ideo-attribution effect. Weiner (1993) proposed that conservatives typically perceive social problems and misconduct to be caused by dispositional characteristics of the perpetrator (e.g., internal, personality-based factors), whereas liberals are more likely to perceive them to be caused by situational factors (e.g., external, social and environmental factors). Thus, conservatives tend to blame defendants' dispositional characteristics for crimes committed and are more punitive toward defendants, whereas liberals tend to take into account situational factors surrounding crimes and are less punitive toward defendants (Weiner, 1993).
This ideo-attribution effect has been well documented (e.g., Skitka & Tetlock 1992;Weiner, 1993), yet Morgan, Mullen, and Skitka (2010) found that conservatives tend to think dispositionally and liberals tend to think situationally when doing so allows them to make decisions that are consistent with their ideals and beliefs. That is, a reversal of the ideoattribution effect may occur if there is enough motivation to maintain consistency between one's salient values and behaviors to lead one to switch from one's natural attribution process to another (i.e., conservatives switch from making dispositional to situational attributions, liberals switch from making situational to dispositional attributions). Our study revealed that conservatives' punishment decisions will exhibit uncharacteristic leniency in response to situational cues (i.e., a same-gender sexual advance) if doing so is in line with their moral values (i.e., punishing violations of traditional sexuality, Skitka & Tetlock, 1993), Similarly, liberals' punishment decisions will not exhibit leniency in response to a situation cue (i.e., a same-gender sexual advance) when doing so is not in line with their moral values. This suggests that jurors' tendency to engage in dispositional versus situational attributions and resulting punitiveness will depend on whether their values are being violated by the specific case facts-even if those values are violated by the victim rather than the defendant.

Law and Policy Implications
Allowing the gay-panic defense invites jurors to rely on their pre-existing attitudes toward homosexuality in reaching a judgment with extreme consequences: whether a defendant is charged with murder or the lesser charge of manslaughter. The latter disqualifies the defendant from much more extreme punishments, including, in some states, the death penalty. Our results indicate that the gay-panic defense might exacerbate anti-gay discrimination in the legal system.
These results therefore empirically confirm the ABA's concern that the gay-panic defense will result in the victim's same-gender sexual behavior being used as a justification for murder (Nichols, 2013). In 2006, California passed the first law requiring jury instructions telling jurors to not rely on bias based on the victim's sexual orientation when a provocation defense is offered (Egelko, 2013). Our results also support the need for these type of jury instructions in cases that do allow the gay-panic defense-particularly given that other research has highlighted the potential for this type of jury instruction to ameliorate some biases (Kraus & Ragatz, 2011).
Some legal scholars have argued that the gay-panic defense should be allowed (as long as there is some evidence to support the defense), not because they endorse recognition of the defense, but because the defense might be more effectively combatted out in the open (Lee, 2008;Perkiss, 2013). That is, given social psychological evidence about implicit bias, gay-panic defenses might be best neutralized by proactively (a) identifying and excluding jurors with antigay attitudes during voir dire, (b) making anti-gay bias salient during trial by either encouraging jurors to ask themselves whether they would accept the same provocation defense if a gay male defendant killed a heterosexual female victim who made a sexual advance or judicial instructions to avoid anti-gay bias (Lee, 2008), or (c) offering scientific evidence that experiencing gay panic does not result in violent behavior to invalidate the defense (Perkiss, 2013). We have provided timely and direct empirical evidence that adding a same-gender sexual advance to a provocation defense will be effective for politically conservative jurors whose moral values are violated by same-gender sexual behavior and, therefore, that these kinds of interventions are needed to combat anti-gay bias affecting jurors' judgments in these cases. These results also have implications for jury selection, as they suggest that prosecuting attorneys might wish to keep conservatives off of the jury when a gay-panic defense is being presented. This research question also raises an interesting normative question for the legal system to consider: What is the legal system's goal in asking jurors to judge whether they believe a given provocation has the potential to create such emotion that a reasonable person could become enraged enough to lose control and kill? Jurors are asked if they believe the defendant's reaction was reasonable, which implies that-even if the law does not explicitly ask them to do so-they might be putting themselves in the defendant's shoes and asking themselves, "What would I have done?" Is it unreasonable, then, for jurors who are morally opposed to same-gender sexual behavior to judge a defendant offering a gay-panic defense as less culpable given that they might envision themselves reacting in the same manner? Are there some bases for these "reasonableness" judgments that the legal system considers to be unacceptable? Does the legal system condone, for example, anti-gay bias when judging whether a defendant's actions were a reasonable response to provocation? Given that our data suggests that anti-gay discrimination does influence these judgments, we believe it is time for the legal system to face the inappropriateness of allowing this defense.

Strengths, Limitations and Future Research
We used a clean, tightly controlled experimental design that allowed for clear cause-andeffect conclusions about a topic of societal importance and ruled out alternative explanations.
Our case was modeled after a real case and our participants were precisely those who would be called for jury duty in our city. Finally, our work was driven by solid theory in several areas of social psychology. Yet of course, our work is not without limitations. No mock jury study will ever replicate an actual jury situation. Our study is vulnerable to typical criticisms of mock jury studies, including a lack of jury deliberation and ecologically valid trial stimuli (e.g., Diamond, 1997). Our methodological decision to use a more ecologically valid sample (i.e., community members rather than students) affected choices about the length of our stimuli and questionnaire.
Further, we did not include more detail about the defense arguments because our theoretical goal was to test the effectiveness of the basic argument inherent in gay-panic defenses because (a) we did not want our effects to be due to idiosyncratic details we chose for the scenario, and (b) we wanted to provide more of a "blank canvas" on which jurors might project their interpretations of the situation based on their subjective biases. Although it would be interesting to investigate the influence of various details surrounding the provocation in future research (i.e., what the victim said specifically, the victim's motivation for starting the fight, etc.), these details were not central to our main research questions. One could also argue that our manipulation included both a manipulation of the victim's behavior (i.e., a same-gender sexual advance) and the gay-panic defense (i.e., a lawyer arguing that the victim's behavior constituted justifiable provocation).
Yet, it would be artificial, even nonsensical, to manipulate them independently, and in some conditions have one without the other (e.g., a condition in which a lawyer argues a gay panic provocation defense without there being a same-gender sexual advance). Future research could, however, investigate whether the same-gender sexual advance is sufficient or whether it has to be used as a defense explicitly. Finally, our conclusions are limited to a gay-panic defense that involved two men, as opposed to women. We would argue, however, that other researchers have already demonstrated that reactions to lesbians as opposed to gay men differ in a gay panic scenario (Kraus & Ragatz, 2011).
We believe that this study constitutes "Stage 1 research" status (Diamond, 1997), given that it is one of the first to investigate the impact of a provocation defense that includes the gay panic component and how political orientation leads jurors to make judgments consistent with their values in the context of such defenses. We hope that this demonstration will inspire future research that investigates boundary conditions of the effects revealed herein, and the effect of other factors that affect cases involving provocation defenses or gay issues more generally, such as negative attitudes toward homosexuality (Wiley & Bottoms, 2013), intersectional issues of sexual orientation and gender, the role of emotion (particularly given that moral outrage has been conceived of as a combination of anger and disgust, Salerno & Peter-Hagene, 2013), dehumanization of the victim, etc.

Conclusion
Provocation defenses invite jurors to decide subjectively whether a defendant's lethal reaction to a victim's behavior was reasonable. The current research has made a theoretical contribution to the literature regarding provocation defenses by demonstrating that jurors' political orientation can reduce jurors' moral outrage toward a defendant who kills a victim whose behavior violates the jurors' values and, ultimately, the punitiveness of their judgments.
The current research has made an applied contribution to the literature by demonstrating, specifically, that conservative jurors-but not liberals-are less morally outraged by a defendant who offers a gay-panic defense relative to an otherwise identical provocation defense that does not include a same-gender sexual advance, ultimately making them more likely to downgrade the defendant's charge from murder to manslaughter. This research highlights the need for further research about other types of biases that provocation defenses might induce jurors to rely upon when deciding punishment. Although there is a large body of research demonstrating racial biases in the legal system, the research demonstrating biases against lesbian, gay, bisexual, transgender, and queer individuals is sparse. This study demonstrates the further need for investigations of the many forms anti-gay discrimination can take in the legal system.

Footnotes
1 Because prior research has demonstrated that women tend to have less negative views of homosexuality than do men (Herek, 1988)-yet are less punitive toward defendants who offer a gay-panic defense than are men (Kraus & Ragatz, 2011)-we tested whether women and men had differential reactions to a gay-panic defense. First, we conducted a logistic regression with dummy codes representing participant gender, gay panic condition, and the interaction entered as predictors of dichotomous verdicts and found that none were significant, Bs  |.96|, ps  .16.
Second, we conducted a logistic regression with the participant gender dummy code, centered political orientation, and the interaction term entered as predictors of dichotomous verdicts and found that none were significant, Bs  |.51|, ps  .33. Third, we conducted a 2 (Provocation defense: Non-gay panic, Gay panic) X 2 (Participant gender) between-subjects analysis of variance on continuous verdict confidence, which revealed that none of the effects were significant, F(1, 68)s  1.57, ps  .21. Finally, we conducted a linear regression testing whether participant gender, participant political orientation, or the interaction between the two predicted continuous verdict confidence and found that none of the effects were significant, all Bs  2.53, all ps  .25. We did not have enough participants to feel confident we could detect a three-way interaction between provocation defense, participant gender, and participant political orientation.
Thus, because participant gender had no main effects or interactive effects with gay panic condition or political orientation on dichotomous verdicts or verdict confidence we felt confident it was not an important moderator and dropped participant gender as a variable and will not discuss this variable further. 2 We find the same results when we run a similar moderated mediation model with moral outrage toward the defendant as the only mediator (i.e., omitting the measure of moral outrage toward the victim). We again find that the conditional indirect effect of provocation type on verdict confidence is significant for conservative but not liberal jurors.  Juror political orientation is represented as 1 SD below the mean (i.e., liberals) and 1 SD above the mean (conservatives). The effect of provocation type was significant for conservative but not liberal jurors.