Anti-Semitism and opposition to Israeli government policies: the roles of prejudice and information

ABSTRACT Past research on the link between anti-Semitism and criticism of the Israeli government has revealed a partial correlation between the two. However, among the political arguments that have inspired such research, one has gone unaddressed: that criticism of the Israeli government along with silence about more egregious human rights abuses committed elsewhere can only arise from anti-Semitism. This article investigates the link between knowledge of conflicts around the world, anti-Semitism, Islamophobia, and criticism of the Israeli and Saudi governments, among US participants. Using a deeper index of anti-Israel opinions than prior research, the results indicate that some criticisms of the Israeli government are more reliable indicators of anti-Semitic sentiment than others. The findings are discussed with regard to how legitimate political criticism can be distinguished from prejudice in the context of the Israeli–Palestinian conflict.

There are only two explanations for this moral anomaly. … One is the nearly worldwide embrace of leftist thought and values. According to this way of thinking, Westerners are almost always wrong when they fight Third World countries or groups; and the weaker party, especially if non-Western, is almost always deemed the victim when fighting a stronger, especially Western, group or country … . The only other possible explanation is that Israel is Jewish.
Yet another possible explanation exists: that the disproportionate critical focus on the Israeli government, to the exclusion of other governments guilty of worse crimes against weaker, "Third World countries or groups", arises from simple ignorance. 1 After all, the media's agenda powerfully influences perceptions of foreign countries (Wanta, Golan, and Lee 2004;Zhang and Meadows 2012), and media outlets in the US and several other countries provide relatively extensive coverage of the Israeli-Palestinian conflict (Segev 2015) largely ignoring objectively worse conflicts, particularly in Africa (Harvey 2012)although this disproportionate coverage has led to relatively "warm" sentiments toward Israel in the US (Besova and Cooley 2009;cf. Anholt Nation Brands Index 2006).
In forming political opinions, knowledge is of inestimable importance (Friedman 2005). Differing amounts of knowledge profoundly affect collective preferences (Althaus 1998), voting (Bartels 1996), support for war (Sirin 2011), and levels of political participation, tolerance, opinion consistency, interestattitude congruence, and attitude-participation congruence (Carpini and Keeter 1997). Widely shared understandings (social representations) concerning a group's or a country's history influence collective identity and how people will react to new political challenges (Liu and Hilton 2005). Therefore, the "moral anomaly" of disproportionate criticism of Israel that Prager observes may be due to one, or a combination of, three factors: an embrace of leftist thought and values, which favours the weaker Palestinians over the stronger Israelis; anti-Semitism, which directs racist hatred toward the Jewish state because it is Jewish; and ignorance, a media effect resulting from ample coverage of the Israel-Palestinian conflict and a lack of coverage of other conflicts around the world.
Previous research has made important contributions to an understanding of the relationship between anti-Semitism and a host of other factors (Tausch 2014), including anti-Israeli government political opinions (Jaspal 2014, 63-65). Kaplan and Small (2006) surveyed a large sample of Europeans, finding a correlation between anti-Semitism and anti-Israeli government political opinions. Using the Anti-Defamation League's (ADL) anti-Semitism scale and a four-item anti-Israel scale, they found that as agreement with each of the four anti-Israel statements increased, so too did the percentage of those qualifying as anti-Semitic increase from twelve, twenty-two, thirty-five, to fifty-six per cent, respectively. However, while the ADL's anti-Semitism scale has eleven items, Kaplan and Small's anti-Israel index had only four items, making it a somewhat coarse measure of anti-Israeli government political opinions. Furthermore, one of the four items asked whether "the Israeli Defense Forces intentionally target Palestinian civilians"which could be interpreted in at least two quite different ways. One way is that individual Israel Defense Forces (IDF) soldiers have intentionally targeted civilians in violation of IDF policy, whichgiven the history of all armed conflicts involving all nationalitieswould be anomalous if it were not true. Another interpretation is that it is the policy of the IDF to target civilians, which given the damage to Israel's reputation it would cause, calls to mind the quip attributed to Tallyrand: it would be worse than a crimeit would be a blunder. The present research used Kaplan and Small's four anti-Israel items, but added eight more, and clarified the civilian-targeting item by adding "as a matter of policy" to the text (see Appendix 1).
A similar study of Canadian respondents likewise found a partial correlation between anti-Semitism and anti-Israeli government political opinions (Baum and Nakazawa 2007). The authors noted that this correlation does not imply that criticism of the Israeli government is anti-Semitic, but suggested that such criticism is more likely to be motivated by anti-Jewish prejudice when it employs traditional anti-Semitic canards or makes radical calls like the dismantling of the State of Israel (see also Kempf 2015). To investigate this suggestion, the present research used a twelve-item anti-Israel index: six items were designed to match the kind of criticisms one can hear on the Israeli Left ("moderate"), and the other six items were designed to evoke traditional anti-Semitic canards or make radical calls ("extreme").
Echoing warnings about the error of conflating anti-Semitism with criticism of the Israeli government (Klug 2003(Klug , 2013, and noting the limitations of Kaplan and Small's methodology, Kempf (2012) delved deeper into the motivations behind anti-Israel/pro-Palestinian opinions among German and Austrian samples. Kempf asked not only about anti-Semitism and anti-Israel political opinions, but also about orientations toward pacifism, human rights, and morality. He found that the majority who took a pro-Palestinian stance were not anti-Semitic; and that within this non-prejudiced majority, the greater their pacifism, rejection of limitations on human rights, concern for victims of human rights violations, moral engagement, and knowledge about the conflict, the stronger their pro-Palestinian position. Contrariwise, the anti-Semitic critics of the Israeli government evinced the opposite pattern: the stronger their pro-Palestinian position, the lesser their pacifism, consistency of human rights orientation, concern for victims of human rights violation, and knowledge about the conflict. Baum (2009) found other correlates of anti-Semitism among Christian and Muslim respondents, including personal and social identity threat, right-wing authoritarianism, neuroticism, and psychoticism. Exploring some of the psychological dynamics behind anti-Semitism, Cohen (2012) used mortality salience (from terror management theory, which predicts that the feeling of existential anxiety prompts greater affirmation of one's worldview) and bogus pipeline (to convince participants that any prejudice they attempted to hide would be detected by psychological techniques; see Roese and Jamieson 1993) manipulations, and found that those reminded of their mortality perceived greater justification for cartoons critical of Israelbut not China. A series of experiments using the same mortality salience and bogus pipeline manipulations found the same mortality salience effects on expressed anti-Semitism, that anti-Semitism evokes hostility to Israel but that hostility to Israel can exist in the absence of anti-Semitism, and that hostility to Israel can in fact lead to anti-Semitism (Cohen et al. 2009). To avoid the effects of social desirability on responses, the present study used the same bogus pipeline manipulation, and applied mortality salience and control treatments to replicate Cohen et al.'s results. Other research has focused on the similarities and differences between anti-Semitism and other forms of prejudice. Zick et al. (2008) provided empirical support for the "group-focused enmity" syndrome, in which all out-groups are devalued. Anti-Semites within their German sample were found to also be negatively disposed to homosexuals, women, homeless people, foreigners, and Muslims, suggesting that underlying anti-Semitism is a general dislike of all out-groups (see also Frindte, Wammetsberger, and Wettig 2005). Similarly, a study of college students in the US found that anti-Semitism correlated with general prejudice and racism (Dunbar 1995). In a study of Dutch secondary school students, Islamophobia was found to be separate from and more prevalent than anti-Semitism, butin a nod to the etymology of the term anti-Semitism was the strongest predictor of negative attitudes to Muslims and Islam, and vice versa (Dekker and van der Noll 2011). For comparison, the present research also included a measure of Islamophobia ) and anti-Saudi Arabia 2 opinions modelled as closely as possible to the statements of anti-Israel opinions.
The present study aims to test the following hypotheses within a sample of internet users in the US: H1, anti-Semitism and anti-Israeli government opinions will be moderately-to-weakly correlated overall, but more weakly correlated with the six "moderate" anti-Israel statements and more strongly correlated with the six "extreme" anti-Israel statements. H2: Kaplan and Small's results will be replicated, in that greater anti-Semitism will predict greater criticism of Israel; but this relationship will be stronger among extreme criticisms, and weaker among moderate criticisms. H3: participants in the mortality salience condition will report greater levels of anti-Semitism and Islamophobia than controls. H4: levels of knowledge for the Israeli-Palestinian conflict will be higher than six other conflicts included in the survey, without regard to level of severity. H5: extent of anti-Israel (anti-Saudi) criticism will be predicted by greater pacifism, human rights orientation, leftwing ideology, a tendency to blame the more powerful side in a conflict, knowledge of conflicts around the world, and anti-Semitism (Islamophobia), and (negatively) by more personal contact with Jews and Israelis (Muslims and Saudis), knowledge of Israel, and right-wing ideology. H6: Extent of anti-Semitism (Islamophobia) will be predicted by right-wing ideology, Islamophobia (anti-Semitism), and anti-Israel opinionsalthough less by moderate than by extreme opinionsand negatively predicted by pacifism, human rights orientation, leftwing ideology, knowledge of conflicts around the world and Israel itself, and personal contact with Jews and Israelis.

Participants
An M-Turk sample of US residents was collected for the survey (N = 951, after eliminating participants who failed the attention check). This was a nonrepresentative convenience sample, more diverse than other internet and college student samples, although its subject pool tends to be younger and more liberal than the general US population, and includes a disproportionately large percentage of whites (Berinsky, Huber, and Lenz 2012). Fifty-four per cent were female; seventy-three per cent were white, five per cent were black, five per cent were Hispanic, five per cent were Asian-American, two per cent were Jewish, one per cent were Arab, and seven per cent stated Other; and their ages ranged from 18 to 76 (M = 36.2 years, SD = 12.49 years). Forty-nine per cent were Christian, thirty-four per cent were atheist or agnostic, 1.5 per cent were Muslim, two per cent were Jewish, two per cent were Buddhist, and eleven per cent stated Other. Asked to place their level of education on a 1-6 scale from "some high school" to "completed graduate school", participants' mean (and median and mode) was 4 (completed college). Asked to place themselves on a 0 Left/Liberal to 100 Right/ Conservative scale, the average placement was 40.72 (SD = 27.9).

Materials
The present study used the ADL's anti-Semitism index and Lee et al.'s (2009) Islamophobia scale, 3 and a twelve-question anti-Israel opinion index using Kaplan and Small's (2006) four statements plus eight additional, six of which reflected extreme positions, and six of which reflected criticisms that can be heard from the Israeli Left. An anti-Saudi government scale was designed with items that were mirror images of the anti-Israeli government items (see Appendix 1).
To test for knowledge of post-Second World War conflicts around the world, participants were asked four simple fill-in-the-blank questions about Israel/Palestine, Colombia, Kashmir, Sri Lanka, Chechnya, Ireland, and the Congo: to name what part of the planet they are located in, one city within them, two of the main combatants involved, and one fact about the conflict. These were scored with a high degree of generosity: for example, credit was given for any geographically correct placement (such as "Western Hemisphere" for Colombia), any city located within the seven states or regions, any partially correct listing of combatants (such as "Tamils and the government" for Sri Lanka), and any arguably correct listing of facts about the conflicts (such as "Muslim Pakistan hates India" for Kashmir).

Procedures
Participants were recruited for a survey on "conflicts around the globe". First, they received a bogus pipeline treatment, modelled after Cohen et al.'s (2009) technique: giving participants an example of the "many tools psychologists use to detect people who lie to create a positive impression of themselves", and warning them that the survey was peppered with such techniques that would expose dishonesty. Next, they were randomly assigned to either a mortality salience or control condition, and wrote down a few sentences describing how they would feel. They were then told they would be asked to answer questions about two randomly selected groups of combatants embroiled in a current conflict (in fact, they were asked only about Jews and Muslims, Israel and Saudi Arabia, in randomized order, using the instruments described above). Participants were also asked to rate the favourability of media coverage of, extent of personal contact with, and valence of experience with Jews and Muslims.
Next, they were asked to place blame for the conflicts using a sliding scale, to rank the conflicts in order of severity, and to note which of the conflicts they remembered learning about from the media. Afterward, they were given White's (2013) estimates of death tolls for the seven conflicts since 1940, and asked whether they would change their severity rankings as a result. Finally, participants were asked to report their level of pacifism, and three questions on their human rights orientation: how important they are, to what extent they may be subordinated to other considerations, and how concerned they are about victims of human rights abuses. Finally, standard demographic questions including ideological self-placement were asked. Where possible, all questions and their answer options were presented in random order.

Results
The correlation between anti-Semitism and anti-Israeli government opinions is positive but weak (r = .266, p = .000), and marginally stronger than that between Islamophobia and anti-Saudi government opinions (r = .242, p = .000). In fact, all four are significantly but weakly correlated: Islamophobia with anti-Semitism (r = .255, p = .000), and anti-Israeli government opinions with anti-Saudi opinions (r = .254, p = .000) and (negatively) with Islamophobia (r = −.211, p = .000). When anti-Israel and anti-Saudi statements are split into moderate and extreme halves, anti-Semitism is only weakly correlated with scores on the moderate anti-Israel index (r = .152, p = .000), but the correlation with scores on the extreme anti-Israel index is over twice as strong (r = .362, p = .000). Islamophobia provides a mirror image: a weak correlation with scores on the moderate anti-Saudi index (r = .126, p = .000), but a correlation over twice as strong with the extreme anti-Saudi index (r = .327, p = .000). While Islamophobia is negatively correlated overall with the full anti-Israel index, they do not correlate at all with the extreme half, only with the moderate statements (r = −.313, p = .000). That is, being more Islamophobic is associated with disagreement with moderate criticisms of the Israeli government, but not with extreme criticisms. Anti-Semitism, on the other hand, is correlated with agreement with extreme criticisms of the Saudi government (r = .223, p = .000), but not with moderate criticisms.
Consonant with prior research, participants' number of anti-Israeli government opinions is correlated with self-reported pacifism (r = .304, p = .000) and human rights orientation (r = .191, p = .000). The number of anti-Saudi government opinions is only weakly correlated with human rights orientation (r = .081, p = .011). Splitting the anti-Israeli and anti-Saudi opinions into extreme and moderate reveals some interesting differences. Agreement with moderate anti-Israel opinions is correlated with pacifism (r = .355, p = .000) and human rights orientation (r = .270, p = .000); agreement with extreme anti-Israel opinions is correlated more weakly with pacifism (r = .133, p = .000), not at all with human rights orientation, and negatively with both knowledge of Israel (r = −.219, p = .000) and knowledge of the other six conflict areas in this study (r = −.182, p = .000). Agreement with moderate anti-Saudi opinions is correlated with human rights orientation (r = .174, p = .000) and knowledge of conflicts (r = .067, p = .039); agreement with extreme anti-Saudi opinions is not correlated with human rights orientation, and correlates negatively with pacifism (r = −.068, p = .035) and knowledge of conflicts (r = −.189, p = .000).
Replicating Kaplan and Small (2006), the survivor distribution of the anti-Semitism (and Islamophobia) index for each of the thirteen levels of the anti-Israel (and anti-Saudi Arabia) index were examined, and separately for the seven levels of the moderate and extreme anti-Israel (and anti-Saudi) indexes. (Survivor distributions are often used to illustrate varying levels of risk along a dimension, like risk of death by age or risk of cancer by tobacco use.) The thirteen curves are significantly different (Wilcoxon Gehan statistic of 86.02, df = 12, p = .000), confirming that measured anti-Semitism differs by the extent of anti-Israel sentiment. This result holds for both sets of extreme (Wilcoxon Gehan statistic of 85.06, df = 6, p = .000) and moderate (Wilcoxon Gehan statistic of 25.80, df = 6, p = .000) anti-Israel curvesthough among the moderate anti-Israel statements, the curves are significantly closer. Contrary to Kaplan and Small's results, these thirteen survivor curves do cross: for some values x of the anti-Semitic index, the fraction of respondents who agree with more than x anti-Semitic statements does not strictly increase with the value of the anti-Israel index. Nonetheless, a substantially similar pattern is apparent. Likewise, the thirteen anti-Saudi opinion curves are significantly different (Wilcoxon Gehan statistic of 75.95, df = 12, p = .000), confirming that measured Islamophobia differs by the extent of anti-Saudi sentiment. So too for both sets of extreme (Wilcoxon Gehan statistic of 97.06, df = 6, p = .000) and moderate (Wilcoxon Gehan statistic of 16.57, df = 6, p = .011) anti-Saudi curvesand again, among the moderate anti-Saudi statements, the curves are significantly closer than those of the extreme anti-Saudi statements. Figure 1 reports the percentage (and total number) of respondents who agree with more than five of the twelve anti-Semitic statements for the twelve different levels of the anti-Israel index. (This is a slightly more sensitive anti-Semitism index, as it has one additional item but the same threshold.) Using the ADL's method of measurement, overall fourteen per cent of participants hold anti-Semitic viewsthe same proportion as Kaplan and Small found in their European sampleand twenty-four per cent agreed with more than five of the Islamophobia items. Just as Kaplan and Small found, only nine per cent of those with anti-Israel index scores of zero hold anti-Semitic viewsthough unlike the steady stepwise increase they found with a four-item anti-Israel scale, here the percentage of participants holding anti-Semitic views stays within a range of 5.9-12.2 per cent through the anti-Israel index value of 7 (which would have to include agreement with at least one "extreme" statement). From eight, as more of the extreme statements are included, the percentage increases respectively to eighteen, thirty-four, fifty-six, and remains at seventy-five for those who agreed with all or all but one of the anti-Israel statements. Slightly more than half of participants agreed with only four or fewer anti-Israel statements, and five or fewer anti-Saudi statements. Applying the ADL cutoff point to the Islamophobia scale, sixteen per cent of those with anti-Saudi index scores of zero hold Islamophobic views, and the percentage of participants holding Islamophobic views stays within a range of 14.5-19.8 per cent through the anti-Saudi index value of 6 (which could possibly include only the six moderate statements). From seven, as more of the extreme statements are included, the percentage increases respectively to twenty-five, forty-two, thirty-six, fifty-seven, seventytwo, and 100. Cohen et al. (2009) found that those exposed to a bogus pipeline treatment along with a mortality salience manipulation expressed agreement with significantly more anti-Semitic statements than controls. However, the present results display no significant differences in mean levels of anti-Semitism, Islamophobia, and anti-Israel or anti-Saudi opinions. Those who received the mortality salience manipulation were slightly more Islamophobic (3.09 vs. 3.00) and held slightly more anti-Saudi opinions (5.40 vs. 5.12) than those in the control groups; while those reminded of their own mortality actually agreed with fewer anti-Semitic (2.47 vs. 2.56) and anti-Israel (4.24 vs. 4.41) statements than those in the control groups. However, these mean differences were all non-significant. Possibly, the effect of the mortality salience manipulation wore off after answering the anti-Semitism (or Islamophobia) index first, according to random assignment. However, including only those who received the anti-Semitism index first, scores on that instrument averaged 2.64 for controls and 2.46 in the mortality salience condition; for those who received the Islamophobia index first, scores on that instrument averaged 2.93 for controls and 3.23 in the mortality salience conditioncomparable differences without statistical significance.
Participants' knowledge of recent conflicts in Israel/Palestine, Colombia, Sri Lanka, Kashmir, Chechnya, Ireland, and the Congo reflected what media effects researchers would predict: that the areas receiving the most media coverage (for reasons of cultural and geographic proximity, trade ties, etc.) were significantly better known than the less-covered conflicts. Asked for a total of four of the most rudimentary pieces of information on each conflict, participants were on average able to offer 3.4 facts on Israel/Palestine, 2.4 on Colombia, 0.94 on Sri Lanka, 0.89 on Kashmir, 1.2 on Chechnya, 2.8 on Ireland, and 1.5 on the Congo. (All t-tests on mean differences between knowledge of the Israeli-Palestinian conflict and the six other conflicts were significant at the p < .001 level.) Furthermore, knowledge of each conflict was significantly correlated (at the p < .001 level) with whether participants reported learning about the conflict in the media: correlations ranged from r = .197 for Ireland on the low end, to r = .506 for Chechnya. Ninety-four per cent of participants reported hearing about the Israeli-Palestinian conflict in the media, forty-one per cent about Colombia, nineteen per cent about Sri Lanka, seventeen per cent about Kashmir, forty-seven per cent about Chechnya, fifty per cent about Ireland, and thirty-five per cent about the Congo.
To investigate the relative contribution of these several factors on the number of anti-Israeli government opinions held by the participants, a series of multiple regression models were designed. The first included only demographic variables (including dummy variables for female, white, and atheists/agnostics): being older, white, and having more education and income predicted holding fewer anti-Israel opinions, while being female and non-religious predicted holding more anti-Israel opinions. The second model added measures of knowledge for the other six conflicts and for Israel: knowing more about Israel predicted fewer anti-Israel opinions, while knowing more about other conflicts around the world predicted more anti-Israel opinions. The third model added scores on the anti-Semitism and Islamophobia indexes: as expected, higher anti-Semitism scores predicted more anti-Israel opinions, but higher Islamophobia scores predicted fewer anti-Israel opinions (meanwhile, race and age faded into insignificance). The fourth model added anti-Saudi opinions (as a proxy for group-focused enmity), perception of media portrayals of Jews (from negative to positive), number of Israeli citizens personally known, and frequency of contact with Jews by valence: having greater (and more positive) contact with Jews and more Israelis personally known predicted fewer anti-Israel opinions, while more anti-Saudi opinions and a more positive perception of media portrayals of Jews predicted more anti-Israel opinions. Lastly, the fifth model added selfreported ideology (Left-Right), levels of pacifism, and human rights orientation, along with a "Blaming the Powerful" 4 composite variable: leftwing ideology, higher levels of pacifism, and a propensity to blame the more powerful side in conflicts around the world predicted more anti-Israel opinions (human rights orientation reached significance only when Left-Right ideology was excluded). While demographic variables were strong predictors in the simpler models, as more variables are added all demographic variables lost significance except for non-religiosity.
Splitting the anti-Israel index into "moderate" and "extreme" halves reveals some interesting differences (see Appendices 2 and 3). Unlike in the regressions for the full index and the moderate half, women and the non-religious were not more likely to agree with the extreme statements in all but one of the five models. Education and income are non-significant in the fuller models of moderate opinions, but they retain significance throughout all five models of extreme opinions, predicting less support. Likewise, greater knowledge of conflicts around the world was a significant predictor of opinions critical of the Israeli government in most of the models for the full index and all models for the moderate index; but it was non-significant in all four models for the extreme index. Meanwhile, knowledge of Israel was a predictor of fewer anti-Israel opinions in the full and extreme indexes, but among moderate anti-Israel opinions, it was non-significant in all models. Leftwing political ideology was a significant predictor of holding both extreme and moderate anti-Israeli government opinions; but among moderate opinions, it is over twice as strong. Similarly, while a tendency to place blame on the more powerful side of conflicts predicts more anti-Israel opinions throughout, pacifism and human rights orientation are significant predictors only of agreement with the moderate opinions. Lastly, anti-Semitism predicts anti-Israel opinions across the board, but it is stronger by over half among the extreme opinions.
Using the same variables (or their equivalents, minus knowledge of Saudi Arabia which was not asked) to predict anti-Saudi opinions is far less successful: while the full model for anti-Israel opinions explains nearly forty per cent of the variance in responses, the equivalent model for anti-Saudi opinions explains less than fifteen per cent of variance. Besides age, income, and educationwhich retain significance in the full model as negative predictorsthe only statistically significant predictors are Islamophobia (β = .317, t(934) = 7.99, p = .000), human rights orientation (β = .126, t(934) = 3.66, p = .000), and anti-Semitism (β = .079, t(934) = 2.40, p = .017).
However, in predicting anti-Semitism and Islamophobia in Table 2 (supplementary material), the same variables explain nearly half of the variance in Islamophobia scores, but less than a third of the variance in anti-Semitism scores. Being older predicts less anti-Semitism, but more Islamophobia; greater income predicts more anti-Semitism while being white and female predicts less, while these two demographic variables are non-significant in predicting Islamophobia. Being non-religious predicts less anti-Semitism, but has no relationship with Islamophobia. Knowledge of global conflicts predicts neither anti-Semitism nor Islamophobia, but knowledge of Israel actually predicts slightly greater anti-Semitism. Having frequent and positive contact with Jews and Muslims predicts lesser anti-Semitism and Islamophobia, respectively, but knowing more Israeli citizens counterintuitively predicts slightly greater anti-Semitism (while knowing more Saudi citizens does not predict Islamophobia). Interestingly, leftwing ideology, pacifism, and a tendency to blame the more powerful side in a conflict predict less Islamophobia, but are non-significant in predicting anti-Semitism; conversely, human rights orientation predicts less anti-Semitism, but is non-significant in predicting Islamophobia.
One striking area of similarity between the two forms of prejudice is that extreme anti-Israel and anti-Saudi opinions are each over twice as strong in predicting anti-Semitism and Islamophobia, respectively, than their moderate counterparts. A perception that the media portrays Jews and Muslims favourably also predicts greater anti-Semitism and Islamophobia, respectively.

Discussion
The weak, positive correlations between anti-Semitism, anti-Israeli government opinions, and anti-Saudi government opinions reinforce prior research suggesting that prejudice against out-groups is a generalized disposition; as John Martilla (quoted in Levitt 2013) observed, "Haters hate; they don't just hate Jews, they hate a whole range of cohorts". However, while Islamophobia was positively correlated with anti-Saudi government opinions and anti-Semitism, it was negatively correlated with anti-Israeli government opinions, suggesting that some may be less likely to criticize the Israeli government due to prejudice toward Muslims (or, conversely, that those who do not fear/ dislike Muslims may feel freer to criticize the Israeli government). One possible explanation is that those with higher levels of Islamophobia view the Israeli government as a bulwark against the perceived threat posed by Muslims, hence less deserving of criticism. H1 receives support. Kaplan and Small's (2006) results for a European sample were largely replicated among this US sample: anti-Semites are more likely to hold anti-Israel opinions than those without prejudice against Jews, and the extent of anti-Israel opinions differentially predicts the likelihood that one holds anti-Semitic views. However, by enlarging the anti-Israel index to comprise equal numbers of "extreme" and "moderate" criticisms, striking differences emerge. Among moderate criticisms, the survival curves are more L-shaped and closer together, while the survival curves for extreme criticisms are further apart and differently shaped. Furthermore, only around ten per cent of those who agreed with seven or fewer of the twelve anti-Israel statements were anti-Semitic by the ADL's criteria; anti-Semites comprised only a third of participants agreeing with nine of twelve statements, but over half to threequarters of the (very few) participants agreeing with more than nine statements. This suggests that unprejudiced critics of the Israeli government can be differentiated from anti-Semites by paying attention to the nature of their criticisms: exaggerations (e.g. "Israel today is the modern moral equivalent of Nazi Germany"), radical calls (e.g. "the state of Israel no longer deserves to exist"), and implying that all Jews are the same (e.g. "to recognize the true face of the Jews, we need only see how they treat the Palestinians") are strong indications that criticism of the Israeli government is coming not from principled opposition to the domination of the weak by the strong, but from hatred or fear of an out-group.
A similar pattern is apparent in the survivor curves for Islamophobia and anti-Saudi government opinions. The moderate criticism curves were closer together and more L-shaped, while the extreme curves were further apart and differently shaped. Fewer than twenty per cent of those who agreed with six or fewer of the twelve anti-Saudi statements agreed with six or more of the Islamophobia items; putative Islamophobes comprised only a third of participants agreeing with nine of twelve statements, but over half to all of the (very few) participants agreeing with more than nine statements. The common pattern across anti-Semitic/anti-Israel and Islamophobic/anti-Saudi results provides some support to the "group-focused enmity" syndrome, suggesting that hatred toward one out-group may not necessarily arise from antipathy unique to that out-group, but from a generalized dislike or distrust of "the Other". H2 receives support.
It is not immediately apparent why the present study would find no significant differences between mortality salience and control groups. One possible explanation is that through repeated exposure to mortality salience manipulations, the M-Turk sample may have become inured to it as they have to some other research techniques (Chandler, Mueller, and Paolacci 2014). However, recent studies have successfully used mortality salience manipulations with M-Turk participants (e.g. Hohman and Hogg 2015). Hohman and Hogg's finding that enhancing self-esteem can negate the group-protective impact of mortality salience suggests the possibility that the present study may have inadvertently enhanced participants' self-esteem by implying a positive comparison between the peace enjoyed by Americans and the violence suffered by the groups of foreigners this study asked about. Another possibility is that the greater Israeli-Palestinian violence at the time of Cohen et al.'s (2009) study compared to the present study (2015-16) affected participants' responses. If that were the case, however, it would be expected that the mortality salience manipulation would at least affect scores on the Islamophobia indexbut no such effect was apparent. Lastly, Cohen et al. used a longer, twenty-three-item anti-Semitism scale which included relatively subtle attitudinal items; it is possible that mortality salience may have made participants more likely to agree with these, but not the statements in the ADL index. H3 does not receive support.
Participants' knowledge of the conflicts covered in this study reflected patterns of mass media coverage. Nearly all recalled hearing about the Israeli-Palestinian conflict, but half or fewer recalled hearing about the six other conflicts. Less than one-fifth heard about Sri Lanka and Kashmir, despite the fact that the conflict in Sri Lanka has claimed more lives over the same time period than the Israeli-Palestinian conflict. Barely more than a third recalled hearing about the conflict in the Congo, despite its being the bloodiest conflict in recent history. These differences were reflected in the number of facts about the conflict areas participants could offer: from a high of 3.4 (of 4) facts on average about Israel/Palestine, down to only 0.89 for Kashmir, even though the Kashmir conflict shares many of the same features (military occupation of a predominantly Muslim population), and has claimed two-thirds of the lives over the same time period as the Israeli-Palestinian conflict. H4 is supported.
Overall, the regression results reveal that anti-Semitism is a strong predictor of anti-Israel opinions. Among all statements, it is the strongest single predictor; but among moderate statements, it is outweighed by political ideology (and even anti-Saudi opinion). Furthermore, when compared to the combined contributions of multiple other factors, it is clearly not the case that anti-Semitism is the predominant driver of anti-Israel opinions. However, among the extreme statements it does play an outsized role. Likewise, in predicting anti-Semitism, anti-Israel opinions are one of several statistically significant variables, and anti-Saudi opinions (among other variables) are a significant predictor of Islamophobia. However, extreme opinions in both cases are more than twice as strong as predictors of prejudice as their moderate counterparts.
Interestingly, the more positively participants rated media portrayals of Jews, the more likely they were to hold anti-Israeli opinions. This may seem counterintuitive at first, but these reports may be less of an objective indicator of media content and more an artefact of the hostile media effect, whereby strong partisans are more likely to perceive a hostile media bias against their own side than nonpartisans (e.g. Gunther and Chih-Yun Chia 2001). Research on the hostile media effect would predict that strong supporters of the Israeli state would perceive media portrayals of Israelis and Jews to be more negative, and that strong critics of the Israeli state would perceive the same media portrayals as being biased in a positive direction. Hence this measurement may be tapping into pro-and anti-Israel orientations rather than actual media content.
Reinforcing previous research, there are several other factors inclining individuals toward criticism of the Israeli government. Leftwing ideology, pacifism, human rights orientation, a tendency to blame the powerful, nonreligiosity, and knowledge of conflicts around the world are all associated with being critical of the Israeli governmentwhich is also associated with being critical of the Saudi government. Knowledge of Israel itself, Islamophobia, having personal contact with Israelis, and having frequent and positive personal contact with Jews are all associated with being less critical of the Israeli government. However, this association between knowledge of Israel and critical opinions disappears within the moderate anti-Israel index, suggesting another way to distinguish criticism emanating from either prejudice or principle: those with scarce knowledge of the country may be more likely to criticize it out of prejudice.
However, these regression results are only very spottily replicated in predicting anti-Saudi opinions. Age, income, and education, which are insignificant in the full Israel model, significantly predict lesser criticism of the Saudi government in its full model, while Islamophobia, anti-Semitism, and human rights orientation are the only significant predictors of greater criticism. H5 receives support regarding anti-Israel opinions, but very little support regarding anti-Saudi opinions.
In investigating the determinants of anti-Semitism and Islamophobia, these same variables provide more explanatory force for Islamophobia than anti-Semitism. In both instances, agreement with extreme criticisms of both Israel and Saudi Arabia are significantly stronger predictors of prejudice than agreement with moderate criticisms. While most of the predicted relationships between Islamophobia and the other variables were borne out by the data, fewer of these evinced a significant relation with anti-Semitism. H6 receives mixed support: more for predicted relationships with Islamophobia, but less for predicted relationships with anti-Semitism.

Conclusion
Denis Prager's conjecture that critics of the Israeli government can be driven by two motivationsanti-Semitism or leftwing ideology, in particular a concern for the weak and powerlessreceives strong support from these results. In addition, the moral anomaly he decries can also be explained by a third factor: ignorance. Even assuming all criticism of the Israeli government to be true, the human rights abuses and war crimes perpetrated by the Israeli state are quantitatively fewer than those in several other parts of the world. Within merely the limited sample and time period included in this study, more human lives have been violently taken in the Congo, Colombia, Sri Lanka, and Chechnya. Yet this being an empirical fact is irrelevant to an understanding of Prager's moral anomaly, if people do not learn of human rights abuses and war crimes in other parts of the world. Moral outrage at this anomaly may be more appropriately directed at the US media, which brings the Israeli-Palestinian conflict into living rooms throughout the US, while keeping audiences largely in the dark about conflictsincluding far graver conflictsin more distant locations, where people are more different in ethnic, cultural, linguistic, and economic terms.
Nonetheless, holding anti-Semitism constant, having above-average knowledge of conflicts around the world actually predicts more criticism of the Israeli government. Other predictors include leftwing ideology, strong pacifist and human rights orientations, a lack of religious faith, and a tendency to take the side of the weakand, being critical of the Saudi government as well (which may result from the conscious recognition of the universal applicability of political and moral principles). Regardless, knowledge of global conflicts is just one of several predictors, and this and several other studies (e.g. Carpini and Keeter 1997) have demonstrated that such knowledge is a rare commodity in the US. The moral anomaly of criticizing Israel's government while ignoring governments guilty of greater crimes may therefore be due to simple ignoranceeither on the part of (some of) Israel's critics, or on the part of those counter-critics who are unaware that many of Israel's critics are drawn by their political principles to criticize any government that violates them. That is, any government they know violates them. Given Israel's greater visibility in the US media, and low overall levels of knowledge about the rest of the world, it is unsurprising that those drawn by their political principles to criticize governments guilty of violating human rights would focus on Israel's. However, given that the US government stands charged with far more counts of the same chargesmilitary aggression and occupation, unlawful imprisonment, and torturelevelled against the Israeli government, an asyet untested means of differentiating legitimate from prejudiced criticism of Israel within the US is whether such critics reserve a proportionate amount of criticism for their own government.
Above all, these results counsel against facile accusations concerning the intentions of the Israeli government's critics. Most of those critical of Israeli policies are not anti-Semites. Only a fraction of the US population harbours anti-Semitic views (Levitt 2013), and while logically this fraction would be overrepresented among critics of Israel, the present and prior research indicate that they comprise only a small part. Inaccurate charges of anti-Semitism are not merely calumny, but threaten to debase the term itself and weaken its connection to a very real, and very dangerous, form of prejudice. In like fashion, the Israeli government's US supporters may not be driven by any laudable feeling of solidarity, sympathy, or admiration, but by Islamophobiayet it would likewise be calumnious to make facile accusations of this sort. Instead, the only way to distinguish legitimate criticism of Israel (or any other state) from hateful prejudice is to engage with the suspected arguments themselves. There are telltale signs and illustrations of prejudice to look for: gross exaggerations, moral equivalences lacking any objective basis, and treating an entire ethnic, national, or cultural group as a monolithic entity stand out. Ignorance of other governments around the world that violate human rights does not.
At the same time, critics of the Israeli government should be at pains to differentiate the actions of the Israeli government and military from the Israeli people and Jews. Critics should focus on specific actors, and entirely avoid overbroad generalizations about groups like Israelis and Jews whose members have heterogeneous opinions on Israeli government policies. One of Cohen et al.'s (2009) most troubling findings was that critical knowledge of the Israeli-Palestinian conflict can lead to anti-Semitism. It is hard to see how this could occur without an illogical conflation of the Israeli government with Israeli citizens or Jews of any citizenship. But the fact that this does seem to occur makes it all the more important for critics to be mindful of the "groupishness" of human psychology, and to prevent criticism of a state from bleeding over into criticism of an entire people. Notes 1. Other possible explanations exist; for instance, some critics argue that foreign nations receiving significant economic, diplomatic, or military support from one's own nation warrant greater scrutiny, since in principle their policies would be relatively easier to influence than those of nations not receiving such support. 2. Saudi Arabia was chosen as the closest Muslim equivalent to Israel's status as the Jewish state; while Saudi Arabia is merely a Muslim state, it is home to Islam's holiest sites and is a pilgrimage destination for millions of Muslims. 3. For ease of comparison, Lee et al.'s (2009) Islamophobia scale was shortened to twelve items by removing the last two Affective-Behavioural and Cognitive items, and by changing "professor" to "boss or supervisor". The ADL anti-Semitism index was increased to twelve items by adding "Jews exert too much influence over the global media". 4. Participants were asked to place blame for each of the seven conflicts on either side using a 0-100 sliding scale; higher scores corresponded to the more powerful side in each conflict: the Israeli government, the Colombian Right, Hindus in Kashmir, Buddhists in Sri Lanka, Russians in Chechnya, the British in Ireland, Tutsis in the Congo.

Disclosure statement
No potential conflict of interest was reported by the author.