Disentangling Islamophobia: The Differential Effects of Symbolic, Realistic, and Terroristic Threat Perceptions as Mediators Between Social Dominance Orientation and Islamophobia

The aim of this paper is threefold. First, based on ongoing theoretical discussions on the dimensionality of Islamophobia, this study analyzes whether Islamophobia empirically constitutes a one-dimensional construct or rather a multidimensional construct consisting of anti-Muslim prejudice and anti-Islam sentiment. Second, the effects of symbolic, realistic, and terroristic (safety) threats on Islamophobia were analyzed concurrently. Finally, within the framework of the revised Integrated Threat Theory (Stephan & Renfro, 2002), and in order to test the mediating effect of threats, SDO is tested as an antecedent of perceived threat and Islamophobia. Respondents from Berlin (N = 355) participated in an online survey. The results indicate that Islamophobia empirically constitutes a two-dimensional phenomenon, consisting of anti-Muslim and anti-Islam sentiment. Whereas symbolic threat is related to both types of Islamophobia, realistic threat is associated only with anti-Muslim prejudice, and terroristic threat is associated only with anti-Islam sentiment. Finally, the results indicate that the relationship between SDO and both dimensions of Islamophobia is mediated by threats. Symbolic threats mediate the relationships between SDO and both dimensions of Islamophobia. Realistic threats mediate the relationship between SDO and anti-Muslim prejudice and terroristic threats between SDO and anti-Islam sentiment.

in Germany that Islam is significantly related to threat perceptions.Nevertheless, systematic empirical studies addressing Islamophobia in Germany are scarcely found (De Nève, 2013).Yet the empirical analysis of Islamophobia is crucial for practical and theoretical purposes.In-depth knowledge of the extent, forms, and possible causes of Islamophobia is relevant for developing countermeasures for the negative consequences of an intergroup conflict.
First, in many of the studies on the subject, Islamophobia is defined and/or measured as a one-dimensional scale, not differentiating between prejudices against Muslims and anti-Islamic sentiment (cf., Kühnel & Leibold, 2007) regardless of an ongoing debate concerning the practical and theoretical implications of such an approach (Benz, 2009;Halliday, 1999;Larsson & Sander, 2015;Pfahl-Traughber, 2012;Richardson, 2012;Shooman, 2011).One of the main critiques is related to the conflation of two distinct social categories, i.e., Muslims (individuals) and Islam (religion and culture) in the definition and operationalization of Islamophobia.In response to these critical discussions, this paper sets out to test whether prejudices against Muslims and Islam constitute distinct phenomena or are indeed empirically non-distinct as implied by the concept of Islamophobia (Conway, 1997;Leibold & Kühnel, 2003).
In German public debate, Muslims and Islam are frequently contextualized as symbolic threats (i.e., threats to values, norms, and culture), realistic threats (i.e., threats to social welfare system, jobs, political power, safety and well-being), and terroristic threats (i.e., threats to physical safety and well-being) (Bielefeldt, 2010;Frindte & Haußecker, 2010;Hafez, 2010b;Kluge, 2010;Schiffer, 2005).From a theoretical perspective, the current Islamophobic tendencies in Germany set the ground for testing a revised version of the Integrated Threat Theory (ITT) (Stephan & Renfro, 2002).The ITT postulates that the more individuals perceive certain social groups as threatening (symbolic and/or realistic), the more likely they are to have prejudice against these groups.Numerous studies have confirmed this hypothesis (Riek, Mania, & Gaertner, 2006;Stephan, Diaz-Loving, & Duran, 2000;Stephan & Renfro, 2002;Stephan, Renfro, & Davis, 2009).More recent research on intergroup threats, however, suggests separating realistic threats (economic, political, and material concerns) from safety threats (physical safety and well-being), as both show empirical distinctiveness and outgroup-specific predictive capacities (Cottrell & Neuberg, 2005;Crawford, 2014).Terroristic threats (similar to safety threats) should therefore be further investigated.Most studies conducted so far in the German and international context on intergroup threat and Islamophobia (Doosje, Zimmermann, Küpper, Zick, & Meertens, 2009;Frindte & Haußecker, 2010;Kühnel & Leibold, 2007;Leibold, 2010;Pettigrew, Christ, Wagner, & Stellmacher, 2007;Pollack & Müller, 2013;Oswald, 2005;Schumann, 2010;Velasco González, Verkuyten, Weesie, & Poppe, 2008) did not analyze all types of threats simultaneously and were thus unable to determine varying effects of each type of threat on Islamophobia.Hence, a systematic approach in which all three types of threat are analyzed concurrently in conjunction with Islamophobia is recommended (Doosje et al., 2009, pp. 226-227).Furthermore, most of the studies measured the direct effect of threats on Islamophobia, but they did not test for any mediation effects of the threats on the relationship between the different antecedent factors and prejudices (see Pettigrew et al., 2007 in this regard).
In order to test the mediating effect of threats, social dominance orientation (SDO) (Sidanius & Pratto, 1999) will be tested as an antecedent of threats and the Islamophobia.Previous studies indicate that SDO is significantly related to Islamophobia (Zick & Küpper, 2006, 2007) and to symbolic threats and realistic threats (Costello & Hodson, 2011;Crowson, 2009;Duckitt, 2006) as well as to terroristic-threat-related negative attitudes towards Islam (Cohrs, Moschner, Maes, & Kielmann, 2005;Oswald, 2005).However, previous research results are ambiguous in predicting specific relationships between SDO and different types of intergroup threat (Cohrs et al., 2005;Duckitt, 2006) and hence, further investigation of SDO in conjunction with all three types of threat simultaneously is needed (Crowson, 2009).This study therefore sets out to fill in this void.More specifically, the aim of this study is to test a mediational model in which three types of threat (realistic, symbolic, and terroristic) mediate the relationship between SDO and Islamophobia.

Islamophobia -The Journey of a Vague Concept
Whereas the existence of Islamophobia has been acknowledged by academia and international institutions (Hafez, 2010a), no agreed-upon position regarding the definition of the term has been established so far (De Nève, 2013;Larsson & Sander, 2015).The two most commonly acknowledged references regarding a definition of Islamophobia in Germany (Pfahl-Traughber, 2012) have been the report by the Runnymede Trust (Conway, 1997) and the definition of Islamophobia as a syndrome of what is referred to as Group-focused Enmity (Zick et al., 2008).In both definitions, prejudiced and stereotypical negative attitudes and discriminatory behavior towards Islam and Muslims are described as the focal point of Islamophobia.
Both concepts have been criticized (Benz, 2009;Halliday, 1999;Larsson & Sander, 2015;Richardson, 2012), with the most frequent critique being that these concepts of Islamophobia conflate prejudices against Muslims with sentiment/criticism/enmity against Islam (Pfahl-Traughber, 2012;Shooman, 2011).This conflation is problematic in several ways.First, from a theoretical perspective, mixing hostile attitudes against individuals on grounds of their membership to a social category (self-attributed or ascribed) with resentment, criticism, or enmity against Islam as a religion or culture lacks theoretical and empirical support.As previous research has shown, aversion against Islam does not necessarily translate into derogative attitudes towards Muslims (Kühnel & Leibold, 2007;Leibold & Kühnel, 2006;Pfahl-Traughber, 2012).Furthermore, the term Islamophobia itself suggests primarily an irrational or exaggerated fear of Islam as a religion, but it does not refer to Muslims as the main target of hostility (Pfahl-Traughber, 2012;Shooman, 2011).Most importantly, the conflation of Muslims and Islam, in other words, mixing the social identity category Muslim with the category Islam (as culture and religion), reduces two distinct social categorizations to one and other.
Based on these ongoing discussions, in some accounts, the term Islamophobia is rejected completely; and a systematic differentiation between prejudices against Muslims and enmity against Islam (Pfahl-Traughber, 2012) as well as the use of alternative terms are suggested (e.g., for anti-Muslim racism: Shooman, 2011;for anti-Muslimism: Pfahl-Traughber, 2012).Nevertheless, this lack of an agreed-upon definition is reflected in the synonymous usage of Islamophobia as anti-Muslim or anti-Islam racism and/or anti-Islamism or anti-Muslimism (De Nève, 2013;Larsson & Sander, 2015;Pfahl-Traughber, 2012).
To bring more clarity into this discussion, the first aim of this study is to test whether Islamophobia, understood as prejudices against Muslims and anti-Islam sentiment, can empirically be verified as two empirically distinct constructs or rather one-dimensional construct.Previous empirical findings (Kühnel & Leibold, 2007;Leibold & Kühnel, 2006;Leibold & Kummerer, 2011;Pfahl-Traughber, 2012) show different approval rates regarding each construct (more negative attitudes regarding Islam than regarding Muslims).Moreover, in agreement with Cottrell and Neuberg's (2005) argument that "traditional conceptions of prejudice -as a general attitude or evaluationproblematically obscure the rich texturing of emotions that people feel toward different groups."(p.1), I argue for operationalizing more specific and explicit items to assess prejudices against Islam and Muslims separately.In contrast to more general constructs such as the feeling-thermometer or the social distance scale (Bogardus, 1933), these newly operationalized items should be in line with commonly addressed negative attitudes in the public debate and media.In doing so, the items will more clearly elicit potential differences in the underlying threat mechanisms of negative attitudes against Muslims (individuals), on the one hand, and negative attitudes against Islam (as religion and culture), on the other hand.Based on these observations, I assume that anti-Muslim prejudice and anti-Islam sentiment are likely to be empirically distinct constructs (H1).

Threat as a Predictor of Prejudice
While social psychological research has developed several fundamentally different theories in order to explain prejudice (Allport, 1954;Brown, 1995;Nelson, 2009), the role of threat as a predictor of prejudice had been widely analyzed (Bobo, 2000;Esses, Dovidio, Jackson, & Armstrong, 2001;LeVine & Campbell, 1972;McLaren, 2003;Quillian, 1995;Riek et al., 2006).One theory that puts threats in a prominent position is the Integrated Threat Theory (ITT) by Stephan andStephan (1996, 2000).As its name already suggests, the ITT integrates several different theoretical approaches in the study of prejudice into a comprehensive model.The original model of ITT suggested four types of threat that are associated with prejudice.These are realistic threats, symbolic threats, negative stereotypes, and intergroup anxiety.All of the constructs have been empirically tested and have been found to predict prejudice (Riek et al., 2006).
Realistic threats are built upon the Realistic Conflict Theory (LeVine & Campbell, 1972;Sherif, 1966) and relate to political and economic power as well as to the physical well-being and safety of the in-group.They can include concerns regarding material goods (e.g., jobs and accommodation) and physical well-being (safety).Several studies confirm the positive relationship between prejudice and perceived realistic threats (Bizman & Yinon, 2001;Quillian, 1995;Stephan, Ybarra, & Bachman, 1999).This study uses measures to assess different types of realistic threats which were commonly associated with Islam and Muslims in German public debates, such as threats to the social welfare system, economy, and access to housing and jobs (Foroutan, 2012).
Symbolic threats are associated with the values, norms, morals, or identity of the corresponding in-group.Islam is frequently pictured in the media as an archaic, barbarian, and sexist religion, which represents a threat to the values and norms of the liberal-democratic German society (Bielefeldt, 2010;Hafez, 2010b;Schiffer, 2005).Prior studies carried out in other contexts have confirmed the positive relationship between symbolic threats and prejudice (Esser, 1996;Hewstone, 2004;Sears, 1988;Stephan, Stephan, & Gudykunst, 1999).In this study, symbolic threats encompass threats to the German and Christian culture, values, and identity.
More recent research on the relationship between intergroup threats and different intergroup outcomes has suggested expanding the types of intergroup threats, i.e. safety threats, beyond the classical symbolic and realistic threats suggested by the original ITT (Cottrell & Neuberg, 2005;Crawford, 2014;Doosje et al., 2009;Myers, Abrams, Rosenthal, & Christian, 2013;Riek et al., 2006).This expansion is thought to more accurately account for situational factors as well as different emotional reactions (e.g., fear, anger, envy, disgust, etc.) to specific intergroup constellations which have been shown to influence how different types of threat perceptions are associated with specific intergroup outcomes (Neuberg & Cottrell, 2002).
In this line of research, Cottrell and Neuberg (2005) argue that in order to more fully account for the variety of distinct intergroup constellations (black vs. white; liberal vs. conservative; male vs. female; majority vs. minority; high-power vs. low-power groups etc.), it is important to account for the specific emotional reactions that are potentially harbored within the constellations themselves.In fact, the ITT has been criticized for conceptualizing threats primarily on a cognitive level as a "cognitive appraisal" without integrating the affective level (Seipel & Rippl, 2005).This omittance also occurs in the revised ITT, although the importance of accounting for the affective level is stated (Stephan et al., 2009).Research findings, however, show that emotions can play a bigger role in explaining prejudice than cognitive factors do (Smith & Ortiz, 2002;Tougas & Beaton, 2002).
By eliciting the complex relationship between threats, emotions, and different intergroup outcomes, Cottrell andNeuberg (2002, 2005) demonstrate that, for example, threats concerning physical safety are empirically distinct from threats concerning jobs and accommodation, and furthermore, that safety threats emerge as a distinct predictor of fear, whereas symbolic and realistic threats are more strongly associated with anger and disgust.In a similar vein, Crawford (2014) demonstrates that safety threat constitutes an empirically distinct construct and shows a better predictive power regarding specific intergroup outcomes such as political intolerance against leftwing targets compared to the predictive power of symbolic and realistic threats.
In the context of Islamophobia, Doosje et al. (2009) specifically suggest analyzing terroristic threats concurrently with symbolic and realistic threats to determine whether terroristic threats are empirically distinct from symbolic and realistic threats, on the one hand, and to show their differential effects on anti-Muslim prejudices, on the other hand.
Following these two lines of research (Cottrell & Neuberg, 2005;Doosje et al., 2009), a third type of threat, terroristic threat, was included as a measure for this study in addition to the original suggestions of Stephan andStephan (1996, 2000).Since the events of 9/11, derogative views on Islam and Muslims in public discussions are frequently explained or justified by an imminent yet not predictable threat posed by terrorist attacks to the physical safety of Germans (Frindte & Haußecker, 2010).Whereas symbolic and realistic threats can be more objectively discussed and evaluated on scientific grounds, terroristic threats are not so easily dismissed, because they constitute a possible but unpredictable event.Furthermore, in line with Cottrell and Neuberg's (2005) theoretical rationale, I assume that terroristic threats fall under the category of safety threats rather than realistic threats, as terrorist attacks should be more closely related to fear than realistic threats (economic resources) or symbolic threats (values, freedom, and rights), which showed stronger associations to anger and disgust as primary emotional reactions (Cottrell & Neuberg, 2005).Therefore, although terroristic threats could be considered as a type of realistic threat in the classical ITT framework (well-being and safety), for analytical and practical reasons, operationalizing it as a category on its own will enable us to empirically compare different types of threats, thus clarifying the assumed empirical distinctiveness of each type of threat and measuring the individual effect of each on the proposed dimensions of Islamophobia.Based upon these analyses, the role of terroristic threats in explaining Islamophobia can be extrapolated and put into context with the other factors.
Regarding prejudice against Muslims, Schumann (2010) finds a positive relationship between perceived terroristic threats and anti-Muslim intergroup bias in her study.Pettigrew et al. (2007) show that symbolic and realistic threats (combined, not individually) mediate the relationship between the variable intergroup contact (direct and indirect) and anti-Muslim prejudice.Leibold (2010) demonstrates that Islam-related threats (symbolic and realistic threats conflated) are significantly associated with anti-Muslim prejudice.Doosje et al. (2009) show that terroristic threats are significantly related with subtle and blatant prejudice and discriminatory intentions against Muslims.Similar results are presented by Oswald (2005) for the US context.
Regarding prejudice against Islam, Pollack and Müller (2013) demonstrate in a representative study in Germany that symbolic threat perceptions are significantly related to prejudice against Islam.Frindte and Haußecker (2010) report a significant relation between perceived terroristic threats (perception of news reports on Islam-related terror in media by the participants of the study) and negative attitudes towards Islam.However, Leibold and Kummerer (2011) demonstrate that symbolic threats and -to a lesser degree -terroristic threats are related to Islamophobia (Muslim and Islam mixed).
Based on these findings, both anti-Muslim prejudice and anti-Islam sentiment have been found to be associated with symbolic and terroristic threats.Unfortunately, no distinct results for realistic threats can be given, as both types of threats were conflated in the study of Pettigrew et al. (2007) and in Leibold's (2010) study.Another study in the Netherlands (Velasco González et al., 2008), however, found no significant relationship between realistic threats and anti-Muslim prejudice, using the more general feeling-thermometer and a social-distance scale as prejudice constructs.Nevertheless, as none of these studies analyzed all three types of threats concurrently and no systematic differentiation between anti-Muslim prejudice and anti-Islam sentiment was made, I expect slightly different results.In line with previous research, I assume that symbolic and terroristic threats will be significantly associated with both anti-Muslim prejudice and anti-Islam sentiment (H2a).However, differences with regard to realistic threats are to be expected.Previous research has shown that realistic threats have emerged as a strong predictor of negative outgroup attitudes against low-status groups (Riek et al., 2006).Research on the public discourse in Germany points to the existence of derogative attitudes towards Muslims, based on their social and economic (de-)evaluation (Foroutan, 2012), and that Muslims are predominantly viewed as inferior by Islamophobic groups, as compared to Jews, who are viewed as superior by anti-Semitic groups (Schiffer & Wagner, 2011).
Therefore, I expect realistic threats to be significantly associated only with anti-Muslim prejudice (H2b).To the best of my knowledge, previous studies have not tested realistic threats associated with a religion, as social psychological research on intergroup processes is primarily concerned with attitudes towards social groups.Although no assumptions based on previous research can be made, I expect that realistic threats (accommodation, jobs, economy, education system) are unlikely to be associated with Islam, since a religion and culture cannot directly "threaten the economy" per se.Muslims, in contrast, can be perceived as potential usurpers of in-group resources.

Social Dominance Orientation as Antecedent of Threats
According to the ITT and the recent studies deriving from this theory (for an overview see Riek et al., 2006), different types of threat have generally been studied as the process mediating the relationship between certain contextual or individual difference variables and prejudice.Among these contextual or individual difference variables, some commonly studied ones are intergroup contact, in-group identification, and status differences (Oswald, 2005;Pettigrew et al., 2007;Riek et al., 2006;Stephan et al., 2000Stephan et al., , 2002;;Tausch, Tam, Hewstone, Kenworthy, & Cairns, 2007).For the revised version of the ITT, Stephan et al. (2009) suggest examining the role of inter-individual difference variables such as personality characteristics, social attitudes or personal ideologies in explaining threat perception and prejudice (Brown, 1995;Zick, 2010).Stephan and Renfro (2002) specifically suggest social dominance orientation (Sidanius & Pratto, 1999) as a further antecedent of threats.Accordingly, this study tests the inter-individual difference construct social dominance orientation as a predictor of different types of threat and the two subtypes of Islamophobia (Sidanius & Pratto, 1999).Previous studies have not analyzed SDO concurrently with symbolic, realistic, and terroristic (physical safety and well-being) threats (Crowson, 2009).Therefore, this study sets out to fill this gap.
The social dominance orientation (SDO) is a relatively new measure which assesses the individual differences in the endorsement of social hierarchies between different groups and hence, the rejection of social equality (Sidanius & Pratto, 1999).The SDO construct is based on the assumptions of the social dominance theory (SDT), which proposes that individuals differ in their endorsement of social hierarchies.The SDT assumes that highstatus groups tend to endorse social-hierarchical structures which favor the dominance of the in-group and demand the subordination of the outgroup.According to the SDT, the endorsement of intergroup differences (high-status vs. low-status groups) is partially a result of ideologies of dominance.Individuals with high SDO have been shown to perpetuate or improve their social status by adopting and approving so-called dominance legitimizing myths, through which the existence of social hierarchies is explained and justified.In turn, high SDO`s have been found to support non-egalitarian policies and attitudes (Zick, 1997).
Whereas the relationship between SDO and different negative intergroup attitudes has been widely tested (Pratto, Sidanius, Stallworth, & Malle, 1994;Sidanius & Liu, 1991;Zick et al., 2008;Zick, Küpper, & Hövermann, 2011), research findings on the relationship between SDO and threat perceptions are still ambiguous.Based on the assumptions of the Dual-Process Model (Duckitt, 2001), it has been argued that SDO, due to its specific motivational mechanisms, should be primarily related to realistic threats, while right-wing authoritarianism (Altemeyer, 1998) is assumed to primarily be related to symbolic and safety threats.Some studies demonstrate that individuals with high SDO are in fact especially prone to realistic threat perceptions (Esses, Dovidio, Jackson, & Armstrong, 2001;Esses, Jackson, & Armstrong, 1998;Duckitt, 2006;Jackson & Esses, 2000) and that realistic threats mediate the relationship between SDO and outgroup evaluation (Vezalli & Giovanni, 2010).Other studies of correlational and experimental design, however, indicate that SDO is also positively associated with symbolic threats (Cohrs et al., 2005;Costello & Hodson, 2011;Crowson, 2009).In this line of findings, Crowson (2009) reports that SDO shows stronger correlation with symbolic threats than with realistic threats.Nevertheless, Crowson points out that in his study, realistic threats did not include items measuring threats to physical safety and well-being and that future studies should address this omission (p.117).The results of this study should therefore shed more light on the question regarding the relation between SDO and the proposed types of threats.
In the context of integrating Islam and Muslims into German society, the role and status of Islam and Muslims in German society is being critically discussed in public debate, which displays highly exclusive tendencies (Foroutan et al., 2014).Individuals with high SDO therefore may perceive the proactive integration efforts and the rising visibility of Islam and Muslims in German society as threatening to the status of the (non-Muslim) in-group.As a result of the threat perceptions, individuals showing a higher rate of SDO could tend to exhibit more prejudice against the perceived outgroup.In light of the results presented above, SDO is expected to be significantly related to Islamophobia.Although SDO has been shown to relate to all three types of threat, mediation effects are expected to be outcome specific.In other words, it is assumed that all three types of threats will mediate the links between SDO and anti-Muslim prejudice (H3a).Regarding the relationship between SDO and anti-Islam sentiment, symbolic and terroristic threats are expected to show mediating effects but not realistic threats (H3b).

Method Participants
The data acquisition for the empirical analyses was done via an online survey (Unipark).
i The survey was advertised through different channels, for example, email distribution lists from universities, school discussion boards, and social media channels.The survey ad and survey starting page indicated that participants should currently be living in Germany and that participation was voluntary and anonymous.As an incentive, a competition (to win an iPod, iPhone, and iPad) was announced.The data acquisition took place between June and August 2014.After removing all non-ethnic German respondents ii and all respondents with any missing values (less than 5%), a total of N = 355 participants completed the survey.In the final sample, 203 (57.2%) of the participants were female, and 152 (42.8%) were male.The ages of the participants ranged from 18 to 36 years (M = 25.43,SD = 4.12).
Social economic status was assessed by education level and cultural capital (number of books possessed).The mean level for education (scale 1 -8) was relatively high (M = 5.35, SD = 1.29), since the sample mainly consisted of students.Cultural capital was assessed by asking about the number of books owned and was measured on a six-point scale ranging from (1) fewer than 10 books to (6) more than 500 books (M = 4.27, SD = 1.22).

Measures
The Cronbach`s α for each of the measured constructs can be found in Table 1.

Predictor Variable
Social dominance orientation was measured according to the shortened SDO scale (Pratto et al., 2013) by three items (e.g., "In setting priorities, we must consider all groups."(reverse coded); "Superior groups should dominate inferior groups.";and "We should not push for group equality.").
iii All items were measured on a five-point Likert scale ranging from (1) strongly disagree to (5) strongly agree.Higher scores equal higher SDO.

Threats
The symbolic and realistic threats were adapted from Stephan et al. (2000Stephan et al. ( , 2002) ) and Velasco González et al.
( 2008).The terroristic threats were self-constructed.Symbolic and realistic threats included four items each, two of which were coined for threats regarding Muslims and two for threats regarding Islam.All the threats were measured on a five-point scale ranging from (1) strongly disagree to (5) strongly agree.All the measures show significant intercorrelation in the expected direction (Table 1).Table 2 indicates the results for the factor analyses with maximum likelihood extraction method and promax rotation.The initial EFA with an eigenvalue = 1 setting yielded five factors (instead of the expected six).A visual inspection of the scree plot revealed a steep slope after the sixth factor.The matrix for the explained variance confirmed the visual inspection by showing that the sixth factor has an initial eigenvalue of .962.The five-factor model explained in total 61.26% of the explained variance.Based on these observations, in a second step, an EFA with a forced factor choice set to six factors was conducted.The six-factor model explained 5% more variance (67.31%) than the five-factor model.Moreover, the six-factor model had a significantly better fit to the data than the five-factor model [χ 2 diff.
(35) = 392.154,p < .001].Therefore, the six-factor model was chosen over the five-factor model.The results for the six-factor model indicate that all items were correlated at least .3with at least one other item, suggesting reasonable factorability.Second, the Kaiser-Meyer-Olkin measure of sampling adequacy was .94,above the recommended value of .6, and Bartlett's test of sphericity was significant (χ 2 (231) = 5403.686,p < .05).The diagonals of the anti-image correlation matrix were all over .5, supporting the inclusion of each item in the factor analysis.The six factors explained 67.4% of variance in total.Because of the presence of Muslims in Germany, Germans have more difficulties finding a job. .97 Because of the presence of Muslims in Germany, Germans have more difficulties finding housing. .94 Because of the presence of Islam, the education system in Germany is threatened. .87 The presence of Islam in Germany threatens our economic prosperity. .78

Anti-Islam Sentiment (α = .94)
Islam is a violence-glorifying religion. .88 Islam is an anti-Semitic religion. .78 Islam is a sexist religion. .76 The Islamic religion is harmful for world peace. .69 Symbolic Threat (α = .91)

Social Dominance Orientation (α = .82)
We should not push for group equality. .87 In setting priorities, we must consider all groups.(reverse coded) .86 Group equality should be our ideal.(reverse coded) .70 Terroristic Threat (α = .83) I am worried that the peace is threatened by radical Islamist groups in Germany. .84 It is only a matter of time before Germany will become a target for Islamist terrorists. .77 Sometimes, I think I could become a victim of an Islamist terrorist attack myself..55 Note.Coefficients smaller than .20 are suppressed.

Measurement Model
A confirmatory factor analysis (CFA) including all the proposed constructs (Amos 22.0) was performed to assess the construct validity of the individual variables.The convergent validity and discriminant validity of the constructs hold when the CFA results indicate a good model fit (Brown, 2006).Following the suggestions of Hu and Bentler (1999), the Root Mean Square Error of Approximation (RMSEA) and the Standardized Root Mean Square Residual (SRMS) were employed as absolute fit indices, and the Comparative Fit Index (CFI) was employed as a comparative fit index.Conservatively speaking, a good fit of the data is indicated by an RMSEA value of less than 0.05, an SRMSR value of less than 0.08, and CFI values of 0.95 or higher (Hu & Bentler, 1999).Table 3 shows that the CFA with a six-factor oblique solution, in which all constructs were allowed to correlate with each other, showed an acceptable absolute and comparative fit to the data.The six-factor oblique model had a significantly better fit Disentangling Islamophobia: SDO and Threats 76 to the data than the three-factor oblique model, in which all variables with factor correlations higher than .70(symbolic, realistic, and terroristic threats as well as anti-Muslim prejudices) were operationalized as a single factor (χ 2 diff.(12) = 774.291,p < .001).The six-factor oblique model had also had a better fit than a three-factor model (χ 2 diff.(12) = 844.651,p < .001), in which all threats (factor one), SDO (factor two) and anti-Muslim prejudice and anti-Islam sentiment (factor three) were constructed as three separate factors.To improve the model fit indices for the final six-factor oblique model, error terms for items of the same latent constructs, but not for items measuring different latent constructs, were allowed to covary, resulting in the respecification of three parameters in total: one residual covariance for symbolic threats one residual covariance for terroristic threats, and one residual covariance for anti-Muslim prejudice.The model χ 2 of 283,286 indicates a lack of an absolute fit (p < .001),which is not uncommon for larger sample sizes.However, all the other fit measures indicate that the model has a good model fit: χ 2 /df = 1.48;CFI = .98;SRMR = .034,and RMSEA = .037 and 90% CI = .039-.052 (fit indices for the unrestrained model are shown in Table 3).The z-statistics obtained for all the factor loadings were statistically significant (p < .001),and the standardized factor loadings were between .52 and .92.
In line with Hypothesis 1, anti-Muslim prejudice and anti-Islam sentiment showed good discriminant as well as good convergent validity.iv Nevertheless, due to the high intercorrelations between some constructs (Table 1), I further tested them for their discriminant validity.Discriminant validity signifies the degree to which measures of two constructs are empirically distinct (Bagozzi, Yi, & Phillips, 1991).With regard to this type of validity, I employed the test proposed by Fornell and Larcker (1981).In this test, a construct is empirically distinct if the average variance extracted (AVE) by that construct's items is greater than the construct's shared variance with other constructs (i.e., the square root of the intercorrelation).All the constructs employed in this study satisfied Fornell and Larcker's (1981)

Structural Model
In a first step, a SEM -excluding all threats-was calculated to test the direct associations of SDO with the dependent Uenal 77 Second, a SEM including all latent variables was tested.The proposed structural model shows a good fit to the data (χ 2 = 295.605;p < .001;df = 208; χ 2 /df = 1.42;CFI = .99;SRMR = .038and RMSEA = .034with 90% CI = .025-.043). Figure 1 shows the results of a model in which all pathways, direct as well as indirect, were estimated simultaneously.Education was entered as a control variable.As assumed, SDO was significantly associated with all threats.Although the effect sizes varied minimally, SDO showed the biggest effect on realistic threats (r 2 = .370,p < .01),followed by terroristic threats (r 2 = .359,p < .01)and symbolic threats (r 2 = .328,p < .05).SDO showed no remaining significant direct relationship with either of the dependent variables in the presence of mediators.Note.Non-significant paths between threats and the dependent variables were included in the model, but are not shown.
Residual variances were allowed to covary among the threat variables as well as among the dependent variables (not shown).
In line with Hypothesis 2a, symbolic threats showed a significant association with both anti-Muslim prejudice and anti-Islam sentiment.Terroristic threats, as proposed, showed a significant relation to anti-Islam sentiment, but, unexpectedly, no significant association with anti-Muslim prejudice.v For realistic threats, the assumed significant association was indicated with anti-Muslim prejudice (H2b).No significant relation between realistic threats and anti-Islam sentiment was found.Education had a significantly negative relationship with symbolic threats (B = -.092,S.E.= .030,p = .007)and realistic threats (B = -.093,S.E.= .022,p = .007),as well as with anti-Islam sentiment (B = -.067,S.E.= .024,p = .014)but not with anti-Muslim prejudice.
To assess the mediating role of the threats on the relationships between the predictor variable and the dependent variables, the total effects of the predictor were further decomposed into direct and indirect effects.Table 4 shows that for anti-Muslim prejudice, the indirect pathways are significant (H3a), indicating full mediation through symbolic and realistic threats but not through terroristic threats.As assumed (H3b), full mediation is indicated for SDO via symbolic and terroristic threats on anti-Islam sentiment but not realistic threats (see Table 4).

Discussion
The goal of this study was first, to analyze whether Islamophobia empirically constitutes a one-dimensional or rather multi-dimensional construct.Second, the effects of intergroup threats (symbolic, realistic, and terroristic) on Islamophobia were analyzed.Finally, SDO is tested as an antecedent of perceived threat and Islamophobia.
It was assumed that all threats would mediate the effects of SDO on anti-Muslim prejudice, but only symbolic and terroristic threats would mediate the effects of SDO on anti-Islam sentiment.The revised ITT (Stephan & Renfro, 2002) was utilized as an analytical framework.To determine the relationships between the antecedent, threats, and Islamophobia, a structural equation model was tested.The assumptions of this study were mostly supported.
First, based on the ongoing theoretical discussions on the definition and dimensionality of Islamophobia (Larsson & Sander, 2015;Kühnel & Leibold, 2007;Pfahl-Traughber, 2012;Shooman, 2011), this study aimed at testing whether Islamophobia can empirically be understood as a one-dimensional or rather multidimensional phenomenon.
Several items that reflected frequently reiterated negative attitudes towards Muslims and Islam in German public debate were assessed.The results of a confirmatory factor analysis revealed two robust factors for prejudices against Muslims and anti-Islam sentiment.Therefore, the results of this study can be interpreted as supporting a multidimensional definition of Islamophobia (Kühnel & Leibold, 2007).
Second, previous studies on Islamophobia and intergroup threats did not analyze all three types of threats (terroristic, symbolic, and realistic) concurrently (Doosje et al., 2009).Therefore, they were not able to determine the effects of each threat individually while accounting for the effects of the others.Following the suggestions of Doosje et al. (2009), this study set out to fill in this void.With regard to Islamophobic attitudes post-September 11, 2001, terroristic threats are a focal point of public debate and policy in Germany (Bielefeldt, 2010;Frindte & Haußecker, 2010).In turn, I argued that terroristic threats do not necessarily fall under the category realistic threats, which are predominantly associated with job and housing security and other economic concerns.The results of this study confirm that terroristic threats could not only be distinguished on a theoretical level, but they were also empirically different from symbolic and realistic threats, based on the results of the confirmatory factor analysis.This finding is in line with previous research on intergroup threats that suggests differentiating between realistic threats and safety threats (Cottrell & Neuberg, 2005;Crawford, 2014).The assumption of Doosje et al. (2009) that terroristic threats would form a unique form of threat is also substantiated in this study.Moreover, terroristic threats explained a significant amount of variance in anti-Islam sentiment.Furthermore, as expected, terroristic threats were associated with anti-Islam sentiment but contrary to expectation, not with anti-Muslim prejudice.
The association of anti-Islam sentiment with terroristic threats might be explained by understanding Islam as a proxy factor, which does not directly commit acts of terrorism yet is perceived as a contributing factor.The nonsignificant relationship between terroristic threats and anti-Muslim prejudice is surprising, as previous research has found otherwise.However, the reason for this might be found in an asymmetrical distribution in the item wordings regarding terroristic threats, anti-Islam sentiment, and anti-Muslim prejudice: three out of three terroristic threat items mentioned Islamists but none referred to Muslims.In the anti-Muslim prejudice items, on the other hand, one item explicitly referred to Islamist terrorists, thus partially mitigating the asymmetry.The anti-Islam sentiment contained four items, all of which referred to Islam.Obviously, this does not imply that Islamist and Islam are equal.Nevertheless, the results have to be interpreted in the light of this asymmetrical distribution.Regardless, when symbolic and realistic threats were excluded from the SEM, terroristic threat did show a significant association with prejudice against Muslims.
As expected, realistic threats were associated with anti-Muslim prejudice but not with anti-Islam sentiment.The significant association between realistic threats and anti-Muslim prejudice is in line with the findings of Riek et al. (2006), who show in their meta-study on intergroup threats and intergroup attitudes that the relationship between threat perceptions and outgroup attitudes seems to be effected by the (perceived) characteristics of the threatening outgroup.Their results indicate that realistic threats are a strong predictor for negative attitudes against low-status outgroups.
Shifting our attention back to the first aim of this study, which is concerned with the dimensionality of Islamophobia, the results on the underlying threat mechanisms of Islamophobia corroborate the notion of a multidimensional model.While symbolic threats were related to both dependent variables, realistic threats were only associated with anti-Muslim prejudices, and terroristic threats were only associated with anti-Islam sentiment.This is partially in line with previous research findings (Frindte & Haußecker, 2010;Leibold, 2010;Pettigrew et al., 2007;Pollack & Müller, 2013;Schumann, 2010).However, previous studies have not found a significant relationship between realistic threats and anti-Muslim prejudices.Nevertheless, the significant association between realistic threats and prejudice against Muslims is also consistent with research on the public perception of Muslims in Germany that Muslims are a social and economic burden to German society (Foroutan, 2012).It could be argued that this perception might have intensified after the heated public debates following the publication of the bestseller book "Deutschland schafft sich ab" ("Germany is Abolishing Itself") by former social democratic politician Thilo Sarrazin in 2010, in which Muslims were portrayed as a social and economic burden to Germany (Foroutan, 2012).In sum, the different threat mechanisms that are associated with each proponent of Islamophobia further corroborate the assumption that Islamophobia is more accurately described as a multidimensional construct.
In the light of these results, a one-dimensional definition and operationalization of Islamophobia seems inadequate to describe negative attitudes against Islam and Muslims and the specific mechanisms that are associated with each of them.Future studies concerned with Islamophobia, at least from a social psychological perspective, would be well advised to further test whether Islamophobia constitutes a one-or rather multi-dimensional construct.A more diverse sample should be obtained in order to further examine the proposed multi-dimensionality.Moreover, comparative studies with samples from other countries should shed further light on the topic at hand.As Islamophobia is a focal point in discussions about immigration and integration in Europe (Sheridan, 2006), a better understanding of the phenomenon should also help to develop more practical approaches for tackling the current intergroup tensions.
Finally, following the suggestions of Stephan and Renfro (2002), social dominance orientation was tested as an antecedent of threats and Islamophobia.Previous research indicates a significant association between SDO and intergroup threat (Cohrs et al., 2005;Crowson, 2009;Imhoff & Bruder, 2014;Oswald, 2005) as well as between SDO and Islamophobia (Imhoff & Bruder, 2014;Zick & Küpper, 2006, 2007).Nevertheless, theoretical assumptions, based on the Dual-Process Model (Duckitt, 2001), in which it has been hypothesized that SDO should primarily relate to realistic threats, were not consistently supported by empirical research (Cohrs et al., 2005).In this line of research, Crowson (2009) suggests to further investigate the association between SDO and threats concerning the physical safety and well-being in conjunction with realistic and symbolic threats.According to Crowson (2009), this study therefore analyzed symbolic, realistic, and terroristic (safety) threats concurrently.
Although the effect sizes of SDO on the threats varied only minimally, SDO was most strongly associated with realistic threats, followed by terroristic threats and symbolic threats.This finding is in line with the assumptions of the Dual-Process Model and previous research associated with it (Duckitt, 2006;Esses et al., 2001;Jackson & Esses, 2000;Vezalli & Giovanni, 2010).The results of this study show, however, the varying effect of SDO on realistic threats (related to jobs, accommodations etc.) and terroristic threats (related to physical safety and wellbeing), the former having a slightly stronger association with SDO.Therefore, adding (terrorism-related) safety threats into the equation, as proposed by Crowson (2009), did not change the quality of relationship between SDO and realistic threats, as safety threats emerged as an empirically distinct construct.In turn, the proposed distinction between realistic threats and safety threats (Cottrell & Neuberg, 2005) is supported for analyzing SDO as antecedent of threats and could also be useful in regard to research on RWA.Future research on the relationship between SDO and different intergroup threats could further analyze the differential effects in other contexts where safety threats are not terrorism-related but associated with other forms of intergroup conflict and also analyse the differential effects in regard to different intergroup outcomes.
Moreover, in agreement with previous findings, SDO significantly predicts Islamophobia directly when threats are excluded from the model (Zick & Küpper, 2006, 2007).These results give further empirical support for the assumptions of the social dominance theory (Sidanius & Pratto, 1999).When threats were included in the model, SDO showed only indirect effects via all threats on Islamophobia.As expected, mediation effects were specific to the outcome variable.Overall, the results of this study provide further empirical support for considering SDO as an antecedent of threat and prejudices, as suggested by Stephan and Renfro (2002).Individuals with high SDO seem to exhibit more prejudice against Muslims and Islam indirectly via threats.It is worth mentioning that this study used the shortened SDO scale (Pratto et al., 2013), which showed acceptable but relatively low reliability including all four items (α = .70).Regardless, one item ("Group equality should be our ideal.")did not fit the data well in the CFA and was therefore dropped out.Future studies could be done to improve the reliability of the SDO measure.
In sum, the revised ITT (Stephan & Renfro, 2002) provided a good theoretical framework for analyzing Islamophobia.
Nevertheless, gaining insight from Cottrell and Neuberg's (2005) sociofunctional threat approach, this study concludes that the proposed distinction between realistic threats and safety threats might also prove useful for future research that utilizes the (revised) ITT, at least in conjunction with terrorism-related safety threats.
On a final note, some limitations of this study should be mentioned.First, it must be noted that due to the crosssectional approach of this study, causal inferences are not possible.However, one study in which longitudinal data was used suggests a causal relationship between threats and prejudice (Schlueter, Schmidt, & Wagner, 2008).Furthermore, the survey took place between June and August 2014, at a time when ISIS was spreading its terror in Iraq and Syria.German media extensively covered the ongoing fighting in this region.Thus, the results of this study have to be interpreted with this situational factor in mind, as it could have influenced the respondents' responses.Moreover, the relatively small sample of this study consisted mainly of students, meaning that the results of this study are not representative for German society as a whole.Replicating this study with a more diverse or representative sample could reveal different results.Additionally, data acquisition was done via an online survey.
The obvious advantage of online surveys is to reach respondents in a short time period and more cost effectively in comparison to the traditional paper-pencil method.Furthermore, online surveys could decrease the tendency for social desirability bias in responses, due to the anonymity of the web, resulting in more candid answers.On the other hand, online surveys must be very brief in order to ensure a high number of finished surveys (for an introduction and discussion on the online survey method, see Dillman, Smyth, & Christian, 2009).Finally, the high percentage of explained variance of the dependent variables in this study (up to 70%) may be exaggerated through the common method variance of the survey procedure.

Conclusion
From a practical perspective, the results of this study suggest that countermeasures and interventions against anti-Muslim and anti-Islam sentiment could critically address the following points.
By differentiating between symbolic, realistic, and terroristic threats, this study allowed for the independent examination of each threat type and their effects on prejudices and sentiments.Interestingly, realistic threats and symbolic threats but not terroristic threats had a significant effect on anti-Muslim prejudices in this sample.Although most of the realistic threats and symbolic threats operationalized in this study can scientifically be challenged as exaggerations (Foroutan, 2012), these perceptions best predicted anti-Muslim prejudices in this sample of welleducated and young individuals.In light of this, measures to reduce anti-Muslim prejudices could aim to reduce symbolic and realistic threat perceptions by providing accurate information on Muslims and Islam in Germany.
Moreover, in order to understand the complex processes involved in the evaluation of intergroup relations on the micro level, it is important to consider the effects of public debate, public policy, and the media on this process.
In the context of this study, it is noteworthy that the majority of people in Germany primarily rely on information from the media when it comes to Islam and Muslims, as a recent representative study demonstrates (Foroutan et al., 2014).Furthermore, analyses on the representation of Islam and Muslims in the media indicate that Islam and Muslims are frequently represented as homogenous entities, which in turn are contextualized in negative framings (e.g., with regard to crime, terrorism, violence, sexism, and intolerance) and depicted as significant and Realistic threats were measured with four items regarding threats concerning economy, housing, accommodation and political issues in relation to Islam and Muslims (i.e., "The presence of Islam in Germany threatens our economic prosperity.";"Because of the presence of Islam, the education system in Germany is threatened.";"Because of the presence of Muslims in Germany, Germans have more difficulties finding a job."; and "Because of the presence of Muslims in Germany, Germans have more difficulties finding housing.").Symbolic threats were measured by four items (i.e., "I am worried that..." (a) "...the occidental culture is endangered by Islam.";(b) "...the Christian norms and values are threatened by Islam.";(c) "...the German norms and values are threatened by the presence of Muslims.";(d) "...our rights and freedom are threatened by the presence of Muslims.")(M = 1.58,SD = 0.78); t(354) = 11.38,p < .001.In sum, the descriptive mean scores suggest relatively low threat perceptions, anti-Islam sentiment, and prejudices against Muslims for this sample of well-educated and relatively young respondents.

Table 2
Factor Loadings Based on a Maximum Likelihood Analysis With Promax Rotation for 22 Items(N = 355) am worried that the German norms and values are threatened by the presence of Muslims.

Table 3
Comparison of Fit Indices of the First Model (Six-Factor Oblique), the Second Model (Three-Factor Oblique), and the Third Model(Three-Factor  Oblique) criterion, showing AVE values above .50.

Table 4
Decomposition of Unstandardized Direct and Indirect Effects on Islamophobia, With Bootstrap Standard Errors in Parentheses