1 Introduction

A key goal of teacher education programs is to develop future teachers who hold inclusive attitudes toward diverse student groups and who are willing to advocate on their behalf. Teacher educators have adopted a number of strategies to foster greater inclusiveness, including discussion of historical injustices perpetuated against minority groups, philosophical and ethical issues surrounding inclusivity, and even social and psychological factors that contribute to between-person and between-group differences (Crowson and Brandes 2013). While the effort to increase inclusivity is a laudable effort, serious questions remain about how inclusive attitudes are developed and maintained in pre-service teachers. This is an important issue to consider, as pre-service teachers with varying personality characteristics, ideological commitments, and background experiences could process messages aimed at fostering more inclusive attitudes quite differently (Dee and Henkin 2002; Unruh and McCord 2010). In order to maximize the effectiveness of instructional interventions aimed at fostering inclusive attitudes, teacher educators need to acquire a better understanding of those for whom their messages are intended (Castro 2010).

The purpose of our study was to address the question of how inclusive attitudes are formed and maintained. In particular, we sought to predict pre-service teachers’ attitudes concerning one particular minority group—students with disabilities—based in part on a set of predictors that have historically been linked to prejudice and discrimination within the social psychology literature. In our proposed path model (see Fig. 1), we predicted that opposition to the practice of inclusion would be predicted by social dominance orientation (SDO), intergroup contact, intergroup anxiety, and use of negative stereotypes. In addition to these social psychological factors, we predicted that pre-service teachers’ self-efficacy for teaching students with disabilities may be related to opposition to the practice of inclusion. At this time, we review the literature that forms the basis of the proposed relationships in our model.

Fig. 1
figure 1

Hypothesized path model

Full size image

1.1 Proximal predictors of opposition to inclusion: self-efficacy, intergroup anxiety, stereotyping

We begin the review of our prediction model by discussing three proximal predictors of attitudes of opposition to inclusion: self-efficacy, intergroup anxiety, and negative stereotyping. Self-efficacy refers to the belief that one can successfully carry out a behavior in order to produce a particular outcome (Bandura 1977). In general, perceived efficacy serves as a motivating factor for engaging in a behavior, with those individuals feeling less efficacious concerning their ability to carry out a behavior being less likely to actually engage in the behavior. According to Skaalvik and Skaalvik (2007), teacher self-efficacy has been defined “as individual teachers’ beliefs in their own abilities to plan, organize, and carry out activities required to attain given educational goals” (p. 612). This factor has been shown to be related to teacher’s occupational commitment and intentions to quit (Klassen and Chiu 2011) and burnout (Schwarzer and Hallum 2008; Skaalvik and Skaalvik 2007). There is also some evidence that teacher self-efficacy may be related to students’ academic achievement (Caprara et al. 2006) and student perceptions of task difficulty and expectancies for future success (Midgley et al. 1989).

To date, little is known about whether and how self-efficacy perceptions among in-service or pre-service teachers translate into prejudiced attitudes toward minority students or support for school-related policies and practices aimed at serving these students. Woodson-Williams (2010) examined the relationship between teacher self-efficacy and prejudiced attitudes toward students classified as English language learners (ELL), controlling for factors such as right-wing authoritarianism (RWA) and SDO. In her multiple regression analysis, practicing teachers’ self-efficacy for teaching ELL students emerged as the strongest predictor of prejudiced attitudes toward these students, followed by SDO, and RWA. Self-efficacy was a strong predictor of negative attitudes towards ELL training programs for teachers (followed by SDO). Low efficacy teachers were also more likely to indicate low intentions to pursue training to learn more about how to teach ELL students.

The findings from Woodson-Williams’ (2010) study are consistent with Mazziotta et al. (2011) assertions about the role that self-efficacy perceptions may play in cross-group interactions. Mazziotta et al. argue that the “personal judgment of one’s own abilities to have a successful cross-group interaction” may play “a crucial role in how people think and feel about the possibility of their own direct cross-group contact and, given the opportunity, whether they will seek out or avoid direct cross-group contact” (p. 258). Arguably, those teachers low in self-efficacy within Woodson-Williams’ study were more avoidance-oriented when it came to interacting with the target minority group (i.e., ELL students), even to the point that they were unwilling to learn more about how to teach them. Assuming that self-efficacy for teaching students with disabilities operates in a similar fashion—i.e., influencing the motivation for intergroup contact within future school and classroom settings—among pre-service teachers, it seems reasonable to expect that efficacy perceptions will negatively predict opposition to inclusion in the current study.

Intergroup anxiety refers to the expectation of “negative psychological, behavioral, and/or evaluative consequences for the self” when considering potential future intergroup contact with members of other groups (Swart et al. 2011, p. 1222). Negative stereotypes, on the other hand, capture a “cognitive component of prejudicial attitudes” (Corenblum and Stephan 2001, p. 253) that reflect negative beliefs concerning the attributes of outgroup members and expectancies that derive from them (Riek et al. 2006). Research on these two factors indicates that both of these factors are implicated in prejudiced responding towards social outgroups, including the desire to be physically distant from those groups (e.g., Curseu et al. 2007).

To this point, we could identify only one study that addresses the relationship between prejudiced beliefs and attitudes and pre-service teachers’ opposition to inclusion. Brandes and Crowson (2009) found that intergroup anxiety (described as “discomfort with disability”) and negative attitudes (which appear to have captured a mixture of affective and cognitive aspects of prejudice) were both positively and significantly correlated with opposition to inclusiveness. Given these findings, we expected pre-service teachers experiencing greater levels of intergroup anxiety and holding more negative stereotypes concerning students with disabilities to be more likely to oppose the practice of inclusion. We expected the effects of these variables to be both direct (reflecting general prejudiced tendencies toward students with disabilities) and indirect via their effects on perceived efficacy for teaching students with disabilities.

1.2 Social dominance orientation as a predictor of opposition to inclusion

Social dominance orientation (SDO) is defined as a “general attitudinal orientation toward intergroup relations, reflecting whether one generally prefers such relations to be equal, versus hierarchical, that is, ordered along a superior-inferior dimension” (Pratto et al. 1994, p. 742). Moreover, individuals high on SDO are those that prefer their own social groups to exhibit dominance over other groups (Sidanius and Pratto 1999). According to Social Dominance Theory (SDT), all societies structure their social groups hierarchically, with some groups sitting further up the status hierarchy than others. In order to perpetuate this hierarchy, certain social myths (i.e., hierarchy-enhancing legitimizing myths) are utilized to justify status differences. Persons scoring higher on SDO are those who create and/or rely on these myths to justify group-based inequality. They are less likely to endorse hierarchy-attenuating myths aimed at reducing inequality (Sidanius and Pratto 1999).

Duckitt and colleagues, arguing from their Dual-Process Cognitive Motivational Model (DPM; Perry and Sibley 2012), have attempted to explain the genesis of SDO, as well as its proposed differential effects on outgroup prejudice (i.e., persons high on SDO are prejudiced against certain groups, but not others)—generally by contrasting SDO with another prejudice-related factor (i.e., right-wing authoritarianism, or RWA; see e.g., Duckitt et al. 2002). According to this model, SDO is the ideological expression of a personality marked by low levels of agreeableness as well as a view of the social world as a competitive jungle (Duckitt and Sibley 2009; see also Perry and Sibley 2012). Unlike persons scoring higher on RWA (i.e., individuals who feel most threatened by social groups that challenge tradition and existing social norms and conventions), those scoring higher on SDO experience outgroups that are lower in status and that activate competitiveness concerns as particularly threatening (Duckitt and Sibley 2007).

Duckitt and colleagues (Duckitt 2006; Duckitt and Sibley 2007) argue that competitiveness concerns may be activated in situations where low status groups actively challenge the existing status hierarchy or in circumstances where there is a perceived need to justify the dominance of one’s own social group over others. Consistent with these assertions, SDO has been shown to correlate more strongly with prejudice against “derogated groups” (i.e., groups that are lower in status) as opposed to groups considered as “dangerous” (i.e., those threatening social norms or the status quo; Duckitt and Sibley 2007). For example, persons high on SDO are more likely to exhibit prejudice against ethnic minorities, immigrants, people with disabilities, and those who are unemployed (see e.g., Bizer et al. 2012; Crowson and Brandes 2010; Duckitt 2006; Duckitt and Sibley 2007; Pratto et al. 1994). Moreover, researchers studying SDO have shown that SDO negatively predicts social policy attitudes aimed at reducing social, educational and income inequalities in society (see e.g., Crowson et al. 2013; Pratto et al. 1994; Sibley and Duckitt 2010).

For the current study, we assumed that pre-service teachers scoring higher on SDO should be more inclined to oppose the practice of including students with disabilities into regular education classrooms. First, we assumed that SDO would exert a direct effect on opposition to inclusion given the abovementioned research indicating that people high on SDO appear to be (a) prejudiced against lower-status social groups (e.g. people with disabilities) that arouse competitiveness concerns and (b) opposed, on principal, to any effort to “level the playing field” for those groups. Students with disabilities fall into the low-status category and, as such, pre-service teachers scoring higher on SDO should be more inclined to oppose the (arguably) social-leveling practice of inclusion.

Second, we assumed that the influence of SDO on opposition to inclusion may also be indirect, via negative stereotyping. As noted above, persons higher on SDO are generally more inclined to endorse hierarchy-enhancing legitimizing myths that justify group-based inequality than those low on this factor. Greater endorsement of these system-justifying ideologies, in turn, leads to increased adoption of social policy attitudes that support the established status differentials in society (Sibley and Duckitt 2010). According to Sidanius and Pratto (1999) and Asbrock et al. (2011), stereotypes may serve as one type of legitimizing myth in order to justify intergroup prejudice. With these considerations in mind, we reasoned that pre-service teachers scoring higher on SDO may score higher on their negative stereotypes concerning people with disabilities (e.g., they are incapable of taking care of themselves or handling their own affairs), in part, to justify their opposition to equality for this group. We further reasoned that opposition to inclusion is a natural expression of this opposition to equality.

Third, we suspected that SDO’s impact on attitudes of opposition to inclusion might be mediated via pre-service teachers’ level of intergroup anxiety related to people with disabilities. We reasoned that pre-service teachers scoring higher on SDO should experience higher levels of anxiety at the thought of interacting with individuals from minority groups (i.e., students with disabilities) for whom they already hold prejudices. The result of this anxiety should be a reduction in their level of self-efficacy concerning their ability to instruct students with disabilities and, by extension, their lowered level of support for inclusive educational practices.

There are several studies that support various aspects of the proposed relationship between SDO and attitudes concerning people with disabilities, including attitudes that do not support inclusion, in the current study. Crowson and Brandes (2010),  Crowson et al. (2013), Perry and Sibley (2012), Bäckström and Björklund (2007) have all reported correlations indicating that persons scoring higher on SDO tend to exhibit more prejudiced attitudes toward people with disabilities. These results were obtained in samples comprised of pre-service teachers, high school students, or community adults. Brandes and Crowson (2009) also demonstrated that the effect of SDO on opposition to inclusion of students with disabilities in schools is mediated by intergroup anxiety in their sample of community adults. Finally, Brandes and Crowson’s (2009) correlational data revealed positive and significant correlations between SDO (on the one hand) and intergroup anxiety (i.e., “discomfort with disability”) and prejudiced attitudes and opposition to inclusion (on the other) in a sample of pre-service teachers. Using mediation analysis, the authors found evidence for both direct and indirect effects of SDO on attitudes toward inclusion via negative attitudes (i.e., prejudice).

1.3 Intergroup contact as a predictor of opposition to inclusion

The contact hypothesis asserts that “positive intergroup contact is capable of reducing intergroup prejudice and improving intergroup relations” (Swart et al. 2011, p. 1221). In general, a substantial body of evidence coming from cross-sectional, longitudinal, and meta-analytic studies (see e.g., Binder et al. 2009; Pettigrew and Tropp 2006; Tropp and Pettigrew 2005) provides strong, converging evidence in favor of the claim that intergroup contact reduces prejudice, with contact exerting larger effects when the quality of contact experiences is high (Hewstone and Swart 2011), as in those cases where contact experiences involve cross-group friendships or other close relations (Tropp and Pettigrew 2005). Notably, the effect of intergroup contact on prejudice may vary depending on whether cognitive or affective indicators of prejudice are utilized. According to Tropp and Pettigrew 2005, contact effects are more apparent when prejudice is operationalized using affective, as opposed to cognitive, dimensions of prejudice (although contact effects have been observed for the latter as well).

Researchers have found compelling evidence that the effect of contact on prejudice is likely indirect (i.e., mediated by various intervening factors), as opposed to direct. Two factors in particular that have been examined as mediators are intergroup anxiety and use of negative stereotypes (see e.g., Binder et al. 2009; Gaunt 2011; González et al. 2008; Levin et al. 2003; Pettigrew and Tropp 2008), with stronger effects generally being observed in the case of the former. The fact that contact is associated with reduced intergroup anxiety may be explained in terms of its effects on the social perceiver’s sense of uncertainty and level of self-efficacy. Specifically, contact experiences may reduce one’s subjective sense of uncertainty and increase his/her sense of efficacy concerning how to interact in intergroup settings (Mazziotta et al. 2011), thereby reducing experienced levels of anxiety at the anticipation of future encounters. The effect of contact on stereotype use, on the other hand, may be explained in terms of the individuating information received as a function of interactions with outgroup members (Tropp and Pettigrew 2005). To the extent that this information is generalized from individual members to the outgroup as a whole, there should be decreased reliance on stereotypes when judging the outgroup.

Based on theory and previous research, we reasoned that prior close contact with people with disabilities should negatively predict pre-service teachers’ level of opposition to the practice of inclusion (with attitudes of opposition to inclusion reflecting, in part, prejudiced attitudes concerning students with disabilities). However, we also reasoned that this relationship should be mediated by intergroup anxiety and stereotype use, with the effects of these variables being both direct and indirect–the latter occurring via perceived self-efficacy for teaching students with disabilities in the future.

2 Hypotheses

Based on our review, we were able to formulate several hypotheses for our study:

  1. 1.

    Self-efficacy for teaching students with disabilities will emerge as a significant negative predictor of attitudes of opposition to inclusion.

  2. 2.

    Negative stereotyping and intergroup anxiety will emerge as significant positive predictors of attitudes of opposition to inclusion. Moreover, the effects of these variables will be both direct and indirect (the latter effects being via self-efficacy).

  3. 3.

    SDO will emerge as a significant positive predictor of attitudes of opposition to inclusion, with its effects being both direct and indirect (the latter being via intergroup anxiety, stereotyping, and self-efficacy).

  4. 4.

    Contact will emerge as a significant negative predictor of attitudes of opposition to inclusion, with its effect being fully mediated via intergroup anxiety, stereotyping, and self-efficacy.

In addition to the abovementioned hypotheses, we also formulated a hypothesis concerning the relationship between intergroup anxiety and negative stereotyping. Specifically, we hypothesized that intergroup anxiety would emerge as a significant, positive predictor of negative stereotyping within our model. This prediction was made based on two theoretical arguments. First, the negative affective consequence (i.e., anxiety) surrounding anticipated intergroup contact with people with disabilities could lead one to rationalize his/her negative affect by consulting available negative stereotypes. Second, heightened intergroup anxiety may distract one’s attention and, thus, promote “reliance on stereotypes to guide judgments” (Aberson and Haag 2007, p. 183). To date, research supports this causal ordering (see e.g., Aberson and Haag 2007; Curseu et al. 2007).

3 Methods

3.1 Participants

Participants were 229 (Male \(\hbox {n}=35\), Female \(\hbox {n}=194\)) pre-service teachers enrolled in one of three sections of a survey course on special education at a mid-sized university in the Southwestern United States. Students enrolled in this course are typically in their junior or senior year of college. Participants’ ages ranged from 19 to 58 (Mean \(=\) 22.34, SD \(=\) 5.185). The racial/ethnic make-up of the sample was as follows: 84.7 % White, 4.8 % Native American or Alaska Native, 3.5 % African American, 3.5 % Hispanic/Latino, 3.5 % Multiracial and/or Multiethnic. Only 16 participants identified themselves as majoring in Special Education, whereas the remainder reported majoring in Early Childhood, Elementary, or Secondary Education.

3.2 Measures

3.2.1 Index of close contact

This index was comprised of six items asking each participant about the number of friends or family members he/she currently has who have disabilities (i.e., physical/orthopedic, intellectual, or learning disabilities), as well as the number of people he/she has been close to over his/her life who has had one or more disabilities. Participants responded using a five-point scale (i.e., \(1=\hbox {none}\), \(2=1\)–3, \(3=4\)–6, \(4=7\)–9, \(5=10\) or more), consistent with the scaling used by Binder et al. (2009). Higher scores on this index reflected greater prior or current exposure to close contact with persons with disability. Cronbach’s alpha for these nine items was .824.

3.2.2 Self-efficacy index

Self-efficacy for teaching students with disabilities was measured by three items developed for this study. Participants rated their level of agreement to the following items on a scale from 1 \(=\) strongly disagree to 7 \(=\) strongly agree: “I believe I am capable of instructing students with specific learning disabilities in my classroom.”; “I believe I am capable of instructing students with physical/orthopedic disabilities in my classroom.”; “I believe I am capable of instructing students with intellectual disabilities/mental retardation in my classroom.” Higher scores reflected greater levels of perceived efficacy. Cronbach’s alpha for these items was .831.

3.2.3 SDO scale

This scale measures a generalized orientation favoring group-based inequality and the desire for one’s own particular group(s) to dominate and be superior over others (Kteily et al. 2012; Pratto et al. 1994). Participants rated their agreement to 16 statements on a scale from 1 \(=\) strongly disagree to 7 \(=\) strongly agree. Example items include, “Some groups are simply inferior to other groups,” “It would be good if groups could be equal” (reverse-scored), and “It’s probably a good thing that certain groups are at the top and other groups are at the bottom.” Higher scores on the scale represented greater levels of social dominance orientation. Cronbach’s alpha for the sample was .875.

3.2.4 Intergroup anxiety index

Participants were presented with three scenarios, requiring them to imagine being in social situations (i.e., dinner parties) where most of the other individuals in those situations had some type of disability (i.e., intellectual disability; specific learning disability; physical/orthopedic disability). For each scenario, participants rated (on a scale from 1 \(=\) not at all to 7 \(=\) a great deal) the degree to which they believed they would experience the following emotions in those situations: “embarrassed,” “comfortable,” “anxious,” “at ease,” “awkward,” and “accepted”. Positive emotion items were recoded and averaged into the 18-item index along with the negative emotion items. Higher scores represented greater intergroup anxiety. Cronbach’s alpha for the items was .938.

3.2.5 Negative stereotyping index

Participants were asked to rate their level of agreement (on a scale from 1 \(=\) strongly disagree to 7 \(=\) strongly agree) with adjectives (“capable,” “weak,” “independent,” “aggressive”) describing individuals with physical/orthopedic disabilities, people with intellectual disabilities, and people with specific learning disabilities. After reverse-coding responses to “capable” and “independent,” items were averaged to produce an overall index of negative stereotyping. Cronbach’s alpha for the index was .897.

3.2.6 Opposition to inclusion scale

This nine-item scale was utilized to measure the degree to which participants are opposed to the practice of including students with disabilities in the classroom. Example items include, “Students with disabilities should be made to attend different classes from those who do not have disabilities,” “Students with disabilities have no business being included in classrooms with students who do not have disabilities,” and “Students with disabilities should be given the opportunity to learn in classrooms with students who do not have disabilities” (reverse coded). Higher scores on the measure reflect greater levels of opposition to inclusion. Cronbach’s alpha for this study was .827.

3.3 Procedure

Participants were recruited at the start of the semester to participate in a survey addressing various attitudes and beliefs related to disability and the practice of inclusion. Participants were offered course-credit for participation in the study. Those opting to participate completed an online survey containing the measures described above. Participants’ were informed that their responses would be confidential and anonymous.

4 Results

4.1 Data preparation and preliminary analyses

Prior to conducting our analyses, we used the regression imputation function in AMOS\(^{\mathrm{TM}}\) 17.0 to generate imputed values for those variables on which there were missing values.

Means and standard deviations for our variables are included in Table 1. On average, participants in our sample were low on SDO, stereotype usage, intergroup anxiety, close contact with people with disabilities, and opposition to inclusion, as reflected by the fact that mean scores on these variables were all below scale midpoints. Participants tended to be somewhat high on self-efficacy, as their average score fell above the scale midpoint on this variable. Although there were some departures from univariate normality on our variables, those departures were nonetheless fairly minor (see Kline 2005).

Table 1 Descriptive statistics of measured variables
Full size table

The correlations among our variables can be found in Table 2. Pre-service teachers scoring higher on SDO were more likely to indicate greater levels of stereotype use, intergroup anxiety, and opposition to inclusion, as well as lower levels of teacher efficacy. Intergroup anxiety was positively correlated with opposition to inclusion and stereotype use, and negatively correlated with intergroup contact and self-efficacy. Intergroup contact was positively associated with self-efficacy and negatively associated with opposition to inclusion and stereotype use. Pre-service teachers who scored higher on self-efficacy were less likely to oppose inclusion or to report stereotype use. Finally, pre-service teachers holding more stereotypes were more likely to oppose inclusion.

Table 2 Correlation analyses of variables
Full size table

4.2 Path analyses

We used AMOS\(^\mathrm{TM}\) 17.0 to carry out the path analyses in this study. Model fit was judged using the \(\chi ^{2}\)/df ratio, the Comparative Fit Index (CFI), and the Root Mean Square Error of Approximation (RMSEA). Good fitting models tend to exhibit \(\chi ^{2}\)/df ratios less than 5.0, CFI values \(>\).95, and RMSEA values at .05 or less (see Kline 2005; Schumacker and Lomax 2004). According to Kline (2005), RMSEA values between .05 and .08 can be considered acceptable.

Table 1 contains the fit statistics for our hypothesized model. Figure 2 contains the standardized path coefficients for the model. As can be seen, our hypothesized model exhibited a modest fit to the data. Although the \(\chi ^{2}\)/df ratio and CFI value fell within acceptable limits, the RMSEA was high. The direct effects of self-efficacy and SDO on opposition to inclusion in the model were statistically significant, with the former being negative and the latter being positive. Intergroup anxiety and negative stereotyping of people with disabilities were both negative predictors of teacher self-efficacy and positive predictors of opposition to inclusion. SDO was a positive predictor of both intergroup anxiety and stereotype use. Contact was a significant negative predictor of intergroup anxiety, but failed to significantly predict stereotype use. The correlation between the exogenous variables, contact and SDO, was slightly negative but not significant. Finally, intergroup anxiety positively predicted stereotype use.

Fig. 2
figure 2

Standardized path coefficients for re-specified model (Model 2). Note All path coefficients statistically significant at \(p < .05\)

Full size image

We re-specified our hypothesized model (see Fig. 2 for Model 2) by removing the correlation between SDO and contact and the direct path from contact to stereotype use. Moreover, we added a path from contact to teacher self-efficacy. Theoretically, this path can be argued on the grounds that greater levels of close contact with people who have disabilities may contribute to a heightened sense of understanding regarding the unique needs of students with disabilities. Prior and current opportunities to interact with people with disabilities may, therefore, lead pre-service teachers to feel more competent in their ability to meet students’ needs.

The re-specified model (Model 2) fit the data very well, as all fit indices fell within optimal limits. Stereotype use and intergroup anxiety emerged as significant negative predictors, whereas contact was a positive predictor of self-efficacy. As in Model 1, self-efficacy, SDO, intergroup anxiety, and stereotype use were significant predictors of opposition to inclusion. SDO and contact were significant predictors of intergroup anxiety, while SDO significantly predicted stereotype use in the model. Finally, intergroup anxiety was again a significant positive predictor of stereotype use.

4.3 Alternative models

We tested three additional models to address possible alternate paths among several variables. In Model 3, we altered the direction of causal influence between intergroup anxiety and stereotype use. Specifically, stereotype use was treated as a predictor of intergroup anxiety. This model was based on the possibility that cognitive stereotypes may increase the sense of anxiety one feels at the thought of interacting with people who have disabilities. The fit of this model was a poorer fit to the data than Model 2, as the value (Table 3) of the Aikaike Information Criterion (AIC) was larger for this model (see Kline 2005).

Table 3 Fit statistics for path models
Full size table

In Model 4, we re-specified the causal relationship between opposition to inclusion and self-efficacy as running from the former to the latter. In other words, the model assumes that pre-service teachers’ general opposition to inclusion may lead them to feel less efficacious about teaching students with disabilities. The fit of this model was very strong. Given the AIC for Model 4 was smaller than that of Model 2, it appears that this model is preferable to the re-specified version of the hypothesized model.

Model 5 represented an amalgamation of Models 3 and 4. Stereotype use was treated as a predictor of intergroup anxiety, whereas opposition to inclusion was a predictor of self-efficacy. This model represented a worse fit to the data than both Model 2 and Model 3.

Based on an examination of the relative fit of all our models, it appears that Model 4 represented the best overall fit to the data. In this model, all paths were statistically significant. See Fig. 3 for the standardized path coefficients for Model 4.

Fig. 3
figure 3

Standardized path coefficients for alternative model (Model 4) Notes. Path from Intergroup anxiety to Self-efficacy was significant at \(p = .044\) (one-tailed). All remaining paths are significant at \(p <.05\) (two-tailed)

Full size image

4.4 Tests of indirect effects

Given that Model 4 was the best fitting model based on our data, we report on the indirect effects tests from that model. We utilized non-parametric bootstrapping (drawing 1,000 samples) to compute standard errors and to carry out inferential tests for each indirect effect. Table 4 contains the unstandardized and standardized indirect effects in the model. All indirects in the model were statistically significant. SDO was a significant indirect predictor of self-efficacy and opposition to inclusion via opposition to inclusion and stereotype use. Moreover, part of the effect of SDO on efficacy flowed through opposition to inclusion. Contact indirectly predicted efficacy and opposition to inclusion via intergroup anxiety. As with SDO, part of the effect of contact was felt through opposition to inclusion. Intergroup anxiety indirectly predicted opposition to inclusion via the mediating factors, stereotype use and opposition to inclusion. The effect of negative stereotyping was mediated by opposition to inclusion.

Table 4 Unstandardized and standardized indirect effects (Model 4)
Full size table

5 Discussion

Many of the findings from this study were consistent with those laid out in our proposed causal model. To begin, our results provided strong evidence that SDO may be an important antecedent of pre-service teachers’ opposition to inclusion concerning students with disabilities. Specifically, we found that pre-service teachers who scored higher on SDO were more likely to oppose inclusion than those who scored lower on this factor. Consistent with our hypotheses, the effects of SDO on opposition to inclusion were both direct and indirect, the latter being mediated by intergroup anxiety and stereotype use. These results replicated previous findings concerning the direct and indirect effects of SDO and intergroup anxiety on opposition to inclusion in samples comprised of community adults and pre-service teachers (Brandes and Crowson 2009; Crowson and Brandes 2010).

Also consistent with our expectations, contact emerged as a significant, indirect, predictor of opposition to inclusion. Nevertheless, the proposed direct effect of contact on stereotype use in the model (see Fig. 1) was not supported by the data. Rather, the effect of contact on stereotype use appeared to be entirely mediated by intergroup anxiety. This in turn translated into an effect of contact on opposition to inclusion that was mediated proximally by intergroup anxiety and more distally by stereotype use.

As expected in our hypothesized model, we found that both intergroup anxiety and stereotype use concerning people with disabilities were significant predictors of teacher self-efficacy and opposition to inclusion in our data. Pre-service teachers scoring higher on intergroup anxiety or stereotype use tended to exhibit lower levels of self-efficacy or were more likely to oppose the practice of inclusion. Moreover, our findings supported the proposed causal relationship between intergroup anxiety and stereotype use. Models 2 and 4, specifying that intergroup anxiety predicts stereotype use, exhibited the strongest fit to the data in comparison to Models 3 and 5, specifying a reversed causal ordering. These results are consistent with previous studies (e.g., Aberson and Haag 2007; Curseu et al. 2007) that suggest affect concerning outgroups “drive” cognitions about those groups.

The predictive relationship between pre-service teachers’ self-efficacy for teaching students with disabilities and their opposition to inclusion attitudes remains, based on our findings, the largest question mark arising from our data. The re-specified version of our hypothesized model (i.e., Model 2) exhibited strong fit to the data, yet the direct effect of self-efficacy on opposition to inclusion was rather weak. Given the moderate zero-order correlation (\(-.\)347) between self-efficacy and opposition, it appears that within that model the effect was largely explained by the intergroup anxiety and stereotyping variables. Model 4, which specified self-efficacy as an outcome, as opposed to cause, of opposition to inclusion demonstrated an even better fit to the data than Model 2. Moreover, the path was of a slightly greater magnitude than that found with the aforementioned model. This suggests that among pre-service teachers, opposition to inclusion attitudes may actually cause lower levels of self-efficacy concerning students with disabilities. Perhaps pre-service teachers justify their opposition to inclusion attitudes by strategically discounting their ability to teach these students.

Clearly, if we assume that Model 4 represents the best fit to the data, then our first hypothesis that self-efficacy will be a negative predictor of opposition to inclusion attitudes is falsified. Nevertheless, given that the fit of Model 4 was only marginally better than Model 2, it is clear that additional tests of the predictive relationship between these two variables should be conducted.

5.1 Limitations and recommendations for future research

Several limitations should be considered with respect to our study. First, the data were cross-sectional in nature, thereby necessitating caution when interpreting the results. In theory, a number of different path models may fit the data as well (or even better) than our preferred model. Moreover, our ultimate selection of Model 4 as “best fitting” hinged on a number of model modifications and model comparisons that may have maximized the fit of that model to our sample data. As such, the reader is advised not to draw firm causal conclusions based on Model 4. Future research should test the proposed relationships in the model using other samples and with data that are longitudinal in nature.

A second potential limitation in our study concerns the generality of our contact, anxiety, stereotyping, efficacy, and opposition to inclusion measures. Although we found relationships among all of these variables within our sample, we were unable to discern whether the relationships might vary as a function of specific disability types of under consideration. For example, it may be the case that the relationship between self-efficacy and opposition to inclusion attitudes would have been larger had participants been asked to explicitly consider students with more, as opposed to less, challenging disabilities. Perhaps the weak relationship we observed between self-efficacy and opposition to inclusion within the current study was due to the generality of the measures. Additionally, the strength of the relationships between intergroup anxiety, stereotyping, and contact, on the one hand, and the efficacy and opposition to inclusion measures, on the other, might have been greater had all measures addressed attitudes and experiences concerning a specific disability category. The possibility that the generality of the abovementioned measures may have influenced the results should be considered in future studies on this topic.

A third potential limitation of the study concerns our conceptualization and measurement of contact. In our study, we tested whether contact is related to negative attitudes and opposition to inclusion, conceptualizing and measuring contact mostly in relation to friends and/or family members who have disabilities. Despite the fact that our study supports the claim that contact with people with disabilities fosters greater inclusivity, our results were nonetheless based on a fairly narrow view of contact—i.e., as it occurs within families and friendships. Clearly, this approach to measuring contact is consistent with Allport’s (1954) assertion that there are optimal conditions for contact (e.g. “equal status contact between majority and minority groups in the pursuit of common goals” and contact “that leads to the perception of common interests and common humanity” (p. 281). Whether simple exposure to persons with disability, irrespective of the kind of relationships that may be involved (e.g., acquaintance versus close-friend or family), results in lower levels of prejudice and opposition to inclusion among pre-service teachers remains an open question that should be addressed in future research. Moreover, the question of whether other, less direct forms of contact—as in the case of extended contact (Gómez et al. 2011) or imagined contact (Crisp and Husnu 2011; Vezzali et al. 2012)—reduces prejudice and opposition to inclusion are other areas that should be considered by researchers interested in this topic.

6 Conclusion

Our study addressed the question of whether SDO, close contact, intergroup anxiety, negative stereotyping, and self-efficacy predict opposition to inclusion among pre-service teachers. Our findings strongly supported much of the initial causal model. Based on our findings, we recommend that teacher educators acknowledge the potential role pre-service teachers’ ideologies, beliefs, contact experiences, and levels of intergroup anxiety may play in how they think about students with disabilities and their general willingness to incorporate these students into their future classrooms (Crowson and Brandes 2013). Acknowledgement of these factors may be useful in developing interventions and experiences (e.g., contact experiences, challenges to stereotypes) that increase pre-service teachers’ knowledge and understanding of students with disabilities and decrease any negative attitudes they may hold. Moreover, teacher educators who recognize these factors may be better equipped in their efforts to adapt persuasive appeals regarding the benefits of inclusion to the background of message recipients (i.e., pre-service teachers).

Although much of the hypothesized model was supported, there remained some ambiguity concerning the question of how self-efficacy and opposition to inclusion may be related. The original model treated self-efficacy as a predictor of opposition to inclusion, whereas the model with the strongest support included self-efficacy as an outcome of opposition to inclusion. In neither model was the relationship between self-efficacy and opposition to inclusion particularly strong. Much more research is needed to better flesh out how these two variables are related.