Challenging Two Mental Illness Stigmas: Personal Responsibility and Dangerousness by Patrick W. Corrigan, David Rowan, Amy Qreen, Robert Lundin, Philip River, Kyle Uphoff'Wasowski, Kurt White, and Mary Anne Kubiak Abstract Two stigmatizing attitudes related to dangerousness and personal responsibility may undermine the oppor- tunities of persons with serious mental illness. This study set out to examine path models that explain how these attitudes lead to discriminatory behavior and to assess the impact of antistigma programs on compo- nents of personal responsibility and dangerousness models.
Two hundred thirteen persons were randomly assigned to one of five antistigma conditions: education on personal responsibility, education on dangerousness, contact with a person with serious mental illness where personal responsibility is discussed, contact where dan- gerousness is discussed, or no change. Persons com- pleted an attribution questionnaire (AQ) representing personal responsibility and dangerousness path models at pretest, posttest, and 1-week followup. They also completed tasks that represented helping behavior.
Goodness of fit indexes from linear structural modeling were mixed for both models but suggested that fear of dangerousness was a key attitude leading to discrimi- natory behavior. Results also showed that subjects who had contact with persons with serious mental illness experienced greater changes than subjects in ... more. less.
the edu- cation or control groups did on measures of attribution and helping behavior. Keywords: Stigma, responsibility, dangerousness, discrimination, stigma change.<br><br> Schizophrenia Bulletin, 28(2):293-309, 2002. The opportunities provided to people with serious mental illnesses like schizophrenia are significantly diminished by public stigma.* Members of the general public are less * We have distinguished between public stigma (ways in which mem- bers of the general public show prejudice and discrimination against people with serious mental illness) and self-stigma (ways individuals with serious mental illness turn against themselves as a result of living in a culture where mental illness stereotypes prevail [Corrigan & Watson, in press]). likely to hire people who are labeled mentally ill (Olshansky et al.<br><br> 1960; Farina and Felner 1973; Bordieri and Drehmer 1986; Link 1987), less likely to lease them apartments (Page 1977, 1983, 1995), and less likely to freely interact with them (Martin et al. 1999). Public stigma is a fairly common phenomenon.<br><br> Studies show that many citizens in the United States (Rabkin 1974; Roman and Floyd 1981; Link 1987; Phelan et al. 1999) and most Western nations (Greenley 1984; Madianos et al. 1987; Bhugra 1989; Brockington et al.<br><br> 1993; Hamre et al. 1994) endorse stigmatizing attitudes about mental ill- ness. One might think that negative attitudes have improved with greater public awareness about mental ill- ness.<br><br> Unfortunately, research suggests the contrary 4 namely, that stigmatizing attitudes have actually worsened over the past 30 years (Pescosolido et al. 1999). Given the negative impact of stigma on persons with serious mental illness, we have two goals in this article.<br><br> First, in study 1, we will develop and evaluate two mod- els 4personal responsibility and dangerousness 4that describe the relationship between stigmatizing attitudes and discriminatory behavior (such as refusing to hire or rent to people with serious mental illness). This kind of research is essential for the development of antistigma programs. The second part of this article, study 2, exam- ines the effects of several stigma-challenging strategies that have been developed, in part, based on the two mod- els developed in study 1.<br><br> Two Models of Mental Illness Stigma Bernard Weiner (1995) developed a model of causal attri- bution that at least partly explains the relationship between stigmatizing attitudes and discriminatory behav- ior. As outlined in figure 1, Weiner believed that attribut- Send reprint requests to Dr. P.<br><br> Corrigan, University of Chicago Center for Psychiatric Rehabilitation, 7230 Arbor Drive, Tinley Park, IL 60477; e-mail: email@example.com. 293 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol. 28, No.<br><br> 2, 2002 P.W. Corrigan et al. Figure 1.<br><br> Hypothetical paths accounting for stigmatizing reactions. Path A represents relationships between attributions of personal responsibility for mental illness, subsequent pity or anger, and the effects of this pity or anger on helping behavior. Path B represents attributions of dangerousness, subsequent fear, and avoidant behavior.<br><br> Path A personal responsibility helping behavior PathB ing personal responsibility for a negative event (e.g., "That person causes his crazy behavior") leads to anger ("I'm sick and tired of that kind of irresponsibility!") and diminished helping behavior ("I'm not going to give him a ride"). Conversely, attributing no blame for a harmful event ("He can't help himself; he's mentally ill") leads to pity ("That poor man is ravaged by mental illness") and the desire to help ("I'll rent him a room until he's back on his feet"). What makes these behaviors discriminating?<br><br> One definition of discrimination is either withholding opportunities from or reacting punitively to someone solely because he or she is a member of a stigmatized out- group (Crocker et al. 1998). These outgroups have included ethnic minorities, women, and people with dis- abilities (including people with serious mental illness).<br><br> Withholding opportunities parallels the refusal of helping behavior in Weiner's model. Are attitudes in general, and attributions in particular, correlated with behavior? After a comprehensive review of the literature, Petty and Cacioppo (1996) noted that the answer to this question is "yes" when attitudes and behav- iors are measured appropriately (Azjen and Fishbein 1977), when attitudes are based on direct experience (Fazio and Zanna 1981), and when behaviors require a preceding deliberative process to initiate (Triandis 1977).<br><br> The specific association between causal attribution, medi- ating anger or pity, and subsequent behavior has been vali- dated in several samples (Weiner et al. 1982; Reisenzein 1986; Schmidt and Weiner 1988; Weiner et al. 1988; Zucker and Weiner 1993; Dooley 1995; Graham et al.<br><br> 1997; Menec and Perry 1998; Rush 1998; Steins and Weiner 1999). Notable among these studies were investi- gations by Reisenzein (1986) and Steins and Weiner (1999) that examined the causal relationships in Weiner's 294 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol. 28, No.<br><br> 2, 2002 model using latent-variable structural modeling tech- niques. We have adopted this methodological strategy and statistical approach to evaluate whether Weiner's attribu- tion model explains the relationship among judgments about a person's responsibility for his or her mental illness, subsequent pity or anger, and discriminatory behavior. Responsibility attributions seem to make sense for explaining the relationship between stigma and discrimina- tory behavior.<br><br> However, these kinds of attributions markedly differ from the typical attitudes about mental illness that emerge in factor analyses of public stigma 4namely, that peo- ple with serious mental illness need to be segregated from society because they are dangerous (Cohen and Struening 1962; Taylor and Dear 1981; Brockington et al. 1993; Link et al. 1999; Pescosolido et al.<br><br> 1999). We have outlined one spec- ulative model elsewhere (Corrigan 2000) and repeat it at the bottom of figure 1. According to this model, attributing a per- son's behavior as dangerous leads to fear; most people respond to violent threats of any kind with apprehension (Johnson-Dalzine et al.<br><br> 1996). Several studies have found a specific relationship between perceiving persons with serious mental illness as dangerous and fearing them (Link and Cullen 1986; Levey and Howells 1995; Angermeyer and Matschinger 1996; Wolff et al. 1996).<br><br> Fear about a person's dangerousness, in turn, yields avoidant behaviors. For exam- ple, one study has shown that a fearful reaction to two politi- cal assassination attempts attributed to persons with schizo- phrenia led to greater social distance between the public and the community of individuals with serious mental illness (Angermeyer and Matschinger 1996). This historical finding has been supported by other research 4namely, perceptions of dangerousness led to avoidance of persons with serious mental illness (Madianos et al.<br><br> 1987; Levey and Howells 1995). Hence, employers failed to hire persons with serious mental illness and landlords did not permit people with psy- chiatric disabilities to move into the landlords' properties. This model is a variation of what Weiner (1995) called primary appraisal: An emotional response like fear yields a behavioral outcome (avoidance) without a mediating attribu- tion.<br><br> In fact, another study looked at similar variables (Boisvert and Faust 1999): A significant relationship was found between violence in mental illness and attributions about the locus of causality (internal vs. external locus of con- trol). Using latent-variable structural modeling techniques, we expect to find the causal relationships outlined at the bottom of figure 1.<br><br> Changing Attitudes About Dangerousness and Personal Responsibility Three strategies relevant to changing mental illness stigma have been gleaned from the social psychology lit- erature: protest, education, and contact (Corrigan and Penn 1999). Protest seeks to suppress stigmatizing atti- tudes about mental illness and behaviors that promote these attitudes through appeals to moral indignation: "Shame on you for thinking about mental illness in a dis- respectful way." Education replaces stigma with more accurate conceptions about the disorders through didactic programs. Contact challenges public attitudes about men- tal illness through direct interactions with people who have these disorders.<br><br> The relative impact of these stigma change strategies was examined in an earlier randomized controlled study (Corrigan et al. 2001c). Results found that protest yielded no significant change in stigmatizing attitudes.<br><br> This is consistent with a suppression rebound effect found in other social psychological research 4that is, trying to sup- press negative attitudes about a group fails to diminish stereotypes but instead, ironically, leads to greater recol- lection of those attitudes (Macrae et al. 1994, 1996). Participation in the education condition led to significant improvement in stigmatizing attitudes.<br><br> Assignment to the contact condition, however, yielded even greater changes than education. In the study discussed in this article, we repeated the education and contact conditions to determine their effects on the elements in figure 1. We blended these processes with content areas that address attitudes that people with serious mental illness are dangerous or responsible for their disabilities.<br><br> Consistent with other research on educa- tion, our didactic program included direct challenges to expectations that persons with serious mental illness are dangerous (Holmes et al. 1999; Penn et al. 1999).<br><br> The contact condition also reviewed misconceptions about dangerousness. We expected these programs to have a marked impact on attributions of dangerousness, conse- quent fear, and avoidance behavior. In addition, we exam- ined the effects of education and contact programs that discuss personal responsibility for mental illness.<br><br> Study 1 Method. In study 1, we examined the path models that represent the impact of judgments of responsibility and dangerousness on behavior. Research participants were drawn from the at-large student body of a local commu- nity college.<br><br> We have used community college students in past research because they tend to be more demographi- cally representative of the population as a whole than are college sophomores from 4-year universities (Corrigan et al. 1999, 2001c). Two hundred sixteen students were informed of the study and asked to participate; 213 agreed and completed all measures.<br><br> Demographics for this sam- ple are summarized in table 1. The sample was almost 295 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol. 28, No.<br><br> 2, 2002 P.W. Corrigan et al. three-quarters female and more than 80 percent unmar- ried.<br><br> Ethnic representation was fairly mixed 4about 40 percent European-American and 40 percent African- American; this diversity is important because other inves- tigations have found that ethnicity predicts prejudice against mental illness (Corrigan et al. 2001a). The sample also showed a wide range of income and education levels.<br><br> Research participants were administered the AQ modeled after Reisenzein (1986) and shown in the appen- dix. Two to three items each sampled the seven constructs in figure 1 (e.g., "How frightened of a person with mental illness would you feel?"). Research participants responded to the items using a nine-point Likert scale.<br><br> Eleven items were taken from Reisenzein 4to assess per- sonal responsibility, pity, anger, and helping behavior 4 with two minor changes. Instead of rating reactions to a person described in test scenarios, research participants were asked to report their attributions about persons with serious mental illness in general. One question from Reisenzein specifically asked the type of helping behavior that research participants might provide: "help in any way Table 1.<br><br> Demographic characteristics participants in this study, n = 213 of Characteristic Age, mean (SD, range) Gender, % female Marital status, % Single Married Separated, divorced, or widowed Ethnicity, % European-American African-American Latino Other Household income, % < 20,000 20,000-39,999 40,000-59,999 60,000-80,000 > 80,000 Education, % High school Some college Associate's degree Bachelor's degree Some graduate school 26.3(12.2, 18-80) 70.4 81.5 10.6 7.9 43.9 41.5 8.5 6.1 31.0 21.1 23.9 16.9 7.0 14.0 67.8 10.3 2.8 4.3 Note. 4SD = standard deviation. Household income percentages do not add up 100% due to rounding.<br><br> that might be necessary, including if necessary first aid" (Reisenzein 1986, p. 1126). We replaced this question with an item relevant to mental illness: "How likely is it that you would help a person with a mental illness?" In a similar manner, three AQ items were generated for each of the constructs in the bottom half of figure 1: danger- ousness ("I would feel unsafe around persons with serious mental illness"), fear ("Persons with serious mental illness terrify me"), and avoidance ("I would try to avoid a person with mental illness").<br><br> These items were also answered using nine-point Likert scales. Research participants were admin- istered AQ items in one of two random orders. Results.<br><br> Path analysis with latent variables was used to test the theoretical model outlined in figure 1 because it is one of the more robust measures of both the size and direc- tion of associations among a set of variables. An explicit distinction is made between latent variables or constructs and manifest variables or the indicator of these constructs; typically, multiple indicators (in this study, individual items of the AQ) are used for each latent variable. Results of these analyses are summarized in the top half (for per- sonal responsibility) and the bottom half (for dangerous- ness) of figure 2.<br><br> In accord with conventional practices in the causal modeling literature (Bentler 1988), circles were used in figure 2 for latent factors and squares designated manifest indicators (individual AQ items). Unidirectional arrows between circles represent causal paths. All analyses were conducted using the SAS System's CALIS procedure (SAS Institute 1990), and the tested models were covariance structure models with multiple indicators for all latent constructs.<br><br> Standard deviations and intercorrelations for the study's manifest variables are summarized in tables 2 and 3. This correlation matrix served as input to the CALIS procedure. The present analysis followed a two-step procedure based in part on an approach recommended by Anderson and Gerbing (1991).<br><br> In the first step, confirmatory factor analysis (CFA) was used to develop a measurement model that demonstrated an acceptable fit to the data 4that is, the AQ items fit into the hypothetical constructs outlined in figure 1. Note that results of the CFA provide psycho- metric support for the AQ. In the second step, the mea- surement model was modified so that it came to represent the theoretical (or path) model of interest.<br><br> We review the measurement and theoretical models for personal respon- sibility and dangerousness separately. Personal Responsibility Model Measurement model. In path analysis with latent variables, a measurement model describes the nature of the relationship between a number of latent variables (or factors) and the manifest indicator variables that hypo- 296 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 Figure 2. Standardized path coefficients for the personal responsibility and dangerousness models.<br><br> These coefficients appear on single-headed straight arrows, t values corresponding with starred (*) coefficients were greater than 1.96 (p < 0.05); ***p < 0.0001. AQ 1 AQ 2 personal responsibility AQ 9 I c ^ 4 AQ 10 1 4 4 helping behavior __- ~-\ t AQ 6 t i AQ 7 - ^ t AQ 8 AQ 12 I AQ 13 f AQ 14 1 thetically comprise the latent factors. The first model from our data comprised four latent constructs: personal responsibility, pity, anger, and helping behavior.<br><br> Measurement models were estimated using the maximum likelihood method. The chi-square value representing goodness of fit for the model was statistically significant, x 2 (38) = 131.2, p < 0.0001; statistically significant findings do not support a measurement model. Typically, the chi-square statistic is used to test the null hypothesis so that the model fits the data when proper assumptions are met.<br><br> In practice, however, the statistic is very sensitive to sample size and departures from multivariate normality and may often result in the rejection of a well-fitting model. A number of other results, however, indicated further problems with this model's fit. Specifically, the pattern of large normalized residuals, parameter signifi- cance tests, and Lagrange multiplier tests showed that one manifest indicator, AQ11 (concern), was apparently affected by both an alternative construct (pity) and the construct that it was expected to influence (helping behavior).<br><br> As a result, the model resulting from this CFA did not fit significantly with our data. Because AQ11 was apparently a multidimen- sional variable, it was eliminated and the measurement model was reestimated. Goodness of fit indexes for the respecified measure- ment model are listed on line 6 of table 4.<br><br> The revised measurement model displayed values greater than 0.9 on the normed fit index (NFI), non-normed-fit index (NNFI), and comparative fit index (CFI), all indicative of an acceptable fit (Bentler and Bonnett 1980; Bentler 1989). Another way to appreciate the size of the fit is to compare these indexes with the values provided by the uncorre- lated factors model (line 4 of table 4). In all cases, the measurement model provided greater values.<br><br> Hence, this model was accepted as the study's final measurement model for personal responsibility. Theoretical model. A path analysis of the theoretical model was completed next, omitting AQ11 from the 297 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 P.W. Corrigan et al.<br><br> Table 2. Standard deviations and intercorreiations for manifest variables: personal responsibility model AQ1 AQ2 AQ3 AQ4 AQ5 AQ6 AQ7 AQ8 AQ9 AQ10 AQ11 (controllable) (responsibility) (sympathy) (P'ty) (sorry for) (angry) (irritated) (aggravated) (certainty) (amount) (concern) SD 1.95 2.08 2.76 2.20 2.60 1.59 1.41 1.57 1.95 1.96 1.97 AQ1 1.00 0.35 0.00 0.08 -0.03 0.10 0.05 0.02 0.07 0.07 0.07 AQ2 1.00 -0.02 -0.05 -0.06 0.11 0.10 0.08 0.01 -0.01 -0.04 AQ3 1.00 0.45 0.64 0.26 0.29 0.30 0.03 -O.03 0.22 AQ4 1.00 0.58 -0.02 0.01 0.05 0.33 0.30 0.47 AQ5 1.00 0.16 0.20 0.24 0.05 0.01 0.40 AQ6 1.00 0.47 0.72 -0.19 -0.22 -0.11 AQ7 1.00 0.58 -0.06 -0.14 -0.05 AQ8 1.00 -0.14 -0.21 -0.07 AQ9 AQ10 AQ11 1.00 0.89 1.00 0.34 0.30 1.00 Note. 4AQ = attribution questionnaire; SD = standard deviation.<br><br> Table 3. Standard deviations and intercorreiations for manifest variables: dangerousness model AQ12 (unsafe) AQ13 (dangerous) AQ14 (threatened) AQ15 (terrify) AQ16 (scared) AQ17 (frightened) AQ18 (hospitalize) AQ19(avoid) AQ20 (don't rent to) SD 2.03 2.02 1.87 1.65 1.85 1.86 1.92 2.42 2.22 AQ12 1.00 0.52 0.67 0.69 0.67 0.68 0.55 -0.36 0.33 AQ13 1.00 0.66 0.48 0.60 0.62 0.50 -0.29 0.20 AQ14 1.00 0.72 0.85 0.85 0.50 -0.42 0.17 AQ15 1.00 0.71 0.73 0.44 -0.37 0.21 AQ16 1.00 0.86 0.47 -0.39 0.22 AQ17 1.00 0.45 -0.38 0.20 AQ18 1.00 -0.34 0.27 AQ19 AQ20 1.00 -0.28 1.00 Note. 4 AQ = attribution questionnaire; SD = standard deviation. analyses in the process.<br><br> Goodness of fit indexes for the personal responsibility theoretical model are provided on line 5 of table 4 and yield mixed support for the model. Values on the NNFI, NFI, and CFI all exceeded 0.9, thereby supporting the data's fit to the model. Another way to determine fit is to compare CFI and NNFI indexes from our data to findings from the path analyses on per- sonal responsibility conducted in earlier studies (lines 1 and 2 of table 4).<br><br> A comparison shows that the CFI from our model was greater than that reported by Steins and Weiner (1999) but the NNFI was less than results from Reisenzein (1986). Moreover, the chi-square to df ratio was greater than 2 (not supporting goodness of fit). Figure 2 summarizes standardized path coefficients for the theoretical model.<br><br> Only the relationship between anger and helping behavior yielded a significant coefficient in the appropriate direction. The other three were nonsignificant. Dangerousness Model.<br><br> Indexes for the theoretical and measurement models for the dangerousness model are also summarized in table 4 and figure 2. Overall, results suggest data fit this model far better than the personal responsibility model. We review the measurement and theoretical models in turn.<br><br> Measurement model. We assumed that the first model for dangerousness comprised three latent constructs: danger- ousness, fear, and avoidance behavior. The chi-square value for the model was statistically significant, X 2 (24) = 64.5, p < 0.05, thereby not supporting goodness of fit.<br><br> However, the NFI, NNFI, and CFI were all above 0.9 (see line 10 of table 4). Moreover, these indexes were well above the NFI, NNFI, and CFI generated by the uncorrelated factors model (line 8, table 4). In addition, the pattern of large normalized residuals and Lagrange multiplier tests failed to identify a manifest indicator that did not fit with its corresponding latent factor.<br><br> Hence, this model was accepted as the study's final measurement model for dangerousness. Theoretical model. A path analysis of the theoretical model was completed next.<br><br> Goodness of fit indexes for the dangerousness theoretical model (line 9, table 4) pro- vided mixed support for the model. Values on the NNFI, NFI, and CFI all exceeded 0.9, supporting goodness of fit. These values also mostly exceeded the corresponding indexes from the studies by Reisenzein (1986) and Steins 298 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 Table 4. Goodness of fit indexes for the personal responsibility and dangerousness model studies, n = 213 Model Reisenzein (1986) model 1.<br><br> Theoretical model Steins and Weiner (1999) model 2. Theoretical model Path A. Personal responsibility 3.<br><br> Null model 4. Uncorrelated factors 5. Theoretical model 6.<br><br> Measurement mode Path B. Dangerousness 7. Null model 8.<br><br> Uncorrelated factors 9. Theoretical model 10. Measurement model Chi- square 82.2 21.1 934.9 104.5 86.7 71.7 1292.5 517.9 74.6 64.5 df 60 3 45 35 31 29 36 27 22 24 NFi 4 4 0.000 0.887 0.907 0.923 0.000 0.599 0.942 0.950 NNFI 0.944 4 4 0.898 0.909 0.926 4 0.479 0.931 0.951 CFI 4 0.880 4 0.921 0.937 0.952 4 0.609 0.958 0.968 Note.<br><br> 4CFI = comparative fit index; NFI = normed fit index; NNFI = non-normed-fit index. Goodness of fit indexes for structural and combined models are listed independently for the two paths. and Weiner (1999) as well as the values from the theoreti- cal model for personal responsibility.<br><br> However, the chi- square to df ratio exceeded 2, thereby not supporting a good fit. All the standardized path coefficients for the the- oretical model (in figure 2) yielded significant coefficients (p < 0.001) in the appropriate direction. Summary.<br><br> Results were mixed in terms of the path model for personal responsibility. The final measurement model supported the factor structure for the four latent constructs 4personal responsibility, pity, anger, and help- ing behavior 4and goodness of fit indexes mostly sup- ported the model. However, only one of the four standard- ized path coefficients was significant; persons who were high in anger were less likely to provide help.<br><br> None of the other relationships were supported. These findings were somewhat sobering given earlier research by Reisenzein (1986) that provided clear support for this model. Does this suggest the model does not fit for mental illness stigma, or might some methodological differences between the studies account for this discrepancy?<br><br> In part, this difference may be due to statistical limitations. The final measurement model for our study did not yield a result with three manifest indicators per latent variable; this shortfall frequently undermines goodness of fit (Hatcher 1994). Moreover, the task in this study differed from the procedure in the investigation by Reisenzein.<br><br> Participants in the earlier study were instructed to rate their attributions, anger, pity, and help in response to a vignette about a specific person. In our study, research participants provided ratings based on their overall impressions of persons with serious mental illness. Perhaps this task, which is more abstract, diminished the relationships that result from attributions about real peo- ple with whom the person interacts.<br><br> Results for the dangerousness model seemed far more consistent. The CFA yielded a factor structure that fit the data well and failed to identify indicator items needing to be dropped from the model. Like the earlier model, indexes representing goodness of fit were mostly support- ive.<br><br> More important, the three standardized path coeffi- cients were highly significant in the predicted direction. These results seem to support our contention that danger- ousness evokes fear, which in turn promotes avoidance of the person perceived to be dangerous: the person with serious mental illness. Study 2 Method.<br><br> In the second study, we designed a 2 (processes: contact vs. education) X 2 (contents: dangerousness vs. personal responsibility) study to test the effects of various stigma change strategies on stigmatizing attributions and subsequent behaviors.<br><br> The same sample described in study 1 participated in study 2. Participants were ran- domly assigned to one of five groups: education on per- sonal responsibility, education on dangerousness, contact with a person with serious mental illness where personal responsibility is discussed, contact where dangerousness is discussed, or no change (the control group). A single leader conducted each condition with four to eight partici- pants in a quiet room with no distractions.<br><br> Each program 299 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol. 28, No. 2, 2002 P.W.<br><br> Corrigan et al. included two parts: a 10-minute presentation immediately followed by a 5-minute discussion. Earlier research has shown that short stigma change programs lead to signifi- cant improvement in attitudes (Penn et al.<br><br> 1994, 1999; Corrigan et al. 2001c). The no-change control presenta- tion concerned hobbies and technology in the 1990s and discussed no issues related to mental illness or physical disability.<br><br> The key ingredient to education programs is contrast- ing myths about mental illness with information that chal- lenges these myths (Corrigan and Penn 1997). Hence, the education presentation reviewed seven myths drawn from the literature (Harding and Zahniser 1994; Penn et al. 1999) and facts that challenge these myths.<br><br> The educa- tion/dangerousness condition included discussions about the relationship between psychoses and violence (Monahan 1992). The education/personal responsibility condition replaced discussion of dangerousness with con- sideration of the controllability of serious mental illness adapted from Weiner (1995). Leaders specifically paired myth (e.g., "Persons with serious mental illnesses like schizophrenia are violent and should be avoided") with research findings (e.g., "Most persons with serious mental illness are no more violent than the average citizen").<br><br> Four group leaders were trained to provide the two education conditions and the one control condition for the study. The scripts for these presentations were written out and read verbatim by the leader. The presentation for the two education and control conditions included slides to illustrate key points.<br><br> The leaders were also provided a set of standardized open-ended questions about the corre- sponding presentation to facilitate discussion. These four leaders rotated through all conditions and were assigned using a Latin square. Post hoc analyses failed to find a leader effect.<br><br> Research participants were also randomly assigned to one of two contact conditions with content that focused on dangerousness or personal responsibility. Research participants in these conditions listened to a 10-minute presentation by one of two persons who discussed their serious mental illness. Both persons were selected because they had at least a 10-year history of psychotic symptoms, suicide attempts, hospitalization for bipolar disorder, and long recovery periods.<br><br> Both persons have made a satisfactory recovery, live independently, work, and report a satisfactory quality of life. These two persons rotated across contact conditions for different cohorts. Research completed on other stigmatized groups has identified several factors that augment the impact of con- tact on public attitudes (Stephan 1987); they were incor- porated into the contact condition.<br><br> Contact effects are enhanced when members of the public meet persons who mildly disconfirm the stereotype (Weber and Crocker 1983; Johnston and Hewstone 1992). Mild disconfirma- tion was communicated to research participants in terms of current status for both contact persons 4that is, despite successful outcomes, the two contact group leaders still struggled with recurring symptoms and discussed this struggle in their presentation. Members of the public tend to disbelieve contacts who grossly disconfirm a stereo- type: "That guy was not really mentally ill." Finally, the effects of contact are facilitated when participants are able to positively interact with contacts (Johnson et al.<br><br> 1984; Worchel 1986); hence, contact group leaders facilitated a 5-minute discussion in which participants questioned the people with mental illness about their experiences. Dependent measures. Completion of the AQ for study 1 was part of the pretest battery for study 2.<br><br> In addi- tion, research participants were administered the Social Distance Scale (SDS) at pretest. The AQ and SDS were readministered immediately following the stigma condi- tion. In addition, research participants were invited to return 1 week later to complete the measures at followup.<br><br> Research participants were debriefed about the study after the third test administration. The SDS is considered a proxy measure of behavior related to attitudes about mental illness 4that is, it repre- sents verbal reports about how research participants might interact with person with serious mental illness (Corrigan et al. 2001c).<br><br> The SDS comprises seven questions about interactions with people with serious mental illness (e.g., "How would you feel about renting a room in your home to a person with severe mental illness?"). Persons respond to each item using a 0 to 3 willingness scale (3 = defi- nitely unwilling). Internal consistency for the SDS total score derived from this sample was satisfactory (a = 0.76).<br><br> In addition to verbal statements about behavior, we wanted to assess direct behavior related to helping per- sons with serious mental illness. Individuals participating in the study were reimbursed $20 at posttest for their effort. After completing all posttests, research participants were instructed in a brief note that they could donate any or all of this money to the National Alliance for the Mentally 111 of the South Suburbs (NAMI-SS); partici- pants were provided a brief description of the advocacy and education efforts of NAMI-SS.<br><br> Research participants were given a receipt on which they could specify the number of dollars they wished to donate. Total donated dollars is an index of helping behavior. Results.<br><br> To determine the effects of randomizing research participants to condition, the frequency of demo- graphic variables summarized in table 1 was examined across the five antistigma conditions; no significant differ- ences were found across groups (p > 0.05). Moreover, no 300 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol. 28, No.<br><br> 2, 2002 significant differences were found in dependent measure pretest scores across groups (p > 0.40). The number and kind of analyses that were possible to determine the effects of the antistigma strategies were fairly complex, as evidenced by tables 5 and 6. The analy- ses are organized into three stages.<br><br> First, we examined the effects of the five conditions on AQ factors and SDS score from pre- to posttest. This section includes a posttest-only examination of the effect of experimental conditions on NAMI-SS donations. Second, we examined the specific effects of two processes by two content areas.<br><br> Finally, we completed followup analyses examining the impact of the five conditions on AQ and SDS variables 1 week later. 1. Pretest and posttest analyses.<br><br> Pre-post differ- ences in dependent variables across conditions were assessed using 2 X 5 (trial by condition) analyses of variance (ANOVAs); results of these analyses are sum- marized in table 5. Significant interactions were found for all AQ factors except pity. Post hoc tests were mixed but generally found that the two contact conditions plus education/personal responsibility led to more significant improvement in AQ scores.<br><br> A significant interaction was also found for the SDS. Participants in the control condition showed less change than those in the other four antistigma conditions. Finally, results of a one-way ANOVA for the amount of money donated to NAMI-SS were significant; both contact conditions and educa- tion/personal responsibility yielded greater helping behavior.<br><br> 2. Content by process interactions. The research question of specific interest in this article was the interac- tion of two different antistigma intervention processes (education vs.<br><br> contact) with two different intervention con- tents (dangerousness vs. responsibility). A series of 2 X 2 X 2 (trial by process by content) ANOVAs were completed to evaluate this question and are summarized in table 6.<br><br> Significant interactions were found for two of the four AQ factors that correspond with the personal responsibility model. Attributions related to the responsibility factor improved significantly for antistigma interventions that included responsibility as content; this serves as a manipu- lation check. Contact also significantly improved the AQ anger scale over education from pre- to posttest.<br><br> All three AQ factors comprising the dangerousness model showed significant trial by interventions interac- tions. Dangerousness, fear, and avoidance factors all sig- nificantly improved after contact (compared to education). Moreover, a significant interaction was found for the SDS, with better results caused by contact.<br><br> Finally, a 2 X 2 ANOVA (process by content) was completed, with NAMI-SS donations as the dependent variable. A signifi- cant interaction was found with contact, which led to greater donations than education. Followup analyses.<br><br> About half the participants (n = 97) returned 1 week later to complete the followup mea- sures. Differences were examined between the followup group and those research participants who did not return to determine whether the followup group was distinct from the entire sample participating in the study. None of the demographic measures differed significantly across groups (p > 0.15).<br><br> All but one of the pretest measures failed to differ significantly across groups; the followup group was higher only on AQ responsibility ratings at pretest (F[l,202] = 6.78, p < 0.01). Moreover, results of 2 X 2 ANOVAs (followup status [yes or no] by trial [pre- vs. post-]) for all the dependent measures in the study were nonsignificant, showing that the two followup groups did not differ in the size of change scores between pre- and posttesting.<br><br> Hence, the followup group members appeared to be similar to those who did not return 1 week later to retake the test battery. Table 5 lists the results of 2 X 5 (trial by condition) ANOVAs examining effects at followup. The middle col- umn of table 5 represents change from pretest to fol- lowup.<br><br> Significant interactions were found for two of four AQ factors in the personal responsibility model: pity and anger. Post hoc tests showed that only contact conditions accounted for this significant change. Two of three AQ factors that composed the dangerousness model showed significant change from pretest to followup: fear and avoidance.<br><br> Post hoc tests showed that this change was only due to one or both of the contact conditions. A signif- icant interaction was also found for the SDS from pretest to followup; change was due to only the contact condition with personal responsibility content. Table 5 also includes the results of 2 X 5 ANOVAs representing changes from posttest to 1 week followup.<br><br> In all but one dependent variable, no significant interactions were found, suggesting that improvement in attributions did not return to baseline during this time. As in the analysis of the pre-post data, 2 x 2 X 2 (process by content by trial) analyses were conducted to determine respective effects from pretest to followup for stigma change processes and content. These results are summarized in table 6.<br><br> In most cases, significant interac- tions were found between antistigma processes and trial, with better improvement because of contact compared to education. Three of four AQ factors in the personal responsibility model yielded significant results. Pity changed significantly from pretest to followup as a result of content; the content conditions that included personal responsibility yielded better results.<br><br> Significant interac- tions were found for anger and help, with contact yielding greater change. Moreover, significant interactions were found for all three AQ factors of the dangerousness model, with each factor yielding greater change because of con- 301 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Table 5. Summary of inferential statistics representing change in attributions and social distance, and NAMI-SS donation behavi ors as a result of five stigma change strategies I Variables Responsibility Group effect Trial effect Interaction Post hoc tests Pity Group effect Trial effect Interaction Post hoc tests Anger Group effect Trial effect Interaction Post hoc tests w Hel P S Group effect Trial effect Interaction Post hoc tests Dangerousness Group effect Trial effect Interaction Post hoc tests Fear Group effect Trial effect Interaction Post hoc tests Avoidance Group effect Trial effect Interaction Post hoc tests Pre vs.<br><br> post F(4,206) = 3.40, p < 0.05 F(1,206) = 2.59, ns F(4,206) = 7.28, p< 0.0001 CTL = EDUd = CONd < EDUr = CONr F(4,206) = 0.76, ns F(1,206) = 1.60, ns F(4,206) = 0.88, ns NA F(4,206) = 1.63, ns F(1,206) = 35.95, p< 0.0001 F(4,206) = 4.45, p< 0.005 CTL = EDUd < EDUr = CONd = CONr F(4,206) = 1.56, ns F(1,206) = 27.51, p< 0.0001 F(4,206) = 3.98, p < 0.005 CTL < EDUd = EDUr = CONd = CONr F(4,206) = 1.27, ns F(1,206) = 85.97, p< 0.0001 F(4,206) = 4.98, p < 0.001 CTL = EDUr < EDUd = CONd = CONr F(4,206) = 0.93, ns F(1,206) = 32.34, p< 0.0001 F(4,206) = 4.43, p< 0.005 CTL = EDUr < EDUd = CONd = CONr F(4,206) = 1.49, ns F(1,206) = 25.15, p< 0.0001 F(4,206) = 4.16, p< 0.005 CTL = EDUr < EDUd = CONd = CONr Pre vs. followup F(4,93) = 3.44, p < 0.05 F(1,93) = 6.57, p< 0.05 F(4,93) = 1.64, ns NA F(4,93) = 2.08, p = 0.08 F(1,93) = 6.33, p< 0.05 F(4,93) = 2.35, p = 0.06 CTL = EDUd = EDUr = CONr < CONd F(4,93) = 1.42, ns F(1,93) = 0.46, ns F(4,93) = 2.62, p < 0.05 CTL = EDUd = EDUr = CONd < CONr F(4,93) = 1.40, ns F(1,93) = 5.31, p< 0.05 F(4,93) = 1.59, ns NA F(4,93) = 3.29, p< 0.05 F(1,93) = 8.21, p< 0.005 F(4,93) = 1.59, ns NA F(4,93) = 3.40, p < 0.05 F(1,93) = 2.53, ns F(4,93) = 2.09, p= 0.09 CTL = EDUr = EDUd < CONd = CONr F(4,93) = 1.37, ns F(1,93) = 0.57, ns F(4,93) = 3.56, p< 0.01 CTL = EDUd = EDUr < CONd < CONr Post vs. followup F(4,96) = 4.50, p< 0.005 F(1,96) = 0.39, ns F(4,96) = 0.49, ns NA F(4,96) = 2.56, p< 0.05 F(1,96) = 1.96, ns F(4,96) = 0.88, ns NA F(4,96) = 1.84, ns F(1,96) = 10.31, p< 0.005 F(4,96) = 0.89, ns NA F(4,96) = 2.23, p = 0.07 F(1,96) = 3.67, p = 0.06 F(4,96) = 2.38, p = 0.06 NA F(4,96) = 3.88, p< 0.01 F(1,96) = 4.73, p< 0.05 F(4,96) = 0.17, ns NA F(4,96) = 3.52, p< 0.05 F(1,96) = 1.37, ns F(4,96) = 0.38, ns NA F(4,96) = 3.74, p< 0.01 F(1,96) = 0.85, ns F(4,96) = 1.04, ns NA I S' Co s R" 28,1 z p o o to ro n o igane by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 (0 (0 2 o c o onat I-S! S Z "o c (0 stance, ial dis o and so (0 c .o "3 JO c (0 a> en c n o en sentin £ o.<br><br> £ ics i stati To " ^ c Fere c o >» CO E E 3 LO O .Q .°5 H inued c o o i tratei (0 0) en a u a E cu fiv "5 4 in £ a a. 3 o fol 05 OSt V a llowup o ui CD Q. 8 D.<br><br> revs. Q. CO CD m ari; p o ^ to CO Q.<br><br> C C ^^ oo r^ "* in in CO O -r^ II n II O) O5_ O5 LLTLL'LL'Z o °8? O) Q. V Q c .<br><br> C O - e co in co n "*. oo "*. 5 "r- T- in 3 JLJLi.S CO CO CO || CD O) O> LL'LL'IL'O oin?<br><br> o o O in o o o o o V 3 Q. . Q.Q .05 -UJ t OOO e l O I!<br><br> c\i II " * in II II in" io" irT U- 1 O O O V CM CM CVJ \ LL LfLL O CO o CO CD O ^ c o Q co c S CD .2 O o 2 - § " * £ ! ! .<br><br> - .2 o co a5 co o "- -c -p o o c z II l Ill ii M lli U S g-fflLL. Hill & < 8 E | S ^ g f W £ O <N S l Oi£g. i Mill l 11 nil 11 d) §11811 o a> c: co o o fi OJ-D r «- S co £ S tact.<br><br> Finally, two significant interaction effects were found when the SDS was used as the dependent variable. Contact produced greater change in SDS scores than education did. Contact combined with content on responsibility yielded the greatest improvement in SDS scores at followup.<br><br> Summary. Results show that although education yields some positive benefits, contact with persons with serious mental illness produces the greatest consistent results. Contact yielded highly significant changes in most ele- ments of both the personal responsibility and dangerous- ness models at posttest.<br><br> Moreover, most of these improve- ments were maintained at followup, while many of the benefits that resulted from education returned to baseline 1 week later. Contact also generated significant changes in the SDS, a proxy measure of behavior. Finally, contact had a much greater impact on a real-world helping behav- ior 4donating money to NAMI-SS.<br><br> Discussion We examined two stigmatizing attitudes that impact per- sons with serious mental illness: Persons with serious mental illness are personally responsible for their disabili- ties, and persons with serious mental illness are danger- ous. Results clearly supported the path between danger- ousness, fear, and social avoidance. These findings parallel conclusions drawn from research from the past four decades (Starr 1955; Link et al.<br><br> 1999; Penn et al. 1999; Pescosolido et al. 1999); the belief that persons with serious mental illness are dangerous is perhaps the most pernicious of stigmatizing attitudes about mental ill- ness.<br><br> More important, results from our study suggest one mechanism for how dangerousness impacts persons with serious mental illness: The public attitude that most per- sons with serious mental illness are dangerous leads to fear. Like most sources of fear, people with serious mental illness are avoided. Citizens are less likely to rent apart- ments to persons with serious mental illness and are more likely to hospitalize them to protect the public.<br><br> One especially noticeable finding in this study was the almost perfect correlation between dangerousness and fear. Research by Blascovich and colleagues (2000, 2001) found that stigmatizing labels are highly correlated with perceptions of threat. Moreover, these perceptions of threat are associated with psychophysiological mark- ers of reactivity and arousal (indicators of fear).<br><br> This kind of reaction occurs even when conscious awareness of being fearful is minimal. Hence, fear seems to be an automatic (not cognitively mediated) response to danger- ousness. Unfortunately, findings reported in this article are based on what participants say they would think and how 303 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 P.W. Corrigan et al.<br><br> Table 6. Summary of analyses that examined interactions between process (education vs. contact) and content (dangerousness vs.<br><br> personal responsibility) factors Variables Pre vs. post Pre vs. followup Responsibility EDUCON x trial DANRES x trial EDUCON x DANRES x trial Pity EDUCON x trial DANRES x trial EDUCON x DANRES x trial Anger EDUCON x trial DANRES x trial EDUCON X DANRES X trial Help EDUCON x trial DANRES x trial EDUCON x DANRES x trial Dangerousness EDUCON x trial DANRES x trial EDUCON x DANRES x trial Fear EDUCON x trial DANRES x trial EDUCON x DANRES x trial Avoidance EDUCON x trial DANRES x trial EDUCON x DANRES x trial Social Distance Scale EDUCON x trial DANRES x trial EDUCON x DANRES x trial F(1,158) = 1.33, ns F(1,158) = 19.10, p < 0.0001; RES < NotRES F(1,158) = 1.98, ns F(1,158) = 0.16, ns F(1,158) = 0.61, ns F(1,158) = 0.53, ns F(1,158) = 9.34, p< 0.005; EDU < CON F(1,158) = 0.55, ns F(1,158) = 0.15, ns F(1,158) = 0.05, ns F(1,158) = 0.36, ns F(1,158) = 2.16, ns F(1,158) = 7.12, p < 0.01; EDU < CON F(1,158) = 0.10, ns F(1,158) = 2.25, ns F(1,158) = 6.85, p < 0.01; EDU < CON F(1,158) = 0.50, ns F(1,158) = 0.39, ns F(1,158) = 6.19, p < 0.05; EDU < CON F(1,158) = 0.36, ns F(1,158) = 0.33, ns F(1,158) = 8.88, p < 0.005; EDU < CON F(1,158) = 0.29, ns F(1,158) = 1.00, ns F(1,68) = 0.84, ns F(1,68) = 0.72, ns F(1,68) = 0.64, ns F(1,68) = 0.67, ns F(1,68) = 4.11, p < 0.05; NotRES < RES F(1,68) = 0.00, ns F(1,68) = 7.98, p < 0.01; EDU < CONd F(1,68) = 2.01, ns F(1,68) = 0.06, ns F(1,68) = 3.06, p = 0.08; EDU < CON F(1,68) = 0.15, ns F(1,68) = 0.89, ns F(1,68) = 4.75, p < 0.05; EDU < CON F(1,68) = 0.81, ns F(1,68) = 0.56, ns F(1,68) = 5.56, p < 0.05; EDU < CON F(1,68) = 1.46, ns F(1,68) = 0.02, ns F(1,68) = 10.56, p < 0.005; EDU < CON F(1,68) = 0.21, ns F(1,68) = 0.01, ns F(1,68) = 9.12, p < 0.005; EDU < CON F(1,68) = 0.64, ns F(1,68) = 7.15, p < 0.01; CON-NotRES > other 3 Note.<br><br> 4ANOVA = analysis of variance; DANRES = content factor: dangerousness vs. personal responsibility; EDUCON = process factor: education vs. contact; NAMI-SS = National Alliance for the Mentally III of the South Suburbs; notRES = person NOT viewed as res ponsi- ble for illness; ns = nonsignificant; RES = person viewed as responsible for illness.<br><br> F test represents 2 (trial) x 2 (process) x 2 (content) ANOVAs for each of the dependent variables grouped as pretest vs. posttest, pretest vs. followup, and posttest vs.<br><br> followup. Po st hoc analyses represent pairwise 2 x 2 analyses between group pretest and either posttest or followup scores. NAMI-SS donation (post test): EDUCON, F(1,168) = 4.11, p< 0.05, EDU < CONd; DANRES, F(1,168) = 0.97, ns; EDUCON x DANRES F(1,168) = 0.17, ns.<br><br> participants say they would behave. Future research needs to include manifest indicators that represent actual behav- ior. Moreover, future research needs to examine other ways in which avoidance is shown (e.g., being unwilling to hire, befriend, or live close to a person with serious mental illness).<br><br> For example, a study that our group is beginning is examining whether employers and landlords will interview people who are identified as mentally ill for a job or an apartment. Finally, future research needs to determine whether attitudes and emotional reactions lead to what an individual might perceive as aversive behav- iors. For example, preliminary analyses of a recently com- pleted study suggest that viewing people with serious men- tal illness as dangerous is associated with beliefs that they should be forced into treatment and should be segregated from the public in institutions (Corrigan et al.<br><br> 2001ft). 304 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol. 28, No.<br><br> 2, 2002 The second question of this study examined how two interacting sets of antistigma interventions affect attri- butes about mental illness and helping behavior. We divided these interventions by crossing antistigma pro- gram processes (education vs. contact) with antistigma program content (personal responsibility vs.<br><br> dangerous- ness). Results show that although education yields some positive benefits, contact with persons with serious mental illness produces the greatest consistent results. Content did not seem to enhance the impact of contact programs.<br><br> This finding parallels our earlier research in which contact was shown to be far superior to education programs in improving stigmatizing attitudes (Corrigan et al. 2001c). Especially of interest in this study was the finding that contact led to significant changes in helping behavior in the form of donating money to NAMI-SS.<br><br> Although these are promising findings, we should not dismiss the demand characteristics of soliciting donations within the context of a study. Future research needs to determine how these findings correspond with helping behavior in the real world. Given the positive findings about contact, future stud- ies need to identify specific mediators that yield the great- est benefits from contact.<br><br> Social psychologists have exam- ined several variables relevant to ethnic prejudice that could be adapted for research on contact and mental ill- ness stigma. One important variable that affects contact is opportunity; members of the majority must have the opportunity to interact with minority group members in order for stigma to change (Sigelman and Welch 1993). Hence, persons with severe mental illness must have for- mal opportunities to contact and interact with the general public.<br><br> Other factors that augment the effects of interper- sonal contact include equal status among participants (Riordan 1978; Cook 1985), cooperative interaction (Johnson et al. 1984; Worchel 1986), institutional support for contact (Williams 1977; Adlerfer 1982), frequent con- tact with individuals who mildly disconfirm the stereotype (Weber and Crocker 1983; Johnston and Hewstone 1992), high level of intimacy (Amir 1969; Brown and Turner 1981; Ellison and Powers 1994), and real opportunities to interact with minority group members (Sigelman and Welch 1993). Each of these suggests specific hypotheses for facilitating contact between members of the general public and persons with severe mental illness.<br><br> Findings from this study also have interesting policy implications for stigma change programs. Findings from this research suggest the prime agent for stigma change, as many advocacy groups have been recommending (Corrigan and Lundin 2001), is the person with serious mental illness. Hence, groups that are endorsing education programs for changing stigma should partner with con- sumers who are willing and able to tell their story.<br><br> This approach is also consistent with the approach of services that foster personal empowerment in people with serious mental illness. In this way, people with serious mental ill- ness assume a central role in challenging the stigma that robs many of them of life opportunities. References Adlerfer, C.P.<br><br> Problems of changing white males' behav- ior and beliefs concerning race relations. In: Goodman, P.S., ed. Change in Organizations.<br><br> San Francisco, CA: Jossey-Bass, 1982. pp. 122-165.<br><br> Ajzen, I., and Fishbein, M. Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84:888-918, 1977.<br><br> Amir, Y. Contact hypothesis in ethnic relations. Psychological Bulletin, 71:319-342, 1969.<br><br> Anderson, J.C., and Gerbing, D.W. Predicting the per- formance of measures in a confirmatory factor analysis with a pretest assessment of their substantive validities. Journal of Applied Psychology, 76:732, 1991.<br><br> Angermeyer, M.C., and Matschinger, H. The effect of violent attacks by schizophrenic persons on the attitude of the public toward the mentally ill. Social Science and Medicine, 43:1721-1728,1996.<br><br> Bentler, P.M. Causal modeling via structural equation sys- tems. In: Nesselroade, J.R., and Cattell, R.B., eds.<br><br> Handbook of Multivariate Experimental Psychology: Perspectives on Individual Differences. New York, NY: Plenum Press, 1988. pp.<br><br> 317-335. Bentler, P.M. EQS Structural Equations Program Manual.<br><br> Los Angeles, CA: BMDP Statistical Software, 1989. Bentler, P.M., and Bonett, D.G. Significance tests and goodness of fit in the analysis of covariance structures.<br><br> Psychological Bulletin, 88:588-606, 1980. Bhugra, D. Attitudes toward mental illness: A review of the literature.<br><br> Ada Psychiatrica Scandinavica, 80:1-12, 1989. Blascovich, J.; Mendes, W.B.; Hunter, S.B.; and Lickel, B. Stigma, threat, and social interaction.<br><br> In: Heatherton, T.F.; Hebl, M.R.; and Hull, J.G., eds. The Social Psychology of Stigma. New York, NY: Guilford, 2000.<br><br> pp. 307-333. Blascovich, J.; Mendes, W.B.; Hunter, S.B.; Lickel, B.; and Kowai-Bell, N.<br><br> Perceiver threat in social interactions with stigmatized others. Journal of Personality and Social Psychology, 80:253-267, 2001. Boisvert, CM., and Faust, D.<br><br> Effects of the label "schizo- phrenia" on causal attributions of violence. Schizophrenia Bulletin, 25(3):479^92, 1999. 305 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Schizophrenia Bulletin, Vol.<br><br> 28, No. 2, 2002 P.W. Corrigan et al.<br><br> Bordieri, J.E., and Drehmer, D.E. Hiring decisions for dis- abled workers: Looking at the cause. Journal of Applied Social Psychology, 16:197-208, 1986.<br><br> Brockington, I.F.; Hall, P.; Levings, J.; and Murphy, C. The community's tolerance of the mentally ill. British Journal of Psychiatry, 162:93-99, 1993.<br><br> Brown, R.J., and Turner, J.C. Interpersonal and intergroup behavior. In: Turner, J.C, and Giles, H., eds.<br><br> Intergroup Behavior. Chicago, IL: University of Chicago Press, 1981. pp.<br><br> 33-65. Cohen, J., and Struening, E.L. Opinions about mental ill- ness in the personnel of two large mental hospitals.<br><br> Journal of Abnormal and Social Psychology, 64:349-360, 1962. Cook, S.W. Experimenting on social issues: The case of school desegregation.<br><br> American Psychologist, 40:452^*60, 1985. Corrigan, P.W. Mental health stigma as societal attribu- tion: Implications for research methods and attitude change.<br><br> Clinical Psychology: Science and Practice, 7:48-67, 2000. Corrigan, P.W.; Edwards, A.; Green, A.; Lickey, S.E.; and Penn, D.L. Prejudice, social distance, and familiarity with severe mental illness.<br><br> Schizophrenia Bulletin, 27(2):219-226, 2001a. Corrigan, P.W., and Lundin, R.K. Don't Call Me Nuts: Coping With the Stigma of Mental Illness.<br><br> Tinley Park, IL: Recovery Press, 2001. Corrigan, P.W.; Markowitz, F.; Watson, A.C.; Rowan, D.; and Kubiak, M.A. "Personal Responsibility and Dangerousness: A Follow-up Study." Unpublished manu- script, 200 \b.<br><br> Corrigan, P.W., and Penn, D.L. Disease and discrimina- tion: Two paradigms that describe severe mental illness. Journal of Mental Health, 6:355-366, 1997.<br><br> Corrigan, P.W., and Penn, D.L. Lessons from social psy- chology on discrediting psychiatric stigma. American Psychologist, 54:765-776, 1999.<br><br> Corrigan, P.W.; River, L.P.; Lundin, R.K.; Penn, D.L.; Wasowski, K.U.; Campion, J.; Mathisen, J.; Gagnon, C; Bergman, M.; Goldstein, H.; and Kubiak, M.A. Three strategies for changing attributions about severe mental illness. Schizophrenia Bulletin, 27(2): 187-196, 2001c.<br><br> Corrigan, P.W.; River, L.P.; Lundin, R.K.; Uphoff- Wasowski, K.; Campion, J.; Mathisen, J.; Goldstein, H.; Gagnon, C; Bergman, M.; and Kubiak, M.A. Predictors of participation in campaigns against mental illness stigma. Journal of Nervous and Mental Disease, 187:378-380, 1999.<br><br> Corrigan, P.W., and Watson, A.C. The paradox of self- stigma and mental illness. Clinical Psychology: Science and Practice, in press.<br><br> Crocker, J.; Major, B.; and Steele, C. Social stigma. In: Gilbert, D.T.; Fiske, S.T.; and Lindzey, G., eds.<br><br> The Handbook of Social Psychology. 4th ed. Vol.<br><br> 2. New York, NY: McGraw Hill, 1998. pp.<br><br> 504-553. Dooley, P.A. Perceptions of the onset controllability of AIDS and helping judgements: An attributional analysis.<br><br> Journal of Applied Social Psychology, 25:858-869, 1995. Ellison, C.G., and Powers, D.A. The contact hypothesis and racial attitudes among black Americans.<br><br> Social Science Quarterly, 75:385^00, 1994. Farina, A., and Felner, R.D. Employment interviewer reactions to former mental patients.<br><br> Journal of Abnormal Psychology, 82:268-272, 1973. Fazio, R., and Zanna, M. Direct experience and attitude- behavior consistency.<br><br> In: Berkowitz, L., ed. Advances in Experimental Social Psychology. Vol.<br><br> 145. New York, NY: Academic Press, 1981. Graham, S.; Weiner, B.; and Zucker, G.S.<br><br> An attributional analysis of punishment goals and public reactions to O.J. Simpson. Personality and Social Psychology Bulletin, 23:331-346, 1997.<br><br> Greenley, J.R. Social factors, mental illness, and psychiatric care: Recent advances from a sociological perspective. Hospital and Community Psychiatry, 35:813-820, 1984.<br><br> Hamre, P.; Dahl, A.A.; and Malt, U. Public attitudes to the quality of psychiatric treatment, psychiatric patients, and prevalence of mental disorders. Nordic Journal of Psychiatry, 48:275-281, 1994.<br><br> Harding, CM., and Zahniser, J.H. Empirical correction of seven myths about schizophrenia with implications for treatment. Ada Psychiatrica Scandinavica, 90:140-146, 1994.<br><br> Hatcher, L. A Step-by-Step Approach to Using the SAS System for Factor Analysis and Structural Equation Modeling. Cary, NC: SAS Institute, 1994.<br><br> Holmes, E.P.; Corrigan, P.W.; Williams, P.; Canar, J.; and Kubiak, M.A. Changing public attitudes about schizo- phrenia. Schizophrenia Bulletin, 25(3):447^56, 1999.<br><br> Johnson, D.W.; Johnson, R.; and Maruyama, G. Goal interdependence and interpersonal attraction in heteroge- neous classrooms. In: Miller, N., and Brewer, M.B., eds.<br><br> Groups in Contact. New York, NY: Academic Press, 1984. pp.<br><br> 187-213. Johnson-Dalzine, P.; Dalzine, L.; and Martin-Stanley, C. Fear of criminal violence and the African American elderly: Assessment of a crime prevention strategy.<br><br> Journal of Negro Education, 65:462-469, 1996. Johnston, L., and Hewstone, M. Cognitive models of stereotype change: HI.<br><br> Subtyping and the perceived typi- 306 by on July 21, 2010 http://schizophreniabulletin.oxfordjournals.org Downloaded from Personal Responsibility and Dangerousness Schizophrenia Bulletin, Vol. 28, No. 2, 2002 cality of disconfirming group members.<br><br> Journal of Experimental Social Psychology, 28:360-386, 1992. Levey, S., and Howells, K. Dangerousness, unpredictabil- ity and the fear of people with schizophrenia.<br><br> Journal of Forensic Psychiatry, 6:19-39, 1995. Link, B.G. Understanding labeling effects in the area of mental disorders: An assessment of the effects of expecta- tions of rejection.<br><br> American Sociological Review, 52:96-112, 1987. Link, B.G., and Cullen, F.T. Contact with the mentally ill and perceptions of how dangerous they are.<br><br> Journal of Health and Social Behavior, 27:289-303, 1986. Link, B.G.; Phelan, J.C.; Bresnahan, M.; Stueve, A.; and Pescosolido, B.A. Public conceptions of mental illness: Labels, causes, dangerousness, and social distance.<br><br> American Journal of Public Health, 89:1328-1333, 1999. Macrae, C.N.; Bodenhausen, G.V.; Milne, A.B.; and Jetten, J. Out of mind but back in sight: Stereotypes on the rebound.<br><br> Journal of Personality and Social Psychology, 67:808-817, 1994. Macrae, C.N.; Bodenhausen, G.V.; Milne, A.B.; and Wheeler, V. On resisting the temptation for simplification: Counterintentional effects of stereotype suppression on social memory.<br><br> Social Cognition, 14:1-20, 1996. Madianos, M.G.; Madianou, D.; Vlachonikolis, J.; and Stefanis, C.N. Attitudes toward mentally ill in the Athens area: Implications for community mental health interven- tion.<br><br> Acta Psychiatrica Scandinavica, 75:158-165, 1987. Martin, J.K.; Pescosolido, B.A.; and Tuch, S.A. Do Attributions and Interpersonal Contact Shape Public Attitudes Toward Mental Health Problems?<br><br> Applying a Traditional Sociological Model of Prejudice. Unpublished manuscript, 1999. Menec, V.H., and Perry, R.P.<br><br> Reactions to stigmas among Canadian students: Testing attribution-affect-help judge- ment model. Journal of Social Psychology, 138:443 4453, 1998. Monahan, J.<br><br> Mental disorder and violent behavior: Perceptions and evidence. American Psychologist, 47:511-521, 1992. Olshansky, S.; Grob, S.; and Ekdahl, M.<br><br> Survey of employment experiences of patients discharged from three state mental hospitals during period 1951-1953. Mental Hygiene, 44:510-522, 1960. Page, S.<br><br> Effects of the mental illness label in attempts to obtain accommodation. Canadian Journal of Behavioural Science, 9:85-90, 1977. Page, S.<br><br> Psychiatric stigma: Two studies of behaviour when the chips are down. Canadian Journal of Mental Health, 2:13-19, 1983. Page, S.<br><br> Effects of the mental illness label in 1993: Acceptance and rejection in the community. Journal of Health and Social Policy, 7:61-68, 1995. Penn, D.L.; Guynan, K.; Daily, T.; Spaulding, W.D.; Garbin, C.P.; and Sullivan, M.<br><br> Dispelling the stigma of schizophrenia: What sort of information is best? Schizophrenia Bulletin, 20(3):567-578, 1994. Penn, D.L.; Kommana, S.; Mansfield, M.; and Link, B.G.<br><br> Dispelling the stigma of schizophrenia: II. The impact of information on dangerousness. Schizophrenia Bulletin, 25(3):437^t46, 1999.<br><br> Pescosolido, B.A.; Monahan, J.; Link, B.G.; Stueve, A.; and Kikuzawa, S. The public's view of the competence, dangerousness, and need for legal coercion of persons with mental health problems. American Journal of Public Health, 89:1339-1345, 1999.<br><br> Petty, R.E., and Cacioppo, J.T. Attitudes and Persuasion: Classic and Contemporary Approaches. Boulder, CO: Westview Press, 1996.<br><br> Phelan, J.C.; Link, B.G.; Stueve, A.; and Pescosolido, B. Public Conceptions of Mental Illness in 1950 and 1996: Has Sophistication Increased? Has Stigma Declined?<br><br> Unpublished manuscript, 1999. Rabkin, J. Public attitudes toward mental illness: A review of the literature.<br><br> Schizophrenia Bulletin, 10(2):9-33, 1974. Reisenzein, R. A structural equation analysis of Weiner's attribution 4affect model of helping behavior.<br><br> Journal of Personality and Social Psychology, 50:1123-1133, 1986. Riordan, C. Equal-status interracial contact: A review and revision of the concept.<br><br> International Journal of Intercultural Relations, 2:161-185, 1978. Roman, P.M., and Floyd, H.H. Social acceptance of psy- chiatric illness and psychiatric treatment.<br><br> Social Psychiatry, 16:21-29, 1981. Rush, L.L. Affective reactions to multiple socia<br><br>