Beyond the Dichotomy: Modernizing Stigma Categorization

Abstract

Even though our knowledge of the cause of disease and disability has grown, stigma still exists. Weiner, Perry, and Magnusson’s seminal study on attributions of stigma has been cited over 500 times since its publication in 1988. The current research sought to replicate and expand this literature in two studies. We used the 10 stigmas from the original study and we added six more (representing common psychological and physical stigmas). In the first study, we examined the classification of stigmas using cluster analysis. We found that instead of dichotomizing stigmas into either psychological or physical, attributions of controllability and stability together resulted in four distinct clusters. Although these were mostly consistent with past literature, the fourth cluster included both psychological and physical stigmas and was rated as moderately controllable and moderately stable. In the second study, we examined how information about responsibility shifts causal attributions, emotional responses, and helping behaviors. Information that an individual was responsible for their stigma led to greater attributions of controllability, less positive emotions, and less help compared to information that an individual was not responsible. More interestingly, the no-information control condition was similar to the responsibility information condition for stigmas that fell into the controllable clusters whereas the control condition was similar to the not responsible information condition for stigmas that fell into the uncontrollable clusters. While parsimony is valued, the psychological/physical dichotomy is not nuanced enough to fully capture the variation in stigmas, which in turn has implications for future research on stigma reduction.

One of the ultimate goals of stigma research is to reduce and eventually eliminate stigma, but in order to do this, we must better understand stigma. Goffman (1963) described stigmas as including “abominations of the body—the various physical deformities” and “blemishes of individual character…for example, mental disorder, imprisonment, addiction, alcoholism” (p. 4). Goffman describes physical and psychological stigmas, however as the research on stigma has progressed so has the definition. Crocker, Major, and Steele (1998) defined stigma as “some attribute, or characteristic, that conveys a social identity that is devalued in some particular contexts....stigmas may be visible or concealed” (p. 505). Many visible stigmas are physical in nature such as blindness or paraplegia whereas many concealable stigmas are more psychological in nature such as drug addiction or depression. Although this distinction between physical and psychological stigmas has withstood decades of research, one of the goals of the current research was to examine if there was a better classification system for stigmas beyond this dichotomy. In service of this goal, we turned to attribution theory.

Attribution theory states that we search for the cause of an outcome and the results of that search dictate emotional responses and expectations for future trajectories, which then dictate subsequent behaviors (Weiner, 1985). Assuming stigmas are negative outcomes, Weiner, Perry, and Magnusson (1988) examined the attributions of both physical and psychological stigmas. When no other information about stigma onset was given, they found that perceptions on onset-uncontrollability were related to physical stigmas, whereas perceptions of onset-controllability were related to psychological stigmas. Furthermore, physical stigmas were usually seen as more stable and less controllable eliciting less blame and anger and more liking, pity, and helping behaviors, whereas psychological stigmas were usually seen as less stable and more controllable eliciting more blame and anger and less liking, pity, and helping behaviors. They did note that not all of the stigmas fit neatly into this physical-psychological split. Specifically, although HIV/AIDS was a physical stigma, it was rated as controllable and stable, while Vietnam War syndrome as a psychological stigma was rated as uncontrollable and unstable. Attributions of controllability and stability were not completely “coupled” for all physical and psychological stigmas therefore it is important to examine where this dichotomy breaks down.

Previous research has searched for moderators and mediators to explain the variability of stigma attributions. For example, research has found many factors that exacerbate stigma thereby leading to increased subsequent prejudicial feelings and discriminatory behaviors. These factors include: living in rural areas, having a lower level of education, holding a conservative political ideology, being from a lower socioeconomic status, lacking familiarity with persons with illnesses, being in an emergency situation, and perceiving the suffering of a stigmatized outgroup member as low (Angermeyer, Matschinger, & Corrigan, 2004; Corrigan, et al., 2001; Gill, Andreychik, & Getty, 2013; Hansen, Bourgois, & Drucker, 2014; O’Sullivan, 2012; Swank, Fahs, & Frost, 2013; Weiner, Osborne, & Rudolph, 2011). Additionally, perceived severity of the illness has also been found to increase discrimination and stigma as it leads to a greater desire for social distance (Kasow & Weisskirch, 2010; Mak, Chong, & Wong, 2014).

As the concept of stigma has evolved, so have the theoretical lenses under which its processes are investigated. The stereotype content model (SCM) posits that stereotype content varies along two primary dimensions of warmth and competence that are fundamental to social perception (Fiske, Cuddy, Glick, & Xu, 2002). These two dimensions link stereotypical beliefs (warmth and competence) with emotional responses like admiration, contempt, pity, and envy. The behavior from intergroup affect and stereotypes (BIAS) map is an extension of the SCM that links stereotype content and emotions with behavioral tendencies of helping and harming (Cuddy, Fisk, & Glick, 2007). Although all three models share overlapping emotions, the BIAS map and attribution theory also share behaviors/behavioral tendencies as outcomes, while suggesting different antecedents. In the former, stereotypic content starts the process, while in the latter an attributional search for the cause of stigma starts the process. Another difference between the models is that the SCM has been widely applied to explain warmth and competence as stable group qualities, whereas not all stigmas are inherently stable according to the attribution model. For example, someone might not always have heart disease or might not always have depression. Instead, some researchers using attributional approaches have distinguished between stigma onset and offset (Corrigan 2000; Weiner et al., 1988). This allows for variability in the perception of controllability of the onset or offset of a stigma. For example, not dieting or exercising while smoking might lead to perceptions of higher onset-controllability for heart disease while someone with a genetic predisposition who otherwise engages in healthy behaviors might lead to perceptions of lower onset-controllability. Similarly, those two people might behave differently after being diagnosed with heart disease leading to different perceptions of controllability of stigma-offset. Both of these classification schemes ultimately aim to explain behavior through emotional responses and it is one of our goals to reduce stigma in terms of both behavioral and emotional expressions.

Much of the research on the stigma of psychological illnesses has used attribution theory as a framework. Corrigan and colleagues assessed attitudes towards people with psychological illnesses and their effect on the allocation of personal resources (Corrigan, Watson, Warpinski, & Gracia, 2004). Psychological illnesses were found to elicit more pity, blame, and less helping behaviors by participants’ lack of willingness to donate money to psychological services. Furthermore, participants were more likely to endorse the use of coercive treatment suggesting that someone with a psychological illness has diminished personal control (i.e., they need to be taken care of because they cannot take care of themselves). In an unrelated study, Corrigan and colleagues examined adolescents’ perceptions of stigma associated with alcohol abuse and mental illness (Corrigan et al., 2005). Interestingly, they manipulated responsibility by differentiating someone with mental illness and someone with mental illness due to a brain tumor. They found that adolescents could make this distinction and judged the person with alcohol abuse the harshest followed by the person with mental illness (not due to a brain tumor). Not all psychological/mental illnesses were rated the same which was consistent with Weiner and colleagues’ (1988) findings that drug addiction (a psychological stigma) was perceived as high on personal control, eliciting more anger and less pity. Taken together, this suggests that participants have variable emotional responses to different psychological stigmas which are mostly driven by attributions of responsibility and blame.

There is less research on the stigma of physical illnesses but it too has used attribution theory. Mosher and Danoff-Burg (2008) examined stigmas related to lung cancer and manipulated smoking status (i.e., controllability) of the target. They found greater anger and less pity toward the smokers compared to the non-smokers which were associated with less helping behaviors towards the former compared to the latter. This suggests that emotional responses toward someone with lung cancer are variable, with information on controllability influencing how people think and act towards those with lung cancer. However, this variability in responses towards lung cancer also speaks to the variability in responses towards cancer in general. Research on cancer in its most general form has found that cancer was rated as low in controllability and high in stability. However, when cancer was compared to other highly stable stigmas (e.g., mental retardation and psychosis) cancer has been rated low on both controllability and stability (Corrigan et al., 2000). In turning to different physical stigmas, Vishwanath (2014) showed that lack of knowledge about juvenile diabetes was not only high, but it also led participants to think that diabetes was rare and partially the fault of the child (responsibility, blame) and to describe it in more negative and harsh terms. In a study of causal attributions of Chronic Obstructive Pulmonary Disease (COPD), an 11-item psychological attributions factor was identified and it predicted higher levels of depression, anxiety, and impairment in health-related quality of life in COPD sufferers (Hoth, Wamboldt, Bowler, Make, & Holm, 2011). Collectively, this suggests that participants may have variable responses to a specific physical stigma depending on the level of specificity used to describe the stigma (i.e., cancer or lung cancer; Kasow & Weisskirch, 2010; Mak et al., 2014) and the other stigmas being used for comparison.

Our knowledge of physical and psychological stigmas has grown in that we now know more than ever about the causes of injury, illness, and disease in the realms of both physical and psychological health (Phelan, Link, Stueve, & Pescosolido, 2000). However, stigma has not necessarily changed (Schomerus et al., 2012) and knowledge of disease does not necessarily ameliorate stigma (Mak et al., 2006). For example, we now know that those who abuse their child were often abused themselves (Thornberry & Henry, 2013); however, even with this knowledge child abusers are still heavily stigmatized today. Cycles of abuse often continue suggesting that child abuse is not as controllable and is more stable than traditional views of this stigma suggest. Furthermore, there is experimental evidence that this contextual information may help shift attributions. Although child abuse was originally categorized as a psychological stigma in Weiner and colleagues’ (1988) paper, and was perceived as controllable (i.e., individuals have a choice over whether or not to abuse their child in the first place) and unstable (i.e., individuals have the ability to stop abusing their child), we know that additional information can and does help in the attributional search for a cause of child abuse. When Weiner and colleagues (1988, Study 2) manipulated the information about stigma onset, participants who were told the individual was responsible liked and pitied the target less which was associated with less charitable donations compared to those who were told the individual was not responsible. Beyond that one study, a meta-analysis by Skelton (2006) examined the literature from 1987 to 1996 for relevant attitude components such as attributions, affective responses, and behavioral intentions when comparing AIDS to leukemia. Although both are physical illnesses, AIDS was rated as more controllable, and people with AIDS were rated as more deserving of infection, more dangerous, and more likely to be avoided compared to people with leukemia.

The increase in knowledge not corresponding to stigma reduction also applies to physical stigmas as well. Heart disease – a physical stigma – was rated as uncontrollable and stable (Weiner et al., 1988). Although, there are some risk factors that cannot be controlled, such as family history, age, and ethnicity, the World Heart Federation (2015) has identified modifiable risks over which one does have personal control (i.e., not stable), such as: poor nutrition, physical inactivity, smoking, and binge drinking. For example, an individual’s nutrition and physical exercise patterns and smoking and drinking frequency may fluctuate, and they may react differently to stressful situations in different contexts (World Health Organization, 2006). Similar to the aforementioned finding on child abuse, Weiner and colleagues (1988; Study 2) found that participants given the responsible information (i.e., heart disease was due to excessive smoking and bad diet) liked and pitied the target less and were angrier toward the target which was associated with less personal assistance and charitable donations compared to those given the not responsible information (i.e., heart disease due to hereditary factors).

Several interventions have been proposed with an attributional framework as a way to reduce stigmatizing attitudes towards those with various illnesses. Biological and genetic attributions have been introduced as a way to help reduce the perceived controllability of persons with psychological illnesses with some support (Phelan, 2002). Yet other researchers have found that attributing the cause of mental illness to genetic and biological factors may actually increase stigma. Specifically, this leads to the perception that recovery is not possible, increases associative stigma towards the relatives of the person with the mental illness, and increases the perception of dangerousness which in turn increases the urge for social distancing (Bennett, Thirlaway, & Murray, 2008; Koschade & Lynd-Stevenson, 2011; Lam & Salkovskis, 2007). In two recent meta-analyses, Kvaale and colleagues have shown that while biogenetic explanations for mental illness reduce blame, they increase endorsement of dangerousness stereotypes, pessimism for recovery (Kvaale, Gottdiener, & Haslam, 2013), and social distance (Kvaale, Haslam, & Gottdiener, 2013). Haslam and Kvaale (2015) claimed that this “mixed blessings” model not only applied to public perceptions of stigma, but also extended to how people view themselves (i.e., if they are patients with mental illness, they were more pessimistic about recovery and see less value in psychological interventions) and how clinicians view their patients (e.g., with less empathy, which is a key factor in the therapeutic relationship). In one intervention emphasizing the malleability of biogenetics, Lebowitz and colleagues found that participants with depression were less pessimistic, more hopeful, and had higher agency compared to those not given the malleability intervention (Lebowitz, Ahn, & Nolen-Hoeksema, 2013). Research on the effectiveness of a variety of intervention techniques suggest that video and contact interventions were the best at increasing feelings of empathy towards people with psychological illnesses and that contact interventions were also effective at decreasing social distance (Matteo, 2013). However, when applying education and contact interventions as a means to reduce stigma towards those with mental illnesses among high school populations, Murphy (2014) found that the attributional model did not accurately assess current attitudes towards mental illnesses especially for those who may have more knowledge or contact with those with illnesses. Similarly, when using biogenetic explanations for mental disorders presented in vignettes, Lebowitz and Ahn (2014) found that mental health professionals displayed less empathy than when psychosocial explanations were used, and this pattern held even when a combination of both explanations were presented suggesting professionals are not immune to these effects. Taken together, this suggests that biogenetic explanations, familiarity, and contact may not be as effective if important features (like emotional responses of empathy or blame) are not taken into account as well as if the contextual illness promotes separation and denigrates sufferers.

Stigmatizing attitudes are far reaching, affecting not only the general public, but also those in professional positions (Lyons & Ziviani, 1995; Mirabi, Weinman, Magnetti, & Keppler, 1985; O’Sullivan, 2012). This is dangerous as stigma often influences a person’s decision to seek out professional help and how well that person is able to respond to treatment (Halter, 2003; Bromley, et al., 2013). Outside of service professionals, research continues to investigate health stigmas according to the mental/physical dichotomy. O’Sullivan (2012) found a preference for rehabilitation efforts with individuals with a physical disability compared to a mental disability in rehabilitation service providers. More recently, Hasson-Ohayon, Hertz, Vilchinsky, & Kravetz (2014) also used this dichotomy to examine attitudes about sexuality towards an individual with either a psychological (schizophrenia) or physical disability (paraplegia). Findings from both studies demonstrated continued use of the psychological/physical dichotomy in stigma research.

All in all, attributions of stability and controllability are variable within the physical and psychological stigma domains, implying stigmas may be more variable than originally thought. Perhaps the categorization of stigmas into a physical/psychological dichotomy does not accurately reflect the perceptions of these stigmas. Therefore the primary goal of the current study was to revisit attributions of stigmas to examine an alternative composition pattern of these stigmas based on controllability and stability (cf. physical/psychological). Assuming we can find a new pattern to classify stigmas, the secondary goal of the current study was to examine how this new pattern maps onto the attributional processes that follow, namely affective reactions, future expectations, and behavioral tendencies.

Study 1

Method

Participants

Seventy participants (81.4% female) were recruited from a large southwestern university’s participant pool (SONA). Students were enrolled in introductory courses (psychology and/or communication) and received research credit to fulfill partial course requirements or extra credit. The mean age of participants was 18.44 (SD = 1.21; range 18–25). The majority of participants identified as White (41.4%), followed by Latino American and Asian/Pacific Islander (21.4% each), African American (7.1%), and bi/multiethnic (2.9%). The majority of participants were freshman (78.6%), followed by sophomores (8.6%), juniors (4.3%), and seniors (1.4%). This study was approved by the university’s IRB and all participants were treated in accordance with American Psychological Association (APA) ethical guidelines. Based on an a priori power analysis (power = .80, alpha = .05), 64 participants were required to detect a medium effect size, so our sample size was sufficient.

Procedure and Materials

Participants completed an online questionnaire using Qualtrics software (Qualtrics Labs, Inc., Provo, UT). Participants responded to 13 questions (dependent variables): responsibility, blame, liking, pity, anger, personal assistance, charitable donations, changeability, and the treatment effectiveness of technical and professional job training, welfare, medical treatments and psychotherapy treatments. All items were presented in the following format: “How RESPONSIBLE do you feel someone with STIGMA is for their illness?” These were rated on 9-point scales anchored at the extremes (e.g., 1 = not at all responsible, 9 = entirely responsible).

All 13 dependent variables were randomly presented for 16 randomly presented stigmas. These included the 10 original stigmas presented in Weiner et al. (1988): Alzheimer’s disease, blindness, cancer, child abuse, drug addiction, HIV/AIDS, obesity, paraplegia, heart disease, and Vietnam War Syndrome – renamed in the current study to Post Traumatic Stress Disorder (PTSD). Six additional physical and psychological stigmas were added for more variety and they included: Chronic Obstructive Pulmonary Disease (COPD), diabetes, stroke, anxiety, major depression, and schizophrenia. We selected COPD, diabetes, and stroke because they were among the most prevalent physical diseases in the US not already part of the 10 original stigmas (Heron, 2015). Similarly, we selected anxiety and major depression because of their high prevalence rates (Substance Abuse and Mental Health Services Administration, 2014) and schizophrenia because, while its prevalence is not as high as other mental disorders, it is a particularly stigmatizing condition (Oliveira, Esteves, & Carvalho, 2015). Participants then answered demographic questions (e.g., age, gender, ethnicity, class standing), were thoroughly debriefed, and compensated for their time upon completion of the study.

Results

Descriptive Statistics

Means and standard deviations for all 13 dependent variables are presented for each stigma in Table 1. Weiner and colleagues (1988) found limited utility in grouping stigmas according to the psychological/physical dichotomy. Instead, they examined how controllability and stability were related for each stigma (see Weiner et al., 1988: Table 3). Using this logic, we ran a cluster analyses to organize the 16 stigmas into meaningful groups based on perceptions of controllability and stability (Burns & Burns, 2009).

Table 1.

Means (and standard deviations) for all 13 dependent variables, by stigma.

Stigma	Controllability-related Variables							Stability-related Variables
	Responsi bility	Blame	Like	Pity	Anger	Assistance	Charitable donations	Change	Technical Job Training	Profession al Job Training	Welfare	Medical treatment	Psycho therapy
Alzheimer’s Disease	1.7 (1.5)	1.7 (1.5)	5.8 (1.7)	7.2 (1.8)	1.5 (1.1)	6.9 (1.9)	6.0 (2.2)	2.3 (1.5)	4.1 (2.4)	4.3 (2.5)	5.5 (2.3)	6.5 (2.2)	5.6 (2.5)
Anxiety	3.3 (2.3)	3.2 (2.2)	5.5 (2.0)	5.4 (2.0)	1.9 (1.4)	5.9 (1.9)	4.3 (2.2)	5.7 (2.1)	5.5 (2.1)	5.7 (1.9)	4.5 (2.0)	7.0 (1.9)	7.3 (2.1)
Blindness	2.0 (2.0)	1.8 (1.6)	6.9 (1.7)	6.9 (2.0)	1.2 (.7)	7.5 (1.8)	6.0 (2.1)	2.6 (2.0)	5.7 (2.5)	5.9 (2.4)	6.0 (2.0)	6.4 (2.2)	5.3 (2.5)
Cancer	2.7 (2.0)	2.4 (1.8)	7.0 (1.8)	7.6 (1.8)	1.4 (.9)	7.3 (1.9)	7.1 (1.8)	3.6 (2.0)	4.5 (2.5)	5.0 (2.4)	6.4 (2.0)	8.3 (1.3)	6.5 (2.2)
Child Abuse	6.0 (3.0)	6.5 (2.7)	2.3 (1.9)	2.7 (2.5)	7.3 (2.8)	3.5 (2.8)	2.6 (2.4)	5.9 (2.5)	4.7 (2.7)	5.0 (2.7)	3.1 (2.2)	4.8 (2.6)	7.3 (2.1)
COPD	3.7 (2.3)	3.7 (2.2)	5.8 (1.9)	5.7 (2.1)	1.9 (1.5)	5.7 (1.9)	4.9 (2.0)	3.9 (1.9)	4.9 (2.2)	5.3 (2.0)	5.4 (2.1)	7.5 (1.7)	4.9 (2.4)
Diabetes	4.3 (2.3)	4.3 (2.1)	6.7 (1.7)	5.5 (1.9)	1.6 (1.2)	5.9 (1.8)	4.9 (2.1)	5.0 (2.1)	4.7 (2.5)	5.2 (2.4)	5.0 (2.0)	7.8 (1.5)	4.8 (2.3)
Drug Abuse	6.6 (2.2)	7.0 (1.8)	3.9 (1.9)	3.9 (2.3)	4.5 (2.3)	4.5 (2.2)	3.2 (1.9)	6.7 (2.2)	5.6 (1.9)	5.7 (2.1)	3.7 (2.3)	6.9 (2.1)	6.9 (2.0)
Heart Disease	3.6 (2.1)	3.6 (2.0)	6.7 (1.7)	6.4 (1.7)	1.5 (1.1)	6.4 (1.8)	5.6 (1.9)	4.5 (2.1)	4.7 (2.6)	5.3 (2.3)	5.7 (2.0)	8.1 (1.3)	4.5 (2.3)
HIV/AIDS	4.6 (2.4)	4.7 (2.3)	6.0 (2.0)	6.4 (2.3)	2.0 (1.6)	5.6 (2.0)	5.4 (2.2)	2.7 (1.7)	4.7 (2.6)	5.2 (2.4)	5.3 (2.3)	7.4 (2.0)	5.0 (2.6)
Major Depression	3.7 (2.5)	3.6 (2.3)	4.6 (2.2)	5.7 (2.0)	2.03 (1.4)	6.0 (2.0)	4.6 (2.2)	6.0 (1.9)	5.4 (2.3)	5.8 (2.3)	4.5 (1.9)	7.0 (2.0)	7.7 (1.7)
Obesity	6.4 (2.0)	6.3 (1.9)	6.0 (1.9)	4.4 (2.1)	2.5 (1.9)	5.1 (2.1)	3.4 (2.0)	7.7 (1.7)	5.1 (2.3)	5.5 (2.3)	3.8 (2.1)	6.8 (1.8)	5.6 (2.4)
Paraplegia	2.6 (2.0)	2.6 (2.0)	6.4 (2.0)	6.8 (2.1)	1.4 (1.1)	6.7 (2.0)	5.7 (2.1)	2.8 (1.8)	5.5 (2.3)	5.5 (2.3)	6.2 (2.1)	6.9 (1.9)	5.8 (2.3)
PTSD	2.6 (2.1)	2.4 (1.9)	6.0 (1.8)	6.8 (1.8)	1.5 (1.1)	6.4 (1.8)	5.8 (2.1)	5.4 (2.2)	5.9 (2.0)	5.7 (2.2)	5.8 (1.8)	6.8 (1.9)	7.9 (1.5)
Schizophrenia	2.2 (1.7)	2.3 (1.7)	4.7 (1.9)	6.4 (2.1)	1.8 (1.4)	5.6 (2.1)	5.1 (2.1)	3.6 (2.0)	4.6 (2.2)	5.4 (2.1)	5.3 (2.2)	7.2 (1.9)	7.1 (2.2)
Stroke	3.1 (2.1)	3.0 (2.0)	6.3 (1.7)	6.6 (1.8)	1.6 (1.2)	6.6 (1.7)	5.5 (2.1)	4.1 (2.2)	4.8 (2.3)	5.0 (2.4)	5.8 (2.0)	7.8 (1.5)	5.6 (2.3)

Note. All scores range between 1 and 9.

Cluster Analysis

First, we created a controllability index by summing ratings of responsibility and blame (see Weiner et al., 1988). Responsibility and blame were significantly and positively correlated for all stigmas (rs ranged from .50 to .86). Next, perceptions of controllability and changeability (i.e., opposite of stability) for each of the 16 stigmas were subjected to a two-step cluster analyses. Hierarchical cluster analysis using Ward’s method suggested a four-cluster solution. Then, the k-means clustering method was used to partition the 16 stigmas into the four clusters based on similar means (see Figure 1). The first cluster was low on controllability (M = 4.38) and low on changeability (M = 2.96) and included Alzheimer’s disease, blindness, cancer, paraplegia, and schizophrenia. The second cluster was mid-level on controllability (M = 9.29) and low on changeability (M = 2.71) and only included HIV/AIDS. The third cluster was high on controllability (M = 12.84) and high on changeability (M = 6.77) and included child abuse, drug abuse, and obesity. The fourth cluster was mid-level on controllability (M = 6.83) and mid-level on changeability (M = 4.92) and included anxiety, COPD, depression, diabetes, heart disease, PTSD, and stroke.

Figure 1.

Four-cluster solution, Study 1.

Cluster 1 overlapped with Weiner and colleagues’ (1988) uncontrollable/stable group of stigmas which were physical in nature, and Cluster 3 overlapped with the controllable/unstable group of stigmas which were psychological in nature. Cluster 2 was also consistent with their original classification in that it was stable and somewhat controllable and only included HIV/AIDS. Cluster 4 is more unique for two reasons. First, it does not fit into the classification scheme as originally proposed by Weiner and colleagues (1988) because it is not uncontrollable and unstable. Rather, it is mid-level on both of those attributions. Second, it is comprised of both physical and psychological stigmas. Splitting stigmas according to the physical/psychological dichotomy would be most detrimental to Cluster 4. Instead, we proceeded to test some of the original relationships between attributions, emotions, helping behaviors, and intervention techniques using this four-cluster solution as a guiding framework.

Hypothesis Testing

We found that the four clusters differed in attributions of controllability (blame and responsibility) and changeability in the aforementioned cluster analysis; cluster means are provided in Table 2. We next tested if clusters differed on elicited emotions and helping behaviors (means presented in Table 2). Clusters differed in levels of liking, F(3, 207) = 88.97, p < .001, partial η² = .56. Clusters 1, 2, and 4 elicited significantly more liking than Cluster 3. Clusters differed in levels of pity, F(3, 204) = 95.20, p < .001, partial η² = .58. Cluster 1 elicited significantly more pity than Clusters 2 and 4 which elicited significantly more pity than Cluster 3. Clusters differed in levels of anger, F(3, 204) = 136.65, p < .001, partial η² = .67. Cluster 1 elicited significantly less anger than Clusters 2 and 4 which elicited significantly less anger than Cluster 3. Clusters differed in levels of personal assistance, F(3, 207) = 61.65, p < .001, partial η² = .47. Cluster 1 elicited significantly more personal assistance than Cluster 4 which elicited significantly more personal assistance than Cluster 2 which elicited significantly more personal assistance than Cluster 3. Clusters differed in levels of charitable donations, F(3, 207) = 85.32, p < .001, partial η² = .55. Cluster 1 elicited significantly more charitable donations than Clusters 2 and 4 which elicited significantly more charitable donations than Cluster 3. Taken together, more positive emotions and helping behaviors were elicited by Cluster 1 followed by Clusters 2 and 4, while more negative emotions and neglect were elicited by Cluster 3.

Table 2.

Means (and standard deviations) for all 13 dependent variables for the control condition for Study 1, by stigma cluster.

Cluster	Controllability-related Variables							Stability-related Variables
	Responsi bility	Blame	Like	Pity	Anger	Assistance	Charitable donations	Change	Technical Job Training	Profession al Job Training	Welfare	Medical treatment	Psycho therapy
Cluster 1	2.2a (1.4)	2.2a (1.4)	6.2a (1.4)	7.0a (1.9)	1.5a (0.8)	6.8a (1.6)	6.0a (1.7)	3.0a (1.3)	4.9a (1.8)	5.2a (1.8)	5.9a (1.7)	7.1a (1.3)	6.0a (1.9)
Cluster 2	4.6b (2.4)	4.7b (2.3)	6.0a (2.0)	6.4b (2.3)	2.1b (1.6)	5.6b (2.0)	5.4b (2.2)	2.7a (1.7)	4.7a (2.6)	5.2a (2.4)	5.3b (2.3)	7.4b (2.0)	5.0b (2.8)
Cluster 3	6.3c (1.9)	6.6c (1.5)	4.1b (1.4)	3.6c (1.7)	4.8c (1.7)	4.4c (1.8)	3.1c (1.6)	6.8b (1.6)	5.1a (2.0)	5.4a (2.0)	3.5c (1.9)	6.2c (1.6)	6.6c (1.8)
Cluster 4	3.4d (1.5)	3.4d (1.4)	6.0a (1.5)	6.0b (1.4)	1.7b (0.9)	6.1d (1.5)	5.1b (1.7)	4.9c (1.4)	5.1a (1.8)	5.4a (1.8)	5.2b (1.5)	7.4a (1.2)	6.1a (1.5)

Note. All scores range between 1 and 9. Cluster means accompanied by different superscript letters within a column are significantly different from one another (p < .05).

Finally, we tested if the clusters would differ on perceived benefits of the various intervention techniques (see Table 2). Clusters did not differ in perceived benefits of technical job training, F(3, 204) = 2.06, p = .106, or professional job training, F(3, 201) = 0.77, p = .511. Clusters differed in perceived benefits of welfare, F(3, 207) = 51.15, p < .001, partial η² = .43. Cluster 1 elicited significantly more benefits from welfare than Clusters 2 and 4 which elicited significantly more benefits than Cluster 3. Clusters differed in perceived benefits of psychotherapy, F(3, 207) = 21.91, p < .001, partial η² = .24. Cluster 3 elicited significantly more benefits from psychotherapy than Clusters 1 and 4 which elicited significantly more benefits than Cluster 2. Clusters differed in perceived benefits of medical treatment, F(3, 207) = 19.84, p < .001, partial η² = .22. Clusters 2 and 4 elicited significantly more benefits from medical treatment than Cluster 1 which elicited significantly more benefits than Cluster 3. Taken together, Cluster 1 was viewed as benefiting the most from welfare probably due to the uncontrollable and stable nature of these physical stigmas, followed by psychotherapy and medical treatment. Cluster 3 was viewed as benefiting the most from psychotherapy probably due to the controllable and unstable nature of these psychological stigmas and was viewed as not benefiting from either welfare or medical treatment. Cluster 2 was viewed as benefiting the most from medical treatment probably due to the physical nature of this stable and somewhat controllable stigma, followed by welfare. Cluster 4 was also viewed as benefiting the most from medical treatment probably due to the somewhat controllable and somewhat changeable nature of those stigmas, followed by both medical treatment and welfare.

Discussion

The goal of Study 1 was to determine if, rather than dichotomizing stigmas into physical or psychological domains as was done previously (Weiner et al., 1988), there was an alternative composition pattern of stigmas based on controllability and stability that may be more informative of thoughts, feelings, and behaviors towards stigmatized individuals. The results of Study 1 suggest that there are four distinct clusters and that these clusters partially support Weiner and colleagues’ (1988) findings. Cluster 1 (uncontrollable, stable) was comprised of physical stigmas and Cluster 3 (controllable, unstable) was comprised of psychological stigmas; both are fully consistent with the aforementioned conceptualization of stigma. Cluster 2 (somewhat controllable, stable) contained only HIV, which is also somewhat consistent with the previous conceptualization of stigma. That is, Weiner and colleagues (1988) also found that HIV/AIDS was its own stigma (i.e., Cluster 2), unlike the other stigmas they studied, and this has not changed in the decades since that study was published. However, Cluster 4 (somewhat controllable, somewhat stable) contained both psychological and physical stigmas and would have been lost had we utilized the physical/psychological dichotomy for categorizing stigmas.

The results of Study 1 also suggest that the four-cluster solution better reflects how attributions map onto emotional conjectures, behavioral tendencies, and future expectations. Regarding emotional conjectures, Clusters 1 and 3 were at the extremes across the board, while Clusters 2 and 4 were somewhere in between. Cluster 1 elicited the most liking, pity and help and the least anger, while Cluster 3 elicited the most anger and the least liking, pity and help. This too is consistent with the general notion that we like and help those with more physical stigmas, whereas we do not like or help those with more psychological stigmas. This pattern also held for the perceptions of benefits from welfare. However, the perceptions of benefits from psychotherapy and medical treatment reflected different patterns. Specifically, Cluster 3 (child abuse, drug abuse, and obesity) was perceived to gain the most benefits from psychotherapy while Cluster 2 (HIV/AIDS) was perceived to gain the least benefits with Clusters 1 and 4 falling in between. Finally, Clusters 2 and 4 were perceived to gain the most benefits from medical treatment while Cluster 3 was perceived to gain the least benefits with Cluster 1 falling in between. It is interesting to note that all groups were above the midpoint of the scale suggesting that all stigmas would benefit from psychotherapy and medical treatment and that all but Cluster 3 would benefit from welfare as well.

Cluster 4 often fell in between Clusters 1 and 3 which shows variability in attributions, reactions, and expectations towards people with those stigmas (i.e., “it depends…”). Additional information about onset responsibility (i.e., controllability) has been shown to play a role in shifting attributional responses and with them emotion and behavioral tendencies as well. The goal of Study 2 then was to replicate and expand on previous research (Weiner et al., 1988, Study 2) in order to better understand this attributional shift when information about onset is provided within the context of our newly developed stigma clusters. To that goal, we expanded on our previous design by adding an information condition where participants were given information that an individual was either responsible or not responsible for their stigma. Participants made the same ratings on attributions, emotions, helping behaviors, and future expectations. For the control (no information) condition, we wanted to replicate our findings from Study 1. For the information condition, we hypothesized that information that someone was responsible for their stigma would lead to higher ratings of responsibility, blame, and anger and less liking, pity, personal assistance, and charitable donations compared to information that someone was not responsible for their stigma. Furthermore, we hypothesized that the control (no information) condition would “default” along the lines of controllability found in Study 1. That is, for clusters high in controllability in Study 1, the no information condition would also fall high on ratings of controllability, whereas for clusters low in controllability in Study 1, the no information conditions would fall low on ratings of controllability. Finally, Study 2 allowed us to ask the following research question: are there any informational biases in attributional shifts in emotions and helping behaviors based on stigma cluster? This could ultimately prove helpful in stigma reduction efforts if information about onset responsibility was more (or less) influential to certain clusters of stigmas.

Study 2

Method

Participants

Seventy-seven participants (75.3% female) were recruited from a large southwestern university’s participant pool (SONA). Students were enrolled in introductory courses (psychology and/or communication) and received research credit to fulfill partial course requirements or extra credit. The mean age of participants was 18.75 (SD = 1.04; range 18–22). The majority of participants identified as White (36.8%), followed by Latino American (25%), Asian/Pacific Islander (22.4%), bi/multiethnic (7.9%), and African American (6.6%). The majority of participants were freshman (54.5%), followed by sophomores (33.8%), juniors (10.4%), and seniors (1.3%). This study was approved by the university’s IRB and all participants were treated in accordance with American Psychological Association (APA) ethical guidelines. Based on an a priori power analysis (power = .80, alpha = .05), 79 participants were required to detect a medium effect size, so our sample size was sufficient.

Procedure and Materials

Participants completed an online questionnaire using Qualtrics software. Participants were randomly assigned to one of two information conditions: a no-information control condition or an information condition where they received information about the onset of the stigma. The no-information control condition was exactly the same as Study 1 – random presentation of 13 dependent variables for all 16 randomly presented stigmas.

In the information condition, stigmatized individuals were presented as being either responsible or not responsible for the onset of their stigmas (see Table 3). For example, the items would now read either “How RESPONSIBLE to you think someone with BLINDNESS (as a result of an industrial accident due to carelessness) is for their illness?” or “How RESPONSIBLE to you think someone with BLINDNESS (as a result of an industrial accident due to a coworker’s carelessness) is for their illness?” Information varied by stigma such that participants were randomly assigned to an information condition for each stigma. That is, no one participant received all responsible or all not responsible information conditions. Otherwise, all 13 questions remained the same, were rated on the same 9-point scales, and were randomly presented for the same 16 stigmas. Participants then answered demographic questions (e.g., age, gender, ethnicity, class standing), and were thoroughly debriefed and compensated for their time upon completion of the study.

Table 3.

Information about stigma onset for the responsible and not responsible conditions, by stigma.

Stigma	Responsible	Not Responsible
Alzheimer’s Disease	As a result of brain dysfunction from a risky sky diving accident	As a result of genetic brain dysfunction
Anxiety	As a result of years of drug use as a teen and young adult	As a result of a family history of anxiety and abnormal neurotransmitter levels
Blindness	As a result of an industrial accident due to carelessness	As a result of an industrial accident due to a coworker’s carelessness
Cancer	Skin cancer from excessive tanning	Skin cancer from a genetic predisposition
Child Abuse	Intentionally abused their own child	Was abused themselves as a child and is currently under emotional distress
COPD	As a result of years of smoking	As a result of being exposed to pollutants at work
Diabetes	Type 2 acquired later in life due to poor diet and no exercise	Type 1 acquired as a child
Drug Abuse	As a result of experimenting with recreational drugs	Developed from prior treatment of pain after an injury
Heart Disease	Due to excessive smoking and a bad diet	Due to hereditary factors
HIV/AIDS	As a result of a promiscuous sex life	As a result of a blood transfusion
Major Depression	As a result of years of drug use as a teen and young adult	As a result of a family history of depression and a thyroid problem
Obesity	Due to excessive eating without exercise	Due to a hereditary thyroid condition
Paraplegia	Due to an injury sustained while texting and driving	Due to an injury sustained while another driver was texting and driving
PTSD	As a result of soldier reenlisting after serving two combat terms in Afghanistan	As a result of soldier’s contract being extended and deployed to Afghanistan
Schizophrenia	As a result of years of drug use as a teen and young adult	As a result of a family history where both parents have Schizophrenia
Stroke	As a result of smoking, eating poorly, and not exercising	As a result of a family history of stroke

Manipulation Check

To ensure that the participants were picking up the responsibility information, we used the ratings of responsibility as a manipulation check. For all stigmas (except for PTSD), the information responsible conditions were rated as more responsible for their illness than the information not responsible conditions (almost all ps < .001; see Supplemental Material for all detailed statistical tests).

Results

No information (control) condition

Just as in the original paper (Weiner et al., 1988), we first present the control condition as a way of replicating our results from Study 1. Clusters differed in levels of responsibility, F(3, 114) = 88.71, p < .001, partial η² = .70, blame, F(3, 114) = 129.47, p < .001, partial η² = .77, and changeability, F(3, 114) = 87.34, p < .001, partial η² = .70. Overall, the clusters fell into the same pattern as in Study 1 (see Table 4 for means for each cluster as well as for contrast information between clusters). Regarding responsibility, Cluster 3 was viewed as significantly more responsible than Cluster 2 which was significantly more responsible than Cluster 4 which was significantly more responsible than Cluster 1. Regarding blame, Cluster 3 was viewed as significantly more to blame than Cluster 2 which was significantly more to blame than Cluster 4 which was significantly more to blame than Cluster 1. Regarding changeability, Cluster 3 was viewed as significantly more changeable than Cluster 4 which was significantly more changeable than Clusters 1 and 2. Consistent with Study 1, without any additional information about stigma onset, Cluster 3 was viewed as the most controllable and the least stable; Cluster 1 was viewed as the least controllable and the most stable; Cluster 2 was somewhat controllable but stable; and Cluster 4 was somewhat controllable and somewhat stable.

Table 4.

Means (and Standard Deviations) for All 13 Dependent Variables for the Control Condition for Study 2, by Stigma Cluster.

Cluster	Controllability-related Variables							Stability-related Variables
	Responsi bility	Blame	Like	Pity	Anger	Assistance	Charitable donations	Change	Technical Job Training	Profession al Job Training	Welfare	Medical treatment	Psycho therapy
Cluster 1	2.1a (0.9)	2.1a (1.0)	6.1a (1.2)	6.7a (1.9)	1.5a (0.8)	6.9a (1.5)	6.0a (1.8)	3.1a (1.3)	4.6a (1.6)	5.1a (1.7)	5.6a (2.1)	6.6a (1.4)	6.1a (1.5)
Cluster 2	4.5b (2.1)	5.0b (2.2)	5.9a (2.0)	5.4b (2.3)	2.1a (1.6)	5.1b (2.3)	4.6b (2.0)	3.1a (2.0)	4.5a (2.5)	5.3ab (2.5)	4.5b (2.3)	6.9abc (1.9)	5.2b (2.1)
Cluster 3	6.6c (1.4)	6.8c (1.2)	4.1b (1.3)	3.3c (1.6)	4.7b (1.3)	4.9b (1.7)	2.9c (1.7)	7.2b (1.2)	5.1a (1.7)	5.8b (1.7)	3.1c (1.5)	5.9c (1.7)	6.5a (1.5)
Cluster 4	3.4d (1.2)	3.3d (1.1)	6.0a (1.2)	5.7b (1.6)	1.8a (1.0)	6.2c (1.5)	5.0b (1.7)	5.0c (1.4)	5.0a (1.2)	5.7b (1.8)	4.7b (1.9)	7.1b (1.3)	6.1a (1.2)

Note. All scores range between 1 and 9. Different letters within a column indicates a significant difference between clusters (p < .05).

We next tested if clusters differed on elicited emotions and helping behaviors. Clusters differed in levels of liking, F(3, 114) = 29.94, p < .001, partial η² = .44, pity, F(3, 114) = 52.77, p < .001, partial η² = .58, anger, F(3, 114) = 94.98, p < .001, partial η² = .71, and charitable donations, F(3, 114) = 41.42, p < .001, partial η² = .52. Clusters fell into the same pattern as in Study 1 (see Table 4 for means for each cluster as well as for contrast information between clusters). Clusters differed in levels of personal assistance, F(3, 111) = 21.55, p < .001, partial η² = .37; however, there was a slight change to this pattern. Cluster 1 still elicited significantly more personal assistance than Cluster 4, which elicited significantly more personal assistance than Clusters 2 and 3 (see Table 3 for means for each cluster). Consistent with Study 1, without any additional information about stigma onset, more positive emotions and helping behaviors were elicited with Cluster 1 followed by Clusters 2 and 4, while more negative emotions and neglect were elicited with Cluster 3.

We next tested if the clusters would differ on perceived benefits of the various intervention techniques. Consistent with Study 1, Clusters did not differ in perceived benefits of technical job training, F(3, 114) = 1.59, p = .195, or professional job training, F(3, 111) = 2.52, p = .062. Clusters differed in perceived benefits of welfare, F(3, 111) = 31.68, p < .001, partial η² = .46, psychotherapy, F(3, 114) = 10.44, p < .001, partial η² = .22, and medical treatment, F(3, 114) = 8.50, p < .001, partial η² = .18. Clusters fell into the same pattern as in Study 1 (see Table 4 for means for each cluster as well as for contrast information between clusters). Consistent with Study 1, without any additional information about stigma onset, Cluster 1 was viewed as benefiting the most from welfare, Cluster 3 was viewed as benefiting the most from psychotherapy, and Clusters 2 and 4 were viewed as benefiting the most from medical treatment.

Information conditions

We tested the effects of information about stigma onset (i.e., responsible [R] or not responsible [NR]) on causal attributions, emotions, helping behaviors, and perceived benefits of various intervention techniques. We also wanted to test if the information effects depended on cluster so we performed a 3 (cluster: 1, 3, 4) x 2 (information: responsible, not responsible) mixed-model ANOVA for each of the 13 dependent variables. We did not include cluster 2 (HIV/AIDS) in these analyses because participants only rated one information condition for this particular stigma (either responsible or not responsible) so the design would not have been balanced. Therefore, we present the results for Cluster 2 (HIV/AIDS) separately at the end. For all tests, there were no significant Cluster x Information interactions, meaning the effects of information did not depend on the cluster. Also, there was a main effect of cluster for causal attributions, emotions, and helping behaviors as well as for welfare and medical treatment and these were consistent with the main effects seen in the control condition of this study and Study 1.

The experimental manipulations were effective in that responsibility and blame were higher for the R information condition compared to the NR information condition, F(1, 29) = 115.01, p < .001, partial η² = .80, and F(1, 29) = 166.64, p < .001, partial η² = .85, respectively. Changeability was also higher for the R information condition compared to the NR information condition, F(1, 29) = 31.97, p < .001, partial η² = .52. For all emotions, there was a significant effect of information, with the NR information condition eliciting more liking and pity and less anger than the R information condition, F(1, 29) = 8.21, p = .008, partial η² = .22, F(1, 28) = 53.04, p < .001, partial η² = .66, and F(1, 29) = 19.67, p < .001, partial η² = .40, respectively. There was a significant effect of information on helping behaviors, with the NR information condition eliciting more charitable donations and personal assistance than the R information condition, F(1, 29) = 21.69, p < .001, partial η² = .43, and F(1, 29) = 20.25, p < .001, partial η² = .41, respectively. Finally, there was a significant effect of information on the perceived benefits of welfare, with more benefits associated with the NR information condition compared to the R information condition, F(1, 29) = 7.58, p = .010, partial η² = .21.

For Cluster 2 (HIV/AIDS), the experimental manipulation was also effective in that responsibility and blame were higher in the R information condition compared to the NR information condition, t(35) = 4.80, p < .001, Cohen’s d = 1.62, and t(36) = 4.85, p < .001, Cohen’s d = 1.62, respectively. However, there were no significant information effects for changeability, emotions, helping behaviors, or perceived benefits of any intervention techniques.

Information relative to no information

Information about stigma onset played differential roles for different clusters of stigmas (see Table 5). Consistent with our hypotheses, the control condition (i.e., no information condition) “defaulted” to the NR information condition for Cluster 1 (low in controllability) and for Cluster 4 (mid-level controllability) across most causal attributions, emotions, and helping behaviors. The control condition “defaulted” to the R information condition for Cluster 3 (high in controllability) across the causal attributions and emotions, but not for the helping behaviors. Finally, the control condition “defaulted” to the NR information condition for Cluster 2 (HIV/AIDS) but only for liking. Inconsistent information is particularly influential in the attributional process and has downstream effects for emotions and helping behaviors, but they did not have an effect on future expectations in that they did not play a role across most of the stability-related variables.

Table 5.

Means for All 13 Dependent Variables for Study 2, by Stigma Cluster.

Note. All scores range between 1 and 9. Brackets indicate a significant difference (p < .05). I_R = information responsible; I_NR = information not responsible; Control = no information.

Discussion

The results from Study 2 complement those from Study 1. First, data from the no information (control) condition in Study 2 were almost identical to data from Study 1. Next, as hypothesized, information about personal responsibility for stigma onset played an important role in subsequent emotions and helping behaviors. Specifically, when participants were given responsible information (e.g., obesity because of excessive eating without exercise) compared to not responsible information (e.g., obesity because of a glandular disorder), subsequent ratings shifted higher for responsibility, blame, changeability, and anger and lower for liking, pity, personal assistance, and charitable donations. Also as hypothesized, attributional shifts occurred when information about stigma onset was inconsistent with expectations about the stigma; that is, for clusters of stigmas seen as high in controllability, information that a person was not responsible for the onset of their stigma was particularly helpful in shifting the emotional responses towards those people in a more positive direction. However, for clusters of stigmas seen as low in controllability, information that a person was in fact responsible for the onset of their stigma was enlightening in the opposite direction – the emotional responses were more negative and the helping behaviors were less likely to happen. Finally, we did not find any interactions between information and cluster and this suggests that stigma cluster may not be biasing information or vice versa. However, upon examination of the pattern of results (see Table 4), it appears that information about stigma onset is more straightforward within a given stigma cluster. Across all clusters, if someone is not responsible for their stigma, we will make the attributional shift in that we are less likely to blame them for said stigma. However, we use information about stigma onset differentially depending on what the information is (i.e., responsible or non-responsible) and what stigma we are examining (i.e., one that is typically viewed as controllable or uncontrollable).

General Discussion

Thoughts, feelings, and behaviors towards those with stigmatizing illnesses are variable depending on perceived controllability and stability of the illness. Previous research suggested that physical stigmas should be seen as more uncontrollable and stable yielding higher positive emotions and more helping behavior while psychological illnesses should be seen as controllable and unstable and yield less positive emotions and less helping behavior (Corrigan, Markowitz, Watson, Rowan, & Kubiak, 2003; Corrigan & Watson, 2002; Weiner et al., 1988). However, the results of the current study suggest this is not consistent for every stigma based solely on the physical/psychological dichotomy. There is an abundance of stigma research; however, the majority focuses on the psychological stigmas (i.e., mental illness) as opposed to physical stigmas with relatively even fewer studies examining both. Due to the large number of both psychological and physical stigmas, the extant literature that has examined both often limits its scope of inclusion by including one psychological and one physical stigma (e.g., schizophrenia and emphysema). Therefore, one way this study contributes to the literature is by providing a macro view of both psychological and physical stigmas that demonstrates how they relate to one another.

The present study also advances the stigma literature by providing a more nuanced perspective of stigmas, acknowledging the coexistence of nature (i.e., biological predisposition) and nurture (i.e., personal factors) and their likely interaction. Most stigmas have “default” levels of controllability and stability (Study 1 and the control condition for Study 2), and these levels can be used to classify stigmas into groups of similar others. This is illustrated in our four-cluster solution from Study 1 and is consistent with, but goes beyond, the physical/psychological dichotomy to help put the attributional process surrounding stigma into a more nuanced perspective. In examining a variety of psychological (mental) illnesses, Corrigan and colleagues (2000) found that stigma varied with varying causal attributions which is consistent with the notion that attributional processes play an important role above and beyond a label of “mental illness” or the dichotomous variables of “psychological” or “physical” illness. After decades of increased public awareness about the biological influences of mental disorders (Pescosolido, 2013; Schomerus, et al., 2012) and Americans’ increased understanding of mental illness today compared to 1950 (Phelan et al., 2000), ratings of social distancing from those with mental illness remains largely unchanged (Pescosolido, 2013).

As with any attributional search, when additional information is available, especially about stigma onset, we can use that information accordingly. For example, is someone obese because of excessive eating or because of a glandular disorder? This accompanying information shifts subsequent thoughts, feelings, and behaviors. Specifically, information about personal responsibility was found to have a negative effect on subsequent emotional reactions (less liking and pity, and more anger) and behavioral intentions (fewer helping behaviors). Interestingly, information about stigma onset was differentially effective in shifting attributions, emotions, and behaviors based on cluster. That is, information about responsibility had a negative effect for clusters typically viewed as uncontrollable (e.g., Cluster 1: Alzheimer’s disease, blindness, cancer, paraplegia, and schizophrenia) whereas information about the lack of responsibility had a positive effect for clusters typically viewed as controllable (e.g., Cluster 3: child abuse, drug abuse, and obesity).

There has been a plethora of research looking at what interventions may be more effective at reducing stigma (Haslam & Kvaale, 2015; Matteo, 2013). Our results suggest that when the perceived onset of the disease is manipulated, subsequent thoughts, attitudes, and behaviors are subject to change. This suggests that there is hope for reducing the prejudice and discrimination towards people with a variety of different illness. Along these lines, a small collection of research suggests targeting the causal attributions of controllability and stability as a stigma reduction technique (Corrigan, 2000; Corrigan et al., 2000). Such efforts would entail replacing erroneous attributions (e.g., “persons with a disability are responsible for the disability and cannot take care of themselves”) with more accurate ones (e.g., “persons with a disability have some control over their behaviors, can live independently, and take care of themselves with sufficient support;” Corrigan et al., 2000). Using biogenetic explanations has shown to have “mixed blessings” (Haslam & Kvaale, 2015). They may reduce blame but they may also increase pessimism for recovery, dangerousness, and social distance. Instead, education about the malleability of biogenetics (e.g., with discussion of epigenetics) might be an additional step needed in these interventions (Lebowitz et al., 2013), but this finding is limited to depression only.

Perhaps a more effective, less expensive, and easier way of reducing stigma would be to target stigmas within the same cluster together. Another method that may be effective in stigma reduction is to target the emotional responses that stigmas within the same cluster trigger. For example, child abuse, drug addiction, and obesity (cluster 3) are negatively perceived as stigmas that are within a person’s control, where change is possible. Additionally, they elicit more anger, less pity, and less helping behaviors than the other stigmas. Intervention efforts would target negative feelings and replace them with positive feelings of elevation for example. These type of campaigns would use media that positively depict uplifting stories about individuals who are stigmatized for being overweight, addicted, dealing with child abuse, etc., which would produce feelings of elevation and directly increase positive attitudes and behavioral intentions towards the individual displayed (Shade, Kim, Jung, & Oliver, 2015). Or in the case of the collection of stigmas in cluster 4 (i.e., anxiety, COPD, depression, diabetes, heart disease, PTSD, and stroke), a combined intervention approach could be especially helpful given the comorbidity of many of the stigmas within this cluster (e.g., depression and heart disease; Nicholson, Kuper, & Hemingway, 2006). Future research should assess whether stigma reduction techniques that are found to be effective for a specific stigma may also be equally effective for another stigma falling in the same cluster. Based on these results, this would suggest that HIV stigma campaigns should remain focused on HIV; however, stigma campaigns targeting drug abuse could team up with those targeting obesity, for example, because these two stigmas might be working in similar ways.

As with all studies, ours was not without limitations. We recruited two samples of undergraduates from one university to complete online self-report measures. Although samples and methods like this are the norm for much research on stigma (Corrigan et al., 2000; Corrigan et al, 2003; Weiner et al., 1988), future research should sample adults outside of the college environment using methods other than self-report to test the generalizability of these effects across people and situations. As with any self-report measures on sensitive topics, participants could have potentially responded in ways that were either politically correct or socially acceptable. However, the online nature of the study allowed for a degree of anonymity that could have countered that bias. Future research using more implicit measures (e.g., implicit association test) may be another way to measure association biases that are less sensitive to social demands. Even though our power analyses suggested that our sample sizes would be sufficient to detect medium-sized effects, future research should aim for larger sample sizes when conducting inferential analyses.

Although we were able to expand the previous research by examining 16 different stigmas (six additional stigmas added to Weiner et al.’s original 10) prevalent in today’s society, this still does not address the full range of stigmatizing conditions. Future research should include additional stigmas that fall along the broad spectrums of both controllability and stability to see how they might fit within our larger constellation of stigma clusters. Finally, stigma is a social construction and is therefore influenced by cultural norms and ideologies (Corrigan, Druss, & Perlick, 2014). As such, some judgments and perceptions can be closely linked with one’s cultural beliefs about the causes and meanings of “stigmatized” groups (e.g., the meaning of fat or obesity is different across cultures; Crandall et al., 2001) and/or their ideology (e.g., social dominance orientation is related to higher stigma towards people with depression; Shamblaw, Botha, & Dozois, 2015). Future research should include more culturally diverse samples, both within the US and in other countries, to see if our four-cluster conceptualization of stigma persists across cultural barriers. It would be beneficial to see how the stigmatizing process generalizes across a variety of groups.