Cultural Models of Substance Use Risk and Attributed Stigma: A Comparison of Young Adults in Brazil and the United States

Abstract

The stigma associated with substance use is well known, but little research has examined stigma attribution, or the tendency to stigmatize, nor is there much cross-cultural research on the topic. We examine cultural models for the risk of substance use associated with stigma attribution in two settings: the United States and Brazil. Study populations of young adults are the focus in each setting. Using methods of cultural domain analysis, cultural consensus analysis, and the analysis of residual agreement, we find similar models in each society. There is a continuum from viewing substance use risk as a biopsychosocial problem to viewing it as a moral issue. In the United States, viewing substance use as a biopsychosocial issue is associated with lower attributed stigma; in Brazil, viewing substance use as a biopsychosocial issue is associated with higher attributed stigma. We argue that social patterns of drug use in each society underlie this difference.

Keywords

stigma attribution cultural models cultural consensus analysis United States Brazil

Stigma, in Goffman’s original sense of a “spoiled identity,” has come to loom large in studies of mental health (Pescosolido, 2013). The stigmatized identity that accompanies mental illness can both interfere with the treatment process and serve as an additional source of stress on an already distressed individual. The stigma associated with substance use and misuse can be particularly acute (Barry, McGinty, Pescosolido, & Goldman, 2014).

Absent from many discussions of stigma is a consideration of stigma attribution; that is, rather than focusing on the stigmatized, this research would focus on the stigmatizer. What factors lead an individual to regard a person who uses or misuses substances as unreliable, dangerous, or unworthy? In recent work, we have examined how a cultural model of the risk of substance use/misuse influences an individual’s tendency to stigmatize, or not, persons with substance use disorders among American college students (Henderson & Dressler, 2017, 2019). Findings indicate, first, that there is contention over the cultural model itself; and, second, how one aligns with those contested features influences the tendency to stigmatize.

American society is far from unique in the high prevalence of substance use disorders. Therefore, the results we have observed among American college students demand comparison with other settings. Our primary aim in this article is to do just that. We carried out work in urban Brazil, sampling a college-aged population, to see how a cultural model of substance use risk might resemble the cultural model observed in the United States, and how cultural competence in that model would be associated with stigma attribution. Here, we directly compare the two cultural models and the patterns of results. This provides greater insight into how an understanding of substance misuse risk is configured within sociocultural settings. It also provides an example of how cognitive culture theory and method, with its strong emic orientation, can be successfully employed in cross-cultural, comparative research. Furthermore, these cross-sectional studies are situated directly in a time and place where beliefs and opinions, especially ones as controversial as drug use and addiction, are in the midst of change.

Substance Use/Misuse in Brazil and the United States

The United States and Brazil are two of the world’s leading drug consumers (Instituto Nacional de Ciencia e Technologia para Politicas Publicas do Alcool e Outras Drogas [INPAD], 2012; United Nations Office on Drugs and Crime [UNODC], 2017). Worldwide, the United States has the highest prevalence of drug use in nearly every measured category (UNODC, 2017). Furthermore, these rates of substance use seem to be holding steady (Substance Abuse and Mental Health Services Administration [SAMHSA], 2015). In Brazil, there are high rates of alcohol misuse, with nearly 10% of the total population dependent on alcohol (INPAD, 2012) and cocaine use, with several studies estimating more than 1 million users nationwide (Duailibi, Ribeiro, & Laranjeira, 2008; Madruga, Miguel, McPherson, McDonell, & Laranjeira, 2017). Brazil’s cocaine consumption is second only to the United States and is 4 times higher than the global average (Abdalla et al., 2016). These high rates are often attributed to Brazil’s spatial proximity to Bolivia, Peru, and Colombia, which are the world’s leading producers of cocaine, and to the long coastline along the Atlantic, which facilitates the importing and exporting of drugs (UNODC, 2017). Unlike other South American countries, cocaine consumption in Brazil has increased in the past decade (Miraglia, 2015), particularly among college-aged individuals (Centro Brasileiro de Informações sobre Drogas Psicotrópicas, 2010).

In the United States, individuals in the 18 to 25 age group use significantly more types and amounts of substances than any other group (Ahrnsbrak, Bose, Hedden, Lipari, & Park-Lee, 2017). Not only do rates of use tend to increase significantly when moving from the 12 to 17 into the 18 to 25 age group, but, in a majority of categories, the use rates decrease between the 18 to 25 and 26 and older groups. This suggests that there is not only an impetus for individuals in the 18 to 25 age group to begin using substances, but there is also a deterrent as people age. Entrance into undergraduate collegiate programs marks the beginning of a critical period that has been associated with a significant increase in the use of alcohol and other substances (Hartzler & Fromme, 2003). University acts as a point of change in influences on young peoples’ understanding of social and moral patterns and behaviors, from a reliance on social networks closely tied to their family of origin to a broader set of social influences, as well as a growing independence in thought.

Researchers have also identified increased rates of substance use among young adults (ages 20-29 years) in Brazil, particularly in relation to alcohol, marijuana, and cocaine (Abdalla et al., 2014; Galduróz, Noto, Nappo, & Carlini, 2005). Unlike in the United States, however, Brazilian students are less likely to leave their parental home to attend university. Indeed, many young Brazilians only leave home after marriage. This means that many of the explanations that have been employed to interpret the American phenomenon of increased young adult substance use do not apply to their Brazilian counterparts. Furthermore, the United States has several nationwide programs dedicated to educating children about drug use/misuse (such as Drug Abuse Resistance Education [D.A.R.E.] or Red Ribbon Weeks), which are potential sources of the shared knowledge embedded within the cultural model of substance misuse risk (Henderson & Dressler, 2017). However, these types of programs are absent from the Brazilian national curriculum. Therefore, there persist questions of not just what is leading to higher rates of substance use among Brazilian college-aged individuals, but also the sharing (or not) of ideas about risk factors associated with substance use disorders.

Cultural Models of Substance Use Risk

Substance use is a behavior that is socially discredited or stigmatized (Goffman, 1963). People are more likely to have negative attitudes toward individuals with substance use disorders than those with other behavioral and emotional disorders (Barry et al., 2014). However, the social and cultural origins of this stigma have received little attention. Such research would address the shared understandings of substance use disorder distributed in the wider society that underlie the tendency to socially discredit individuals with this disorder. This is the question of stigma attribution (Corrigan, 2000). The principal goals guiding our research are to determine how cultural models of substance misuse risk shape individuals’ understanding of these disorders and influence stigma attribution across different societies (Henderson & Dressler, 2017).

Previous research on this question was conducted by Link and colleagues (Link, Phelan, Bresnahan, & Pescosolido, 1999) and by Pescosolido and associates (Martin, Pescosolido, & Tuch, 2000). Link et al. (1999) use the concept of “cultural beliefs” in their research on stigma. They argue that the most important beliefs are those about causality. In the General Social Survey, respondents were asked whether specific forms of mental illness, including substance abuse, were caused by a chemical imbalance, genetics, stress, an individual’s upbringing, bad character, or divine will. Substantial proportions of respondents endorsed each of these risk factors, with the exception of divine will. With the same social survey data, Martin et al. (2000) found that individuals positing genetics and stress as causes of mental illness reported feeling more comfortable in social interaction with someone with mental illness, including substance use disorder.

Although important and intriguing, these studies leave many questions unexplored. The items employed in the General Social Survey appear to be investigator initiated, rendering moot the question of whether or not these same risk factors would emerge when using an emic approach. It also begs the question of what other sorts of factors might emerge in the domain of substance abuse causation. Furthermore, Martin et al. (2000) treat these causes as discrete and independent, losing the opportunity to observe how these and other elements of the cultural domain might be configured or patterned. Finally, is an understanding of a configuration of potential causes shared within a social setting in which the risk of substance abuse is highly salient, and does this in turn influence stigma attribution?

To address these questions, we explored a cultural model of substance misuse risk among American college students (Henderson & Dressler, 2017), using methods from cultural domain analysis (Borgatti, 1999), cultural consensus analysis (Romney, Weller, & Batchelder, 1986), and the analysis of residual agreement (Dressler, Balieiro, & dos Santos, 2015). Students first free listed potential risk factors for substance abuse. These were reduced to the 28 most commonly listed factors. A separate sample performed an unconstrained pile sort. Multidimensional scaling (MDS) was used to plot item similarities and differences in two dimensions, and cluster analysis was used to identify major groupings, of which there were five. These included clusters of risk factors that we labeled, based on respondent explanations, as Biological, Self-Medication, Familial, Social, and Hedonistic (see “Results” below for more details).

A separate sample then rated these items on a 4-point scale of the degree of “influence” each potential risk factor might have with respect to substance use/misuse, and they completed a stigma attribution scale. Cultural consensus analysis indicated agreement on the rating of these risk factors, although the level of consensus was fairly modest. Next, the data were analyzed for residual agreement (Dressler et al., 2015). Residual agreement analysis identifies subgroups that deviate in consistent ways from the overall consensus. In our data, this meant that while respondents agreed overall that these were important risk factors, one subgroup emphasized biological and self-medication factors as more important, whereas the other subgroup emphasized social and hedonistic factors as more important, and this formed a continuum of culturally defined influence from the biological to the hedonistic. When correlated with stigma attribution, those who emphasized hedonistic/social factors stigmatized substance users more than those who emphasized biological/self-medication factors (Henderson & Dressler, 2017). Subsequent research showed that this association was stronger among respondents with a more conventional or conservative outlook with respect to morality (Henderson & Dressler, 2019).

These results are consistent with the tension within American society between seeing substance abuse disorder (and mental illness in general) as the result of a moral and characterological failure by the victim, versus a disease. The continuum from emphasizing biological/medical factors to emphasizing hedonistic/social factors describes this tension between disease and morality.

Our current research has taken this same theoretical and methodological approach to Brazil to explore how cultural context may influence the results. In the following, we will present both the results from the U.S. research and the Brazil research. The two are linked by a common dependent variable: stigma attribution. This will enable us to directly compare the cultural models derived in each setting and to examine how individual orientations within these cultural models are associated with stigma attribution.

Research Settings

The U.S. research setting was the campus of The University of Alabama. It is a comprehensive research university with a student population greater than 38,000, located in a community of more than 100,000 persons. Although there is significant industry in the community (including steel, health care, and automobile manufacturing), the university dominates the local economy. Like many universities, there is a lively social scene, ranging from the Greek-letter organizations to the local campus “strip” with its bars and live bands. Needless to say, students arriving at the university are confronted with an array of possibilities for experimenting with mind-altering substances to which they may or may not have been exposed before. They confront potential experimentation with a social freedom they also may or may not have had previously. And, as noted above, they arrive with an already-developed cultural model of substance use resulting from years of prior socialization in both schools and religious institutions.

The research setting in Brazil is the city of Ribeirão Preto, a community of more than 600,000 persons in the north of the state of São Paulo (for general descriptions of the city, see Dressler (2018) and Dressler, Balieiro, & dos Santos (2017)). It is known in Brazil as a relatively affluent community, although the level of socioeconomic inequality found there is the same as the rest of the country. Although the historical foundation of the local economy is agriculture, in the past 50 years it has grown as a center of manufacturing, finance, health care, and, especially, education. Home to a campus of the University of São Paulo, there are also campuses of several private universities.

The number of private universities in Brazil has grown substantially in recent years. Admission to public institutions is notoriously difficult, so the private schools address a demand for higher education that has increased dramatically in Brazil in the past 30 years. At the same time, they are clearly profit-making ventures. Nevertheless, substantial numbers of students coming from socioeconomic backgrounds that do not prepare them well for admission to public universities find the funds to satisfy their desires for higher education. Also, as noted above, the typical American pattern of leaving one’s parental home and traveling to a (at least somewhat) distant university is not unknown in Brazil, but it is definitely less frequently pursued. As such, Brazilian university students have a much greater tendency to remain in their home communities and live in their parental homes while at university.

Method

We employed what Schrauf (2018) refers to as a “cross-cultural sequential exploratory design.” This mixed-method research design entails the collection of qualitative and then quantitative data in (at least) two societies, leading to a statistical comparison of results. Our studies were designed to facilitate direct comparison between college-aged students in the United States and Brazil. The first study was completed during 2015 at the University of Alabama, with undergraduate students as respondents. This study received approval from the Institutional Review Board for the Protection of Human Subjects of The University of Alabama. All data for the second study in Ribeirão Preto were collected and analyzed in 2017 and were approved by the Research Ethics Committee of Paulista University in Ribeirão Preto.

The samples in each of the two studies consisted primarily of college students between the ages of 18 and 25, although the Brazilian samples were slightly older due to differences in the Brazilian education system. While Americans tend to transition directly from secondary school to university, Brazilians often have a break in educational enrollment while they are studying to gain university acceptance. Even still, this age range in both contexts captures a particularly salient and liminal period in individuals’ lives, where they are making the transition into autonomous adulthood. In the United States, participants were recruited from ongoing classes at the University of Alabama. In Brazil, participants were recruited through professors and students at two local universities and also at popular young adult hangout locations, such as a local shopping mall in Ribeirão Preto.

The first phase of each study involved a face-to-face interview where participants completed a free-listing task. Each of the study participants was asked to list factors that affect an individual’s likelihood of developing a dependence on drugs or alcohol. The lists for each individual were recorded and synonyms were identified and reconciled in each of the samples. The independent samples were then analyzed for frequency and salience in ANTHROPAC (Borgatti, 1996) and the most important items were selected for inclusion in the two cultural models.

In the second phase, separate samples of students participated in an unconstrained pile-sort task. For this task, each participant was asked to place each of the previously identified and retained items into groups based on perceived similarity. The only restriction was that participants were not allowed to create a single pile consisting of every item in the model. Participants often kept a running commentary throughout the sorting process and were asked to give descriptions of their piles upon completion. These interviews were transcribed and analyzed for salient themes regarding the types of and relationships between risk factors. The pile sorts were coded and analyzed with ANTHROPAC to create a proximity matrix for each informant and an aggregate proximity matrix for each sample. The individual proximity matrices are numerical representations of the understanding that each individual possesses regarding the relative similarities and differences of each element in the cultural model. The matrices can be examined with cultural consensus analysis, to confirm the coherence of the underlying structure of the cultural model (Borgatti, 1994; Boster & Johnson, 1989). Hierarchical cluster analysis and nonmetric MDS were then used to further analyze the aggregate proximity matrices within each society, to visualize the similarities and differences in meaning among the terms.

Finally, participants in the third phase of the research completed a series of Likert-type response scales. On the first scale, participants were asked to rate the degree of influence each risk factor has in the ultimate development of a dependence on drugs/alcohol. These ratings were analyzed within each culture for cultural consensus, and cultural competence coefficients were derived for each respondent.

We also investigated whether residual agreement analysis (Dressler et al., 2015) could detect additional patterned variation beyond the overall shared understanding of substance use/misuse, because this variation has been previously shown to be related to levels of attributed stigma (Henderson & Dressler, 2017). Residual agreement is carried by the second factor in cultural consensus analysis. The loadings on the second factor do not have a direct interpretation as first factor loadings (cultural competence coefficients) do; however, the sign of the coefficient identifies persons who are relatively homogeneous in their responses, after their overall cultural competence has been controlled for. This can be used to identify subgroups that agree in particular about the ratings of subsets of items relative to the overall cultural consensus, which is how we previously identified emphases on biomedical interpretations versus moral interpretations of risk factors (refer to Dressler et al., 2015, for further details of this method of residual agreement analysis). The independent variables—cultural competence coefficients and residual agreement coefficients—are thus culture-specific.

The second scale was a measure of stigma attributed toward individuals with substance use disorder. This serves as the dependent variable and is the same for each sample. Corrigan et al.’s (2002) scale, designed to assess stigma attribution toward persons with mental illness in general, was adapted to assess stigma attribution toward substance users. This 20-item scale measures stigma attribution in terms of how responsible persons are seen to be for their disorder and how dangerous persons with the disorder are seen to be. Sample items include the following: “I would feel unsafe around persons with substance use disorders”; “How responsible do you think a person is for their substance use disorder?” and “How much concern do you feel for persons with substance use disorders?” In Brazil, the scale was translated and back-translated by the authors and a Brazilian research assistant.

In the United States, all of the free lists were face-to-face interviews, whereas the pile sorts and rating tasks were collected by self-report in a classroom setting. In Brazil, all of the free lists were collected in face-to-face interviews, whereas 13.3% of pile sorts and rating tasks were collected in face-to-face interviews and the remainder were collected using an Internet application. This latter technique of data collection was used because of the relatively short time available to be in the field. Gravlee, Maxwell, Jacobsohn, & Bernard (2018) have compared cultural domain analyses from face-to-face interviews with those collected via an Internet application and found the results to be largely similar. We also confirmed no difference in results between face-to-face and Internet-collected data; nevertheless, this difference in data collection must be considered in any interpretation of the results.

Results

Descriptive statistics for each sample are presented in Table 1.

Table 1.

Descriptive Statistics.

	United States	Brazil
Sample size	212	48
Gender (% female)	68.2	70.4
Age (average, SD)	18.9, 1.06	23.4, 5.47
Ethnicity (% White)	79	76.1
Rural versus urban community of origin (% rural)	40.1	2.8
Know someone with addiction (% yes)	88.2	76

The stigma attribution scale, the dependent variable, was analyzed within each sample using exploratory factor analysis. A principal components factor analysis with varimax rotation was performed on the 20-item scale in each of the independent samples to determine the underlying multivariate structure. The Kaiser–Meyer–Olkin measure of sampling adequacy for each analysis was greater than .7, and Bartlett’s test was significant for each. There were four factors in the American sample and five factors in the Brazilian sample with eigenvalues greater than 1. A scree plot of the eigenvalues indicated no clear cutoff for the U.S. sample; however, for the Brazilian sample, the scree plot clearly indicated that a three-factor solution was more appropriate. More importantly, visual inspection of the factor solutions indicated that the first two factors of each solution were very similar, each accounting for more than 50% of the variance in each sample. These first two factors were retained for further analysis.¹

Item loadings on these first two factors for each of the two studies are presented in Table 2. The first factor includes 10 items with high (>0.50) loadings and refers to feelings that individuals with substance use disorders are dangerous. There were five American items and six Brazilian items with high loadings on the second factor, with four of these items being identical. This factor captured the degree to which a respondent felt both sympathy toward and empathy for an individual with substance use disorder. Factor scores on each of these two factors were calculated for each individual in the two studies and are used as dependent variables in this study.²

Table 2.

Factor Analysis of the Stigma Attribution Scale.

Scale items	Factor 1: Dangerousness		Factor 2: Sympathy
Scale items	Southern United States	Brazil	Southern United States	Brazil
I would feel unsafe around persons with substance abuse disorders	.821	.791	−.163	.157
How scared of a person with a substance abuse disorder would you feel?	.810	.888	−.215	−.062
I would feel threatened by a person with a substance abuse disorder.	.800	.890	−.253	−.124
How irritated would you feel by a person with a substance abuse disorder?	.779	.685	.126	−.191
Persons with substance abuse disorders terrify me.	.768	.859	−.135	−.020
How dangerous do you feel a person with a substance abuse disorder is?	.751	.770	−.233	.073
I would feel aggravated by persons with substance abuse disorders.	.746	.689	.061	.179
How frightened of a person with a substance abuse disorder would you feel?	.737	.843	−.290	−.025
How angry do persons with substance abuse disorders make you feel?	.716	.654	.085	−.183
I think persons with substance abuse disorders pose risk to other people unless they are treated.	.695	.632	−.167	.109
How controllable do you think substance abuse disorders are?	.344	.047	.386	−.284
How responsible do you think a person with a substance abuse disorder is for their present condition?	.306	.074	.208	.319
I feel pity for persons with substance abuse disorders.	−.021	−.280	.689	.107
If I were a landlord, I probably would rent an apartment to a person with a substance abuse disorder.	−.399	−.311	−.135	−.504
I would try to avoid a person with a substance abuse disorder.	.440	.299	.198	.571
How much sympathy would you feel for a person with a substance abuse disorder?	.355	.122	.748	.704
How much concern do you feel for persons with substance abuse disorders?	.194	−.322	.701	.532
How sorry do you feel for persons with substance abuse disorders?	.143	−.050	.809	.696
How likely is it that you would help a person with a substance abuse disorder?	.099	−.121	.528	.721
How certain do you feel that you could help a person with a substance abuse disorder?	.069	−.024	.305	.776

The list of items within the domain of substance abuse risk for the U.S. sample and for the Brazilian sample are shown in Table 3. Items are listed in the order of salience (Smith’s s) within each sample. There were 28 items retained for use in the American sample and 29 items retained in the Brazilian sample. Of these, 12 were precisely the same in each sample. These items largely revolved around stress, depression, social isolation, having friends who use, and social class. An additional six items were approximate matches, dealing with peer pressure, family difficulties, and purely hedonistic impulses. An additional four items were quite similar, referring to family history, lack of will power, and a belief in self-control. Finally, there were six items from the American sample and seven from the Brazilian sample that appear to be culture-specific. For the American sample these seem to mainly refer to early childhood trauma, whereas for the Brazilian sample these are a continuation of hedonistic items.

Table 3.

Free List Items.

United States		Brazil
Exact matches
Depression	44.00	37.50	Depression
Stress	38.00	25.00	Stress
Desire for acceptance	26.00	25.00	Desire for acceptance
Low social class	23.00	25.00	Financial problems
High social class	23.00	43.75	Having a lot of money
Easy access	23.00	43.75	Easy access
Loneliness	21.00	25.00	Feeling lonely or isolated
Friends who use	21.00	18.75	Friends who use
Environment	21.00	37.50	Environment
Curiosity	15.00	56.25	Curiosity
Addictive properties	15.00	50.00	Addictive properties
To rebel	0.00	12.50	To rebel
Approximate matches
Peer pressure	51.00	100.00	Influence of friends
Poor home life	44.00	50.00	Family problems
Enjoyment	38.00	18.75	Search for pleasure
Coping mechanism	23.00	25.00	Search for relief
Emotional instability	21.00	25.00	Emotional problems
Lack of family/friend support	10.00	43.75	Lack of family structure/dialogue
Similar matches
Genetics	33.00	31.25	Family history of addiction
Lack of will power	13.00	31.25	Having a weak head
Repetitive use	13.00	25.00	Belief you won’t be addicted
Addictive personality	10.00	18.75	Predisposition to addiction
U.S.-specific items		Brazil-specific items
Social media	26.00	68.75	At parties or clubs
Post-traumatic events	23.00	68.75	Lack of knowledge
Low self-esteem	21.00	68.75	Good sensation
Childhood exposure to family use	21.00	62.50	To escape reality
Boredom	15.00	37.50	To feel better
Pain	10.00	31.25	To relax
—	—	0.00	Lack of god

Next, unconstrained pile sorts of the items were analyzed. Following Borgatti (1994), we ran a consensus analysis on the individual proximity matrices first to confirm that there was an agreed-upon structure for the similarities and differences in meaning among the items within each sample. The eigenvalue ratio for the American sample was 17:1 and for the Brazilian sample was 12:1. These indicate substantial cultural consensus within each sample for the overall structure of the domain.

Figure 1 shows the results of the nonmetric MDS and hierarchical cluster analyses of the aggregate proximity matrices formed from the pile sorts of each set of items within each sample. The two-dimensional solution for each sample was satisfactory (stress = 0.169 for the American sample and stress = 0.164 for the Brazilian sample). In the hierarchical cluster analysis, cluster solutions were selected at the point that intra-cluster heterogeneity increased substantially. The results from the two samples are quite similar, with five clusters of items dominating the American results and four clusters dominating the Brazilian results. Both samples recognize the family, social influences, self-medication, and hedonistic impulses as substance use/misuse risk factors. The major difference is that the American sample includes an explicitly biological cluster of items. The Brazilian item that comes closest to these biological items is “family history,” although this term does not refer explicitly to a potential genetic component to family history; furthermore, it clusters with other family items.³ (See the “Discussion” section below for a much more detailed consideration of the similarities and differences in the MDS model.)

Figure 1.

Multidimensional scaling and cluster analysis results: U.S. sample on the left; Brazilian sample on the right.

Next, samples rated the degree of influence of each item as a risk factor for substance use/misuse on a 4-point scale, from “not at all influential” to “very influential.” In the American sample, a cultural consensus analysis indicated a modest agreement on the degree of influence of each item, eigenvalue ratio = 3.368; mean competence (±SD) = 0.476 (±0.211). In the Brazilian sample, however, there was no consensus, eigenvalue ratio = 1.891; mean competence (±SD) = 0.352 (±0.219). A closer examination of the mean ratings for each sample indicated a fairly narrow range of variation of each rating, with a majority of items being rated as at least somewhat influential. In other words, the tendency seemed to indicate that nearly all items could potentially be considered risk factors. This skew toward the upper end of the scale could mean that, although the informants were consistent in their ratings, the data itself are not suitable for the interval method of cultural consensus analysis (Romney, 1999).

The effect of this on the analysis was confirmed in the following way. Each sample’s ratings were dichotomized so that “not at all influential” and “a little influential” were collapsed as “little or no influence,” whereas “somewhat influential” and “very influential” were collapsed as “some or substantial influence.” Items were combined in this way to try to approximate equal distributions on each side of the dichotomy. Then, a cultural consensus analysis using the “match” method was carried out. This method is less sensitive to the tendency for respondents to rate many items in the same way, relative to the other methods of cultural consensus analysis (Romney, 1999). This analysis confirmed a broad consensus within each sample (for the American sample, the eigenvalue ratio = 4.026, mean competence [±SD] = 0.531 [±0.231]; for the Brazilian sample, the eigenvalue ratio = 6.061, mean competence [±SD] = 0.549 [±0.321]). Therefore, in each sample, there is a broad agreement that nearly all of the risk factors included in this cultural model influence the development of substance use disorders to some extent.

To identify residual agreement, each sample was divided into two groups on the basis of the second factor loadings from the “match” method of cultural consensus analysis. In the American sample, because of the relatively large sample size, the third of the sample with the largest loadings with negative signs on the second factor were retained, along with the third of the sample with the largest loadings with positive signs on the second factor. The Brazilian sample was divided into two groups on the basis of the sign of the loading (positive vs. negative). Again, the sign of the residual agreement coefficients identifies subgroups of respondents who are relatively homogeneous in their ratings of items, beyond the overall cultural consensus. Comparing the groups on how they, on average, deviate from the overall cultural consensus identifies what unique configuration of items they emphasize (Dressler et al., 2015).

To retain sensitivity to these potential differences, the full 4-point response interval was used in the calculation of deviations. First, the average rating of each item was calculated for the total sample. Then, the total group average was subtracted from each of the individual ratings for each subgroup. Each set of subgroup deviation scores was then averaged, so as to identify the items that were rated either higher or lower by the subgroup than by the total sample. These subgroup deviations from the consensus ratings are then plotted so as to pinpoint exactly where the two groups diverge. These plots are shown in Figure 2.

Figure 2.

Residual agreement analysis results: U.S. sample on the left; Brazilian sample on the right.

For Americans, one subgroup identified by residual agreement coefficients tended to rate genetics and self-medication as more important than the overall sample, whereas the other subgroup rated social and hedonistic items as more important. This forms a continuum from emphasizing “biomedical” explanations of substance use risk to emphasizing “moral” explanations of substance use risk.

For the Brazilians, one subgroup of respondents identified by residual agreement coefficients rated risk factors such as family history, depression, and financial problems as more important than the overall sample, whereas the other subgroup of respondents rated factors such as going to parties/clubs, curiosity, seeking good sensations, and a lack of knowledge as more important than the overall sample. These results suggest a continuum from emphasizing “psychosocial” explanations of risk to emphasizing “moral” explanations of risk.⁴

Finally, both sets of respondents’ overall cultural competence in the cultural consensus model, and their residual agreement coefficients, were correlated with the two factors describing stigma attribution. These results are shown in Table 4. For the U.S. sample, there is no correlation of cultural competence with either dangerousness or sympathy; however, there is a positive correlation of residual agreement with dangerousness. Respondents who emphasize moral over biomedical explanations of risk also view persons who use/misuse substances as more dangerous. And, there is an inverse correlation with sympathy; persons who emphasize a biomedical explanation of substance use view persons who use substances with greater sympathy.

Table 4.

Correlations Between Cultural Competence, Residual Agreement, and Stigma Attribution.

	Perceived dangerousness	Sympathy
U.S. sample (n = 212)
Cultural competence	.012	.150**
Residual agreement	.231***	−.298***
Brazilian sample (n = 48)
Cultural competence	.359**	.030
Residual agreement	−.267*	.206

Note. Note that levels of statistical significance vary for correlations of similar magnitude due to sample size differences between societies.

p < .10. **p < .05. ***p < .01.

For the Brazilian sample, both cultural competence in the cultural consensus model and residual agreement are associated with viewing substance users as dangerous. The correlation of residual agreement and stigma attribution is, however, in the opposite direction from that of the U.S. sample. That is, persons who emphasize moral over psychosocial explanations of substance use risk are less likely to view the user as dangerous. In the Brazilian sample, the correlation of residual agreement and sympathy is not statistically significant (although the direction of the correlation is again in the opposite direction from the corresponding correlation for the U.S. sample).

These results are presented graphically in Figure 3 for residual agreement and dangerousness. In these graphs, each sample has been broken into three equal groups. For the American sample, the tendency toward greater stigma attribution going from a biomedical to a moral emphasis is direct (F = 3.93, df = 2, 207; p = .02), with the difference between the clearly biomedical explanation group and the clearly moral explanation group significant (p = .007). For the Brazilian sample, the tendency toward greater stigma attribution going from a psychosocial to a moral emphasis is inverse (F = 2.34; df = 2, 45; p = .10), with the difference between the clearly psychosocial explanation group and the clearly moral explanation group significant (p = .039).

Figure 3.

Association of stigma attribution and residual agreement in Brazil and the United States.

Discussion

A cultural model of substance misuse risk was investigated through cultural domain analysis, cultural consensus analysis, and residual agreement analysis in both the southern United States and Ribeirao Preto, Brazil. We have previously reported the results of the American study and demonstrated that differences in the understanding of risk factors associated with substance use/misuse were directly related to levels of attributed stigma (Henderson & Dressler, 2017). Thus, the aim of this article was to further this investigation by introducing a comparative field site in urban Brazil, which is experiencing a substance use crisis of similar scope.

The terms in which stigma is attributed, which we labeled dangerousness and sympathy, were very similar in each sample, as evidenced by the factor analysis of the stigma scale. This, in itself, is noteworthy, as few studies have been designed in such a way to allow for this direct comparison.

Furthermore, both studies presented here found evidence of shared, albeit internally contested, cultural models of substance misuse risk. The American cultural model contained five major types of risk factors (Social, Familial, Self-Medication, Hedonistic, and Biological), whereas the Brazilian model included four (Social, Familial, Self-Medication, and Hedonistic).

Although these clusters bear identical names, this should not imply that the clusters represent precisely identical sentiments; rather, they depict how these clusters are culturally constituted within these two locations. For instance, the social and familial clusters in each sample appear to be quite similar in terms of their content; however, there are subtle, yet important, differences in the connotations for included items. For Americans, the Social cluster represented a way that the individual was drawn into and trapped within the world of substance use (e.g., peer pressure; social media), whereas the Familial cluster was a potentially toxic environment that an individual could escape should they choose to do so (e.g., environment; to rebel). The opposite is true of the Brazilian clusters, where the social factors are temporal and situational (e.g., environment; at parties/clubs), whereas the familial factors are more permanent and inescapable (e.g., family history of addiction; predisposition to addiction).

A similar difference exists between the risk factors included within the Hedonistic and Self-Medication clusters of causes. For Americans, the Hedonistic cluster combines to represent someone who wants to use substances (e.g., curiosity; boredom)—that is, they are self-indulgent despite knowing of the consequences of their actions. In contrast, the Brazilian hedonist lacks the knowledge or experience to reject their indulgence in substance use and, thus, it references the seeking of fun/pleasure through substance use without the negative connotations of intention (e.g., lack of knowledge; having a weak head). Furthermore, although American’s differentiate between the factors that cause psychological distress (e.g., depression; stress) and the factors that relieve that distress (e.g., enjoyment), Brazilians kept these factors together to formulate a single cluster. Regardless, these clusters represent the use of alcohol/drugs to treat psychological distress within each research site. Finally, Americans differentiate between the biological factors of individuals and substances that may lead to substance use, whereas Brazilians allocate these same items into the familial and hedonistic clusters, respectively.

Moving into the rating tasks, it was noteworthy that in both cases the influence ratings tended to be skewed in the direction of “influential”; that is, in both samples, nearly all the risk factors identified in the free lists were viewed by most of the samples as at least somewhat important in the development of substance use disorder.

We then used residual agreement analysis (Dressler et al., 2015) to investigate whether there was a more deeply seated contention found in the cultural consensus analysis results. In each sample, the second factor loadings for individuals, which indicates their positions in terms of residual agreement, identified clusters of individuals who were more in agreement with one another in terms of the relative influence of the risk factors in the cultural model. The first cluster of individuals was found to have a medical interpretation of the cultural model, as the American group more highly rated the Biological and Self-Medication risk factors and the Brazilian group more highly rated the Self-Medication and Familial risk factors. The second cluster of individuals in each sample had a moral interpretation of addiction, as they emphasized the importance of the Social and Hedonistic risk factors.

However, adherence to the medical model led to less stigma for Americans but led to greater stigma for Brazilians. Thus, despite nominally similar facets of the cultural model, the models were understood and then utilized in very different ways. In the U.S. data, there is no correlation of cultural competence with stigma. The association is, rather, found exclusively in the correlation of the residual agreement coefficients with dangerousness. In the Brazil data, the correlation of residual agreement and dangerousness is also observed (in the opposite direction, of course); however, there is a positive correlation of overall cultural competence and dangerousness. This is actually an artifact of the fact that cultural competence and residual agreement are correlated in the Brazil data (r = –.70, p < .01). In other words, the overall ratings of influence are being driven by the respondents who tend to privilege the importance of the psychosocial risk factors. In these data, the residual agreement analysis clarifies precisely what is driving the cultural model.

On a larger scale, differences in the understanding of the cultural models can be tied to the specifics of the substance use crises that are occurring in each country. In Brazil, scholars and government agencies tend to focus on the country’s high rates of crack cocaine use (INPAD, 2012), even going so far as to call it a “crack epidemic” (Abdalla et al., 2014; Nappo, Sanchez, & Ribeiro, 2012). However, it is important to note that despite the high incidence rates, the prevalence of crack cocaine usage is still quite small (0.7%; Carlini et al., 2010). Thus, it is a relatively small proportion of Brazilian society that is a part of the growing problem of crack consumption. Worldwide, and in Brazil, crack cocaine is associated with lower class consumption, as it is typically cheaper and more readily accessible than powder cocaine. So, it is likely that the risk factors that we have termed “psychosocial” are perhaps more strongly associated with being lower class in Brazil. And, we should note, being lower class in Brazil also means having darker skin color (Dressler, Balieiro, & dos Santos, 1999).

The emphasis in the Brazilian sample on psychosocial risk factors and self-medication indicate a sort of generic “stress model” for the risk of substance misuse. But, although wealthier people can afford to seek treatment for depression or counseling for past traumatic events, it is poor people who have little option for relief, thus turning to drugs/alcohol for self-medication. In some ways this reconceptualization of lower class substance use as a medically and socially explainable phenomenon is a step forward in terms of potential stigma reduction. That is, these individuals do not merely use substances because they are inherently “bad” people; rather, there is a culturally agreed-upon explanation for the ways in which their negative life circumstances and stress associated with those circumstances lead to substance use and abuse. However, when considering who is a “danger” to the large middle-class segment of the population we interviewed for this study, it is still a particular segment of the population—in terms of social class and skin color—who present some degree of threat to their daily lives. In this way, the results presented in this article could in fact be confounded by the high levels of stigma associated with lower social classes and particular racial categories in Brazil.

At the same time, rates of substance use/misuse are also high among individuals of higher socioeconomic classes, as the greater purchasing power lend greater access to a wider range of substances. This is especially the case for younger Brazilians (Baldin, Sanudo, & Sanchez, 2018; Locatelli, Sanchez, Opaleye, Carlini, & Noto, 2012). These are the people who are arguably the most influenced by the Brazilian Social cluster. These risk factors primarily focus on the social environment where substance use is likely to occur. Friends are seen as having influence on an individual’s substance use, but they have the most influence in particular places where substance are easily accessible. As one of our Brazilian respondents put it, “When you go to a party, everyone drinks, and then you automatically drink.” The same sentiment was expressed in relation to drugs: “Even at parties, young people drink a lot of alcohol, sometimes for fun and such, and when the drugs end up being there—‘ah, I’ll try it!’” This cluster captures the idea of socially driven (and socially acceptable) experimentation, especially among young people. Thus, substance use by individuals influenced by these risk factors can be thought of as being confined to a particular place and even a particular time period in someone’s life. Although rates of substance use are high across the socioeconomic spectrum, the type of substance use and, furthermore, the cultural meanings attached to those substances are very different.

In contrast, the United States is presently in the midst of the “opioid epidemic,” a problem that originated in and was propagated by the medical community. This crisis has affected everyone—poor and rich—to the extent that it has led to an increase in the overall mortality rate in the United States (Case & Deaton, 2015). In this way, the “medical” narrative of substance use has fully infiltrated the cultural psyche. Case and Deaton (2015) refer to these deaths as “deaths of despair,” in that they reflect the associated problems of reduced labor force participation, reduced marriage rates, and increases in reports of poor health and poor mental health. Thus, the self-medication risk factors directly evoke this culturally specific narrative that has developed out of the opioid crisis: Individuals who use substances for these reasons are not dangerous to others, but only to themselves. Furthermore, these individuals deserve help, as they were drawn into addiction by factors beyond control. That is, they are not bad people, only the victims of bad circumstance.

The American Social cluster primarily focused on risk factors that arose from people “you surround yourself with” or “outside influences.” That is, it captures the idea that a person’s social network influences an individual in either a positive or a negative way, with negative influences being those that lead to the use of drugs/alcohol. This cluster of causes also embodies the zeitgeist of recent drug education in the United States, with popular programs such as D.A.R.E., the Red Ribbon Campaign, and others relying heavily on the narrative of “peer pressure” and “just say no.” What underlies this narrative is the notion that once an individual does not “say no” to the pressure to use drugs/alcohol, they will keep saying “yes” when presented with other opportunities to use substances. Caving to peer pressure evokes the idea that a person is getting drawn into an alternative lifestyle, from which they may never return. Furthermore, it is through concepts such as “peer pressure” that drug users have gained a reputation of recruiting new members, as they continue the cycle with other nonsuspecting innocents. Thus, it is these people who contribute to the mass hysteria associated with substance use, as they are dangerous (both to themselves and to others) and are undeserving of help because they themselves are to blame for their problems.

The distribution of these clusters of risk factors in the MDS plot also provides evidence for these cultural differences. Both plots contain a clear divide between external and internal risk factors. In the American MDS, this divide manifests such that the Social and Familial clusters reside in the top half of the plot, which the other more individualistic clusters reside in the bottom. Here, internal refers to risk factors that are present “inside the self” that receive little to no influence from others. Although there are provisional differences between family and friends, both refer to external groups of people who have the power to influence substance use. In contrast, the external/internal division in the Brazilian MDS manifests as a separation between the Self-Medication/Familial risk factors and the Social/Hedonistic risk factors. “Inside the self” is indistinguishable from “inside the family.” Therefore, it is perhaps no surprise that overemphasis on similar sets of risk factors in these two cultural settings led to different outcomes regarding perceptions of dangerousness.

From a methodological perspective, these results lend support to the use of cultural domain analysis in cross-cultural, comparative research. A major critique of ethnoscience, the earliest formal approach to a fully emic, cognitive research strategy, was that the potential for comparative research was, by definition, lacking, as the focus was on local knowledge and local meaning (Harris, 1976; Helfrich, 1999; Wang & Huang, 2016). The response by ardent ethnoscientists of the time was that, with the accumulation of sufficient data from diverse settings, it would eventually be possible to develop theoretical constructs useful for cross-cultural comparison (a plea for data that harkened back to Boas). The same charge can be leveled against a research strategy based in contemporary cognitive culture theory employing cultural domain analysis. (Recently, one of us presented research based on cultural domain analysis at an interdisciplinary conference; a prominent sociologist in the audience criticized the research for having high “internal” validity, but to be totally lacking in any “external” validity.)

An emic, cognitive theoretical orientation hews closely to what has elsewhere been referred to as the “anthropological prime directive”: to see the world as others see it (Dressler, 2018). An equally important anthropological directive, however, is to utilize the array of variation worldwide as our laboratory for testing hypotheses about human behavior (Ember & Ember, 2009). The results presented here suggest that these goals may not be as incommensurate as they have been portrayed. In each study presented here, the approach adhered rigorously to methods capable of eliciting data in terms meaningful to the participants. At the same time, these results proved to be sufficiently comparable to offer insights into the configuration and use of cultural models of substance use/misuse across the two settings.

These, of course, are still preliminary findings. Although the focus on college-age respondents is valid due to the salience of issue of substance use and abuse in that age group, the question of stigma and its attribution spans more diverse settings in each society. Further research should employ the results presented here as a foundation for expanding the study of stigma attribution. A second limiting characteristic of the research is that the mix of techniques (face-to-face interviews; self-report via paper and pencil; self-report via the Internet) may have influenced the results. We cannot offer a plausible methodological hypothesis for why this mix of techniques would have created associations in the opposite direction across the two societies, but the possibility cannot be ruled out.

Conclusion

The findings from this comparative study suggest that the tension between viewing individuals with substance use disorder as either victims of a biopsychosocial disease or as failed persons is not unique. The ways in which that understanding is configured, however, differs across societies, based on historical and social influences on the development of substance use, which in turn is associated with stigma attribution. Evidence suggests that stigma, and hence stigma attribution, can suppress the seeking of treatment and that it serves as an additional stressor for already distressed individuals. Further cross-cultural research on stigma attribution could thus contribute to both a better understanding of and an amelioration of a serious worldwide health problem.

Footnotes

Acknowledgements

François Dengah, Kathryn S. Oths, and two anonymous reviewers for Cross-Cultural Research provided helpful comments on an earlier draft of this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported here was supported by funds from the Graduate School, The University of Alabama, and by the Dr. Milady Murphy Endowed Scholarship in Anthropology.

Notes

Author Biographies

Nicole L. Henderson is a doctoral candidate in the program in biocultural medical anthropology, Department of Anthropology, The University of Alabama. She has conducted research on cultural models, stigma, substance use/misuse, and foodways in Brazil and the United States.

William W. Dressler is professor emeritus in the Department of Anthropology, The University of Alabama. He is a biocultural medical anthropologist with research interests in cultural models and health and, in particular, the relationship between culture and the individual.

References

Abdalla

R. R.

Caetano

Massaro

Mitsuhiro

Pinsky

Laranjeira

R. R.

Madruga

C. S.

(2016). Illegal stimulants use in Brazil: Epidemiological aspects and possible reasons for high consumption of crack/cocaine and amphetamine-type stimulants. In Preedy

(Ed.), Neuropathology of drug addictions and substance misuse (pp. 1085-1093). Amsterdam, The Netherlands: Elsevier.

Abdalla

R. R.

Madruga

C. S.

Ribeiro

Pinsky

Caetano

Laranjeira

(2014). Prevalence of cocaine use in Brazil: Data from the II Brazilian National Alcohol and Drugs Survey (BNADS). Addictive Behaviors, 39, 297-301.

Ahrnsbrak

Bose

Hedden

Lipari

Park-Lee

(2017). Key substance use and mental health indicators in the United States: Results from the 2016 National Survey on Drug Use and Health. Rockville, MD: Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration.

Anders

Alario

F.-X.

Batchelder

W. H.

(2018). Consensus analysis for populations with latent subgroups: Applying multicultural consensus theory and model-based clustering with CCTpack. Cross-Cultural Research, 52, 274-308.

Baldin

Y. C.

Sanudo

Sanchez

Z. M.

(2018). Efetividade de uma intervenção digital na diminuição da prática de binge drinking entre frequentadores de baladas [Effectiveness of a digital intervention in reducing binge drinking among club patrons]. Revista de Saúde Pública, 52(2), 1-12.

Barry

C. L.

McGinty

E. E.

Pescosolido

B. A.

Goldman

H. H.

(2014). Stigma, discrimination, treatment effectiveness, and policy: Public views about drug addiction and mental illness. Psychiatric Services, 65, 1269-1272.

Borgatti

S. P.

(1994). Cultural domain analysis. Journal of Quantitative Anthropology, 4, 261-278.

Borgatti

S. P.

(1996). Anthropac 4. Natick, MA: Analytic Technologies.

Borgatti

S. P.

(1999). Elicitation techniques for cultural domain analysis. In Schensul

J. J.

LeCompte

M. D.

Nastasi

B. K.

Borgatti

S. P.

(Eds.), Vol. 3. Ethnographer’s toolkit: Enhanced ethnographic methods (pp. 115-151). Walnut Creek, CA: AltaMira Press.

10.

Boster

J. S.

Johnson

J. C.

(1989). Form or function: A comparison of expert and novice judgments of similarity among fish. American Anthropologist, 91, 866-889.

11.

Carlini

Noto

Sanchez

Carlini

Locatelli

Abeid

(2010). IV Levantamento nacional sobre o consumo de drogas psicotrópicas entre estudantes do ensino fundamental e médio das redes pública e privada de ensino de 27 capitais Brasileiras [National survey of psychotropic drug consumption among elementary and secondary school students in 27 Brazilian capitals]. Centro Brasileiro de Informações Sobre Drogas Psicotrópicas.

12.

Case

Deaton

(2015). Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century. Proceedings of the National Academy of Sciences of the United States of America, 112, 15078-15083.

13.

Centro Brasileiro de Informações sobre Drogas Psicotrópicas. (2010). VI Levantamento nacional sobre o consumo de drogas psicotrópicas entre estudantes do ensino fundamental e médio das redes pública e privada de ensino nas 27 capitais Brasileiras. São Paulo, Brazil: Federal University of São Paulo.

14.

Corrigan

P. W.

(2000). Mental health stigma as social attribution: Implications for research methods and attitude change. Clinical Psychology: Science and Practice, 7, 48-67.

15.

Corrigan

P. W.

Rowan

Green

Lundin

River

Uphoff-Wasowski

. . .Kubiak

M. A.

(2002). Challenging two mental illness stigmas: Personal responsibility and dangerousness. Schizophrenia Bulletin, 28, 293-309.

16.

Dressler

W. W.

(2018). Culture and the individual: Theory and method of cultural consonance. New York, NY: Routledge.

17.

Dressler

W. W.

Balieiro

M. C.

dos Santos

J. E.

(1999). Culture, skin color, and arterial blood pressure in Brazil. American Journal of Human Biology, 11, 49-59.

18.

Dressler

W. W.

Balieiro

M. C.

dos Santos

J. E.

(2015). Finding culture change in the second factor: Stability and change in cultural consensus and residual agreement. Field Methods, 27, 22-38.

19.

Dressler

W. W.

Balieiro

M. C.

dos Santos

J. E.

(2017). Cultural consonance in life goals and depressive symptoms in urban Brazil. Journal of Anthropological Research, 73, 43-65.

20.

Duailibi

L. B.

Ribeiro

Laranjeira

(2008). Profile of cocaine and crack users in Brazil. Cadernos de Saúde Pública, 24, s545-s557.

21.

Ember

C. R.

Ember

(2009). Cross-cultural research methods. Lanham, MD: AltaMira Press.

22.

Galduróz

J. C. F.

Noto

A. R.

Nappo

S. A.

Carlini

E. A.

(2005). Household survey on drug abuse in Brazil: Study involving the 107 major cities of the country—2001. Addictive Behaviors, 30, 545-556.

23.

Goffman

(1963). Stigma: Notes on the management of spoiled identity. Englewood Cliffs, NJ: Prentice Hall.

24.

Gravlee

C. C.

Maxwell

C. R.

Jacobsohn

Bernard

H. R.

(2018). Mode effects in cultural domain analysis: Comparing pile sort data collected via internet versus face-to-face interviews. International Journal of Social Research Methodology, 21, 165-176.

25.

Harris

(1976). History and significance of the emic/etic distinction. Annual Review of Anthropology, 5, 329-350.

26.

Hartzler

Fromme

(2003). Heavy episodic drinking and college entrance. Journal of Drug Education, 33, 259-274.

27.

Helfrich

(1999). Beyond the dilemma of cross-cultural psychology: Resolving the tension between etic and emic approaches. Culture & Psychology, 5, 131-153.

28.

Henderson

N. L.

Dressler

W. W.

(2017). Medical disease or moral defect? Stigma attribution and cultural models of addiction causality in a university population. Culture, Medicine, and Psychiatry, 41, 480-498.

29.

Henderson

N. L.

Dressler

W. W.

(2019). Cultural models of substance misuse risk and moral foundations: Cognitive resources underlying stigma attribution. Journal of Cognition and Culture, 19, 78-96.

30.

Instituto Nacional de Ciencia e Technologia para Politicas Publicas do Alcool e Outras Drogas. (2012). Segundo levantamento nacional de álcool e drogas [Second national survey of alcohol and drugs]. Retrieved from http://inpad.org.br/wp-content/uploads/2014/03/Lenad-II-Relat%C3%B3rio.pdf

31.

Lacy

M. G.

Snodgrass

J. G.

Meyer

M. C.

Dengah

H. J. F.

Benedict

(2018). A formal method for detecting and describing cultural complexity: Extending classical consensus analysis. Field Methods, 30, 241-257.

32.

Link

B. G.

Phelan

J. C.

Bresnahan

Stueve

Pescosolido

B. A.

(1999). Public conceptions of mental illness: Labels, causes, dangerousness, and social distance. American Journal of Public Health, 89, 1328-1333.

33.

Locatelli

Sanchez

Opaleye

Carlini

Noto

(2012). Socioeconomic influences on alcohol use patterns among private school students in São Paulo. Revista Brasileira de Psiquiatria, 34, 193-200.

34.

Madruga

C. S.

Miguel

A. Q.

McPherson

McDonell

M. G.

Laranjeira

(2017). Crack/cocaine consumption and urban violence in Brazil: Data from the Brazilian National Alcohol and Drugs Survey. Drug and Alcohol Dependence, 100, e125-e126.

35.

Martin

J. K.

Pescosolido

B. A.

Tuch

S. A.

(2000). Of fear and loathing: The role of “disturbing behavior,” labels, and causal attributions in shaping public attitudes toward people with mental illness. Journal of Health and Social Behavior, 41, 208-223.

36.

Miraglia

(2015). Drugs and drug trafficking in Brazil: Trends and policies. Center for 21st Century Security and Intelligence Latin America Initiative. Retrieved from https://www.brookings.edu/wp-content/uploads/2016/07/Miraglia-Brazil-final.pdf

37.

Nappo

S. A.

Sanchez

Z. M.

Ribeiro

L. A.

(2012). Is there a crack epidemic among students in Brazil? Comments on media and public health issues. Cadernos de Saúde Pública, 28, 1643-1649.

38.

Pescosolido

B. A.

(2013). The public stigma of mental illness: What do we think; what do we know; what can we prove? Journal of Health and Social Behavior, 54, 1-21.

39.

Romney

A. K.

(1999). Cultural consensus as a statistical model. Current Anthropology, 40, S93-S115.

40.

Romney

A. K.

Weller

S. C.

Batchelder

W. H.

(1986). Culture as consensus: A theory of culture and informant accuracy. American Anthropologist, 88, 313-338.

41.

Schrauf

R. W.

(2018). Mixed methods designs for making cross-cultural comparisons. Journal of Mixed Methods Research, 12, 477-494.

42.

Substance Abuse and Mental Health Services Administration. (2015). Behavioral health trends in the United States: Results from the 2014 National Survey on Drug Use and Health. Retrieved from https://www.samhsa.gov/data/sites/default/files/NSDUH-FRR1-2014/NSDUH-FRR1-2014.pdf

43.

United Nations Office on Drugs and Crime. (2017). World drug report 2017. Retrieved from http://www.unodc.org/doc/wdr2016/WORLD_DRUG_REPORT_2016_web.pdf

44.

Wang

Huang

Y.-H. C.

(2016). Contextuality, commensurability, and comparability in comparative research: Learning from Chinese relationship research. Cross-Cultural Research, 50, 154-177.