Dressler, William W., Mauro C. Balieiro, Rosane P. Ribeiro, and José Ernesto dos Santos. (2007) A...

This article was published in an Elsevier journal. The attached copyis furnished to the author for non-commercial research and

education use, including for instruction at the author’s institution,sharing with colleagues and providing to institution administration.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copy

Social Science & Medicine 65 (2007) 2058–2069

A prospective study of cultural consonance and depressivesymptoms in urban Brazil

William W. Dresslera,�, Mauro C. Balieirob, Rosane P. Ribeiroc, Jose E. dos Santosd

aDepartment of Anthropology, The University of Alabama, P.O. Box 870210, Tuscaloosa, AL 35487-0210, USAbDepartment of Psychology, Paulista University, Ribeirao Preto, BrazilcFaculty of Nursing, University of Sao Paulo, Ribeirao Preto, BrazildFaculty of Medicine, University of Sao Paulo, Ribeirao Preto, Brazil

Available online 3 August 2007

Abstract

Cultural consonance refers to the degree to which individuals, in their own beliefs and behaviors, approximate the

prototypes for belief and behavior encoded in shared cultural models. In previous cross-sectional studies, lower cultural

consonance in several cultural domains was associated with worse health outcomes, including greater psychological

distress. The current paper extends these findings in three ways. First, the effect of cultural consonance on depressive

symptoms is tested in a prospective study. Second, it is hypothesized that the effect of cultural consonance in a specific

cultural domain will depend on the degree of cultural consensus within that domain: the higher the cultural consensus, the

greater the effect of change in cultural consonance in that domain on depressive symptoms. Third, it is hypothesized that

cultural consonance will have an inverse effect on depressive symptoms independent of the occurrence of stressful life

events (a well-known risk factor for depression). We tested these hypotheses in a study conducted in urban Brazil, and

found that change in cultural consonance (assessed as a general construct) was associated with depressive symptoms at a

2-year follow-up. Furthermore, cultural consonance in the domains in which there was highest cultural consensus—the

domains of family life and lifestyle—was more strongly associated with depressive symptoms at follow-up than cultural

consonance in domains with lower cultural consensus. Finally, all of these effects were independent of stressful life events.

These results lend further support to the importance of cultural consonance in relation to human health.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Brazil; Cultural consonance; Depressive symptoms

Introduction

It goes virtually without saying that culturalfactors are important in the etiology and expression

of mental health problems. What has provenproblematic is the development of theory andmethod to examine how culture, conceptualized asa property of a social group, translates into effectson individuals. Dressler and colleagues (Dressler,Balieiro, Ribeiro, & Santos, 2005; Dressler, Borges,Balieiro, & Santos, 2005) have proposed just such atheory, termed a ‘‘theory of cultural consonance,’’and a method to operationalize the key con-struct. Cultural consonance is the degree to which

ARTICLE IN PRESS

www.elsevier.com/locate/socscimed

0277-9536/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.socscimed.2007.06.020

�Corresponding author. Tel.: +1 205 348 1954;

fax: +1205 348 7937.

E-mail addresses: [email protected],

[email protected] (W.W. Dressler), [email protected]

(M.C. Balieiro), [email protected] (R.P. Ribeiro),

[email protected] (J.E. dos Santos).


individuals, in their own beliefs and behaviors,approximate the prototypes for those beliefs andbehaviors encoded in shared cultural models. High-er cultural consonance (assessed in several domains)was associated cross-sectionally with lower arterialblood pressure and fewer expressed symptoms ofpsychological distress in a series of studies (Dressler,2005).

The purpose of this study is to expand upon thesefindings in three ways. First, in this paper, theeffects of cultural consonance on depressive symp-toms are examined prospectively, based on a 2-yearfollow-up of a sample from an urban Braziliancommunity. Second, hypotheses are proposed andtested regarding the relative effects of culturalconsonance in different domains. Specifically, it ishypothesized that low cultural consonance in thecultural domains in which there is higher culturalconsensus or sharing will have a stronger effect ondepressive symptoms. And third, the effects ofcultural consonance relative to the effects of amajor known risk factor for depression, stressful lifeevents, are examined. Depressive symptoms havebeen examined widely in research on culture andhealth and hence provide an appropriate outcomefor this prospective test of a theory of culturalconsonance.

Cultural consonance and health

There are a number of problems with conven-tional concepts of culture applied to the study ofculture and health. The problem of linking cultureas a term that refers to an aggregate (i.e., a socialgroup) to the beliefs and behaviors of individuals isforemost. Linking these levels of analysis in aconceptually satisfying way, employing researchmethods that unambiguously show such a link,has proven elusive in social scientific research. Thisproblem is addressed by a theory of culturalconsonance (Dressler, 2005).

A theory of culture derived from cognitiveanthropology provides the foundation for theconcept of cultural consonance. In Goodenough’s(1996) definition, culture is that which one needs toknow in order to function adequately in a socialgroup. It is this knowledge that is both learned andshared that enables us to appropriately interpretothers’ behaviors and that directs our own beliefsand behaviors. This definition locates culture in theindividual, because not to do so leads to complexontological problems. At the same time, elaborating

the notion of sharing to mean also ‘‘distributed’’extends the locus of culture to the group level(Sperber, 1985). Culture is the knowledge thatconstructs the world around us on the basis of ashared understanding that this is, indeed, the worldand the way it works (D’Andrade, 1984; Searle,1995). Individuals are variably able to reproducethat group knowledge, based on their own sociali-zation and their position in the social structure.Given the incomplete nature of any individual’sknowledge, the overall structure of the environmentof meaning within which social life proceeds is afunction of how shared knowledge is distributedacross individual minds.

This shared knowledge is usefully conceptualizedas encoded at the individual level in a series oflinked and overlapping cultural models. Culturalmodels are schematized representations of anorganized area of shared knowledge—or a culturaldomain—that can be observed in everyday speechand that include the elements of that domain andthe associations among those elements (D’Andrade,1995, p. 151).

This theoretical orientation can be joined to themethod of cultural consensus analysis developed byRomney, Weller, & Batchelder (1986). Culturalconsensus analysis is a statistical technique forevaluating the degree of sharing among a set ofinformants regarding the structure of a culturaldomain. Using cultural consensus analysis, therelative degree to which individual informantsreproduce that shared knowledge in their ownresponses, referred to by Romney et al. as ‘‘culturalcompetence,’’ can be quantified. Given a basic levelof sharing, it is possible to estimate the content ofthe shared knowledge—the ‘‘culture’’—that, inessence, is the best estimate of the responses toquestions about that cultural domain that a reason-ably culturally competent (or knowledgeable)informant would produce.

The cultural consensus model thus joins theaggregate and individual levels of analysis. At thesame time, it evaluates how well an individualknows a particular cultural domain, relative to theaggregate knowledge structuring that domain. Itdoes not evaluate what that informant herselfbelieves or how she behaves relative to thataggregate knowledge. At this point, the constructof cultural consonance becomes important, linkingthe individual’s own beliefs and behaviors to theshared cultural model in that domain. The more anindividual’s own beliefs and behaviors correspond

ARTICLE IN PRESSW.W. Dressler et al. / Social Science & Medicine 65 (2007) 2058–2069 2059


to the prototypes for belief and behavior encoded inshared cultural models, the higher her culturalconsonance (Dressler, 2005).

We hypothesized that low cultural consonancewould be a chronically stressful experience thatwould be associated with worse health outcomes. Ina study conducted in the African-American com-munity in the southern United States (Dressler &Bindon, 2000), and two conducted in urban Brazil(Dressler, Balieiro et al., 2005, 2007; Dressler,Borges et al., 2005), cultural consonance in variouscultural domains was associated with elevatedarterial blood pressure and symptoms of psycholo-gical distress. These associations persisted aftercontrolling for other known correlates of bloodpressure and psychological distress. Additionally,there was a tendency for individuals to be consis-tently more or less consonant across a number ofdomains, the result being a kind of generalizedcultural consonance. Psychological distress wasassociated with this generalized cultural conso-nance, but arterial blood pressure was specificallyassociated with cultural consonance in domainsdefining social distinction and affiliation (Dressler,Balieiro et al., 2005, 2007; Dressler, Borges et al.,2005).

Study aims

The current study was undertaken to extend thesefindings in three ways. First, all previous studieswere conducted using a cross-sectional researchdesign. One weakness of this design is temporalambiguity; that is, did low cultural consonanceleads to psychological distress, or are distressedindividuals less able to organize their lives effec-tively to become culturally consonant? The dataanalyzed here are from a 2-year follow-up of asample for which the cross-sectional analysis hasalready been reported (Dressler et al., 2007). In thatcross-sectional analysis, we found cultural conso-nance in five different cultural domains to beassociated with reported symptoms of psychologicaldistress. Here, we will determine if cultural con-sonance at Time 1, and change in cultural con-sonance from Time 1 to Time 2, is associated withdepressive symptoms at Time 2, controlling forprior levels of depressive symptoms.

Second, while the idea of generalized culturalconsonance is itself an interesting one, variousorientations in anthropological theory have empha-sized that different dimensions of culture have

different causal potential. Here, we propose thatthe degree of cultural consensus (or sharing) in acultural domain is a measure of its salience as anorganizing principle within a society and that lowcultural consonance in the cultural domains withthe highest cultural consensus will be experiencedas more stressful and will have the largest impacton reported depressive symptoms at the 2-yearfollow-up.

Finally, the effects of cultural consonance will becompared to another correlate of depressive symp-toms. Depressive symptoms are associated withboth chronic and acute social stressors (Paykel,2003, p. 62). As noted above, low cultural con-sonance is hypothesized to be a chronic stressor.Stressful life events, on the other hand, are acutesocial stressors that have been found to beimportant risk factors for the onset of episodes ofdepression (Paykel, 2003, p. 62). In order to rule outthe alternative hypothesis that low cultural con-sonance is a consequence of or confounded withstressful life events, we will examine the effect ofcultural consonance on depressive symptoms con-trolling for stressful life events.

This study was conducted using a mixed qualita-tive–quantitative, ethnographic–epidemiologic de-sign. The development of measurements for thequantitative, hypothesis-testing phase of the re-search depended on the results of the initialethnographic phase in which exploratory anddiscovery-oriented methods were used to examinethe shared meanings defining relevant culturaldomains within the research community. In thefollowing, the research setting and this initialethnographic phase of cultural domain analysis willbe described first, and the operational specificationof research hypotheses flowing from that analysiswill be delineated. Then the methods and results ofepidemiologic survey research to test those hypoth-eses will be presented.

The research setting and cultural domain analysis

Ribeirao Preto is a city of approximately 500,000people in the Brazilian state of Sao Paulo. It is botha financial and service center for the agriculturalindustries in the area (including the production ofsugar cane, citrus, coffee, and cattle), and a center ofeducation and health care for the region.

To adequately sample from this large andcomplex community, four socioeconomic strata,represented by residential neighborhoods, were

ARTICLE IN PRESSW.W. Dressler et al. / Social Science & Medicine 65 (2007) 2058–20692060


selected for study. These have been described indetail in other papers (Dressler, 2005; Dressler et al.,2007), so will only be briefly described here. Thepoorest neighborhood began almost 20 years ago asa favela or so-called squatter settlement. In 1994, themunicipality created a housing project, removed thefavelados (residents of a favela) to it, and then razedthe favela. The residents thus achieved legal status,but they also assumed responsibilities for payingrent and bills for municipal services. Given theireconomic marginality, many were unable to do soand have since been replaced by persons of slightlyhigher means who could afford to buy houses in theneighborhood. Nevertheless, this still is a lower classneighborhood, with the majority of residentsemployed as unskilled or semi-skilled laborers, farmworkers, and domestic servants.

The second neighborhood is a conjunto habita-

cional, a lower-middle class community that is aplanned neighborhood built in a cooperative agree-ment between the municipality and a developer.Residents qualify for a lottery to buy homes withlow-interest financing on the basis of the security oftheir employment status. This conjunto was begun inthe late 1980s, and since has evolved from a strictlyworking class neighborhood into a mixed residentialarea of working people (especially factory workers,bus drivers, and similar occupations) and lower-level white-collar workers.

The third neighborhood is a solidly middle class,mixed residential and business area that wasfounded at the turn of the 20th century, shortlyafter the city was founded. Occupational pursuitsare diverse, ranging from the owners of smallbusinesses to middle-level management in largerconcerns, as well as teachers and nurses.

The fourth neighborhood is an upper-middleclass residential area, the residents of which areprimarily upper-level professionals (doctors, law-yers, university professors), upper-level manage-ment, and owners of large businesses.

The cultural domains examined in this researchwere chosen on the basis of a combination of theoryand ethnographic intuition, and because they areorganized spheres of discourse commonly encoun-tered in everyday conversation in Brazil. Thedomains included lifestyle, social support, familylife, national identity, and food; however, forlogistical reasons, we did not examine the domainof food in the follow-up study, so that will not bediscussed or employed as a variable here. Thecultural domains of lifestyle and social support

represent individual experience of social class andsocial affiliation (Dressler & Bindon, 2000). Familylife is a cultural domain of considerable importancein daily life in Brazil, as evidenced both inethnographic data and in the scholarly literature(da Matta, 1985). National identity is also a topic ofinterest among Brazilians and a focus of Brazilianscholarly literature (da Matta, 1991).

The research was conducted in three stages:(a) cultural domain analysis, which provided bothan understanding of the domains and a foundationfor the development of measures of culturalconsonance in each domain; (b) cross-sectionalsurvey research, to collect data on cultural con-sonance and health outcomes; and (c) a follow-upinterview to assess the effects of cultural consonanceprospectively.

Cultural domain analysis (Bernard, 2006), thefirst stage of data collection, consisted of an in-depth examination of each cultural domain usingthe interviewing and data-analytic techniques ofcognitive anthropology, including freelisting; un-constrained pile sorts; rating and ranking tasks; theanalysis of these data with multidimensional scal-ing, cluster analysis, and correspondence analysis;focus group and open-ended individual interviews;and, ultimately, a cultural consensus analysis ineach domain. Freelists, pile sorts, and ranking taskswere used to generate elements of each domain andto explore the dimensions of meaning structuringcultural models in each domain. These interviewswere carried out iteratively with samples rangingfrom 15 to 45 respondents, depending on thedomain. The final cultural consensus analyses wereconducted with a sample of 66 respondents, selectedpurposively to reflect major dimensions of socialvariation in Brazil. Focus groups were conductedfor each cultural domain in each neighborhood,resulting in a total of 20 focus groups. Individualinterviews were conducted on each domain with twopersons in each neighborhood, for a total of 40individual interviews. Qualitative data from focusgroups and individual interviews (both analyzedusing standard indexing software) were used tofurther explore the content of each cultural model,relative to the organization of cultural modelsrevealed by the more structured techniques ofcultural domain analysis. In the following, thecultural domain analyses are discussed in generalterms; a more specific discussion can be found inDressler, Borges et al. (2005) and Dressler, Balieiroet al. (2005).



The cultural domains of lifestyle and socialsupport were examined in previous research inBrazil and the US (Dressler, 2005; Dressler &Bindon, 2000). Lifestyle refers both to the acquisi-tion of culturally desired material goods and torelated behaviors, especially leisure-time activities.In the literature in the social sciences, lifestyle hasbeen regarded as a means of displaying one’s placein the socioeconomic hierarchy (Bourdieu, 1984). Incultural domain analyses, lifestyle is culturallyconstructed as things necessary to lead a ‘‘goodlife.’’ Furthermore, the emphasis is not on conspic-uous consumption, but rather on what we term akind of ‘‘domestic comfort.’’ In pile sorting tasks,respondents emphasized a distinction between whatwas truly necessary for a good life (e.g., a home,furnishings for the home, basic appliances, enoughmoney for a few ‘‘extras’’) versus things that mightbe pleasant to have, but not truly necessary (e.g.,jewelry, consumer electronics). Respondents alsoemphasized leisure time activities that both broughtthem together with family and friends, and thatprovided opportunities for self-improvement. Infocus group discussions, this comfortable domesticlife was linked to notions of ‘‘tranquility’’ and‘‘relaxation.’’ In the test for cultural consensus,there was strong agreement on the lifestyle neces-sary to have a good life (eigenvalue ratio ¼ 6.59;mean competence ¼ .717.12).

Social support refers to help and assistance thatindividuals can anticipate within their social networkin times of felt need. When asked to freelist problemsfor which help is often sought, the problems listedincluded problems within the family, issues such asneeding to borrow money, and problems at theworkplace. Supporters included those within inti-mate relationships such as family and friends, as wellas co-workers, church members, and professionalssuch as a doctor. In focus group discussions,respondents emphasized the necessity of the entiresupport system (not just support from select indivi-duals) to help them cope with everyday problems inlife. In the test for cultural consensus, respondentswere asked to rank potential supporters from most-to-least important for each problem. There wasstrong agreement on these rankings (eigenvalueratio ¼ 6.53; mean competence ¼ .677.14).

For the domain of family life, respondents listedthe characteristics of families they did and did notadmire. This resulted in a set of elements arrayedalong an evaluative dimension. At the positive end ofthe dimension the elements fell into either a category

of love and affection within the family, or into acategory of structure and organization. At thenegative end of the dimension, the elements felleither into a category that described rude or boorishbehavior within the family (mal-educado in Portu-guese) versus more serious family problems such asviolence and addiction. In focus group discussions,respondents augmented these elements by includingthe importance of gendered social roles within thefamily, and emphasized the impact of social changeon traditional conceptions of the family. When askedto rank the importance of elements ‘‘for having afamily,’’ there was strong consensus (eigenvalueratio ¼ 7.92; mean competence ¼ .827.09).

Finally, the cultural domain of national identitydescribes what makes Brazilians distinctly Brazilianin their own eyes. Respondents were asked tofreelist Brazilian characteristics, and from thebeginning this was the most highly contestedcultural domain, mainly because some respondentsnamed characteristics that were regarded as un-flattering to other respondents. Additionally, in thefocus group discussions, there was considerableemphasis placed on regional differences in Brazilianidentity. The contested nature of the domain wasconfirmed in the cultural consensus analysis, inwhich there was consensus regarding these char-acteristics, but it was relatively low (eigenvalueratio ¼ 3.97; mean competence ¼ .577.19).

These cultural domain analyses served as thefoundation for the development of measures ofcultural consonance (described below). Cross-sec-tionally, cultural consonance in all domains wasassociated with psychological distress (Dressler etal., 2007). Furthermore, the measures of culturalconsonance factored into two independent dimen-sions, one describing a generalized cultural con-sonance in social factors, and one describingcultural consonance in the four dimensions of thecultural model of food (again, not included in theprospective data set). Both factors of generalizedcultural consonance were associated with psycholo-gical distress cross-sectionally (Dressler et al., 2007).

The first question that will be addressed here is ifa generalized cultural consonance is associated withdepressive symptoms assessed 2 years later. Second,we predict that as individual variables, the relativestrengths of the effects of different measures ofcultural consonance will be a function of theiroverall consensus. Specifically, the predictive effectsof individual measures of cultural consonanceshould be ordered as follows: cultural consonance



in family life, cultural consonance in lifestyle,cultural consonance in social support, and culturalconsonance in national identity. Third, all effects ofcultural consonance will be independent of theeffects of stressful life events over the 2-year period.

Methods

Simple random samples of households wereselected from each of the neighborhoods usingcomplete enumerations of occupied dwellings basedon municipal service records. Within each house-hold, the head of household, spouse, and one childover the age of 18 were invited to participate in theresearch. This resulted in a sample of 271 persons,which is a response rate of 72.3% (based on thenumber of people completing the research protocoldivided by the number of people who were invitedto participate). The research protocol was reviewedand approved by both the Institutional ReviewBoard for the Protection of Human Subjects of TheUniversity of Alabama, and the Ethics Committeeof the Faculty of Medicine of Ribeirao Preto, theUniversity of Sao Paulo.

For the 2-year follow-up, 210 members of theoriginal sample agreed to be re-interviewed (77.5%of the original sample). In the original datacollection, in addition to the initial interview inwhich sociocultural and psychological data werecollected, we carried out two dietary interviews anda clinical interview. We did not have sufficientresources to collect all of these data at follow-up,so the interviews on cultural consonance in foodwere dropped; hence, those variables are notincluded here.

The dependent variable in the research reportedhere is a Brazilian Portuguese version of the Centerfor Epidemiologic Studies Depression Scale (CES-D), translated and validated by da Silveira & Jorge(2000). The CES-D has acceptable reliability in thissample (Cronbach’s alpha ¼ .88).

Three covariates are employed: age (assessed asreported age in years); gender (coded as women ¼ 0and men ¼ 1); and, socioeconomic status (SES).SES was assessed as a composite scale of householdincome (reported in number of minimum salariesper month), education (reported as years of educa-tion completed), and occupational status. The latterwas coded using the scale reported in Pastore (1982)developed in Brazil. A single principal componentaccounts for 60.7% of the shared variance in these

three variables, and the principal component scoreis used as the measure of SES.

Each measure of cultural consonance was calcu-lated in general in the same way. Using the outputfrom cultural consensus analysis that provides theculturally best responses to the questions posed ineach domain, we designed questions to be asked ofindividuals about their own beliefs and behaviors.Then, the degree to which their reports of theirbeliefs or behaviors matched the answers from thecultural consensus analysis was calculated tooperationalize cultural consonance in that domain.In the following, we briefly describe the derivationof these measures; a detailed discussion of eachmeasure can be found in Dressler, Borges et al.(2005) and Dressler, Balieiro et al. (2005).

The measurement of cultural consonance inlifestyle assessed the degree to which individualscorrespond to the model in their own behavior. Formaterial goods, respondents indicated whether ornot they owned each item. On leisure activities,respondents reported on a four-point scale (fromnever to several times per week) the frequency withwhich they engaged in those behaviors. To calculatecultural consonance in lifestyle, the number ofmaterial items owned by, and the number of leisurebehaviors reported by, the individual were summedthat were rated in the cultural consensus analysis asbeing at least ‘‘important’’ in having a good life.This sum was then divided by the total to yield aproportion. The closer this proportion is to 1.0, themore the respondent’s own lifestyle matches that ofthe consensus model.

To measure cultural consonance in social sup-port, respondents were provided with sets of cardsfor the seven potential supporters and were asked torank order them in terms of seeking support fromthem for each of the nine problems. To calculatecultural consonance in social support, the datamatrix was transposed so that each respondentbecame a column and each row was that respon-dent’s ranking of a particular supporter for aparticular problem. A column was added that wasthe corresponding cultural consensus ranking. Wethen calculated correlation coefficients between therankings by each respondent and the rankings fromthe consensus analysis. These correlations are usedas the measure of cultural consonance in socialsupport.

To measure cultural consonance in family life weassessed individuals’ perceptions of their families.Respondents were presented with a statement about



the family, phrased explicitly in terms of their ownfamily, and asked to agree or to disagree with thatstatement. For each concept from the culturaldomain analysis, we generated a statement describ-ing the family that would be acceptable in everydayspeech. We generated a minimum of one sentencefor each concept, although for several of the moresalient items we generated more than one statement.Examples of items include: ‘‘In my family we feelclose to one another,’’ and ‘‘At times I wish myfamily was more organized.’’ All items were writtenoriginally in Portuguese. To arrive at the totalscale score, we weighted an individual’s responsesby the importance of each particular conceptestimated from the cultural consensus analysis.The scale has high internal consistency reliability(alpha ¼ .89).

For national identity, as in the domain of familylife, we wrote items that represented personal beliefsin this domain and asked respondents to agree ordisagree with the statements (again, all items werewritten originally in Portuguese). In scaling theitems, we first experimented with a unidimensionalscale of all of the items, but this had very lowinternal consistency. Then, a series of exploratoryfactor analyses (principal components analysis withvarimax rotation) were carried out, successivelydeleting items that had low communalities. It wasapparent that the items referring to the mostpositive characteristics of Brazilian life had verylow variability (i.e., people endorsed the mostpositive items). There was, however, a consistentfactor that combined most of the unfavorable viewsof Brazilians. These included perceptions of othersas unhelpful; of relying on the Brazilian jeito (a wayof bending the rules); and, taking advantage ofothers (see da Matta, 1991). A scale of the eightitems has acceptable internal consistency reliability(alpha ¼ .69) and represents a scale of ‘‘culturalcynicism.’’ That is, those individuals who endorsemore of the items have a more cynical view ofBrazilians and Brazilian life, but it is a distinctlyculturally constructed cynicism.

All four of the cultural consonance variables loada single principal component that explains 46% ofthe variance among the measures. Principal compo-nent loadings are as follows: cultural consonance inlifestyle ¼ .78, cultural consonance in socialsupport ¼ .62, cultural consonance in familylife ¼ .58, and cultural cynicism ¼ �.77. Theprincipal component score is used as a measure ofgeneralized cultural consonance.

Stressful life events were collected using aninventory of events developed and validated inBrazil (Savoia, 2000). This inventory consists of 10major events that are culturally regarded as negative(such as death of a family member, maritalseparation, and loss of a job) The number of eventsoccurring in the year prior to the follow-up inter-view were counted.

Results

Descriptive statistics are shown in Table 1. Whenthe original sample (n ¼ 271, data not shown) andthe follow-up sample (n ¼ 210) were compared onall variables used in this analysis, there were nodifferences. A number of the variables used in theanalysis differ between Time 1 and Time 2 for thefollow-up subsample, with some values increasingover the time period and others decreasing.

To test the prospective effects of cultural con-sonance on depressive symptoms, a regression

ARTICLE IN PRESS

Table 1

Descriptive statistics for the follow-up sample

Variable Descriptive statistics

(n ¼ 210)

Age 40.6 (711.3)

Sex (% men) 37.0

SES .01 (71.0)

Stressful life events 2.2 (71.7)

Depressive symptoms

Time 1 12.4 (710.1)

Time 2 11.9 (79.7)

Cultural consonance in lifestyle

Time 1 .67 (7.15)

Time 2a .69 (7.16)

Cultural consonance in social

support

Time 1 .49 (7.18)

Time 2b .51 (7.16)

Cultural consonance in family life

Time 1 106.5 (724.8)

Time 2a 102.7 (721.7)

Cultural cynicism

Time 1 11.6 (73.7)

Time 2a 11.2 (73.4)

Generalized cultural consonance

Time 1 .02 (7.9)

Time 2 .00 (71.0)

aTime 1 versus Time 2, po.05.bTime 1 versus Time 2, po.10.

W.W. Dressler et al. / Social Science & Medicine 65 (2007) 2058–20692064


model was used (Singer & Willett, 2003). Thegeneral form of the model is as follows:

Y 2 � Y 1 ¼ aþ b1Y time 1 þ b2X time 1

þ b3X time 2�time 1 þ e,

where ‘Y’ refers to the dependent variable and ‘X’refers to an independent variable. In formatting theregression equation in this way, change over time inthe dependent variable is examined as a function ofchange over time in the independent variable, whilecontrolling for initial levels of both independent anddependent variables, meaning that both the effectsof initial levels of the independent variables, as wellas their effects as a function of change over time,can be examined.

Table 2 shows the regression of the change in theCES-D in relation to both the initial levels ofgeneralized cultural consonance and the change ingeneralized cultural consonance over time. General-ized cultural consonance at Time 1 has an effect ondepressive symptoms at Time 2, controlling forinitial levels of depressive symptoms. Also, asgeneralized cultural consonance increases over the2-year follow-up, CES-D scores decrease, control-ling for initial levels of both cultural consonanceand depressive symptoms. The occurrence of stress-ful life events in the year prior to the follow-upinterview also has a significant effect on change inCES-D scores over the 2 years, but this effect is

independent of the effect of generalized culturalconsonance.

Table 3 shows the regression of the change inCES-D scores in relation to both the initial levels ofcultural consonance and the change in culturalconsonance over time, but with the four measures ofcultural consonance entered separately (in thisanalysis only three of the tolerance values associatedwith the cultural consonance variables—includingboth the Time 1 measures and the change mea-sures—were less than .50, and those three were inthe .40–.50 range; there is thus no problem withcollinearity). There is a small effect of initial levelsof cultural consonance in family life on the CES-Dat follow-up. Otherwise, change in cultural con-sonance in family life and change in culturalconsonance in lifestyle over the 2-year follow-upare associated with change in CES-D scores,controlling for initial levels of each cultural con-sonance variable and depressive symptoms. Again,these effects are independent of the effect of stressfullife events.

ARTICLE IN PRESS

Table 2

Regression of change in depressive symptoms from Time 1 to

Time 2 on generalized cultural consonance, stressful life events,

and covariates

Variables Unstandardized

regression

coefficient

Standardized

regression

coefficient

Age �.011 �.015

Sex �2.457* �.137*

SESa .799 .094

CES-Dtime 1b

�.483** �.569**

Cultural

consonancetime 1

�1.472* �.168*

Stressful life events .766* .152*

Cultural

consonancetime2 – time 1

�2.798** �.200**

Intercept 5.124* –

Multiple R .556** –

Multiple R2 .309 –

*po.05.

**po.01.aSES, socioeconomic status.bCES-D, Center for Epidemiologic Studies Depression Scale.

Table 3

Regression of change in depressive symptoms from Time 1 to

Time 2 on individual measures of cultural consonance, stressful

life events, and covariates

Variables: Unstandardized

regression

coefficient

Standardized

regression

coefficient

Age .033 .004

Sex �2.477* �.138*

SESa .923 .108

CES-Dtime 1b

�.476** �.560**

CCFamily lifetime 1c

�.049+ �.139+

CCLifestyletime 1d

�6.562 �.120

CCSocial supporttime 1e 5.473 .118

Cultural cynicismtime 1 .270 .116

Stressful life events .840** .167**

CCFamily lifetime2�time1 �.119** �.271**

CCLifestyletime2�time 1 �12.006* �.137*

CCSocial supporttime

2�time 1

4.610 .087

Cultural cynicismtime

2�time 1

.224 .082

Intercept 7.871 –

Multiple R .593** –

Multiple R2 .351 –

*po.05.

**po.01.aSES, socioeconomic status.bCES-D, Center for Epidemiologic Studies Depression Scale.cCCFamily Life, cultural consonance in family life.dCCLifestyle, cultural consonance in lifestyle.eCCSocial support, cultural consonance in social support.

W.W. Dressler et al. / Social Science & Medicine 65 (2007) 2058–2069 2065


Discussion

This paper had three aims: (a) to determine ifchange in generalized cultural consonance over a2-year period was associated with change indepressive symptoms, (b) to predict and test whichindividual measures of cultural consonance wouldbe most important in determining change indepressive symptoms, and (c) to determine if anyof these effects were independent of the effect of amajor known risk factor for depression, stressfullife events.

With respect to the first aim, the analysisconfirmed that change in generalized culturalconsonance over the 2-year period was associatedwith change in depressive symptoms. Controllingfor the initial level of cultural consonance, asgeneralized cultural consonance increased over thetime period, depressive symptoms reported at thefollow-up interview declined from those reported atthe initial interview, controlling for the initial levelof depressive symptoms (and for demographicvariables with known associations with depressivesymptoms). Furthermore, this effect is independentof the effect of stressful life events.

With respect to the second aim, we predictedthat the importance of the individual measures ofcultural consonance would be a function of theiroverall level of consensus. Based on culturalconsensus analyses, this meant that the rankorder of the importance of the effect of changeover time in cultural consonance on depressivesymptoms would be as follows: cultural consonancein family life, cultural consonance in lifestyle,cultural consonance in social support, and culturalcynicism. The results largely supported thisprediction. As predicted, after controlling for initiallevels of cultural consonance in family life, changeover time in cultural consonance in family life hadthe strongest effect on changing depressive symp-toms. Also, as predicted, change over time incultural consonance in lifestyle had the nextstrongest effect on changing depressive symptoms.With respect to cultural consonance in socialsupport and cultural cynicism, however, the sup-port for the predictions was less clear, bothbecause neither of these variables was statisticallysignificant in this analysis, and because the stan-dardized regression coefficients for each culturalconsonance variable were approximately equal. Thesupport for the second prediction is thereforequalified.

There does, nevertheless, appear to be some meritin regarding the overall cultural consensus within acultural domain as an index of the causal potentialwith respect to cultural consonance within thatdomain. The foundation for this prediction deservesadditional discussion. We have argued that lowcultural consonance is a chronically stressful ex-perience for persons precisely because knowledge ofthe cultural domains is so widely shared. Culturalmodels both guide our own beliefs and behaviorsand shape our expectations for others’ beliefs andbehaviors. Where there is a wide consensus, there isprobably very little conscious reflection upon ourexpectations of the structure of other peoples’beliefs and behaviors within a domain. This isprecisely what would make the gap betweenexpectation and experience so apparent in mundanesocial interaction. ‘‘Everyone’’ knows how everydaylife is supposed to be structured within a particulardomain, but the individual with low culturalconsonance in that domain sees herself, and is seenby others, not to correspond to that model. As wehave argued elsewhere, this is likely to lead to thekind of psychological dilemma that Antonovsky(1981) described as a lack of ‘‘sense of coherence.’’

Of course, not ‘‘everyone’’ knows or agrees uponthe cultural model. As we have shown elsewhere(Dressler, 2005; Dressler, Balieiro et al., 2005;Dressler, Borges et al., 2005), there are significantdifferences between groups defined on the basis ofage, gender, and education in cultural competence inspecific cultural domains, although there is noconsistent pattern. For example, in the domain offamily life there are significant differences in culturalcompetence between men and women; women havehigher cultural competence (mean competence ¼ .85)than men (mean competence ¼ .79). But this is not tosay that there is not a high level of competenceamong men, it only means that women’s competenceis that much higher (which, on the basis of our focusgroup data, we interpret as reflecting the importanceof women in that domain). Also, there is aninteresting form of contention over the model. Boththe cultural consensus analysis and our focus groupdata show that there is a lively dispute over therelative importance of family organization versus theemotional climate of the family in terms of creating agood Brazilian family. But this does not mean thatpeople do not agree that structure and affect are bothimportant; they merely dispute which is mostimportant. A larger discussion of this point wouldtake us too far afield; it is important to emphasize,



however, that a strong cultural consensus does notpreclude systematic variability and contention withinthat consensus, and that this systematic variabilityand contention does not change the fact that thiscultural model structures the environment of mean-ing and understanding within which people live theirdaily lives. (It is worth noting, too, that if theregression models in Table 3 are run separately formen and women, there are no differences.)

It might be objected that our emphasis on culturalconsensus and the stressful nature of a lack ofconsonance with that consensus ignores individualagency. It might be argued that people know theconventional cultural model, but that they reject itin favor of an alternative and more personalconstruction of meaning, and that we should beexamining consonance with those personal mean-ings. Curiously, if it were possible to separate outpersonal and cultural meaning here, the likely resultwould be stronger effects of cultural consonance inthe models. If an additional measure of some kindof personal consonance was included, what thiswould do in a statistical model is account for someof the error variance, i.e., the residual varianceassociated with those people who have low culturalconsonance and low depressive symptoms, or highcultural consonance and high depressive symptoms.Accounting for that error variance would likelyreduce the standard error associated with theestimate of the effect of cultural consonance, andthe estimate of the effect would at least be morestatistically reliable. So, our point here is that evengiven the fact of intracultural diversity and the factthat culture is contested, and granting the possibilitythat consonance with personal (as opposed tocultural) meaning may be important, individualconsonance with a shared model of culture haspowerful causal potential with respect to psycholo-gical well-being.

Returning to the causal ordering of the variables,there are good reasons why the domains of familylife and lifestyle are so important. Classic Brazilianethnography places considerable emphasis on thefamily as a core institution in Brazilian society (daMatta, 1985; Wagley, 1971). While the family itselfis changing substantially in Brazil, its salience as aninstitution is not. This was very clear in our focusgroup discussions. The focus groups converged withthe more structured cultural domain analyses inidentifying the salient elements of the cultural modelof the family—specifically structure and affect—aswell as contention over the relative importance of

these elements. And several additional themesemerged as we observed discussants employ thesemodels in the focus group. One of these was thenotion of the family being basic, almost primordial,with respect to the structure of society. Focus groupparticipants also emphasized the role of the motherwithin the family in virtually metaphysical terms.According to the discussants, it is the ‘‘nature’’ ofthe woman within the family to bring a special kindof ‘‘understanding’’ and ‘‘comprehension’’ to thenurturing of children, which in turn leads thosechildren along the morally correct path in socializa-tion. This central role of the family in Braziliansociety is key to the significance of culturalconsonance in family life as a predictor of psycho-logical well-being. A higher cultural consonance infamily life means to the individual, and to thosewith whom that individual interacts, that he or shehas achieved something of fundamental importancein Brazilian life.

The cultural domain of lifestyle is interestinglydifferent from that of family life, in that to thedegree that focus group participants emphasized theimportance of the family, they downplayed theimportance of lifestyle. Focus group participantswere quite critical of what would be regarded asconspicuous consumption; rather, as we suggestedsome years ago (Dressler, Santos, & Balieiro, 1996),the emphasis in the discussion was on a domesticcomfort, not on status display. When discussinglifestyles, focus group participants did emphasizethat a certain basic economic stability was necessaryto live comfortably, but in doing so participantsrepeatedly returned to terms like ‘‘tranquil,’’‘‘calm,’’ and ‘‘relaxed’’ when discussing a style oflife. Again, achieving this tranquility depends inpart on achieving a certain level of material well-being, but the point of material well-being is toprovide the context within which a tranquil life canbe lived.

At one level, both cultural consonance in familylife and cultural consonance in lifestyle contrastwith the domains of social support and nationalidentity (cultural cynicism) in terms that da Matta(1985) identified in discussing Brazilian society. Hecontrasted three ‘‘spaces’’ of Brazilian social life: a

casa (the home), a rua (the street), and outros

espac-os (other spaces). The home refers to theloving bonds of the family and, in terms we mightborrow from our focus group participants, the spacein which one’s tranquil and relaxed lifestyle can beconstructed and enjoyed. The street refers to the



impersonal world of economics and politics, withinwhich one is anonymous and, hence, vulnerable.And, other spaces refer to the metaphysical, theunseen, and the unknown. As cultural domains,family life and lifestyle describe a casa, while thebonds of social support as well as the nature ofBrazilian identity take one out into a rua, if not intooutros espac-os. It can be argued, therefore, that thegreater causal impact of cultural consonance infamily life and cultural consonance in lifestyle is dueprecisely to the centrality of these domains inBrazilian social life (although it is possible thatthese cultural domains have an importance inhuman social life that transcends the specificBrazilian cultural context; this is a question forfuture research).

These measures of cultural consonance identify aseparate causal pathway in the development ofdepressive symptoms. This is evidenced by theindependence of their effects from the effect ofstressful life events. The effect of stressful life eventsobserved here is comparable to the effect of thisvariable observed in many epidemiologic studies ofdepressive symptoms. But the influence of culturalconsonance is separate from that risk. As othershave noted (Paykel, 2003), both chronic (lowcultural consonance) and acute (stressful life events)stressors are associated with higher depressivesymptoms. In some instances, low cultural con-sonance and certain kinds of stressful life eventsmay be related (for example, loss of employmentmay lead to lower cultural consonance in lifestyle).But in general, the effects of a chronically stressfulcircumstance like low cultural consonance areseparable from the effects of acute stressors like lifeevents.

It is worth noting, too, that cultural consonance isassociated prospectively with depressive symptomsindependently of other known correlates of depres-sion, such as low SES. This is particularly interest-ing in that it might be argued, for example, thatcultural consonance in lifestyle is a proxy for SES.But the results obtained here indicate that it is notmerely having material resources that are importantwith respect to individual well-being; rather, it is theinvestment of those material resources in a lifestylethat is collectively valued and that meaningfullycommunicates to others one’s position in the socialstructure. In future work, it will be interesting tocompare and contrast the effects of culturalconsonance in various domains with other knowncorrelates of the risk of depression.

There are several limitations of this study. First,the sample size is relatively modest; nevertheless,there was ample power to detect the effects ofcultural consonance in family life and culturalconsonance in lifestyle. Post hoc statistical poweranalyses show that the regression model employedhere has power beyond the .80 level for bothanalyses. Second, we did not assess stressful lifeevents at both time periods, relying instead onaccurate recall at the second interview. There is littleevidence, however, that individuals over-report thelarge-scale events (e.g., death of a family member,divorce) that make up the scale used here. Third,like all regression models, the results here assumethat any unmeasured variables not included in theanalysis would be uncorrelated with cultural con-sonance, and this may not be a reasonable assump-tion. And, finally, a 2-year follow-up period wasused purely for convenience, but this may not be theoptimal period for evaluating the prospective effectsof cultural consonance. This can be addressed infuture research.

Ultimately, these results continue to support theimportance of cultural consonance as a significantfactor in individual well-being. These results areparticularly important in showing that change overtime in cultural consonance is associated withchange in depressive symptoms, and in suggestinghow cultural consonance in different domains mayhave different causal potential.

Acknowledgments

This research was supported by a research grantfrom the National Science Foundation (BCS-0090193). Roy D’Andrade and Kathryn S. Othsoffered helpful comments on an earlier draft of thispaper.

References

Antonovsky, A. (1981). Health, stress, and coping. San Francisco,

CA: Jossey-Bass.

Bernard, H. R. (2006). Research methods in anthropology (4th

ed.). Lanham, MD: Altamira Press.

Bourdieu, P. (1984). Distinction: A social critique of the judgment

of taste (Richard Nice, Trans.). Cambridge, MA: Harvard

University Press.

da Matta, R. (1985). A casa e a rua. Sao Paulo, Brazil: Editora

Brasiliense.

da Matta, R. (1991). Carnivals, rogues, and heroes: An

interpretation of the Brazilian dilemma (J. Drury, Trans.).

Notre Dame, IN: Notre Dame University Press.



D’Andrade, R. G. (1984). Cultural meaning systems. In

R. A. Schweder, & R. A. Levine (Eds.), Culture theory:

Essays on mind, self and emotion (pp. 88–119). Cambridge:

Cambridge University Press.

D’Andrade, R. G. (1995). The development of cognitive anthro-

pology. Cambridge: Cambridge University Press.

da Silveira, D. X., & Jorge, M. R. (2000). Escalade rastreamento

populacional para depressao (CES-D) em populac- oes clınica

e nao-clınica de adolescentes e adultos jovens. In

C. Gorenstein, L. H. S. G. Andrade, & A. G. Zuardi (Eds.),

Escalas de avaliac- ao clınica em psiquiatria and psicofarmaco-

logia (pp. 125–138). Sao Paulo, Brazil: Lemos-Editorial.

Dressler, W. W. (2005). What’s cultural about biocultural

research? Ethos, 33, 20–45.

Dressler, W. W., Balieiro, M. C., Ribeiro, R. P., & Santos, J. E.

D. (2005). Cultural consonance and arterial blood pressure in

urban Brazil. Social Science & Medicine, 61, 527–540.

Dressler, W. W., Balieiro, M. C., Ribeiro, R. P., & Santos, J. E.

D. (2007). Cultural consonance and psychological distress:

Examining the associations in multiple cultural domains.

Culture, Medicine and Psychiatry, 31, 195–224.

Dressler, W. W., & Bindon, J. R. (2000). The health consequences

of cultural consonance: Cultural dimensions of lifestyle, social

support and arterial blood pressure in an African American

community. American Anthropologist, 102, 244–260.

Dressler, W. W., Borges, C. D., Balieiro, M. C., & Santos, J. E.

D. (2005). Measuring cultural consonance: Examples with

special reference to measurement theory in anthropology.

Field Methods, 17, 331–335.

Dressler, W. W., Santos, J. E. D., & Balieiro, M. C. (1996).

Studying diversity and sharing in culture: An example of

lifestyle in Brazil. Journal of Anthropological Research, 52,

331–353.

Goodenough, W. H. (1996). Culture. In D. Levinson, &

M. Ember (Eds.), Encyclopedia of cultural anthropology

(pp. 291–299). New York: Henry Holt.

Pastore, J. (1982). Inequality and social mobility in Brazil.

Madison, WI: University of Wisconsin Press.

Paykel, E. S. (2003). Life events and affective disorders. Acta

Psychiatrica Scandanavica, 418(Suppl.), 61–66.

Romney, AK., Weller, S. C., & Batchelder, W. H. (1986). Culture

as consensus: A theory of culture and informant accuracy.

American Anthropologist, 88, 313–338.

Savoia, M. G. (2000). Instrumentos para avaliac- ao de eventos

vitais e de estrategias de enfrentamento (coping) em situac- oes

de estresse. In C. Gorenstein, L. H. S. G. Andrade, &

A. G. Zuardi (Eds.), Escalas de Avaliac- ao Clınica em

Psiquiatria and Psicofarmacologia (pp. 377–386). Sao Paulo,

Brazil: Lemos-Editorial.

Searle, J. (1995). The construction of social reality. New York:

Free Press.

Singer, J. D., & Willett, J. B. (2003). Applied longitudinal analysis:

Modeling change and event occurrence. Oxford: Oxford

University Press.

Sperber, D. (1985). Anthropology and psychology—Towards an

epidemiology of representations. Man, 20, 73–89.

Wagley, C. (1971). An introduction to Brazil (revised ed.).

New York: Columbia University Press.


Dressler, William W., Mauro C. Balieiro, Rosane P. Ribeiro, and José Ernesto dos Santos. (2007) A...

Documents

Transcript of Dressler, William W., Mauro C. Balieiro, Rosane P. Ribeiro, and José Ernesto dos Santos. (2007) A...