Abstract
Theoretical and methodological developments guided by a cognitive theory of culture have advanced our understanding of cultural processes over the past 40 years. The theoretical construct of cultural models, developed in the 1970s, provided a more precise definition of culture. The cultural consensus model, introduced in 1986, enabled investigators to verify and analyze the degree to which culture was shared and how it was distributed. Subsequent advances in the cultural consensus model, especially the analysis of residual agreement, provided a more complete approach to describing intracultural diversity. Finally, the concept and measurement of cultural consonance, introduced in 1996, demonstrated how shared cultural models link to social practice. This article provides a brief overview of these trends in the study of culture, with an emphasis on how this theory and these methods have been applied in research.
Virtually all anthropology students are taught the definition of culture as all things human introduced by E. B. Tylor in the 19th century; few learn he also suggested that “The quality of mankind which tends most to make the systematic study of civilization possible is that remarkable tacit consensus…which so far induces whole populations…to settle down to the same general level of art and knowledge” (Tylor 1874:10, emphasis added). That culture is shared became a fundamental assumption. A recent approach to the study of culture has been to reduce the scope of the concept to render it operational and to increase its power as an explanatory construct (Blount 2011). This theoretical work led to the contemporary concept of “cultural models” to describe both how culture is encoded as knowledge in individual minds and how it is shared, creating collective cultures (D’Andrade 1995).
In the mid-1980s, Romney et al. (1986) introduced “cultural consensus analysis,” which, for the first time in anthropology, provided a theoretically and methodologically satisfying way of verifying and describing the sharing of culture. Beginning also in 1986, extensions of the cultural consensus model to better describe intracultural variation were introduced (Boster 1986; Boster and Johnson 1989). Later, Dressler (1996) proposed the concept and measure of “cultural consonance” or the degree to which individuals incorporate shared cultural knowledge into their own beliefs and behavior. This linked cultural consensus to social practice.
This article provides a brief overview of these developments. It aims to show how they converge in the application of cognitive culture theory to a diverse set of questions in the social sciences.
Cultural Consensus
The cultural consensus model systematizes what ethnographers have always done: Based on informants’ responses to a set of questions, the ethnographer observes how an understanding of the world is shared and distributed to infer what cultural models guide social practice. Cultural consensus analysis flows from basic ethnography. The first step is to identify a cultural domain or an organized sphere of discourse—literally, some topic of sufficient salience in a society that people shape everyday narrative and interaction around. After identifying a domain, the elements that make up the domain are elicited (e.g., what are the names of illnesses that populate the domain of “sickness?”), along with meaningful dimensions linking those elements (e.g., are illnesses linked and differentiated by symptoms? severity? chronicity? causation? treatment?). In short, how is the domain configured? There are specialized interviewing techniques for cultural domain analysis (see Borgatti 1999).
Once the configuration of the cultural domain is described, a structured data collection instrument is prepared to test for cultural consensus. This instrument may consist of true–false questions or a rating or ranking task. It is assumed that these questions provide an adequate sampling of the knowledge organizing the cultural domain because the immediate focus is not on the knowledge itself but rather on the people who possess that knowledge. In cultural consensus analysis, the first step is to calculate the degree of similarity in responses between each pair of respondents.
The initial question addressed in cultural consensus analysis is: Is the consistency in responses among informants sufficient to infer that they are drawing on a shared, underlying cultural model? Factor analytic methods are employed to estimate overall agreement. If the first eigenvalue is several times larger than the second (the ratio of 3:1 was considered sufficient), then there is consensus among respondents, suggesting a common, underlying cultural model. A large first eigenvalue indicates that respondents have high loadings on the first factor, and hence, a high correlation with the latent variable identified. That latent variable is the aggregate knowledge encoded in the cultural model. Therefore, the first factor loading indicates how strongly the respondent agrees with the overall model and is referred to as “cultural competence.” The correlation between any pair of respondents is the product of their respective cultural competence. Hence, individuals in a society are seen to agree with one another to the extent that they each agree with the underlying cultural model (see Romney et al. [1986] and Weller [2007] for more details).
Once consensus has been confirmed, the underlying, aggregate model, or the “culturally best” answers to the questions (referred to as the “cultural answer key”), can be estimated. The best estimate of the underlying knowledge base is not a simple average of what people respond but rather a weighted average, giving higher weight to individuals with higher cultural competence.
The cultural consensus model has been used widely in various fields. A sampling of cultural domains examined include: health (Alang 2018; Alemi et al. 2017; Brown et al. 2018; Chavez et al. 1995a, 1995b; Copeland 2011; Gartin et al. 2010; Mundorf et al. 2017; Ross et al. 2011; Weller and Baer 2001; Weller et al. 1993, 2002), food and food systems (Johnson and Griffith 1996; Newkirk et al. 2009; Oths et al. 2003), ethnobotanical and ethnobiological knowledge (Atran et al. 2005; Godoy et al. 2009; Medin et al. 2007; Rocha 2005; Ross 2002), organizational culture (Caulkins and Hyatt 1999; Collins and Dressler 2008; Jaskyte and Dressler 2004, 2005; Keller and Lowenstein 2011; Rinne and Fairweather 2012; Smith et al. 2004), fisheries management (Carothers et al. 2014; Frawley et al. 2019; Johnson and Griffith 2010; Naves et al. 2015), race and ethnicity (Crafa et al. 2019; Dengah et al. 2019; Gravlee 2005), social organization and life goals (Dressler et al. 2017; Kennedy et al. 2013; Koster 2011; Snodgrass et al. 2014; Torres 2009), and tourism (Bae and Chick 2017; Paris et al. 2015). This is not an exhaustive list of studies, but it does provide a sample of the diverse problems explored with cultural consensus analysis.
One major aim in studies of cultural consensus is to examine the distribution of cultural models (Gatewood 2012). Several approaches have been employed. Baer et al. (2004) examined the distribution of cultural models of AIDS, comparing physicians and lay persons in Mexico and the United States. They found an overall shared consensus model of AIDS across the two societies; however, physicians and lay persons within each society shared more than they did with corresponding samples in the other society.
Baer et al.’s (2004) analyses retained a focus on the aggregate, comparing groups of respondents. In their study of innovation in organizations, Jaskyte and Dressler (2004, 2005) also employed aggregate results. For each of a sample of 20 organizations, consensus data on organizational values were collected. Each organization was then assessed in terms of the degree of consensus regarding these values (using both the eigenvalue ratio and mean cultural competence), and these parameters were in turn examined in relation to an aggregate measure of organizational innovation.
There has also been interest in examining cultural consensus in relation to individual-difference variables. The only indicator of cultural consensus for this type of analysis is the individual cultural competence coefficient, which has been used as both an independent and a dependent variable in different studies. For example, Hopkins (2011) found, in a study of a village in Mexico, that persons who were more central in the social network of the community also had higher cultural competence in ethnobotanical knowledge. McDade et al. (2007) found in the Bolivian Amazon that children of mothers with higher cultural competence in ethnobotanical knowledge had lower immune system challenge. And Copeland (2018a) found that higher cultural competence in a model of HIV management was associated with better health status in the marginal communities of Nairobi.
One challenge in studying the distribution of cultural competence is that in highly shared models, there is relatively little variation in cultural competence coefficients. Hruschka and Maupin (2013) offer some guidance in this respect.
Residual Agreement
In the same year as the cultural consensus model was introduced, Boster (1986) introduced the concept of “residual agreement.” As noted, the correlation between two respondents is the product of their respective cultural competence. To the extent the expected correlation underestimates the observed correlation, there is agreement beyond the overall cultural consensus. Boster (1986) originally employed a comparison of observed and expected matrices of correlations among respondents to test for residual agreement. Among Amazonian horticulturalists, he found stronger agreement within families in the naming of cultigens than among the overall community.
Boster and Johnson (1989) demonstrated that residual agreement is carried by the second factor in cultural consensus analysis. Although there is no simple interpretation of the second factor loading for a respondent, the second factor in cultural consensus analysis divides respondents into two more-or-less equal groups: those with positive loadings and those with negative loadings. Respondents with same-sign loadings on the second factor tend to agree with one another more. This does not mean that there is no overall cultural consensus; rather, within the context of overall cultural consensus, subgroups of respondents tend to privilege certain aspects of that knowledge over other aspects.
A number of techniques have been used to examine residual agreement. Most commonly, the mean or modal responses of the subgroups have been directly compared (Hruschka et al. 2008; Ross and Medin 2005).
Dressler et al. (2015) introduced a technique for describing these tendencies, taking into account overall cultural consensus. Since residual agreement refers to divergence from the overall cultural consensus, the difference between each respondent’s answer and the cultural answer key can be calculated (this works best for rating scales). The subgroups’ (i.e., respondents with positive versus negative second factor loadings) difference scores can be averaged, and the mean difference scores of one subgroup plotted versus the other subgroup. Alternately, separate cultural consensus analyses can be conducted within each residual agreement group, and the difference between the overall cultural answer key and the subgroup cultural answer key can be calculated for each item (which works best for rankings). In samples with strong residual agreement, there is a large inverse correlation (r > –.8) between the mean difference scores, indicating that one set of respondents rates that subset of items as more important than the overall cultural consensus, while the other set of respondents rates that subset of items as less important than the overall cultural consensus.
In Boster and Johnson’s (1989) original article employing residual agreement coefficients (second factor loadings), they found that novice and expert fishermen differed in their sorting of fish into categories, with the novice fishermen using simpler criteria (e.g., shape, size) and experts using more complex criteria (e.g., behavior). Dressler et al. (2015) found significant residual agreement in four cultural domains in Brazil (lifestyle, social support, family life, and national identity), each of which also had high overall cultural consensus. Using the residual agreement coefficients as dependent variables, they found that residual agreement in different domains was associated with different background variables (including age, education, and time period). Dengah (2013) examined a cultural model of religious life among Brazilian Pentecostal Protestants in two different churches. While there was an overall cultural consensus, in one church there was a stronger emphasis placed on a contemplative religiosity, whereas in the other there was a stronger emphasis on ecstatic religious practices, including glossolalia and religious healing. Schultz (2019) examined a cultural consensus model of material lifestyles among the Tsimane’ of lowland Bolivia, finding that, in the context of an overall consensus on the importance of specific material goods in the Tsimane’ way of life, one subgroup emphasized purely market (or Western) material goods, while the other emphasized a syncretic model of market and traditional material goods.
Henderson and Dressler (2017, 2019, 2020) used residual agreement coefficients as independent variables. Examining cultural models of risk for substance misuse among university students in the United States and Brazil, they found that many risk factors were seen as potentially important, but there was residual agreement regarding which were most important. Some students emphasized biopsychosocial factors more, while others emphasized moral and character flaws. Where students fell along this continuum was associated with a tendency to stigmatize substance users.
Cultural Consonance
People do not simply know things, they do things as well. The concept of cultural consonance or how well individuals match, in their personal beliefs and behaviors, the shared prototypes for belief and behavior encoded in cultural models captures this social practice (Dressler 1996, 2018). The measurement of cultural consonance flows directly from cultural domain analysis employing the cultural consensus model. Once a shared cultural model has been identified, the cultural answer key can be used to formulate questions that pertain directly to the individual’s beliefs and behaviors rather than to the community at large (Dressler, Borges et al. 2005).
The precise measurement of cultural consonance is tailored to the specific cultural domain. For example, the concept of “lifestyle” is often employed to refer to the material goods and leisure-time activities associated with middle-class comfort. In Brazil, I found a strong cultural consensus regarding the lifestyle important for “having a good life.” The consensus emphasized domestic comfort and not conspicuous consumption. To measure cultural consonance, I simply asked survey respondents whether they possessed the material goods and engaged in the leisure-time activities that were rated as at least “somewhat important” for having a good life as defined by the cultural consensus model. The resulting scale of cultural consonance in lifestyle was approximately normally distributed; the average respondent possessed or engaged in about two-thirds of the items. It is important to emphasize, however, that the cultural expectation is to have all of the items or engage in all of the activities rated as somewhat important in the consensus model; therefore, for many individuals, there was a clear mismatch between cultural expectation and social practice (Dressler 2018:120–43).
Measuring cultural consonance in other domains can require different approaches. For example, in the domain of family life, there was a strong cultural consensus around the characteristics of a good Brazilian family. These included both a specific emotional climate within the family and the structure and organization of the family. For many of the emotionally charged characteristics (e.g., “love,” “union,” “understanding”), a straightforward question to a respondent asking whether these traits are found in their families is likely to elicit social desirability responses. Therefore, I opted for oblique, reverse-coded questions such as “At times in my family I wish we felt more love for one another.” For each of 13 main elements of the cultural domain, statements of this type were generated, answered on a 4-point Likert-type response scale of agree–disagree. Then, each item was weighted by the relative importance of that element in the cultural consensus analysis. The resulting scale had an internal consistency reliability of α = .89 (Dressler et al. 2005).
Cultural consonance has generally been studied in relation to health outcomes, the hypothesis being that low cultural consonance is a chronically stressful experience. Psychological distress is associated with lower cultural consonance in: religion (Dengah 2014; Read-Wahidi and DeCaro 2017), lifestyle (Maltseva 2018; Dressler et al. 2007a; Reyes-García et al. 2010a; success in online gaming (Snodgrass et al. 2011, 2013, 2014), family life (Balieiro et al. 2011; Dressler et al. 2007b), and managing HIV seropositivity (Copeland 2018b). Higher blood pressure is associated with lower cultural consonance in lifestyle and social support (Dressler and Bindon 2000; Dressler et al. 1997, 1998, 1999, 2005; Sweet 2010). Immunocompetency is associated with lower cultural consonance in HIV management (Copeland 2018a, 2018b) and social support (Dressler et al. 2016). Body mass is associated with cultural consonance in lifestyle (Dressler et al. 2008, 2012; Reyes-García et al. 2010b). Finally, Dressler et al. (2007a, 2017, 2019) found that cultural consonance in the domains of lifestyle, social support, family life, national identity, and education combine to form a construct of cultural consonance in life goals, which is associated with less depression and psychological distress.
Cultural consonance has generally been measured in relation to overall cultural consensus, but Andrews (2018, 2019) recently examined cultural consonance with subgroup models from residual agreement. Among Mexican immigrant women in the southern United States, she found residual agreement in the cultural domain of la buena vida (“the good life”). There was an overall cultural consensus regarding the items potentially defining a good life, which included both material goods associated with a middle-class lifestyle and somewhat more abstract goals, such as being a good person, being a productive member of the community, and providing an education for one’s children. With residual agreement analysis, she found that one group of respondents emphasized material goals in migration, while the other emphasized more abstract goals. She then calculated two measures of cultural consonance, one with each residual agreement model. The two measures of cultural consonance were highly correlated (r = .95), and overall cultural consonance was associated with lower depression and lower diabetes risk.
Discussion
The cultural consensus model and its extensions to residual agreement and cultural consonance have enabled social scientists to examine empirically some fundamental assumptions regarding culture and to incorporate it as a variable in multivariate research designs. The associations of cultural competence, residual agreement, and cultural consonance with a variety of outcome variables has proven to be robust (Dressler 2018). Furthermore, it has been suggested that the intersection of competence, residual agreement, and consonance provided the structure of the cultural niche in human evolution (Dressler 2019).
There are numerous directions for future research employing this approach. One important question involves the intersection of cultural models of differing social distribution. There are cultural models (e.g., family life) of very wide distribution in society. At the same time, each of us is influenced by cultural models of much more narrow distribution, such as those based on religion and ethnicity (among other factors). How the dimensions of culture discussed here intersect across different cultural domains in relation to outcomes is an important area of research.
There have also been recent innovations introduced in the basic cultural consensus model, specifically in relation to better understanding intracultural variation (Anders et al. 2018; Lacy et al. 2018). These newer models have yet to be tested in ethnographic research, and determining their utility is an important next step.
The introduction of the cultural consensus model was truly paradigmatic in anthropology, and the study of cultural consensus and cultural consonance will continue to contribute important new insights in social science.
Footnotes
Acknowledgments
Kathryn S. Oths and three anonymous reviewers offered helpful comments on previous drafts 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) received no financial support for the research, authorship, and/or publication of this article.