Cultural Perceptions of Weight in African American and Caucasian Women

Abstract

Objective

To determine if African American (AA) and Caucasian women grouped variables related to race and weight into discrete clusters and if there were discernable response patterns with unique subgroup characteristics.

Methods

Women (N=277, 48% AA) completed a card sorting task, ranking 28 variables. We used multidimensional scaling to determine perceived similarities and differences between variables, and latent class analysis to identify subgroups responding similarly.

Results

We identified 5 clusters of variables and 4 response patterns, which were demographically and anthropometrically distinct.

Conclusions

These results can be used for empirical cultural tailoring of behavioral weight loss interventions.

Despite ongoing efforts, the burden of obesity continues to disproportionately impact African American women.¹ This disparity exists independent of income, education, and other social factors that may be typically associated with obesity in white women.² A common construct used to explain the difference in weight status for African American women is culture.³ Culture has been defined as a group of shared beliefs, meanings, and perceptions based on a common experience. ⁴ However, by nature of individual environments and experiences, immersion in a particular culture and the extent to which common beliefs, meanings, and perceptions are shared are variable. This variability leads to heterogeneity within any given cultural group.

Although heterogeneity within African American culture is well recognized, the idea of culturally tailoring an intervention for weight reduction has generally been approached with broad strokes and generalizations. Investigators have had difficulty in consistently moving beyond “surface structure” elements as described by Resnicow to find broadly applicable “deep structure” intervention targets that have true salience for a wide audience.⁴ As a result, the outcomes typically achieved in culturally tailored interventions have not been any better than those achieved with nontailored approaches.⁵

A key challenge for culturally tailoring weight-loss interventions with the goal of eliminating disparities in obesity is the development of a systematic approach to understanding the heterogeneity within this group. Application of culturally tailored strategies may have sufficient rationale to employ, but to strengthen the scientific approach and improve outcomes, we need to effectively define subgroups of individuals within the larger population and associated, specific behavioral targets to an extent that provides confidence that the intervention will be more salient for members of the target audience. The approach also needs to be sensitive to the influence that social and economic factors can have on cultural expression.

In 2007, we described the results of a study that employed the nominal group technique (NGT) comprising African American and white women who were asked how their race affected their weight.⁶ Overall, each group had a unique perspective on how their race affected their weight: African American women had more permissive influences for weight gain whereas white women had influences that opposed weight gain. Although many ideas generated by the groups appeared to be broadly applicable regardless of race, such as easy access to junk food/fast food and lack of exercise, the 2 racially self-identified groups of women provided completely distinct groups of responses based on their perceptions of race and weight.

The objective of our current research is to use an empirical approach to help define putative or supposed cultural variables that can be subsequently mapped to specific behaviors and used as the basis of culturally tailored intervention strategies. More specifically, we wanted to identify how independent samples of African American and white women perceived and cognitively linked the previously derived cultural variables that were associated with perceptions of race and body weight. In this manuscript, we report the results of a card-sorting task completed by 277 African American and white women, along with the resulting empirical representation of how these participants organized the cultural arrays. Using latent class analysis to identify unique subgroup memberships based on the card-sorting task, we hypothesized that the women would organize cultural variables into distinct clusters and that there would be discernible response patterns and associated group characteristics amongst the study population.

METHODS

Participants

Study participants were recruited from the Birmingham, Alabama, metropolitan area by using advertisements in local publications, posting bulletins on the University of Alabama at Birmingham (UAB) campus, and collaborating with other researchers on the UAB campus. To be eligible to participate, women aged 19 and older needed to identify themselves as either African American or white. All eligible participants were informed that the card-sorting activity would last approximately 20 minutes and they would receive $15 compensation for their time. This study was approved by the University of Alabama at Birmingham Institutional Review Board. Each participant provided informed consent.

Measures

Upon completion of the informed consent, the participant completed a demographic questionnaire to gather information on age, race, educational level, marital status, and employment. Physical measurements (ie, weight, height, waist circumference) were collected after completion of the card-sorting task described below. Body weight was measured in light clothing without shoes by a digital scale. Height was measured using a wall-mounted stadiometer. Body mass index (BMI) was calculated as weight (kg)/height (m²).

Procedure

The card-sorting activities were conducted one-on-one (ie, one participant with one research assistant of the same race). The research assistant informed the individual participant that the research team had collected a list of 28 statements from groups of women on factors that affected them or their weight.⁶ Examples of the statements included “Not getting enough exercise” or “Believing that being thin is the ideal body shape.” A single statement was typed on each of the 28 3″ × 5″ cards. The research assistant asked each participant to read each card to herself and then to sort the cards into different piles based on her perception of how each variable/statement influenced her weight. All research assistants were trained using a standard protocol, and the principal investigator was responsible for ensuring consistency across interviewers and maintaining the training of the research assistants.

Data Analysis

A cognitive mapping approach involving multidimensional scaling and hierarchical cluster analysis was used to derive an empirical representation of the cognitive structures that participants used to organize the array of culture variables.^7,8 We chose cognitive mapping as a technique to organize the card-sorting information because it would provide a clear empirical understanding of how African American and white women might cognitively construe the various cultural concepts in the context of managing their body weight. A latent class analysis (LCA) was then used to identify subgroups of participants who shared a common profile based on how they perceived the cultural variables with respect to their weight (Latent GOLD software, version 4.5).

Our first objective was to determine how the participants perceived the differences and similarities among the cultural factors affecting them and their weight. We modeled the card-sorting data using a nonmetric multidimensional scaling (MDS) program based on squared Euclidean distances (SPSS software, version 19).⁷ The MDS results represent the participants’ perceptions of the relative similarity of factors that are a set of interelement distances plotted in a derived multidimensional space. The axes defining the multidimensional space empirically are assumed to represent the decisional criteria used in the sorting process.

Our second objective was to use a hierarchical cluster analysis to model heterogeneity among African American and white participants with respect to how they ranked the similarities and differences of the 28 card sort factors. We used the card sort rankings as the basis for deriving the class membership. The parameters of the model, including the number of classes, the proportion of the sample within each class, and the probability of endorsing each of the cultural variables, were estimated using maximum likelihood estimation with the expectation maximization algorithm. The model parameters indicate the proportion of the total sample included within each class and the response probabilities for each indicator given class membership.

Our final objective was to understand how membership in a specific class was associated with socio-demographic variables and body weight. We further characterized the resulting subgroups identified through the LCA approach by comparing the individuals with respect to socio-demographic factors, including age, race, marital status, employment, education, income, self-reported health status, and body weight/weight status (BMI). LCA is a model-based clustering procedure used to identify mutually exclusive categorical latent (unobserved) classes of cases within a population based on a pattern of responses to a set of variables used for the purposes of segmentation.⁸ We determined the adequacy of different models based on several goodness-of-fit measures including log-likelihood ratio of χ² (LR), Akaike’s information criteria (AIC), and Bayesian information criterion (BIC). We examined the pattern of conditional probabilities to determine the variables most important in constituting classes. We then statistically evaluated the ability of these indicators to discriminate among the identified classes using Wald statistics.

RESULTS

Cognitive Mapping

We derived 5 clusters from the 28 cultural variables. The clusters are shown in Table 1 and graphically in Figure 1. Based on a 2-dimensional solution, the research team interpreted the horizontal dimension as representing cultural specificity. The variables along this dimension can be viewed as ranging from culturally nonspecific to culturally specific. By culturally nonspecific, we imply that the given theme (ie, the nomenclature used to describe the cluster of variables) might be perceived as generally relevant to body weight by both African American and white women. However, a culturally nonspecific theme does not mean that the influence is of the same magnitude and/or direction (ie, positive/negative) for both groups. On the other hand, culturally specific themes can be viewed as generally being more or less relevant for one group compared to the other. The vertical axis was interpreted as representing determinants of behavior with variables on this axis ranging from externally uncontrolled influences to internally motivated individual choice. By externally uncontrolled influences we mean that participants perceive limited choice or control over the circumstance(s) that may influence their weight. Comparatively, internally motivated individual choice means that the participant perceives some, although not absolute, level of control over the situation that may ultimately have an impact on her weight. The horizontal axis was anchored on the culturally nonspecific border by a cluster of thematically similar variables including not getting enough exercise, eating too much junk food, and having poor food selection habits. On the culturally specific end of this dimension, there were variables including not exercising because of hair care issues, believing that men like women with curves, and believing that men prefer thin women. The vertical dimension (determinants of behavior) was anchored at one extreme (uncontrolled influence) by variables such as genes affecting weight or stress eating and at the other extreme (internally motivated choice) by variables including not exercising because of hair care issues, limited food options due to money or education, and poor food selection.

Table 1.

Descriptive Statistics for Clusters and Variables Within Clusters (N=277)

Clusters and Variables Within Clusters	Segment				Total
	1	2	3	4
Cluster 1: (Lifestyle behaviors)	5.61	5.37	6.08	4.24	5.46
13. Not getting enough exercise	6.01	6.13	6.18	4.67	5.88
10. Eating fast foods/junk foods	5.62	5.48	6.41	4.35	5.60
12. Making poor food selections	5.79	5.36	6.14	4.07	5.50
11. Preparing unhealthy food (high-fat, high-salt, fried foods)	5.14	4.42	5.63	3.67	4.84
Cluster 2: (Perceived behavioral controls)	4.22	5.16	4.53	4.60	4.63
16. Overeating when stressed	4.75	5.47	5.05	4.09	4.93
4. Not liking the way my body looks	4.07	5.34	4.29	5.09	4.64
27. Having self-motivation to be healthy	4.09	5.03	4.41	4.93	4.57
23. I think that my genes	3.91	4.96	4.34	4.28	4.38
Cluster 3: (Interpersonal transitional influences)	4.19	4.42	4.79	3.59	4.32
14. Being concerned about the diseases related to being overweight and associated medical costs	4.38	4.81	4.76	3.86	4.52
15. Having multiple roles gets in the way of making health choices	3.98	4.68	4.66	3.63	4.30
24. My food traditions	4.21	4.21	4.89	3.42	4.27
22. Learning bad food habits from my family	3.98	4.09	4.91	3.51	4.19
Cluster 4: (Contextual decisions)	3.87	2.90	3.78	3.28	3.48
28. Taught not to be wasteful with your food	4.02	3.60	4.04	3.19	3.78
7. Always talking about and/or trying new diets	4.02	3.09	4.00	3.65	3.70
9. Having diet programs targeted at me	3.88	3.27	3.88	3.44	3.64
21. Not understanding the importance of maintaining a healthy body weight	4.16	2.75	3.88	3.49	3.59
20. Not having enough money to buy healthier foods	3.72	2.91	3.53	3.21	3.36
17. Not having enough education on how to eat healthy	3.68	2.47	3.86	3.21	3.32
26. Avoiding exercise so that I won’t need to redo my hair	3.63	2.12	3.24	2.67	2.95
Cluster 5: (Social pressure)	3.20	3.54	2.66	4.37	3.33
6. Feeling pressure to be thin because being heavy is not acceptable	3.58	4.01	2.72	4.95	3.68
3. Believing that being thin is the ideal body shape	3.42	3.84	2.89	4.72	3.60
1. Expecting to have a perfect body	3.32	3.61	2.78	4.63	3.45
18. Having low self-esteem	3.10	3.78	2.59	4.77	3.41
19. Having limited clothing choices	3.27	3.42	2.91	3.84	3.30
2. Believing that being successful depends on being thin and looking beautiful	3.15	3.60	2.53	4.40	3.30
5. Believing that men prefer thin women	3.00	3.57	2.21	4.51	3.18
25. Believing that men like women with curves	3.09	3.08	2.54	3.88	3.06
8. Believing that my family expects me to be thin	3.04	2.92	2.70	3.91	3.05

Figure 1.

Results of Cluster Analysis

When considering the overlap of the 2 dimensions, themes in quadrant I could be considered culturally nonspecific choices. Themes in quadrant II could be considered culturally specific choices. Quadrant III includes themes that could be characterized as culturally nonspecific influences, and quadrant IV includes culturally specific influences.

Cultural Variable Clustering

The 5 clusters of cultural variables were reviewed by the research team with regard to the specific content of each cluster, detailed components that were a part of the derivation of the theme, and its relative position on the 2-dimensional map. Consensus was reached among the team on representative cluster descriptors. Cluster 1 included variables directly related to lifestyle behaviors, such as food choices and exercise habits. This cluster was primarily situated in quadrant I of the cognitive map, representing culturally nonspecific choices. These variables are likely to impact African Americans and whites universally and are at the level of individual choices directly proximal to the behavior or outcome (eg, African American or white women make poor food choices, and those choices affect their weight).

Cluster 2 included variables summarized as perceived behavioral controls affecting choices or beliefs/attitudes about weight. Variables in this cluster included belief that genes affected weight, stress leading to eating, and self-motivation to be healthy. This cluster was primarily in quadrant III, representing culturally nonspecific influences. This mapping suggests that African American and white women could be equally influenced by these factors, although the direction of influence could vary by race.

Cluster 3 represented a series of variables dealing with interpersonal transitional decisions. The variables in this cluster appeared to represent women at various stages of life and the life-stage dependent factors that influence weight and weight-related behaviors. The factors described were typically in the context of relationships with those within the immediate social network (eg, family, friends, co-workers). Concern about medical cost and associated diseases was prospective in nature, whereas learned food habits and traditions were retrospective considerations. Current roles being played by women were more contemporaneous. This clustering was near the center of both dimensions, having a higher orientation toward individual choices and low cultural specificity.

Cluster 4, contextual decisions, was contained in quadrant II, representing a spectrum of culturally specific choices. As such, we observed less overlap of the variables in this cluster between African Americans and whites. Cluster 4 was anchored in quadrant II by “avoiding exercise so that I won’t need to redo my hair” and ranged to “always talking about or trying new diets.” These variables represented a range of individual-level choices that had a predominant applicability to one ethnic group compared to another.

Cluster 5 was identified as social pressure. Examples of variables found in cluster 5 include “feeling pressure to be thin because being heavy is not acceptable” and “believing that my family expects me to be thin.” It was primarily located in quadrant IV and contained themes associated with culturally specific influences on weight and weight-related behaviors that originate from the larger, self-identified societal context.

Latent Class Analysis- Socio-demographics and Responses to Card-sorting Task

The results of the LCA are shown in Figure 2. A 4-segment solution demonstrated the best fit to the data (BIC(LL)=7556.51, AIC(LL)=7154.25). For descriptive purposes and a common frame of reference, we characterized participants within each segment in terms of their responses to an obesogenic environment. The research team used descriptors from the clusters that were typically most distinctive in characterizing members of the segment as well as relationships between clusters that provided suggestions regarding the context for the pattern of responses. As a result of this process, the 4 segments were named: (1) Lifestyle Challenged by Contextual Decisions, (2) Managing Lifestyle With Challenges From Perceived Behavioral Controls, (3) Lifestyle Challenged by Interpersonal Transitions, and (4) Accepting Social Standards With Influence From Perceived Behavioral Controls.

Figure 2.

Results of Latent Cluster Analysis

Participants in Segment 1, identified as Lifestyle Challenged by Contextual Decisions, had the highest mean age and composed a significant proportion of African Americans (72%; Table 2). A majority of participants in this segment were single and working full time and tended to have an annual household income of less than $30,000. Most were overweight or obese, with less than 20% being considered normal weight. Within this segment, participants scored highest on the Lifestyle Behaviors and lowest on the Social Pressures factors. They also shared the highest cluster score among the 4 segments for the theme of Contextual Decisions. In consideration of the socio-demographic composition of this segment and their collective profile of responses across cultural variable clusters, women in this segment, who appear to be highly influenced by their contextual environment (eg, built environment), are likely to gain weight in response to the obesogenic environment. This segment may be further inclined to weight gain as a result of lower levels of resources available to overcome barriers to engaging in healthy lifestyle behaviors, which they recognize as being highly important to them in impacting their weight.

Table 2.

Characteristics of Participants (N=277)

		Segment				Total
		1	2	3	4
Segment Size, %		29.18	28.49	26.72	15.61	100.00
Segment Size, n		81	77	76	43	277
Clusters
Lifestyle behaviors		5.61	5.37	6.08	4.24	5.46
Perceived behavioral controls		4.22	5.16	4.53a	4.60a	4.63
Interpersonal transitional decisions		4.19	4.42	4.79	3.59	4.32
Contextual decisions		3.87a	2.90	3.78a	3.28	3.48
Social pressures		3.20	3.54	2.66	4.37	3.33
Demographic Characteristics
Age (years)		38.2±9.91a	34.85±9.45b	38.14±9.09a	35.30±10.17ab	36.76±9.71
Education, %	High School	17.52	7.69	13.24	13.81	13.00
	Some college	47.71	17.57	44.60	29.01	35.38
	College	14.59	25.17	24.55	18.89	20.94
	Graduate	20.17	49.57	17.61	38.29	30.69
Ethnic, %	White	28.14	88.51	37.05	55.49	51.99
	African American	71.86	11.49	62.95	44.51	48.01
Marital Status*, %	Unmarried	59.87	45.05	50.78	54.29	52.35
	Married	40.13	54.95	49.22	45.71	47.65
Employee Status, %	Not full-time employee	12.05	30.39	13.40	19.34	18.91
	Full-time employee	86.69	69.61	85.28	80.66	81.09
Annual Family Income, %	Less than $30,000	48.03	18.78	36.34	40.39	36.16
	$ 30,000 – $60,000	34.16	39.80	33.60	37.25	36.90
	More than $60,000	15.40	40.82	27.35	18.71	26.94
General Health*, %	Excellent	16.40	26.31	13.56	20.71	19.13
	Very good	35.69	37.19	36.88	28.89	35.38
	Good	40.68	32.87	39.72	38.82	37.91
	Fair	7.23	3.63	9.84	11.59	7.58
BMI (kg/m2), %	Underweight (<18.5)	1.60	0.85	0.00	0.09	0.74
	Normal (18.5–24.99)	18.60	42.53	14.46	43.28	28.68
	Overweight (25–29.99)	37.85	31.05	38.59	30.83	35.66
	Obesity (>30)	38.58	24.57	45.10	25.52	34.93

Note.

Means that do not share subscripts differ at p < .05 in the Fisher LSD significant difference comparison.

^*The proportion is significantly different across 4 segments (p < .05).

Participants in Segment 2, Managing Lifestyle with Challenges from Perceived Behavioral Controls, were the youngest on average, had the highest proportion with graduate-level education, and were predominantly white (88%; Table 2). This segment had the highest proportion of individuals with >$60,000 annual household incomes but also had the highest proportion that was unemployed. Forty-three percent were normal weight, and approximately 63% rated their health as very good or excellent. Within this segment, women scored highest on the Lifestyle Behaviors cluster, similar to women in the Lifestyle Behaviors Challenged by Contextual Decisions segment. However, they were the lowest of any segment on the Contextual Decisions cluster. Relative to other segments, individuals in this segment ranked the Perceived Behavioral Controls variable cluster the highest value. In summarizing this segment, it appears that they recognize the importance of key lifestyle behaviors in determining body weight (“Not understanding the importance of maintaining a healthy body weight” was ranked lowest within this segment), but they also subscribe to external controls that influence their behaviors or outcomes (eg, Stress eating had the highest average ratings within this segment). They appear to have higher education, income, and other resources that might be used to overcome the barriers that they perceive as being a part of the obesogenic environment. The confluence of these attitudes, beliefs, and resources may be a key factor in their having the lowest observed body weight and highly rated self-reported health.

Participants in Segment 3, Lifestyle Challenged by Interpersonal Transitions, were comparable to those in the Lifestyle Challenged by Contextual Decisions segment (Segment 1) in many respects. A majority of participants in this segment were African American (63%), with some college education or less, and largely employed (Table 2). A larger proportion was obese, with 45% having a BMI ≥30 kg/m2. They had the highest score of all 4 segments on Lifestyle Behaviors; they also had the lowest score on Social Pressures. However, in contrast to the segment Challenged by Contextual Decisions, they were highest among the 4 segments on the Interpersonal Transitional Decisions. Although from a socio-demographic perspective, we might expect a similar response to the obesogenic environment for this segment of women as those in the Segment 1, the pattern of responses suggests that the pathways by which the high prevalence of obesity is reached is different from those Challenged by Contextual Decisions. The women in the Lifestyle Challenged by Interpersonal Transitions segment appear to place a high value on their roles in social circles (work, family, friends, community), and these roles may have a significant influence on their ability to implement healthy lifestyle behaviors. Furthermore, this segment perceives very little social pressure to achieve thinness (lowest scores on believing that men prefer thin women and feeling pressure to be thin because being heavy is not acceptable), which may further promote the effects of an obesogenic environment.

Participants in Segment 4, Accepting Social Standards With Influence From Perceived Behavioral Controls, had more of an equal distribution of white and African American women compared to other segments (Table 2). They were generally employed, and a majority were unmarried. The highest proportion of Segment 4 incomes was in the less than $30,000 category. The Accepting Social Standards segment also had the highest proportion of participants in the “Fair” health category. However, similar to participants in the Managing Lifestyle With Challenges From Perceived Behavioral Controls segment (Segment 2), a relatively higher proportion was normal weight (43%). Within this segment, the cluster of Perceived Behavioral Controls was most relevant; and across segments, this segment had the highest score on the Social Pressures cluster. Comparable to participants in Segment 2, Contextual Decisions was the least relevant area for influencing body weight and weight-related behaviors. In summarizing this segment, it appears that societal norms for body weight, body image, and beauty have a particular salience for these individuals. However, they do not have the resources of the individuals in the Managing Lifestyle With Challenges From Perceived Behavioral Controls segment to perform the healthy lifestyle behaviors or feel sufficiently educated on weight and associated topics; but perhaps because of the societal pressure they perceive, and a poorer body image perception, engaging in healthy lifestyle behaviors is not as much of a challenge for those in the Accepting Social Standards segment.

DISCUSSION

In this study, we presented 277 community-dwelling African American and white women with empirically derived ideas about how race is related to weight and weight-related behaviors. In response to the card-sorting task, the participants’ responses were organized into distinct clusters, representing 5 thematic areas. These clusters included Lifestyle Behaviors, Perceived Behavioral Controls, Interpersonal Transitional Influences, Contextual Decisions, and Social Pressure. When graphically illustrated, these clusters create a multidimensional map framed by behavioral determinants along one axis and cultural specificity along the other. We observed 4 unique response patterns that characterized subgroups of the 277 women. The patterns provide unique insights into the complex nature of the interactions between socio-demographic factors that are associated with weight and the cultural beliefs and attitudes that arbitrate decisions proximal to weight-related behaviors.

This work has attempted to quantify what has heretofore been a largely qualitative effort: understanding the impact of culture on weight and weight-related behaviors in African American and white women. Cultural differences have been cited as playing a major role in the obesity disparity, but the extent and exact nature of culture’s role have been unclear. It has been difficult to identify the heterogeneity within a self-identified racial group; it has been challenging to understand how income or education affects/interacts with cultural perceptions; it has also been tacitly assumed that beliefs of one racial group are primarily limited to the members of that group. We believe this work has taken a significant step forward in providing some empirically derived rationale for observed differences in behaviors and outcomes, providing nuanced understanding of this complex issue.

With the current methodology, we are trying to formulate a framework, based upon which cultural tailoring may be more effective. For example, the PEN-3 model is a framework that addresses culture in public health interventions and includes the 3 domains of relationships and expectations, cultural empowerment, and cultural identity.⁹ Currently the application of models that might guide cultural tailoring approaches such as the PEN-3 model do not provide guidance regarding audience segmentation, thereby leaving assumptions to be made regarding the heterogeneity of the audience and ultimate salience of the intervention at the individual level. The assumptions regarding the target audience extend beyond cultural beliefs and attitudes to include broader socioeconomic factors that can impact aspects of intervention delivery and implementation. Audience segmentation can be achieved using measures of acculturation or ethnic identity.^10–12 The limitation of such measures is that the segmentation happens exclusively within one ethnic group, limiting the generalizability of the resulting intervention(s). With this current study, we have been able to complete a more inclusive segmentation that includes white and African American women, demonstrating the shared experiences and perceptions while creating 4 unique segments of the population. Moving forward from this point, we would hypothesize that application of a given model for cultural tailoring such as PEN-3 would provide a higher level of salience at the individual level and have a greater generalizability beyond one self-identified ethnic group or social class.

We consider today’s society to be an obesogenic environment with constant pressures that promote excess caloric consumption combined with numerous opportunities to limit energy expenditure, ultimately leading to a positive energy balance.¹³ A number of factors likely determine our responses to the obesogenic environment, including level of exposure to obesogenic influences, available resources, along with our biologic predisposition to weight gain. Despite the individual variation in these factors, certain patterns of obesity prevalence within specific groups have developed over time, with a higher prevalence being noted in African American women and lower-income white women.¹ Reasons for a seemingly higher susceptibility of African American women to the obesogenic environment have often been attributed to factors such as acceptance of larger body size, lower income or education levels, and lack of access to health-promoting environments;¹⁴ explanations for the relationship between lower socioeconomic status and obesity in white women include fewer opportunities for upward mobility due to discrimination (ie, obesity leads to low socioeconomic status) or fewer resources to engage in weight gain prevention or weight reduction activities (ie, low socioeconomic status leads to obesity).¹⁵ In either case, it has been difficult to substantiate these pathways to obesity due to the complex behaviors and biology involved.

This study uses an empirical approach to suggest plausible hypotheses regarding how the divergence in behavioral response to the obesogenic environment may begin. Women in Segment 1, identified as Lifestyle Challenged by Contextual Decisions, are faced with the challenge of trying to maintain a healthy lifestyle while managing on a lower level of resources. They perceive that the resources available to them are not adequate to some extent to support a healthy lifestyle AND maintain their current sociocultural identity. As a result, they may make choices based largely on the most dominant context. For example, an African American woman with a lower income may be on a fixed budget that includes a limited amount for hair care. If exercise is going to lead to additional expenditures for hair care (above and beyond what may be required to participate in the exercise activity), this may be a limiting factor. In this instance, financial considerations combine with cultural context to create a barrier to a healthy lifestyle (“Not getting enough exercise” was the highest individual item within the Lifestyle Behaviors segment for cluster 1). The urgency of achieving a lower body weight is also limited, as this segment had lower scores for the theme of Social Pressure. The lack of influence from society does not lead them to reorder their priorities to achieve thinness, per se. Although this segment was largely comprised African American women, only 43% of the African American women in the sample were captured by this characterization.

The other segment that captured a significant percentage of African American women (35.8%) was Segment 3, Lifestyle Challenged by Interpersonal Transitions. This segment, however, had more racial balance and points to many of the commonalities shared between African American and white women in responding to the obesogenic environment. Similar to those in the segment Lifestyle Challenged by Contextual Decisions, the theme of Lifestyle Behaviors was very salient for this segment; however, past and present relationships or experiences may have a significant impact on their ability to engage in healthy lifestyle behaviors. The women who are challenged by the interpersonal transitions work actively to resolve perceptions of how others see them with perceptions of themselves. Based on their responses, they feel the least amount of pressure to achieve some standard of beauty. On the other hand, they feel that being a mother or wife means that food must be prepared a certain way, consistent with family traditions. In addition, although this segment had higher levels of income and education, simply having more income and education in this context may not necessarily provide members of this segment the level of resources required to actively engage in a healthy lifestyle. For example, an educated single mother may have enough income to pay for a gym membership, but not enough social or family support to get help with child care so that she can exercise consistently.

Segment 2, Managing Lifestyle Behaviors With Challenges From Perceived Behavioral Controls, indicated as did segments 1 and 3 that the theme of Lifestyle Behaviors was very relevant to weight for them; however, this segment appeared to be influenced by perceptions of factors that contributed to behaviors (stress leads to eating) or outcomes (genes). Because this segment was more educated and had the highest levels of income, they seem to be less affected by the income/education-related contextual decisions that pose barriers for Segment 1. The large majority of this segment was white; despite our efforts to understand the relationship of culture and socio-demographic factors, it is still challenging to disentangle class and culture in the United States at the present time. In this situation, it is difficult to know if the social setting comes first and, as a result, influences beliefs and attitudes related to healthy eating and physical activity. Alternatively, might a woman who believes that she is genetically challenged to maintain a normal weight make different choices about her use of resources to manage her weight, thereby putting more of a priority on engaging in a healthy lifestyle? Additional studies may further elucidate answers to these types of questions.

More than any other segment, Segment 4 (Accepting Social Standards With Influence From Perceived Behavioral Controls) subscribed to cultural norms related to body image and standards of beauty. Food and exercise issues were much less salient for this cluster. The racial diversity of this segment highlights the limitations faced by simple methods of categorization and segmentation. For example, over 14% of the African American sample fell in this segment.

The findings of this study are limited in several key ways. The associations between the segments, their response patterns, and body weight cannot be considered causal. Although plausible pathways by which a variety of social factors combined with sociocultural beliefs and attitudes lead to a given body weight can be hypothesized from these results, additional study is needed to further quantify the impact of these factors on weight over time. Additionally, due to the cross-sectional nature of this study, we cannot state that these segments have stable constituency over time. Individual resources, perceptions, beliefs, and attitudes may change over time; and as a result, the response pattern of individuals to the factors may change. The findings from this sample are also limited to this particular audience; thus, the generalizability of these results will be limited. However, the reproducible methodology provides an opportunity to determine the relevance of these clusters and segments to other groups of women who may be less educated, live in rural areas, or be of a different age range.

In conclusion, this study used a card-sorting task to identify 4 distinct segments from a population of African American and white women who had unique patterns of response to culturally related ideas about body weight and weight-related behaviors. This is the first attempt to our knowledge to characterize cultural perceptions related to weight in a biracial sample using an empirical approach such as multidimensional scaling and latent class analysis. To think of these issues as being exclusively cultural or socioeconomic is likely an oversimplification of a complex issue. As these results suggest, most decisions for women to engage in a behavior related to weight are likely influenced by both components to varying degrees. Additional study will be necessary to determine how this process can be used to promote similar levels of behavior change amongst African American and white women engaged in behavioral interventions for weight control.