Beliefs matter: Cultural beliefs and the use of cervical cancer-screening tests

In this article we examine the influence of cultural beliefs on behavior or, more specifically, beliefs about cervical cancer risk factors and the use of Pap exams. Individual Latinas' (Hispanic women) holding of beliefs similar to Latinas' generally (cultural consonance) did not significantly influence their use of Pap exams. Rather, structural factors such as medical insurance, age, marital status, education, and language acculturation explained Latinas' use of this medical service. However, when Latinas held beliefs similar to those of Anglo women, then they were significantly more likely to have had a Pap exam within the past two years. Latinas whose beliefs were closer to those of physicians were significantly less likely to have had the exam recently. Arriving at these findings involved both ethnographic interviews and survey research. That these beliefs proved to be significant influences on behavior suggests not only the important ways that beliefs matter but that ethnographic methods for examining those beliefs also matter. [Latinas and cervical cancer, Pap exams, culture and behavior, ethnography and survey research]

t is taken for granted in anthropology, and cultural studies for that matter, that humans learn the meanings, maps, knowledge, and discourses more or less shared by a people and thus become "cultural subjects," "which equips them with cultural 'know-how' enabling them to function as culturally competent subjects" (Hall 1997:22). Rather than deny this basic observation, the objective here is to show that acquiring cultural know-how and functioning as culturally competent subjects are complicated by competing cultural models. The ostensible reason for raising this problem is that in a previous analysis we found that Latinas (or Hispanic women), especially immigrants, held beliefs about the relative importance of risk factors for cervical cancer that diverged much more from those of physicians than those held by Anglo women (Chavez et al. 1995). We were left wondering to what extent these patterns of beliefs were associated with behavior, specifically the use of Pap exams, a screening test for cervical cancer. In other words, to what degree do cultural beliefs matter in the use of medical services?
We found that beliefs matter, but the story is complex.
For Latina immigrants, arriving in the United States requires a negotiation of the cultural knowledge about disease and prevention they arrived with, the predominant popular cultural knowledge, and knowledge promoted by physicians. These competing and overlapping models also promote specific behaviors, such as the need to access preventive medical services. It is often assumed that acquiring knowledge similar to physicians' knowledge about health promotion and disease prevention will increase the well- being of any population. Our findings, however, suggest that Latinas whose cultural beliefs were closer to those of Anglo women were significantly more likely to have had a Pap exam within the past two years, while those whose beliefs were closer to those of physicians were significantly less likely to have had the exam recently. Given this observation, our objectives here are twofold: (1) to describe how we arrived at these findings, a process that involved both ethnographic interviews and survey research, and (2) to provide an explanation for why these findings "make sense." A number of issues of wider concern frame this research. The first concerns an overemphasis on the role of culture-the cultural explanation-for understanding politically sensitive issues such as the utilization of medical services. During the 1950s and 1960s, cultural beliefs (such as fatalism, lack of future orientation, ingrained cultural values) were used in a simple and naive way to explain behavior, an approach that has become criticized, and rightly so, as "blaming the victims" for their disadvantaged social status. We do not wish to reiterate the critiques here (see, for example, Rosaldo 1989). Suffice it to say that since the 1960s, the pendulum shifted to more structural explanations for the use of medical services, such as the availability of services, poverty, medical insurance, work contingencies, lack of child care, and other factors associated with the political economy of health care (Chavez 1986;Chavez and Torres 1994;Doyal 1995;Morsey 1996;Young and Garro 1982).' Language and other communication-related barriers were also engulfed in this structural framework.
We argue that although structural factors may account for much, even most, of the underuse of medical services by some less powerful members of a society, they do not account for all the variation in behavior that humans exhibit.2 Fortunately, statistical techniques allow for the relative impact of a number of variables on a dependent variable, such as the use of medical services. We recommend careful assessment of the relative contribution of cultural and structural factors on the use of medical services, which should also tell us something about the relationship between culture and behavior more generally.
Another issue is that anthropologists have focused more on the cultural models side of the culture-practice relationship than on the behavioral side. For example, tremendous strides have been made in systematizing the study of cultural models, primarily through consensus analysis (Boster 1986; Garro 1986Garro , 1988McMullin et al. 1996;Romney et al. 1986;Romney et al. 1987;Weller 1987;Weller andRomney 1984, 1988;Weller et al. 1993). The examination of the specific ways in which adhering to a cultural model influences actual behavior, however, has received less attention. An important exception has been Dressler's work on the influence of culture on blood pressure.
In a series of important publications (Dressler 1996(Dressler , 1999Dressler and Bindon 2000;Dressler et al. 1998), Dressler has developed the concept of cultural consonance.
Cultural consonance assesses the degree of correspondence between an individual's beliefs or behaviors and the cultural ideal (Dressler 1996:7). Each individual's correlation to the agreed-on cultural ideal is a measure of his or her cultural consonance. Individuals vary from high levels of cultural consonance to low levels. Importantly, Dressler next uses multiple regression analysis to examine the predictive strength of cultural consonance for blood pressure in relation to the predictive strength of other measures, such as occupation, education, and income.
A problem that arises when examining the influence of cultural beliefs on health-seeking behavior concerns sample size and sampling methods. Consensus analysis was designed for use by anthropologists and is superb for analyzing levels of agreement among relatively small samples. To tease out the influence of cultural beliefs on behavior, especially relative to other variables such as income, education, and medical insurance, typically requires large samples and the use of random sampling. In an effort to bridge the culture model-practice divide, the research presented here describes the logic and methods we  (Bernard 1988;Weller and Romney 1988). Although we have presented the methods and findings of the ethnographic interviews elsewhere (Chavez et al. 1995;Martinez et al. 1997;McMullin et al. 1996), a brief summary of the results is necessary to explain how we moved from the ethnographic findings on cultural beliefs to an examination of the influence of cultural beliefs on behavior.

Freelisting
Our approach combined open-ended questions with the systematic data collection techniques of "freelisting" and ranking (Weller and Romney 1988). Open-ended ques-tions and freelisting helped to elicit qualitative data from which we developed thematic categories of perceived risk factors. In the freelisting task, we asked respondents to list everything that could cause or increase the chance of getting cervical cancer and then to discuss the reasons for the listings. We found that this language did not presuppose an understanding of the academic concept of "risk" on the part of the lay informants and clearly implied "risk" to the physician informants. This method does not impose a set of beliefs on the interviewees. Also, by employing this style of questioning, we did not establish a set of risk factors a priori. The interviewees themselves provided the risk factors; they came out of the interviewees' own ways of thinking about cervical cancer. These items, therefore, might vary according to the groups studied.
To determine which beliefs were most common, we reviewed all listed risk factors for an initial set of respondents (approximately one-third of each of the five groups), determined the risk factors mentioned most frequently by each group, and then reviewed these factors to ascertain the number of separate risk factors listed. This last step was necessary because more than one group listed some of the same risk factors. It also allowed us to sift out the more idiosyncratic beliefs, such as those listed by only one person. This process yielded 24 different risk factors for cervical cancer.

Ranking
To ascertain the relative importance of the risk factors for each individual, we printed each risk factor on an index card and asked the respondents to rank order them from most important to least important for increasing a woman's chances of getting cervical cancer. For the initial group of respondents from whom we established the list of risk factors for ranking, we accomplished the ranking task during a second interview. For the remainder of the respondents, the interviewers included the ranking task during the first and only interview. This rank ordering procedure allowed us to determine the extent of agreement among the interviewees concerning the relative importance of the risk factors.4 We could also examine variations within individual groups as well as determine differences between groups in the study.

Ethnographic Findings
Based on the analysis of the ranked 24 risk factors for cervical cancer (  Chavez et al. 1995. of the risk factors that Latina immigrants gave high rankings to (abortions, lack of medical attention, physical trauma to the vaginal area and rough sex, birth control pills, lack of care after giving birth, sex during menstruation, and not following a proper diet after giving birth) physicians ranked lower or as unimportant factors. Conversely, some of risk factors that physicians believed were important (starting sexual relations at a young age, smoking cigarettes, sexual relations, having many children, and poverty) Latina immigrants ranked lower or as less important. Interestingly, both Latina immigrants and physicians ranked relatively high-although Latina immigrants ranked them consistently lower than physicians-other sex-based risk factors: many lovers/sexual partners, vaginal infections that are sexually transmitted, and a husband/ boyfriend who is infected and gives the woman a disease.
Physicians exhibited a high level of agreement among themselves. In the freelisting task, almost all the physicians (90 percent) mentioned exposure to sexually transmitted diseases as a risk factor for cervical cancer, citing human papilloma viruses (HPVs) specifically (Chavez et al. 1995). They based their perceptions on recent epidemiological research that indicates that certain "high-risk" types of HPV may cause cervical cancer while "low-risk" types do not (Cannistra and Niloff 1996). Most of the physicians (93 percent) also cited multiple sexual partners, and a majority (63 percent) mentioned sexual activity at an early age. As one physician noted: "Human papilloma virus is the big thing now. Multiple sexual partners, and that's a kind of generic coverall that just increases the risk by increasing your exposure to sexually transmitted diseases" (quoted in Martinez et al. 1997:346).
As a result of the seeming certainty of sexual risks, and the risk posed by sexually transmitted diseases, the pervasive assumption among physician interviewees was that cervical cancer itself is a sexually transmitted disease. As one physician put it, "The major risk is the number of sexual partners. Cervical cancer is a sexually transmitted disease" (quoted in Martinez et al. 1997:346). Physicians' rankings reflect this emphasis on sexual behaviors as risk factors for cervical cancer because of the perceived connection between sexual behavior and the transmission of HPVs, what they believed is the underlying causative agent for cervical cancer. Physicians gave their highest three rankings to vaginal infections, many lovers/sexual partners, and beginning sex at an early age, respectively.
Chicanas ranked highly a combination of factors, some of the same factors ranked by Mexican immigrants (lack of medical attention, vaginal infections that are sexually transmitted, husband/boyfriend who is infected) and some of the factors ranked highly by Anglo women (most notably, heredity or family history In addition, Latinas, especially immigrants, place a negative value on nonnormative behavior, which includes behavior that may be viewed as morally questionable (Martinez et al. 1997). By engaging in such behaviors, women (and husbands) increase the risk of acquiring an illness such as cancer, as indicated by this quote from a 33- year-old Mexican immigrant: When men have relations with other women and come and do it with their wives, they are going to cause them to have a disease. Men give their wives diseases, but they do not analyze what they do, and unfortunately in this country we are in there is more prostitution. There are women who do it for nothing more than to pay the rent, that's all. But now even when the man does not fool around, now also the woman goes out with men other than her husband, and they get infected, and then they have children.
This attention to what for them is nonnormative and morally questionable behavior serves as a cultural context for understanding Latina immigrants' rankings of cervical cancer risk factors (Martinez et al. 1997). They ranked high many lovers/sexual partners, vaginal infections that are sexually transmitted, birth control pills, husbands/boyfriends who are infected with disease, lack of care after giving birth, sex during menstruation, and not following a proper diet after giving birth.
Abortion is a risk factor that is morally questionable for these women and results in physical trauma to the body. For example, a 52-year-old Mexican woman commented: Maybe it [cervical cancer] is because they stop the baby from coming, and they yank it out using herbs, like women who do it with teas, and there remains like a sore or wound. Just think about how a germ can get in there. Because sometimes it's one's own fault to be practically rotting because of a stupidity like that. It's preferable to have baby and not yank it out because a sore remains there. Afterwards, if she makes love too soon and her partner is not clean in that area, a bunch of dirty junk is going into her wound. [quoted in Martinez et al. 1997:354] Her explanation of the possible risk of abortion for cervical cancer is quite complex, combining the physical trauma of "yanking" and creating a sore, and all because of, in her judgment, "stupidity." In addition, engaging in intercourse too soon (before the wound is healed) with a partner who is "not clean" can further aggravate the situation. She, like many of the women we interviewed, combined several themes in her understanding of the risk factors for cervical cancer. This multiple indemnity of abortions may account for its high ranking among Latina immigrants. A lack of medical attention, we found, reflected Latinas' belief that women should seek medical attention and, therefore, that not doing so increases the risk of illness, even cancer. But Latinas also discussed the obstacles to medical care, placing the blame not on the individual but on the structure of health care delivery and their lack of medical insurance, low incomes, need to work, child care problems, and language difficulties.
We also found little evidence for Latinas', especially Latina immigrants', knowledge about HPVs. Such knowledge, based on the epidemiological literature, is central to the biomedical model of cervical cancer risk factors. About one out of five Mexican immigrants, Chicanas, and Anglo women cited infections as a risk factor for cervical cancer, especially when such infections were associated with illicit sexual activity. HPVs were not mentioned among immigrants and rarely among U.S.-born Latinas and the Anglo women. In addition, the logic of the sexual transmission of HPVs is also behind the physicians' high ranking of the risk factor of beginning sexual relations at an early age.
Latinas did not rank this sex-related risk factor high, underscoring their emphasis on a logic of nonnormative and morally questionable behavior for understanding cervical cancer risk factors rather than knowledge about HPVs. For this reason, we argue that although Latina immigrants and physicians both ranked many lovers/many sexual partners and vaginal infections that are sexually transmitted relatively high, they did so for different underlying reasons. This is a case of concordance at one level, ranking, but discordance at the level of the underlying logic for the ranking. It suggests the need to carefully examine not just ranking outcomes but also the reasons and understandings for  1995). Another potential limitation of the study was that it would not find hard-to-reach members of the population, the homeless and those engaged in street comer employment and migrant agricultural labor. This may be more of a bias, however, for male rather than female Latinos, who are less likely to be homeless or seek day work by standing on street comers (Chavez 1992). Sample size goals varied by population. For Anglo women, we sought a minimum of 400 interviews and completed 422. We sought a larger subsample of Latina respondents in order to examine variation within the population. Therefore, we continued random dialing until we had at least 800 interviews (n = 803). Our survey randomly selected both households and respondents within households-the woman 18 years or older who had the most recent birthday. The cooperation rate was 78.5 percent (Hubbell et al. 1995).8 Latina respondents could choose to answer the questions in Spanish or English. We pilot tested the questionnaire, tested its content validity, and translated it from English, to Spanish, to English. The final questionnaire included inquiries about demographic characteristics and medical care access as well as knowledge, attitudes, and practices related to breast and cervical cancer. It also included a previously validated five-point "acculturation" scale (Marin et al. 1987) that measures acculturation primarily on the use of Spanish or English (e.g., read with, speak with, think with, used as child, and speak with friends).9 All the Anglo women were born in the United States.
In 1993, we also conducted a mail survey of physicians practicing in Orange County, resulting in 369 respondents.
We targeted physicians who specialized in family medicine, internal medicine, and obstetrics/gynecology. After our initial contact, we attempted to improve the response rate by mailing up to three follow-up letters to the physicians. We also included $5 with the initial letter. Our persistent efforts resulted in a 51 percent response rate.  (24) from El Salvador, the survey also included Latina immigrants from other countries. In short, the countywide survey has a more diverse sample of Latinas than that found among our ethnographic interviewees.
Physicians (Table 3) in the ethnographic interviews and broader survey were similar in terms of age, years practicing medicine, and ethnicity. The specialties of the physicians were also found in about the same proportions. There were proportionately fewer female physicians in the survey. Table 4 provides a summary of the respondents' sociodemographic characteristics. The major difference is between Latina immigrants and both U.S.-born Latinas and Anglo women. Latina immigrants, with a mean age of 33, were younger than both U.S.-born Latinas (mean age 37) and Anglo women (mean age 44). Latina immigrants had received fewer years of education (mean nine years) than U.S.-born Latinas (mean 13 years) and Anglo women (mean 15 years). Latina immigrants had been in Orange County, on average, three-fourths of the time they had  About two-thirds (66.9 percent) of Latinas had private or public medical insurance, but immigrants were much less likely to be insured than U.S.-born Latinas. Almost all Anglo women (91.9 percent) had access to medical insurance. Importantly, Latinas (70.9 percent) were less likely to have had a Pap exam in the two years previous to the interview than were Anglo women, most of whom (90 percent) indicated this high level of preventive care. Latina immigrants were much less likely than U.S. women to have had a Pap exam within this time frame. These patterns suggest that Latinas do, indeed, face obstacles to the use of Pap exams in Orange County that Anglo women do not.
This stage of the research required developing a set of survey questions that built on what we had learned about Latinas', Anglo women's, and physicians' cultural beliefs about cervical cancer risk factors. The problem was that the surveys' sampling methods (random telephone survey among women in Orange County and a survey mailed to physicians) placed constraints on the number of questions and the question format. The ranking tasks we used to systematically elicit responses from informants in face-to-face interviews could not be used in a large telephone-based survey of women or a questionnaire mailed out to physicians. Survey respondents would not be able to physically rank order 24 cards.
We decided to use paired comparisons of a smaller number of risk factors. The number of items respondents can reasonably be expected to compare, given the time constraints of phone interviews and the possible mental fatigue caused by a large number of such comparisons, is limited. We settled on six risk factors for cervical cancer: "abortions," "heredity or family history of cervical cancer," "starting sexual relations at an early age," "lack of medical attention," "many sexual partners," and "vaginal infections." We then paired each risk factor with each of the other risk factors, for a total of 15 paired comparisons.
Interviewees were asked to indicate which of the paired risk factors was, in their perception, more important for causing or increasing a woman's chances of developing cervical cancer. Latinas and Anglo women interviewed by phone were questioned orally, while physicians responded to written questions pairing each of the variables.
We selected these risk factors because each one was among the highest ranked risk factors for one of the differ ent groups of women and physicians in the ethnographi interviews. For example, Mexican and Salvadoran immigrants gave greater emphasis to "abortions" (ranking it and 1, respectively) than the other groups. This risk factor indicates the emphasis that Latina immigrants gave to nonnormative and morally questionable behaviors as healt risks as well as the notion that physical stress and trauma increase a woman's chances of getting cervical cancer.
Mexican immigrants and U.S.-bom Latinas (Chicanas gave greater emphasis to "lack of medical attention" tha the other groups (1 and 3, respectively).
Anglo women and Chicanas emphasized "heredity or family history" (ranking it 1 and 2, respectively) to a greater extent than physicians and Latino immigrants (Mexicans ranked it 15). This indicates the importance o heredity for explaining health and disease in popular U.S culture.
Physicians emphasized "vaginal infections," "many sexual partners," and "starting sexual relations at an early age" (ranking them 1, 2, and 3, respectively). This ordering reflects the physicians' adherence to an epidemiologica model that finds sexual behavior risk factors paramount for cervical cancer, primarily because of the spread of HPVs Vaginal infections pose a general risk that subsumes a range of possible infection agents, including sexually transmitted diseases (e.g., HPVs). We used the more gen eral "vaginal infections" to avoid having the word sex (a We hypothesized that if we selected risk factors that were salient among the different groups of ethnographic interviewees, then perhaps they would effectively characterize differences in perceptions about risk factors among the broader groups of randomly sampled respondents. It would be especially noteworthy if the 24 risk factors used in the ethnographic interviews could, after analysis, be reduced to only six items. However, by reducing 24 items to 6 items we have maximized the differences between the groups. For example, eliminating 18 risk factors has the potential of removing risk factors that may have been ranked similarly by individuals in the various groups. Moreover, each of the six items should, if successful, be  Table 5 shows the average ra various groups in our sample gave priority to the sex-base partners, sex at an early age, underlying logic of the physici the sexual transmission of HP sexually active, the more like sexual partners, which, in turn infection by HPVs. They gave followed by heredity, leaving  Vaginal infections 3 2 2 3 3 1 3 3 Many sexual partners 4 3 4 5 5 5 4 1 Heredity 5 5 5 1 1 4 1 5 Sex at an early age 6 6 6 6 6 6 6 2 Do Beliefs Matter? Do these differences in beliefs about cervical cancer risk factors matter? More specifically, does it matter if individual Latinas rank cervical cancer risk factors similarly (consonance) to the average or ideal ranking of those risk factors by Latinas generally? Does it matter if Latinas are closer to Anglo women's or to physicians' beliefs about the risk factors for cervical cancer? Latinas whose rankings are similar to the physicians' rankings would rank high the sex-based risk factors, whereas Latinas whose rankings are more similar to Anglo women's rankings would rank high heredity as a risk factor for cervical cancer. Does it make a difference to perceive, on the one hand, the chances of acquiring cervical cancer as increasing with one's own sexual behavior or to perceive, on the other hand, one's chances as increasing as a result of heredity, a factor over which one has virtually no control (and thus no personal blame)?
To examine these questions, we undertook logistic regression analyses. Table 6 lists the variables and their values used in the analyses. The dependent variable in the logistic regressions is the use of Pap exams. Latinas who had never had a Pap test or had not had one for more than two years before the interview are classified as low users. Latinas who had a Pap exam within the two years prior to the interview are classified as high users.
In model 1 (see Table 7), we examined individual Latinas' correlation (consonance) to an ideal, or average, Latina ranking of the six risk factors for cervical cancer. Individual Latinas ranged from -.89 to 1.0 in their correlations to an ideal Latina ranking of vaginal infections, abortions, lack of medical care, heredity, many sexual partners, and sex at an early age (in this descending order). We recoded this variable, and all the other interval-and ordinal-level variables, into a dichotomous variable, with the value "0" equaling correlations below the median and the value "1" for median correlations or higher." In other words, Latinas in the 1 range had greater consonance with the average or ideal Latina ranking of the six risk factors.
Model 2 (see Table 8) examines Latinas' correlation to an ideal, or average, physician ranking of the six risk factors for cervical cancer. Latinas' correlations with the physicians' rankings ranged from -1.0 to .940, with a median of -.269. Latinas whose rankings were more similar to the average physician rankings prioritized the sexual behavior risk factors and downplayed heredity and abortions as risk factors. We use individual Latinas' correlations to the physicians' ideal ranking to test the taken-for-granted hypothesis that ranking risk factors similar to physicians positively influences Latinas' use of Pap exams (a premise of most medical interventions).
Model 3 (see Table 8) examines Latinas' correlation to an ideal, or average, Anglo women's ranking of the six risk factors. Latinas' correlations to Anglo women's rankings ranged from -.89 to 1.0, with a median of .371. Latinas who had high correlations with Anglo women's beliefs ranked high heredity, lack of medical care, and vaginal infections, in that order, and ranked low many sexual partners, abortions, and sex at an early age, also in that order.
We examined the effect of these independent variables in separate logistic analyses in which all the other dependent and independent variables remained the same. The other independent variables include medical insurance (public or private), years of schooling, age, income, and an acculturation index that has been used with Latino populations that is primarily a language use index (Marin et al. 1987). The high correlation between the variable foreign- Including these variables allowed us to examine their influence on the use of Pap exams and to test for the independent effect of cultural beliefs on behavior.
We also included marital status as an independent variable. Marital status may have an independent influence on Pap exam use for at least two reasons. First, in the ethnographic interviews, Latinas often mentioned that they received their first Pap exams when they went to doctors during pregnancy. However, this effect is also partially  captured in the nativity variable, for the immigrants were younger than the U.S.-born Latinas and Anglo women and were more likely to have at least one child under 18 years of age in the household and to be focused on household activities rather than employed. For Latinas, therefore, acquiring a Pap exam often occurs when seeking other forms of medical care, especially prenatal care, rather than as primary preventive care for cervical cancer. This differs from the case of Anglo women, for whom preventive care is a much more central feature of their health-seeking model.
In addition, the literature on Latinos emphasizes the role of family and gender relations, which is another source of cultural beliefs. Latina immigrants often bring with them a set of general cultural beliefs about gender relations, patriarchy, sexuality, and morality (Dorrington 1995;Hondagneu-Sotelo 1994;Martinez et al. 1997;Paz 1961;Pefia 1991). Latina immigrants may adhere to these beliefs to varying degrees; for example, there are probably differences that are related to class background and local cultural milieus (Del Castillo 1993;Melhuus and Stolen 1996;Zambrana 1995;Zavella 1987). With such qualifications in mind, Hondagneu-Sotelo (1994) has observed that Latino immigrants' traditional gender ideals define both masculine and feminine behavior. The ideology of machismo characterizes men as independent, sexually assertive, and informally polygamous (Gutmann 1996;Pefia 1991;Zinn 1995). The ideology for women prescribes dependence, subordination, selfless devotion to family and children, and sexual chastity. Although undergoing change, traditional notions of patriarchy and family honor have traditionally hinged on unmarried women's virginity and on married women's fidelity to their husbands (Hondagneu-Sotelo 1994;Horowitz 1983;LeVine 1993).
Marital status, therefore, may be an important influence on the use of Pap exams-but for reasons similar to our understanding of Latinas' beliefs about risk factors. Married women, whose sexual activities fall within normative behavior, may be more inclined to have Pap exams than unmarried women. The cultural beliefs around gender also provide some insight into the relative importance that Latina immigrants, in the ethnographic interviews, gave to the role of the male spouse/partner in bringing home disease (infections).

The Influence of Beliefs on Behavior
The results presented in Table 7 indicate that Latinas with a median or higher correlation (consonance) with the Latinas' ideal ranking of the six risk factors for cervical cancer are about 36 percent more likely to have had a Pap exam in the two years prior to the interview than Latinas whose rankings were below the median correlation.  are not the only predictors of Pap exam use, nor are they the strongest predictors; but, importantly, they are significant even when accounting for the influence of the socioeconomic variables. Just as in model 1, medical insurance, being married, language acculturation, education, and age are significant predictors of Pap exam use. Notice, however, that the odds ratios are slightly lower than in Table 7, as a result of the beliefs variable also having predictive strength. Latinas who are closer to the physicians' ideal ranking of cervical cancer risk factors are about 60 percent as likely to have had a Pap exam recently than Latinas whose rankings were less similar to the physicians' rankings.
In model 3 (Table 8), Latinas with beliefs about risk factors similar to Anglo women's are 67 percent more likely to have had a Pap exam recently than Latinas expressing less consonance. This influence is statistically significant and occurs with the same control variables in the analysis.
Once again, medical insurance, being married, language We believe that these findings must be viewed from the perspective of the analysis of Latinas' cultural beliefs about cervical cancer risk factors that we have developed based on ethnographic interviews (Chavez et al. 1995;Martinez et al. 1997 do not raise the specter of nonnormative, morally questionable behavior. That married women are more likely than nonmarried women to seek Pap exams could also be explained, at least partially, by this interpretation. Some unmarried women may not seek Pap exams because they are not expected to be engaged in sexual activity outside of marriage; seeking a Pap exam is an implicit admission of sexual activity, thus raising issues of guilt and shame." Married Latinas, who are demographically young, are also more likely to have young children, with prenatal care visits providing an opportunity for physicians to provide a Pap exam. It is possible that Latinas who are closer to physicians' rankings also adhere to the physicians' causal model of cervical cancer, which focuses on the sexual transmission of HPVs. Just as in the ethnographic interviews, we found little evidence for such an interpretation. For example, physicians based their high rankings of starting sexual relations at an early age on the epidemiological connection between HPV and cervical cancer. Neither Latinas nor Anglo women made this connection, as evidenced by their consistently ranking it the lowest of all the risk factors. Although acquiring "infections" is perceived as an important risk factor, specific knowledge about HPVs appears to be lacking.14 We offer these interpretations of Latinas' beliefs and the underlying logic for those beliefs as a guide for future research on the use of Pap exams by Latinas and other women in the United States and in other national contexts. These findings have implications for interpreting the acculturation variable, which is constructed mainly from questions about language use. It does not answer the question, Acculturation to what beliefs? More acculturated, and U.S.-born, Latinas appear to be closer to the beliefs (rankings) of Anglo women about cervical cancer risk factors than they are to physicians' rankings." Thus, while the acculturation variable may predict behavior, it provides only marginal insight into specific beliefs about cancer risk factors and even less about the influence of these beliefs on behavior. Substituting an acculturation variable for a more comprehensive understanding of beliefs is not, in our opinion, supported by these findings (i.e., the ranking correlation variables are significant predictors of Pap exam use independent of the acculturation variable). Developing an understanding of specific cancer-related beliefs provides important, and significant, information about behavior, that is, the use of a cancer-screening test.
We end on a note about the strength of ethnographic research. We began our research with ethnographic interviews of a relatively small number of women in a restricted geographic area of Orange County, California. Through surveys, we applied what we learned about beliefs to a much larger population that was dispersed throughout the county. We began with Mexican and Salvadoran immigrants, but the larger survey included immigrants from a number of Latin American countries. We also began with U.S.-born women of Mexican descent (Chicanas), but the survey included a diverse group of U.S.-born Latinas. We also greatly expanded on the number of physician respondents. The consistency in beliefs about cervical cancer risk factors across the various samples suggests the generalizability of ethnographic findings on beliefs to much larger populations. Moreover, our initial interpretations of Latinas' beliefs were supported by the findings of the larger survey. Latina immigrants embed beliefs about cervical cancer within a larger context of social relationships and normative behavior. That these beliefs proved to be significant influences on behavior suggests not only the important ways that beliefs matter but that ethnographic methods for examining those beliefs also matter. Notes Johnson Foundation 1987. 2. Critical medical anthropologists have also sought a balance between cultural and structural factors in understanding access to medical services (Lock and Scheper-Hughes 1996;Singer et al. 1992). For more on cultural factors in cancer research, see American Cancer Society 1985, Balshem 1991, Erwin et al. 1992, Erwin et al. 1999, Hunt 1998, Loehrer 1993, Loehrer et al. 1991, Michielutte and Diesker 1982, Patterson 1987, Perez-Stable et al. 1992, and Posner and Vessey 1988. A more general dialectical anthropology also seeks to integrate the study of structure, cultural processes, and human agency (Basch et al. 1994;Comaroff and Comaroff 1991;Roseberry 1989).
3. Given the sensitive nature of the topic, specially trained female researchers interviewed female respondents. Interviews ranged between about two and four hours and were conducted in Spanish or English, depending on the respondent's preference. All interviews were transcribed verbatim in the language in which the interviews were conducted.
4. The analysis of ranked items builds on recent research in cultural consensus analysis, which is concerned with determining the extent to which a group of people shares "cultural knowledge" within a specific domain of culture (e.g., plant identification, kinship, and diseases) (Boster 1986;Garro 1986Garro , 1988Weller and Romney 1988;Weller et al. 1993). When interviewees achieve a specific level of agreement on the elements within a cultural domain, they are said to have an agreed-on "cultural model" (Romney et al. 1986;Romney et al. 1987;Weller and Romney 1988 7. Some Anglo women also discussed trauma as a risk factor, and other research (Wardlow and Curry 1996) has also found such beliefs among African American women.
8. The cooperation rate, defined as the number of completed interviews divided by the sum of the completed interviews and refusals by eligible women, was 1,225/(1,225 + 336) or 78.5 percent.
9. For a discussion of methods and a summary of general findings, see Chavez et al. 1997aChavez et al. , 1997bHubbell et al. 1997;and Mishra et al. 1998. 10. Average agreement scores for each of the eight groups were calculated by using the square root of the average Pearson correlation coefficient for each respondent with all other respondents within each group (Weller 1987 11. We also undertook the logistic regression analyses presented here using the ordinal-and interval-level variables rather than the recoded dichotomized variables, and the findings did not differ significantly. The same variables were significant. We report the findings on the dichotomized variables because it is easier to interpret the odds ratio statistic. 12. The high levels of insurance coverage and Pap exam compliance among Anglo women, that is, the lack of variation on these variables, do not lend themselves to multivariate analysis. For heuristic purposes, however, we did similar logistic analyses for Anglo women's beliefs. We found that being more similar or more different (consonance) to an average or ideal Anglo ranking did not significantly predict the use of Pap exams. What did predict such behavior among Anglo women were income and age. Age had a negative influence on the use of Pap exams, which may be related to the Anglo women's median age of 41, with many no longer of childbearing age. Although Pap exams are still important preventive exams for postmenopausal women, these findings suggest that older Anglo women are less likely to have such exams. Medical insurance was not statistically significant, which may be because of the lack of variation on this variable; most Anglo women have medical insurance of some type. We also found that ranking the six risk factors closer to an ideal physician ranking does not significantly predict Anglo women's use of Pap exams. Income and age continued to be strong predictors. For examples of Latina immigrants stating that they did not need a Pap exam because they were not engaged in sexual activities, see Martinez et al. 1997. 14. Additional evidence about a lack of knowledge concerning HPVs surfaced when, in 1998, we again conducted ethnographic interviews with 20 Mexican immigrant women in northern Orange County (for a discussion of the general findings, see Hubbell et al. n.d.) who were all low users of Pap exams, for this was the group that most interested us. Among the questions we asked were "Have you ever heard of human papilloma virus?" and "Some people have told us that cancer of the cervix is caused by the human papilloma virus. Do you think that's true?" Of the 20 Mexican immigrant women, only one had heard of HPVs, at a clinic she had visited a few days before the interview. As to whether she believed that HPVs cause cervical cancer, she replied, "I don't know that." All of the other women said that they did not think HPVs caused cervical cancer or that they did not know.
15. The acculturation variable has a .48 correlation to Latinas' correlations with Anglo women's rankings but a -.14 correlation to Latinas' correlations with physicians' rankings.