Bias, Research

Updated About content Print Article Share Article
views updated



In the behavioral sciences, the difficulties of studying complex, changing interactions among living beings led to investigations of possible sources of bias. For example, the gender, race, class, and even presence of a researcher during an interview have been shown to influence the responses of the interviewee (Oakley). Researchers sought to apply the scientific method to problems in the behavioral sciences, in an attempt to eliminate bias.

Like all scholars, scientists hold, either explicitly or implicitly, certain beliefs concerning their enterprise. Most scientists try to use what they assume to be the best information to collect data and draw theories to elucidate the laws and facts that will be constant, providing that experiments have been done correctly. But the individuals who make observations and create theories are people who live in a particular country at a certain time in a definable socioeconomic condition, and their situations and mentalities impinge on their discoveries. Aristotle "counted" fewer teeth in the mouths of women than in those of men— adding this dentitional inferiority to all the others he asserted characterized women (Arditti). Galen, having read the book of Genesis, "discovered" that men had one less rib on one side than women did (Webster and Webster). Neither is true, and both would be refuted easily by observation of what would appear by today's standards to be easily verifiable facts. Although they also could count, they took these to be "nonclassical cases" because what we take to be facts can vary depending upon the theory or paradigm—the specific problematics, concepts, theories, language, and methods— guiding the scientist.

Because most scientists, feminists, and philosophers of science recognize that no individual can live a life and be entirely neutral or value-free since science and values are both very important. To some, "objectivity is defined to mean independence from the value judgments of any particular individual" (Jaggar, p. 357). The paradigms themselves, however, also are far from value-free. The values of a culture, in the historical past and the present society, heavily influence the ordering of observable phenomena into a theory. The worldview of a particular society, time, and person limits the questions that can be asked, and thereby the answers that can be given. Therefore, the very acceptance of a particular paradigm that appears to cause a "scientific revolution" within a society may depend at least in part upon the congruence of that theory with the institutions and beliefs of the society (Kuhn).

Elizabeth Potter (2001) documented Boyle's choice of the mechanistic model to explain his Law of Gases both because it comported well with the data and because it supported the status quo of conservative religion and monarchy of the seventeenth century with regard to class and gender compared to the competing animistic model seen as more radical socially. Scholars suggest that Darwin's theory of natural selection was ultimately accepted by his contemporaries (whereas they did not accept similar theories as described by Alfred Russel Wallace and others) because Darwin emphasized the congruence between the values of his theory and those held by the upper classes of Victorian Britain (Rose and Rose). Social Darwinists used Darwin's theory to base the political and social rights to their wealth and power held by men and the upper classes in biological determinism. In this manner Darwin's data and theories reinforced the natural superiority of wealthy men, making his theories acceptable to the leaders of Victorian English society. Fausto-Sterling's research (1999) revealed how different societies at particular historical periods have also used varying biological and genetic data as determinants for the social construction of gender and race.

Not only what is accepted, but what and how we study, have normative features. Helen Longino (1990) has explored the extent to which methods employed by scientists can be objective, in the sense of not being related to individual values, and lead to repeatable, verifiable results while contributing to hypotheses or theories that are congruent with nonobjective institutions and ideologies such as gender, race, and class that are socially constructed in the society: "Background assumptions are the means by which contextual values and ideology are incorporated into scientific inquiry" (p. 216). For example, scientists may calculate rocket trajectories and produce bombs that efficiently destroy living beings without raising the ethical questions of whether the money and effort for this research to support the military could be better spent on other research questions that might be solved by using similar objective methods.

Unintended Research Bias

Given the high costs of sophisticated equipment, maintenance of laboratory animals and facilities, and salaries for qualified technicians and researchers, little behavioral or biomedical research is undertaken without governmental or foundation support. The choice of problems for study in medical research is substantially determined by a national agenda that defines what is worthy of study, that is, worth funding. As Marxist (Zimmerman et al.), African-American (Campbell, Denes, and Morrison), and feminist (Harding, 1998) critics of scientific research have pointed out, the scientific research undertaken in the United States reflects the societal bias toward the powerful, who are overwhelmingly white, middle/upper class, and male. Members of Congress and the individuals in the theoretical and decisionmaking positions within the medical and scientific establishments that set priorities and allocate funds for research exemplify these descriptors. The lack of diversity among Congressional and scientific leaders may allow unintentional, undetected flaws to bias the research in terms of what we study and how we study it. Some have characterized the diversion of scarce resources away from public health measures known to prevent diseases for the masses towards the multibillion dollar Human Genome Project as an example of placing the interests of the powerful above those of the general public, since gene therapy and designer genes are likely to benefit fewer, wealthier people.

Examples from research studies demonstrate that unintentional bias may be reflected in at least three stages of application of the scientific method: (1) choice and definition of problems to be studied; (2) methods and approaches used in data gathering, including whom we choose as subjects; and (3) theories and conclusions drawn from the data.

CHOICE AND DEFINITION OF PROBLEMS TO BE STUDIED. Many diseases that occur in both sexes have been studied in males only and/or used a male-as-norm approach. Cardiovascular diseases serve as a case in point. Research protocols for large scale studies (MRFIT; Grobbee et al.; Steering Committee of the Physicians' Health Study Group) of cardiovascular diseases failed to assess gender differences. Women were excluded from clinical trials of drugs, they said, because of the desire to protect women or fetuses (and fear of litigation) from possible teratogenic effects on fetuses. Exclusion of women from clinical drug trials was so pervasive that a meta-analysis surveying the literature from 1960 to 1991 on clinical trials of medications used to treat acute myocardial infarction found that women were included in less that 20 percent and the elderly in less than 40 percent of those studies (Gurwitz, Nananda, and Avorn).

Many of these studies, including the Physicians' Health Study, were flawed not only by the factors of gender and age but also by factors of race and class. Susceptibility to cardiovascular disease is known to be affected by lifestyle factors such as diet, exercise level, and stress, which are correlated with race and class. Since physicians in the United States are not representative of the overall male population with regard to lifestyle, the results may not be applicable to most men. The data from these studies should not have been generalized to the population as a whole. (Some argued they directed studies to the group that they care about most, namely, people like themselves.)

Designation of certain diseases as particular to one gender, race, or sexual orientation not only cultivates ignorance in the general public about transmission or frequency of the disease; it also results in research that does not adequately explore the parameters of the disease. Most of the funding for heart disease has been appropriated for research on predisposing factors for the disease (such as cholesterol level, lack of exercise, stress, smoking, and weight) using white, middle-aged middle-class males. Much less research has been directed towards elderly women, African-American women who have had several children, and other high-risk groups of women. Virtually no research has explored predisposing factors for these groups, who fall outside the disease definition established from the dominant perspective.

Recent data indicate that the initial designation of AIDS as a disease of male homosexuals, drug users, and Haitian immigrants not only has resulted in homophobic and racist stereotypes but also has particular implications for women of color. In 1981 the first official case of AIDS in a woman was reported to the Centers for Disease Control and Prevention (CDC). By 1991, $80 million had been spent since the inception of the Multicenter AIDS Cohort Study (MACS), designed to follow the natural history of HIV among gay and bisexual males (Faden, Kass, and McGraw). Although by 1988, the case reports for women were higher than the number for men in 1983, the year the MACS began (Chu, Buehler, and Berelman), it was not until the final quarter of 1994 that the first study on the natural history of HIV infection in women began. In 1998, the CDC reported that AIDS remains the leading cause of death among black females aged 25 to 44, and the second leading cause of death overall among those aged 25 to 44 (CDC, 1998). The majority of women diagnosed with AIDS are black or Hispanic.

These types of bias raise ethical issues. Healthcare practitioners treat the majority of the population, which consists of females, minorities, and the elderly, based on information gathered from clinical research in which women and minorities have not been included. Bias in research thus leads to further injustice in healthcare diagnosis and treatment. Understanding this bias led to changes in policies in the 1990s. Investigators now receiving federal money must give a compelling reason if their studies fail to include both men and women, young and old, as well as individuals of diverse races. Although this increases the cost of research, since the sample must be larger, cost alone does not stand as a compelling reason.

APPROACHES AND METHODS USED IN DATA GATHERING. Using the white, middle-aged, heterosexual male as the "basic experimental subject" not only ignores the fact that females may respond differently to the variable tested; it also may lead to less accurate models even for many men. For example, the standard dosage of certain medications is not only inappropriate for many women and the elderly, but also for most Asian men, because of their smaller body size and weight. Certain surgical procedures such as angioplasty and cardiac bypass result in higher death rates for women (Kelsey) and Asian men and may require modification for the same reason (Chinese Hospital Medical Staff; Manley et al.).

When women of color are used as experimental subjects, clinicians often hold stereotypical and racist views that limit accurate diagnosis. For example, numerous research studies have focused on sexually transmitted diseases in prostitutes in general (CDC, 1987; Cohen et al; Rosser, 1994) and African-American women as prostitutes in particular. Several studies have also revealed that practitioners recognize and report at higher rates crack-cocaine abuse in African-American women and alcohol abuse in American Indian women, compared to white women seeking prenatal care. An American Civil Liberties Union study revealed that in forty-seven out of fifty-three cases brought against women for drug use during pregnancy in which the race of the woman was identifiable, 80 percent were brought against women of color (Pattrow, p. 2).

Frequently it is difficult to determine whether these women are treated disrespectfully and unethically due to their gender or whether race and class are more significant variables. From the Tuskegee syphilis experiment (1932–1972), in which the effects of untreated syphilis were studied in 399 men over a period of 40 years (Jones), it is clear that men who are black and poor may not receive appropriate treatment or information about the experiment in which they are participating. Scholars (Clarke and Olesen) explore the extent to which gender, race, and class become complex, interlocking variables that may affect access to and quality of healthcare.

Using only a particular discipline's established methods may result in approaches that fail to reveal sufficient information about the problem being explored. This may be a difficulty for research surrounding medical problems particularly important to the elderly, women, men of color, and homosexual males. Pregnancy, childbirth, menstruation, menopause, lupus, sickle-cell disease, AIDS, and gerontology represent healthcare issues for which the methods of one discipline are clearly inadequate.

Methods that cross disciplinary boundaries or include combinations of methods traditionally used in separate fields may provide more appropriate approaches. For example, heart disease is caused not only by genetic and physiological factors but also by social/psychological factors such as smoking and stress. Jean Hamilton (1985) has called for interactive models that draw on both the social and the natural sciences to explain complex problems. Some of the biological solutions such as Depo-Provera or Norplant implants (Washburn) favored for addressing teen pregnancy in some African-American and American Indian populations will be less effective without accompanying strategies based upon research from the social and behavior sciences on raising self-esteem, increasing education, and dealing with underlying family dynamics. Stripped of the complex of social, economic, educational, and family dynamics issues that may contribute to teen pregnancy, Norplant implants and Depo-Provera may prevent a particular pregnancy. Without information about family planning, counseling to deal with family problems, and education and job skills, however, such approaches do not solve the basic problems causing the teen pregnancy.

THEORIES AND CONCLUSIONS DRAWN FROM THE DATA. Emphasis upon traditional disciplinary approaches that are quantitative and maintain the distance between observer and experimental subject supposedly removes the bias of the researcher. Ironically, to the extent that these "objective" approaches are synonymous with a particular approach to scientific phenomena, they may introduce bias. As a corrective to such bias to a science that is too narrow, Sandra Harding proposes the notion of "strong objectivity" which recognizes the cultural, social, and historical forces that shape the questions asked by scientists, their approaches, and the theories and conclusions drawn from their data (1993, 1998).

Theories may be presented in androcentric, ethnocentric, or class-biased language. An awareness of language should aid experimenters in avoiding the use of terms such as "tomboyism" (Money and Erhardt), "aggression, " and "hysteria, " which reflect assumptions about sex-appropriate behavior (Hamilton). Researchers should use evaluative terms such as "prostitute" with caution. Often the important fact for AIDS research is that a woman has multiple sex partners or is an IV drug user, rather than that she has received money for sex. The use of such terms as "prostitute" may induce bias by promoting the idea that women are vectors for transmission to men when, in fact, the men may have an equal or greater number of sex partners to whom they are transmitting the disease. Even more important, by emphasizing AIDS in "prostitutes, " healthcare practitioners are able to distance themselves and their patients from the risk of AIDS. This may also lead to practitioners treating prostitutes as less than human and underdiagnosing AIDS in women who are not prostitutes. Focus on group characteristics such as "prostitute" or "poor, black, unmarried woman" repeats the initial mistake of identifying the disease by group rather than by behavioral risk.

Once a bias in terminology is exposed, the next step is to ask whether that terminology leads to a constraint or bias in the theory itself. Theories and conclusions drawn from medical research may be formulated to support the status quo of inequality for oppressed groups. Not surprisingly, the androcentric bias in research that has led to exclusion of women from the definitions and approaches to research problems may result in differences in management of disease and access to healthcare procedures based on gender. In a 1991 study in Massachusetts and Maryland, John Z. Ayanian and Arnold M. Epstein (1991) demonstrated that women were significantly less likely than men to undergo coronary angioplasty, angiography, or surgery when admitted to the hospital with a diagnosis of myocardial infarction, unstable or stable angina, chronic ischemic heart disease, or chest pain. This significant difference remained even when the variables of race, age, economic status, and other chronic diseases (such as diabetes and heart failure) were controlled. A similar study (Steingart et al.) revealed that women have angina before myocardial infarction as frequently and with more debilitating effects than men, yet they are referred for cardiac catheterization only half as often. Gender bias in cardiac research has therefore been translated into bias in management of disease, leading to inequitable treatment for life-threatening conditions in women. Women exhibited higher death rates from angioplasty (Kelsey et al.) and thrombolytic therapy (Wenger, Speroff, and Packard).

Recognizing the possibility of bias is the first step toward understanding the difference it makes and combating it. Perhaps white male researchers have been less likely to see flaws in and question biologically deterministic theories that provide scientific justification for their superior status in society because they gain social power and status from such theories. Researchers from outside the mainstream (women and people of color, for example) are much more likely to be critical of such theories because they lose power from those theories.

In order to eliminate bias and recognize the cultural, social, and historical forces impacting their research, the community of scientists needs to include individuals who serve as members on review panels and as leaders to review studies from backgrounds of as much variety and diversity as possible with regard to race, class, gender, and sexual orientation (Rosser, 2000). Only then is it less likely that the perspective of one group will bias research design, approaches, subjects, and interpretations. Since the scientific method itself is supposed to be "self-correcting, " if results are continually tested and subject to critical review, these biases are likely to be exposed.

sue v. rosser (1995)

revised by author

SEE ALSO: AIDS; Feminism; Genetic Discrimination; Metaphor and Analogy; Prisoners as Research Subjects; Privacy and Confidentiality in Research; Race and Racism; Research Policy; Sexism


Arditti, Rita. 1980. "Feminism and Science." In Science and Liberation, pp. 350–368, ed. Rita Arditti, Pat Brennan, and Steven Cavrak. Boston: South End Press.

Ayanian, John Z., and Epstein, Arnold. 1991. "Differences in the Use of Procedures between Women and Men Hospitalized for Coronary Heart Disease." New England Journal of Medicine 325: 221–225.

Campbell, George, Jr.; Deanes, Ronni; and Morrision, Catherine, eds. 2000. Access Denied: Race, Ethnicity and the Scientific Enterprise. Oxford: Oxford University Press.

Centers for Disease Control. 1987. "Antibody to Human Immunodeficiency Virus in Female Prostitutes." Morbidity and Mortality Weekly Report 36: 157–161.

Centers for Disease Control. 1998. "Guidelines for Evaluating Surveillance Systems." Morbidity and Mortality Weekly Report 37: 1–18.

Chinese Hospital Medical Staff and University of California School of Medicine. 1982. Conference on health problems related to the Chinese in America. San Francisco, May 22–23.

Chu, S.Y.; Buehler, J. W.; and Berelman, R. L. 1990. "Impact of the Human Immunodeficiency Virus Epidemic on Mortality in Women of Reproductive Age, United States." Journal of the American Medical Association 264: 225–229.

Clarke, Adele E., and Olesen, Virginia L., eds. 1999. Revisioning Women, Health, and Healing: Feminist, Cultural, and Technoscience Perspectives. New York: Routledge.

Cohen, J., et al. 1988. "Prostitutes and AIDS: Public Policy Issues." AIDS and Public Policy Journal 3: 16–22.

Faden, R.; Kass, N.; and McGraw, D. 1996. "Women as Vessels and Vectors: Lessons from the HIV Epidemic." In Feminism and Bioethics: Beyond Reproducion, pp. 252–281, ed. S. M. Wolf. New York: Oxford University Press.

Fausto-Sterling, Anne. 1999. "Gender, Race, and Nation: The Comparative Anatomy of Hottentot Women in Europe, 1815–1817." In Deviant Bodies: Critical Perspectives on Difference in Science and Popular Culture, pp. 19–48. NY: Routledge.

Grobbee, Diederick E.; Rimm, Eric B.; Giovannucci, Edward; et al. 1990. "Coffee, Caffeine, and Cardiovascular Disease in Men." New England Journal of Medicine 323(15): 1026–1032.

Gurwitz, J. H.; Nananda, F. C.; and Avorn, J. 1992. "The Exclusion of the Elderly and Women from Clinical Trials in Acute Myocardial Infarction." Journal of the American Medical Association 268(2): 1417–1422.

Hamilton, Jean. 1985. "Avoiding Methodological Biases and Policy-Making Biases in Gender-Related Health Research." In vol. 2 of Women's Health: Report of the Public Health Service Task Force on Women's Health Issues, pp. 54–64, ed. Ruth L. Kirschstein and Doris H. Merritt. Washington, D.C.: U.S. Dept. of Health and Human Services, Public Health Service.

Harding, Sandra. 1993. The Racial Economy of Science. Bloomington, IN: Indiana University Press.

Harding, Sandra. 1998. Is Science Multicultural? Postcolonialisms, Feminisms, and Epistemologies. Bloomington, IN: Indiana University Press.

Jaggar, Alison M. 1983. Feminist Politics and Human Nature. Totowa, NJ: Rowman & Allanheld.

Jones, James H. 1981. Bad Blood: The Tuskegee Syphilis Experiment. New York: Free Press.

Kelsey, Sheryl F. et al., and Investigtors from the National Heart, Lung, and Blood Institute Percutaneous Transluminal Coronary Angioplasty Registry. 1993. "Results of Percutaneous Transluminal Coronary Angioplasty in Women: 1985–1986 National Heart, Lung, and Blood Institutes Coronary Angioplasty Registry." Circulation 87(3): 720–727.

Kuhn, Thomas S. 1970. The Structure of Scientific Revolutions. 2nd edition. Chicago: University of Chicago Press.

Longino, Helen. 1990. Science as Social Knowledge: Values and Objectivity in Scientific Inquiry. Princeton, NJ: Princeton University Press.

Manley, Audrey; Lin-Fu, Jane S.; Miranda, Magdalena; et al. 1985. "Special Health Concerns of Ethnic Minority Women." In vol. 2 of Women's Health: Report of the Public Health Service Task Force on Women's Health Issues, pp. 37–47, ed. Ruth L. Kirschstein and Doris H. Merritt. Washington, D.C.: U.S. Dept. of Health and Human Services.

Money, John, and Erhardt, Anke. 1972. Man and Woman, Boy and Girl: The Differentiation and Dimorphism of Gender and Identity from Conception to Maturity. Baltimore: Johns Hopkins University Press.

Multiple Risk Factor Intervention Trial Research Group. 1990. "Mortality Rates after 10.5 Years for Participants in the Multiple Risk Factor Intervention Trial: Findings Related to A Priori Hypotheses of the Trial." Journal of the American Medical Association 263(13): 1795–1801.

Oakley, Ann. 1981. "Interviewing Women: A Contradiction in Terms." In Doing Feminist Research, pp. 30–61, ed. Helen Roberts. London: Routledge & Kegan Paul.

Pattrow, Lynn M. 1990. "When Becoming Pregnant Is a Crime." Criminal Justice Ethics, (Winter/Spring): 41–47.

Rose, Hilary, and Rose, Stephen. 1980. "The Myth of the Neutrality of Science." In Science and Liberation, pp. 17–32, ed. Rita Arditti, Pat Brennan, and Steve Cavrak. Boston: South End Press.

Rosser, Sue V. 1994. Women's Health: Missing from U.S. Medicine. Bloomington, IN: Indiana University Press.

Rosser, Sue V. 2000. Women, Science, and Society: The Crucial Union. New York: Teachers College Press.

Steering Committee of the Physicians' Health Study Group. 1989. "Final Report on the Aspirin Component of the Ongoing Physicians' Health Study." New England Journal of Medicine 321: 129–135.

Steingart, Richard M.; Packer, Milton; Hamm, Peggy; et al. 1991. "Sex Differences in the Management of Coronary Artery Disease." New England Journal of Medicine 325(4): 226–230.

Washburn, Jennifer. 1996. "The Misuses of Norplant: Who Gets Stuck?" Ms. 7(3): 232–236.

Webster, Douglas, and Webster, Molly. 1974. Comparative Vertebrate Morphology. New York: Academic Press.

Wenger, N. K.; Speroff, L.; and Packard, B. 1993. "Cardiovascular Health and Disease in Women." New England Journal of Medicine 329: 247.

Zimmerman, B., et al. 1990. "People's Science." In Science and Liberation, pp. 299–319, ed. Rita Arditti, Pat Brennan, and Steve Cavrak. Boston: South End Press.

More From