Public Reactions to Innovations in Science: Genomics, Race, and Identity

 Jennifer L. Hochschild and Maya Sen 

Government Department, Harvard University

October 30, 2010

Paper Prepared for the 32^nd Annual Research Conference of the Association for Public Policy Analysis and Management November 4-6, 2010, Boston MA

 DRAFT:  Please do not quote or cite without permission from the authors.  We welcome comments and suggestions; please send them to hochschild@gov.harvard.edu or msen@fas.harvard.edu.  Many thanks to Patrick Moynihan for substantial advice on designing the survey, Natalie Padilla and Claire Wheeler for exceptional research assistance, and the staff at Knowledge Networks for their professionalism in implementing the survey.  Thanks also to Jeremy Freese, Penny Visser, and anonymous reviewers for the wonderful public service of TESS (Time-sharing Experiments in the Social Sciences). This research was sponsored by the National Science Foundation through TESS and by the Robert Wood Johnson Foundation through a Health Policy Investigator Award. 

 Abstract: Although science and technology are touching people's lives in ways unimaginable only decades ago, political scientists and policy analysts are still exploring how the public understands and assesses new, highly technical scientific information.  This study uses a new public opinion survey to examine Americans’ reactions to and understanding of one scientific innovation: the use of genomics technology to trace ancestry, typically defined as race or ethnicity.  

 This arena has three analytic virtues.  First is its importance: genetics research may soon revolutionize medical practice in the United States, and possibly decisions in the criminal justice system as well as the way Americans understand race.  Second is its novelty: elite or partisan opinion on genomic science has yet to coalesce, and policies of support or regulation are just beginning to be developed. Our study can thus capture the early stages of opinion formation on a new issue. Third is its popular appeal: many Americans are being introduced to genomic science through racial ancestry tests, as seen in popular television shows or direct-to-consumer ads.

 Our goal is to refine existing models of public trust in science and technology by adding a new substantive focus, and placing two analytic elements at center stage: racial or ethnic identity as a lens through which other individual characteristics are channeled, and the relationships among emotional, cognitive, and salience responses to scientific innovation. More broadly, we argue that people with different immutable characteristics (such as race, gender, and age) respond to scientific innovation in intelligibly different ways, and that types of response to scientific innovation are related but vary in intelligible and important ways.  We posit, although we cannot show it in this paper, that all of these reactions inform support for science funding or regulation.

In a broad sense, scientific and technological innovations are moving in two directions at once.  The proliferation of nuclear power, the rise of the internet and computer-based communications, and global warming bring people closer together than they might perhaps wish, and have the potential to affect all humans more or less equally.  But other issues, in particular genomic science,[1] affect people as distinct, discrete, embodied individuals, whose particularities are the key to the science and its impact. Genomics may end up having world-wide applications and implications but unlike developments such as global warming, to use one example, it will connect with ordinary citizens in radically and intimately different ways.  Thus public support for some kinds of scientific research (e.g. global warming) may bear little relationship to public support for other kinds of scientific research (e.g. genomic medicine).  That context implies that policy analysts need to focus specifically on public support of genomics, especially since nonscientists’ understanding of, trust in, and reliance on the new science may differ from their views on other types of science, and may be associated with policy choices that are just now being hinted at.

Scholars initially sought to explain public responses to scientific innovation through a focus on scientific literacy or knowledge in general, as the central causal force.  Later researchers have added noncognitive factors, such as political ideology or religiosity, and have paid close attention to context through messages from the media or elites. In this paper we add three elements to that analytic trajectory: what we believe to be the first survey on views about genetic ancestry tests, evidence that racial identity serves as a framing device within which other variables are differentially associated with trust in science, and evidence that responses to scientific innovation differ systematically depending on whether one probes cognitions, affect, salience, or other responses to science. We also return, rather to our surprise, to the field’s original emphasis on knowledge.

We focus on Americans’ reactions to genetic ancestry testing because of three analytic virtues.  First is its importance: genetics research may soon revolutionize the way that medicine is practiced in the United States through new diagnoses and treatments and through at-home DNA tests.  Second is its novelty: elite opinion on genomic science has yet to coalesce, and policies of support or regulation are just beginning to be developed. The public therefore has few cues -- and no consistent cues -- about its moral or partisan valence or substantive validity, so our study can capture the early stages of opinion formation. Third is its popular appeal: many Americans are being introduced to genomic science through racial ancestry tests, as seen in a variety of popular television shows and other media.

            After a brief discussion of the literature on the cause of public views about scientific research, we develop our own analytic framework. We then test the new framework through analysis of a survey that we commissioned on public attitudes toward genetic ancestral testing. Our survey is missing crucial information – most importantly, attitudes about policy issues relating to science – but it provides a rich array of evidence on the components that are likely to be associated with such attitudes. [2]  We conclude the paper by discussing the broader analytic claims that this survey makes plausible, and the policy relevance of our results.

Research on public attitudes toward science can be roughly organized into three strains of argument, or perhaps three stages of development -- the scientific literacy (or knowledge-deficit) model, the low information rationality (or “cognitive miser”) model, and the value-predisposition model.   

Scientific Literacy: Starting with research by Fuyuan Shen (Shen 1975), scholars studying public opinion toward emerging science and technology developments have focused on scientific literacy, operationalized as measures of the public’s knowledge about scientific facts and the nature of scientific inquiry [for example, (Brossard and Nisbet 2006); (Bodmer 1985)]. Although Shen proposed that different types of scientific literacy could be important in different contexts (that is, scientific literacy for consumers versus scientific literacy for cultural understanding), subsequent work focused on the type of scientific literacy needed for effective citizenship (Durant et al. 1989; Durant et al. 1992).  Operationalizations vary, but Jon Miller has established the premise that an acceptable level of public scientific literacy is the sophistication needed to understand science articles in major newspapers such as the New York Times (or Le Monde or the Guardian) (Miller 1983; 1986; 1987; 1995; 1998; 2000; 2004).

 Surveys generally implement the concept of scientific literacy through a battery of questions about processes (e.g., the scientific method or probability theory) and specific areas of knowledge (e.g., definitions of  DNA or a molecule). The surveys show several things. First, Americans are alarmingly “illiterate” about scientific processes.[3]  “[O]ver the last four decades, the percentage of US adults with a minimal level of understanding of the meaning of scientific study has increased from 12 percent in 1957 to 21 percent in 1999,” while “the percentage of US adults who understand the basic idea of an experiment has increased from approximately 22 percent in 1993 to 35 percent in 1999.”  Second, the picture is even dimmer with regard to specific scientific topics.  For example, “[i]n 1997, 11 percent of US adults were able to provide a correct classification of a molecule, and 13 percent were able to provide a correct explanation in 1999.”  With regard to genomics and DNA, in particular, “[i]n 1990, approximately 24 percent of US adults were able to provide an explanation of DNA that included its role in heredity.  By 1999, the percentage of adults giving a response that clearly identified DNA as being responsible for heredity increased to 29 percent” [all quotations in this paragraph are from (Miller 2004): pp].

Third, the surveys showed that several attributes of respondents are consistently associated with levels of scientific literacy (e.g., Miller 2004).  Men tend to be more scientifically literate than women, and younger people more than older people. Both formal and

informal education is strongly related to scientific literacy, as are taking more college-level science courses and visiting science museums or reading science magazines.

Although the link is seldom explicit, this line of research is motivated in part by the belief that increased scientific literacy is associated with increased support for scientific endeavors. However, this logic is subject to strong criticism on two fronts.  Low scientific literacy is not, in fact, linked with a lack of interest in or enthusiasm for science and technology. In fact, despite minimal scientific literacy, “[f]or the last 15 years, approximately 70 percent of US adults have reported that they are very interested in new medical discoveries” and most Americans consistently report that they are “enthusasistic” about science (Miller 2004).  Levels of literacy are also poorly linked to direct measures of policy views, such as endorsement of governmental funding or other types of support of scientific research. 

Even more problematically, people’s opinions on scientific issues differ in ways that are not captured by levels of scientific literacy. For example, one meta-analysis of two hundred articles found variance in attitudes toward scientific topics among people with similar levels of scientific literacy (Allum et al. 2005).  More damning, other studies have a negative link between scientific literacy and support for specific areas of science.  Evans and Durant (1995), for example, find that support for particular policies (e.g., manipulation of human embryos) is actually negatively correlated with scientific literacy.  The evidence on the scientific literacy model is thus mixed at best.

Low-Information Rationality:  In this view, policy attitudes toward science are more closely linked with other attitudes and beliefs than with knowledge [(Ho et al. 2008);  Brossard and Nisbet 2006].  The underlying premise is that the public is “miserly” about acquiring information; people form policy preferences largely on the basis of cues from friends, family, and co-workers, or from elites and the media (Fiske and Taylor 1991; Nisbet 2005). Thus if the public (or a segment of the public) receives consistent messages endorsing (or disputing) the efficacy or importance of science and technology, opinion will tend toward (dis)approval for scientific endeavors.  If the messages are mixed, the public’s views will look (and perhaps be) essentially random or shallowly rooted (Zaller 1992). This model resembles political scientists’ argument that most voters use simple and easily accessible heuristics (donkey or elephant, a New Deal), or comments from an accessible elite,  in deciding who to vote for (Popkin 1993).

            Knowledge is not absent in these models; deference to scientific authority, which is associated with education, has both direct and indirect impacts on the public’s views of agricultural biotechnology.  Similarly, “science knowledge does play a modest role” (Brossard and Nisbet 2006).  But scientific literacy, or education more generally, tends to play a relatively small role in this model compared with heuristic devices.

Value Predisposition Models: Recent scholarship has merged aspects of both the scientific literacy and the low-information rationality models, and added new variables to explain support for science and technology.  The “value predisposition model” [as labeled by Shirley Ho and her co-authors (Ho et al 2009; Ho et al 2010)] holds that “the public might employ a dual-process approach by relying simultaneously on values and knowledge when forming attitudes about science and technology.”  Deeply held religious or ideological beliefs, for example, might be associated with individuals’ views of science and technology, particularly if the issue at hand has acquired a religious or political valance.  Religiosity thus is an important predictor of whether the public believes that nanotechnology is “morally acceptable” (Scheufele1 et al 2008). Religiosity and political conservatism combine to moderate the otherwise positive effects of knowledge in beliefs toward stem cell research (Ho, Brossard, and Scheufele 2008).  Alternatively, wealthier people or people living in cities might have more favorable views of science funding than their opposites.[4] 

Developing the Models: Outcomes, Framing, and Topic

Outcomes: These models, and their proponents, are not as sharply distinguished as we have suggested here, which makes them difficult to test in relationship to one another.  In addition, they, or rather in the research from which the models are derived, vary frustratingly with regard to the appropriate outcome of interest.  Dependent variables range from scientific literacy per se to normative views about science, endorsement of a particular scientific or technological innovation, or general support or approval of the enterprise. In our view, for purposes of policy analysis the appropriate outcomes of interest are support for various levels of government funding of science and support for various levels of government regulation of science (ranging in both cases roughly from “a lot” to “none”).[5]  Those are the measures that we will include in the national survey soon to go into the field, although space constraints kept us from including them in the survey that this paper analyzes.

            In lieu of appropriate outcome measures, here we take a step back and examine the array of intermediate outcome measures that we expect to be associated with levels of support for government funding and regulation of science. We posit that a given level of support for government funding and regulation of science is likely to be connected with a combination of different kinds of views, not merely knowledge, values, or deference to scientific authorities.  Thus our first hypothesis:

H₁: Support for government funding and regulation of science and technology is associated with a combination of three distinct reactions to innovation: trust in science, emotional responses to science, and beliefs about the importance of science.

A more refined version of H₁ would deconstruct “a combination of” into a determination of whether these views are equally important, additive, or interactive. That refinement awaits the full survey, which will include a full complement of the ultimate outcome measures. For now, we will examine patterns into which various views about one scientific arena fall.

Conceptual lens: The extant models of support for science posit an array of independent variables associated with a particular outcome; as noted above, the independent variables include knowledge, scientific literacy, religiosity, political ideology, and others.  Some researchers also include ascriptive characteristics such as gender or age in their models.  We posit that, at least in the realm of DNA ancestry testing, one characteristic – self-identified race or ethnicity – is not merely an additional independent variable but a framing structure, or lens, through which all of the other independent variables operate.  Thus our second hypothesis:

H₂: In a given scientific arena, one or another ascriptive characteristic may function as a lens through which independent variables are associated in different ways with the outcomes of interest. In the case of genetic ancestry tests, self-identified race or ethnicity plays this role.

We are not the first to examine whether race plays a distinctive role in explaining views about scientific innovation. More analytically, Schnittker, Freese, and Powell (Schnittker et al. 2000) examined whether views on the genetics components of mental illness vary across racial groups; they found that African Americans were more likely than whites to reject the notion that mental illnesses are caused by genetics.  Shostak and her colleagues found that, “contrary to expectations,” Whites are not more likely than Blacks to place a lot of importance on genetics in explaining social or health outcomes (Shostak et al. 2009).  Kessler and colleagues (Kessler et al. 2007) found that Blacks associate genetics more with family history rather than with specific medical concepts and language.  Most generally, the General Social Survey (GSS) has asked about “confidence in the scientific community” in every year of the survey from 1973 through 2008. Throughout that period, close to half of Whites expressed “a great deal of” confidence, and most of the rest expressed “some” confidence in the scientific community; barely 5 percent have “hardly any” confidence. Among Blacks, however, only about a quarter have a great deal of confidence and roughly 15 percent express hardly any confidence in the scientific community. 

And of course, there is a substantial academic literature showing that people of different races (and sometimes ethnicities) perceive and explain various features of American society differently from those of other groups ((Hochschild 2001); (Hochschild 1995); OTHERS). But we have not yet seen studies that seek to meld the rich research on support for science with the rich literature on race-specific views. That is the second contribution to which we aspire in this paper.

Topic: Finally, note the increase of public awareness of genetic ancestry testing; we believe, although we cannot prove, that it is the entryway for many Americans into an awareness of and views about genomic science more generally.  The George Lopez Tonight Show has featured a running series of DNA ancestry tests of Hollywood luminaries such as Snoop Dogg, Jessica Alba, and Larry David.  Henry Louis Gates Jr. has hosted two PBS series involving genetic ancestry testing, and has published several books about them (Gates 2007; 2008).  More generally, we counted (what we think are) all newspaper articles published in the United States on the topic of genetic testing and race.  Figure 1 shows the number of articles per year from 1988 through 2009.[6]  They have increased from around 100 articles per year in most of the 1990s to over 400 in the late 2000s.  The number peaks around 2006, coinciding with Gates’ documentary on the use of genetic testing of prominent African Americans. But journalistic interest in the topic remains relatively strong through 2008 and 2009.

Figure 1: Newspaper articles per year addressing race or ancestry and DNA  

Of course, we cannot say whether public knowledge of this arena is increasing at the same pace as media attention. But as of 2004, almost a fifth of Americans claimed to have heard or read “a great deal” about genetic testing and another two-thirds claimed to have heard “something” about it (General Social Survey 1973+); by 2010, 8 percent said that they have undergone “any type of genetic testing” (AARP 2010). So arguably this is an arena in which public awareness, and perhaps knowledge and direct experience, are growing, at least from a base of complete ignorance.

Data

The data for our analyses come from a 2010 survey of 1,095 adults conducted online by Knowledge Networks.  Respondents were equally divided into five self-defined groups: nonHispanic Whites, nonHispanic Blacks, Hispanics, Asian Americans, and Multiracials.  Appendix table A1 provides summary statistics for each race or ethnic group for the variables of interest in this paper.

The survey asked about a particular type of DNA ancestry test, which works as follows: a consumer takes a DNA sample (typically through a cheek swab), encloses it in a prepackaged envelope, and returns it to a (usually commercial) testing company.  The testing company reports the results, most often online. The results provide a general profile of the test-taker’s continent-wide ancestry (for example, “75% European, 20% South Asian, 5% Native American” or “97% African”).[7]

Because we were asking about abstract concepts in an arena in which few people have direct experience, we relied on hypothetical vignettes (Finch 1987).  These vignettes featured imaginary individuals, of the same gender as the respondent, who have received the results of a DNA ancestry test.  Each respondent addressed eight vignettes.  Half reported that the test results accorded with the respondent’s own self-identified race or ethnicity (“reinforcing” or “confirming” results), and half reported that the test results showed a mixed racial or ethnic background (“contradictory” or “refuting” results).  Note that for Multiracials, the mixed-race prompt was a confirming result whereas a mono-racial prompt was a refuting result.) For example, a Black woman respondent would receive the following two vignettes:

“Isabella is a woman who identifies as African American.  She has taken a DNA test that indicates that her female lineage can be traced primarily to Africa,” and

“Emily is a woman who identifies as African American.  She has taken a DNA test that indicates that her female lineage is spread across Europe or the Middle East, Africa, North America, Latin America or Spain, and Asia.”

For each of the eight vignettes, a map of the world appeared with the relevant continent(s) highlighted on the same screen.

The vignettes were modified in a variety of ways not relevant to this paper, so we ignore the remaining six vignettes here. What matters for our purposes is that each respondent was asked to imagine that he or she was the person in the vignette, and to respond to three questions: “If you were Isabella/Emily,

• Would the result make you feel pleased… displeased?” (7 point scale ranging from very pleased to very displeased);
• Would the result be believable… not believable?” (7 point scale ranging from very believable to very unbelievable); and
• Would this result matter a lot/not matter at all to your identity?” (2 point scale)

Thus we have responses addressing emotions, cognitions, and salience to two vignettes, one of which reinforces, and one of which challenges, the self-identified group identity of the respondent.[8]  We also have an array of demographic and other characteristics of the respondents, provided by Knowledge Networks since these respondents are part of their ongoing online survey pool.

In this exploratory study, we analyze the survey through top-line descriptive results, cross-tabulations of the three answer categories, and regression analyses both with and without race as a framing structure.

Top-Line Results

We begin with summaries of the data, for the sample as a whole and separately for each group.  Tables 1 and 2 display the results for the prompts that confirm and disconfirm, respectively, the respondents’ prior racial or ethnic self-definition. We collapsed the response categories for ease of interpretation.

Confirming Vignettes: Not surprisingly, very few respondents (7 percent) are displeased when DNA ancestry tests concur with their own identity, and a large majority (70 percent) believes the test results. Perhaps more surprising is the fact that for three tenths of respondents, a test confirming their identity would matter in shaping that identity. 

Table 1 here

However, when one compares across groups, more surprising results emerge. Whites are least likely to be pleased with single-race results (38 percent), and Blacks are most likely to be pleased (63 percent) – a result that completely reverses the logic of the “one drop of blood” rule imposed by Whites on Blacks roughly a century ago.  One-drop laws were promulgated in many American states, and in most states they were revised several times to be made more stringent, in order to maintain White racial purity and to ensure that all people with “mixed blood” would be socially constructed as Black, or at least as nonWhite.  Thus in many states, having a single Black grandparent or even great-grandparent would lead a person to be officially classified and socially perceived as Black.  Throughout the period in which one-drop rules were being promulgated, some Blacks protested this implied (or stated) insult, and many recognized their own racially mixed ancestry.  But a century later, White respondents are less likely to be gratified by a test that confirms their own ancestry as European while Black respondents are more likely to be gratified by a test that confirms their own ancestry as primarily African.  Multiracials are also relatively unlikely than other groups to be pleased with DNA ancestry test results that confirm them as racially mixed – an outcome that warrants further exploration.

Note also that for more Blacks (45 percent) and arguably Hispanics (36 percent) than for Whites (18 percent) or Multiracials (23 percent), this test matters a lot to their own identity.  Below we consider the relationship among emotions, cognitions, and salience; for now it is worth noting that salience varies more across racial groups than does gratification at the test results, and gratification varies more than does trust in the test’s outcome. These three dimensions of response to new scientific information are not synonymous.

Contradicting Vignettes: Next, consider the same array of responses to the vignette that contradicts the respondent’s self-chosen identity. That is, for all monoracial respondents, the named person in the vignette is found to have ancestry from many different continents; for Multiracials, the vignette character was assigned ancestry from only one continent, randomly chosen.

Table 2 here

Overall, fewer respondents would be pleased with a DNA ancestry test result that disputed their self-identified race or ethnicity (51 percent pleased with reinforcement, compared with 36 percent pleased with refutation). Similarly, fewer would find the refuting results believable (70 percent believe the reinforcing results, whereas 56 percent believe the refuting results). Surprisingly the same proportions would allow the DNA ancestry test to affect their identity when the test refutes their self-definition (29 percent) as when the test reinforces it (31 percent).  If these data have external validity, these proportions suggest that genomics will have an important, though complicated, impact on American racial dynamics.

For the contradicting vignette as for the confirming one, we find intriguing differences across racial or ethnic groups. Blacks, Hispanics, and especially Multiracials are more likely to be pleased with refuting test results than are Whites and Asian Americans.  For Multiracials, that result arguably stems from uncertainty about one’s own racial heritage; for Blacks and perhaps Hispanics, that result combined with the results from panel A of table 1 suggest that those groups are especially engaged with DNA ancestry tests no matter what they show.

Blacks are also the most likely to report that the genetic ancestry test results would matter a lot to their identity.  That holds not only for confirming tests as we saw above, but also for contradicting results; they are the only group to show this joint outcome.  Perhaps ancestry matters a great deal to Black Americans, or perhaps they are especially likely to be persuaded by the science of genomic testing, or both.

Comparison of “believable” responses across the two vignettes deepens the paradox of Blacks embracing, and Whites rejecting, the old “one drop of blood rule.”  There is a 20 percentage point difference between the proportion of Blacks who trust confirming and contradicting results (71 percent to 51 percent), compared with an 11 percentage point spread among Whites (66 percent trusting confirmation, and 55 percent trusting contradiction).  That is, Blacks are even more likely to believe a result showing monoracial than mixed-race ancestry than are Whites, despite the fact that most African Americans do in fact have nonBlack ancestors.  Hispanics and Asian Americans have a spread similar to that of Blacks, while for Multiracials the proportions are reversed – they are more inclined to believe results that contradict their self-definition (70 percent) than to believe results that confirm it (64 percent).  That may again reflect Multiracials’ uncertainty about their own ancestral heritage.

Regression Analyses 

These summary statistics suggest that race or ethnicity plays an important role in individuals’ response to scientific innovation, at least when the innovation engages with racial issues. As the literature on support for science reminds us, however, such results may reflect differences in value dispositions, use of heuristics, or knowledge rather than an outlook shaped by racial identity itself.  They may also be associated with demographic characteristics beyond race.  To explore these possibilities, we conducted ordered and binary logit regressions, with each response category regressed on:

• Age: a continuous variable noting the respondent’s age at time of the survey;
• Gender: an indicator variable for whether the respondent is female (1) or male (0);
• Race: an indicator variable for whether the respondent is Black, Hispanic, Asian American, or Multiracial, with White as the omitted variable;
• Household income: a 20-point scale ranging from “under $5,000” to “$175,000 or more;”
• Metropolitan status: an indicator variable for whether the respondent lives in a metro area (1) or in a rural area (0);
• Employed: an indicator variable for whether the respondent is employed for pay (1) or is not currently employed for pay (0);
• Education: a 14-point scale ranging from no formal education to professional or doctorate degree;
• Religious attendance: a 7-point scale ranging from “More than once a week” to “Never,” with the larger number indicating more attendance;
• Ideology: a 7-point political scale ranging from “extremely conservative” to “moderate, middle of the road” to “extremely liberal.” Conservative is coded as positive, liberal as negative;
• Partisanship: a 7-point scale ranging from “Strong Republican” to “Undecided/Independent/Other” to Strong Democrat.  Republican is coded as positive, liberal as negative.
• Internet access: an indicator variable for whether the respondent has internet access (1) or does not have internet access (0);
• Tech expert: a four-point scale measuring how strongly the respondent agrees or disagrees that “Others often rely upon me to stay up-to-date about the latest technology.”

Age and Gender are demographic categories analogous to Race.  Household income, Metropolitan status, and Employed are social characteristics of the respondent.  Education is the indicator of scientific literacy.  Religious attendance, Ideology, and Partisanship are value predispositions. Internet access and Tech expert are indicators of the respondent’s access to information from media or experts so that informative heuristics are readily available to him or her.

The response categories comprised a seven-point ordinal categorical variable for the emotional and cognitive responses categories, and a two-point dichotomous variable for the salience measure.   We include interaction variables for race by education, in order to explore the possibility that race functions as a lens through which other especially important characteristics operate.  For example, well-educated Blacks might be less supportive of scientific innovation than are well educated Whites, since the former may be aware of the Tuskegee experiment or early twentieth century “racial science.” 

            The pooled regressions are weighted using the survey weights, so any inferences from these should be applicable to the population of adult Americans. The regressions conducted separately by racial group are not weighted, so inferences are purely within the sample.  Similar to the R-squared measure in linear regressions, the deviance measure (also called "residual deviance") is used to evaluate the overall fit of logistic models.  Generally, the smaller the deviance, the better the fit. 

Hypotheses 1 and 2 generate several expectations.  First, variables that operationalize knowledge, value predispositions, access to heuristics, social characteristics, and demographic characteristics will be associated differently with the three types of response to scientific innovation (emotional, cognitive, and salience).  Given that the extant research literature shows mixed results in the relative importance of these variables, we do not have clear predictions about which is most strongly associated with a particular type of response. 

Second, at least in a scientific arena that focuses on race, respondents’ racial identity will play a distinct, perhaps unique, role in relationship to the three response categories. Furthermore, all of these results may differ depending on whether the genetic ancestry test reinforces or contradicts one’s initial self-definition.

Regression Analyses: Confirming Vignette:  Consider first the vignette that reinforces the respondents’ prior racial identity.  Table 3 presents the ordered and ordinary logit regression outputs, separately for the emotional, cognitive, and salience response items.[9]  A positive direction in the dependent variable indicates, respectively, increased pleasure, increased trust, and increased salience of the scientific test.  We omit (Column A) and then include (Column B) dummy variables for each racial or ethnic group with nonHispanic Whites as the baseline group. We then interact these dummy variables with education (Column C).[10] 

Table 3 here

Consider first columns 1A, 2A, and 3A.  One demographic characteristic, Age, has an impact on two of the three outcome measures older adults are slightly less pleased with DNA ancestry tests that confirm their racial identity, but also slightly less likely to allow them to matter in determining that identity. The identity of older people is apparently more fixed and less responsive to new scientific information than that of younger adults.  Among social characteristics, those with higher Household Income are more likely to trust the test’s results but less likely to allow them to affect their own racial or ethnic identity. Education, the first of the factors that is prominent in the literature on support for science, functions  roughly as Miller and other proponents of scientific literacy would expect.  Compared with poorly educated respondents, the well-schooled are more pleased with confirming DNA test results and more inclined to believe the science. With one exception, value predispositions have no association with emotional, cognitive, or salience responses to DNA ancestry testing.  We have no explanation for why people who attend religious services more often see DNA ancestry tests as important to their racial or ethnic identity than do the less religious; it may simply be a statistical fluke.  If our operationalization of access to heuristics is appropriate, cognitive miserliness also plays an unimpressive role.  People who consider themselves to be relative experts in technology use (Tech expert) are less likely than others to trust DNA ancestry testing, although internet access and technological expertise are not otherwise associated with reactions to the tests.

Now consider columns 1B, 2B, and 3B – using the same set of variables but adding in each racial or ethnic group separately to the list of independent variables.  Age is no longer associated, even minimally, with reactions to the tests but women are considerably more pleased with DNA test results that confirm their racial identity than are men.  Household income retains the same associations with the three test results. The relationship between Education and being pleased with or trusting the test results is strengthened, while the value predispositions remain irrelevant. Technology experts are even a little less likely to trust DNA testing, but otherwise, use of heuristics is irrelevant.

            In short, the role of independent variables other than Race does not change a lot when racial categories enter the analysis. However, the racial categories themselves do matter; what we saw in the initial summary statistics remains the case when controls are added. Compared with Whites, three of the four groups (Blacks, Hispanics and Asian Americans, but not Multiracials) are more likely to be pleased with DNA test results that confirm their racial identity, are more likely to trust the results, and are more likely to agree that the results matter. These effects are larger than the effects for any other variables; at least in the arena of racially-oriented genomic science, the most useful tool for understanding Americans’ attitudes will be their racial identity.  The race variables have more impact on the outcomes than any of those more prominent in the literature on support for science, which suggests that scholars need to pay more attention to salient demographic characteristics when tracing the public’s views of particular scientific innovations.

Finally, consider columns 1C, 2C, and 3C, which retain all of the previous independent variables and add interactions between each racial group (other than Whites) and education.  Adding the interaction term allows us to explore whether education, which has already been shown to be important and which is central to the literature of support for science, has a different effect across different racial groups. Although the effect is not consistent, education does play a distinctive role in one crucial case: among Black respondents, increased education means are positively linked with the salience of DNA test results confirming prior identity. This effect is stronger for Blacks than for similarly educated whites. 

Simulations: Confirming Vignettes: We do not try to interpret the coefficients on the race dummy variables in these regressions, as these coefficients have a substantive meaning only when the educational variable is equal to zero (i.e., if a respondent has no formal education).  That is as rare an occurrence in this survey as in real life.  Instead, in order to illustrate the importance of the racial or ethnic categories and to add some context to otherwise difficult-to-interpret logistic coefficients, we performed simulations using the non-interacted models in table 3, columns 1B, 2B, and 3B.  (We do these simulations using Zelig, a social science statistics package for the R statistical language (Imai et al. 2006).  Consider first the respondents' emotional responses (column 1B). Holding all other attributes at their mean or median and using the results from the statistical regressions, hypothetically changing the respondent's racial self-identification from White to Black would result on average in a statistically significant 17.5 percent increased likelihood that the respondent would be "completely pleased" with a confirming test result.  The results are also statistically significant for Hispanics, for whom "switching" their racial self-identification from White to Hispanic would result in an estimated 11.6 percent increased likelihood of being "completely pleased" with these test results.  For Asian Americans and for Multiracials, the results are weaker, with the slight increases in the "completely pleased" response not being statistically significant.

The same pattern holds when we examine the cognitive responses (column 2B).  Holding all other attributes at their mean or median, and using the results from the statistical regressions, hypothetically changing the respondent's racial self-identification from White to Black would result on average in a statistically significant 14.2 percent increased probability  that the respondent would find confirming test results "completely believable."  The results are also statistically significant for Hispanics and for Asian Americans.  For Hispanics, "switching" their racial self-identification would result on average in a 7.6 percent increased probability of finding the results "completely believable" and, for Asian Americans, it would result on average in a 8.3 percent increased probability.  The results for Multiracials were not statistically significant.

Lastly, we simulated how respondents would change their assessment of DNA ancestry tests' salience if their race were different.  A switch in racial self-identification from White to Black would be associated with, on average, a 25.3 increased percent probability of saying that the tests "matter a lot," holding all other variables at their mean or median. Switching from White to Hispanic would be linked with a 22.7 percent increase, while moving from White to Asian American would be associated with a 15.7 percent increase. All of these results are statistically significant.  For Multiracials, however, the estimate is not only close to zero (-1.4 percent), but also not statistically significant.

Regression Analyses: Disconfirming Vignette: We turn next in Table 4 to a parallel analysis, this time focusing on vignette results that challenge or contradict the respondents’ self-defined group identity.  The variables are the same, as is the analytic strategy.

Table 4 here

The overall pattern of results is roughly similar to that of Table 3 (the confirming vignettes), suggesting that respondents are reacting to this new scientific information in general more than to the particularities of the information itself.  There are, however, a few intriguing differences that mostly show stronger associations between explanatory variables and one or another of the outcomes for contradicting than confirming DNA ancestry tests.  Age has no association with responses to the contradicting vignette, but Gender has a strong and consistent effect.  Women are more pleased with contradicting results, trust them more than men do, and see them as more salient to their identity.  We lack an explanation for this strong and consistent finding, but it implies that scholars need to do more to figure out what “gender” means with regard to support for science than has heretofore  been done.

Setting aside race for the moment, we see that one of the three social characteristics, Household income, is associated with cognitive responses to and salience of disconfirming vignette.  Respondents with higher incomes trust disconfirming results less and are less likely to permit the results to shape their own identity. We had seen in table 3 that affluent respondents trust confirming results more, and in that case also they are less likely to permit the results to shape their own identity. The pattern of these outcomes perhaps reflects a level of self-confidence among the affluent that poorer people lack, but they certainly suggest that if any Americans are going to embrace DNA ancestry testing, it will be those with lower rather than higher incomes.

              Education is again associated with emotional and cognitive responses to the new information, as the traditional literature on support for science would expect. Well-educated respondents are more pleased with both confirming and disconfirming results, and they are more inclined to trust both sets of results.

            As with the confirming vignette, value predispositions play little explanatory role, with the exception of ideology in relation to trust in the results. That is, liberals have more confidence in disconfirming (but not in confirming) results of a DNA ancestry test than do conservatives. Finally, the two variables intended to capture the availability of heuristics for cognitive misers, Internet access and Tech expert, play opposite roles. People with little internet access are more pleased with disconfirming results, while people with relatively high levels of technological expertise see the results as more salient.  As with gender, we have no very compelling explanation, especially since the results for these two variables do not combine with the same variables in the confirming vignette to form a coherent pattern.

            Consider next the model in which race is added, that is, columns 1B, 2B, and 2C.  As with the confirming vignette, race or ethnicity is associated with all three reactions to the contradicting vignette even with a variety of controls added.  Race may have slightly less impact in this scenario, however, as indicated by smaller coefficients.  Three of the four nonWhite groups (Blacks, Hispanics, and Multiracials) are more pleased with disconfirming results than are Whites; Blacks and Hispanics were also pleased with confirming results, which suggests an overall engagement with the enterprise of DNA ancestry testing rather than a focus on a particular result.  Hispanics and Multiracials are more likely than Whites to trust disconfirming evidence; Hispanics were also more likely to trust confirming evidence. Blacks and Asian Americans, in contrast, are not  more likely than Whites to believe disconfirming DNA ancestry tests although they were more likely to believe the confirming results.  Blacks, Hispanics, and Asian Americans are all more likely than Whites to allow disconfirming, as well as confirming, results to affect their identity “a lot.”  Overall, the survey suggests that DNA ancestry testing will have a greater impact on nonwhite than on White Americans.

                       The interactions between race and education (columns 1C, 2C, and 3C) again show few results.  However, Asian Americans with high levels of education are less likely to be pleased with an analysis contradicting their racial identity than are comparably schooled Whites. 

Simulation: Disconfirming Vignette: To highlight the importance of the racial categories across the response categories and the different vignettes, we again performed simulations, this time using the non-interacted models in Table 4, Columns 1B, 2B, and 3B.  As we did for the confirming vignettes, we first consider respondents' emotional responses. Holding all other attributes at their mean or median, and using the results from the statistical regressions, hypothetically changing the respondent's racial self-identification to from White to Black would result on average in a statistically significant but small 4.4 percent increased likelihood that the respondent would be "completely pleased" with a contradicting test result. For Hispanics, Asian Americans, and Multiracials, the results are even smaller and none are statistically significant. 

                       The same pattern holds for the cognitive response category. For Hispanics, "switching" their racial self-identification would result on average in a 4.0 percent increased probability of finding disconfirming results "completely believable" and, for Asian Americans, it would result on average in a 3.8 percent drop in probability.  For African Americans, the change in probability is close to zero. None of these results are statistically significant.  Multiracials are different however; hypothetically changing a respondent’s self-identification from White to Multiracial is linked on average with an increased 11.2 percent increase in finding disconfirming results believable. This result is statistically significant.

                       Lastly, we simulated how respondents would change their assessment of disconfirming genetic ancestry tests' salience if their race were different.  Here, the results from the contradicting analysis resemble those from the confirming analysis.  A switch in racial self-identification from White to Black would be associated with, on average, a 21.5 increased percent probability of saying that the tests "matter a lot," holding all other variables at their mean or median. Switching from White to Hispanic would be linked with a 13 percent increase, while moving from White to Asian American with a 11.8 percent increase. All of these results are statistically significant.  For Multiracials, however, the estimate is not only close to zero (-1.9 percent), but also not statistically significant.

Regression Analyses: Race as a Lens: The results so far suggest that race is an important predictor of how people view and formulate emotional, cognitive, and salience evaluations of scientific topics, or at least this particular topic.  To further explore the role of racial self identification, we regressed the three outcome variables separately on each group: nonHispanic Whites, nonHispanic Blacks, Hispanics, Asian Americans, and Multiracials. Motivating this approach is the argument that research on support for scientific innovation should look more carefully at ascriptive characteristics of potential supporters (or opponents).  Older blacks, for example, might view research on the racial elements of genomics differently than younger whites because of their memory of the Tuskegee experiment or of discrimination in the medical arena; Asian American women may have a different take on ancestral testing than Hispanic women because of norms about racial mixture or purity, and so on.  Using race as a lens through which to analyze the other independent variables enables us to determine the nature and magnitude of these sorts of differences. 

Table A2 (in the Appendix) presents the results of this analysis for the survey vignettes confirming respondents’ prior racial self-identification. These group comparisons are far from methodologically perfect, as the number of respondents in each group is small and confidence intervals could easily overlap. (Note also that these regressions are not weighted, meaning that the inferences apply to the subsample of individuals included in the survey.) Nevertheless, they are a start.

As table A2 shows, some independent variables do vary in importance from one race or ethnicity to another.  Education, Age, and Partisanship are associated with Whites’ emotional response to confirming DNA ancestry tests, but none of these were important predictors for the other four groups. In this case at least, the literacy and value predisposition models hold only for white Americans. With regard to cognitive response, living in a Metropolitan area matters for White respondents, while for Blacks, Education and Household income are more important.  No variable appear significantly related to trust among Hispanics, Asian Americans, and Multiracials. With regard to the salience of the confirming ancestry test to one’s racial identity, no variables are significant for Whites and Multiracials, and those that matter for Blacks, Hispanics, or Asian Americans differ across the groups.

Looked at horizontally rather than vertically, of the fifteen possible relationships for any one variable (five groups with three outcomes for each group), no variable is statistically significant more than three times, and the overall pattern looks essentially random. Our point is not that we have identified a coherent pattern of predictors for any group; we have not.  Our point is rather that starting from respondents’ race or ethnicity may be necessary in order to understand how the factors usually involved in explaining support for science are, in fact, related to various outcome measures.

Table A3 presents the parallel results for vignettes that conflict with respondents’ prior group self-identification.  Again, characteristics vary in importance from group to group and none stand out as particularly important.  Gender and Religious attendance are significant predictors of Blacks’ gratification about conflicting test results, while Religious attendance and Employment status are for Hispanics.  No variable is significantly associated with pleasure at these results among White, Asian American, and Multiracial respondents.  Education matters for Whites’ and Blacks’ trust in the test results, but not for the other groups.  Of the three value predispositions, Partisanship has no strong association with any outcome, while Religious attendance and Ideology are intermittently important.  The two measure of ready access to heuristics play little role, although as always we note that those are imperfect measures.

Among Blacks, but no other group, four variables that were associated with one or another outcome in the confirming vignette are also associated in the same way and to the same degree with the contradicting vignette.  That suggests a consistent pattern of responses to DNA ancestry testing in general in that group, not only a response to the particular results.  But apart from that intriguing lead, we again conclude that we have not identified a coherent pattern of predictors.  Instead, we have identified reasons to be concerned about the current literature on support for science and reasons to focus more carefully on race, or perhaps other ascriptive characteristics, in trying to understand how and when knowledge, value predispositions, cognitive miserliness, or social circumstances relate to views on scientific innovation.

Relations among Outcome Measures

We turn finally to the outcome measures themselves, in order to develop our earlier point that research on public support for science needs to be clearer and more theoretically informed about just what “support” means.  As we noted earlier, we are hampered by not having a direct measure of public support for science in this survey; we are not the first to discover the perils of trying to use data for a purpose for which the data were not gathered.  Nevertheless, the fact that we have three outcome measures for each vignette provides some purchase on the question of how to conceptualize public support.

Tables 5 and 6 show cross-tabulations of the emotional and cognitive responses to the vignettes, further disaggregated by whether the respondent felt that the rest results would matter a lot or would matter not at all. Table 5 addresses the confirming vignette and table 6 the contradicting one.  It is clear that emotional and cognitive responses are closely associated; respondents who find the test results believable are most likely to be pleased with them, and vice versa.  Similarly, those who do not trust the test results are also likely to be displeased by them.  We are agnostic about the causal direction of this relationship; it probably works in both directions. 

Adding the third response category of salience complicates the picture, but in sensible ways. A higher proportion of those reporting that the test results matter also say that they find the results believable. That result holds in both the confirming and disconfirming vignettes.  In the confirming vignette, those who say the test results matter to their racial identity are also more pleased with the results than those who say the test results do not matter.  In the contradicting vignette, there is no difference in pleasure between those for whom the salience is high and those for whom it is low.

These general patterns point to a key theme: different types of questions around the general issue of support for science are associated in intelligible ways, but they are nonetheless distinct.  We have shown that different response categories are predicted by different independent variables (and that those differences may themselves vary by race or ethnicity).  Now we are also showing that, although pleasure, cognition, and salience are closely related, significant segments of the population have nonobvious responses – they may be pleased even if they do not trust the science, or may trust the science but not be pleased by the results, or may find the results not believable but still assert that they matter, or may find the results believable but still claim that they do not matter.  These complex responses could arguably affect their overall support for scientific innovation. However, determining how all of this is associated with our ultimate outcome of interest, support for scientific funding and regulation, awaits the next survey.

Discussion and Conclusion

 This paper had two purposes: to bring analytic attention to public attitudes toward the emerging science of genomics, and to extend research on the nature of and characteristics associated with public attitudes toward science.  Thus we focused on two issues within the realm of DNA ancestry testing, the independent variables that could be related to attitude formation and the outcome variables, or attitudes, themselves.  Here we draw conclusions about each issue in turn.

Independent variables: This exploratory survey suggests that ascriptive characteristics such as age, gender, or race may be more important than most (though not all) scholars have thought in shaping individuals’ attitudes toward science and technology.  Race or ethnicity is singularly important as a lens for making sense of genetic ancestry tests, regardless of the actual test results.  Blacks, Hispanics, and Asian Americans are more likely than Whites or Multiracials to be emotionally engaged with and to trust science that confirms their racial identity.  They are also more likely to think that it matters.  This effect remains, to some degree, when they consider science that conflicts with their existing identity. 

In addition, although neither we nor anyone else working in this field so far as we know anticipated this result, gender is also associated with individuals’ attitudes about racial ancestry tests.  Women are more emotionally engaged than men are when the science confirms prior racial identities; when the science conflicts with their self-identification, women are not only more emotionally engaged but also more likely to trust the results and to see them as salient. We cannot (yet!) offer a compelling explanation of these results, but we note their importance to our broader thesis: ascriptive characteristics may act as a lens through which individuals respond to scientific developments. 

This thesis exists in some tension with current literature on support for or understanding of science, which has focused on knowledge, value predispositions, and cognitive shortcuts.  We found at most intermittent support with regard to value predispositions, somewhat more sustained support for education, and little support for the use of heuristics (although our operationalization of the latter is weak, given the available items in the survey). This failure of the usual explanatory factors to be very compelling may be due to the newness of genomics as a public policy area, and its lack of a clear political or policy valence.  That is, adults may have learned very little about genomics through schooling; they have heard little about it through the media or other experts or elites; and even sophisticated Americans lack clear ideas of how it fits into religious beliefs, political ideology, or any other set of values. (That newness is, in fact, what has drawn us into this arena of research and what led us to focus on DNA ancestry tests for this initial survey.)  Even more than in most policy arenas, this topic may fall into Zaller’s category of people basically sampling one of the ideas that they happen to have available to them when they respond to the survey prompts.

Overall, we conclude our investigation of independent variables with the old saw, “more research is needed.” Race and gender are surprisingly important; standard variables are surprisingly weak; not much of the variance is explained in any case; vignettes with opposite messages are sometimes treated in the same way.  We are persuaded of the need to treat certain independent variables such as race and gender (and in some arenas, age) as a lens or prior condition(s) through with other more conventionally used independent variables operate, but beyond this the research field is open.

Outcome variables: This research also raises issues about the measures used to understand public opinion toward science and technology.  Different characteristics of the respondents become important in relation to their answers to one kind of question or another.  Both Technical Expertise and Household Income matter for cognitive responses and salience, but not for emotional responses to scientific innovation. Education matters for emotional and cognitive responses but not for salience. Ideology matters for cognitive responses to vignettes that contradict one’s racial identity, but not otherwise.

These differences matter for the ultimate outcome of interest, public views on support for and regulation of scientific innovation because the outcome variables are somewhat independent of one another.  They are linked but not so tightly that policy analysts or political actors can ignore the framing of the question that citizens are being asked to answer.  That is, if people are asked how much confidence they have in the scientific community (as many surveys do), the political and policy implications might differ from those emerging from a question asking about one’s gratification with or the salience of a new piece of knowledge.

This general point is well known among survey researchers, and is apparent in sets of questions about an attitude, the conviction with which that attitude is held, and the importance of that issue to the respondent.  However, the literature on public support for science has not sorted out clearly just what dependent variable(s) is of prime importance.  That is, to our knowledge, no study has explored how attitudes toward science vary according to the particular question being asked.  Instead, for the most part work in this field treats different responses as fungible for purposes of analyzing opinion formation.  A full exploration of these issues awaits more study.

We conclude with a note of frustration; the TESS survey did not ask directly about public support for funding or regulation of innovation in genomic science.  So we cannot draw conclusions about the links among race or other ascriptive characteristics, intervening variables such as education or value dispositions, reactions to the science, and views on appropriate public policy. Those links will be testable in the survey that we are now developing.

References

AARP (2010). Genetic Testing Survey

Allum, N.et al. (2005). Science Knowledge and Attitudes across Cultures: A Meta-Analysis. American Association for the Advancement of Science

Bodmer, W. (1985). The Public Understanding of Science. London: Royal Society.

Brossard, Dominique and Matthew Nisbet (2006). "Deference to Scientific Authority among a Low Information Public: Understanding U.S. Opinion on Agricultural Biotechnology." International Journal of Public Opinion Research 19(1): 24-52.

Durant, J.et al. (1992). "The Public Understanding of Science in Britain: The Role of Medicine in the Popular Presentation of Science." Public Understanding of Science 1: 161-82.

Durant, J.et al. (1989). "The Public Understanding of Science." Nature Genetics 340: 11-14.

Finch, Janet (1987). "The Vignette Technique in Survey Research." Sociology 21(1): 105-14.

Fiske, Susan and Shelley Taylor (1991). Social Cognition, 2nd Ed. New York: McGraw-Hill.

Gates, Henry Louis, Jr. (2007). Finding Oprah's Roots: Finding Your Own. New York: Crown.

Gates, Henry Louis, Jr. (2008). In Search of Our Roots: How 19 Extraordinary African Americans Reclaimed Their Past. New York: Crown Publishers.

General Social Survey (1973+). p. i. James Davis and Thomas Smith. Chicago IL: National Opinion Research Corporation

Ho, Shirley et al. (2008). "Effects of Value Predispositions, Mass Media Use, and Knowledge on Public Attitudes toward Embryonic Stem Cell Research." International Journal of Public Opinion Research 20(2): 171-92.

Hochschild, Jennifer (1995). Facing up to the American Dream: Race, Class, and the Soul of the Nation. Princeton NJ: Princeton University Press.

Hochschild, Jennifer (2001). "Where You Stand Depends on What You See: Connections among Values, Perceptions of Fact, and Political Prescriptions." Citizens and Politics: Perspectives from Political Psychology. J. Kuklinski. New York: Cambridge University Press: 313-40.

Hochschild, Jennifer and Maya Sen (2010). What Do DNA Ancestry Tests Reveal About Americans’ Identity?  Examining Public Opinion on Race and Genomics American Political Science Association. Washington D.C.

Imai, Kosuke et al. (2006) "Zelig: Everyone's Statistical Software." http://GKing.Harvard.Edu/zelig.

Kessler, L et al. (2007). "Knowledge About Genetics among African Americans." Journal of Genetic Counseling 15(xx): 191-200.

Miller, Jon (1983). "Scientific Literacy: A Conceptual and Empirical Review." Daedalus 112(2): 29-48.

Miller, Jon (1986). "Reaching the Attentive and Interested Publics for Science." Scientists and Journalists: Reporting Science as News. S. Friedmanet al. New York: Free Press: 55-69.

Miller, Jon (1987). "The Scientifically Illiterate." American Demographics 9(6): 26-31.

Miller, Jon (1995). "Scientific Literacy for Effective Citizenship." Science/Technology/Society as Reform in Science Education. R. E. Yager. New York: State University of New York Press: 185-204.

Miller, Jon (1998). "The Measurement of Civic Scientific Literacy." Public Understanding of Science 7(xx): 203-23.

Miller, Jon (2000). "The Development of Civic Scientific Literacy in the United States." Science, Technology, and Society: A Sourcebook on Research and Practice. D. Kumar and D. Chubin. New York: Plenum Press: 21-47.

Miller, Jon (2004). "Public Understanding of, and Attitudes toward, Scientific Research: What We Know and What We Need to Know." Public Understanding of Science 13(273-294).

Nisbet, Matthew (2005). "The Competition for Worldviews: Values, Information, and Public Support for Stem Cell Research." International Journal of Public Opinion Research 17(xx): 90-112.

Popkin, Samuel (1993). "Information Shortcuts and the Reasoning Voter." Information, Participation, and Choice. B. Grofman. Ann Arbor: University of Michigan Press: 17-35.

Schnittker, Jasonet al. (2000). "Nature, Nurture, Neither, Nor: Black-White Differences in Beliefs About the Causes and Appropriate Treatment of Mental Illness." Social Forces 72(xx): 1101-32.

Shen, Fuyuan (1975). "Scientific Literacy and the Public Understanding of Science." Communication of Scientific Information. S. Day. Basel: Karger.

Shostak, Saraet al. (2009). "The Politics of the Gene: Social Status and Beliefs About Genetics for Individual Outcomes." Social Psychological Quarterly 72(1): 77-93.

Zaller, John (1992). The Nature and Origins of Mass Opinion. New York: Cambridge University Press.

References

Hochschild, Jennifer (1995). Facing up to the American Dream: Race, Class, and the Soul of the Nation. Princeton NJ: Princeton University Press.

Hochschild, Jennifer (2001). "Where You Stand Depends on What You See: Connections among Values, Perceptions of Fact, and Political Prescriptions." Citizens and Politics: Perspectives from Political Psychology. J. Kuklinski. New York: Cambridge University Press: 313-40.

Popkin, Samuel (1993). "Information Shortcuts and the Reasoning Voter." Information, Participation, and Choice. B. Grofman. Ann Arbor: University of Michigan Press: 17-35.

Zaller, John (1992). The Nature and Origins of Mass Opinion. New York: Cambridge University Press.

Table 1: Emotional, Cognitive, and Salience Responses to Confirming Vignette

Panel A: Emotional response --confirm

Panel B: Cognitive response -- confirm

Panel C: Salience -- confirm

Table 2: Emotional, Cognitive, and Salience Responses to Contradicting Vignette

Panel A: Emotional response--contradict

Panel B: Cognitive response--contradict

Panel C: Salience--contradict

 Table 3: Confirming Vignette: Emotional, Cognitive, and Salience Responses

Regressions are weighted logits (ordered and binary).  Boldface  indicates significance at the .05 level.

Table 4: Contradicting Vignette: Emotional, Cognitive, and Salience Responses

Regressions are weighted logits (ordered and binary).  Boldface  indicates significance at the .05 level.

Table 5: Emotional and Cognitive Responses to Confirming Vignette

Panel A: Among those for whom the test result “Matters a lot”

Panel B: Among those for whom the test result “Matters not at all”

Note: Shaded areas are column percentages.

Table 6: Emotional and Cognitive Responses to Contradicting Vignette

Panel A: Among those for whom the test result “Matters a lot”

Panel B: Among those for whom the test result “Matters not at all”

Note: Shaded areas are column percentages.

APPENDIX

Table A1: Summary Statistics for Regression Analyses

For each variable, respondents not asked this item, respondents who refused to answer, and respondent for whom the variable is “not applicable or not familiar” are all excluded. Thus the percentages in table A1 may not add to 100 percent; in almost every case, only a few percent of the respondents are not included here.