|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
FRAMING, PERSUASION, AND ATTITUDES TOWARD SAME-SEX MARRIAGE
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS
UNIVERSITY OF FLORIDA
This document is dedicated to the graduate students of the University of Florida.
I thank my committee, Dr. Kendal Broad and Dr. Chuck Peek, for their support and
guidance in this project. Dr. John Henretta provided much assistance as well. Numerous
other sociology graduate students and faculty allowed me to conduct my experiment in
their classes, including Clay Hipke, Alex Goldman, Dr. Leonard Beeghley, Dr. Monika
Ardelt, Dana Berkowitz, Dan Dexheimer, John Foster, Ray Hinojosa, and Victor
Romano. Robin Dungey helped ensure that I need not be homeless while conducting my
research. Finally, thanks are due to Danielle Doughman.
TABLE OF CONTENTS
A C K N O W L E D G M E N T S ......... ................................................................................... iv
L IST O F T A B L E S ................................................................... .............. vi
ABSTRACT .............. ..................... .......... .............. vii
1 BACKGROUND .................. .................. ................... .......... .................
2 LITER A TU R E R EV IEW ............................................................... ...................... 4
3 HYPOTHESES ....................................................... ............. ....... ...... 10
4 M E TH O D S A N D SA M PLE ........................................ ..........................................15
5 A N A L Y SIS ........................................................ 2 1
6 R E S U L T S .............................................................................2 4
7 DISCU SSION AND CON CLU SION S ............................................. ....................33
APPENDIX: THE SUCCESSFUL ARGUMENT .................................. ...............42
L IST O F R E F E R E N C E S .......................................................................... ....................43
B IO G R A PH IC A L SK E TCH ...................................................................... ..................47
LIST OF TABLES
1 Background Characteristics of the Sample ................................... ............... 19
2 Means for Dependent Variables for Different Argument Groups............................24
3 Predictors of Attitudes toward Rights and Responsibilities of Marriage ...............25
4 Predictors of Attitudes toward "Same-Sex Marriages".................. .........................28
5 Compare the Argument You Received to Other Pro-SSM Arguments ...................29
6 Predictors of Self-Reported Effects of the Study on Subject...............................32
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Arts
FRAMING, PERSUASION, AND ATTITUDES TOWARD SAME-SEX MARRIAGE
Chair: Kendal Broad
Cochair: Chuck Peek
Major Department: Sociology
This thesis considers what kinds of arguments can affect heterosexuals' attitudes
regarding same-sex marriage rights, examining one primarily narrative-based argument
alongside several more abstract, cognitive-rational arguments. Students in 12 different
sociology classes were given one of six different survey packets. The first section asked
about background characteristics and opinions. The second section, which was the only
section that varied among respondents, was one of six different arguments concerning
same-sex marriages (or, for the control group, an argument about global warming). The
arguments offered concerning same-sex marriages were all "pro" arguments based on
publicly available material found on the Human Rights Campaign's website
(www.hrc.org), but they differed greatly in the types of appeals they made. The final
section of the survey asked about opinions on same-sex marriages, in a variety of ways.
Regression models revealed that those subjects who received a story about a lesbian
couple and how they and their children would benefit from same-sex marriage rights (the
most narrative and personalized of all the appeals) became more supportive of same-sex
marriage rights than other subjects. The results support the idea that narrative appeals are
more likely to alter attitudes toward same-sex marriages than cognitive appeals, and
provide some support for the idea that inducing people to think of same-sex marriage in
terms of rights and benefits yields significantly more positive attitudes. The findings may
also be evidence of the persuasiveness of appeals that include discussions of benefits to
Over the course of the last 35 years, the lesbian, gay, bisexual and transgender
(LGBT) rights movement has sought to eliminate discrimination against, and advance the
rights of, gay and lesbian Americans. During the 1990s, public discussion of whether
LGBTs should have the right to marry people of the same sex emerged, with Hawaii and
Vermont serving as political flashpoints due to the unusually liberal policies they
adopted. 1996 saw the full nationalization of the issue in the "Defense of Marriage Act."
In 2003 the Massachusetts Supreme Court ruled that same-sex partners should be allowed
to marry, and several municipalities across the country followed suit. In the context of a
particularly close Presidential election campaign, this thrust the issue of same-sex
marriage very prominently into the American political scene. Numerous state-level
referenda restricting or banning recognition of same-sex relationships appeared on 2004
election ballots across the country, and the sitting President pledged to push for a
Constitutional Amendment against same-sex marriage.
Over the course of the last year and a half, proponents and opponents of same-sex
marriage have sought to make their cases and persuade people of their positions in
various ways.1 A great deal of research has suggested that how an issue is "framed" or
positioned in an evaluative context can induce people to think in very different ways
about it (Goffman 1974, Gamson 1992, Brewer 2002, Lakoff 2004). According to Snow
1 LGBTs are certainly not all in favor of making same-sex marriage rights a primary issue in the
movement's agenda. For a good discussion of internal debates on this topic, see Goldstein (2003).
and Benford (1992), a frame is "an interpretive schemata that simplifies and condenses
the 'world out there' by selectively punctuating and encoding objects, situations, events,
experiences, and sequences of action within one's present or past environment" (137).
Thus, cultural conservatives have attempted to influence public opinion by framing same-
sex marriage as an issue involving radical change in historical definitions of marriage, the
legitimization of homosexuality, redefinition of "family" and traditional gender roles, and
the recruitment of children into a "homosexual lifestyle." Proponents of allowing such
marriages have referenced equality, discrimination, changing definitions of marriage, and
the costs of being denied marriage rights, among other things. A few cultural
conservatives have even endorsed same-sex marriage on the grounds that it may promote
monogamy and "stable relationships" among LGBTs. However, it seems clear that the
dominant frames that opponents of same-sex marriage have relied upon involve
traditional morality, while proponents' primary frames generally involve discrimination
and equality. But until now, there has been no systematic study of the persuasiveness of
the different kinds of arguments deployed in this debate or of the success of different
ways of framing the issue.
This paper examines several different arguments that have been used by proponents
of same-sex marriage rights, and asks whether some arguments are more (or less)
successful than others at winning support for same-sex marriage. In a study of a
successful referendum campaign to roll back LGBT civil rights protections in Maine,
Weithoff (1998) argued for the importance of studying persuasion on LGBT rights issues:
GLBTs need to appeal to a wider audience than their immediately identifiable
constituents. While arguments to mobilize supporters are certainly important, the
GLBT community also needs to create argument schema that are attractive to
individuals not directly related to the gay civil rights movement (Pochna & Vegh,
1998). Because the majority of citizens are not seriously committed to either a
conservative or progressive perspective, their mental schemas on this issue are
likely to be somewhat permeable. Consequently, these people can readily embrace
new information and arguments when they are presented in a logically compelling
way (Kelly, 1955). Effective counter-arguments and competing narratives could
potentially change these voters' perspectives (Fairhurst & Sarr, 1996) and,
consequently, their proclivity to vote for pro-GLBT rights initiatives (or oppose
anti-GLBT rights initiatives) in future elections. (Weithoff 1998: 77-8)
This paper considers whether a narrative approach is more successful than abstract,
logical approaches, and further, whether different kinds of "rational" abstract frames have
different levels of success. While the subject area of the study is specific (that is, same-
sex marriage), its findings may be more broadly applicable to other issues involving
framing, persuasion, and social-political opinions.
This paper draws on two broad literatures: that concerning opinions about gays and
lesbians and homosexuality, and that concerning framing, persuasion, and opinion
A great deal is known about American public opinion on gays and lesbians and
homosexuality. The General Social Survey (GSS) has consistently asked a question that
reads: "What about sexual relations between two adults of the same sex--do you think it
is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?"
Further, it has regularly asked whether homosexuals1 should be able to make a speech in
town or teach in a college, and whether a book in favor of homosexuality should be
removed from the public library. Loftus (2001) relies on 25 years of data from the GSS
to analyze trends in opinion. In analyzing the trends shown in the GSS data from 1973
through 1998, Loftus relied upon answers to the three civil liberties questions as the
dependent variable in one set of her regressions, and on answers to the first question
discussed (acceptability of homosexuality) as the dependent variable in her other set of
regressions. On the whole, Americans are much more supportive of the three civil
liberties for LGBTs than they are willing to say homosexuality is "not wrong." Loftus
summarizes her findings concerning attitude trends as showing that
1 The term lionmosc \tu.I IIs" is considered negative by some for its historical association with the view of
same-sex attraction as a disorder. However, it is used at times in this paper when discussing opinion and
attitude survey questions that use the term "homosexuals," because expressed attitudes toward
liomoilc\ uiks" may be different than those toward "gays and lesbians" (or other terms).
Americans' attitudes toward the morality of homosexuality became slightly more
liberal from 1973 to 1976, became increasingly conservative through 1990, and
have become more liberal since 1990. Over the same 25-year period, willingness
to restrict the civil liberties of homosexuals declined slightly, the only departure
being a brief increase in negative attitudes in the late 1980s. (Loftus 2001: 778)
Yang (1997) finds a similar positive trend over time, though he points out that as of the
mid-1990s Americans remained, shockingly, roughly evenly divided over outlawing
Attitudes toward the acceptability of homosexuality are correlated with a number of
other factors. For example, Bowman and O'Keefe (2004) present a breakdown of
attitudes by various subgroups showing that (for example) in a Gallup 2003 survey,
females, whites, young people, the highly educated, those who rarely or never attend
church, East Coast and West Coast residents, Democrats and Independents, liberals, and
the well-off were far more supportive of allowing homosexuality to be legal than males,
blacks, older people, those with little education, frequent churchgoers, Midwesterners and
Southerners, Republicans, conservatives, and those with low family incomes. Herek
(1988; 2002a; 2002b) and Herek and Capitano (1999) have found that women are more
tolerant than men, and that gay males are more despised than lesbians. Altemeyer (2001)
examined attitudes among psychology students and their parents at one university and
found that the trend of greater acceptance of homosexuality was most likely due in large
2 Yang's data on this Gallup survey question ends in 1994, but Bowman and O'Keefe (2i" "14 followed it
through 2004. The question is worded as follows: "Do you think homosexual relations between consenting
adults should or should not be legal?" Support for allowing homosexual relations between consenting
adults to be legal, as measured by the Gallup poll, was higher in the years following Yang's analysis,
peaking at 60 percent in early 2003. It fell in the aftermath of the Lawrence v. Texas decision striking
down sodomy laws, and as of early 2004, more respondents [49%] were opposed to having same-sex
relations be legal than were supportive [46%] (Bowman and O'Keefe 2004). However, survey questions
worded slightly differently have found less intolerance on this issue, suggesting that some respondents may
misinterpret the question as asking whether same-sex relationships should have legal standing.
part to increased contact with people whom the subjects knew to be gay. Other research
using college students as subjects has suggested that variables such as "female sex, liberal
sex-role attitudes, lower religiosity as measured both by beliefs and attendance .. and
having positive contact with gay, lesbian, and/or bisexual persons" (Hinrichs and
Rosenberg 2004) and the personality variable "openness to experience" (Cullen, Wright,
and Alessandri 2002) are important predictors of attitudes toward LGBTs.
But how do these attitudes change with exposure to new stimuli? Social
psychologists have devoted considerable attention to persuasion. As Foumier, Martin,
and Nadeau (2002) note, various social psychologists have examined the roles played by
subject characteristics, source characteristics, message characteristics, and social context
in the outcome of attempts at persuasion (2). With regard to message characteristics in
political and social persuasion, one of the most well-known academic findings is that
narrative-format appeals and appeals that elicit emotion tend to be more successful that
abstract rational appeals (see, for example, Hovland, Janis and Kelley 1953; Cobb and
Kuklinksi 1997; Brader 2005), although abstract appeals do have some effect. According
to Perloff (2003),
Social psychologists argue that people are frequently more influenced by concrete,
emotionally interesting information than by "dry, statistical data that are dear to the
hearts of scientists and policy planners" (Nisbett et al., 1976, p. 132). Vivid case
histories personalized stories or narrative evidence exert particularly strong
effect on attitudes (Taylor & Thompson, 1982). According to this view, vivid
case histories evoke stronger mental images than abstractly presented arguments,
are easier to access from memory, and are therefore more likely to influence
attitudes when the individual is trying to decide whether to accept message
recommendations (Rook, 1987). Narratives are let's face it more interesting
than statistical evidence (Green & Brock, 2000). As stories, they engage the
imagination and are "intuitively appealing to humans, as we are all essentially
storytellers and avid story recipients" (Kopfman et al., 1998, p. 281). (182-183)
The distinction between narrative and abstract arguments is clearer than a related
dichotomy in social psychological research on persuasion: "emotional" or "affect-based"
arguments versus "rational" arguments. One critique of the emotional/rational dichotomy
is that there has never been a widely accepted operational definition of what constitutes
an "emotional" or "rational" argument; most of the definitions have been ad-hoc
common-sense definitions (Hovland, Janis, and Kelley 1953, Edwards 1990). In practice,
narrative-format arguments often are personalized and therefore are good at eliciting
emotions, while abstract argument involve complex cognition and are perhaps more
"rational." But there has been relatively little consideration given to different types of
arguments within the "emotional-narrative" and "cognitive-rational" spheres. That is, are
some types of abstract rational arguments-say, those based on moral justifications rather
than utilitarian justifications, or those that invoke the idea of equality, or those that use
legal reasoning-more effective than others? Some scholarly attention has been devoted
to the persuasive effects of arousing fear (Pratkanis and Aronson 1992, Perloff 2003) and
arousing sympathy, but there has been little direct comparison of these two techniques.
Kahnemann and Tversky's prospect theory (1979, 1984), suggesting that people tend to
be risk-averse, may play a role in how people evaluate policy arguments-that is, policies
that are framed as changes from the status quo are more likely to be negatively evaluated.
While Cobb and Kuklinski (1997) found that "con" arguments were more successful than
"pro" arguments regarding passing NAFTA and passing a national health insurance
policy, they also found that simple arguments on policy issues are not necessarily more
successful than complex arguments.3
3 Cobb and Kuklinksi note that their working definition of easy versus hard arguments is not entirely clear,
but suggest that "easy" arguments tend to be short, likely to invoke emotion, relatively concrete, and do not
Studies by Edwards (1990) and Edwards and von Hippel (1995) suggested that
those attitudes that are primarily rooted in affect are most likely to be changed by
"emotional" appeals, whereas those attitudes that are more cognitively rooted can be
changed either by cognitive by affect-based appeals. Stangor et al. (1991) found that
attitudes toward stigmatized minority groups, including gays and lesbians, were better
predicted by emotional responses to the groups than by cognitive evaluations of the
groups. These two findings together suggest that appeals to emotions may be more
successful in changing negative attitudes about gays and lesbians. However, Brounk
(1996), in a particularly relevant study to this paper, investigated whether "affect-based"
(narrative and personalized) or "cognition-based" (abstract) appeals were more likely to
alter subjects' attitudes toward homosexuals. Contrary to his hypothesis, he found that
both cognition-based and affect-based appeals increased student subjects' favorability
toward LGBTs to the same degree.
Regardless of whether an issue is cast in a narrative or in an abstract logical format,
a burgeoning literature on "framing" suggests that attitudes on political and social issues
can be influenced greatly by what people see an issue as being "really about." According
to Nelson and Kinder (1996), "frames are constructions of the issue: they spell out the
essence of the problem" and "suggest how it should be thought about" (1057). Perhaps
the best-known example of the power of framing is Nelson, Clawson and Oxley's (1997)
experimental study of news coverage of controversy over whether a municipality should
allow a KKK rally: those subjects who watched a news story depicting the conflict as
being about civil liberties were markedly more tolerant of it than those subjects who
require much cognitive effort to comprehend, while "hard" arguments tend to be longer, more abstract, do
not immediately invoke emotion, and demand more complex cognition.
watched a story discussing the issues of public order that might arise during the rally. It
is easy to see how this might be relevant to attitudes on same-sex marriage policy.
Whether one believes that the controversy is "really about" the integrity of the
Constitution, or legal rights, or cultural acceptance, or church-state separation issues,
could certainly affect how one feels toward it.
As of yet, there is no study that investigates framing and persuasion in the building
of favorability toward particular policies that address the interests of a disliked minority
group. That is, some previous work has investigated whether affect-based arguments are
particularly effective at changing attitudes toward a group (with mixed and contradictory
findings), but not whether narrative arguments are especially good at building support for
policies that might benefit that group. Other work (e.g. Cobb and Kuklinksi 1997) has
tested "hard" versus "easy" appeals and their effects on policy preferences but not with
regard to appeals and policies that deal with the interests of a stigmatized minority group.
In particular, the fact that many straight Americans appear to have very negative beliefs
about homosexuality at the same time that they endorse civil liberties and
antidiscrimination protection for LGBTs suggests that same-sex marriage may be an
especially interesting issue to investigate. Is same-sex marriage seen as a civil liberties
issue, or a "lifestyle endorsement" issue? Does this vary depending on how the issue is
framed? And how do these factors affect the ways that people form opinions about it?
This paper tests three primary hypotheses and four secondary hypotheses through
an experiment. The first hypothesis tests a basic assumption outlined in social
psychology, communications, and political psychology research (e.g. Cobb and Kuklinksi
1997, Gilbert 1991, McGuire and Papageorgis 1962, Perloff 2003): that many people
tend to be persuaded, at least to some degree and at least temporarily, by arguments they
read or hear about an issue if those arguments are not contradicted by counter-arguments.
Even those subjects who seem likely to oppose a particular argument (for example,
subjects who oppose same-sex marriage) may be moved by it somewhat if they have not
been "immunized" against persuasion in advance by learning about their opponents'
argument and potential rebuttals to it (McGuire and Papageorgis 1962; Anderson and
McGuire 1965; Billeaudeaux et al 2005). Therefore, the first hypothesis is:
HI: Those subjects who read any argument in favor of same-sex marriage will be
more supportive of same-sex marriage than those subjects in the control group, who read
an argument on an unrelated topic.
The second hypothesis draws on a rich tradition of research suggesting that appeals
that take at least a partly narrative form, that are personalized rather than abstract, and
that are structured to easily elicit emotion are more effective than abstract arguments that
rely more on cognitive-rational processes (Brader 2005, Perloff 2003):
H2: Those subjects who read an argument in favor of same-sex marriage rights
based around a story about a same-sex couple and their children will be more supportive
of allowing same-sex marriages than those subjects in the control group and those
subjects who read primarily cognitive-rational arguments in favor of allowing same-sex
The third hypothesis derives from national survey data indicating that Americans
are more favorable toward extending the specific protections involved in marriage to
LGBTs than they are toward "same-sex marriage," perhaps because they are concerned
that same-sex marriages are a cultural endorsement of homosexuality. If this is true, it
seems likely that framing same-sex marriage as an issue involving specific rights and
benefits will be particularly successful in building support for it. Thus, H3 is:
H3: Those subjects who read an argument listing some of the specific rights and
benefits that go with "marriage" will be more favorable toward allowing same-sex
marriages than those subjects in the control group and those subjects who read other
arguments in favor of allowing same-sex marriages.
The first two hypotheses will be tested with reference to four different dependent
variables, while the third will be tested with reference to three of the four dependent
variables. These dependent variables are described in the next section.
The four secondary hypotheses concern interaction effects. Specifically, they
concern the possibility that some types of arguments might produce very different
changes in attitudes depending on who reads the argument. For example, the argument
that "same-sex marriage would undermine traditional gender roles" would probably play
5 Some subjects received an argument that civil marriage should not be governed by religious definitions of
marriage because of separation of church and state and because civil marriage is distinct from religious
marriage. However, these subjects are not grouped in with those who received the list of rights in this
hypothesis because these subjects' argument contained no discussion of the rights and benefits of marriage,
increasing the chance that these subjects thought of "same-sex marriage" as at least a government
endorsement of homosexuality rather than as a policy to extend particular rights to same-sex couples.
quite differently with a traditionalist than a radical. Therefore, the first interactive
H4: African-Americans will respond differently than other subjects to the argument
about marriage having changed historically over time using an analogy comparing gay
rights to black civil rights and same-sex marriage to interracial marriage.
The direction of this difference is not specified. This is because it seems equally
plausible that the civil rights analogy will either (a) be particularly persuasive to blacks
because it hits close to home (i.e., if the comparison succeeds) or (b) be particularly
unpersuasive to blacks if they feel that the problems faced by gays are not as serious as
those faced by blacks.6
The second interactive hypothesis concerns gender. Gilligan (1993) argues that
women's moral reasoning processes are different than those of men. That is, women, at
least in modem Western cultures, think more relationally and are more attuned to
6 According to Lewis (2003), the widely held belief that blacks are more "anti-gay" than whites is
borne out by the analysis of 31 different surveys over a 30-year span, but with a twist: "Despite their
greater disapproval of homosexuality, blacks' opinions on sodomy laws, gay civil liberties, and
employment discrimination are quite similar to whites' opinions, and African Americans are more likely to
support laws prohibiting antigay discrimination" (Lewis 2003: 59). This finding appears to provide
powerful support for the idea that attitudes toward homosexuals and homosexuality are conceptually
distinct from attitudes toward civil rights and civil liberties for homosexuals. While blacks' greater support
for antidiscrimination laws for gays is likely rooted in their historical experience of oppression, it is
interesting to note that at least some prominent black leaders have spoken out strongly against gay rights
movements' comparison of the gay rights struggle to the black civil rights struggle and against the
comparison of the same-sex marriage issue to interracial marriage. For example, Jesse Jackson has
criticized this comparison, and the prominent Atlanta black minister Eddie Long has gained an enormous
following in large part by criticizing the idea of same-sex marriage. Some commentators have suggested
that President Bush's focus on same-sex marriage leading up to the 2004 election was in part motivated by
his hope that the issue could pull socially conservative blacks away from blacks' traditional strong loyalty
to the Democratic Party (Bolce, De Maio, & Muzzio 1993; McKissack 2004). While Bush's bid to attract
blacks was at best marginally successful (he drew 11% of black voters in 2004 versus 9% in 2000)
(McKissack 2004), it remains unclear how racial attitudes will interact with further efforts to advance the
rights of LGBT Americans.
personalized appeals, whereas men are more attuned to and persuaded by "rational" or
abstract arguments. This hypothesis is
H5: The story will have a disproportionately positive effect on women as compared
The third interactive hypothesis concerns the interaction of religious intensity with
the argument that religious definitions of marriage should not govern what is considered
marriage for civil purposes:
H6: Those subjects who are more strongly religious will react differently to the
argument that, due to the separation of church and state, religious definitions should not
govern civil marriage, than those subjects who are less religious.
As in the first interaction, the direction of this interaction is not specified. At first
this may seem surprising: surely we would expect that the highly religious would find the
argument for church and state separation less appealing than less religious people!
However, it is also possible that the highly religious subjects will display a greater
attitude change in favor of same-sex marriage rights than less religious subjects if (a) they
are opposed to same-sex marriages at the outset and (b) they are persuaded by the idea
(contained in the argument) that same-sex marriage advocates are merely trying to alter
the civil definition of marriage, not the religious definition.
The final hypothesis tested is:
H7: Those subjects with a gay or lesbian close friend or relative will respond
differently to the story than those subjects who do not have a gay close friend or relative.
Again, this hypothesis does not specify a direction, and again this may seem
surprising. It seems intuitive that those subjects with gay close friends or relatives would
respond more favorably to the story, since the story personalizes the issue and these
subjects could easily mentally "substitute in" their relative or friend for a character in the
story. However, it might also be the case that an argument that personalizes the issue
does no good for those people to whom it has already been personalized. In other words,
those subjects with a gay close friend or relative might already have performed the
mental operations that the story is intended to generate, and the story might therefore
have no additional effect on them or at least could have less of an effect on them than
on subjects without a gay close friend or relative.
METHODS AND SAMPLE
To test the effects of various arguments about same-sex marriages on respondents, I
designed a three-part survey instrument. The first part of the survey included about 25
questions including items concerning basic demographic variables (age, race, parents'
income, etc.), general measures of attitudes toward homosexuality and homosexuals,
questions about the respondents' sexual orientation and sexual experience, and measures
of other variables that might be expected to have significance in predicting attitude about
same-sex marriage (political ideology, type of area in which the respondent grew up,
The second part of the survey instrument was a one-page long argument in favor of
same-sex marriages. However, there were different versions of this second part. I drew a
variety of arguments from the website of the nation's most prominent pro-LGBT rights
group, The Human Rights Campaign (www.hrc.org), modified them slightly (primarily
so that each argument was roughly 2/3 of a single-spaced page to one page in length), and
randomly inserted one of them into each survey packet so that each respondent received
only one of the arguments as a stimulus. I pulled five different arguments off of the HRC
website to create my "argument groups." I added in a control group that received an
argument about global warming rather than same-sex marriage.
7 Four of the five arguments in favor of same-sex marriages were pulled from the Human Rights
Campaign's booklet "Answers to Questions about Marriage Equality" found at
Ihp \ \ \ .hrc.org/Template.cfm?Section=Center&Template=/TaggedPage/TaggedPageDisplay.cfm&TP
LID=63&ContentID=17353. The fifth argument, the story, was pulled from HRC's materials at
hup \ \ \\ .hrc.org/millionformarriage/hrc_adcenter/jo_teresa.html. This argument is reproduced in the
Of the five arguments in favor of same sex marriage rights that were selected, each
argued from a different angle. One, STORY, was a story about a lesbian couple and their
kids, and the risks posed to the kids by the fact that their parents could not marry. This
appeal was mostly personalized and narrative (though it also contained a component
stating that the story was an example of why same-sex marriage rights would be
desirable), whereas the other arguments were more purely rational and abstract, and less
personalized. NOTRELIG was an argument stating that civil marriage and religious
marriage are distinct and that (due to the separation of church and state) religious
definitions of marriage should not control what constitutes civil marriage. This argument
also stated that allowing same-sex marriages would in no way require any church or
religion to perform these kinds of marriages. Another argument, RIGHTS, was a list of
rights that same-sex partners are denied as long as they cannot marry, followed by the
argument that same-sex marriages should be allowed. The fourth argument, CONST,
was an argument stating that the US Constitution has never been amended to discriminate
against a particular group, and that it would be inappropriate to use it to discriminate.
The final argument, MARCHANG, was about how marriage as it now is has not been
"around forever, in every society." This argument discussed the facts that marriage for
love is a recent innovation and that interracial marriage was forbidden in many parts of
the US until relatively recently. The control group, CONTROL, received an argument
rebutting the "climate skeptics" and stating that global warming is real. In every case
(except for the control group), the arguments did not merely present information, but also
explicitly argued in favor of allowing same-sex marriage.
appendix. The global warming argument for the control group was pulled from Greenpeace's website:
The final part of the survey consisted on measures of opinion toward same-sex
marriage and civil unions, as well as a few questions asking about the effects of the
experiment and some questions asking respondents about the effectiveness of different
arguments. There were several open-ended questions in this section, but the majority of
questions were Likert scale-style questions with assigned values ranging from 1 through
3 or 1 through 5. (The items used in this paper as dependent variables are described
The three parts were combined into a single packet that was given to respondents.
Surveys were obtained from a total of 523 respondents from 12 different Sociology
classes (none of them taught by this author), including 2 classes at a nearby Community
College and 2 upper-division classes. Of the 523 surveys obtained, results from thirteen
students could not be used because the students who had filled them out were under age
18. Two students filled out the beginning questions about age and major, but nothing
else, rendering their surveys useless for the study. Respondents from the 2 upper-
division courses were excluded. This is because including these classes would have
meant including students who had higher levels of exposure to Sociology, and might
react in different ways to the survey than introductory-level students. Respondents who
indicated their sexual orientation was either "gay/homosexual" or "bisexual" were also
excluded. Only 7 students, or about 1% of the remaining part of the sample, chose one of
these options, and these 7 students were clearly (in this sample) more supportive of same-
sex marriage rights than the heterosexual students.8 The remaining sample consisted of
8 The 1% figure is surprisingly low, perhaps due in part to the fact that this question appeared on the first
page of the survey, making LGBT students reluctant to answer honestly for fear of having their answer
In 6 cases this author came to the participating class and administered the survey;
on 2 occasions a colleague other than the class instructor was hired to do so; and 4 times
the instructor of the class administered the study. The response rate was low for the
initial two classes, most likely due to the fact that the survey was presented to students as
a voluntary exercise at the end of class. However, when the timing of survey was moved
to the beginning of classes rather than the end, response rates skyrocketed to about 90%
of the students who were present.
The sample used was a convenience sample, not a random probability sample, and
the views of the subjects cannot be generalized to the population as a whole. The most
obvious problems with doing so include the following: First, the fact that young people
have relatively liberal attitudes on LGBT rights issues has been extensively documented
by numerous sources (e.g. Loftus 2001; Bowman and O'Keefe 2004), and young people
may also differ from a full-age-range sample in other ways. Second, all respondents are
currently enrolled in an educational institution, and the majority of respondents are
enrolled in an educational institution that is considered by many to be the state's best
public university. The sample likely has higher educational aspirations, a higher level of
academic attainment, and a higher socioeconomic class background that the population as
a whole. Third, the sample consisted entirely of students who had chosen to take a class
in Sociology. The clearest effect of this on sample composition is that there were roughly
twice as many women as men sampled, as Sociology classes at this university tend to be
composed mostly of women. Women are generally more positive toward LGBTs than
men. All of the characteristics identified in this paragraph might reasonably be predicted
seen by other students. According to Black et al (2000), about 2.5% of Americans self-identify as gay or
to result in finding more liberal attitudes toward same-sex marriage and homosexuality in
the sample than in the U.S. population.
Table 1. Background characteristics of the Sample (Maximum N = 438)
Mean or Percent St. dev.
% Female 67.9% N/A
% Nonwhite 34.5% N/A
Mean age 19.1 years 1.8
Mean religious strength 5.5 (1 to 10 scale) 2.6
Mean social liberalism 6.1 (1 to 10 scale) 2.3
Mean for "homosexuality wrong"9 3.0 (1 to 5 scale) 1.8
% with gay close friend or relative 56.9% N/A
% Democrat 57.5% N/A
% Republican 40.0% N/A
However, the fact that the sample is unrepresentative of the U.S. population should
not cause us too much concern, for several reasons. First, and most importantly, while it
is not a valid source of data about opinions of the population, it is useful as a source of
data about persuasion and opinion change. (This use is subject to some caveats,
discussed in the conclusions section.) Second, though this paper does not do so, it is
possible to assign weights to the sample to correct for some of the ways in which it is not
representative. This could not correct for the age or education skews of the sample, but
could easily be used to correct other problems that may be present in terms of
representativeness. Third, the sample is an excellent source of data about the opinions of
well-educated young people in a politically centrist state. This is important because it
9 "Mean for 'homosexuality is wrong'" refers to the mean of 2 questions asking "Which of the following
best summarizes your attitude toward a man having sex with another man/woman having sex with another
woman?" Answer choices, based on the GSS options for a similar question, ranged from "It is always
wrong" (1) to "It is not wrong at all" (5).
provides a guide to the opinions of those people who will likely occupy high-status and
high-influence roles in society, and thus have greater influence than would be expected
merely from their numbers alone, several decades from now.10
'1 The usefulness of the sample is perhaps enhanced by the fact that the distribution of political party
identification in the sample nearly exactly mirrors the distribution of 2004 Presidential votes in the county
in which the university is located. The sample, like the county, was about 60% Democrat and 40%
Republican; this similarity may indicate that the classes sampled are less "liberally biased" and
unrepresentative of students at the university than some might assume.
To analyze the data, four dependent variables were created based on the questions
asked in part 3 of the survey packet. The first dependent variable combined scores from
4 closely related questions. The 4 questions asked subjects to support, oppose, or
indicate that they were neutral on the following proposals: extending Social Security
benefits to same-sex partners, extending health insurance benefits to same-sex partners,
extending inheritance rights to same-sex couples, and extending child support
responsibilities to same sex partners. A scale, ranging in value from 4 to 12, was created
based on these questions, with 12 indicating support for all 4 policies. (Cronbach's alpha
for the scale was a quite high value of .90, indicating that the 4 variables are indeed
related in subjects' responses patterns. The value would have been improved by
removing the child support question, but only very slightly, to .92.) This scale served as
a measure of subjects' support for the rights and responsibilities that go with marriage -
regardless of how they might react to the words "marriage" or "civil unions."
The second dependent variable created was a scale based on 5 questions. These
questions were: 1. Would you support or oppose a federal Constitutional Amendment that
would define marriages as something that can only take place between a man and a
woman, and would make same-sex marriages illegal? 2. Which of the following comes
closest to your position on same-sex marriages/civil unions? [Same-sex couples should
be allowed to legally marry; same sex couples should be allowed to form civil unions but
not marry; same-sex couples should get no legal recognition.] 3. Do you strongly favor,
favor, oppose, or strongly oppose allowing gays and lesbians to marry legally? 4. Do you
strongly favor, favor, oppose, or strongly oppose allowing gay and lesbian couples to
enter into legal agreements with each other that would give them many of the same rights
as married couples? 5. If your home state drafted a law defining marriages as a union
that can only occur between a man and a woman, and made same-sex marriages illegal,
would you support or oppose this law?11 All of these questions offered answer choices
that could be scored from either 1 to 3 or from 1 to 5. The scoring was adjusted so that
the questions with 3 answer choices were instead arrayed from 1 to 5. A scale was then
created, ranging from 5 to 25, with 25 indicating the greatest support for same-sex
marriages. This variable served as a measure of support for same-sex marriages when
described as such, as distinct from the specific policy items that might be included under
such terms. That is, while the first dependent variable measures support for granting
same-sex couples the tangible rights of marriage, this one measures support for the idea
of "same-sex marriages." (Question 4, which was intentionally worded identically to a
question appearing in some national surveys, was probably originally intended to be a
question measuring support for civil unions. However, it was included in this scale
because it uses the phrase "married couples" and likely was interpreted accordingly.
1 Two other questions intended to measure attitudes on this topic were included in the survey, but not in
this scale. The first asked, "Do you think defining marriage as a union only between a man and a woman is
an important enough issue to be worth changing the Constitution for, or isn't it that kind of issue?" This
question was asked because it has been included in national surveys, but the pattern of responses suggested
that a huge number of respondents misunderstood the question perhaps because it is a leading question.
Another question asked, "Do you think that allowing two people of the same sex to legally mary will
change our society for the better, will it have no effect, or will it change our society for the worse?"
Subjects were substantially more negative on this question than on the other indicators of attitudes about
same-sex marriage rights, with the mean slightly on the negative side (2.09 on a 1 through 3 scale, with 3
meaning "worse"). More than 20% of subjects selected "don't know" or did not answer this question, and
inclusion of this variable in the scale would have meant eliminating those 20% from the analysis. Separate
logistic regressions for the "effects on society" variable alone revealed that the story did not have any
discernable effects on beliefs about the effect of same sex marriages on society. This finding will be
discussed in more detail below.
Regressions tested, but not shown here, for a scale that omitted this question showed
results very similar to the 5-item scale.) Cronbach's alpha was again very high, .96.
The two remaining dependent variables directly asked for subjects' evaluations of
the experiment and its effects. The third dependent variable, COMPARE, asked subjects
to compare the quality of the argument they read in the study to "other arguments" they
had "read or heard in favor of same-sex marriages," regardless of their own position on
the topic. This was scored on a 1-to-5 scale (with an extra answer choice reading "the
argument that I read was not about same-sex marriages" to screen out the control group).
The final dependent variable, EFFECTS, asked subject to rate on a 1-to-5 scale the effect
that participating in the study had on them. Answer choices ranged from "It made me
much more supportive of same sex marriages" to "It made me much less supportive of
Some subjects did not answer every question. For continuous independent
variables, the mean was substituted in for any missing values, and a dummy was created
to flag those cases where substitution had occurred. For the dependent variables, means
were not substituted in. Instead, cases with missing values were ignored in the analysis.
The means for each different variable, broken up by argument group, are presented
in Table 2 (below). This table shows that the variance in means between argument
groups is very slight when control variables are not accounted for.
Table 2: Means for Dependent Variables for Different Argument Groups (Maximum
Total N = 399)
Attitude on Attitude on Argument Effects on Effects on
Rights "Marriage" quality you Society*
CONTROL 10.17 17.42 3.33 3.04 1.89
MARCHANG 10.13 17.10 3.29 3.10 2.02
NOTRELIG 10.21 17.39 3.63 3.09 1.81
CONST 10.39 18.68 3.70 3.14 2.02
RIGHTS 9.83 16.95 3.61 3.10 1.81
STORY 10.60 17.54 3.81 3.23 1.93
Overall Mean 10.23 17.52 3.60 3.12 1.91
Scale Range 4 to 12 5 to 25 1 to 5 1 to 5 1 to 3
* I reversed the original direction of this variable's scoring for consistency with the other scales. For every
dependent variable presented here, higher scores indicate greater support, greater positive effects, or higher
Table 3, below, presents several different models for the first dependent variable,
the subject's attitude toward granting same-sex partners some of the legal rights and
responsibilities that come with marriage. The first regression, Model 1, predicts the
subject's attitude based on some theoretically salient independent variables. The second
regression predicts attitudes based on the argument groups alone. Model 3 combines the
first two models' predictors, and Model 4 goes a step further by adding in race as a
predictor. The final model, Model 5, tests the 4 interaction terms as well as all the other
predictors from prior models.
Table 3: Predictors of Attitudes toward Rights and Responsibilities of Marriage (n = 399)
STORY x F
Model 2 Model 3
Values reported are unstandardized/standardized coefficients.
* p < or =.10, **p< or =.05, ***p< or =.01.
Note 1: This dependent variable ranges in value from 3 to 12. The variable is a scale based on respondents'
indication that they "support," are "indifferent" to, or are "opposed" to the extension of each of the
following to same-sex couples: Social Security benefits, inheritance rights, health insurance benefits, and
child support responsibilities. The mean score was about 10.2, indicating a very high level of support for
these 4 things.
Note 2: FEMALE = Respondent's sex; SOCLIB = Social liberalism on 1 to 10 scale; ANYGAY = does
subject have any gay relatives or close friends; HOWRELIG = how religious from 1 to 10; GAYWRONG
= level of disagreement with the statement that 2 people of the same sex having sex is wrong (scale from 1
through 5); CHOICE = belief that homosexuality is chosen or biological (1 to 5 scale); INCOME = parents'
income category (1 to 5 scale) ; MARCHANG = Arg. that marriage has changed historically incorporating
a civil rights analogy; CONST = Arg. that Constitution has never been Amended to discriminate; NRELIG
= Arg. that religious definitions should not govern civil marriage; RIGHTS = list of rights same sex couples
would get only with marriage; STORY = story about a same sex couple and their kids and why they might
want marriage rights; STORY x F = interaction of sex with the story; RELIG x NRELIG = interaction of
religious strength with the argument that religious shouldn't govern civil marriages; RACExMCHANG =
interaction of race with civil rights analogy argument; ANYGAYxSTORY = interaction of the story with
knowing a gay person well. All dummy variables (female, the arguments, the race terms, and the
interaction terms) are coded such that the listed designation (e.g. female) is assigned the value 1.
Model 1 reveals that a handful of predictors can explain a moderate amount of the
variance in scores. The R-squared is .42. It is notable that in this model and all others,
the dummy variable designating whether the subject has a gay relative or close friend is
an insignificant predictor of his or her attitude on the topic (p = .60). Significant
predictors include female sex (p<.01), social liberalism (p <.01), religious strength (p =
.07), belief that homosexuality is wrong (p < .01), and belief that homosexuality is
chosen (p < .01). The strong predictors of attitudes on this topic are not surprising;
however, there is a fair amount of unexplained variance in the model. An R-squared of
about .4 would not be considered problematic for most social science models; however, it
seems a bit unusual here given that we might expect attitude toward the acceptability of
homosexuality, belief in the extent to which being gay is a choice, religious strength,
gender, the subject's close contact with a gay person, and SES to nearly-perfectly predict
attitude toward extending the rights and responsibilities of marriage to same-sex couples.
This finding will be discussed further below.
Model 2 reveals that the received arguments alone cannot explain expressed
attitudes on this topic. The R-squared for this model is virtually indistinguishable from
zero. However, adding in the original controls in Model 3, we see that the argument that
took the form of a story is a significant predictor of expressed attitudes (p=.01). The
standardized coefficient for this variable is .12, indicating that whether or not the subject
received the story has an effect on expressed attitude of about the same size as gender.
All the other arguments appear as insignificant, indicating that they are indistinguishable
from the control group in their effects. Model 4, which adds in dummy variables for
race, reveals that Native Americans are significantly more negative in their attitudes on
this dependent variable than other racial groups (p<.01) while Asian Americans status is a
borderline significant predictor of greater favorability (p = .10). (The small number of
Native American in the sample, however, means this finding should be treated with
skepticism.) The story retains its significance (p =.01). Finally, Model 5 tests for
interaction effects. The only significant interaction was that of African American status
with the argument that used an interracial marriage analogy (marginally significant with p
The second dependent variable was a scale measuring general favorability toward
extending "marriage" to same-sex couples. Table 4, below, presents 5 models testing the
predictors of this outcome. The models here are very similar to those for the first
dependent variable. Notably, there is a much lower level of unexplained variance than
for the first dependent variable, and the R-squareds for the four models that include
control variables are very high. In the first model, the significant predictors are identical
to those for the first dependent variable's Model 1. Sex (p < .01), social liberalism (p
<.01), religious strength (p =.02), belief that homosexuality is wrong ( p<.01), and belief
that homosexuality is chosen (p < .01) are all significant predictors. Once again, close
personal acquaintance with a gay person and parents' income level are insignificant in
their effects. Model 2, testing the arguments alone, shows no significant effects, but in
Models 3 and 4, the argument that featured a story yields borderline significant effects
(p=.07, p=.06), though these effects are smaller than those of most of the control
variables. No other arguments yield significant effects. Model 4, testing the effects of
race, yields somewhat different results than Model 4 for the first dependent variable. In
this model, Asians are slightly more positive toward same-sex marriages than whites. No
significant effects appear for African Americas, Native Americans, Hispanics, or subjects
of other races. None of the interaction terms tested are significant.
Table 4: Predictors of Attitudes toward "Same-Sex Marriages" (n = 375)
STORY x F
Model 2 Model 3
Values reported are unstandardized/standardized coefficients.
* p < or =.10, ** p < or =.05, ***p < or =.01.
The third dependent variable asked subjects to compare the quality of the argument
that they read to other arguments in favor of same-sex marriages that they might have
heard or read outside of the study. Table 5, below, presents the results for these models.1
1 The dummy variables for the argument groups are tested against the reference group consisting of those
subjects who received the global warming argument and selected a rating for their argument even though
they were instructed not to do so. The mean rating for students who received the warming argument and
rated their argument anyway was 3.33, indicating a slight positivity bias. The question this variable was
based on was asked of only about 60% of respondents.
None of the models are strong predictors of subject-rated argument quality, and the R-
squared for all models are strikingly low. In Model 1, gender is the only marginally
significant control predictor (p = .10) of "how good" an argument was perceived to be.
Model 2's test of the arguments by themselves reveals the story as borderline significant
(p=.06) and relatively large in its effect: on a 1 through 5 scale with little variance and
clustering around the mean, receiving the story was associated with nearly a .5 point
increase in argument quality rating. In Model 3 the story appears as the only borderline
significant predictor of quality ratings (p=.06), though in Model 4 its significance drops
(p=. 10) and Asian-American status appears as a marginally significant predictor (p=.09)
of less favorable argument quality evaluation. Model 5 reveals no significant
interactions. Overall, the models reveal that none of the variables tested were very strong
predictors of argument quality rating. However, whether or not the subject received the
story had a more significant effect on argument quality rating than any background
variables that might be expected to bias subjects' ratings.
The final dependent variable was the self-reported effect of participating in the
experiment. While most subjects reported that the experiment did not affect them, about
15% of subjects who answered this question indicated that it did affect their attitudes on
the topic. A very small number of subjects (8 out of 394, or just over 2% of those who
answered this question) appeared to "backlash" against the study, in that they stated that
they became more negative toward same-sex marriage rights. Many more subjects
reported that they became slightly (41 of 394, or 10.5%) or considerably (9 of 394, or
2.4%) more favorable toward same-sex marriages. While a plurality of those subjects
reporting positive effects were subjects who received the story, there were some subjects
in every group, including the control group, who reported greater favorability due to
participation. 7.4% of the control group subjects reported becoming more favorable
toward same-sex marriages; between 10% and 14% of each of the four "rational"
argument subject pools reported the same, and 20.2% of the subjects who received the
story became more favorable toward same-sex marriages.
Table 5: Compare the Argument You Received to Other Pro-SSM Arguments (n=247)
Model 1 Model 2 Model 3 Model 4 Model 5
Intercept 3.15*** 3.33*** 2.98*** 3.15*** 3.15***
FEMALE .20/.11* .15/.08 .14/.08 .20/.11
SOCLIB .03/.08 .03/.08 .05/.12 .04/.12
ANYGAY -.06/-.04 -.06/-.04 -.06/-.03 -.12/-.08
HOWRELIG .00/-.01 .00/-.01 -.01/-.03 -.01/-.04
GAYWRONG .02/.04 .02/.04 .00/.01 .01/.02
CHOICE .02/.02 .01/.01 .01/.02 .01/.01
INCOME .03/.03 .01/.01 -.01/-.02 -.01/-.01
MARCHANG -.04/-.02 -.01/-.01 -.09/-.04 -.15/-.07
CONST .30/.14 .31/.15 .24/.11 .22/.10
NRELIG .36/.17 .34/.16 .27/.12 .11/.05
RIGHTS .28/.13 .28/.13 .19/.09 .19/.09
STORY .48/.24* .50/.25* .44/.22* .49/.24
Hispanic -.24/-.08 -.23/-.08
African-American -.12/-.05 -.19/-.08
Asian -.36/-.11* -.33/-.10
Native American .00/.00 .02/.00
Other .58/.06 .62/.07
STORY x F -.22/-.10
R-squared/Adjusted .02/-.01 .04/.03 .07/.02 .08/.02 .10/.01
Values reported are unstandardized/standardized coefficients.
* p < or =.10, ** p < or =.05, **p < or =.01.
Models for the self-reported effects of the study appear in Table 6 (below). Model
1 resembles the first model for each of the previously discussed dependent variables, with
female sex and religious intensity appearing as the significant control variables (p = .02
for each), but with a weak R-squared value of .06. However, in Model 2, the story has a
significant effect (p=.02), even in the absence of any control variables, while other
arguments are indistinguishable from the control group. Model 3 tells much the same
story, with the story again appearing as significant (p=.02), and belief that homosexuality
is wrong showing up as marginally significant (p=.09). The standardized coefficient for
the story, .14, is larger than that of any other variable, indicating that the single most
important factor in subjects' evaluation of the experiment's effects on them was whether
or not they received the story. It is noteworthy that the effects of this variable are
stronger than any of the control variables that might tend to "bias" subjects' evaluations
of the study's effects on them. For example, whether or not a subject received the story
was almost twice as influential as gender in predicting self-reported effects of the study.
Model 4, adding in race variables, reveals that belief in the immorality of homosexuality
becomes clearly significant (p=.03) and the story becomes more significant (p=.01),
while subjects designating themselves as Native American and "other" have significantly
less positive evaluations of the experiment's effects than others (p=.02 and p<.01,
respectively). Finally, Model 5 shows that the interaction terms for African American
times the MARCHANG argument and for having an LGBT close friend or relative times
the story are at least borderline significant (p=.03 and p=.07). At least in terms of self-
reported effects, being African American increased receptivity to the argument that used
a racism/civil rights analogy (relative to other arguments), while having a close LGBT
contact increased receptivity to the story (relative to other arguments).
Table 6: Predictors of Self-Reported Effects of the Study on Subject (N=394)
STORY x F
Model 2 Model 3
Values reported are unstandardized/standardized coefficients.
* p < or =.10, ** p < or =.05, ***p < or =.01.
DISCUSSION AND CONCLUSIONS
The sample displayed a very high level of support for same sex marriage rights.
For the first dependent variable, the mean score on a scale ranging from 3 to 12 was 10.2,
indicating that subjects were very supportive of extending 3 rights and one responsibility
of marriage to same-sex partners. For the second dependent variable, which ranged from
5 to 25, the mean score was 17.5, indicating a fairly high level of support for same sex
marriage and civil union rights though not nearly as high as the level of support
indicated for the first dependent variable. This finding clarifies why there is a good deal
of unexplained variance for the first dependent variable but not the second. Even those
students who were relatively conservative often supported the 3 rights and one
responsibility that comprised the first dependent variable, whereas they were much less
likely to support "marriage."
This result suggests that subjects drew a distinction between (a) the specific rights
and responsibilities that go along with marriage and (b) "marriage." Although there are
numerous rights that go along with marriage but not with state-level civil unions (such as
the right to visit a partner in the hospital outside one's home state), subjects did not
appear attuned to this fact. Therefore, many subjects were willing to extend the rights
that go with marriage to same sex couples, but not willing to endorse "same-sex
marriage." This finding is consistent with Loftus' (2001) and Yang's (1997) findings of
multidimensional attitudes, and suggests that some subjects may oppose "same-sex
marriage" because the phrase connotes to them some endorsement of homosexuality, yet
support the legal aspects of marriage because they oppose discrimination.
It is interesting that the sample clearly supported same-sex marriage rights and
leaned toward support for "same-sex marriages" at the same time that it showed a slightly
negative attitude toward the effects same-sex marriages would have on society (in the
"effects on society" question). The subjects who were opposed to same-sex marriages
were extremely likely to say such marriages would have a negative effect on society,
while the supporters of same-sex marriages were divided between believing such
marriages would improve society and would have no effect on it. For example, of the 77
subjects who indicated the least support for same-sex marriages on the 5-question scale
(those whose score on the 25-point scale was between 5 and 10), 67 subjects indicated
allowing same-sex marriages would make society worse, 9 chose "don't know" or did not
answer, and 1 indicated "no effect." Of the 80 most supportive subjects (all of whom
scored 25 out of 25), 51 chose "make society better," 23 chose "no effect," and 6 chose
"don't know" or did not answer. It appears that even many supporters of same-sex
marriages see their benefits as accruing primarily or even exclusively to the parties
directly involved rather than to society at large.
Many of the predictors that have been shown to be significant in predicting
attitudes toward the moral acceptability of homosexuality were shown to be significant in
predicting attitudes toward same-sex marriage rights. Gender, social liberalism, religious
strength, belief that homosexuality is wrong, and belief that homosexuality is chosen
were significant in their effects. Other variables were shown to be insignificant. None of
the interactions tested were consistently significant across the dependent variables; this
may be due to the lack of interaction or to confounding interactive effects that, when
aggregated, canceled themselves out. The interaction of African-American status with
the argument using an analogy to interracial marriage was borderline significant (p=.09)
in predicting disproportionately strong support for the rights of marriage and significant
(p=.03) in predicting disproportionately positive self-reported effects of the study. A
close relationship with an LGBT person was borderline significant (p=.07) in predicting
disproportionately positive self-reported effects of the study for those subjects who
received the story. The other interactions (of gender with the story and of religious
strength with the argument about religion) were not significant in any models. Thus,
there are mixed results for H4 and H7, while H5 and H6 are rejected. Further, models
tested but not shown in this paper revealed that including a dummy variable for
community college status did not add to the predictive power of the models.1 There was
also no apparent independent effect of parents' income level on attitudes. Finally, and
surprisingly, having a gay close friend or relative did not have any effect on subjects'
attitudes. This finding is contradictory to much previous research on the effects of
contact with a gay or lesbian person on attitudes toward LGBTs. It may be that even
those subjects with a close LGBT friend have not thought about the actual benefits of
marriage rights to that friend. The story tested was not "just" a story; the fact that it had
clear policy implications may have catalyzed thinking about the benefits of marriage
rights in a way that simply having a LGBT friend does not.
1 Despite the fact that community college students were dramatically more conservative in their attitudes
than university students, this appears to be accounted for by the social liberalism variable. That is,
community college students are less favorable toward same-sex marriages because they are less socially
liberal in general than university students.
The most interesting finding of the study is that the argument consisting of a story
had significant effects on some dependent variables, while other, more purely "rational"
arguments, did not. Thus, H1 (that each of the arguments would have an effect) is
rejected, H2 (that the story would be the most effective) is largely confirmed, and H3
(that the list of rights would have the greatest effect) is rejected. This supports prior
research suggesting that emotional appeals are more likely to succeed, even when they
concern policy issues, than rational appeals. Of course, the distinction between these two
types of appeals is not as clean as it might appear at first, and no argument is purely
"rational" or "emotional." However, this finding adds to our understanding of persuasion
by suggesting that affective appeals can do more than just humanize disliked groups;
affective appeals outperform more cognitive kinds of appeals in building support for
policies that benefit stigmatized groups. It is particularly interesting that in this sample,
having an LGBT close friend or relative was not significantly associated with attitudes
toward same-sex marriage, but a personalized story about a lesbian couple was. This
may be because unless the "relative or close friend" was particularly interested in getting
married, the topic of same-sex marriage may never have come up in conversation, and
because marriage is not an immediately salient topic to many college students. Further, it
is interesting that the story was clearly effective in building support for extending same-
sex couples the rights of marriage, but only borderline significant in building support for
"same-sex marriage." This may be due to the fact that the story dealt with reasons that
the rights of marriage were important, but did not address the significance of the term
What can we make of the fact that subjects were markedly more supportive of
extending several of the rights of marriage to same-sex couples than of "same-sex
marriage" at the same time that the subjects who read the argument describing the rights
that go along with marriage did not display higher levels of support for same-sex
marriage than other subjects? This finding, which initially appears counterintuitive and
troubling, is in fact quite revealing. The word "marriage" itself apparently carries a
cultural meaning that is quite difficult to challenge using rational-abstract appeals.2 It is
noteworthy that the argument listing the rights of marriage and the argument that civil
marriage is different than religious marriage both failed to elicit greater favorability
toward same-sex marriages.3 Accepting "same-sex marriage" may be difficult for some
heterosexuals otherwise supportive of LGBT rights because even when they read about
the pragmatic aspects of it (or about the fact that civil marriage is distinct from religious
marriage), they still see legal same-sex marriage as an endorsement of homosexuality.
Thus, framing same-sex marriage as a "rights issue" is not enough if one uses the
culturally laden term "marriage." Rather, it appears necessary to rely on a partly
narrative appeal to overcome the traditionalism surrounding the word "marriage."
It is likely that this study underestimates the gap in persuasiveness between
primarily abstract and primarily narrative arguments. The sample consists primarily of
2 For this reason, columnist Leonard Pitts (2005) has recently argued that supporters of same-sex marriage
should simply use the phrase "civil unions." This ignores the fact that civil unions are state-level policies
while marriage confers both state and federal level benefits; however, it is in theory possible to redefine
"civil unions" such that they encompass federal-level legislation. While supporters of such a strategy likely
find it pragmatic, it is susceptible to charges that it accepts second-class cultural status for LGBTs.
3 It may be that those who received the "rights" argument worried that same-sex marriages still conveyed
cultural or religious endorsement of homosexuality, while those who received the "NOTRELIG" argument
did not realize that same-sex marriages have specific pragmatic benefits. Perhaps combining these two
appeals would be more effective.
people who have unusually strong educational credentials. These subjects likely have, on
the whole, a far greater familiarity with reading complicated materials that lack direct
emotional punch than the population at large, and should therefore be more susceptible to
abstract-rational kinds of persuasion. The fact that the appeal with the greatest emotional
punch was more successful than other appeals even in this sample suggests that it might
be dramatically more effective in the population at large. Sears' (1986) classic article
discussing the pitfalls of experimental research using college students as subjects warns
that college students generally possess "unusual cognitive skills," and maintains that
Customary procedures in laboratory studies should produce a strongly cognitive
set. ... A college student in a testlike situation [such as in experimental research]
knows not to respond with simple evaluative preferences; rather, what is called for
is paying close attention, dispassionate judgment, a search for the "right" answer,
critical thinking, and close attention. Social psychology's use of relatively well-
educated subjects, selected for their superior cognitive skills, along with research
sites, procedures, and tasks that promote dispassionate, academic information-
processing, should help produce empirical evidence that portrays humans as
dominated by cognitive processes, rather than by strong evaluative predispositions.
(Sears 1986: 524)
This pattern did not appear in this study. The failure of primarily abstract appeals, and
the success of a primarily narrative appeal, in a subject pool that is heavily cognitively-
oriented, indicates that narrative appeals may be even more effective in the general
public. It also suggests that the story might more clearly affect support for "marriage"
(and not just for the rights of marriage) in the population as a whole than in this subject
4 It might be argued that some control group participants' claim of becoming more favorable toward same-
sex marriages and some control group participants' rating the argument they received are two pieces of
evidence supporting Sears' warning that college student subjects behave how they think they are supposed
to behave. However, the evidence that this happened in this study is scant. It seems much more likely that
reading questions about same-sex marriage caused students to think about it even if they did not read an
argument about it. Further, even if the critique is true, the comparison of different argument groups to the
control group reduces the significance of this critique to virtually nil.
On a related note, it may well be important that the story contained three children
as characters, and discussed the way that the lesbian couple's children were harmed by
the fact that the couple was not allowed to marry. Several open-ended questions asked
subjects for the "best argument they had heard in favor of' and the "best argument they
had heard against" same-sex marriages. Many subjects, whether or not they had received
the story as their argument, brought up kids in their answers, suggesting that children
occupy an especially important place in discussions of same-sex marriages. Best (1990)
has suggested that children are particularly sympathetic characters in narratives of any
kind, and that "atrocity tales" involving children are an excellent means of generating
concern and action. The story may have been successful at raising concern about what
would happen to the children in a world without same-sex marriage rights, in a way
similar to the way that atrocity tales generate concern.
The implication that part of the success of the story was due to the fact that it
successfully invoked concern for children is noteworthy for several reasons. First, Craig
et al. (2005), in a study of heterosexual Floridians' ambivalence on LGBT rights issues,
found that attitudes clustered together around two different sets of considerations: one set
of considerations considers LGBT adult roles, and the other concerns interactions with
children. In Craig et al's study, Floridians were substantially more supportive of
allowing LGBTs to serve in the military, have privacy in their own homes, have
protections against employment discrimination, and have family health insurance rights
than they were of allowing LGBTs to teach school, marry, adopt children, and join
groups like the Boy Scouts. Further, in the open-ended section, many subjects referenced
the supposed effects of same-sex marriages on children usually as a reason same-sex
marriages should not occur. These two findings together indicate that the proverbial
cultural deck is likely stacked against using children as part of an argument in favor of
same-sex marriage rights. Even though the story therefore called to mind a frame (effects
on children) that might have been expected to negatively impact attitudes toward same-
sex marriages, it appears that the direction of the story's argument (that same-sex
marriage rights would benefit children) was more important than the fact that its frame
has generally been used to argue against same-sex marriage rights. While this is
consistent with Brewer (2002)'s finding in an experimental study of LGBT rights frames
that the same frame (in his study, an "equality" frame or a "morality" frame) can be used
to support policies that go in opposite directions, it is inconsistent with his finding that his
subjects' "exposure to the frames encouraged participants to use value language not only
in ways suggested by the frames but also in ways that challenged the frames" (303).
Brewer writes that
Although exposure to a value frame may encourage some citizens to use value
language in a way that is connected with the frame's interpretation of the
relationship between the issue and the value, it may encourage other citizens to use
value language in a way that challenges this interpretation. Gamson's (1992) focus
group research suggests that citizens can draw upon popular wisdom, "counter-
frames" that criticize dominant media frames, and their own reasoning skills when
they encounter frames contained within media coverage. If this is so, then citizens
who encounter value frames may borrow the "value words" in those frames to
make their own points about an issue. (305)
It is striking that exposure to the "children" frame generally did not lead
respondents to oppose same-sex marriage on the grounds that it would be bad for
5 This finding is especially noteworthy given that conservative subjects could have taken the logically
coherent position that same-sex couples should not be allowed to marry and should not be allowed to
adopt, thus overcoming the argument that allowing same-sex couples to adopt but not marry is bad for their
children. It may be that hearing effects on children invoked to support same-sex marriage was new and
These findings suggest three promising routes for future work on this topic. First,
do abstract arguments on this topic that discuss children have a greater level of success
than other kinds of abstract arguments? It may be the case that children are so privileged
in this discourse that their presence in arguments is enough to overcome the
disadvantages that cognitive-rational arguments generally face. Secondly, what types of
narrative appeals are the most persuasive? Would the story have succeeded if it did not
have children, who are probably highly sympathetic characters, in it? What if the story
had been about a gay male couple, rather than two women? Following this line of
investigation might be useful not merely for understanding persuasion on same-sex
marriage, but also for developing a more finely honed understanding of how exactly
narrative appeals appear to be more persuasive and successful than more abstract appeals.
Finally, in an experiment similar to Brewer's (2002), how would subjects respond to pro-
and anti- same-sex marriage frames that deployed children as projected beneficiaries or
victims of same-sex marriages? Such an investigation would likely offer useful insights
into whether and how using frames in unexpected, novel ways can overcome the
tendency of some frames to be associated with particular "directions."
novel and subjects were persuaded by this position in part because they had never thought about counter-
arguments against it, as a great deal of evidence suggests that when it comes to resisting persuasion,
forewarned is forearmed (e.g. McGuire and Papageorgis 1962; Anderson and McGuire 1965, Billeaudeaux
et al 2005). The finding that the children frame did not provoke a backlash should not be interpreted to
mean that no subjects who received this argument wrote about negative effects of same-sex marriages on
children in the open-ended section. Rather, it means simply that the frame induced greater favorability
toward same-sex marriages; this favorability apparently outweighed any tendency of the frame to evoke
fear of GLBT parents "corrupting" their children.
THE SUCCESSFUL ARGUMENT
Please read the following text of a story in support of allowing same-sex marriages (from
www.hrc.org). When you are done reading, continue on to Part 3.
A story of two same-sex partners
Jo and Teresa live in Maryland. They've been in a loving, committed relationship for 19
years and have three children Jake, 12; Matthew, 9; and Bena, 2. Between skinned
knees, soccer practices and never enough time in the day, they face all the same joys and
frustrations as other parents but without the same protections.
Because the government won't give them legal protections, Jo and Teresa's children don't
qualify for full Social Security survivors' benefits if one of them dies, even after a
lifetime of paying taxes. And if one of the kids gets sick, in some states they could even
be denied the right to visit them in the hospital because they aren't "family." And Jo and
Teresa aren't eligible for COBRA health coverage for each other, or for family medical
leave to care for a sick loved one. If Teresa, who is not legally considered the children's
parent, should pass away, whatever she leaves to the children will be taxed more than it
would be if she were married to Jo. And if Teresa and Jo split up, Teresa will have no
legal obligation to pay child support to the children-regardless of how much money she
Why talk about this now? Because extreme politicians want to amend the U.S.
Constitution to deny any legal protections for gay partners and their families. Meanwhile,
gay and lesbian couples in long-term, committed relationships aren't eligible for
government-issued civil marriage licenses and the legal protections they provide.
Although no government should ever tell religious institutions who they can marry, the
government should not discriminate in providing civil marriage licenses to any devoted
Jo and Teresa deserve the same protections other families have and the ability to raise
their children without discrimination.
LIST OF REFERENCES
Anderson, L. R. and McGuire, W. J. 1965. "Prior Reassurance of Group Consensus as a
Factor in Producing Resistance to Persuasion." Sociometry 28: 44-56.
Altmeyer, B. 2001. "Changes in Attitudes toward Homosexuals." Journal of
Homosexuality 42 (2): 63-75.
Billeaudeaux, A., Domke, D., Hutcheson, J., and Garland, P. 2005. "The Bush
Administration, Inoculation Strategies, and the Selling of a 'War.'"
03/pdf files/InnoculationPaper.pdf, last accessed May 2005.
Bolce, L., De Maio, G., and Muzzio, D. "The 1992 Republican 'Tent': No Blacks
Walked In." Political Science Quarterly 108 (2): 255-270.
Bowman, K., and O'Keefe, B. 2004. "Attitudes about Homosexuality and Gay
Marriage." American Enterprise Institute: Studies in Public Opinion.
www.aei.org, last accessed April 2005.
Brader, T. 2005. "Striking a Responsive Chord: How Political Ads Motivate and
Persuade Voters by Appealing to Emotions." American Political Science Review
Brewer, P. 2002. "Framing, Value Words, and Citizens' Explanations of their Issue
Opinions." Political Communication 19: 303-316.
Brounk, T. 1996. Changing Negative Attitudes toward Gay and Lesbian People: the
Impact of Cognition versus Affect. Ph.D. Thesis, Ohio State University- Columbus.
Cobb, M., and Kuklinksi, J. 1997. "Changing Minds: Political Arguments and Political
Persuasion." American Journal of Political Science 41 (1): 88-121.
Craig, S., Martinez, M., Kane, J., and Gainous, J. 2005. "Core Values, Value Conflict,
and Citizens' Ambivalence about Gay Rights." Political Research Quarterly 55: 5-
Cullen, J., Wright, L., and Alessandri, M. 2002. "The Personality Openness Variable as
It Relates to Homophobia." Journal ofHomosexuality 42 (4): 119-134).
Edwards, K. 1990. "The Interplay of Affect and Cognition in Attitude Formation and
Change." Journal of Personality and Social Psychology 59: 202-216
Edwards, K., and W. von Hippel. 1995. "Hearts and Minds: The Priority of Affective
versus Cognitive Factors in Person Perception." Personality and Social
Psychology Bulletin 21: 996-1011.
Foumier, P., Martin, P., and Nadeau, R. 2002. "Who Can be Talked Out of Their
Positions and How? Persuasion and High-Stakes Political Choices." Paper
presented at the Canadian Political Science Associations 2002 meetings. Available
online at http://www.fas.umontreal.ca/POL/Martinp/FMNCPSAO2.pdf, last
accessed April 2005.
Gamson, W. 1992. Talking Politics. Boston: Cambridge University Press.
Gilbert, D. 1991. "How Mental Systems Believe." American Psychologist 46: 106-119.
Goffman, E. 1974. Frame Analysis: An Essay on the Organization ofExperience.
Boston: Northeastern University Press.
Gilligan, C. 1993. In a Different Voice: Psychological Theory and Women's
Development. Cambridge: Harvard University.
Goldstein, R. 2003. "The Radical Case for Gay Marriage: Why Progressives Must Join
This Fight." Village Voice, September 3-9, 2003.
http://www.villagevoice.com/news/0336,goldstein,46732,1.html, last accessed
Greenpeace. 2005. "Countering the Skeptics."
http://archive.greenpeace.org/climate/industry/reports/sceptics.html, last accessed
Herek, G. 1988. "Heterosexuals' Attitudes toward Lesbians and Gay Men: Correlates
and Gender Differences." Journal of Sex Research 25 (4): 451-477.
Herek, G. 2000. "The Psychology of Sexual Prejudice. Current Directions in
Psychological Science 9: 19-22.
Herek, G. 2002a. "Gender Gaps in Public Opinion about Lesbians and Gay Men."
Public Opinion Quarterly 66 (1): 40-66.
Herek, G. 2002b. "Heterosexuals' Attitudes toward Bisexual men and Women in the
United States." Journal of Sex Research 39 (4): 264-274.
Herek, G., and Capitano, J. 1999. "Sex Differences in How Heterosexuals think about
Lesbians and Gay Men: Evidence from Survey Context Effects. Journal of Sex
Research 36 (4): 348-360.
Hovland, C., Janis, I., and Kelley, H. 1953. Communication and Persuasion:
Psychological Studies of Opinion Change. New Haven, CT: Yale University Press.
Kahnemann, D., and Tversky, A. 1979. "Prospect Theory: An Analysis of Decision
Under Risk." Econometrics 47: 263-291.
Kahnemann, D., and Tversky, A. 1984. "Choices, Values, and Frames." American
Psychologist 39: 341-350.
Lakoff, G. 2004. Don't Think of an Elephant: Know Your Values and Frame the
Debate The Essential Guide for Progressives. New York: Chelsea Green.
Lewis, G. 2003. "Black-White Differences in Attitudes toward Homosexuality and Gay
Rights." Public Opinion Quarterly 67 (1): 59-78.
Loftus, J. 2001. "America's Liberalization in Attitudes toward Homosexuality, 1973 to
1998." American Sociological Review 66 (5): 762-782.
McGuire, W. J. and Papageorgis, D. 1962. "Effectiveness of Forewarning in Developing
Resistance to Persuasion." Public Opinion Quarterly 26: 24-34.
McKissack, F. 2004. "Bush, Blacks and Values." Fort Wayne Journal-Gazette.
Nelson, T., Clawson, R., and Oxley, Z. 1997. "Media Framing of a Civil Liberties
Conflict and Its Effect on Tolerance." American Political Science Review 91: 567-
Nelson, T., and Kinder, D. 1996. "Issue Framing and Group-Centrism in American
Public Opinion." Journal ofPolitics 58: 1055-1078.
Perloff, R. 2003. The Dynamics of Persuasion: Communications and Attitudes in the
21t Century. New York: Lawrence Erlbaum and Associates.
Pitts, L. 2005. "Stop Calling It Gay Marriage." Detroit Free-Press, June 10.
www.freep.com/voices/columnists/pittsl0e_20050610.htm, last accessed June
Pratkanis, A., and Aronson, E. Age ofPropaganda: The Everyday Use and Abuse of
Persuasion. New York: Henry Holt.
Snow, D., and Benford, R. 1992. "Master Frames and Cycles of Protest." Pages 133-
155 in A.D. Morris and C.M. Mueller, Frontiers in Social Movement Theory. New
Haven: Yale University Press.
Stangor, C ., Sullivan, A., & Ford, T. 1991. "Affective and Cognitive Determinants of
Prejudice." Social Cognition 9: 359-380.
Weithoff, C. 2002. "Naming, Blaming, and Claiming in Public Disputes: The 1998
Maine Referendum on Civil Rights Protection for Gay Men and Lesbians."
Journal ofHomosexuality 44: 61-82.
Yang, A. 1997. "Trends: Attitudes toward Homosexuality." The Public Opinion
Quarterly 61 (3): 477-507.
Raj Ghoshal is the most important rising academic of the 21st Century. He
completed a B.A. in political science in May 2001 at the New College, where his senior
thesis examined "threat construction" and mass media depictions of Osama bin Laden.
His M.A. thesis in sociology from the University of Florida (2005) examines persuasion
in relation to attitudes about same-sex marriages. In his free time he enjoys chicken soda.