Adult Romantic Relationships as Contexts of Human Development: A Multimethod Comparison of Same-Sex Couples With Opposite-Sex Dating, Engaged, and Married Dyads

Developmental Psychology
Copyright 2008 by the American Psychological Association
2008, Vol. 44, No. 1, 91–101
0012-1649/08/$12.00
DOI: 10.1037/0012-1649.44.1.91
Adult Romantic Relationships as Contexts of Human Development: A
Multimethod Comparison of Same-Sex Couples With Opposite-Sex Dating,
Engaged, and Married Dyads
Glenn I. Roisman, Eric Clausell, Ashley Holland, Keren Fortuna, and Chryle Elieff
University of Illinois at Urbana-Champaign
This article presents a multimethod, multi-informant comparison of community samples of committed
gay male (n
30) and lesbian (n
30) couples with both committed (n
50 young engaged and n
40 older married) and noncommitted (n
109 exclusively dating) heterosexual pairs. Specifically, in this
study the quality of same- and opposite-sex relationships was examined at multiple levels of analysis via
self-reports and partner reports, laboratory observations, and measures of physiological reactivity during
dyadic interactions. Additionally, individuals in same-sex, engaged, and marital relationships were
compared with one another on adult attachment security as assessed through the coherence of partici-
pants’ narratives about their childhood experiences. Results indicated that individuals in committed
same-sex relationships were generally not distinguishable from their committed heterosexual counter-
parts, with one exception—lesbians were especially effective at working together harmoniously in
laboratory observations.
Keywords: same-sex couples, married couples, dating couples, engaged couples, observational study
Committed romantic relationships (and marriages in particular)
mature, or malevolent contexts of development is a question that
represent fundamental contexts of human development, not only
can be addressed empirically. Toward that end, this study joins a
for children but also for the adults involved in them (e.g., Clarke-
relatively small set of comparative, observational studies that have
Stewart & Dunn, 2006; Cummings & Davies, 1994). Nonetheless,
examined the quality of same- and opposite-sex relationships at
there is much debate as to whether committed same-sex relation-
multiple levels of analysis.
ships are inherently inferior to heterosexual unions in terms of
Typically, developmental research related to the claim that
promoting healthy psychological adjustment. It is important to
committed same-sex relationships are unlike heterosexual unions
emphasize that although few social scientists espouse this perspec-
has focused on comparisons of the children of gay, lesbian, and
tive, this is a common lay view that is likely in part attributable to
heterosexual parents. Findings in this area of research have been
the fact that, until relatively recently, homosexuality was viewed
highly consistent. To date, this literature suggests that children
as a form of psychopathology in the Diagnostic and Statistical
reared by individuals with same- versus opposite-sex sexual pref-
Manual of Mental Disorders (DSM) and the International Classi-
erences are generally not distinguishable (see Allen & Burrell,
fication of Diseases (ICD) used by mental health professionals (for
1996; Stacey & Biblarz, 2001), even when such studies have been
a review see Mendelson, 2003). Importantly, however, the as-
based on nationally representative data sets (Wainright, Russell, &
sumption that same-sex unions are atypical, psychologically im-
Patterson, 2004). A more fundamental question, especially given
how few gay and lesbian couples actually have children (Gates &
Ost, 2004), is whether committed same-sex couples differ from
engaged and married couples in ways that suggest that the devel-
Glenn I. Roisman, Eric Clausell, Ashley Holland, Keren Fortuna, and
opment of adults who are involved in same-sex relationships is
Chryle Elieff, Department of Psychology, University of Illinois at Urbana-
undermined. Although there is a long history of research on the
Champaign.
psychology of gay and lesbian individuals, there is in fact only a
This study was sponsored by a Wayne F. Placek Award from the
modest literature specifically devoted to adult gay and lesbian
American Psychological Foundation and several grants (including a Beck-
man Award) from the Research Board at the University of Illinois at
relationships, and smaller still are the number of investigations that
Urbana-Champaign. The dating couples study was sponsored by a National
provide direct comparisons of same- and opposite-sex couples.
Science Foundation Grant to R. Chris Fraley. The authors gratefully
Nonetheless, a few pioneers, including Blumstein and Schwartz
acknowledge this financial support as well as Chao-Hsun (Ray) Hu’s
(1983) and Kurdek (e.g., 1986, 1995a, 1996, 1998, 2004) have
efforts related to the coding of observational data presented in this report.
conducted comparative studies focused on same- and opposite-sex
This article is dedicated to the memory of Nicole Laws-Carroll, an honors
romantic relationships in adulthood using interview and question-
student whose contributions to our laboratory and our lives will not soon be
naire methods. In a review of findings drawn from gay and lesbian
forgotten.
couples, Kurdek (1995b) identified the following six key princi-
Correspondence concerning this article should be addressed to Glenn I.
ples that govern same-sex relationships: (a) Many gay men and
Roisman, Department of Psychology, University of Illinois at Urbana-
Champaign, 603 East Daniel Street, Champaign, IL 61820. E-mail:
lesbians identify themselves as being involved in a committed
roisman@uiuc.edu
relationship; (b) One salient difference between gay and lesbian
91

---

92
ROISMAN, CLAUSELL, HOLLAND, FORTUNA, AND ELIEFF
relationships is that lesbians tend to be more sexually exclusive
analysis via self-reports and partner reports of adjustment, labora-
than gay men; (c) Gay men, and particularly lesbians, are more
tory observations of interpersonal behavior, and measures of phys-
likely to endorse an “ethic of equality” compared with their het-
iological reactivity during dyadic interactions.
erosexual counterparts; (d) As with heterosexual relationships, gay
In our studies we routinely acquire measures of participants’
and lesbian partnerships show reliable changes over time; (e) Few
autonomic responses (i.e., skin conductance and heart rate) during
differences emerge when comparing global levels of satisfaction
their interactions with romantic partners not only due to empirical
for gay, lesbian, and heterosexual couples; and (f) Predictors of
evidence (cited earlier) that these forms of physiological activation
relationship stability and satisfaction are consistent across gay,
are intimately tied to interpersonal functioning (e.g., Gottman, &
lesbian, and heterosexual couples.
Levenson, 1992; Levenson & Gottman, 1985; Tsai & Levenson,
Such longitudinal research by Kurdek (e.g., 2004) on same-
1997) but also in light of Fowles’ (1980, 1988) claim for the
versus opposite-sex couples has thus led many social scientists to
existence of two antagonistic motivational systems whose activa-
conclude that gay, lesbian, and heterosexual relationships are quite
tion can be inferred from unique physiological correlates. Specif-
similar to one another in many respects (Herek, 2006). Nonethe-
ically, Fowles (1980) argued that engagement of the Behavioral
less, it is important to point out that research in this area has not yet
Inhibition System, which has often been characterized as being
fully tapped into the methodological sophistication that now typ-
involved in the effort to inhibit behavior, is reflected in increases
ifies cutting-edge interpersonal relationships research. For exam-
in electrodermal activity, whereas the Behavioral Activation Sys-
ple, it is now well-established that the observed quality of adults’
tem, which is involved in approach-related (hyper)activation, is
relationships as well as measures of cardiac (e.g., heart rate) and
signaled by heart rate reactivity. Using this theoretical framework,
electrodermal (e.g., skin conductance) reactivity during interper-
we have recently shed light on the possible developmental foun-
sonal interactions have proved to be reliable predictors of relation-
dations of the patterns of autonomic response so routinely ob-
ship satisfaction and dissolution (high levels of reactivity tend to
served among distressed couples in research on the psychophysi-
be associated with negative interpersonal outcomes; Gottman &
ology of marriage and divorce (Gottman & Levenson, 1988) in that
Levenson, 1992; Levenson & Gottman, 1985; Roisman, 2007; Tsai
adults insecure with respect to their early experiences with care-
& Levenson, 1997). Such measures provide objective assessments
givers appear to show the largest increases in heart rate and skin
of the quality of adults’ relationships as developmental contexts
conductance during their marital and premarital interactions rela-
that go significantly beyond simply asking adults to describe
tive to resting conditions (Roisman, 2007).
themselves and their relationships (see Gottman, Levenson, Gross,
Finally, in addition to assessing relationship quality at multiple
et al., 2003; Gottman, Levenson, Swanson, et al., 2003).
levels of analysis, in this study we also compared individuals in
In fact, surprisingly few comparative studies have been pub-
engaged, married, and committed same-sex relationships to one
lished providing observational data relevant to committed gay,
another on adult attachment security, as assessed through the
lesbian, and heterosexual relationships. In two exceptions, Gott-
coherence of these adults’ narratives about their childhood expe-
man, Levenson, Swanson, et al. (2003) found that same-sex cou-
riences. Our laboratory has consistently assessed adult attachment
ples actually demonstrated more positive (and fewer negative)
security using the resource-intensive, semistructured Adult Attach-
behaviors during their interactions than did married couples,
ment Interview (AAI; George, Kaplan, & Main, 1985) because of
whereas Julien, Chartrand, Simard, Bouthillier, and Be´gin (2003)
(a) an interest in examining how adults’ represented experiences
found no evidence for differences among committed gay male,
are reflected in their adult relationships, (b) empirical evidence that
lesbian, and married couples in terms of the observed quality of
secure adults are more likely to have higher quality romantic
their interactions. Although such findings run counter to claims
relationships than their insecure counterparts (Babcock, Jacobson,
that same-sex unions are typically of relatively low quality, sam-
Gottman, & Yerington, 2000; Bouthillier, Julien, Dube, Belanger,
ples from these studies were modest in size and thus this ground-
& Hamelin, 2002; Cohn, Silver, Cowan, Cowan, & Pearson, 1992;
breaking work awaits replication. In addition, both Gottman, Lev-
Creasey, 2002; Crowell et al., 2002; Paley, Cox, Burchinal, &
enson, Swanson, et al. (2003) and Julien et al. (2003) contrasted
Payne, 1999; Roisman, Madsen, Hennighausen, Sroufe, & Collins,
gay and lesbian couples with married dyads only. As such, it is
2001; Wampler, Riggs, & Kimball, 2004), and (c) our view that
presently unclear how same-sex couples compare with a broader
developing a coherent narrative about childhood experiences is a
range of committed (i.e., engaged) and noncommitted (i.e., dating)
salient life task of adulthood (Roisman, 2006). It is important to
heterosexual couple types.
note that adult attachment security also represents the single stron-
In order to expand the comparative literature on same- versus
gest predictor of attachment security in the next generation (van
opposite-sex relationships, this report presents a multimethod,
IJzendoorn, 1995). Given the challenges that gay and lesbian
multi-informant comparison of samples of committed gay male
adults face in “coming out” to family members as well as other
(n
30) and lesbian (n
30) couples with a range of committed
forms of adversity in their families of origin (Savin-Williams,
(n
50 young engaged and n
40 older married) and noncom-
2001; Weston, 1992), we wanted to determine whether gay men
mitted (n
109 exclusively dating) heterosexual pairs, all drawn
and lesbians were at any disadvantage in this developmentally
from the same community using identical recruitment methods
salient domain of psychological adaptation. For example, evidence
(this is an important design point because, in Gottman, Levenson,
that individuals in same-sex relationships are more likely to pro-
Swanson, et al., 2003, same- and opposite-sex dyads were drawn
duce incoherent narratives about early life experiences could sug-
from different communities, which may in part explain the differ-
gest that, relative to their heterosexual counterparts, sexual minor-
ences they identified between same- vs. opposite-sex couples).
ities may face serious challenges in terms of successfully engaging
More specifically, in the present study the quality of same- and
adult relationships. To date, however, there are no published data
opposite-sex relationships was examined at multiple levels of
on the AAI narratives of gay and lesbian adults.

---

SPECIAL SECTION: SAME-SEX AND OPPOSITE-SEX COUPLES
93
In sum, consistent with emerging evidence from developmen-
set of relationship types drawn from the same small Midwestern
tal psychology that adults’ romantic relationships represent key
community. Specifically, we have conducted research involving
developmental contexts of adulthood (e.g., Roisman, Masten,
109 dating couples (all heterosexual and dating one another ex-
Coatsworth, & Tellegen, 2004) and motivated by societal ques-
clusively but with no expectation of long-term commitment, ages
tions about whether same-sex relationships undermine adult
18 to 25 years), 50 younger engaged couples (ages 18 to 30 years
development, this article presents a multimethod comparison of
involved in their first engagement; Roisman, 2007; Roisman et al.,
committed gay male and lesbian couples with well-defined
2007), 40 older married couples (age 50 or older and married for
groups of heterosexual committed (i.e., young engaged and
at least 15 years; Roisman, 2007), and 60 same-sex dyads (30 gay
older married) and noncommitted (i.e., young dating) couples
male and 30 lesbian couples; age 18 or older, mutually committed,
drawn from the same community. Using state-of-the-art meth-
and coupled for at least a year). In all of our dyadic studies,
ods drawn from interpersonal relationships research, our goal
participants received $25 each ($50 per couple) and were recruited
was to provide rare empirical data relevant to the developmental
from the same community in an identical manner using on- and
resources available within a broad range of couple types. Al-
off-campus flyers and Listservs.
though we expected to identify relatively few differences be-
Please note that this work represents a secondary analysis of a
tween committed same- and opposite sex dyads, because of
set of studies conducted in our laboratory. Generally, we have been
previous evidence that couples become more effective at col-
interested in examining how developmental experiences and other
laborating as they mature (Carstensen, 1992, 1993; Carstensen,
individual differences contribute to interpersonal functioning
Gottman, & Levenson, 1995) and those that do not are more
within distinct kinds of relationships (e.g., among married, en-
likely to divorce (Gottman, & Levenson, 1992), we did antic-
gaged, and dating couples). We began our work on same-sex
ipate that committed dyads (i.e., gay male, lesbian, engaged,
couples, however, largely to inform debate about how such rela-
and married) would have higher quality interactions compared
tionship types might be similar or distinct from heterosexual
with noncommitted (i.e., heterosexual dating) couples. In addi-
relationships. In order to be able to compare same-sex couples with
tion, we attempted to replicate findings from Gottman, Leven-
participants from other committed relationship types widely rec-
son, Swanson, et al. (2003) indicating that same-sex couples
ognized by legal and/or social mechanisms not currently available
interact more harmoniously during conflict interaction tasks
to same-sex couples in most parts of the United States (i.e.,
than do committed heterosexuals as measured by objective
engagement, marriage), we decided to focus on same-sex dyads in
coders. We had no reason to expect differences among com-
which both participants (a) had been romantically involved with
mitted couples (same- and opposite-sex) in terms of mean levels
one another for a year or more and (b) agreed mutually that they
of relationship satisfaction or autonomic reactivity during their
were in a committed relationship (although not used as a criterion
dyadic interactions, other than those attributable to well-
to define commitment, 85% 51/60 were cohabiting at the time of
documented age-related declines in reactivity (e.g., Carstensen,
the study, and two additional couples were temporarily living
Gottman, & Levenson, 1995). Nonetheless, these measures, as
together. As some studies have used cohabitation as a marker of
well as assessments of adult attachment security, were included
commitment among same-sex couples Gottman, Levenson, Swan-
in this study in order to test the assumption that committed
son, et al., 2003; Julien et al., 2003 , we conducted follow-up
same-sex relationships are lower quality contexts of human
analyses focused on the subset of cohabiting same-sex participants.
development than a range of committed and noncommitted
The pattern of results was the same as reported here). In addition,
heterosexual relationship types.
to be able to determine whether demographic differences between
couples types accounted for any differences identified between
Method
same-sex and opposite-sex couples, we recruited a demographi-
cally diverse group of same-sex couples. Table 1 summarizes key
Participants
demographics of each of the groups by sex (i.e., age, relationship
The Relationships Research Laboratory at the University of
length, and percentage White). Table 1 also lists information about
Illinois has studied romantically involved couples from a diverse
the percentage of participants who earned a Bachelor’s degree or
Table 1
Demographics of Participants by Relationship Type and Sex
Dating
Engaged
Married
Same-sex
Indicator
Male
Female
Male
Female
Male
Female
Gay
Lesbian
Age in years, M(SD)
21 (1.8)
20 (1.5)
23 (2.7)
22 (3.0)
58 (7.6)
56 (6.2)
33 (10.5)
35 (11.5)
Range
18–25
18–25
19–29
18–30
50–77
50–73
19–54
20–61
Rel. length in months, M (SD)a
16 (15.3)
41 (22.4)
379 (104.1)
85 (85.3)
71 (57.1)
Range
1–69
4–108
201–621
13–324
12–232
% White
76
73
88
86
98
98
83
90
% Bachelor’s degreeb
40
28
44
50
75
68
63
80
Note.
Rel.
relationship.
a Couples were asked how long they had been involved romantically except married couples, who were asked how long they had been married. b %
Bachelor’s degree is an estimate for dating men and women as this question was not asked at in-take but to the approximately 50% of participants who
were followed up longitudinally after a year.

---

94
ROISMAN, CLAUSELL, HOLLAND, FORTUNA, AND ELIEFF
higher by group. (These data were not used as covariates because
Measures
the younger participants were often still in the midst of completing
their educations).
Reported quality.
In our studies of engaged, married, gay
male, and lesbian couples, Berscheid, Snyder, and Omoto’s (1989)
Emotional Tone Index (ETI) was used to assess the frequency with
Procedures
which respondents experience 27 different emotions in their ro-
mantic relationships using a 7-point scale (dating couples did not
All of our studies used an identical procedure. Prior to visiting
complete this questionnaire). Twelve positive and 15 negative
the University of Illinois, participants separately completed a
emotions are included in the ETI, with both intense and less
packet of questionnaires. Upon arrival to the laboratory, the AAI
intense feeling states represented. Differences between each par-
(see below) was administered by research assistants trained by
ticipant’s average rating on the positive and the negative emotions
Glenn I. Roisman, a trained and reliable AAI coder (interviews
were computed to derive the relative frequency with which adults
were administered to participants in separate rooms). After the
experience positive and negative emotions in their relationships. In
participants completed the AAI, research assistants administered
addition, participants used a partner form of the measures to
several additional self-report measures, including a questionnaire
describe their partners’ experience of each of these emotions,
listing common problem areas in relationships (e.g., money, com-
which was coded using the same method described above. In our
munication, in-laws). Participants were instructed that this form
more recent studies of dating, gay male couples, and lesbian
would be the only questionnaire that their partner would see and
couples, we have administered the well-known Dyadic Adjustment
that they should use a 10-point scale to describe the degree to
Scale (DAS; Spanier, 1976). The DAS is a 32-item questionnaire
which each domain listed was currently a problem area in their
that has been used with a wide variety of couples to assess
relationship (from 1
not a problem to 10
is a serious
adjustment/satisfaction (note that some questions were modified
problem).
slightly to be appropriate for dating and noncohabiting couples).
Upon finishing the relationship-problem questionnaire, partici-
All items of the DAS were composited to create a total dyadic
pants were reunited with their partners to complete the last part of
adjustment score. Alpha reliability data for the ETI (positive and
the session, a standard relationship interaction task, in a comfort-
negative emotions, self and partner) and DAS (self-report) are
able living room environment. After participants had an opportu-
provided in Table 2.
nity to talk with their partner, physiological sensors were attached
Observed quality.
In all of our studies of romantically in-
and participants were asked to rest for approximately 4 min, during
volved couples, we have consistently rated both the overall dyadic
which a nontalking baseline was acquired for the final 3 min. A
quality of adults’ interactions as well as the positive and negative
research assistant entered the room after the rest period with the
affect of each individual during the interactions. Dyadic quality
problem inventories participants had completed previously and
was coded by trained graduate research assistants from videotapes
instructed couples to identify a problem area (i.e., disagreement) in
of the couple interactions using the 7-point Overall Quality rating
their relationship. After couples decided on an issue using the
drawn from a system created by Aguilar et al. (1997), which
problem inventories, participants were given 10 min to discuss and
evaluates the degree to which relationships facilitate disclosure of
attempt to resolve this problem. Couples were next instructed to
emotions and ideas, appear to sustain personal development, and
take 5 min to talk about areas of agreement.
help individuals with task demands. Coders also used the positive
Surface sensors measuring skin conductance levels and cardiac
and negative affect ratings from the Interactional Dimensions
activity were adhered to participants’ torsos and fingers by a
Coding System (IDCS; Kline et al., 2005). Both affect ratings were
research assistant prior to the beginning of the dyadic interaction.
made separately along a 9-point scale for each partner, with lower
Physiological recordings were subsequently monitored from an
scores reflecting less and higher scores indicating more positive/
adjoining room during the interpersonal observation as well as
negative emotion as reflected in each participant’s face, voice, and
throughout the rest period just prior to the interaction. A research
body. In order to create a measure of emotional tone comparable
assistant was present in the living room environment only long
to the self-report emotional tone index described above (see
enough to give directions to the participants and answer any
Roisman, 2007), negative affect scores were subtracted from
questions they might have.
positive affect ratings to create an index of observed emotional
tone. As such, the range of possible values for this variable is
9
affectively negative behaviors to 9
affectively positive
Apparatus
behaviors.
Audiovisual.
Remotely controlled, high-resolution color video
The interrater reliabilities (intraclass correlations; ICCs) of all
cameras were used in conjunction with VCRs to record couples’
scales were adequate (
.60) across all relationship types (see
dyadic interactions (AAIs were audiotaped). Cameras were em-
Table 2). Note that in each study, ICCs were based on a subsample
bedded in two bookshelves facing the participants. Lavaliere mi-
of double-coded interactions (in each case, final scores were ar-
crophones clipped on participants’ clothing were used to record
rived at by consensus). Fifty-two percent of the videos were
their conversations during the interactions.
reliability cases in the dating couples study, 80% were reliability
Physiological.
A system consisting of two Pentium comput-
cases in the engaged couples study, 100% were reliability cases in
ers, Snapmaster software (HEM Data Corp., Southfield, MI), and
the married couples study, and 15% were reliability cases in our
bioamplifiers (James Long, Inc., Caroga Lake, NY) was used to
study of same-sex couples (less overlap was used in our study of
acquire continuous recordings of both participants’ physiological
same-sex couples because our coders were highly experienced by
responses during the dyadic interaction and at rest.
that time, having already coded engaged and married couples).

---

SPECIAL SECTION: SAME-SEX AND OPPOSITE-SEX COUPLES
95
Table 2
Reliability of Measures by Relationship Type and Sex
Dating
Engaged
Married
Same-sex
Indicator
Male
Female
Male
Female
Male
Female
Gay
Lesbian
Reported quality
ETI positive (S)
—
—
.89
.88
.94
.93
.89
.95
ETI negative (S)
—
—
.86
.91
.82
.91
.91
.88
ETI positive (P)
—
—
.94
.91
.95
.92
.92
.92
ETI negative (P)
—
—
.93
.91
.91
.91
.93
.90
DAS adjustment
.88
.86
—
—
—
—
.84
.90
Observed quality
Dyadic quality
.80
.84
.79
.73
.87
Positive affect
.85
.83
.76
.78
.70
.68
.60
.72
Negative affect
.76
.78
.64
.75
.66
.76
.96
.76
AAI Q-set
% over .6 threshold
—
84
85
82
88
Mean
—
.79
.75
.73
.78
Range
—
.60–.91
.60–.90
.60–.83
.61–.91
Note.
ETI
Emotional Tone Index (questionnaire); DAS
Dyadic Adjustment Scale (questionnaire); AAI
Adult Attachment Interview; (S)
self-report; (P)
partner report; —
not available. All coefficients above are standardized alphas, except observational data, which are intraclass
correlations (all ps
.001) and AAI data, as noted (see Method section for details). Partner-report ETI refers to ratings of the individual listed at the top
of each column by partner.
Because a number of videos were treated as group-coded training
experiences (Main & Goldwyn, 1998). In the AAI, participants are
cases, the percentage of cases coded by consensus was 61%,
asked to describe their early relationships with parents and revisit
100%, 100%, and 38% for dating, engaged, married, and same-sex
episodes of salient separations and possible rejection, abuse, and
samples, respectively.
loss. Participants then discuss the effects of these experiences on
Physiological reactivity.
In all of our studies of romantic
their development.
relationships, measures of physiological responding were sam-
In each study, the coherence of participants’ discourse was
pled from participants’ electrodermal and cardiac systems dur-
assessed using Kobak’s (1993) AAI Q-set by coders reliable with
ing a 3-min rest period and the 10-min disagreement epoch.
Main and Goldwyn’s (1998) AAI Coding System. The AAI Q-set
Electrodermal response was measured by skin conductance
(Kobak, 1993) consists of 100 descriptive cards that are sorted into
level (SCL). A constant-voltage device was used to pass a small
a forced normal distribution across nine piles from least to most
voltage between electrodes attached to the palmar surface of
characteristic. We estimated intercoder reliability by double-
the last phalanxes of the second and fourth fingers of the
coding a subset of AAIs, achieving
80% reliability of .6 or
nondominant hand. SCL was measured in microsiemens. The
greater (Spearman-Brown formula), the standard criterion in re-
specific cardiac measure obtained was heart rate (HR). Elec-
search of this kind (see Table 2). A third coder rated transcripts for
trode stickers were placed in a bipolar configuration on oppo-
which initial coders were discrepant and the most highly correlated
site sides of each participant’s chest (a ground lead was posi-
sorts were ultimately averaged to increase reliability (means and
tioned on the sternum) and cardiac inter-beat intervals (IBI)
ranges are provided on Table 2). The percentage of double-sorted
were measured as time in milliseconds between successive
(reliability) AAIs was as follows: engaged (25%), married (25%),
R waves of the electrocardiogram (EKG). Heart rate was cal-
and same-sex (44%). In the final step of data reduction, Pearson
culated second by second using the following standard formula:
correlations were computed between each of the composited sorts
HR
(1/IBI)
60,000 ms. As such, HR reflects beats per
and a prototypic “secure versus insecure” prototype sort developed
minute. For SCL and HR, change in physiological respond-
by Kobak and his colleagues (see Kobak, Cole, Fleming, Ferenz-
ing was calculated by subtracting mean levels of physiologi-
Gillies, & Gamble, 1993). Prototypically secure items include
cal responding during the 3-min rest period from mean lev-
“responds in a clear, well-organized fashion” and “is credible and
els during the disagreement epoch of the interpersonal
easy to believe.” On the basis of this analysis, participants were
interactions, a practice commonly used in physiological re-
assigned scores ranging from
1.00 to 1.00 on security, with
search that has been defended conceptually and statistically by
higher scores indicating greater resemblance to the prototypically
Rogosa (1995; see also Llabre, Spitzer, Saab, Ironson, &
secure individual.
Schneiderman, 1991).
Adult attachment security.
We administered the AAI (George
Results
et al., 1985; Hesse, 1999), a well-validated, semistructured inter-
view regarding early experiences with caregivers in all of our
Analytic Plan
studies (relevant data from our study of dating couples, however,
are not yet coded). The AAI is designed to examine whether adults
One of the analytic challenges of this study was in avoiding
are able to construct a coherent narrative regarding their childhood
violations of the independence assumption inherent in most

---

96
ROISMAN, CLAUSELL, HOLLAND, FORTUNA, AND ELIEFF
statistical tests, particularly for nondistinguishable (i.e., gay
groups (e.g., older married couples in long-term marriages). In
male and lesbian) couples. For simplicity of presentation, in
each analysis, we examined whether the significant differences
initial analyses the scores of gay males and lesbians were
between the two groups could be attenuated to nonsignificance
averaged within couples to avoid nonindependence (the one
by adding the three demographic variables into a univariate
exception was for dyadic quality analyses, where this was
ANOVA as covariates in a single block. Details regarding these
unnecessary). Using an analysis of variance (ANOVA) frame-
analyses are available from Glenn I. Roisman by request.
work, groups of males and females from all relationships types
In order to provide confirmation of the conclusions reached
were then compared with one another on relationship quality
using ANOVA, we also present results of multilevel modeling
indicators (i.e., self- and partner-reported, observed, and phys-
using hierarchical linear modeling software HLM 6.02 (Bryk &
iological) and on adult attachment security (see Table 3). Males
Raudenbush, 1992). Specifically, we estimated models in which
and females were treated as separate groups in initial ANOVA
observations from individual partners (Level I) were nested in
analyses both because of documented sex differences on some
couples (Level II). The Level I model was a within-couple regres-
variables that were the focus of this investigation (e.g., heart
sion equation that used information from both partners to estimate
rate reactivity; Smith, Gallo, Goble, Ngu, & Stark, 1998) and in
an intercept for each dyad, which reflects an average of the
order to provide a complete descriptive account of our data. For
outcome score for each couple (to be predicted in a Level II
each analysis, results of least significant difference (LSD) post
regression by couple type; see Kurdek, 2004). Because HLM
hoc tests (e.g., all pairwise contrasts) are provided. (All signif-
models require that the analyst identify a referent couple type for
icant effects are p
.05, unless otherwise specified.) Note that
all comparisons, we present two sets of analyses examining com-
the ANOVA analyses presented here are conservative in that
parisons of all groups first to gay male couples and next to lesbian
they use degrees of freedom for statistical tests based on the
couples. Relationship type was represented using a set of dummy
number of same-sex couples rather than individuals while re-
codes that were entered simultaneously in the Level II equation
taining a simple and familiar analytic framework that allows for
(the referent couple type served as an implicit comparison group;
direct comparisons of gay males and lesbians with heterosexual
thus, positive unstandardized coefficients indicate that a given
males and females separately.
couple type had higher scores on the outcome of interest compared
As described earlier, heterosexual couples were identified
with the goal of studying distinct relationship types that, by
with gay male/lesbian dyads and negative effects indicate that a
definition, vary on key demographic parameters. As such, in
given couple type had lower scores on the outcome of interest
supplementary ANOVA analyses, we examined the effects of
compared with gay male/lesbian dyads). In HLM analyses, we also
these background characteristics (i.e., age, relationship length,
examine whether the effects of individual-level (i.e., sex, ethnicity,
and percentage White) on statistically significant differences
and age) and couple-level (i.e., relationship length) covariates
identified between individuals from same-sex couples and
accounted for relationship type differences on dependent mea-
opposite-sex groups. We examined the effects of demographic
sures.
variables on differences between individuals in same-sex versus
Note finally that in all ANOVA and HLM analyses, missing
opposite-sex couples only (a) because the primary focus of this
data were imputed within studies using an EM algorithm (
1% of
article was on these putative differences and (b) the sampling
data from any study were imputed), with one exception—
for the same-sex couples study was designed so that participants
physiological analyses for the engaged sample were based on a
represented a wide range of ages and relationship durations—
subset of 40 cases for whom complete physiological data were
unlike our other samples, in which we built in range restrictions
collected (we did not impute these values to be consistent with data
for purposes of identifying demographically homogenous
in Roisman, 2007).
Table 3
Indicators of Relationship Quality and Attachment Security by Relationship Type and Sex
Dating
Engaged
Married
Same-sex
Indicator
Male
Female
Male
Female
Male
Female
Gay
Lesbian
Reported quality
ETI (self)
—
—
3.9 (1.2)
4.1 (1.3)
3.5 (1.3)
3.8 (1.3)
3.4 (1.1)
3.6 (1.4)
ETI (partner)
—
—
3.2 (1.7)
4.0 (1.3)
3.0 (1.6)
3.2 (1.5)
2.9 (1.3)
3.4 (1.3)
DAS adjustment
119 (14.1)
121 (12.5)
—
—
—
—
114 (7.9)
116 (9.9)
Observed quality
Dyadic quality
3.4 (1.3)
4.1 (1.2)
4.3 (1.0)
4.3 (1.2)
4.8 (1.2)
Emotional tone
0.5 (3.1)
0.2 (3.2)
0.8 (2.3)
1.0 (2.2)
0.2 (1.8)
0.6 (2.1)
1.0 (2.4)
1.6 (2.1)
Physiological reactivity
Heart rate
5.3 (4.3)
4.9 (4.4)
4.2 (4.4)
4.3 (3.9)
2.9 (3.5)
3.9 (3.1)
4.1 (2.8)
3.6 (2.8)
Skin conductance
4.3 (2.5)
3.8 (2.0)
4.6 (2.5)
3.7 (2.1)
2.7 (1.6)
2.9 (1.8)
4.3 (1.8)
4.1 (1.7)
AAI Security
—
—
.12 (.49)
.37 (.44)
.24 (.43)
.22 (.41)
.38 (.29)
.38 (.30)
Note.
ETI
Emotional Tone Index (questionnaire); DAS
Dyadic Adjustment Scale (questionnaire); AAI
Adult Attachment Interview; —
not
available. Partner-report ETI refers to ratings of the individual listed at the top of each column by partner. All values are means (with standard deviations
in parentheses).

---

SPECIAL SECTION: SAME-SEX AND OPPOSITE-SEX COUPLES
97
Group Differences on Relationship Quality and Adult
post hoc tests revealed that married men showed less heart rate
Attachment
reactivity than dating men and women (effect size rs
.29 and
.24, respectively). In addition, lesbians showed less heart rate
Reported quality.
Individuals composing gay male, lesbian,
reactivity than dating men (effect size r
.23), an effect that was
engaged, and married couples were compared by sex on self- and
accounted for by demographic variables.
partner-reported emotional tone in two ANOVAs. The overall F
The overall F test for electrodermal reactivity was also signif-
test for self-reported emotional tone was nonsignificant, F(5,
icant, F(7, 437)
4.7, p
.001, and post hoc tests revealed that
239)
1.67, p
.14. In contrast, the overall effect for partner-
married men showed less electrodermal reactivity than gay men,
reported emotional tone was statistically significant, F(5, 239)
lesbians, dating men, dating women, engaged men, and engaged
2.98, p
.05, and post hoc tests were computed. These tests
women (effect size rs
.36, .29, .41, .26, .43, and .39, respec-
revealed that engaged women were seen by their partners as
tively). Similarly, married women showed significantly less elec-
experiencing more positive relative to negative emotion than all
trodermal reactivity than gay men, lesbians, dating men, dating
other groups (engaged men, married men, married women, and gay
women, and engaged men (effect size rs
.31, .18, .36, .36, and
males; effect size rs were .25, .32, .27, .39, respectively) except
.32, respectively). In all cases, the effects of demographic variables
lesbians (no other differences emerged among individuals in mar-
(especially age) accounted for differences between individuals in
ried and same-sex couples). Follow-up analyses demonstrated that
same-sex relationships and older married individuals.
differences involving individuals from same-sex relationships
Adult attachment security.
An ANOVA was computed to con-
were not accounted for by the effects of age, relationship duration,
trast gay males, lesbians, engaged men and women, and married men
or ethnicity.
and women on the security AAI Q-set dimension, reflecting the
In addition, gay men, lesbians, dating males, and dating females
coherence of adults’ narratives about their childhood experiences
were compared on overall self-reported dyadic adjustment as mea-
(AAI data on the dating couples sample are not yet available). The
sured by Spanier’s (1976) Dyadic Adjustment Scale, F(3, 277)
overall F test for security was significant, F(5, 239)
3.0, p
.05,
2.91, p
.05. Note, first, that both gay men and lesbians reported
and post hoc tests revealed that engaged men produced significantly
levels of adjustment virtually identical to the norms reported by
less coherent attachment-related narratives (i.e., were less secure) than
Spanier (1976) for married men and women (M
114.8, SD
gay males, lesbians, and engaged women (effect size rs
.20, .26,
17.8; based on a sample of 218 individuals). That said, dating men
and .30, respectively). Effects involving gay men and lesbians were
and women reported significantly higher levels of satisfaction than
not accounted for by demographic variables.
gay men (effect size rs
.21 and .32, respectively), an effect that
remained significant when controlling for demographic variables.
Multilevel Modeling Results
Observed quality.
Two ANOVAs were computed to examine
group differences on observations of dyadic quality as well as of
Using the approach outlined earlier, parallel analyses were con-
individuals’ emotional tone during their laboratory interactions.
ducted using an HLM framework to replicate results related to the
The overall F test for dyadic quality was significant, F(4, 258)
comparison of lesbian and gay couples with opposite-sex couple
10.6, p
.001. Using the post hoc test, as expected, dating couples
groups. Results are summarized in Tables 4 and 5 (Table 4 contrasts
were rated as having significantly lower quality interactions com-
gay male dyads with all other relationship types; Table 5 contrasts
pared with gay male, lesbian, engaged, and married couples (effect
lesbian dyads with all other couple types). Note that effects of couple
size rs
.27, .36, .34, and .49, respectively). In addition, lesbian
type are provided both with and without demographic covariates
couples had significantly higher quality interactions than engaged
(unstandardized effects in parentheses are coefficients without any
couples (effect size r
.28). (No other group differences were
covariates in the models). As would be expected, HLM analyses
significant, and significant differences described above involving
produced a similar pattern of results as identified above. Most criti-
individuals from same-sex relationships were not accounted for by
cally, as can be seen in Tables 4 and 5, both gay male and lesbian
demographic variables.)
couples demonstrated significantly more positive emotional tone than
In terms of observed emotional tone, F(7, 457)
4.3, p
.05,
did dating couples (effects that held controlling for demographic
post hoc tests indicated that dating males emitted significantly
variables). Similarly, lesbian couples (see Table 5) demonstrated
more negative (vs. positive) affect than gay men, lesbians, engaged
significantly more positive emotional tone than married couples,
men, engaged women, and married women (effect size rs
.23,
effects (as in the ANOVA analyses) that were accounted for by
.27, .20, .26, and .37, respectively). Similarly, dating women
demographic differences between the groups.
showed significantly more negative (vs. positive) emotional tone
than gay men, lesbians, engaged men, and engaged women (effect
Summary of Results
size rs
.18, .21, .21, and .32, respectively). Interestingly, lesbi-
ans showed more positive emotional tone than married men (effect
This study yielded the following four primary conclusions:
size r
.34). Note that significant differences between (a) lesbians
1. Gay males, lesbians, married men, and married women have
and married men, (b) dating men and gay men, and (c) dating
similarly positive views of their relationships. Interestingly, dating
women and gay men were accounted for (i.e., reduced to nonsig-
couples and engaged women tend to have even more positive
nificance) by the effects of demographic variables.
appraisals of their interpersonal experiences than those in longer
Physiological reactivity.
Two ANOVAs were computed to
term relationships, a finding that may reflect the fewer stressors
contrast groups on change in heart rate and electrodermal response
that exist for such couples, the greater ease with which such
from baseline (i.e., physiological reactivity). The overall F test for
relationships can be dissolved and, in the case of DAS analyses,
heart rate reactivity was significant, F(7, 437)
2.2, p
.05, and
that some domains of conflict tapped by the instrument are simply

---

98
ROISMAN, CLAUSELL, HOLLAND, FORTUNA, AND ELIEFF
Table 4
Unstandardized Coefficients for Demographic and Relationship Type Predictors of Relationship Quality Indicators and Attachment
Security From HLM Models Comparing Gay Male Dyads With All Other Couple Types
Level I variables
Level II variables
Indicator
Age
Ethnicity
Sex
Rel. length
Datinga
Engageda
Marrieda
Lesbiana
Reported quality
ETI (self)
0.00
0.08
0.27*
0.00
—
0.46 (0.55*)
0.50 (0.20)
0.06 (0.19)
ETI (partner)
0.01
0.13
0.56**
0.00
—
0.55 (0.65*)
0.88 (0.21)
0.03 (0.50)
DAS adjustment
0.14
4.63*
1.57
0.03
5.69* (6.00**)
—
—
0.61 (1.81)
Observed quality
Emotional tone
0.00
0.86*
0.37*
0.00
1.42* ( 1.36**)
0.24 ( 0.08)
0.78 ( 0.58)
0.16 (0.62)
Physiological reactivity
Heart rate
0.11**
1.06*
0.12
0.00
0.25 (0.97)
1.03 (0.10)
1.47 ( 0.73)
0.27 ( 0.49)
Skin conductance
0.05**
0.22
0.47*
0.00
0.71 ( 0.23)
0.48 ( 0.15)
0.03 ( 1.53**)
0.33 ( 0.22)
AAI security
0.01**
0.24**
0.12
0.00
—
0.31** ( 0.14)
0.13 ( 0.16*)
0.12 ( 0.01)
Note.
HLM
hierarchical linear modelling; Rel.
relationship; ETI
Emotional Tone Index (questionnaire); DAS
Dyadic Adjustment Scale
(questionnaire); AAI
Adult Attachment Interview; —
not available. Partner-report ETI refers to ratings of the individual listed at the top of each
column by partner. Coefficients in parentheses represent effects without Level I and II covariates. Positive coefficients indicate that couple type scored
higher on the outcome than gay males. Negative coefficients indicate that couple type scored lower on the outcome than gay males. Male
1, Female
2 for sex; non-White
0, White
1 for ethnicity.
a Versus gay.
* p
.05.
** p
.01.
less relevant or salient for dating compared with more committed
married men and women proved to be the least physiologically
couples.
reactive couples we have studied. No other group differences
2. Committed couples (i.e., gay male, lesbian, married, and
emerged in this domain, except that lesbians showed less heart-rate
engaged) have higher quality laboratory interactions with their
reactivity than dating men.
partner than noncommitted (i.e., heterosexual dating) couples.
4. Although all the groups we have studied on average produced
Although almost no differences were identified among committed
coherent narratives about their childhood experiences, engaged
couples in terms of the observed quality of their interactions, in
men were less secure than gay men, lesbians, and engaged women.
partial replication of Gottman, Levenson, Swanson, et al. (2003),
lesbians appear to be particularly skilled at working harmoniously
Discussion
with their partners.
3. Corroborating consistent evidence for age-related declines in
In this study, we examined the claim that committed same-sex
physiological reactivity during interpersonal interactions, older
relationships are atypical, psychologically immature, or malevo-
Table 5
Unstandardized Coefficients for Demographic and Relationship Type Predictors of Relationship Quality Indicators and Attachment
Security From HLM Models Comparing Lesbian Dyads With All Other Couple Types
Level I variables
Level II variables
Indicator
Age
Ethnicity
Sex
Rel. length
Datinga
Engageda
Marrieda
Gaya
Reported quality
ETI (self)
0.00
0.08
0.27*
0.00
—
0.52 (0.36)
0.44 (0.01)
0.06 ( 0.19)
ETI (partner)
0.01
0.13
0.56**
0.00
—
0.57 (0.14)
0.85 ( 0.29)
0.03 ( 0.50)
DAS adjustment
0.14
4.63*
1.57
0.03
6.29* (4.19*)
—
—
0.61 ( 1.81)
Observed quality
Emotional tone
0.01
0.86*
0.37*
0.00
1.58** ( 1.98**)
0.40 ( 0.70)
0.94 ( 1.19*)
0.16 ( 0.62)
Physiological reactivity
Heart rate
0.11**
1.06*
0.12
0.00
0.02 (1.46*)
0.76 (0.59)
1.74 ( 0.24)
0.27 (0.49)
Skin conductance
0.05**
0.22
0.47*
0.00
1.05* ( 0.01)
0.82 (0.07)
0.36 ( 1.31**)
0.33 (0.22)
AAI security
0.01**
0.24**
0.12
0.00
—
0.19* ( 0.13)
0.01 ( 0.15*)
0.12 (0.01)
Note.
HLM
hierarchical linear modelling; Rel.
relationship; ETI
Emotional Tone Index (questionnaire); DAS
Dyadic Adjustment Scale
(questionnaire); AAI
Adult Attachment Interview; —
not available. Partner-report ETI refers to ratings of the individual listed at the top of each
column by partner. Coefficients in parentheses represent effects without Level I and II covariates. Positive coefficients indicate that couple type scored
higher on the outcome than lesbian couples. Negative coefficients indicate that couple type scored lower on the outcome than lesbian couples. Male
1,
Female
2 for sex; non-White
0, White
1 for ethnicity.
a Versus lesbian.
* p
.05.
** p
.01.

---

SPECIAL SECTION: SAME-SEX AND OPPOSITE-SEX COUPLES
99
lent contexts of human development by directly observing and
ceptions of their relationship-related experiences, such data pro-
systematically comparing individuals in such relationships with
vide a relatively narrow database with respect to addressing ques-
heterosexuals drawn from noncommitted dating couples as well as
tions about the relative quality of different kinds of committed
with committed engaged and married dyads. We believe that this
partnerships. In contrast, studies that use multiple methods (in this
report is most profitably viewed in the broader context of longi-
study, self-reports, observations, and measures of interpersonal
tudinal research (e.g., Kurdek, 2004), demonstrating that, com-
physiological reactivity) provide a broader basis for comparison
pared with married individuals, committed gay males and lesbians
and thereby allow for a more complete understanding of same-
are neither less satisfied with their relationships nor report higher
versus opposite-sex relationships (Gottman, Levenson, Swanson,
levels of the kinds of personal attributes that mitigate against the
et al., 2003; Julien et al., 2003).
quality and longevity of adult relationships.
Second, it is important to emphasize just how limited the com-
The current study adds to this literature by demonstrating that,
parative literature is currently with respect to studies that have
controlling for demographic differences, gay males and lesbians in
tapped into the kind of methodological sophistication that is com-
our studies were generally not distinguishable from their commit-
monplace in research on other relationship types. To our knowl-
ted heterosexual counterparts on measures of self- and partner-
edge, this article in fact represents only the third published com-
reported relationship quality, as well as in how they interacted with
parative observational study focused on mean-level differences
one another—and responded physiologically—while attempting to
among gay male, lesbian, and heterosexual couples. That the
resolve conflict in their relationships. Similarly, despite the signif-
results of this burgeoning literature are coherent across these few
icant challenges faced by many sexual minority youth in their
studies—and consistent with work by Kurdek and others using
families of origin (Savin-Williams, 2001; Weston, 1992), gay
“insider” methods (e.g., self-reports, interviews) —is encouraging.
males and lesbians proved to be among the most secure adults we
Regardless, there is clearly much research to be done in this area,
have interviewed in our laboratory, as reflected in the coherence of
especially in scaling up work of this kind such that relevant
their narratives about childhood experiences. This latter finding is
analyses are based on a wide variety of state-of-the-art methods
especially important in that attachment security has been identified
applied to more representative samples of same- and opposite-sex
as a critical interpersonal resource for adult development (Hesse,
couple types. On this point in particular, the current study is not
1999). To the extent that differences were identified between
without important limitations. For example, our samples of het-
committed same- and opposite-sex couples, the findings suggested
erosexual dating, engaged, married, gay male, and lesbian couples
that lesbians were especially effective at working together harmo-
were clearly modest in size. In addition, this study relied on
niously (in partial replication of Gottman, Levenson, Swanson, et
secondary analyses of data from couples that were not demograph-
al., 2003).
ically matched. Finally, we make no claim that these community
Rather than reflecting sexual orientation, however, the finding
samples are representative.
that lesbian couples interacted more harmoniously than other cou-
We also recognize that, despite our goal of examining the
ple types we have studied may in part be attributable to the
strengths and resources of various kinds of adult relationship
interpersonal dynamics of interacting with someone of the same
forms, some may view this analysis as implicitly adopting what
sex (see Stacey & Biblarz, 2001, for a similar argument in relation
sociological scholars (e.g., Stacey & Biblarz, 2001) have referred
to the high quality parenting of lesbians). In fact, we have recently
to as the hetero-normativity assumption. As Baumrind (1995)
examined this possibility in a study of same- versus opposite-sex
emphasized, differences among gay males, lesbians, and hetero-
interactions between strangers during a challenging collaborative
sexuals (when they emerge) need not be conceptualized as deficits,
task (i.e., 3-D puzzle building). Using modified versions of the
particularly in a pluralistic society that might be expected to
IDCS emotional tone scales described in the current work—and
embrace differences that do not threaten development. That said,
consistent with the results of the present study—we found that
we emphasize that the methods we use in our laboratory reflect
same-sex stranger dyads appear to demonstrate advantages in
dimensions of variability that are not linked to heterosexual (or
observations of the quality of their interactions. More specifically,
homosexual) ideals but rather are empirically validated assess-
both males and females in same-sex stranger dyads showed more
ments of factors that promote (and reflect) successful adult rela-
positive relative to negative affect during their interactions than
tionships, methods widely used in the scientific study of these
did males or females in opposite-sex dyads (unpublished analyses
critical interpersonal contexts (Gottman & Levenson, 1992; Lev-
are available from Glenn I. Roisman). Although clearly not defin-
enson & Gottman, 1985; Tsai & Levenson, 1997). That such
itive, such results suggest that the emerging finding that same-sex
sophisticated research tools yield evidence that committed same-
couples in some cases have more positive interactions than
sex couples are much like their engaged and married counterparts
opposite-sex dyads (e.g., Gottman, Levenson, Swanson, et al.,
further calls into question the assumption to the contrary—an
2003) may have little to do with sexual orientation per se.
assumption that seems increasingly inconsistent with available
We hope that the major impact of this work will be to draw
data.
further attention to two interrelated issues that have been some-
what underemphasized in reviews of comparative research on
References
same- versus opposite-sex relationships (e.g., Herek, 2006). First,
as noted by Gottman, Levenson, Gross, et al. (2003) in their
Aguilar, B., Christian, S. D., Collins, W. A., Cook, J., Hennighausen,
seminal study, the vast majority of work in this area has involved
K. H., Hyson, D. M., et al. (1997). Romantic relationship assessment
the administration of self-report questionnaires and interviews to
observational rating scales. Unpublished coding manual, Parent-Child
same- and opposite-sex (typically married) couples. Although such
Project, University of Minnesota.
“insider” information is critical in ascertaining participants’ per-
Allen, M., & Burrell, N. (1996). Comparing the impact of homosexual and

---

100
ROISMAN, CLAUSELL, HOLLAND, FORTUNA, AND ELIEFF
heterosexual parents on children: Meta-analysis of existing research.
relationship satisfaction and relationship dissolution. Journal of Homo-
Journal of Homosexuality, 32, 19 –35.
sexuality, 45, 23– 43.
Babcock, J. C., Jacobson, N. S., Gottman, J. M., & Yerington, T. P. (2000).
Gottman, J. M., Levenson, R. W., Swanson, C., Swanson, K., Tyson, R., &
Attachment, emotional regulation, and the function of marital violence:
Yoshimoto, D. (2003). Observing gay, lesbian and heterosexual couples’
Differences between secure, preoccupied, and dismissing violent and
relationships: Mathematical modeling of conflict interaction. Journal of
nonviolent husbands. Journal of Family Violence, 15, 391– 409.
Homosexuality, 45, 65–91.
Baumrind, D. (1995). Commentary on sexual orientation: Research and
Herek, G. M. (2006). Legal recognition of same-sex relationships in the
social policy implications. Developmental Psychology, 31, 130 –136.
United States: A social science perspective. American Psychologist, 61,
Berscheid, E., Snyder, M., & Omoto, A. M. (1989). The relationship
607– 621.
closeness inventory: Assessing the closeness of interpersonal relation-
Hesse, E. (1999). The Adult Attachment Interview: Historical and current
ships. Journal of Personality and Social Psychology, 57, 792– 807.
perspectives. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attach-
Blumstein, P., & Schwartz, P. (1983). American couples: Money, work,
ment: Theory, research, and clinical applications (pp. 395– 433). New
sex. New York: William Morrow.
York: Guilford.
Bouthillier, D., Julien, D., Dube, M., Belanger, I., & Hamelin, M. (2002).
Julien, D., Chartrand, E.,