KELLY E.

KNIGHT Sam Houston State University

Assortative Mating and Partner Influence on

Antisocial Behavior Across the Life Course

Why do individuals select romantic partners is the selection of the child’s other parent. This
who use drugs, are criminals, or have mental person contributes half of the child’s genome
health problems, a choice that eventually puts and, hopefully, a reasonable share of the par-
them and their children at increased risk for enting and socialization, as well as ultimately
negative developmental outcomes? Results are contributing to the family’s genetic immortal-
presented from a systematic literature review ity (Lykken & Tellegen, 1993). Although the
on assortative mating for antisocial behavior selection of a mate has profound consequences,
and on the subsequent influence partners have the process is not well understood. Even less
on each other. All cross-sectional, retrospec- is known about individuals who select dysfunc-
tive studies except one supported assortative tional or antisocial mates who, for example,
mating over partner influence. In contrast, all use drugs, are criminals, or have mental health
prospective studies supported partner influence. problems that eventually put them and their
Given that prospective data are generally bet- children at increased risk for negative devel-
ter than retrospective data, partner influence opmental outcomes. There is some consensus
is recognized here as an important finding, among researchers that people choose mates
previously hidden or discounted in the litera- who are similar to themselves, a phenomenon
ture because of its reliance on retrospective, often referred to as assortative mating.
cross-sectional designs. Theoretical perspec- Broadly defined, assortative mating is ‘‘the
tives, social homogamy, heterotypic assortment, nonrandom coupling of individuals based on
and methodological issues are also examined. their similarity to each other on one or more
characteristics’’ (Watson et al., 2004, p. 1030).
The family is among the most important social- The term mating is taken from biology to indi-
ization domains in a child’s life (Simons, Gordon cate an individual’s pairing with the opposite
Simons, & Ebert Wallace, 2004), and research sex to produce a biological child. However,
has demonstrated that the choices parents make in social science research, the operationalized
can exert both positive and negative influ- definition of assortative mating has less to do
ences on their child’s development (Thornberry, with biology and more to do with close rela-
Freeman-Gallant, & Lovegrove, 2009). Perhaps tionships and the type of data collected about
the most important choice a parent ever makes, partners and their relationships occurring at
which occurs well before his or her child is born, different developmental stages. Definitions of
mates can include adolescent couples, cohab-
iting young adults, spouses, same-sex partners,
Sam Houston State University, College of Criminal Justice, and at other times biological parents. In addition,
Box 2296, Huntsville, TX 77341 ([email protected]). some authors distinguish between primary and
Key Words: antisocial behavior, assortative mating, mate secondary assortative mating. Primary assorta-
selection, partner influence. tive mating describes similarity based on the
198 Journal of Family Theory & Review 3 (September 2011): 198–219
DOI:10.1111/j.1756-2589.2011.00095.x
Assortative Mating and Partner Influence 199

direct selection of phenotypes (i.e., observ- topic’s interdisciplinary relevance to psycholo-

able characteristics and behavior). Secondary gists, sociologists, geneticists, and criminolo-
assortative mating occurs when concordance is gists alike. Considering the importance of this
due to shared demographic characteristics or topic, not only for individuals themselves but
social environment. Often, the concepts are con- also for their children, the family, and society at
flated with other processes related to partner large, the goal of this article is to assess and inte-
similarity.1 These mechanisms can be grouped grate relevant theoretical and empirical literature
into two broad areas: selection and socialization developed across multiple disciplines to address
(Yamaguchi & Kandel, 1993). the following broad question: Through assor-
In general, selection refers to the processes tative mating processes, does an individual’s
that initially help two individuals find each other antisocial behavior in adolescence and young
and includes similar terms, such as assortative adulthood influence his or her mate selection,
mating, homophily, homogamy, and endogamy. and in turn, does that individual’s mate selection
Homophily refers to the contact among similar influence his or her later antisocial behavior? The
others that occurs at greater rates than among first section of the article draws on the theoreti-
dissimilar others (McPherson, Smith-Lovin, & cal literature to explain the mechanisms that may
Cook, 2001). Homogamy describes a partner- be responsible for assortative mating and partner
ship with or marriage to someone similar in influence on antisocial behavior. The second and
status, whereas endogamy involves marriage third sections of the article present findings from
within one’s own group (Kalmijn, 1998). Social a comprehensive literature review of the evi-
homogamy refers to the process whereby indi- dence related to assortative mating for antisocial
viduals in similar social groups are more likely to behavior and the subsequent influence partners
form relationships (Rhule-Louie & McMahon, have on each other over time. The article con-
2007). In contrast, intermarriage, heterogamy, cludes with a summary of findings, methodolog-
and exogamy are terms used to describe marriage ical limitations, and future research directions.
outside one’s group (Kalmijn, 1998). Socializa-
tion, however, is a broad term used in this context THEORETICAL PERSPECTIVES
to describe the influence partners have on each
other once their union begins. The literature Assortative Mating Theory
includes several analogous terms to describe this Theoretically, there are various mechanisms that
phenomenon, including contagion, corruption, can account for assortative mating and partner
causation, cohabitation effects, phenotypic con- influence. The following section outlines both
vergence, infection, reciprocal exchange, social structural and individual perspectives, as well as
amplification, and partner influence. genetic-based hypotheses of assortative mating.
Overall, these concepts are often conflated. Then, theories that help explain the influence that
For example, an author may find support for partners have on each other’s antisocial behavior
assortative mating because partners appear sim- are discussed.
ilar on their drug use when measured well into A structural explanation for assortative mat-
their relationship (which is actually a measure ing that focuses on the larger social context is
of partner similarity). However, such an assess- often referred to as social homogamy. Com-
ment cannot determine whether partners were munity structural factors, such as ethnic het-
similar in their drug use before their courtship, erogeneity, residential stability, socioeconomic
a factor that may have increased the probability status, and level of urbanization, are correlated
of their coupling (i.e., selection), or whether one with antisocial behavior (Sampson & Groves,
or both of the individuals’ drug use changed 1989), promote propinquity, and limit the scope
over time as a result of the other partner’s of possible mates (Matsueda & Heimer, 1997).
influence (i.e., socialization). Furthermore, the Academic, occupational, and religious institu-
research on the extent, mechanisms, and con- tions, as well as the family, promote both group
sequences of selection into and socialization of identification and sanctions that shape mating
antisocial behavior is rather piecemeal, given the preferences and provide a shared environment
in which individuals are likely to socialize and
select a similar other (Kalmijn, 1998).
1 The terms partner similarity, concordance, and homo- At the individual level, antisocial behavior
phily are used interchangeably. is associated with a host of problem behaviors
200 Journal of Family Theory & Review

(Fergusson, Horwood, & Lynskey, 1994) that the degree of similarity, and prefer those who
may promote assortative mating. Often, antiso- resemble their mother’’ (Bereczkei et al., 2002).
cial individuals have lower levels of intelligence
(Lynam, Moffitt, & Stouthamer-Loeber, 1993;
Raine et al., 2005), a history of maltreatment Partner Influence Theory
(Ireland, Smith, & Thornberry, 2002), and lim- Structural, individual, and genetic factors may
ited or compromised advice and support from play important roles in understanding assortative
family and friends (Quinton, Pickles, Maughan, mating with respect to antisocial behavior. From
& Rutter, 1993). They may have less self-control the literature, it appears that antisocial individu-
(Gottfredson & Hirschi, 1990) and are thus less als are likely to select into relationships with each
likely to perceive opportunities, consider alter- other. Once partnerships have been established,
natives, and make effective choices when select- however, do partners influence each other’s anti-
ing mates (Quinton et al., 1993). Combined, social behavior through socialization processes?
these limitations lower an antisocial person’s Perspectives from criminology—namely, social
own desirability when looking for his or her learning, social control, and developmental the-
partner, which by default may increase his or ories—can begin to address this question. They
her chances of pairing with a similar individual. illuminate the mechanisms underlying partner
Consensual validation theories, in general, influence by considering how interactions with
may also help explain the mechanisms respon- important others help explain not only the
sible for assortative mating. Individuals are etiology of antisocial behavior but also the
thought to seek out similar others to validate mechanisms by which it changes over time.
their perspective of the world. That is, they For example, social learning theories assume
select compatible environments and companions that children come into the world tabula rasa
supportive of their values and behavior. Such and are, in large part, socialized by their parents’
agentic moves facilitate understanding and com- behavior (Buchanan, 1996) through observation,
munication, increase participation in joint activi- modeling, and reinforcement (Bandura, 1977).
ties, and promote a common lifestyle (Capaldi & Later in adolescence and young adulthood,
Crosby, 1997; Kalmijn, 1994, 1998; Yamaguchi peers begin to play an increasingly important
& Kandel, 1997). Ultimately, ‘‘it is rewarding to role in shaping behavior (Haynie, Giordano,
be around an individual who engages in actions & Manning, 2005). Sutherland’s (1947) differ-
that one regularly enjoys, but unpleasant to spend ential association theory posits that antisocial
time with a person who behaves in ways that one behavior is learned through interactions within
finds foreign and uninviting’’ (Simons, Stewart, adolescent peer groups. An application of this
Gordon, Conger, & Elder, 2002, p. 404). theory would argue that similarity between part-
Genes are thought to influence antisocial ners stems from dating within these homophilous
behavior and to play a role in assortative mat- peer groups, in which favorable definitions of
ing. Mate selection is undoubtedly driven by antisocial behavior are learned and reinforced.
personal preferences, which are often based In addition to social learning theories, social
on phenotypes, or observable, genetically influ- control perspectives posit that individuals are
enced characteristics. Some evolutionary per- more likely to engage in antisocial behavior
spectives, such as genetic similarity theory when their bonds to prosocial people and insti-
(Rushton, 1989), assert that genetically simi- tutions are weak; they are then less influenced
lar individuals detect and prefer one another by prosocial norms (Hirschi, 1971). Conversely,
to ensure reproductive benefits and gene sur- individuals with strong attachments to prosocial
vival (Reynolds, Baker, & Pedersen, 2000). In others avoid jeopardizing those bonds; this pro-
opposition to direct genetic detection theories, cess is key to influencing their behavior toward
Bereczkei, Gyuris, Koves, and Bernath (2002, conformity. An application of this theory might
p. 681) posit a ‘‘genetically canalized learning suggest that when two antisocial individuals are
process’’ of assortative mating based on attach- romantically involved with each other, they, too,
ment processes, whereby children internalize avoid jeopardizing bonds with their mates. Ulti-
their opposite-sex parent’s phenotype as a tem- mately, couples may conform to each other to
plate for acquiring mates with shared genes. ensure that bonds are maintained.
Adult men ‘‘match the mental image of their Thornberry, Lizotte, Krohn, Farnworth, and
mother to females as potential mates, estimate Joon Jang (1991) extended the social control
Assortative Mating and Partner Influence 201

perspective in their interactional theory of delin- place, and, thus, before they are able to influence
quency to posit a developmental approach that each other. However, it is nearly impossible to
views antisocial behaviors as dynamic and time design a study that follows both members from
varying (Thornberry & Krohn, 2005). Antiso- adolescence into late adulthood because, at the
cial behavior in one developmental stage can initiation of a study, few adolescents know with
subsequently influence and reinforce future out- whom they will mate (i.e., have a child).2 This
comes. Movement along an antisocial trajectory problem does not affect other studies that exam-
is related to movement along other life-course ine stable, time-invariant variables. For example,
trajectories and transitions into age-graded roles, research has shown that, by and large, individ-
such as dating, marriage, and family (Elder, uals positively assort on race (Kalmijn, 1998),
1998). Theoretically, an individual’s antiso- which can be measured long before or after
cial trajectory has mutually reinforcing conse- couples have met and does not change.
quences for mate selection. For example, young The preferred design, to collect prospective
adults who have not outgrown normative ado- data on both individuals in a dyad before mate
lescent delinquency will have few conventional selection, is nearly impossible to accomplish
options, and their pool of possible mates will unless an entire population is studied. Alterna-
be limited to other antisocial individuals. In tive research designs could include (a) cross-
turn, mate selection itself should affect subse- sectional studies that collect data after mate
quent antisocial behavior because it is subject selection, which at best measure concordance
to the influence that partners can have on each between dyad members but not assortative mat-
other over time. In summary, the confluence of ing per se; (b) longitudinal studies that use
selection (i.e., assortative mating via structural, prospective early data for the focal subject col-
individual, and genetic factors) and socializa- lected before mate selection and prospective
tion (i.e., partner influence via social learning data for the partner collected after mate selec-
and social control mechanisms), coupled with tion, although this measures only an indirect
the developmental consequences of antisocial association between partners; (c) longitudinal
behavior over time (as proposed by interactional studies that use prospective early data for the
theory) helps explain the similarity seen between focal subject collected before mate selection but
mates. Simply put, similarity may arise from rely on retrospective data for the mate collected
a number of reinforcing factors. For instance, after mate selection; and (d) studies that use
selection into a relationship with a similarly official criminal records, which are intrinsically
antisocial person may be partly a result of the prospective but frequently biased and underesti-
developmental consequences of participating in mate assortative mating.3 Overall, the research
antisocial behavior (e.g., when prosocial ties are designs can sidestep the methodological diffi-
no longer available). In addition, similarity may culty inherent in studies of assortative mating,
increase from the socialization that occurs once
a relationship has been established and antisocial
behavior is reinforced over time, not by peers, 2 The only exception found in the literature is the Simon
but by partners. et al. (2008) study of middle school couples all attending
the same school. Data from a subsample of students (n =
78) who did not have relationships with each other during
EXAMINING THE LITERATURE the first wave of data collection but who were later involved
Although partner similarity makes theoretical in romantic relationships at the second wave were used
to estimate assortative mating. However, the study, like so
sense, it is important to examine what the liter-
many others, suffered from methodological limitations. Only
ature has shown so far. Unfortunately, finding 30% of respondents from the original sample were included,
clear evidence that distinguishes between selec- participants were very young, and the average length of
tion and socialization effects is virtually impos- relationship was 3 months. Moreover, findings from one
sible because antisocial behavior is a dynamic school are not generalizable to the larger population.
3
and time-varying variable (Thornberry & Krohn, Data collected from official records measure response
2005). Despite good intentions, most studies to criminal acts rather than actual antisocial behavior,
much of which is not recorded by the criminal justice
assess some form of partner concordance as a system (Thornberry & Krohn, 2000). In addition, individuals
result of a methodological limitation inherent in (especially women) with a criminal record are less common
these studies: the need to measure the behaviors and are often located at the extreme end of the antisocial
of both individuals before mate selection takes continuum.
202 Journal of Family Theory & Review

but doing so involves methodological trade-offs each area, results are largely sorted by the stud-
that convolute the evidence for assortative mat- ies’ design (e.g., prospective vs. retrospective).
ing for antisocial behavior. In addition, the literature is used to address
With this limitation in mind, the rest of three research questions. Figure 1 corresponds
this article presents a comprehensive litera- to these questions by depicting a longitudinal
ture review, organized into two primary areas: model divided into three sections (i.e., assorta-
assortative mating and partner influence. Within tive mating, contagion, and mutual influence),

FIGURE 1. PRIMARY MEASURES OF PARTNER SIMILARITY.

Time

Before Mate Selection After

t-3 t-2 t-1 t t+1 t+2 t+3

Assortative Mating

ASB
Focal Subject:

ASB
Mate:

Contagion

ASB
Focal Subject:

ASB
Mate:

Mutual Influence

ASB ASB ASB

Focal Subject:

ASB ASB ASB

Mate:

Notes: Solid boxes represent prospective measures of antisocial behavior. Dashed boxes represent retrospective measures
of antisocial behavior. All arrows indicate positive associations.
Assortative Mating and Partner Influence 203

each illustrating one of the primary measures of such cases, attrition and selection issues poten-
partner similarity. In each section of the figure, tially limit the generalizability of the findings.
time (t) is represented from left to right. Mate The types of dyads were usually some combi-
selection is depicted by a solid, vertical box that nation of young couples, cohabiting partners,
divides time into the time before and the time spouses, and parents. Some 40 different types of
after a couple begins a relationship together. antisocial behavior were measured—most com-
Clear boxes represent developmentally specific monly delinquency or crime, substance use, and
antisocial behavior (ASB) for a focal subject psychiatric disorders. In some studies a broad
or his or her mate. Arrows indicate positive construct of antisocial behavior was used that
associations between the partners’ behavior. consisted of several indicators. Later in this text,
To conduct the literature review, the titles these measures are reported simply as antisocial
and abstracts of approximately 250 articles and behavior (e.g., Capaldi, Kim, & Owen, 2008).
books were screened from bibliographies, the Wherever possible, the specific measure of anti-
Internet, and computerized databases. The social social behavior is reported (e.g., prevalence of
science subject area of CSA Illumina was used drug use).
to search 27 databases, including PsycINFO, The goal of most studies included estimat-
Sociological Abstracts, and Criminal Justice ing assortative mating. However, many studies
Abstracts. Key search terms included assor- simply measured partners’ similarity at one time
tative mating, homogamy, partner similarity, point, usually after marriage. Other strategies
antisocial behavior, crime, delinquency, alcohol included correlating retrospective measures that
use, drug use, marriage, romantic relationships, assessed age of onset and were purported to have
and couples. In total, 81 articles and books occurred before mate selection; however, the
were deemed potentially relevant because they limitations of retrospective measures have been
examined some form of partner similarity; they well documented (Yarrow, Campbell, & Burton,
were obtained for further review. This strat- 1970). Some of the better estimates involved
egy kept the search broad. However, 55 studies associations between one partner’s adolescent
were eliminated because they (a) were older and the other partner’s adult measures, a method
than 30 years, (b) relied on a single respondent referred to in this text as a contagion proxy. The
to assess both members of the dyad’s behavior, most comprehensive studies included estimates
(c) measured only general personality or mild of (a) partner similarity, (b) assortative mating
affective disorders, (d) sampled fewer than 25 measured using the contagion proxy or retro-
respondents, (e) used a biased sampling strat- spectively, and (c) partner influence. Findings
egy (e.g., treatment sample without controls, from the three types of assessments are compared
or snowball sample), (f) were review articles, to determine whether partner similarity is a func-
or (g) were duplicate studies (with similar data tion of selection, socialization, or some combina-
already reviewed here). No qualitative studies tion of both. To paint the clearest picture possible
were found. In the end, 26 studies were retained while still being descriptively succinct, the fol-
for in-depth review. lowing details are described whenever a specific
These studies can be divided into four study is discussed: sample size, participants’
main types: (a) life-course studies that followed ages, relationship length, dyad type (e.g., dating
either community-based or high-risk adoles- couples, biological parents), study design (e.g.,
cents through adulthood, (b) shorter school- prospective, retrospective), type of data analyzed
based studies of adolescents, (c) marriage and (e.g., self-reports, official records), analytic
substance-use studies, and (d) studies primarily method (e.g., correlations, logistic regression),
using diagnostic interview schedules. Studies findings when significant, and statistical controls
with diagnostic measures included community- and gender differences if reported.
based samples, treatment patients with matched The following six sections of this article
controls, and designs involving twins and their summarize the findings related to assortative
spouses or parents. Many of the studies were mating only. Studies are organized by their
longitudinal in design, but their analyses in pub- design features (i.e., studies with prospective
lications were based on data collected from only or retrospective measures, studies comparing
one time point. Sample sizes ranged from 79 data from different developmental stages, studies
to 4,318 dyads. Often, studies used subsamples without controls for social homogamy, studies
of dyads from larger, longitudinal studies. In of partners’ parents’ history of alcoholism,
204 Journal of Family Theory & Review

and studies comparing heterotypic behavior). minimize the inclusion of partner influence
Descriptions of these studies are often shorter effects, as those couples have had less time
than descriptions in the subsequent section to influence each other. Next are findings from
on partner influence because the analyses are four prospective studies that represent young
generally less complex (e.g., correlations). couples in relatively new relationships, listed
from youngest subjects to oldest.
INFLUENCE OF ANTISOCIAL BEHAVIOR
ON MATE SELECTION
Young couples and prospective measures.
Among the youngest couples, Furman and
This section begins by addressing the first Simon (2008) estimated partner similarity in
of three research questions: Are respondents’ adolescent relationships of 3 months or longer
histories of antisocial behavior (measured in duration (n = 83 dyads,4 mean age = 15
prospectively before mate selection) correlated years, average relationship length = 11 months,
with their mates’ histories of antisocial behavior SD = 9.26) and, using the Achenbach Youth
(measured retrospectively after mate selection) Self-Report, found significant concordance on
and, if so, in what direction? This research externalizing (r = .23, p < .055 ) but not on
question involves examining studies that relied internalizing behaviors. With a slightly older
only on prospective data for one mate and sample, Capaldi and Crosby (1997) also found
retrospective data for the other (see the first significant homophily among a young sample
section of Figure 1). Studies with this design (n = 118, mean age = 18, average relation-
can compensate for the inherent inability to ship length = 11 months), using a construct of
measure the behaviors of both individuals prospective antisocial behavior that included
before mate selection, as discussed previously. official offenses, self-reported delinquency, and
Methodologically, this design is perhaps the observed aggression (r = .44). These couples
best bet for estimating assortative mating, given were in relatively less stable relationships; only
that antisocial behaviors vary over time and 4% were married, whereas the rest were living
are susceptible to the influence of others. together (26%) or dating (70%).
Unfortunately, none of the studies examined Krueger, Moffitt, Caspi, Bleske, and Silva
assortative mating with this type of data. This (1998) used the Dunedin sample to study assor-
means that no studies used the best possible tative mating among 360 young adult couples
design and evidence for assortative mating is (mean age = 21 years) in long-term relation-
further limited, which is surprising because such ships (average relationship length = 26 months,
a design is feasible. 48% married or cohabiting). To reduce measure-
ment error, confirmatory factor-analytic models
Assortative Mating estimated the similarity of partners on three
indices of antisocial behavior. Results indi-
Although evidence was not available to answer cated that couples were highly assorted on
the first research question, 25 of the 26 studies self-reported antisocial behavior (.54) and peer
assessed the similarity of partners’ antisocial delinquency (.54). Dyads were only moder-
behavior after the couple had already met. All ately assorted on five attitudinal measures of
reported significant associations, which provides the consequences of antisocial behavior (aver-
overwhelming support for partner similarity. age r = .32). Even less assortment was found on
However, this strategy is only a crude proxy two personality correlates of antisocial behavior,
for assortative mating because estimates are negative emotionality (.17) and constraint (.13).
contaminated with the influence couples have Last, Leonard and Mudar (2003) estimated
on each other. With one exception (Simon, partner similarity among relatively older, new-
Aikins, & Prinstein, 2008), couples had already lywed couples (n = 519, mean age = 27 years,
been together for some time when at least
one partner’s antisocial behavior was measured.
Given that relationship length and age are 4
All samples sizes represent dyads (e.g., a sample of
generally correlated, findings from studies using n = 83 dyads involves 166 individuals).
somewhat older, established couples are even 5
All correlations are significant, p < .05. To save space,
more muddled by partner influence effects. In specific p values are not reported. Likewise, 95% confidence
contrast, studies with younger samples could intervals are not reported.
Assortative Mating and Partner Influence 205

median relationship length = 18 months) and twins and their spouses on self-reported alcohol
found significant covariate-adjusted correla- consumption (n = 2,897) and cigarette smok-
tions on self-reported antisocial behavior (r = ing (n = 914). Significant tetrachoric correla-
.23), depressive symptoms (r = .13), alco- tions were found for lifetime prevalence of
hol involvement (r = .47), and peer drinking nicotine dependence (r = .31), regular smok-
(r = .74), but not on parental alcoholism. In ing (r = .30), alcoholism (r = .15), and regular
summary, these four studies demonstrated sig- drinking (r = .38). Likewise, Maes, Neale et al.
nificant partner similarity on a variety of anti- (2006) found a significant tetrachoric correlation
social behaviors and their correlates (e.g., peer between twins and their spouses (n = 4318)
delinquency and drinking, depression, attitudes, on lifetime self-reported smoking (r = .38).
personality). Couples in relatively new relation- Altogether, these studies demonstrate signifi-
ships were examined, which minimizes partner cant partner similarity, and their large sample
influence confounds. Note, however, that rela- sizes lend additional strength to the findings.
tionships in adolescence and young adulthood However, because of their reliance on lifetime
are often short term and experimental (Haynie prevalence measures, they did not directly assess
et al., 2005). The median age at first marriage assortative mating per se.
in the United States is approximately 25 years These estimates would be more compelling
for women and 27 years for men (Fields, 2004). if the authors had ascertained age of onset and
The mean age of mothers at first birth is in the could isolate reports of antisocial behaviors to
range of 25 years (Mathews & Hamilton, 2002). adolescence only. Two such studies used this
Ultimately, estimates based on young couples somewhat improved method. Taylor, McGue,
may be of less consequence; these relation- and Iacono (2000) found significant parental
ships are more likely to dissolve, and antisocial correlations (using structural equation mod-
behaviors are subject to maturational effects. els) for retrospectively assessed, self-reported
Altogether, the findings indicate that partner adolescent delinquency (n = 486; r = .23 and
similarity among samples of young couples is a r = .35, for men and women, respectively).6
common but imperfect proxy that demonstrates In addition, Maes, Silberg, Neale, and Eaves
support for assortative mating. (2007) found similarity between biological par-
ents (using maximum-likelihood estimations of
Twin studies and retrospective measures. In correlations) on retrospective measures of con-
addition to using relatively young couples to duct disorder occurring before age 18 (n = 920;
prospectively estimate partner similarity, some r = .18). Although retrospective reports have
studies used data from twins and their families, in their limitations, these two studies provide bet-
which reliance on cross-sectional, retrospective ter estimates of assortative mating because they
measures was more common. Frequently, the attempted to measure behaviors that occurred
goal of these studies was to improve heritability well before parents had met each other. In
estimates; less attention was given to precisely summary, these findings indicate that signifi-
estimating assortative mating. None of the twin cant partner similarity was found on a variety of
studies reported the age of dyad members or antisocial behaviors (including psychiatric disor-
relationship length. Nevertheless, findings from ders, substance use and abuse, and delinquency),
five twin studies are reported here because they all retrospectively measured before couples had
constitute a body of research that attempts to met.
address assortative mating, primarily in the
context of genetic studies. Contagion proxy. A better method implemented
Foley et al. (2001) reported similarity in many of the life-course studies involves
between parents’ (n = 544) self-reports on life- testing for an association between focal subjects’
time prevalence of psychiatric disorders, specifi- adolescent behavior and their partners’ adult
cally depression (2.4%, confidence interval [CI] behavior, referred to here as a contagion
[1.6, 3.9]), depression comorbid with paternal
alcoholism (1.3%, CI [0.6, 2.3]), and heterotypic
diagnoses of maternal depression and paternal 6 Hicks, Krueger, Lacono, McGue, and Patrick (2004)
alcoholism (1.5%, CI [0.8, 2.6]) or paternal sim- also analyzed the data and found significant parental
ple phobia (1.4%, CI [0.7, 2.5]). Agrawal et al. similarity on general vulnerability to externalizing disorder
(2006) also retrospectively measured female (r = .51, CI = .41–.61).
206 Journal of Family Theory & Review

proxy (see the second section of Figure 1). more (using teacher reports and self-reports).
Life-course studies are uniquely situated to They used a retrospective measure of their
use this technique because the focal subjects focal subject’s childhood conduct disorder
are followed prospectively through adolescence and, using logistic regression, found that it
well into adulthood. The partner then joins the was associated with having a deviant partner,
study after the relationship has already begun prospectively measured in adulthood (odds ratio
and, therefore, can provide only prospective [OR] = 2.42, CI [1.39, 4.49]). Altogether,
measures of his or her current behavior. the four life-course studies provide support
Findings from three such studies are presented for the argument that birds of a feather flock
to answer the following research question: What together. Although using a contagion proxy
effect does a respondent’s history of antisocial to estimate assortative mating is limited by
behavior (measured prospectively before mate its reliance on measures assessed in different
selection) have on his or her mate’s current developmental stages, it is perhaps the best bet,
antisocial behavior (measured prospectively given the research to date. Through prospective,
after mate selection)? longitudinal measures, these findings indicate
In their life-course study, Simons, Stewart that focal subjects’ behavior is indeed related,
et al. (2002) assessed young couples (n = 236, albeit indirectly, to their partners’ behavior on
mean age = 21 years, relationship length not measures of delinquency, crime, violence, and
reported) in continuing and exclusive romantic physical abuse. It is important to note, however,
unions (56%), cohabiting partnerships (22%), that despite several significant associations, most
or marriages (21%). Using self-report data, couples do not assort on antisocial behavior.
the focal subjects’ adolescent delinquency was
significantly correlated with an equivalent mea- Social homogamy. In the evidence presented so
sure of their partners’ adult criminal behavior far, partners appear similar to each other on mea-
(r = .30 and r = .47 for focal men and women, sures of antisocial behavior, whether reported
respectively). Structural equation modeling pro- early in the relationship, retrospectively, or in
vided comparable results (.38 and .26). In a different developmental stages. However, assor-
similar assessment of self-report data, Moffitt, tative mating may not be the only mechanism
Caspi, Rutter, and Silva (2001) reassessed the responsible for partner similarity. Perhaps cou-
Krueger et al. (1998) sample of young couples ples share similar demographic characteristics or
(n = 360, relationship length not reported) but social environments that increase the probabil-
this time correlated the focal subjects’ adoles- ity for antisocial behavior. Research has shown
cent antisocial behavior with their adult partners’ significant evidence for social homogamy in
past-year delinquency (r = .33 and r = .33), general (for reviews, see Kalmijn, 1994, 1998;
violence (r = .27 and r = .34), perpetration of McPherson et al., 2001). However, only seven
physical abuse against focal subjects (r = .32 of the cross-sectional studies and none of the
and r = .26), and criminal charges (r = .21 and prospective studies found for this literature
r = .38) (all for men and women, respectively). review (n = 26) controlled for variables related
Kim and Capaldi (2004) also reassessed Capaldi to social homogamy (e.g., age, race, education,
and Crosby’s (1997) sample using a smaller sub- socioeconomic status, religion) in their analyses.
sample (n = 79, initial mean age = 21 years, When included, these variables tended to reduce
mean relationship length = 4.5 years) of pre- but did not eliminate associations found between
dominately married and cohabiting couples who partners’ antisocial behavior. For example,
remained together at their second and third McLeod (1995) used logistic regression to
waves of assessments. Using a multimethod, examine a community-based sample (n = 586,
multiagent construct of antisocial behavior, the age and relationship length not reported) and
focal subjects’ adolescent behavior significantly found significant concordance among spouses
correlated with their partners’ adult antisocial on retrospective, self-reported drug dependence
behavior (r = .23). In summary, these three (OR = 3.34). However, adding age and educa-
studies provide support for the contagion proxy. tion controls reduced odds ratios from 3.34 to
In a slightly different version of this 2.50. Sakai et al. (2004) showed that, among
technique, Quinton et al. (1993) assessed focal a sample of parents with children in treat-
subjects and their partners (n = 319) in their ment who were matched with community
mid-20s who were cohabiting for 6 months or controls (n = 357, mean age = 44, duration
Assortative Mating and Partner Influence 207

not reported), correlations of retrospective, alcohol-use disorders or major depression, net of

self-reported substance dependence and anti- sociodemographic factors. In addition, Leonard
social personality disorder symptoms decreased and Mudar (2003; n = 519; mean age = 27
but remained significant when adjusted for age years, median relationship length = 18 months)
and race (e.g., r = .40 to r = .38, r = .33 did not find a significant correlation on self-
to r = .28, respectively). In summary, social reported parental alcoholism. Overall, there is
homogamy is likely to inflate assortative mating little evidence to suggest that couples assort on
estimates. Studies need to include demographic the basis of their parents’ history of alcoholism;
controls to disaggregate social homogamy from two studies found support, whereas two studies
assortative mating findings. did not. Consider, however, that findings are
mired by reliance on a single respondent’s
Parents’ history of alcoholism. Related to social retrospective reports or results that report an
homogamy are studies on couples’ family back- indirect association (i.e., similarity between
ground characteristics. These studies did not parent’s and spouse’s alcoholism).
measure assortative mating as it is defined here
but are reported because they provide evidence Heterotypic assortment. Findings regarding
that there is something similar in the back- assortative mating based on the family’s his-
grounds of individuals that brings them together. tory of antisocial behavior are inconclusive.
This proxy for assortative mating is not precise However, family background characteristics are
but provides an indirect measure of behaviors or often implicated in the continuity of antisocial
characteristics that occurred well before part- behavior across the life course. Persistent anti-
ners met. Altogether, four studies examined social behavior often starts in early childhood
concordance on family history of alcoholism. (Moffitt, 1997), sometimes as a manifestation
Note, however, that dyad members reported on of coercive parent-child interactions (Patterson,
their parents’ behavior (i.e., investigators relied Capaldi, & Bank, 1991). Heterotypic continuity
on a single respondent); thus, results should in antisocial behaviors (e.g., poor temperament
be interpreted with caution. Among a sample in childhood, delinquency in adolescence, sub-
of newlyweds (n = 416, mean age = 23 years, stance use in adulthood) may actually represent
relationship length not reported), Boye-Beaman, different expressions of the same latent trait over
Leonard, and Senchak (1991) used logistic the life course (Sakai et al., 2004). Therefore, it
regression to find a positive association between is important to determine whether couples are
spouses’ retrospective, self-reports of having assorting on behaviors that appear different but
an alcoholic parent, net of sociodemographic are similarly antisocial (e.g., finding an associ-
characteristics (OR = 1.38 and 1.46, for men ation between an individual’s criminal behavior
and women, respectively). In addition, Gleiber- and his or her partner’s drug use). The seven
man, Harburg, DiFranceisco, and Schork (1992) studies that examined heterotypic concordance
examined a community sample of spouses (n = are discussed in the following paragraphs.
184, mean age = 40 years, relationship length Two prospective studies report significant
not reported) and found significant retrospec- heterotypic concordance. Kim and Capaldi
tive associations between wives’ self-reports of (2004; n = 79, initial mean age = 21 years,
alcohol use and husbands’ reports of mother’s and mean relationship length = 4.5 years) found
alcohol use (only x 2 = 15 reported), but corre- significant heterotypic similarity between long-
lations between both spouses’ parental history term partners’ measures of antisocial behavior,
of alcohol use were not reported. In summary, depressive symptoms, and physical and psy-
two studies found indirect support for partner chological aggression using correlations and
similarity based on family history of alcohol use. multimethod and multiagent constructs (r =
However, the following two studies did not .25 − .56). With self-report data and structural
find support for concordance on parental history. equation models, Leonard and Mudar (2003;
In a longitudinal study of male university n = 519; mean age = 27 years, median rela-
students and employees (n = 327, mean age = tionship length = 18 months) found that, at
35 years, mean relationship length = 8 years) marriage, husbands’ alcohol involvement was
using logistic regression, Schuckit, Smith, Eng, associated with wives’ peer drinking (.34), and
and Kunovac (2002) did not find concordance vice versa (.25). However, antisocial behavior
on spouses’ retrospective self-reports of parental was not associated with either variable.
208 Journal of Family Theory & Review

Five studies using retrospective diagnostic of psychiatric disorders (including depression;

interviews examined heterotypic similarity, but anxiety; substance use; and phobic, conduct, and
only three found support for cross-concordance. antisocial personality disorders). These findings
Galbaud du Fort, Boothroyd, Bland, Newman, indicate that the evidence for heterotypic assort-
and Kakuma (2002; n = 519, mean age = 45 ment is minimal. It is unclear whether couples
years, relationship length not reported) found are more likely to assort on behaviors that appear
that wives’ depression was associated with different but are similarly antisocial. Ultimately,
husbands’ antisocial behavior, net of other there is not yet enough evidence to suggest
significant diagnoses and age (OR = 2.16, CI that, within a couple, differences in the types of
[1.11–4.19], using contingency tables). Using antisocial behavior committed by each partner
logistic regression, McLeod (1995; n = 586, actually represent two distinct expressions of the
age and relationship length not reported) same latent antisocial trait.
found that among lifetime and premarital
diagnoses of anxiety, depression, and substance
dependence, the only heterotypic concordance Limitations: Different Types of Relationships
was between anxiety and substance dependence and Definitions of Mates
(adjusted OR = 2.05, CI [1.28, 3.38] and Although support for heterotypic assortment was
OR = 2.49, CI [1.51, 4.09], for wives’ and weak, other factors complicate the degree to
husbands’ risk, respectively). Foley et al. (2001; which support for assortative mating could be
n = 544, age and relationship length not found. For instance, the definitions of mates
reported) assessed similarity on parents’ lifetime were often conflated, thus making it difficult
prevalence of psychiatric disorders and found to ascertain specific evidence for couples in
parental similarity on heterotypic diagnoses of different types of relationships, which ranged
maternal depression and paternal alcoholism from adolescent couples to biological parents.
(1.5%, CI [0.8, 2.6]) and of maternal depression In addition, relationship types ranged widely
and paternal simple phobia (1.4%, CI [0.7, in each sample. For example, Simons, Stew-
2.5]). Altogether, three studies found support art et al. (2002) included dating, cohabitating,
for heterotypic similarity using retrospective and married couples in their sample. Despite the
diagnostic interviews. broad use of the term assortative mating, the only
In contrast, two studies did not find support studies that actually examined biological parents
for cross-concordance. Sakai et al. (2004) did not (mates, in the literal sense) were the twin and
find evidence of heterotypic assortment among treatment-based designs (n = 4). Furthermore,
a sample of parents with children in treatment, Maes, Silberg et al. (2007; age and relation-
who were matched with community controls ship length not reported) was the only study to
(n = 357, mean age = 44 years, duration of rela- assess differences between biological and non-
tionship not reported). Among measures of con- biological parents. On retrospective measures of
duct disorder, substance dependence, and antiso- conduct disorder, significant correlations were
cial personality disorders, as well as their respec- found for twin families with both biological par-
tive symptom counts, significant homotypic ents (n = 920; r = .18) but not for families with
correlations were found among both treatment a nonbiological father (n = 94). Such distinc-
and control parents (r = .16–.40), with only a tions are important because assortative mating
few exceptions. However, heterotypic associa- (and its consequences) may be greater among
tions were no longer significant when multiple couples who are parents or involved in some type
regression analyses controlled for homotypic of long-term relationship. Conversely, antisocial
behavior. Likewise, Low, Cui, and Merikangas individuals are more likely to become parents at
(2007; n = 255, mean age = 40 years, average a younger age (Thornberry, Krohn, Lizotte, &
length of marriage = 14 years), using logis- Smith, 1998) and less likely to attain supportive
tic regression analyses, did not find support and stable relationships (Quinton et al., 1993).
for heterotypic, spousal similarity among sub- Sakai et al. (2004; n = 357, mean age = 44
stance use and anxiety disorders. In summary, years, relationship length not reported) found
only four of the seven studies that reported significantly higher correlations (using Fisher’s
on heterotypic concordance found significant Z transformation) on self-reported symptoms of
similarity between partners. Behaviors varied antisocial personality disorder between biolog-
but were predominately retrospective measures ical parents who were no longer married than
Assortative Mating and Partner Influence 209

for married biological parents (coefficients not relationship is either short lived, sometimes
reported). Unfortunately, most studies (n = 12) resulting in a child, or involves some type of
estimated assortative mating using marital sam- ongoing partnership. The next section examines
ples. Thus, the degree to which antisocial indi- the extent to which the literature shows that
viduals and antisocial parents, in particular, partners are similar because they are able to
exhibit assortative mating may be underesti- influence each other over time.
mated.
INFLUENCE OF MATE SELECTION
Summary of Assortative Mating Findings ON ANTISOCIAL BEHAVIOR

Several important findings emerge from the Despite several methodological issues, this
literature. Overall, assortative mating for anti- review reports substantial similarity between
social behavior is substantial, if the definition couples on a variety of antisocial measures.
of assortative mating is relaxed to include After the initial assortment, however, other
assessments of partner similarity based on mechanisms responsible for partner concordance
(a) prospective measures among young couples include contagion and mutual influence. Defined
in new relationships, (b) retrospective measures here, contagion is the unidirectional effect that
of lifetime prevalence, (c) retrospective mea- one partner has on the other. Estimating it
sures of adolescent behavior, and (d) prospective involves using measures from different devel-
measures of focal subjects’ behaviors in ado- opmental periods. For example, respondents’
lescence and a prospective measure of the adolescent drug use is used to predict their part-
partners’ behavior in adulthood (i.e., conta- ners’ adult drug use. Although the influence
gion proxy). Theoretically, an alternative (and is indirect (because the respondents’ behav-
perhaps methodologically better) examination ior is measured well before they meet their
would involve (e) correlating the focal subjects’ partners), it has the advantage of isolating unidi-
prospective adolescent behavior with their part- rectional effects. Findings regarding contagion
ners’ retrospective behavior, but unexpectedly, were described earlier because several studies
such studies do not exist. Surprisingly, (f) few used estimates of contagion as a proxy for assor-
studies control for social homogamy in their tative mating. To review, four studies provided
estimates; those that do reduce but do not elimi- support for the contagion proxy on a number
nate assortative mating associations. In addition, of measures, including delinquency, crime, vio-
(g) similarity on parents’ history of alcoholism lence, and physical abuse.
is not evident, and (h) support for heterotypic Mutual influence, in contrast, is the bidi-
assortment is minimal. Last, (i) assortative mat- rectional effect that occurs when both partners
ing may be underestimated because antiso- influence each other (see the third section of
cial individuals are underrepresented in marital Figure 1). This distinction is important because
samples. mates will appear more similar than they actu-
On the whole, partners are similar on a number ally were before mate selection. Sorting out
of antisocial behaviors (including delinquency, these effects is essential to fully understanding
crime, substance use, and psychiatric disorders). the consequences of mate selection. Therefore,
Nevertheless, the methodological issues dis- in addition to examining how antisocial behavior
cussed earlier (i.e., the need to measure behav- influences mate selection via assortative mating
iors of both individuals before mate selection, processes, a second goal of this article is to con-
the age and length of relationships commonly sider how partners, through mutual influence,
assessed, retrospective data, contagion proxies affect each other’s antisocial trajectories once
comparing behaviors in different developmen- mate selection has occurred. Findings from the
tal stages, studies without controls for social literature are presented to answer the follow-
homogamy, minimal support for heterotypic ing question: After mate selection, what effect
similarity, and underrepresentation of antiso- does respondents’ current antisocial behavior
cial individuals in marital samples) weaken the (measured at time + 1) have on their mates’
evidence for assortative mating theory. antisocial behavior (measured at time + 2) and,
However, assortative mating is just one of in turn, what effect does the mates’ current anti-
the possible mechanisms influencing partner social behavior (measured at time + 2) have
concordance. After mating selection occurs, the on the respondents’ later antisocial behavior
210 Journal of Family Theory & Review

(measured at time + 3), controlling for prior (b = −.27, p < .05). Her T1 antisocial behav-
antisocial behavior? ior influenced his T2 psychological aggression
toward her when his T1 antisocial behavior was
low but not when his T1 antisocial behavior
Mutual Influence
was high. His behavior at T1 did not influ-
After an examination of the literature, no ence her behavior at T2. In a follow-up study
studies of mutual influence using the definition (Capaldi et al., 2008), female partners’ antiso-
described in this research question were cial behavior also significantly predicted male
found. Few studies had prospective longitudinal focal subjects’ arrests.7 Overall, these findings
data on both partners over several time suggest that women can, at times, influence the
periods, which means that establishing temporal psychological aggression and criminal behavior
order is difficult and the ability to interpret of their male partners over time.
the causal direction of influence is reduced. Second, Leonard and Mudar (2003) used
However, 15 studies indirectly estimated partner prospective data (n = 519, mean age = 27
influence, each using one of four methods: (a) 3 years, median relationship length = 18 months)
studies used prospective longitudinal data to to fit structural equation models and found sig-
estimate time-lagged, cross-partner associations; nificant associations between newlywed spouses
(b) 2 longitudinal studies assessed concurrent on T1 self-reported alcohol involvement at the
associations between partners’ behavior after beginning of marriage (.46). In addition, hus-
mate selection while controlling for prior bands’ T1 alcohol involvement predicted their
antisocial behavior; (c) 2 studies examined wives’ T2 alcohol involvement 1 year later (.15)
prospective, cross-sectional data; and (d) 8 but to a lesser extent. Here, men, but not women,
retrospective studies estimated the extent to influenced their partners’ behavior.
which similarity increased over time. The Using a different method, the third study
findings from the 15 studies are outlined in estimated time-lagged associations using data
the following section. For each study, the collected before and after mate selection, which
assortative mating (loosely defined) findings are is similar to the way that a contagion proxy is
summarized. Then, estimates for influence are assessed. However, the findings are described
described. Finally, the mechanisms most likely here because it is the only study in which data
responsible for partner similarity—assortative were collected before couples actually met and,
mating, influence, or combinations of both—are more important, because findings for influence
discussed. Gender differences are discussed are stronger than for assortative mating. In more
when reported. detail, Simon et al. (2008) prospectively studied
adolescent couples (n = 79, age not reported,
Prospective longitudinal data. Three studies mean relationship length = 13.63 weeks, SD =
estimated time-lagged, cross-partner associa- 19.10) in one middle school, Grades 6–8. The
tions occurring after mate selection using only measure that couples positively assorted on
prospective, longitudinal data. First, Kim and (as shown by assessing intraclass correlations at
Capaldi (2004) found that partners of various T1 before dating) was self-reported sadness (r =
types (n = 79, mean age = 21 years, mean rela- .38). However, for peer-rated sadness (b = .52,
tionship length = 4.5 years) were similar on
adult antisocial behavior and depressive symp-
toms, as well as on a number of other het- 7 In this follow-up study (Capaldi et al., 2008), using data
erotypic characteristics, using multimethod and
from a 12-year period (n = 119–158, mean age range:
multiagent data. In addition, the authors used 18–28 years, mean relationship length: 49–216 weeks,
hierarchical regression models to estimate het- depending if Wave 1–5), female partners’ T1 antisocial
erotypic mutual influence. Controlling for other behavior significantly predicted male focal subjects’ T2
behaviors, female partners’ depressive symp- arrests (counts and prevalence within a 1-year period), net of
toms (but not antisocial behavior) at Time 1 (T1) arrest history, deviant peer associations, age, attachment to
significantly predicted male focal subjects’ psy- female partner, both partners’ substance use and depressive
symptoms, and relationship stability. The relationships held
chological (but not physical) aggression toward even when models were run separately for men at risk for
her at Time 2 (T2; b = .24, p < .05). In addition, persistence (i.e., with at least one arrest at T1) and for men at
a significant, negative interaction term was found risk for onset in young adulthood (i.e., with no prior arrests
between the partners’ T1 antisocial behaviors at T1).
Assortative Mating and Partner Influence 211

p < .01), relational aggression (b = .29, p < Second, Moffitt et al. (2001) reported sim-
.01), relational victimization (b = .25, p < .05), ilar findings (n = 360, mean age = 21 years,
and self-reported depression (b = .23, p < .05) relationship length not reported). As already
(but not physical aggression and physical discussed, evidence for assortative mating was
victimization), hierarchal regression models found with a contagion proxy (r ranging from
found that significant interactions between .21 to .38) among couples of various types. Then,
adolescents’ T1 and their partners’ T1 behavior using hierarchical multiple regression, partners’
predicted adolescents’ later T2 behavior, but adult antisocial behavior predicted focal sub-
only when their partners’ T1 behavior was jects’ adult antisocial behavior, net of adolescent
high. Overall, this finding suggests little behavior, for both men and women (b = .33,
assortative coupling, but significant (though t = 4.87 and b = .30, t = 5.29, respectively).
indirect, as couples did not know each other However, an adolescent antisocial behavior ×
at T1) influence effects; adolescents who were partners’ adult antisocial behavior interaction
initially low changed more as a result of their effect occurred for women (b = .18, t = 2.25),
partners’ behavior. Gender differences were not but not men, again suggesting that the rela-
examined. tionship between adolescent and adult antisocial
In summary, findings from these three behavior is stronger for women with antisocial
prospective, longitudinal studies indicate that partners. For men, there is continuity in their
influence processes are at play on some but not behavior regardless of their partners’ behavior.
all of the behaviors measured—women influ- In summary, findings from the two longitudinal
enced men’s psychological aggression and crim- studies add to the overall evidence that men are
inal behavior; men influenced women’s alcohol more likely than women to influence their part-
use; and young adolescent couples influenced ners, especially when their partners are antisocial
each other’s sadness, depression, and relational in adolescence.
aggression and victimization. However, none
Prospective cross-sectional data. Just two
of the studies controlled for prior antisocial
studies included here used only prospective,
behavior.
cross-sectional data.8 First, Haynie et al. (2005)
In addition to the three time-lagged, cross- analyzed Add Health (self-report) data on
partner analyses, two additional studies esti- adolescents and their partners (n = 2945 and
mated concurrent associations while controlling 1321, mean age = 16 years, mean relation-
for prior antisocial behavior. First, Simons, ship length = 9 months) using negative bino-
Stewart et al. (2002) fit structural equation mod- mial regression. Although peers’ delinquency
els (n = 236, mean age = 21 years, relationship had a stronger effect, romantic partners also pre-
length not reported) and found that, for male dicted each other’s delinquency for both minor
focal subjects, self-reported adolescent delin- (b = .1) and serious delinquency (b = .1) mod-
quency predicted having an antisocial partner as els, net of control variables associated with
a young adult (.38 contagion proxy). In turn, social homogamy (race, socioeconomic status,
having an antisocial partner in adulthood pre- and family structure). When considering gender
dicted criminal behavior (.21). For female focal interactions, again, young men (but not women)
subjects, a similar effect was found (.26 and influenced their partners’ minor delinquency
.45, respectively), but the coefficient for adult (b = 0.08). 9
women was twice as large as that for men, In the second study, Herrera, Wiersma,
which suggests that romantic relationships exert and Cleveland (2008) analyzed Add Health
more influence on women’s criminal behav-
ior. In addition, using ordinary-least-squares
regression, a significant main effect of antiso- 8 Arguably, the studies estimated partner similarity rather

cial partner on criminal behavior was found for than partner influence but are still described in this section
female focal subjects only (b = .44, t = 8.29), because they are the only prospective, cross-sectional studies
as was a significant delinquency × antisocial that included regression and interaction models rather than
correlations to estimate associations.
partner interaction effect (b = .23, t = 3.01). 9 The coefficient b = 0.08 = 1 − (exp (.08)) = 8% means
These findings show that the association between that each standard-unit-deviation increase in the male
adolescent delinquency and adult crime is strong partner’s behavior translates into an 8% increase in the
when women have antisocial partners. female focal subject’s behavior.
212 Journal of Family Theory & Review

(self-report) data from the Romantic Pairs presented earlier in their use of time-lagged,
subsample (n = 1275, mean age = 22 years, cross-partner estimates occurring after mate
relationship length not reported). Not surpris- selection. However, the authors relied on ret-
ingly, their findings from hierarchal regression rospective measures collected after marriage
models show a positive association between that reported on behaviors occurring before
partners’ intimate partner violence (IPV) for marriage. Specifically, the authors examined
both men and women (b = .20 and b = .28, 545 husband-wife dyads (mean age = 29, aver-
p < .001, respectively). In addition, women’s age relationship length = 5 years) and initially
general violence was positively associated with found support for premarital assortment on
their perpetration of IPV, but only for women in self-reported illicit drug use (κ = .34). Then,
relationships with men who had perpetrated IPV they compared (less conservative) log-linear and
against them (b = 2.79, p < .01). However, (more conservative) latent trait models (that con-
men’s general violence was also positively asso- trolled for individuals’ latent predispositions for
ciated with their perpetration of IPV, but only for drug use). Results from the less conservative
men in relationships with women who did not analyses revealed T1 similarity (.24; before mar-
perpetrate IPV against them (b = .40, p < .01). riage), even greater T2 similarity (.44; within the
Overall, this study demonstrates the subtle but last year of marriage), and, last, marital influ-
gendered nature of partner influence, at least in ence of wives’ T1 drug use on husbands’ T2
the context of partner violence. However, it is drug use (.14)—all of which suggested support
unclear whether violence is conditioned by part- for both assortative mating and influence. How-
ners’ behavior and gender or, instead, whether ever, results from the more conservative models
assortative mating is at play. Altogether, results did not provide support for influence. Following
from the two prospective, cross-sectional studies up 5 years later with the same couples, Yam-
should be interpreted with caution; the temporal aguchi and Kandel (1997) essentially replicated
order and direction of influence by gender is not their findings, looking at marijuana use only.
clear. Overall, the authors argued that assortative mat-
Nevertheless, all seven prospective studies ing rather than marital influence accounts for
showed that partners are similar on antisocial partner similarity on drug use.
behavior. The mechanisms responsible for con- The third study, McLeod (1995), used logis-
cordance most likely include assortative mating, tic regression to examine a community-based
but partner influence, too, appears to play an sample (n = 586, age and relationship length
important role (in sadness, depression, psycho- not reported) and found significant concordance
logical aggression, relational aggression and within marital couples with retrospectively diag-
victimization, minor and serious delinquency, nosed anxiety and alcohol or drug dependence
crime, violence, IPV, and antisocial behavior). (OR = 2.05–3.34). Similar analyses confined
When considering gender, four of the seven stud- to behaviors reported to have occurred before
ies provided evidence suggesting that, compared marriage supported premarital assortment. In
to women, men are more likely to influence their addition, analyses of parental and childhood risk
partners’ behavior. factors predicting partners’ psychiatric problems
provided indirect support for assortment but not
Retrospective cross-sectional data. In addition conclusively. Three different tests of marital
to the seven prospective studies already pre- influence were not significant. Overall, findings
sented, eight studies examined partner influence revealed more support for assortative mating
using retrospective measures, and their strate- than for marital influence.
gies varied widely. However, all the studies In the fourth study, Maes, Neale et al.
in one way or another addressed the question (2006), using retrospective prevalence data
of whether partners become more similar over from twins and their families (n = 4,318,
time. Increased similarity, as a function of length age and relationship length not reported),
of time spent together, was taken as evidence for found significant spousal correlations on self-
partner influence. In summary, they all found reported smoking (r = .38) but ruled out
support for assortative mating, but only one spousal influence as a mechanism of similarity
study reported support for influence. because the association between spouses was not
Yamaguchi and Kandel’s (1993) analytic mediated by duration of marriage (findings not
strategy was similar to that of the studies reported). In the fifth study, Agrawal et al. (2006;
Assortative Mating and Partner Influence 213

n = 914 and 2,897, age and relationship length (OR = 4.96, CI [.88, 27.8]), and childhood
not reported) also fit twin models and found little conduct disorder (CCD; OR = 4.02, CI [2.03,
support for marital influence using retrospective, 7.96]). In adulthood, similarity was found for
self-report measures of substance abuse. antisocial behavior (AAB; OR = 20.1, CI [5.97,
The sixth study, by Price and Vandenberg 67.5]) but not for antisocial personality disorder
(1980), demonstrated concordance in married (APD = CCD + AAB), which is a life-course
couples (n = 134, mean age = 44 years, rela- measure of antisocial behavior from childhood
tionship length not reported) on retrospective through adulthood. Overall, spousal similarity
measures of self-reported drinking (r = .50) was stronger for adult antisocial behavior than
and smoking (r = .31), starting when spouses for childhood conduct disorder, which indicates
began dating. Here, spousal similarity appeared support for influence. However, additional anal-
to be a function of length of marriage; hier- yses demonstrated that similarity in adult antiso-
archical regression models predicted wives’ cial behavior and in child conduct disorder were
alcohol consumption by adding a husbands’ independent of each other (i.e., the couples who
alcohol × length of marriage interaction term were similar for juvenile symptoms were not
(b = .68, p < .001), which suggested that hus- the same couples similar for adult symptoms).
bands had more influence on their wives’ Ultimately, the authors argued that assortative
alcohol consumption as years of marriage mating, rather than marital influence, was evi-
increased. dent, because analyses from their larger sample
Only one study, Low et al. (2007), simply revealed that having adult-only antisocial behav-
compared the behaviors of each spouse before ior was similar between married and unmarried
marriage with lifetime measures (using diagnos- individuals. Altogether, the eight studies pro-
tic interviews, family reports, and case histories). vided more support for assortative mating than
With a cross-sectional, clinical, and matched for influence on psychiatric disorders, substance
community sample (n = 255, mean age = 40 use, and antisocial behavior. The only study to
years, average length of marriage = 14 years), address gender suggested that husbands’ alcohol
the authors first used logistic regression mod- use influenced wives’ use over time. However,
els and found a significant association between given these studies’ reliance on retrospective
spouses on substance-use disorders, net of con- measures, their findings have less weight.
trols (OR = 7.6, CI [1.9, 30.4]), but not on
anxiety or heterotypic disorders. However, pro-
viding evidence for assortative mating rather Summary of Partner Influence Findings
than for marital influence, findings revealed that At face value, the literature suggests that
87.3% of focal subjects and 80% of spouses assortative mating, rather than partner influence
had a diagnosis before marriage. In addition, the or social homogamy, is the primary mechanism
proportion of spouses’ relatives (siblings or par- responsible for partner similarity. However,
ents) with substance-use disorders was 35.1% methodological limitations—namely, the need
for concordant couples, which was significantly to assess the behaviors of both individuals before
different from couples with one (21.4%) or no mate selection and the reliance on retrospective,
spousal diagnoses (17.4%). cross-sectional measures—affect the validity of
In the eighth and last study, Galbaud du this finding. All the cross-sectional, retrospective
Fort et al. (2002) asked pairs of spouses (or studies except one (Price & Vandenberg, 1980)
cohabiting couples) to complete the Diagnos- supported assortative mating over influence (n =
tic Interview Schedule to provide retrospective 8). Caspi and Herbener (1990, 1993), although
information on lifetime psychiatric symptoms not reviewed here because the authors measured
(n = 519, mean age = 45 years, relationship personality rather than antisocial behavior, also
length not reported). To measure spousal asso- found little support for influence (measured as
ciations for psychiatric symptoms and disor- similarity over time). They argued that couples
ders, they calculated odds ratios for prevalence may share a common reinforcement history,
from two-by-two contingency tables. Spousal and if rewards and punishments are the same
similarity for juvenile behaviors included tru- from one situation to the next (i.e., before and
ancy (OR = 5.11, CI [2.66, 9.85]), lying after marriage), then little change is expected.
(OR = 3.21, CI [1.43, 7.23]), early substance In contrast, all the prospective studies found
use (OR = 6.43, CI [3.15, 13.1]), vandalism support for partner influence (n = 7).
214 Journal of Family Theory & Review

The difference between retrospective and mating, ultimately arguing that the mechanisms
prospective results is a new and important find- are diverse and complex. Theoretical tests of
ing that emerged only through this systematic partner influence were a bit more prevalent,
review of the literature. Given that prospec- showing some support for both social learn-
tive data are generally better than retrospective ing and social control mechanisms (Capaldi
data, prospective findings should be given more et al., 2008; Haynie et al., 2005; Simons, Stewart
weight. Furthermore, partner influence processes et al., 2002).
can influence the retrospective data that are used By and large, the goal of many studies was to
to find support for assortative mating, thus mak- ascertain evidence, in and of itself, for assortative
ing assortative mating effects appear stronger mating or partner influence. Methodologically,
than partner influence effects. Conversely, influ- however, the data were often conflated with
ence processes do not affect prospective data other processes related to partner concordance,
collected before mate selection, because mates thus making it difficult to empirically assess the
have yet to meet. Although Rhule-Louie and mechanisms and consequences of mate selec-
McMahon (2007) found support in the literature tion. Figure 1 illustrates how antisocial measures
for assortative mating, partner influence, too, should be temporally disaggregated and demon-
is recognized here as an important mechanism strates that more is involved than simply find-
responsible for the concordance found between ing an association between partners’ concurrent
partners. This finding has been largely hidden antisocial behavior. In response, three research
or discounted by the literature because of a his- questions corresponding to Figure 1 were devel-
torical reliance on retrospective, cross-sectional oped to disentangle the support for assortative
designs. In addition, four studies suggested that mating, contagion, and mutual influence process
men are more likely than women to influence by paying special attention to how (retrospec-
their partner’s antisocial behavior on measures tively or prospectively), when (before or after
of delinquency, crime, alcohol use, and violence. mate selection), and what types of antisocial
behaviors were measured.
CONCLUSION
Assortative Mating
The overall goal of this article has been to assess
and integrate the empirical literature on part- The first research question attempted to address
ner similarity to determine whether, through the methodological problems inherent in the
assortative mating processes, an individual’s study of assortative mating by assessing the
mate selection influences his or her later anti- similarity between two individuals before they
social behavior. Theoretically, the confluence of actually meet and thus before they are able to
selection (i.e., assortative mating via structural, influence each other. Unfortunately, no studies
individual, and genetic factors) and socialization examined assortative mating by correlating the
(i.e., partner influence via social learning and focal subjects’ prospective adolescent behav-
social control) processes, coupled with the devel- ior with their partners’ retrospective adolescent
opmental consequences of antisocial behavior behavior. Nevertheless, assortative mating for
over time (as proposed by interactional theory), various antisocial behaviors is evident if the defi-
helps explain partner concordance. However, nition is relaxed to include concurrent, retrospec-
the majority of the research did not empiri- tive, and contagion proxies that actually measure
cally test theoretical explanations of assortative partner similarity. Little similarity is observed on
mating or partner influence. Barring the inclu- respondents’ parental history of alcoholism, and
sion of basic demographic variables to examine support for heterotypic assortment is minimal.
social homogamy (which received little support Few studies controlled for social homogamy,
in this literature review), virtually no studies and those that did reduced but did not eliminate
explored, for example, whether structural (e.g., assortative mating associations.
socioeconomic status, ethnic heterogeneity) or
individual (e.g., self-control, intelligence, deci-
Contagion
sion making) factors best explain assortative
mating on antisocial behavior. McLeod (1995) The second research question, estimating con-
did examine childhood and adult risk factors but tagion, addressed two methodological issues by
found little explanatory support for assortative assessing the association between focal subjects’
Assortative Mating and Partner Influence 215

prospective adolescent behavior and their part- Many studies use clinical samples or subsamples
ners’ prospective adult behavior. First, the conta- of larger studies, in which selection bias is
gion proxy has been used to estimate assortative evident. In addition, male subjects are often
mating. Although this strategy is limited because oversampled because they are more likely to
measures are assessed in different developmen- commit antisocial behavior. Likewise, samples
tal stages, the use of prospective, longitudinal with responses from both members of the dyad
measures is an improvement on cross-sectional are small, possibly because dyadic data are
data. Second, contagion has also been used as difficult and expensive to collect. Frequently
a proxy for partner influence because it has (but not in studies included in this review),
the added advantage of temporally isolating the one partner is asked to report on the behaviors
unidirectional effects that one partner has on of both individuals, which results in inflated
the other. However, the association is indirect correlations and has a problem of shared-method
because partners have yet to meet. Regard- variance (Lorenz, Conger, Simon, Whitbeck,
less of whether the contagion proxy is used & Elder, 1991). Respondents’ perceptions are
to assess assortative mating or partner influ- biased because they may project their own values
ence, prospective findings indicate that focal and assume similarity (Jussim & Osgood, 1989).
subjects’ adolescent behavior is related to their Second, measurement issues make it diffi-
partner’s adult behavior on measures of delin- cult to compare findings. Antisocial behavior
quency, crime, violence, and physical abuse. is a broad term used to encompass a spectrum
of problem behaviors across the life course,
Mutual Influence including delinquent attitudes and peers, sub-
stance use, delinquent behavior, criminal justice
The third research question, on mutual influ- involvement, intimate partner violence, and clin-
ence, was an attempt to address how partners ical diagnoses (anxiety, depression, conduct
affect each other’s antisocial trajectory once
and adult personality disorders). Measurement
mate selection had occurred. No studies were
techniques range from collecting self-reported
found, presumably because few had prospective
general delinquency to retrospectively diagnos-
longitudinal data on both partners over several
ing psychiatric disorders and reviewing official
periods. However, the influence that one part-
criminal records. Likewise, definitions of mates
ner had on the other has been estimated using
several different strategies and measures. All of range from adolescent couples to biological par-
the cross-sectional, retrospective studies except ents. Few, if any, studies examine assortative
one supported assortative mating over partner mating among same-sex couples.
influence. In contrast, all the prospective studies Third, contextual changes further complicate
found support for partner influence. Given the research. Over time, increases in educational
limitations inherent in the study of assortative attainment expand the network of potential part-
mating and the fact that, in general, prospective ners (Kalmijn, 1998). College students, for
data are better than retrospective data, partner example, are often geographically separated
influence is recognized here as an important from their parents, thus weakening the family’s
mechanism responsible for partner similarity; influence on dating and marital choices (Rosen-
this mechanism was previously hidden or dis- feld, 2007). Other factors include changes in
counted in the literature as a result of a reliance women’s participation in the workforce, access
on retrospective, cross-sectional designs. to contraception, and average age at marriage.
There have also been considerable changes in
the importance of women’s earning potential and
Methodological and Contextual Issues their position in the marriage market (Sweeney &
Although this finding is an important and new Cancian, 2004). Contextual issues such as these
addition to the literature—because results from make it difficult to generalize from these findings
studies were disaggregated and compared by because early research (e.g., in the 1950s) was
their design features—other methodological and different from that of today. Combined, these
contextual issues may also affect the validity issues—generalizability, measurement, and con-
of the research on assortative mating and textual changes—make it difficult to assess
partner influence, overall. First, samples are findings and limit the understanding of assor-
rarely representative of the general population. tative mating and partner influence.
216 Journal of Family Theory & Review

Nevertheless, this literature review and future at heightened risk because they are more likely
research will push the field forward. For to have not one, but two, antisocial parents.
example, research has shown that marriage and
other types of romantic bonds promote desis-
REFERENCES
tance from crime and other forms of antisocial
behavior (Maume, Ousey, & Beaver, 2005; Agrawal, A., Heath, A. C., Grant, J. D., Pergadia,
McCarthy & Casey, 2008; Meeus, Branje, & M. L., Statham, D., Bucholz, K. K., et al. (2006).
Assortative mating for smoking and for alcohol
Overbeek, 2004; Sampson & Laub, 1990, 1993).
consumption in female Australian twins and their
However, the mechanisms responsible for the spouses. Behavior Genetics, 36, 553 – 566.
good-marriage effect are not clear (Giordano, Bandura, A. (1977). Social learning theory. Oxford,
Cernkovich, & Holland, 2003; Simons, Stewart UK: Prentice Hall.
et al., 2002; Warr, 1998), and the convention- Bereczkei, T., Gyuris, P., Koves, P., & Bernath, L.
ality or deviance of the spouse usually has (2002). Homogamy, genetic similarity and im-
not been examined. Future research that exam- printing: Parental influence on mate choice prefer-
ines processes related to partner concordance ences. Personality and Individual Differences, 33,
677 – 690.
in their models will clarify the marriage-crime
Boye-Beaman, J., Leonard, K. E., & Senchak, M.
relationship. (1991). Assortative mating, relationship develop-
In addition, further research on assortative ment, and intimacy among offspring of alcoholics.
mating and partner influence will help bring Family Dynamics of Addiction Quarterly, 1(2),
about a better understanding of the role of gen- 20 – 33.
der in antisocial behavior. Theoretically, power Buchanan, A. (1996). Cycles of child maltreatment:
imbalances, which differentially shape men’s Facts, fallacies, and interventions. West Sussex,
and women’s mating preferences (Shackelford, UK: Wiley.
Schmitt, & Buss, 2005), antisocial behavior Capaldi, D. M., & Crosby, L. (1997). Observed and
reported psychological and physical aggression
(Caspi, Lynam, Moffitt, & Silva, 1993), and the in young, at-risk couples. Social Development, 6,
context of their romantic relationships (Haynie 184 – 206.
et al., 2005), should translate into gender dispar- Capaldi, D. M., Kim, H. K., & Owen, L. D. (2008).
ities in partner similarity on antisocial behavior. Romantic partners’ influence on men’s likelihood
In the literature reviewed here, estimates of gen- of arrest in early adulthood. Criminology, 46,
der differences were minimal, in part because 267 – 299.
few studies actually examined them. Caspi, A., & Herbener, E. S. (1990). Continuity and
Overall, research designed to further under- change: Assortative marriage and the consistency
of personality in adulthood. Journal of Personality
standing of the onset, persistence, and change and Social Psychology, 58, 250 – 258.
in antisocial behavior across the life course has Caspi, A., & Herbener, E. S. (1993). Marital assort-
shown that several factors help explain its vari- ment and phenotypic convergence: Longitudinal
ability. Individual, family, school, employment, evidence. Social Biology, 40(1–2), 48 – 60.
community, and structural factors all play an Caspi, A., Lynam, D., Moffitt, T. E., & Silva, P. A.
important role in predicting the probability that (1993). Unraveling girls’ delinquency: Biological,
an individual will become involved in anti- dispositional, and contextual contributions to ado-
social behavior (Thornberry & Krohn, 2005; lescent misbehavior. Developmental Psychology,
29, 19 – 30.
U.S. Department of Health and Human Ser- Elder, G. H. (1998). The life course and human
vices, 2001). In addition, the importance of development. In W. Damon & R. M. Lerner
peers during adolescence has been a signifi- (Eds.), Handbook of child psychology: Vol. 1.
cant focus of criminological research (Haynie Theoretical models of human development (5th
et al., 2005). This literature review advances ed., pp. 939 – 991). Hoboken, NJ: Wiley.
the field by acknowledging that research needs Fergusson, D. M., Horwood, L. J., & Lynskey, M. T.
to account for the selection and influence of (1994). The comorbidities of adolescent problem
not only peers but also partners, who play an behaviors: A latent class model. Journal of
Abnormal Child Psychology, 22, 339 – 354.
increasingly significant role as individuals grow Fields, J. (2004). America’s families and living
older. The current findings combined with future arrangements (2003 Current Population Reports
research will aid understanding of the effects of P20-553). Washington, DC: U.S. Census Bureau.
assortative mating and partner influence on cou- Foley, D. L., Pickles, A., Simonoff, E., Maes, H. H.,
ples and their children who, unfortunately, are Silberg, J. L., Hewitt, J. K., et al. (2001). Parental
Assortative Mating and Partner Influence 217

concordance and comorbidity for psychiatric between partners and risk for aggression in roman-
disorder and associate risks for current psychiatric tic relationships. Journal of Family Psychology,
symptoms and disorders in a community sample of 18, 82 – 96.
juvenile twins. Journal of Child Psychology and Krueger, R. F., Moffitt, T., Caspi, A., Bleske, A., &
Psychiatry, 42, 381 – 394. Silva, P. A. (1998). Assortative mating for antiso-
Furman, W., & Simon, V. A. (2008). Homophily cial behavior: Development and methodological
in adolescent romantic relationships. In M. J. implications. Behavior Genetics, 28, 173 – 186.
Prinstein & K. A. Dodge (Eds.), Understanding Leonard, K. E., & Mudar, P. (2003). Peer and partner
peer influence in children and adolescents drinking and the transition to marriage: A
(pp. 203 – 224). New York: Guilford Press. longitudinal examination of selection and influence
Galbaud du Fort, G., Boothroyd, L. J., Bland, R. C., processes. Psychology of Addictive Behaviors, 17,
Newman, S. C., & Kakuma, R. (2002). Spouse 115 – 125.
similarity for antisocial behaviour in the gen- Lorenz, F. O., Conger, R. D., Simon, R. L., Whitbeck,
eral population. Psychological Medicine, 32, L. B., & Elder, G. H., Jr. (1991). Economic
1407 – 1416. pressure and marital quality: An illustration of the
Giordano, P. C., Cernkovich, S. A., & Holland, D. D. method variance problem in the causal modeling
(2003). Changes in friendship relations over the of family processes. Journal of Marriage and the
life course: Implications for desistance from crime. Family, 53, 375 – 388.
Criminology, 41(2), 63 – 89. Low, N., Cui, L., & Merikangas, K. R. (2007). Spo-
Gleiberman, L., Harburg, E., DiFranceisco, W., & usal concordance for substance use and anxiety
Schork, A. (1992). Familial transmission of alco- disorders. Journal of Psychiatric Research, 41,
hol use: V. drinking patterns among spouses. 942 – 951.
Behavior Genetics, 22, 63 – 79. Lykken, D. T., & Tellegen, A. (1993). Is human
Gottfredson, M. R., & Hirschi, T. (1990). A general mating adventitious or the result of lawful choice?
theory of crime. Stanford, CA: Stanford University A twin study of mate selection. Journal of
Press. Personality and Social Psychology, 65, 56 – 68.
Lynam, D., Moffitt, T., & Stouthamer-Loeber, M.
Haynie, D. L., Giordano, P. C., & Manning, W. D.
(1993). Explaining the relation between IQ and
(2005). Adolescent romantic relationships and
delinquency: Class, race, test motivation, school
delinquency involvement. Criminology, 43,
failure, or self-control? Journal of Abnormal
177 – 210.
Psychology, 102, 187 – 196.
Herrera, V. M., Wiersma, J. D., & Cleveland, H. H.
Maes, H. H., Neale, M. C., Kendler, K. S.,
(2008). The influence of individual and partner
Martin, N. G., Heath, A. C., & Eaves, L. J. (2006).
characteristics on the perpetration of intimate Genetic and cultural transmission of smoking
partner violence in young adult relationships. initiation: An extended twin kinship model.
Journal of Youth and Adolescence, 37, 284 – 296. Behavior Genetics, 36, 795 – 808.
Hicks, B. M., Krueger, R. F., Lacono, W. G., Mc- Maes, H. H., Silberg, J. L., Neale, M. C., &
Gue, M., & Patrick, C. J. (2004). Family Eaves, L. J. (2007). Genetic and cultural trans-
transmission and heritability of externalizing mission of antisocial behavior: An extended twin
disorders: A twin-family study. Archives of parent model. Twin Research and Human Genetics,
General Psychiatry, 61, 922 – 928. 10, 136 – 150.
Hirschi, T. (1971). Causes of delinquency. Berkeley: Mathews, T. J., & Hamilton, B. E. (2002). Mean age
University of California Press. of mother, 1970–2000. National Vital Statistics
Ireland, T. O., Smith, C. A., & Thornberry, T. P. Report, 51, 1 – 16.
(2002). Developmental issues in the impact of Matsueda, R. L., & Heimer, K. (1997). A sym-
child maltreatment on later delinquency and drug bolic interactionist theory of role-transitions, role-
use. Criminology, 40, 359 – 396. commitments, and delinquency. In T. P. Thorn-
Jussim, L., & Osgood, D. W. (1989). Influence berry (Ed.), Developmental theories of crime and
and similarity among friends: An integrative delinquency (pp. 163 – 213). New Brunswick, NJ:
model applied to incarcerated adolescents. Social Transaction.
Psychology Quarterly, 52(2), 98 – 112. Maume, M. O., Ousey, G. C., & Beaver, K. (2005).
Kalmijn, M. (1994). Assortative mating by cultural Cutting the grass: A reexamination of the link
and economic occupational status. American between marital attachment, delinquent peers
Journal of Sociology, 100, 422 – 452. and desistance from marijuana use. Journal of
Kalmijn, M. (1998). Intermarriage and homogamy: Quantitative Criminology, 21, 27 – 53.
Causes, patterns, trends. Annual Review of McCarthy, B., & Casey, T. (2008). Love, sex, and
Sociology, 24, 395 – 421. crime: Adolescent romantic relationships and
Kim, H. K., & Capaldi, D. M. (2004). The association offending. American Sociological Review, 73,
of antisocial behavior and depressive symptoms 944 – 969.
218 Journal of Family Theory & Review

McLeod, J. D. (1995). Social and psychological bases Sakai, J. T., Stallings, M. C., Mikulich-Gilbertson,
of homogamy for common psychiatric disorders. S. K., Corley, R. P., Young, S. E., Hopfer, C. J.,
Journal of Marriage and the Family, 57, 201 – 214. et al. (2004). Mate similarity for substance
McPherson, M., Smith-Lovin, L., & Cook, J. M. dependence and antisocial personality disorder
(2001). Birds of a feather: Homophily in social symptoms among parents of patients and controls.
networks. Annual Review of Sociology, 27, Drug and Alcohol Dependence, 75, 165 – 175.
415 – 444. Sampson, R. J., & Groves, W. B. (1989). Community
Meeus, W., Branje, S., & Overbeek, G. J. (2004). structure and crime: Testing social-disorganization
Parents and partners in crime: A six-year theory. American Journal of Sociology, 94,
longitudinal study on changes in supportive 774 – 802.
relationships and delinquency in adolescence and Sampson, R. J., & Laub, J. H. (1990). Crime and
young adulthood. Journal of Child Psychology and deviance over the life course: The salience of
Psychiatry, 45, 1288 – 1298. adult social bonds. American Sociological Review,
Moffitt, T. E. (1997). Adolescence-limited and life- 55, 609 – 627.
course-persistent offending: A complementary pair Sampson, R. J., & Laub, J. H. (1993). Crime in the
of developmental theories. In T. P. Thornberry making: Pathways and turning points through life.
(Ed.), Developmental theories of crime and Crime and Delinquency, 39, 396 – 396.
delinquency (pp. 11 – 54). New Brunswick, NJ: Schuckit, M. A., Smith, T. L., Eng, M. Y., & Kuno-
Transaction. vac, J. (2002). Women who marry men with
Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. alcohol-use disorders. Alcoholism: Clinical and
(2001). Sex differences in antisocial behaviour: Experimental Research, 26, 1336 – 1343.
Conduct disorder, delinquency, and violence Shackelford, T. K., Schmitt, D. P., & Buss, D. M.
in the Dunedin longitudinal study. New York: (2005). Universal dimensions of human mate pref-
Cambridge University Press. erences. Personality and Individual Differences,
Patterson, G. R., Capaldi, D., & Bank, L. (1991). An 39, 447 – 458.
early starter model for predicting delinquency. In Simon, V. A., Aikins, J. W., & Prinstein, M. J. (2008).
D. J. Pepler & K. H. Rubin (Eds.), Earlscourt sym- Romantic partner selection and socialization
posium on childhood aggression (pp. 139 – 168). during early adolescence. Child Development, 79,
Hillsdale, NJ: Erlbaum. 1676 – 1692.
Price, R. A., & Vandenberg, S. G. (1980). Spouse Simons, R. L., Gordon Simons, L., & Ebert
similarity in American and Swedish couples. Wallace, L. (2004). Families, delinquency, and
Behavior Genetics, 10, 59 – 71. crime: Linking society’s most basic institution to
Quinton, D., Pickles, A., Maughan, B., & Rutter, M. antisocial behavior. Los Angeles: Roxbury.
(1993). Partners, peers, and pathways: Assortative Simons, R. L., Stewart, E., Gordon, L. C., Conger,
pairing and continuities in conduct disorder. R. D., & Elder, G. H. (2002). A test of life-course
Development and Psychopathology, 5, 763 – 783. explanations for stability and change in antisocial
Raine, A., Moffitt, T. E., Caspi, A., Loeber, R., behavior from adolescence to young adulthood.
Stouthamer-Loeber, M., & Lynam, D. (2005). Criminology, 40, 401 – 434.
Neurocognitive impairments in boys on the Sutherland, E. H. (1947). Principles of criminology
life-course persistent antisocial path. Journal of (4th ed.). Oxford, UK: Lippincott.
Abnormal Psychology, 114, 38 – 49. Sweeney, M. M., & Cancian, M. (2004). The chang-
Reynolds, C. A., Baker, L. A., & Pedersen, N. L. ing importance of white women’s economic
(2000). Multivariate models of mixed assortment: prospects for assortative mating. Journal of Mar-
Phenotypic assortment and social homogamy for riage and Family, 66, 1015 – 1028.
education and fluid ability. Behavior Genetics, 30, Taylor, J., McGue, M., & Iacono, W. G. (2000).
455 – 476. Sex differences, assortative mating, and cultural
Rhule-Louie, D. M., & McMahon, R. J. (2007). Prob- transmission effects on adolescent delinquency: A
lem behavior and romantic relationships: Assorta- twin family study. Journal of Child Psychology
tive mating, behavior contagion, and desistance. and Psychiatry, 41, 433 – 440.
Clinical Child and Family Psychology Review, 10, Thornberry, T. P., Freeman-Gallant, A., & Love-
53 – 100. grove, P. J. (2009). The impact of parental stressors
Rosenfeld, M. J. (2007). The age of independence: on the intergenerational transmission of antisocial
Interracial unions, same-sex unions, and the behavior. Journal of Youth and Adolescence, 38,
changing American family. Cambridge, MA: 312 – 322.
Harvard University Press. Thornberry, T. P., & Krohn, M. D. (2000). Self-report
Rushton, P. J. (1989). Genetic similarity, human method for measuring delinquency and crime.
altruism, and group selection. Behavioral and In D. Duffee (Ed.), Measurement and analysis
Brain Sciences, 12, 503 – 518. of crime and justice: Criminal justice 2000
Assortative Mating and Partner Influence 219

(pp. 33 – 83). Washington, DC: National Institute Warr, M. (1998). Life-course transitions and desis-
of Justice. tance from crime. Criminology, 36, 183 – 216.
Thornberry, T. P., & Krohn, M. D. (2005). Apply- Watson, D., Klohnen, E. C., Casillas, A., Nus Simms,
ing interactional theory to the explanation of E., Haig, J., & Berry, D. S. (2004). Match makers
continuity and change in antisocial behavior. and deal breakers: Analyses of assortative mating
In D. P. Farrington (Ed.), Integrated develop- in newlywed couples. Journal of Personality, 72,
mental and life-course theories of offending 1029 – 1068.
(pp. 183 – 209). New Brunswick, NJ: Transaction. Yamaguchi, K., & Kandel, D. (1993). Marital homo-
Thornberry, T. P., Krohn, M. D., Lizotte, A. J., & phily on illicit drug use among young adults:
Smith, C. A. (Eds.). (1998). Taking stock: An Assortative mating or marital influence? Social
overview of findings from the Rochester youth Forces, 72, 505 – 528.
development study. New York: Kluwer Aca- Yamaguchi, K., & Kandel, D. B. (1997). The
demic–Plenum. influence of spouses’ behavior and marital
Thornberry, T. P., Lizotte, A. J., Krohn, M. D., dissolution on marijuana use: Causation or
Farnworth, M., & Joon Jang, S. (1991). Testing selection. Journal of Marriage and the Family, 59,
interactional theory: An examination of reciprocal 22 – 36.
causal relationships among family, school, and Yarrow, M. R., Campbell, J. D., & Burton, R. V.
delinquency. Journal of Criminal Law and (1970). Recollections of childhood: A study of
Criminology, 82, 3 – 35. the retrospective method. Monographs of the
U.S. Department of Health and Human Services. Society for Research in Child Development,
(2001). Youth violence: A report of the surgeon 35(5), 83.
general. Rockville, MD: Author.
Copyright of Journal of Family Theory & Review is the property of Wiley-Blackwell and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.