understanding associations among family support, friend support,...

Personal Relationships, 22 (2015), 79–91. Printed in the United States of America.Copyright © 2014 IARR; DOI: 10.1111/pere.12063

Understanding associations among family support,friend support, and psychological distress

BRIANA N. HORWITZ,a CHANDRA A. REYNOLDS,b AND SUSAN T. CHARLESc

aCalifornia State University, Fullerton; bUniversity of California, Riverside; and cUniversityof California, Irvine

AbstractEmotional support from family and friends is associated with lower psychological distress. This study examinedwhether genetic and environmental influences explain associations among family support, friend support, andpsychological distress. Data were drawn from the Midlife Development in the United States (MIDUS) study andincluded 947 pairs of monozygotic (MZ), same-sex dizygotic (DZ), and opposite-sex DZ twins. Results showed that agenetic factor explains the relation between friend support and psychological distress, independent of family support.Alternatively, a nonshared environmental factor accounts for an association between family support, friend support, andpsychological distress. Thus, heritable factors shape a distinct relation between friend support and psychologicaldistress, but unique experiences contribute to a link between family support, friend support, and psychological distress.

Greater emotional support from friends isconsistently linked with lower levels of psy-chological distress (e.g., Cohen, 2004; Lepore,Evans, & Schneider, 1991; Norton et al.,2005; Ritsner, Modai, & Ponizovsky, 2000).A frequent explanation for this association isone focusing on environmental sources, suchthat friends engage in behaviors that help toalleviate individuals’ psychological distress.A recent genetically informed study, how-ever, demonstrated that the relation betweenfriend support and psychological distress isattributable to genetic influences (Horwitz,Reynolds, Neiderhiser, & Charles, in press).This suggests that the same set of genes thatinfluence friend support also contribute to

Briana N. Horwitz, Department of Psychology, CaliforniaState University, Fullerton; Chandra A. Reynolds, Depart-ment of Psychology, University of California, Riverside;Susan T. Charles, Department of Psychology and SocialBehavior, University of California, Irvine.

Funding for the Midlife Development in the UnitedStates study was provided with support from the John D.and Catherine T. MacArthur Foundation Research Net-work on Successful Midlife Development.

Correspondence should be addressed to Briana N. Hor-witz, Department of Psychology, California State Uni-versity, Fullerton, 800 N. State College Blvd., Fullerton,CA 92831, e-mail: [email protected].

psychological distress. A new question thatarises from this finding is: What types of genet-ically based mechanisms explain this link? Forexample, heritable tendencies to evoke moreor less support from others may be responsiblefor an association between general aspects ofsocial support (i.e., with family and friends)and psychological distress. Alternatively,heritable tendencies to seek out supportivefriends may account for a distinct relationbetween friend support and psychologicaldistress that is not shared with other forms ofsocial support, including family support. Thus,additional research is needed to assess whethergenetic contributions to the link between friendsupport and psychological distress are sharedwith family support. This study examinedwhether genetic and environmental influencesexplain associations between family support,friend support and psychological distress in asample of adult twins in the United States.

Genetically informed designs are usedto examine potential genetic influences onphenotypes. Phenotypes refer to any observ-able characteristic, such as hair color, cognitiveability, personality (e.g., extraversion),reported social support, and reported psy-chological distress. In twin studies, genetic

79

80 B. N. Horwitz, C. A. Reynolds, and S. T. Charles

factors reflect the influence of the whole geno-type on the measured phenotype, rather thanthe effects of any particular genes (Reiss &Leve, 2007). The twin-based method has beenposited to be particularly useful for examininghow genetic factors shape complex traits,which are likely influenced by many (poly-genic) genes (McGue & Bouchard, 1984). Asdescribed in detail in the Method section, twinstudies compare covariances for a trait or setof traits among genetically identical (monozy-gotic [MZ]) twins and fraternal (dizygotic[DZ]) twins. For example, twin designs canbe used to assess whether identical twins andtheir co-twins are more alike on a particularphenotype compared to fraternal twins andtheir co-twins. If identical twins are morealike on the given phenotype, this indicates theinfluence of genetic effects on that phenotype.

Twin-based methods have been used tounderstand how mental health outcomes,including psychological distress, are shaped.Indeed, previous research has shown thatapproximately 44% of the individual vari-ance in psychological distress is ascribableto genetic effects (Rijsdijk et al., 2003). Thissuggests that psychological distress is, at leastin part, a heritable characteristic.

Likewise, twin-based methods have beenused to understand how social support arises.This research has shown that genetic factorsexplain approximately 17%–38% of the indi-vidual variance in both family support andfriend support (e.g., Kendler & Baker, 2007).When genetic influences are found to con-tribute to social support, this may reflect theinfluence of genotype–environment correla-tion (rGE) on social support (e.g., Horwitz& Neiderhiser, 2011). Genotype correlationrefers simply to a correlation between an indi-vidual’s genotype and environment (Plomin,DeFries, & Loehlin, 1977; Scarr & McCart-ney, 1983). For example, an underlying heri-table characteristic (e.g., personality) may becorrelated with an individual’s exposure to cer-tain environmental experiences (e.g., the typeof social support that the individual receives).

Two types of rGE that are relevant forunderstanding how genetic factors shape socialsupport include evocative and active rGE.Evocative rGE is defined as a phenomenon

where individuals’ heritable characteristicsare likely to elicit certain responses fromothers. For example, previous research hassuggested that people with higher levels ofheritable aggressive or externalizing behaviorstend to elicit more negative responses fromothers (e.g., Ganiban et al., 2009; Marceauet al., 2013). Evocative rGE is relevant forunderstanding both friend and family supportbecause family and friends may react in similarways to individuals’ heritable characteristics.Alternatively, active rGE occurs when indi-viduals select relationships that are correlatedwith their own heritable characteristics. Priorwork has suggested that adolescents’ heri-table characteristics (e.g., personality traits)are related to the particular peer groups theyselect (e.g., Brendgen, 2012). For example,adolescents with higher levels of heritableaggressive personality may tend to select peerswith similar aggressive tendencies (Brendgen,2012). Active rGE is relevant for explainingfriend support because friends are selected butnot family support because family membersare ascribed. To illustrate, individuals withhigher levels of heritable extraversion maytend to select more supportive friends.

Genetically informed studies also providethe capability of investigating the extent towhich genetic factors explain associationsamong phenotypes. Using this approach, sev-eral previous studies have shown that geneticfactors explain links between social supportand symptoms of psychological distress (e.g.,Bergeman, Plomin, Pederson, & McClearn,1991; Spotts & Pederson, 2005; Yuh & Reiss,2008). These studies often use general mea-sures of social support that include both familyand friend support in one composite measure.Thus, it is difficult to determine from thiswork whether genetic influences contribute toassociations between symptoms of psycholog-ical distress and general aspects of support, ormore specific forms of support from family orfriends. A more recent study demonstrated thatthe link between friend support and psycholog-ical distress is also ascribable to genetic effects(Horwitz et al., in press). This finding helpedto clarify that the same heritable characteristicsthat are important for shaping friend supportalso contribute to psychological distress.

Social support and psychological distress 81

Yet questions remain about what geneticallybased mechanisms explain about the relationbetween friend support and psychological dis-tress. Specifically, genetic influences may giverise to associations between general aspectsof support (i.e., from family and friends) andpsychological distress. A plausible explanationfor these genetic influences would be evocativerGE. For example, the same heritable charac-teristics in individuals that tend to elicit lesssupport from both family and friends may alsolead to greater levels of psychological distress.Alternatively, genetic influences may shape adistinct relation between friend support andpsychological distress that is independent ofsupport from family members. These geneticinfluences may reflect the role of active rGE.That is, individuals’ heritable characteristicsmay tend to influence their friend choices and,in turn, their friend support and psychologicaldistress. Thus, a genetically informed designis needed to investigate whether genetic influ-ences contribute to a relation between familysupport, friend support, and psychological dis-tress, or to a unique association between friendsupport and psychological distress.

In addition to genetic influences, twin stud-ies also provide information about the role ofenvironments, including shared and nonsharedenvironments, in shaping phenotypes. Sharedenvironmental influences, described in moredetail in the Method section, are ascertainedin twin studies by the degree of similaritythat twin pairs share with each other that doesnot stem from genetic influences (e.g., twins’rearing environment). For example, growingup in a highly supportive environment whereparents encouraged and nurtured friendshipsmay give rise to strong supportive relationshipsfrom both family and friends and lower levelsof psychological distress (Shaw et al., 2004).In this way, shared environmental influencesmay shape links among general aspects ofindividuals’ social support from family andfriends and psychological distress. Nonsharedenvironmental influences may also account foran association between family support, friendsupport, and psychological distress. Nonsharedenvironmental influences are those uniqueenvironmental experiences that twins do notshare with their co-twins (e.g., negative life

events unique to the individual). For example,job loss may lead to increased psychologicaldistress and social withdrawal (Russell, 1999),which may influence how people perceivesupport from family and friends.

This study examined the degree to whichgenetic and environmental (both shared andnonshared) factors contribute to the associa-tion between family support, friend support,and psychological distress. The sample wasdrawn from the Midlife Development in theUnited States (MIDUS) study, which was thesame sample of adult twins that was used in theinvestigation by Horwitz and colleagues (inpress). This study builds on Horwitz and col-leagues’ work by examining whether geneticcontributions to the link between friend sup-port and psychological distress are explainedby or distinct from support from family mem-bers. The extent to which environmentalfactors explain associations between familysupport, friend support, and psychologicaldistress was also assessed.

Method

Participants

The sample consisted of 998 twin pairs fromthe MIDUS study (Kessler, Gilman, Thornton,& Kendler, 2004). Data on the twins were col-lected primarily from 1995 to 1996. Both mem-bers of each twin pair participating in the studymet the overall study eligibility criteria and cri-teria specific to the twin sample. These criteriaincluded being at least first-degree relatives ofthe original contact or his or her spouse or part-ner, between 25 and 74 years, having a residen-tial telephone number, living in the continentalUnited States, and speaking English. Zygositywas determined using self-report questions thatasked about similarity in traits such as eye andhair color and issues such as whether the twinswere mistaken for one another as children.Such techniques are generally more than 90%accurate (Lykken, Bouchard, McGue, & Telle-gan, 1990). Twin pairs were randomly desig-nated as either Twin 1 or Twin 2.

In the current analyses, twin pairs wereexcluded if they had missing or unknownvalues for zygosity or sex or if they came from


families with data from two or more pairs.After these exclusions, this study included947 twin pairs, including 362 MZ (n= 168MZ males, n= 194 MZ female pairs), 335same-sex DZ (n= 129 DZ males, n= 206 DZfemale pairs), and 250 opposite-sex DZ pairs.The sample included more women (55%) thanmen (45%) and ranged in age from 25 to 74years (M = 45, ± 12.05). The ethnic break-down included Caucasian= 92.0%, AfricanAmerican= 4.5%, Native American/Eskimo=0.8%, multiracial= 0.3%, other= 1.0%, andunreported= 1.6% individuals. Most indi-viduals were married (71%), with the restnot married (28%), or who did not reporton their marital status (1%), and over halfthe sample had at least 1–2 years of col-lege education (58%). Finally, 12.94% ofthe sample reported symptoms that classi-fied them as having had a major depressivedisorder, dysthymia, or generalized anxietydisorder within the past year. This is consis-tent with a recent study using the nontwinMIDUS sample which also reported that12.20% of the sample had a major depressivedisorder, dysthymia, or generalized anxietydisorder (Charles, Piazza, Mogle, Sliwinski, &Almeida, 2013). These MIDUS twin samplecharacteristics are summarized in Table 1.

Measures

Family support

Family support (six items, α= .89) measuredthe degree of emotional support provided byfamily members, including brothers, sisters,parents, and children who do not live with them(Brim, Ryff, & Kessler, 2004). On a 4-pointLikert scale from 1 (a lot) to 4 (not at all),respondents rated how much family membersreally care about you, understand the way youfeel about things, you can rely on them for helpif you have a serious problem, and open upto them if you need to talk about your wor-ries. Items were reverse-scored and averagedtogether to create a composite scale, in whichhigher scores indicated greater family support.

Friend support

Friend support (four items, α= 0.81) assessedthe degree of emotional support provided by

Table 1. Sample demographic information

Category Percent

ZygosityMZ 38.2Same-sex DZ 35.4Opposite-sex DZ 26.4

GenderWomen 55.0Men 45.0

EthnicityCaucasian 92.0African American 4.5Native American/Eskimo 0.8Multiracial 0.3Other 1.0Not reported 1.6

Marital statusMarried 71.0Not married 28.0Not reported 1.0

EducationOver 12 years 55.4For 12 years 28.6Under 12 years 11.8Missing 4.3

Major depressive disorder or dysthymiawithin past yearYes 12.2No 87.8

Note. DZ= dizygotic; MZ=monozygotic.

close friends (Brim et al., 2004). On a 4-pointLikert scale from 1 (a lot) to 4 (not at all),respondents rated how much close friendsreally care about you, understand the wayyou feel about things, can you rely on themfor help if you have a serious problem, andopen up to them if you need to talk aboutyour worries. Items were reverse-scored andaveraged together to create a composite scale,in which higher scores indicated higher levelsof friend support.

Psychological distress

The Nonspecific Psychological Distress Scale(Kessler et al., 2002; Mroczek & Kolarz,1998; six items, α= 0.87) was used to measurepsychological distress. Respondents were


asked to rate for how long during the last30 days from 1 (none of the time) to 5 (all ofthe time) they felt so sad that nothing couldcheer them up, nervous, restless or fidgety,hopeless, that everything was an effort, andworthless. This scale was developed usingitem response models and factor analysis,yielding a single factor structure represent-ing current, general psychological distress.Each item was selected because of its highutility in identifying symptoms common tomany DSM-IV-TR (Diagnostic and Statis-tical Manual of Mental Disorders, 4th ed.,Text Revision) disorders, including depressedmood, motor agitation, fatigue, nervousness,and worthless guilt. The measure was validatedin eight administrations using samples fromdifferent populations (Kessler et al., 2002;Mroczek & Kolarz, 1998). The measure wasdeveloped as a screening tool for quickly iden-tifying people with mental health issues in thegeneral population (Kessler et al., 2002). Itemswere averaged together to create a compositescale, with higher scores indicating higherlevels of psychological distress. To correctfor positive skewness, psychological distresswas transformed prior to analyses using logtransformation.

Data analysis

Twin design

Twin designs include pairs of MZ twins whoshare 100% of their genes and DZ twins whoshare, on average, 50% of their segregatinggenes. Initial genetic, shared environmen-tal, and nonshared environmental estimates,based on twin correlations (i.e., intraclasscorrelations), can provide a preliminary under-standing of their contributions to a trait.Specifically, comparing intraclass correlationsbetween MZ twins with those between DZtwins allows for initial estimates of these influ-ences for a single phenotype. For example,suppose a study reveals that MZ twins havean intraclass correlation of 0.40 for a singlephenotype and the DZ twins have an intraclasscorrelation of 0.25. The MZ correlation of 0.40reveals that 0.60 or 60% of the variability ofthe phenotype in question is ascribable to non-shared environmental effects. This is because

MZ twins share 100% of their genetic material,so any factors contributing to their dissimilarityare attributable to the nonshared environment.Of the remaining 40%, researchers can com-pare estimates of the MZ twins with thoseintraclass correlations of the DZ twins toevaluate heritable and shared environmentalcontributions. Higher correlations for MZ ver-sus DZ twins indicate genetic contributions.The total genetic contribution is estimated bydoubling the difference between MZ and DZpairs, that is, (0.40 – 0.25)× 2= 0.30, or 30%of the variance. Hence, the remaining 10%of the variance shared by twin pairs stemsfrom shared environmental variance (i.e.,0.40 – 0.30= 0.10). These initial estimates,within trait and even across traits, are usefulin guiding expectations from formal modelestimation, but that is appropriately conductedon covariance matrices and takes into accountsample sizes, missing data, and so on.

Biometric model fitting

A biometric model-fitting approach, Choleskydecomposition, was used to estimate simul-taneously the relative additive genetic (A),shared environmental (C), and nonshared envi-ronmental (E) contributions to the associa-tions between family support, friend support,and psychological distress by maximum likeli-hood estimation of the raw data in Mx (Neale,Boker, Xie, & Maes, 2003). This approachextends the univariate model (described above)to examine the associations between pheno-types and variables. Mx is a statistical soft-ware program commonly used for structuralequation modeling (Neale et al., 2003). Asdisplayed in Figure 1, this model includesthree variables: family support (Variable 1),friend support (Variable 2), and psychologi-cal distress (Variable 3). Family support wasentered prior to friend support and psycho-logical distress to test the extent to whichfamily support account for the genetic covari-ance between friend support and psychologi-cal distress. Furthermore, friend support wasentered prior to psychological distress, consis-tent with theoretical explanations that friendsupport influences psychological distress (e.g.,Cohen, 2004).


Family Relationship

Quality(Twin 1)

A1 E1

C3C2C1

E3A3E2A2

a11

e11

c11

a21

e21

a31

e31

a22

e32e22

a32

a33

e33

c21

c31

c22c32

c33

Friendship Quality

(Twin 1)

Lower PsychologicalDistress (Twin 1)

A1 E1

C3C2C1

E3A3E2A2

a11

e11

c11

a21

e21

a31

e31

a22

e32

e22

a32

a33 e33

c21

c31

c22c32

c33

MZ, DZ = 1.0

MZ = 1.0, DZ = 0.5

Family

Relationship Quality

(Twin 2)

Friendship Quality

(Twin 2)

Lower Psychological

Distress (Twin 2)

Family Support

(Twin 1)

Friend Support

(Twin 1)

Lower Psychological Distress (Twin 1)

Family Support

(Twin 1)

Friend Support

(Twin 1)

Lower Psychological Distress (Twin 1)

Figure 1. Cholesky decomposition. This figure includes latent genetic (A), shared environmental(C), and nonshared environmental (E) factors for both twins in a pair. A, C, and E influences on theassociations between family support, friend support, and psychological distress= path estimatesa11, a21, a31, c11, c21, a31, e11, e21, e21. A, C, and E influences on the association between friendsupport and psychological distress, independent of family support= path estimates a22, a32, c22,c32, e22, e32. Residual A, C, and E variance in psychological distress= path estimates a33, c33,e33.

Prior to model fitting, psychological distresswas reverse-scored (higher levels of family andfriend support= lower level of psychologicaldistress) to be positively associated with fam-ily and friend support. This model includesgenetic (A1, A2, and A3), shared environmen-tal (C1, C2, and C3), and nonshared environ-mental (E1, E2, and E3) factors. As depictedin Figure 1, A1, C1, and E1 explain vari-ance in family support. The relative contribu-tions of A1, C1, and E1 to family support areestimated by paths a11, c11, and e11. A1, C1,and E1 may also explain variance in friendsupport (paths a21, c21, and e21) and psycho-logical distress (paths a31, c31, and e31). Thefactors A2, C2, and E2 explain variance infriend support. The relative contributions ofthese factors to friend support are estimated bypaths a22, c22, and e22. A2, C2, and E2 also

account for variance in psychological distressvia the paths a32, c32, and e32. Finally, A3,C3, and E3 account for residual variance inpsychological distress, independent of familyand friend supports, denoted by the paths a33,c33, and e33.

The extent to which the genetic, sharedenvironmental, and nonshared environmen-tal factors associated with family and friendsupport account for variance in psycholog-ical distress is computed as: (a21

2 + a312

+ c212 + c31

2 + e212 + e31

2). The degree towhich genetic and environmental influencesrelated to friend support explain psychologicaldistress, independent of family support, iscomputed as (a32

2 + c322 + e32

2). Residualvariance in psychological distress unrelated toboth family and friend support is computed as(a33

2 + c332 + e33

2).


Tests of model fit

A full model was tested, where all possiblegenetic, shared environmental, and nonsharedenvironmental paths were estimated. The pathsthat were not significantly different from zerowere then dropped systematically, as indicatedby 95% confidence intervals. Model differ-ences were explored using a nested modelapproach that compared constrained models(where parameters were dropped that were notsignificantly different from zero) with the fullmodel. Differences in χ2 values of the mod-els (Δχ2) in relation to differences in theirdegrees of freedom were tested to determinewhether constrained models resulted in a sig-nificant decrement in model fit compared to thefull model. If the Δχ2 value for a nested modelto the full model is nonsignificant, the nestedmodel accurately represents the data and is,therefore, a preferable fit compared to the fullmodel. Each nonsignificant path was droppedindividually. If there was a significant decre-ment in model fit (despite nonsignificant confi-dence intervals in the full model), the path wasconsidered significant. Paths in larger groupswere also dropped until they arrived at the bestfitting model.

The Bayesian information criterion (BIC) fitstatistic was also calculated (Raferty, 1995). Itwas did so because the χ2 statistic, which islikely to reject a model that fits the data well butimperfectly, is highly sensitive to sample sizeand is more likely to favor saturated models(Mulaik et al., 1989; Neale & Cardon, 1992).The Mx-provided BIC statistic evaluated the χ2

statistic for a model minus the product of thedegrees of freedom and the natural log likeli-hood of the sample size. Negative or smallerBIC values across a series of models fitted tothe same data are suggestive of a preferablefit, reflecting model parsimony for large sam-ple sizes.

Model assumptions

Twin models rely on several key assumptionsthat, if violated, could impact genetic and envi-ronmental estimates (e.g., Johnson, 2007). Onekey assumption (i.e., the equal environmentassumption) is that environmental effects arethe same across twin types. Violations of this

assumption could occur if MZ twins are morelikely to be treated similar to DZ twins, and ifthese environmental influences lead to greatersimilarity in MZ than in DZ twins’ friend sup-port and psychological distress. The validityof this assumption has been supported in twinstudies of psychiatric disorders (e.g., Hettema,Neale, & Kendler, 1995; Neiderhiser et al.,2004). Another assumption is that assortativemating (i.e., when individuals select partnerswith similar traits) is not present. Assortmentfor similar heritable traits has implications forsubsequent generations (i.e., genetic transmis-sion) because this form of assortment leadsto correlated genetic influences among parentsand, as a consequence, increases genetic relat-edness among their offspring than would beotherwise expected. In contrast, friend selec-tion does have not implications for future gen-erations in terms of direct genetic transmissionand, therefore, is not relevant for this study.Finally, twin models rely on the assumptionthat the effects of Gene×Environment interac-tion (G×E) in any of the key phenotypes ofinterest are negligible.

Results

Corrections for the effects of twins’ age, sex,marital status, and frequency of contact withco-twins were made by computing standard-ized residuals from the regression of scoreson these variables (McGue & Bouchard,1984). Table 2 displays summary statistics,including the means, standard deviations,and ranges for family support, friend sup-port, and psychological distress for the entiresample. Next, twins were selected randomlyto compare the means of friend support andpsychological distress between MZ and DZtwin groups. Results showed no differencesin means between MZ and DZ groups forfamily support, t(839)= 1.35; friend support,t(842)= –1.48; or psychological distress,t(839)= –1.71, p> .05. This suggests that MZand DZ twin groups do not differ, on aver-age, in their friend support or psychologicaldistress. Furthermore, correlation analysesdemonstrated significant associations betweenthese variables. Specifically, greater familysupport was correlated with greater friend


Table 2. Summary Statistics for Family Support, Friend Support, and Psychological Distress

Variable N Minimum to maximum M SD

Family support 1,646 –2.02 to 0.63 0.00 0.15Friend support 1,646 –2.16 to 0.97 0.00 0.64Psychological distress 1,654 –0.19 to 0.59 0.00 0.15

Note. This table presents means, standard deviations, and ranges for family support, friend support, and psychologicaldistress for the entire sample. These summary statistics reflect correction for positive skewness using log transformation(for psychological distress) and adjustment for the effects of twins’ age, sex, marital status, and frequency of contact withco-twins for each variable.

support (r = 0.32, p< .0001), and lower psy-chological distress was correlated with greaterfamily support (r = –22, p< .0001) and greaterfriend support (r = –21, p< .0001).

Intraclass correlations

Results from the intraclass correlations sug-gested that genetic influences are presentin the phenotypes of family support, friendsupport, and psychological distress. This wassuggested using MZ correlations for familysupport (r = 0.34, p< .0001), friend support(r = 0.18, p< .0001), and psychological dis-tress (r = 0.36, p< .0001) that were nearlytwice the magnitude of DZ correlations forfamily support (r = 0.21, p< .0001), friendsupport (r = 0.10, p= 0.013), and psycholog-ical distress (r = 0.20, p< .0001). Therefore,genetic influences account for approximately26% ([0.34 – 0.21]× 2= 0.26]) of the variancein family support, 16% of the variance infriend support ([0.18 – 0.10]× 2= 0.16), and32% ([0.36 – 0.20]× 2= 0.32) of variance inpsychological distress.

Model-fitting results

Prior to biometric model fitting, the presenceof sex limitation was tested, a phenomenonwhere the degree of genetic and environmentalinfluences on phenotypic associations varies bysex. A model where paths for men and womenvaried with one that constrained genetic andenvironmental paths to be the same in malesand females (Neale, Roysamb, & Jacobson,2006). The results showed no significant evi-dence of the presence of sex limitation in theassociations between family support, friend

support, and psychological distress; thus, anal-yses from biometric model fitting that con-strained estimates to be equal across same-sexand opposite-sex twin groups were presented.

Table 3 summarizes the parameter esti-mates and 95% confidence intervals from boththe full model (model fit: –2LL= 1356.85,df = 3315, BIC= –10609.39) and the bestfitting model (model fit: –2LL= 1359.79,df = 3319, BIC= –10621.54, Δχ2 = 2.94,p= .568). In the best fitting model, all sharedenvironmental paths (c11, c21, c31, c22, c32,and c33) could be dropped without resultingin a significant degradation in model fit (seeTable 3 and Figure 2). As such, findings fromthis investigation suggest that associationsbetween family support, friend support, andpsychological distress are not explained bycontributions from the shared environment.

Results from the best fitting model furthershowed that genetic factors accounted for 36%of the total variance in psychological distress[a31

2 + a322/(a31

2 + a322 + e31

2 a332)= (0.04+

0.32)/(0.04+ 0.32+ 0.01+ 0.62= 0.36]. A ge-netic factor (A1) that was related to familysupport, but unrelated to friend support con-tributed to psychological distress, suggestingthat genetic influences contribute to an associa-tion between family support and psychologicaldistress, independent of friend support. Fac-tor A1 explained 11% of the genetic variance[a31

2/(a312 + a32

2)= 0.04/(0.04+ 0.32)= 0.11]and 4% of the total variance in psycholog-ical distress (a31

2 = 0.212 = 0.04). Anothergenetic factor (A2) that was related to friendsupport, but unrelated to family support,accounted for a majority (88%) of thegenetic variance in psychological distress[a32

2/(a312 + a32

2)= 0.32/(0.04+ 0.32)= 0.88]


Table 3. Parameter estimates, 95% confidence intervals, and fit statistics from the full and bestfitting models

Variable 1:Family support

Variable 2:Friend support

Variable 3:Lower psychological

distress

Path Estimate (CI) Path Estimate (CI) Path Estimate (CI)

Full model latent factor

Variance explained by family supportA1 a11 0.58 (0.38, 0.68) a21 0.11 (–0.20, 0.31) a31 0.16 (–0.15, 38)C1 c11 0.16 (–0.40, 0.40) c21 0.21 (–0.44, 0.44) c31 0.20 (–0.50, 0.50)E1 e11 0.80 (0.74, 0.86) e21 0.28 (0.19, 0.38) e31 0.12 (0.04, 0.21)

Variance explained by friend supportA2 a22 0.33 (–0.49, 0.49) a32 0.34 (–0.64, 0.64)C2 c22 0.10 (–0.37, 0.37) c32 0.16 (–0.45, 0.45)E2 e22 0.86 (0.81, 0.91) e32 0.01 (–.06, .09)

Residual variance in psychological distressA3 a33 0.41 (–0.59, 0.59)C3 c33 0.00 (–0.45, 0.45)E3 e33 0.79 (0.73, 0.84)

Reduced-model latent factor

Variance explained by family supportA1 a11 0.57 (0.47, 0.64) a21 — a31 0.21 (0.08, 0.33)C1 c11 — c21 — c31 —E1 e11 0.82 (0.77, 0.88) e21 0.38 (0.32, 0.43) e31 0.11 (.04, 0.19)

Variance explained by friend supportA2 a22 0.28 (0.19, 0.37) a32 0.57 (0.48, 0.64)C2 c22 — c32 —E2 e22 0.88 (0.84, 0.91) e32 —

Variance explained by family supportA3 a33 —C3 c33 —E3 e33 0.79 (0.73, 0.84)

Note. Latent factors and paths correspond to Figure 1. Each path name has three components: a lowercase letter, followedby two subscripted numbers. The lowercase letter and first subscripted number refer to a specific latent factor: A, C, orE. The second subscripted number refers to the specific manifest variable (identified by its order in the model) that islinked to the latent factor via the path. For example, path a21 links latent factor A1 to the second variable in the model A1.For the most parsimonious model, dashes (–) indicate parameters that could be dropped without resulting in a significantincrease in the χ2 value. CI= 95% confidence intervals.

and 32% of the total variance in this phenotype(a32

2 = 0.572 = 0.32). As such, genetic influ-ences on the link between friend support andpsychological distress are distinct from familysupport. Furthermore, a majority (63%) of thetotal variance in psychological distress wereascribable to nonshared environmental effects[e31

2 + e332/(a31

2 + a322 + e31

2 + e332)= (0.01

+ 0.62)/(0.04+ 0.32+ 0.01+ 0.62)= 0.63]. Anonshared environmental factor (E1) that wasrelated to family support also contributed tofriend support and psychological distress.This indicates that nonshared environmentalinfluences explain the association betweenfamily support, friend support, and psy-chological distress. Factor E1 accounted


Family

support

A1

.57

(.47, .64)

.82

(.77, .88).38

(.32, .43)

.21

(.08, .33)

.28

(.19, .37)

.88 (.84, .91)

.57

(.48, .64).79

(.73, .84)

.11

(.04, .19)

E1 A2 E2 E3

Friend

support

Lower

Psychological Distress

Figure 2. Parameter estimates and 95% confidence intervals from the best fitting model for onetwin in a pair. Path estimates and 95% confidence intervals in parentheses correspond to resultsfrom the reduced model in Table 2. Path estimates not significantly different from zero are notincluded here.

for 1% of the nonshared environmen-tal variance in psychological distress[e31

2/(e312+ e33

2)= 0.01/(0.01+ 0.62)= 0.01],as well as 1% of the total variance in this phe-notype (e31

2 = 0.11= .01). An additionalnonshared environmental additional factor(E3) contributed the remaining variance in psy-chological distress. This factor was responsiblefor a majority (98%) of both the nonsharedenvironmental variance in [e33

2/(e312 + e33

2)=0.62/(.01+ 0.62)= .01] and total variance(e33

2 = 0.79= 0.62) in psychological distress.This finding indicates that residual variancein psychological distress is explained by thenonshared environment.

Discussion

This study demonstrated that genetic con-tributions explain an association betweenemotional support from friends and psy-chological distress that is independent ofemotional support from family members.This suggests that genetically based mech-anisms that cannot explain family supportare important for understanding how the linkbetween friend support and psychologicaldistress arises. Findings from the present studyalso revealed that nonshared environmentalinfluences explain a relation between family

support, friend support, and psychologicaldistress. This suggests a new mechanismrelating distress and social support, such thatnonshared environmental influences shapelinks between general aspects of social supportfrom different interpersonal relationships andpsychological distress.

This study showed that the genetic linkbetween friend support and psychologicaldistress is not shared with social support fromfamily members. Active rGE may explain thedistinct genetically based association for friendsupport because friendships are selected, butfamily relationships are ascribed. An activerGE explanation for the current findings isconsistent with previous studies that havesuggested that adolescents’ heritable charac-teristics influence what particular peer groupsthe adolescents tend to select (e.g., Brendgen,2012). Thus, individuals’ heritable character-istics (e.g., heritable personality traits) mayinfluence their choices of certain friendshipsover others and, in turn, their friend supportand psychological distress. As personalitytraits were not measured in this study, it ispossible only to speculate what traits mayplay a role in individuals’ friend choices.Extraversion is one plausible candidate giventhat previous research has shown that individu-als with higher levels of heritable extraversion


may tend to select friends who are also moreextraverted (Selfhout et al., 2010). In turn,these individuals may experience higher levelsof support from the friends they have selectedand lower psychological distress.

A nonshared environmental factoraccounted for an association between familysupport, friend support, and psychologicaldistress. This suggests that nonshared environ-mental influences operate in shaping generalaspects of social support in relation to psy-chological distress. Accordingly, previousresearch showed that nonshared environmentalfactors accounted for similarities in the qualityof children’s relationship with their familyand friends (Pike & Atzaba-Poria, 2003). Withrespect to the current findings, nonshared envi-ronmental influences may include individuals’unique environmental experiences duringadulthood. Such unique experiences (e.g.,negative life events) may operate by shap-ing psychological distress and perceptionsof support from family and friends, in turn.For example, psychological distress triggeredby job loss is often associated with socialwithdrawal (Russell, 1999). Thus, individuals’own negative life events may lead to increasedpsychological distress and social withdrawal,which may influence how individuals perceivetheir social support networks. Nonshared envi-ronmental factors may also involve the impactof unique experiences during childhood. Forexample, differential parental treatment ofeach twin has been linked to later psycholog-ical distress (Reiss et al., 1995). Therefore,individuals’ own experiences in childhoodmay give rise to later psychological distressand their perceptions of family and friends.

Implications

An implication of these results is that inter-ventions focused on reducing psychologicaldistress may need to consider how heritablecharacteristics, which are unrelated to familysupport, shape perceived and received friendsupport. In addition, future research needsto assess specific heritable characteristicsunderlying the unique association betweenfriend support and psychological distress.Furthermore, when nonshared environmental

factors are found to contribute to associa-tions between measures of social supportand adjustment, this suggests the importantrole of environments that are unique to eachindividual in these associations (Leve, Harold,Ge, Neiderhiser, & Patterson, 2010). As such,these results also indicate that environmentsunique to each individual are important in theconsideration of how support from both familyand friends is tied to psychological distress.Additional research is needed to examine whatparticular aspects of each person’s uniqueenvironments are critical for shaping hisor her family support, friend support, andpsychological distress.

Limitations

Certain limitations of this study need to beaddressed. First, this study used a community-based sample, which warrants caution aboutgeneralizing findings to clinical populations.For example, the extent to which geneticfactors shape the association between family,friendships, and psychological distress mayvary between community and clinical popula-tions. At the same time, the strength of usinga community-based sample is the ability toshed light on how family support and friendsupport are linked to psychological distressin the general population. Second, given thatpsychological distress is a general measure ofmental health (Kendler et al., 1994), caution iswarranted about generalizing findings to linksbetween family support, friend support, andspecific mental disorders. Third, the closenessof the twins may vary by the size of the twins’family; therefore, future studies should includefamily size as a control variable. Fourth, owingto the self-report nature of the data, associa-tions between family support, friend support,and psychological distress are potentiallyvulnerable to inflation (Lindell & Whitney,2001). At the same time, one possible waythat common method variance operates is viagenetically influenced self-report tendencies.For example, the same genetic influences thatshape tendencies to report lower levels ofsupport may also influence greater psycho-logical distress. Genetically informed designsthat use multi-informant and observational


reports of the study variables would affordmore stringent tests of their intercorrelations.Fifth, these analyses used cross-sectional data;thus, conclusions about the direction of effectsbetween family support, friend support, andpsychological distress and whether the relativeinfluences from genetic and environmentaleffects on this association vary over timecannot be drawn.

This study demonstrated that the associa-tion between friend support and psychologicaldistress is accounted for by genetic influencesthat are distinct from family support. Thissuggests novel genetically based mechanismson the link between friend support and psycho-logical distress. For example, these heritablecharacteristics may play a role in shapingindividuals’ selection of certain friends overothers and, in turn, their friend support andpsychological distress. In addition, nonsharedenvironmental contributions explained theassociation between family support, friendsupport, and psychological distress. As such,environmental influences that are unique toeach individual (e.g., their unique relationshipswith family and friends) are also important forunderstanding how social support from bothfamily and friends is related to psychologicaldistress.

ReferencesBergeman, C. S., Plomin, R., Pederson, N. L., & McClearn,

G. E. (1991). Genetic mediation of the relationshipbetween social support and psychological well-being.Psychology and Aging, 6, 640–646.

Brendgen, M. (2012). Genetics and peer relations: Areview. Journal of Research on Adolescence, 22,419–437.

Brim, O. G., Ryff, C. D., & Kessler, R. C. (Eds.). (2004).How healthy are we? A national study of well-beingat midlife. The John D. and Catherine T. MacArthurfoundation series on mental health and development.Studies on successful midlife development. Chicago,IL: University of Chicago Press.

Charles, S. T., Piazza, J. R., Mogle, J., Sliwinski, M. J.,& Almeida, D. M. (2013). The wear and tear of dailystressors in mental health. Psychological Science, 24,733–741.

Cohen, S. (2004). Social relationships and health. Ameri-can Psychologist, 59, 676–684.

Ganiban, J. M., Ulbricht, J. A., Spotts, E. L., Lichtenstein,P., Reiss, D., Hansson, K., & Neiderhiser, J. M. (2009).Understanding the role of personality in explainingassociations between marital quality and parenting.Journal of Family Psychology, 23, 646–660.

Hettema, J. M., Neale, N. C., & Kendler, K. S. (1995).Physical similarity and the equal-environment assump-tion in twin studies of psychiatric disorders. BehavioralGenetics, 25, 327–335.

Horwitz, B. N., & Neiderhiser, J. M. (2011).Gene-environment interplay, family relationships,and child adjustment. Journal of Marriage andFamily, 73, 804–816.

Horwitz, B. N., Reynolds, C. A., Neiderhiser, J. M., &Charles, S. (in press). Friend support and psychologicaldistress in a United States adult twin sample. PersonalRelationships.

Johnson, W. (2007). Genetic and environmental influenceson behavior: Capturing all the interplay. PsychologicalReview, 114, 423–440.

Kendler, K. S., & Baker, J. H. (2007). Genetic influenceson measures of the environment: A systematic review.Psychological Medicine, 37, 615–626.

Kendler, K. S., Neale, M. C., Kessler, R. C., Heath, A. C.,& Eaves, L. J. (1994). Parental treatment and the equalenvironment assumption in twin studies of psychiatricillness. Psychological Medicine, 24, 579–590.

Kessler, R. C., Andrews, G., Colpe, L. J., Hiripi, E.,Mroczek, D. K., Normand, S. L. T., ... Zaslavsky,A. M. (2002). Short screening scales to monitorpopulation prevalences and trends in non-specificpsychological distress. Psychological Medicine, 32,959–976.

Kessler, R. C., Gilman, S. E., Thornton, L. M., & Kendler,K. S. (2004). Health, well-being, and social respon-sibility in the MIDUS twin and sibling subsamples.In O. G. Brim, C. D. Ryff, & R. C. Kessler (Eds.),How healthy are we? A national study of well-being atmidlife. The John D. and Catherine T. MacArthur foun-dation series on mental health and development. Stud-ies on successful midlife development (pp. 124–152).Chicago, IL: University of Chicago Press.

Lepore, S. J., Evans, G. W., & Schneider, M. L. (1991).Dynamic role of social support in the link betweenchronic stress and psychological distress. Journal ofPersonality and Social Psychology, 61, 899–909.

Leve, L. D., Harold, G. T., Ge, X., Neiderhiser, J. M., &Patterson, G. (2010). Refining intervention targets infamily-based research. Perspectives on PsychologicalScience, 11, 516–526.

Lindell, M. K., & Whitney, D. J. (2001). Accounting forcommon method variance in cross-sectional researchdesigns. Journal of Applied Psychology, 86, 114–121.

Lykken, D. T., Bouchard, T. J., McGue, M., & Tellegan,A. (1990). The Minnesota Twin Registry: Some initialfindings. Acta Geneticae Medicae et Gemellologiae(Roma), 39, 35–70.

Marceau, K., Horwitz, B. N., Narusyte, J., Ganiban, J.M., Spotts, E. L., Reiss, D., & Neiderhiser, J. M.(2013). Gene-environment correlation underlying theassociation between parental negativity and adoles-cent externalizing problems. Child Development, 84,2031–2046.

McGue, M., & Bouchard, T. J., Jr. (1984). Adjustmentof twin data for the effects of age and sex. BehaviorGenetics, 14, 325–343.


Mroczek, D. K., & Kolarz, C. M. (1998). The effect ofage on positive and negative affect: A developmentalperspective on happiness. Journal of Personality andSocial Psychology, 75, 1333–1349.

Mulaik, S. A., James, L. R., Alstine, J. V., Bennett, N.,Lind, S., & Stilwell, C. D. (1989). Evaluation ofgoodness-of-fit indices for structural equation models.Psychological Bulletin, 105, 430–445.

Neale, M. C., Boker, S., Xie, G., & Maes, H. (2003). Mx:Statistical modeling. Richmond, VA: Department ofPsychiatry.

Neale, M. C., & Cardon, L. R. (1992). Methodology forgenetic studies of twins and families. Boston, MA:Kluwer Academic.

Neale, N. C., Roysamb, E., & Jacobson, K. (2006). Mul-tivariate genetic analysis of sex limitation and G×Einteraction. Twin Research and Human Genetics, 9,481–489.

Neiderhiser, J. M., Reiss, D., Pederson, N. L., Lichtenstein,P., Spotts, E. L., Hansson, K., ... Ellhammer, O. (2004).Genetic and environmental influences on mothering ofadolescents: A comparison of two samples. Develop-mental Psychology, 40, 335–351.

Norton, T. R., Manne, S. L., Rubin, S., Hernandez, E.,Carlson, J., & Bergeman, C. (2005). Ovarian cancerpatients’ psychological distress: The role of physicalimpairment, perceived unsupportive family and friendbehaviors, perceived control, and self-esteem. HealthPsychology, 24, 143–152.

Pike, A., & Atzaba-Poria, N. (2003). Do sibling and friendrelationships share the same temperamental origins? Atwin study. Journal of Child Psychology and Psychia-try, 44, 598–611.

Plomin, R., DeFries, J. C., & Loehlin, J. C. (1977).Genotype–Environment interaction and correlation inthe analysis of human behavior. Psychological Bul-letin, 84, 309–322.

Raferty, A. E. (1995). Bayesian model selection in socialresearch. Sociological Methodology, 25, 111–163.

Reiss, D., Hetherington, M. E., Plomin, R., Howe, G. W.,Simmen, S. J., Henderson, S. H., ... Law, T. (1995).Genetic questions for environmental studies: Differ-

ential parenting and psychopathology in adolescence.Archives of General Psychiatry, 52, 925–936.

Reiss, D., & Leve, L. D. (2007). Genetic expressionoutside the skin: Clues to mechanisms of Geno-type×Environment interaction. Development andPsychopathology, 19, 1005–1027.

Rijsdijk, F. V., Snieder, H., Ormel, J., Sham, P., Gold-berg, D. P., & Spector, T. D. (2003). Genetic andenvironmental influences on psychological distressin the population: General Health Questionnaireanalyses in UK twins. Psychological Medicine, 33,793–801.

Ritsner, M., Modai, I., & Ponizovsky, A. (2000). Thestress-support patterns and psychological distress ofimmigrants. Stress and Health, 16, 139–147.

Russell, H. (1999). Friends in low places: Gender, unem-ployment and sociability. Work, employment and soci-ety, 13, 205–224.

Scarr, S., & McCartney, K. (1983). How people maketheir own environments: A theory of genotype greaterthan environment effects. Child Development, 54,424–435.

Selfhout, M., Burk, W., Branje, S., Denissen, J., Van Aken,M., & Meeus, W. (2010). Emerging late adolescentfriendship networks and Big Five personality traits: Asocial network approach. Journal of Personality, 78,509–538.

Shaw, B. A., Krause, N., Chatters, L. M., Connell, C. M., &Ingersoll-Dayton, B. (2004). Emotional support fromparents early in life, aging, and health. Psychology andAging, 19, 4–12.

Spotts, E. L., Neiderhiser, J. M., Ganiban, J., Reiss,D., Lichtenstein, P., Hansson, K., ... Pederson, N.L. (2005). Accounting for depressive symptoms inwomen: A twin study of associations with interper-sonal relationships. Journal of Affective Disorders, 82,101–111.

Yuh, J., Neiderhiser, J. M., Spotts, E. L., Pederson, N. L.,Lichtenstein, P., Hansson, K., ... Reiss, D. (2008). Therole of temperament and social support in depressivesymptoms: A twin study of mid-aged women. Journalof Affective Disorders, 106, 99–105.

understanding associations among family support, friend support,...

Documents