INFLUENCE OF MOTHER'S DEPRESSION ON HER...

A R T I C L E

INFLUENCE OF MOTHER’S DEPRESSION ON HER REPORTS OF FATHER

INVOLVEMENT AND CHILD BEHAVIORAL PROBLEMS: A LATENT STATE-TRAIT

APPROACH

MARYNA RASKIN, NATHAN E. FOSSE, AND M. ANN EASTERBROOKSTufts University

ABSTRACT: Research on father involvement has shown positive effects on child development. Because fathers in high social risk samples may behard to recruit or retain in studies, the literature often has relied on maternal report of father involvement. A major limitation of this approach is thatunobserved traits of the reporting mothers may distort the real associations between father involvement and children’s development. Using maternaldata from a large, longitudinal sample (N = 704) of low-income, young mothers, we evaluated the degree to which a stable depressive trait affectedthe link between mother-reported measures of father involvement and child problems. Three waves of maternal depression data were used to fit alatent state-trait model of depression, allowing for separate estimates of occasion-specific symptoms and stable depressive trait. A latent regressionanalysis which did not control for this trait revealed a link between father involvement and child problems similar in magnitude to the links reported inthe literature. However, this association disappeared once we accounted for the effect of maternal depressive trait. Results suggest that studies usingmaternal reports of both father and child behaviors should control for such confounding effects. We elaborate on these findings in the conclusion andoffer suggestions for future research on the role of fathers in child development.

RESUMEN: La investigacion sobre la aportacion del papa demuestra efectos positivos en el desarrollo del nino. Como resulta difıcil reclutar o reteneren un estudio a papas en grupos muestras de alto riesgo social, la literatura investigativa a menudo confıa en reportes maternos sobre la participaciondel papa. Una considerable limitacion en este acercamiento es que las caracterısticas no observadas de las madres que reportan pudieran distorsionarlas verdaderas asociaciones entre la participacion del papa y el desarrollo del nino. Usando informacion de las madres en un grupo muestra longitudinalextenso (n = 704) de madres jovenes de bajos recursos, evaluamos el grado al cual una caracterıstica depresiva estable afecto el lazo entre las medidasreportadas por la madre acerca de la participacion del papa y los problemas del nino. Se usaron tres procederes de informacion de la depresion maternalpara adecuar un modelo de caracterıstica estable latente de depresion, lo cual permitirıa estimados separados de sıntomas de una ocasion especıficay caracterısticas depresivas estables. Un analisis que no fue parte del control de esta caracterıstica revelo un vınculo entre participacion del padre yproblemas del nino. Sin embargo, esta asociacion desaparecio una vez se tomo en cuenta el efecto de la caracterıstica depresiva maternal. Los resultadosrevelan que los estudios que usan reportes maternos tanto de conductas del papa como del nino deben proveer control para tales confusos efectos. En laconclusion, elaboramos sobre estos resultados y ofrecemos sugerencias para la futura investigacion sobre el papel de los papas en el desarrollo del nino.

RESUME: Les recherches sur l’implication du pere ont eu des effets positifs sur le developpement de l’enfant. Parce que les peres dans des echantillonsa haut risque social peuvent s’averer difficule a recruter ou a garder dans des etudes, les recherches reposent souvent sur les compte-rendus que fontles meres du degre d’implication du pere. Une limite importante a cette approche se trouve dans le fait que les traits inobserves du rapport des merespeuvent deformer les liens reels entre l’implication du pere et le developpement de l’enfant. En utilisant des donnees maternelles d’un grand echantillonlongitudinal (n = 704) de jeunes meres de milieu defavorise, nous avons evalue le degre auquel un trait depressif stable affectait le lien entres les mesuresrapportees par les meres de l’implication du pere et les problemes de l’enfant. Trois vagues de donnees de depression maternelle ont ete utilisees pourcorrespondre a un modele de depression d’etat-de trait latent, permettant d’avoir des estimations separees de symptomes occasionnels-specifiques etd’un trait depressif stable. Une analyse qui n’a pas controle ce trait a revele un lien entre l’implication du pere et les problemes de l’enfant. Cependant,ce lien a disparu une fois que nous avons inclu l’effet du trait depressif maternel. Les resultats suggerent que les etudes utilisant des compte-rendusmaternels des comportements d’a la fois le pere et l’enfant devraient controler de tes effets confondants. Nous elaborons sur ces resultats dans laconclusion et offrons des suggestions pour les recherches a venir sur le role des peres dans le developpement de l’enfant.

Direct correspondence to: Maryna Raskin, Department of Child Development, 177 College Avenue, Tufts University, Medford, MA 02155; e-mail:[email protected].

INFANT MENTAL HEALTH JOURNAL, Vol. 36(1), 88–103 (2015)C© 2014 Michigan Association for Infant Mental HealthView this article online at wileyonlinelibrary.com.DOI: 10.1002/imhj.21487

88

Maternal Report of Father Involvement • 89

ZUSAMMENFASSUNG: Die Forschung zur vaterlichen Beteiligung hat auf positive Effekte fur die kindliche Entwicklung verwiesen. Weil Vater insozialen Hochrisiko-Stichproben schwer zu rekrutieren oder in Studien zu halten sind, setzt die Literatur oft auf den mutterlichen Bericht uber dieBeteiligung des Vaters. Eine wichtige Einschrankung dieses Ansatzes ist, dass unbeobachtete Merkmale der berichterstattenden Mutter die wirklichenZusammenhange zwischen der vaterlichen Beteiligung und der Entwicklung der Kinder verzerren konnen. Anhand von mutterlichen Daten aus einergroßen Langsschnittstichprobe (n = 704) von jungen Muttern mit niedrigem Einkommen bewerteten wir, in welchem Ausmaß, ein stabiles depressivesMerkmal die Verbindung zwischen mutterlichen Berichten uber die vaterliche Beteiligung und die Probleme des Kindes beeinflusst. Drei Wellen vonDaten zur mutterlichen Depression wurden verwendet, um ein latentes State-Trait-Modell der Depression anzulegen, dass getrennte Schatzungen furanlassspezifische Symptome und stabile depressive Merkmale erlaubt. Eine Analyse, die nicht fur dieses Merkmal kontrollierte, zeigte eine Verbindungzwischen vaterlicher Beteiligung und Problemen des Kindes. Jedoch verschwand diese Assoziation wenn wir den Effekt der mutterlichen Depressivitat(Trait) berucksichtigten. Die Ergebnisse legen nahe, dass Studien, die mutterliche Berichte uber Verhaltensweisen von Vatern und Kindern nutzen,fur solche storenden Effekte kontrollieren sollten. Wir gehen auf diese Erkenntnisse in der Schlussfolgerung naher ein und bieten Anregungen fur diezukunftige Forschung zur Rolle von Vatern in der kindlichen Entwicklung.

ABSTRACT: ��:�� !� ��" ��#$%&

'��(�� )��*� �+,)-,�

�.��!� ��(n = 704)��/&0�� ,!��(��!� ��)��1��(��,!�3��/&0��,!��-��2/3a latent state-trait model��(��4��occasion-specific symptoms��,!��.��516%&3� ��1��

��,!��7��-��!�

��,��8,��confounding effects�516%&3(9�.��7�� 9��:��!� ��(��

��: ��,��,��,��(N =704),��,�� –��,��,��,��,��,��,��

* * *

During the last several decades, researchers and practitionershave devoted increased attention to the role of fathers in the devel-opment and well-being of children (Cabrera & Tamis-LeMonda,2013; Downer, Campos, McWayne, & Gartner, 2008). Studies havedocumented that fathers play a significant role in children’s devel-opmental functioning, including socioemotional, behavioral, andacademic domains (Cabrera, Shannon, & Tamis-LeMonda, 2007;Leidy, Schofield, & Parke, 2013; McWayne, Downer, Campos,& Harris, 2013). Most research documenting the benefits of in-volved father–child relationships has arisen primarily from studiesof low-risk families; that is, married, White couples with moder-ate and high socioeconomic status (SES) (Easterbrooks, Raskin, &McBrian, 2014; Roggman, Bradley, & Raikes, 2013). However, the

influence of fathers on their young children’s development appliesquite broadly across a range of socioeconomic and family struc-ture configurations (e.g., married, divorced, nonresidential). Thereis some suggestion that father involvement is even more criticalin low-income or in higher risk contexts (Fitzgerald, Bockneck,Cabrera, & Tamis-LaMonda, 2013; Howard, Lefever, Borkowski,& Whitman, 2006; Martin, Ryan, & Brooks-Gunn, 2010).

CHALLENGES IN RESEARCH ON HIGH-RISK FATHERS

A major challenge in studying fathers in socially disadvantagedfamilies (e.g., unwed couples, low SES, or adolescent parents)is that these fathers often are hard to recruit or retain in studies

Infant Mental Health Journal DOI 10.1002/imhj. Published on behalf of the Michigan Association for Infant Mental Health.

90 • M. Raskin, N.E. Fosse, and M.A. Easterbrooks

(Carlson & McLanahan, 2010; Costigan & Cox, 2001; Hernandez& Brandon, 2002). Many unwed fathers are less likely to live withtheir children or to be involved in parenting, given that their rela-tionships with the mothers of their babies are often unstable (Hof-ferth & Goldscheider, 2010; Kalil, Ziol-Guest, & Coley, 2005).If the primary study contact is the mother, she may not provideaccess to contact information about the father of the child. Further,employment hours of parents working outside the home may makeit relatively more difficult for fathers to participate (since womenare more likely to curtail their employment when they have youngchildren). As a result, there is a concern that the current literatureon fathers may be biased toward lower risk families and more in-volved fathers (Roggman et al., 2013). In particular, fathers whoare nonresidential or unmarried are underrepresented in studies thatuse paternal reports of their own parenting; therefore, this researchlikely does not characterize all men’s involvement in the lives oftheir children.

For reasons outlined earlier, studies of nonresident fathersoften have to rely on maternal reports of parental involvement(Cabrera, Mitchell, Ryan, Shannon, & Tamis-LeMonda, 2008).Maternal reports of fathers’ involvement can be an effective re-search tool (Martin, Brazil, & Brooks-Gunn, 2013; Mitchell et al.,2007). However, the validity of maternal report of father involve-ment is a methodological and conceptual concern in fathering re-search (Hernandez & Coley, 2007). Especially among unmarriedcouples living apart, discrepancies between maternal and parentalreports can be large (Coley & Morris, 2002; Kitterød & Lyngstad,2014).

CONCERNS ABOUT REPORTER DISCREPANCIES

The fact that maternal and paternal reports on the same measuresare often discrepant has been a known issue in child develop-ment literature, extending to topics far beyond the question offather involvement. Mothers’ and fathers’ ratings of their childrenalso tend to be quite different (Achenbach, Edelbrock, & Howell,1987; Grigorenko, Geiser, Slobodskaya, & Francis, 2010; Thurber& Sheehan, 2012). For example, fathers tend to rate children’sinternalizing behavior and psychiatric symptoms lower than domothers (Huberty, Austin, Harezlak, Dunn, & Ambrosius, 2000;Jensen, Xenakis, Davis, & Degroot, 1988; Seiffge-Krenke & Koll-mar, 1998), especially when reporting on sons (Luoma, Koivisto,& Tamminen, 2004).

Various explanations for the discrepancies in mother and fa-ther reports have been proposed. Parents might differ in their will-ingness to report problems; their answers might be influenced bythe quality of their relationship with the child, the amount of timethey spend with the child, or the situations in which they typicallyinteract with the child (Treutler & Epkins, 2003; Youngstrom,Loeber, & Stouthamer-Loeber, 2000). Gender roles and social-ization may affect both the actual behaviors of the children andhow these behaviors are perceived by the parents of each gen-der (Thompson, MacLaren, Harris, & Kain, 2009). De Los Reyesand Kazdin’s (2005) theoretical model combines these differencesunder an overarching category of informant attributional bias. At-

tributional biases are cognitive processes that affect memory recalland influence how individuals perceive others.

The concept of rater attributional bias is at the core of thedepression-distortion hypothesis (Jensen et al., 1988), one of themost investigated explanations for informant discrepancies in childand adolescent psychiatric literature. This hypothesis postulatesthat a caregiver’s depression promotes negative bias and causesparents to overstate or overreport problems in their children. Thisframework is grounded in cognitive theories of depression, whichsuggest that affected individuals are characterized by attentionalbiases for negative material, mood-congruent memory recall, andcognitive deficits in how they process information (Dalgleish,Hill, Golden, Morant, & Dunn, 2011; Gotlib & Joormann, 2010).These cognitive difficulties have been proposed to be a stable traitof depression-vulnerable individuals, as they characterize evenasymptomatic individuals who were depressed in the past (Ehring,Tuschen-Caffier, Schnulle, Fischer, & Gross, 2010). In clinicalliterature, this trait has been referred to as a “depressogenic cog-nitive style” (Hankin, Fraley, & Abela, 2005) or a “depressive at-tributional style” (Abela & Seligman, 2000; Peterson et al., 1982;Sweeney, Anderson, & Bailey, 1986), and is characterized by neg-ative processing and impaired judgment that manifests across mul-tiple self-referent domains.

The depression-distortion hypothesis generally has been ap-plied to a depressed mother’s overreporting of child behavior prob-lems (Mowbray, Lewandowski, Bybee, & Oyserman, 2005; Ord-way, 2011; van der Toorn et al., 2010); however, it also might ap-ply to mothers’ perceptions of father involvement. Some authorshave suggested that a mother’s perceptions of paternal involve-ment and her satisfaction with the relationship could be affectedby her emotional well-being (Coley & Morris, 2002; Fagan & Lee,2010; Lamb & Lewis, 2010). Several correlational studies havefound associations between greater maternal depression and lowermaternal perceptions of father involvement (Malik et al., 2007;Mezulis, Hyde, & Clark, 2004).

STATISTICAL APPROACHES TO ACCOUNTING FORREPORTER EFFECTS

Modern statistical techniques allow researchers to separately modelvarious sources of variance, including the variance that is due toreporter effects (called method-related variance; see Eid & Diener,2006). These modeling techniques are grounded in classical testtheory, which postulates that research measurements can be de-composed into the true score and measurement error (Campbell& Fiske, 1959; Carmines & Zeller, 1979; DeVellis, 2011; Kline,2005). The true score is the quantity that would have been obtainedif there was no measurement error; it is a theoretical (unobserved)construct and is commonly referred to as the “latent” variable.Measurement error is a broad term that refers to deviations fromthe true scores due to two sources: nonrandom (i.e., confounding)and random (i.e., chance) influences.

Researchers model latent variables by measuring multipleindicators of the same construct, which are commonly referred toas “observed” or “manifest” variables. An example of manifest



variables would be a set of questionnaire items that a researcherdevelops to tap into the same latent construct (e.g., a questionnairein which all items pertain to symptoms of depression). Multipleindicators of the same phenomenon also could be obtained fromstructurally different raters (e.g., behavior problems reported bythe child’s mother, father, and teacher) or from the same ratersover multiple occasions.

It is conjectured that manifest variables designed to measurethe same construct will be intercorrelated and that the varianceshared by these items is a reflection of a common cause (i.e., theinfluence of the latent construct; T.A. Brown, 2006). The remain-ing variance can be further decomposed into (a) method-specificvariance (e.g., effect of the reporter, if more than one reporter); (b)occasion-specific variance (e.g., effect of time, if measurementsoccurred on more than one occasion); (c) indicator-specific vari-ance (e.g., the effect of specific questions chosen to assess thebehavior); and (d) random error (Little, 2013).

For illustrative purposes, we present a simplified model ofvariance decomposition for a study involving two raters (e.g.,mother and father) reporting on the same three-item questionnaireabout child behavior problems (see Figure 1). As shown in thefigure, the construct of child behavior problems is shared by eachreporter and is modeled as a “true trait” by extracting the com-mon variance from all six indicators. The variance due to specificvantage points of the mother and father perceptions is modeled asseparate factors to account for any variance in the three item scoresthat is unique to the rater (Little, 2013). Finally, the variance dueto the indicator-specific effects also is partitioned out. This prac-tice is recommended because the same questionnaire items shareidiosyncratic components with themselves that may not be sharedwith remaining items (Geiser & Lockhart, 2012; Pohl & Steyer,2010). As such, if this indicator-specific variance is not modeled,it may lead to correlated error terms, which would interfere withmodel estimation (LaGrange & Cole, 2008).

The modeling of rater effects is motivated by the concernthat reports by different informants are confounded by systematic,person-specific characteristics (Pohl & Steyer, 2010). However,rater effects can be separated from the true score and other sourcesof variance only if a researcher has data from multiple raters (as inFigure 1). For this reason, the use of multiple informants has beenproposed as a best practice since Campbell and Fiske’s (1959) sem-inal publication on the multitrait-multimethod (MTMM) matrix(Cole, Ciesla, & Steiger, 2007; Grigorenko et al., 2010). However,often researchers do not have access to multiple raters reportingon the same construct. A major concern in such cases is that thetrue scores on all constructs in the study are “contaminated” (i.e.,confounded) by the variance related to the informant, which leadsto overestimated associations between constructs (Little, 2013).In other words, the strength of association between two variableswill be stronger than it is in reality because both variables shareadditional common variance attributable to person-specific char-acteristics of the informant.

Although it is impossible to investigate the degree of rater biasin single-reporter studies the way that it can done in multimethod

studies, assessing the same reporter over multiple occasions canhelp mitigate the concern about confounding by person-specificvariance (Eid, Geiser, & Nussbeck, 2009). Specifically, by col-lecting longitudinal data on the characteristics believed to bias theinformant’s responses, a researcher can explicitly model, and thencontrol for, the stable person-specific variance due to this trait.This task can be accomplished by using latent state-trait (LST)modeling.

LST models are a class of longitudinal models that attempt topartition variance of a variable into components that remain stableover time (traits) and factors that change over time (states). Thesemodels allow researchers to determine the degree to which differ-ences in scores on a given measure depend on occasion-specific ef-fects versus stable person-specific effects. The LST theory (Steyer,Ferring, & Schmitt, 1992; Steyer, Majcen, Schwenkmezger, &Buchner, 1989; Steyer & Schmitt, 1990; Steyer, Schmitt, & Eid,1999) recognizes that self-reports may be distorted systematicallyby stable response dispositions that remain consistent across situa-tions whereas fluctuation in occasion-specific effects, on the otherhand, reflects variance that is independent of this trait. LST the-ory has been widely applied in research on personality, emotion,subjective well-being, psychopathology, and other fields of psy-chology (Geiser & Lockhart, 2012).

An illustrative example of an LST model is shown inFigure 2. As shown, the model decomposes the variance in thescores measured at three time points into the variance that is dueto stable individual differences (trait) and the variance that is dueto the situation effects at each occasion of measurement (state).Figure 2 also shows three indicator-specific factors. Multiple ap-proaches to modeling indicator-specific variance have been pro-posed (e.g., Geiser & Lockhart, 2012; LaGrange & Cole, 2008);here, we illustrate the approach that has been developed to over-come technical and interpretative problems of MTMM designs(Geiser, Eid, & Nussbeck, 2008; Geiser & Lockhart, 2012). Inthis technique (called the M − 1 approach), one indicator is cho-sen as a reference method against which other method factors arecontrasted. As such, the nonreference indicator-specific factors aremodeled to represent the part of the variance of these indicators thatis not shared with the state factors (see Coffman & MacCallum,2005; Little, Cunningham, Shahar, & Widaman, 2002).

THE PRESENT STUDY

In this article, we seek to address the issue of possible report biasinduced when mothers provide information on both the fathersand their children. Statistically, we are guided by the LST model-ing framework, which postulates that stable response dispositions(traits) can be modeled explicitly as the variance shared by in-dicators assessed over time. Theoretically, this article is guidedby the cognitive theories of depression, and specifically by thedepression-distortion hypothesis which postulates that a stable de-pressive trait may negatively bias the way a caregiver describesboth her child’s behavior and her partner’s father involvement.



FIGURE 1. A conceptual model depicting six manifest variables measuring one trait by two methods. Indicator-specific variance is modeled for Indicators 2 and 3,keeping Indicator 1 as a reference.

We use data on maternal depressive symptoms measured threetimes over the course of 2 years to evaluate our proposed model.Specifically, our conceptual model (illustrated in Figure 3) postu-lates that the symptoms mothers report on any given occasion aresimultaneously influenced by the occasion (e.g., mood during theinterview) and the stable depressive cognitive bias (trait) said to de-scribe depression-vulnerable individuals. By explicitly modelingthe variance due to these factors, we will be able to control for theireffect on mother-reported father involvement and child behavioralproblems as well as on the link between these constructs. Giventhe earlier literature reviewed, we hypothesize that higher fatherinvolvement will predict lower child problems 1 year later. Second,we hypothesize that after we control for the mother’s depressivecognitive trait, the magnitude of the association between fatherinvolvement and child problems will be substantially reduced.

METHOD

Sample and Procedures

Data were derived from a longitudinal randomized control trialevaluation of a statewide newborn home-visiting program for first-

time parents ages 20 years and under at childbirth. The programis an adaptation of the Healthy Families America and is designedto (a) prevent child abuse and neglect by supporting positive, ef-fective parenting; (b) promote optimal child health, growth, anddevelopment; (c) encourage parental educational attainment, job,and life skills; (d) prevent repeat teen pregnancies; and (e) promoteparental health and well-being.

Eligibility criteria for study participation included being fe-male, 16 years of age or older, not having previously receivedHealthy Families America services, English or Spanish fluency,and being cognitively able to provide informed consent. Womenseeking Healthy Families Massachusetts services (N = 837) wererandomly assigned to either the home-visiting group or the con-trol group (who received referrals to other service programsfor young parents and monthly child development information).A total of 704 mothers (61% home-visiting group, 39% con-trol group) participated in evaluation activities (response rate =84%), which included, at a minimum, an agency data release oran initial (Time 1, T1) telephone interview. Most participantsalso completed follow-up telephone interviews at 12 months’(Time 2, T2) and at 24 months’ postenrollment (Time 3, T3).

FIGURE 2. A conceptual latent state-trait model depicting 12 manifest variables measuring one trait factor and three state factors. Indicator-specific variance is modeledfor Indicators 2, 3, and 4, keeping Indicator 1 as a reference.



FIGURE 3. A latent model predicting child behavior problems from father involvement (as reported by the child’s mother), after controlling for maternal depressogeniccognitive style.

The number of mothers who completed the telephone interview atT1 was 684; of them, 82% (n = 564) participated at T2, and 87%(n = 594) at T3. Details on the methodology of the larger evaluationstudy are presented elsewhere Jacobs, Easterbrooks, Mistry, Bum-garner, Fauth, Goldberg, Greenstone, Raskin, Contreras, Coskun,Fosse, Kotake, & Scott (in press). Next, we briefly describe mea-sures used in the present study.

Measures

Demographic characteristics (T1). Demographic characteristicsof the mothers, fathers, and children were collected throughmaternal interviews and included mothers’ and fathers’ age andethnic background; child age and sex; fathers’ education, employ-ment, military service, incarceration, and number of children withother partners; and mothers’ current relationship status and levelsof support received from the father.

The Center for Epidemiological Studies-Depression (T1, T2, andT3; CES-D, Radloff, 1977). The 20-item CES-D assesses symp-toms experienced during the past week (e.g., “I felt that I couldnot shake off the blues even with help from my family or friends”)rated on a Likert scale of 0 (not at all) to 3 (a lot). The CES-Dhas demonstrated strong psychometric properties in both clinicaland epidemiological studies (Radloff, 1977) with diverse groups(Naughton & Wiklund, 1993), including adolescents 14 years ofage and older (Sharp & Lipsky, 2002) and postpartum women(Radloff, 1991; Weinberg et al., 2001).

An overall score, reflecting severity of symptoms, typicallyis created by summing the 20 items. However, studies have foundsupport for a four-factor structure of the measure, dividing theitems into the following four subscales (Radloff, 1977; Wong,2000): (a) Depressed Affect (seven items: e.g., “I felt depressed.”“I had crying spells.”); (b) Lack of Well-Being (four items: e.g.,“I felt hopeful about the future;” reverse-coded), “I was happy”



(reverse-coded); (c) Somatic Symptoms (seven items: e.g., “I didnot feel like eating; my appetite was poor.” “My sleep was rest-less.”); and (d) Interpersonal Difficulties (two items: e.g., “Peoplewere unfriendly.” “I felt that people dislike me.”). For each sub-scale, a sum is taken for the included items. Radloff (1977) foundall four subscales to have satisfactory reliability, with the relia-bility of the Depressed Affect subscale being the highest. In oursample, Cronbach’s αs ranged from .66 to .88 and were similar tothose reported by Radloff (1977). We used scores on the subscalesand not the raw items in latent modeling. This approach, calledparceling, has been demonstrated to have important psychomet-ric advantages over item-level data (see Coffman & MacCallum,2005; Little et al., 2002). Parceling has a normalizing tendency, asit yields an indicator that has more scale points that do the originalitems and is thus more continuous in nature (Little et al., 2002).

Father involvement (T1). During the intake interview, mothers re-ported on the following aspects of fathers’ involvement in theirlives: (a) the amount of time he spent with the mother in the pastmonth, rated on a Likert scale ranging from 0 (none) to 5 (daily);(b) mother’s satisfaction with the quality of shared time on a Likertscale ranging from 0 (not at all) to 3 (very); and (c) three di-chotomous (“yes” or “no”) questions about whether he providedfinancial or material support, emotional support, and physical helpwith pregnancy or parenting. Answers to the dichotomous ques-tions were summed to obtain a count of types of supports offeredby the father that ranged from 0 to 3.

These three indicators (presence, quality of shared time, anddegree of support) were rescaled to range from 0 to 15 and usedas manifest variables in the latent model of father involvement.Additional variables also collected at intake (maternal relationshipstatus: single or in a relationship with father or another partner, andwhether she co-resided with the father) were used in the descriptiveanalyses to contextualize the nature of father involvement.

The Brief Infant-Toddler Social and Emotional Assessment (T3)(BITSEA; Briggs-Gowan & Carter, 2006). The BITSEA was usedto assess child socioemotional and behavioral problems or de-lays, and competence at T3. This 44-item questionnaire covers abroad range of socioemotional behaviors in young children (1–3years), including internalizing, externalizing, and regulatory do-mains; and competence (Briggs-Gowan & Carter, 2002). Problembehavior items inquire about behaviors that are both typical of de-velopment (e.g., aggression, sadness, or fear) and those that are notdevelopmentally appropriate (e.g., self-injurious behaviors).

To complete the BITSEA, parents endorse problem and com-petency statements (e.g., “Has trouble falling asleep or stayingasleep”) using a Likert scale of 0 (not true/rarely), 1 (somewhattrue/sometimes), and 2 (very true/often). In two instances, parentsindicate how worried they are about a particular problem usinga Likert scale of 1 (not at all worried), 2 (a little worried), 3(worried), and 4 (very worried).

Four subscales were calculated: Externalizing (six items: e.g.,“Hits, shoves, kicks, or bites children (not including brother or sis-

ter”), Internalizing (eight items: e.g., “Cries or hangs onto youwhen you try to leave”), Dysregulation (eight items: e.g., “Cries orhas a tantrum until he or she is exhausted”), and Competence [11items, e.g., “Can pay attention for a long time (other than watch-ing TV)”]. Higher externalizing, internalizing, and dysregulationscores indicate greater levels of socioemotional or behavioral prob-lems; lower competence scores indicate a possible deficit/delay.Scores are created by summing the item responses from respectivesubscales. In our sample, Cronbach’s αs were: α = .65 for theExternalizing Problems subscale; α = .52 for the InternalizingProblems subscale; α = .58 for the Dysregulation Problems sub-scale; and α = .66 for the Competence subscale.

The BITSEA was standardized on 600 children (Briggs-Gowan & Carter, 2006). Age and sex differences were found,specifically for competence, and authors recommended percentile-based cut scores to create Age-Band × Sex norm groups for thetotal problem and competence scores. No norm groups were pro-vided for the subscales; we controlled for the effects of child ageand sex on the subscales.

Analytic Approach

We began our analyses by exploring mother, child, and fathercharacteristics in our sample. To contextualize these descriptives,we conducted bivariate analyses (χ2 statistic and t tests) comparingmothers who reported daily father presence to all other mothers.These analyses were conducted in Stata 13/SE (Kohler & Kreuter,2012).

We then proceeded to fit a series of latent models to test ourstudy hypotheses. Figure 3 illustrates our conceptual model. Wefirst performed a latent regression analysis to test whether father in-volvement at T1 predicted child behavioral problems at T3 (Model1; illustrated in the top part of Figure 3). A latent regression anal-ysis estimates the association between dependent and independentvariables that are latent. We then fit an LST model of maternaldepression (Model 2; see the bottom part of Figure 3). We chosethe Depressive Affect subscale as the reference indicator, giventhat it was shown to be the most reliable factor. In our final step,we fit the entire model pictured in Figure 3, which tested the latentregression model predicting child behavioral problems from fatherinvolvement, controlling for the effect of maternal depressive trait(Model 3). The analytic steps described earlier follow the logic ofhierarchical regression analysis, in which predictors are entered asblocks and then nested models are compared (specifically, Models1 and 3). This analytic approach reflected the hypotheses that fa-ther involvement will predict child problems, but that the strengthof the relation between these mother-reported measures would be-come less robust once we control for mothers’ depressive bias.Both latent regression analyses controlled for the effects of child’sage and sex.

In the latent framework, results are evaluated in three steps.The first step involves examining the structural components of theproposed models (i.e., the measurement model). Specifically, fac-tor loadings are evaluated to examine whether manifest indicators



supported the proposed unobserved (latent) constructs. The sec-ond step involves evaluating the structural model (i.e., the latentpath model testing the study hypotheses). Finally, model fit statis-tics are explored to identify models with the best fit to the data.We used the following model fit statistics. First, the χ2 statisticwas examined, which tests whether estimated matrices are sta-tistically equal to the observed matrices (A nonsignificant χ2 isdesired). Because this test is sensitive to sample size (Little, 2013)we evaluated several additional indices: the Comparative Fit Index(CFI), the Tucker–Lewis Index (TLI), the root mean square error ofapproximation (RMSEA), and the standardized root mean squareresidual (SRMR). These are commonly interpreted relative to cut-off values for acceptable fit (Geiser, 2013). Lower values of theRMSEA and the SRMR (<.08) and higher values of the CFI andthe TLI (>.90) indicate good fit (Hu & Bentler, 1999; Little, 2013).We conducted all latent analyses in Mplus Version 7.11 (Muthen& Muthen, 1998–2012).

Missing Data

The primary reason for missing data was wave attrition or in-termittent missingness (18% at T2, 13% at T3). Bivariate analy-ses used listwise deletion to handle missingness. Full-informationmaximum likelihood (FIML) estimation was used to account formissing data in latent model analyses. In FIML, a likelihood func-tion for each individual is estimated based on the variables that arepresent (Enders, 2008, 2010).

RESULTS

Descriptive Analyses

As shown in Table 1, ours was a diverse sample of young women,a large proportion of whom (40%) already were parenting at T1.Mothers who reported daily father presence did not differ fromthe rest of the mothers on such characteristics as mother’s placeof birth, membership in the home-visiting group, and depressionat T1; child’s sex, age, and level of internalizing and externalizingproblems; and father’s age, level of education, employment, currentmilitary service, and number of other children. Some differencesbetween the groups were found. Mothers who reported daily fa-ther presence were slightly older, were less likely to be Black orHispanic, and were less likely to report their relationship statusas single or to be in a relationship with another partner. Fatherswho were described as present were more likely to be White orHispanic and less likely to be incarcerated; they were more likelyto coreside with the mother, and to provide financial, emotional,and physical support. Most mothers (61%) who described fathersas present daily were very satisfied with the quality of shared time.

Latent Models

Table 2 summarizes estimates and model fit statistics of the la-tent regression model predicting mother-reported child behavior

problems from father involvement (Model 1), the latent state-traitmodel of depression (Model 2), and the full model, in which childproblems are predicted by father involvement, controlling for theeffect of mothers’ depressive trait (Model 3).

The results shown in Model 1 indicate that the three observedindicators of father involvement capture quite well common vari-ance in the latent factor (factor loadings range = .85–.92). Thethree manifest variables of the latent factor of child behavioralproblems (internalizing, externalizing, and dysregulation) also hadmoderately high loadings (.61, .59, and .83, respectively). The la-tent regression analysis revealed that lower father involvement atT1 significantly predicted more child behavioral problems at T3,p < .05.

Model 2 in Table 1 summarizes the results of the LST anal-ysis of the longitudinal depression data. As shown, indicators hadsatisfactory (moderate to high, .52–.95) loadings onto the statefactors, which in turn loaded highly onto the trait factor, indicat-ing high stability over time in mothers’ depressive symptomology(.72 for depression state factor at T1, .77 for T2, and .59 for T3).Converted into R2, these factor loadings indicate that at each timepoint, 52, 59, and 35%, respectively, of true individual differencesin depression states were explained by a stable trait, and the remain-ing 48, 41, and 65%, respectively, were due to occasion-specificeffects.

Factor loadings for indicator-specific factors ranged from .28to .54, which means that between 8 and 29% of the observedvariability in depression data was due to the indicator-specific ef-fects. These effects are relatively high, indicating that a modelwithout these factors likely would have had correlated residuals,which would lead to estimation problems. The highest loadingswere for the lack of well-being factor (.40, .54, and .52, respec-tively), which is to be expected given that this subscale was com-prised of the only four items of the CES-D that are positivelyworded and must be reverse-coded. Method factors were inter-correlated, suggesting that mothers who endorsed high values onthe somatic subscale items also endorsed high values on the itemsdescribing interpersonal problems and low values on the itemsdescribing lack of well-being; and mothers who had high val-ues on interpersonal problems had low values on lack of well-being.

Model 3 in Table 1 summarizes the results of the conceptualmodel, exploring the effect of father involvement on child prob-lems after controlling for the effect of maternal depressive trait.As shown, after such accounting, the association between fatherinvolvement and child problems was no longer significant. In par-ticular, the point estimate of the association was reduced from aβ = −.11, p < .05, in Model 1 to β = −.07 in Model 3, al-though the SE of the point estimate remained unchanged at .05.There was a strong, statistically significant association betweenmaternal depressive traits and child problem behaviors, β = .35,p < .001. Moreover, mothers with a higher depressive trait alsoreported lower father involvement, β = −.16, p < .01. In otherwords, maternal depressive trait was related to both the predictorand the outcome in the model and, as such, acted as a confound



TABLE 1. Descriptive and Inferential Statistics of Study Variables (N = 704a)

Father Involvement With Mother (Time 1)

Total Yes, daily No, not dailyN = 704 n = 322 n = 330 χ2/t (df)

Mother Characteristics (Time 1) n (%)/M (SD) n (%)/M (SD) n(%)/M (SD)Time1 Intake Mother age (in years) 18.72 (1.32) 18.84 (1.24) 18.61 (1.39) t(650) = −2.18, p < .05Mother Ethnicity χ2(3) = 9.63 (3), p < .05

White (Non-Hispanic) 238 (38.0%) 136 (42.2%) 112 (33.9%)Black (Non-Hispanic) 124 (19.0%) 50 (15.5%) 74 (22.4%)Hispanic 225 (34.5%) 104 (32.3%) 121 (36.7%)Other (Non-Hispanic) 55 (8.4%) 32 (9.9%) 23 (7.0%)

Place of BirthU.S. Mainland 536 (82.2%) 273 (84.8%) 263 (79.7%)Puerto Rico 43 (6.6%) 22 (6.8%) 21 (6.4%)Foreign Born 73 (11.2%) 27 (8.4%) 46 (13.9%)

Pregnant or Parent StatusPregnant 260 (39.9%) 189 (58.7%) 203 (61.5%)

In Home-Visiting Group 396 (60.7%) 197 (61.2%) 199 (60.3%)Child Characteristics (Time 3)

Child Age (Months) 24.58 (6.67) 24.52 (6.55) 24.48 (6.42)Child Sex

Boy 344 (53.3%) 174 (54.5%) 170 (52.1%)Father of Baby Characteristics (Time 1)

Father Age 20.96 (4.12) 21.08 (4.41) 20.85 (3.82)Father Ethnicity χ2(3) = 20.36, p < .001

White (Non-Hispanic) 179 (27.9%) 109 (34.8%) 70 (22.4%)Black (Non-Hispanic) 140 (21.8%) 51 (16.3%) 89 (28.4%)Hispanic 65 (10.1%) 126 (40.3%) 118 (37.7%)Other (Non-Hispanic) 257 (40.1%) 27 (8.6%) 36 (11.5%)

Completed High School/GED 232 (48.5%) 126 (48.5%) 106 (48.6%)Completed 1 Year of College 42 (8.8%) 21 (8.1%) 21 (9.6%)Employed 351 (57.5%) 196 (61.4%) 154 (53.7%)In the Military 16 (2.5%) 6 (1.9%) 10 (3.2%)Incarcerated 21 (3.3%) 0 (0.0%) 21 (6.7%) χ2(1) = 22.28, p < .001No. of Other Children 1.55 (1.11) 1.52 (1.02) 1.56 (1.18)

Mother–Father Relationship (Time 1)Father Lives With Mother 181 (27.8%) 172 (53.6%) 9 (2.7%) χ2(1) = 209.95, p < .001Mother Is Single 174 (26.8%) 13 (4.0%) 161 (48.9%) χ2(1) = 166.99, p < .001Mother Is With Another Partner 43 (6.6%) 5 (1.6%) 38 (11.6%) χ2(1) = 26.26, p < .001How Much Time Father Spends With Mother χ2(2) = 652, p < .001

None 142 (21.8%) 0 (0.0%) 142 (43.0%)< Daily 188 (28.8%) 0 (0.0%) 188 (57.0%)

Daily 322 (49.4%) 322 (100.0%) 0 (0.0%)Mother Satisfied With the Quality of the Time Father Spends With Her χ2 (3) = 194.63, p < .001

Not Spending Time Or Not At All Satisfied 156 (24.8%) 2 (0.7%) 154 (48.0%)Somewhat 62 (9.9%) 33 (10.7%) 29 (9.0%)Pretty Much 140 (22.3%) 85 (27.7%) 55 (17.1%)Very 270 (43.0%) 187 (60.9%) 83 (25.9%)

Father Provides Financial Support 401 (62.0%) 276 (85.7%) 125 (38.5%) χ2(1) = 153.26, p < .001Father Provides Emotional Support 504 (77.8%) 318 (98.8%) 186 (57.1%) χ2(1) = 163.00, p < .001Father Provides Physical Help 462 (72.8%) 307 (98.1%) 155 (48.1%) χ2(1) = 199.75, p < .001

aThe total N in our study was 704. However, due to missingness and wave attrition, the ns in the columns may not add up to 704.

of the association between father involvement and child outcomes.Depression state factor at T1 was uncorrelated with father involve-ment at T1, suggesting that when variance due to the depressiontrait is modeled, occasion-specific fluctuation in depression is not

related to how mothers characterize father involvement. In con-trast, depression state at T3 was strongly correlated with childproblems at T3. Even when variance due to the depression traitis modeled, occasion-specific influences on depression symptom



TABLE 2. Results of the Latent Regression Analysis Predicting Child Behavioral Problems at Time 3 From Father Involvement at Time 1, as Reportedby the Child’s Mother (n = 692)a

Estimate

Parameter Model 1 Model 2 Model 3

Measurement ModelsFather Involvement

Support .92 (.01)∗∗∗ .92 (.02)∗∗∗

Presence .89 (.01)∗∗∗ .89 (.01)∗∗∗

Quality of Shared Time .85 (.02)∗∗∗ .85 (.02)∗∗∗

Child Behavior ProblemsExternalizing .59 (.04)∗∗∗ .63 (.04)∗∗∗

Internalizing .61 (.04)∗∗∗ .63 (.04)∗∗∗

Dysregulation .83 (.04)∗∗∗ .78(.03)∗∗∗

Depression State (Time 1)Depressive Affect .93 (.02)∗∗∗ .93 (.02)∗∗∗

Lack of Well-Being .56 (.03)∗∗∗ .58 (.03)∗∗∗

Somatic Symptoms .76 (.02)∗∗∗ .76 (.02)∗∗∗

Interpersonal Problems .61 (.03)∗∗∗ .62 (.03)∗∗∗









Depressive TraitDepression State (Time 1) .72 (.04)∗∗∗ .73 (.04)∗∗∗

Depression State (Time 2) .77 (.05)∗∗∗ .77 (.05)∗∗∗

Depression State (Time 3) .59 (.04)∗∗∗ .59 (.04)∗∗∗

Method Factor (LW):.40 (.04)∗∗∗ .38 (.05)∗∗∗

Lack of Well-Being (Time 2) .54 (.05)∗∗∗ .54 (.05)∗∗∗

Lack of Well-Being (Time 3) .52 (.05)∗∗∗ .52 (.05)∗∗∗

Method Factor (SS):Somatic Symptoms (Time 1) .28 (.05)∗∗∗ .26 (.05)∗∗∗

Somatic Symptoms (Time 2) .33 (.05)∗∗∗ .32 (.05)∗∗∗

Somatic Symptoms (Time 3) .32 (.05)∗∗∗ .33 (.05)∗∗∗

Method Factor (IP):Interpersonal Problems (Time 1) .35 (.06)∗∗∗ .32 (.06)∗∗∗

Interpersonal Problems (Time 2) .52 (.07)∗∗∗ .53 (.08)∗∗∗

Interpersonal Problems (Time 3) .30 (.05)∗∗∗ .30 (.05)∗∗∗

Path ModelsFather Involvement → Child Problems −.11 (.05)∗ −.07 (.05)Depressive Trait → Child Problems .35 (.06)∗∗∗

Depressive Trait → Father Involvement −.16 (.06)∗∗

CovariancesMethod Factor (SS) With

Method Factor (LW) −.25 (.09)∗∗ −.26 (.09)∗∗

Method Factor (IP) .39 (.10)∗∗∗ .31 (.10)∗∗

Method Factor (IP) WithMethod Factor (LW) −.18 (.08)∗ −.17 (.09)∗

Depression State (Time 1) WithFather Involvement −.07 (.06)

Depression State (Time 3) WithChild Problems .42 (.06)∗∗∗

χ2 48.78∗∗∗ 42.08 277.06∗∗∗

df 17 39 148RMSEA 0.06 0.01 0.04CFI/TLI 0.98/0.97 0.99/0.99 0.97/0.96SRMR 0.03 0.02 0.05

LW = Lack of Well-Being; SS = Somatic Symptoms; IP = Interpersonal Problems; RMSEA = root mean square error of approximation; CFI/TLI = Comparative FitIndex and the Tucker–Lewis Index; SRMR = standardized root mean square residual.aLatent regressions (Models 1 and 3) controlled for the effect of child’s age and sex on child problems and child sex on father involvement.∗p < .05. ∗∗p < .01. ∗∗∗p < .001.



were still highly correlated with how mothers characterize theirchildren. In other words, mothers’ reports on their children areimpacted by both the stable characteristics of the mothers andsituational effects.

DISCUSSION

There is bountiful evidence that healthy child development de-pends not only on mothers but on fathers as well (Cabrera &Tamis-LeMonda, 2013). Changing social norms over the lastseveral decades have led to shifts in both the amount of timethat fathers spend in childrearing and the qualitative aspects offathers’ participation in coparenting (Cabrera et al., 2007; Leidyet al., 2013). The study of father research has evolved considerablyfrom simply focusing on the effects of parental divorce and fatherabsence on children to considering the long-term effects of fatherson the development of their children in large-scale longitudinalstudies such as the Fragile Families and Child Wellbeing Study(McLanahan & Garfinkel, 2012; Reichman, Teitler, Garfinkel, &McLanahan, 2001; Sigle-Rushton & McLanahan, 2004). Such in-creased scientific attention to the role of fathers has been paralleledby broad policy support for programs to promote responsible fa-ther involvement, such as the Supporting Healthy Marriage Project(Dion, 2005) and President Obama’s 2009 call for a national con-versation to promote responsible fatherhood (Braswell, n.d.).

Despite the great gains made to understand the role of fathersin supporting healthy development of the children, the field offathering research is complicated by several methodological chal-lenges. One of the main challenges is gaining access to represen-tative samples. Given that fathers from higher risk backgroundsare more difficult to recruit and retain in surveys, many studysamples may be skewed toward resident fathers. This issue is ofconcern, given the dramatic shifts toward higher proportions ofchildren born outside of marriage (41% of births) in the UnitedStates today (Carlson & Meyer, 2014). Nonmarital childbearingis more common among disadvantaged families, in which fathersare nonresident, often young, and experience economic difficulties(Cancian & Haskins, 2014; Furstenberg, 2014).

What is currently known about the role of low-income fatherspoints to great benefits of father involvement for the developmentof the children, including lower risk for cognitive developmen-tal delays (Tamis-LeMonda, Shannon, Cabrera, & Lamb, 2004)and increased social competence (Kelley, Smith, Green, Berndt, &Rogers, 1998). In addition, it has generally been concluded thatlow-income and young fathers tend to be less involved with theirchildren (Fagan & Lee, 2010; Roopnarine, Fouts, Lamb, & Lewis-Elligan, 2005). Still, there is considerable variability in patternsof involvement among disadvantaged, nonresident fathers. Severalstudies have revealed that young fathers hold responsible views ofparenting as a set of behaviors comprising financial, social, andemotional obligations (Bishop, Gay, Ledesma, & Garofalo, 2005;Lemay, Cashman, Elfenbein, & Felice, 2010). A study of adoles-cent mothers (Easterbrooks, Raskin, & McBrian, 2014) found thatfathers were generally present for their children, but the link be-

tween father involvement and the development of the child wasmore complex than initially thought.

Clearly, parenting experiences of fathers may not be as uni-form as has been suggested by prior research; however, method-ological challenges related to accessing representative samplesmake it harder for researchers to shed light on such complexities.One major obstacle is recruitment of fathers into studies and theneed to rely on other informants (e.g., mothers). Although stud-ies in which fathers are participants do exist (e.g., Cabrera et al.,2004; Summers et al., 1999), it is more common to assess low-income, nonresident fathers using maternal report. However, a ma-jor concern with this analytic strategy is that such reports maybe distorted by perceptions and attributions held by the mother.Specifically, paternal-involvement literature that relies on mater-nal reports of father involvement suffers from a lack of accountingfor the possible confounding effect of some unobserved character-istics of the mothers’ reporting on the fathers. Such unobservedtraits may inflate the real associations between father involvementand children’s development.

Our study aimed to address this concern. We investigated thelink between mother-reported father involvement and child devel-opment in a sample of young, low-income, primarily unmarriedfamilies. Using three waves of maternal data from a large sampleof low-income, at-risk young mothers, we evaluated the degreeto which maternal-report bias related to depression might affectthe link between mother-reported measures of father involvementand child developmental problems. Our study used indicators ofmaternal depression measured over time to model a stable mater-nal depressive trait, independent of occasion-specific symptoms.Results show that after our model accounted for the effect of thedepressive trait on father involvement and child development, theeffects of father on child development were reduced by nearlyhalf and to statistical nonsignificance, even as the SEs remain un-changed. In other words, the results of our study suggest that afterwe draw out the variance in the measures of father involvement andchild behavioral problems due to the effect of a maternal depressivetrait, we no longer see an association between these constructs.

Several themes emerged from the analyses and are discussednext, including (a) demographic characteristics associated withmother’s tendency to describe the father as more involved, (b)the role of maternal depression as confounds in studies whichrely on mother-reported father involvement, (c) clarifications aboutthe contributions of fathers to the development of the child, and(d) additional methodological considerations about the ways ofmeasuring father involvement.

Correlates of Higher Mother-Reported Father Involvement

Our investigation began with descriptive analyses of the charac-teristics of the fathers whom mothers described as involved on adaily basis. Not surprisingly, relationship status and satisfactionwith the levels of father support were associated with how motherstended to report the frequency of father involvement. Specifically,mothers who described fathers as involved on a daily basis were



more likely to coreside with the fathers, not be single, and reportgreater levels of father support and higher satisfaction regardingthe quality of the time spent with him. These results suggest thatmothers’ reports of father involvement are not independent fromher perceptions about the quality of the relationship and, as such,offer support to our call to consider the issue of confounding inmaternal reports of paternal involvement.

Depressive Trait as a Study Confound

In our study, we used a measure of a stable depressive trait asa proxy for reporter bias, although there are many other stablecharacteristics that also should be explored in future studies. In ourcase, we were guided by the theory that mothers who have a depres-sive trait might have a tendency to underreport father involvement,but overreport the level of child problems. This nondependencebetween the measures of father involvement and child outcomewould then inflate the degree of the association between them in away that lower involvement would be a stronger predictor of de-velopmental problems than it is in reality. The results of our studyconfirmed this hypothesis. In fact, after controlling for the effectof maternal depressive trait, we found that the link between fatherinvolvement and child outcome disappeared.

The literature on the high prevalence of depression amongpostpartum women supports our concern about relying on mater-nal reports. Depression following childbirth affects 13% of women,with estimates ranging from 3 to 25% (Surkan, Gottlieb, Mc-Cormick, Hunt, & Peterson, 2012). Women in high social risksamples are especially vulnerable, with rates of depression as highas 26% in women in poor, urban communities (Moses-Kolko &Roth, 2004). Young, low-income mothers are at even higher riskfor depression than are older mothers, with rates ranging from 30to 60% (Beers & Hollo, 2009; J. Brown, Harris, Woods, Buman,& Cox, 2012; Mollborn & Morningstar, 2009; Tzilos, Zlotnick,Raker, Kuo, & Phipps, 2012). These data suggest that researchrelying on maternal reports of both children and father in the earlyyears of parenting should control for the potential effect that de-pression might have on mothers’ responses.

Role of Fathers in Child Development

Research on fathers has documented that they play a significant rolein children’s development. For example, father–child interactionhas been shown to play a critical role in young children’s socialrelationships with peers, through mechanisms of self- and emotionregulation (Leidy et al., 2013). Fathers’ influence on functioning inthe cognitive domain also has substantial documentation, as earlyas the child’s infancy (Pancsofar & Vernon-Feagons, 2010).

At the same time, recent attempts to systematically exam-ine the strength of associations between father involvement andchildren’s development have found the effects to be inconsistentand small, depending on the way that father involvement wasoperationalized and on the domains of child behaviors assessed(Adamsons & Johnson, 2013; McWayne et al., 2013). For exam-ple, McWayne et al.’s (2013) meta-analysis of 21 studies pub-

lished in 1998 to 2008 exploring the effect of fathers on children’searly learning revealed small to moderate associations (mean ef-fect sizes = .08–.30). In a meta-analysis of 52 studies investigatingnonresident fathers, Adamsons and Johnson (2013) found strongesteffects of father involvement on social functioning (mean effectsize = .15), with much smaller effects on behavioral, academic,and cognitive domains (effect sizes of .05, .04, and .03, respec-tively).

Note that in our first latent regression model, in which we didnot control for the effect of maternal depressive trait, the effectsize of father involvement on child outcome was consistent withthe effect sizes found in other studies (Adamsons & Johnson, 2013;Amato & Gilbreth, 1999). Given how many studies in Adamsonsand Johnson’s (2013) meta-analysis used maternal reports of fatherinvolvement, the fact that our first model found a similar effect sizeraises a concern about the issue of unaccounted reporter bias inprevious studies.

Although we expected the magnitude of the association be-tween father involvement and child problems to be substantiallyreduced after controlling for the mother’s depressive trait, we weresurprised to see the link disappear completely. Both our findingsand the recent meta-analyses might suggest that the contributionof father involvement to the development of the child is trivial.However, we caution readers from making this conclusion. Wesuspect that studies might be failing to detect a stronger effect offather involvement due to concerns with the validity and reliabilityof current measures of father involvement.

Measurement of Father Involvement

Various conceptualizations of father involvement have been usedin the field, such as presence and hours of direct contact, directchildcare tasks, quality of physical care, nature of dyadic play andstimulation, quality of father–child interactions, and involvementin child-related activities (Bornstein, 2002; Brooks-Gunn & Mark-man, 2005; Lamb, Pleck, Charnov, & Levine, 1985; Pleck, 2010).Our measure of father involvement was based on indicators of theamount of time and extent of support that the fathers offered to themother and her satisfaction with it. While this measure assessessome important dimensions of paternal involvement, a more elabo-rated assessment of father involvement may provide the sensitivityto detect a stronger contribution. Studies employing a multidimen-sional view of father involvement, examining time spent with thechild as well as other avenues of support (e.g., quality of inter-actions, taking responsibility for transportation and appointments,etc.) have provided a more nuanced understanding of the circum-stances and impact of father involvement.

Further, measures of father involvement which are based ondirect contact are not sensitive to the fact that fathers’ access tothe child might be greatly limited by their employment, incarcer-ation, or military service, or when mother–father conflict is high,since mothers often serve as the “gatekeepers” and “gateopen-ers” in providing access and engagement between fathers and theirchildren (Allen & Hawkins, 1999; Parke, 2002; Roy, Buckmiller,



& McDowell, 2008; Swisher & Waller, 2008). A father who is notengaged in direct care may still be involved in other ways (e.g., bysupporting the mother financially or emotionally) that have mea-surable impacts on the child (Jia & Schoppe-Sullivan, 2011; Leidyet al., 2013; McHale, Fivaz-Depeursinge, Dickstein, Robertson, &Daley, 2008). Therefore, it is possible that studies that have in-cluded both direct and indirect assessments of father involvementwould find the larger total father-involvement effects.

Limitations

The strength of our study includes the large size and ethnic di-versity of the sample. However, our analysis captures only theexperiences of socially disadvantaged teenaged mothers in theNortheastern United States. The fact that our findings may notgeneralize to other social groups or contexts is a limitation. Addi-tional research should examine the variability in the effect of proxyreports that may depend on characteristics of participants and/ortheir contexts (Hewlett, 1992; Lamb, 1987). A different method-ological limitation of our study concerns our measure of maternaldepression. Although it is widely used in the mental health field asa screening tool and is associated with a clinical diagnosis (Radloff& Locke, 1986), high symptom scores on the CES-D are not nec-essarily indicative of a diagnosis of depression. However, it maybe more generalizable to nonclinical populations because it cap-tures the more mild forms of dysthymia that are common amongpostpartum mothers.

REFERENCES

Abela, J.R., & Seligman, M.E. (2000). The hopelessness theory of depres-sion: A test of the diathesis-stress component in the interpersonaland achievement domains. Cognitive Therapy and Research, 24(4),361–378.

Achenbach, T.M., Edelbrock, C., & Howell, C.T. (1987). Empiricallybased assessment of the behavioral/emotional problems of 2- and3-year-old children. Journal of Abnormal Child Psychology, 15(4),629–650.

Adamsons, K., & Johnson, S.K. (2013). An updated and expanded meta-analysis of nonresident fathering and child well-being. Journal ofFamily Psychology, 27(4), 589–599.

Allen, S.M., & Hawkins, A.J. (1999). Maternal gatekeeping: Mothers’beliefs and behaviors that inhibit greater father involvement in familywork. Journal of Marriage and the Family, 61, 199–212.

Amato, P.R., & Gilbreth, J.G. (1999). Nonresident fathers and children’swell-being: A meta-analysis. Journal of Marriage and the Family, 61,557–573.

Beers, L.A.S., & Hollo, R.E. (2009). Approaching the adolescent-headedfamily: A review of teen parenting. Current Problems in Pediatricand Adolescent Health Care, 39(9), 216–233.

Bishop, V., Gay, F., Ledesma, J., & Garofalo, R. (2005). Exploring theadolescent and young adult minority fatherhood experience. Journalof Adolescent Health, 36(2), 119–120.

Bornstein, M.H. (2002). Handbook of parenting (2nd ed.). Mahwah, NJ:Erlbaum.

Braswell, K. (n.d.). Responsible Fatherhood Toolkit: Re-sources from the Field. Retrieved January 29, 2014, fromhttps://www.fatherhood.gov/toolkit/home

Briggs-Gowan, M.J., & Carter, A.S. (2002). Brief Infant-Toddler Socialand Emotional Assessment (BITSEA): Manual (Version 2.0). NewHaven, CT: Yale University.

Briggs-Gowan, M.J., & Carter, A.S. (2006). Brief Infant-Toddler Socialand Emotional Assessment (BITSEA). San Antonio, TX: PsychCorp.

Brooks-Gunn, J., & Markman, L. (2005). The contribution of parenting toethnic and racial gaps in school readiness. The Future of Children,15(1), 139–168.

Brown, T.A. (2006). Confirmatory factor analysis for applied research.New York: Guilford Press.

Brown, J., Harris, S., Woods, E., Buman, M., & Cox, J. (2012). Longitu-dinal study of depressive symptoms and social support in adolescentmothers. Maternal and Child Health Journal, 16(4), 894–901.

Cabrera, N., Mitchell, S., Ryan, R., Shannon, J., & Tamis-LeMonda,C. (2008). Low-income, nonresident father involvement with theirtoddlers: Variation by father’s race and ethnicity. Journal of FamilyPsychology, 22(3), 643–647.

Cabrera, N., Shannon, J., & Tamis-LeMonda, C. (2007). Fathers’ in-fluence on their children’s cognitive and emotional development:From toddlers to pre-K. Applied Developmental Science, 11(4),208–213.

Cabrera, N.J., Shannon, J.D., Vogel, C., Tamis-LeMonda, C., Ryan, R.M.,Brooks-Gunn, J. et al. (2004). Low-income fathers’ involvement intheir toddler’s lives: Biological fathers from the Early Head StartResearch and Evaluation Study. Fathering: A Journal of Theory,Research, & Practice about Men as Fathers, 2, 5–26.

Cabrera, N., & Tamis-LeMonda, C.S. (Co-Eds.). (2013). Handbook offather involvement: Multidisciplinary perspectives (2nd ed.). NewYork: Taylor & Francis.

Campbell, D.T., & Fiske, D.W. (1959). Convergent and discriminant vali-dation by the multitrait-multimethod matrix. Psychological Bulletin,56(2), 81–105.

Cancian, M., & Haskins, R. (2014). Changes in family composition: Impli-cations for income, poverty and public policy. Annals of the AmericanAcademy of Political and Social Science, 654, 31–47.

Carlson, M.J., & McLanahan, S.S. (2010). Fathers in fragile families. InM.E. Lamb (Ed.), The role of the father in child development (5thed., pp. 241–269). Hoboken, NJ: Wiley.

Carlson, M.J., & Meyer, D.R. (2014). Family complexity: Setting thecontext. Annals of the American Academy of Political and SocialScience, 654, 6–11.

Carmines, E.G., & Zeller, R.A. (1979). Reliability and validity assessment.In Quantitative applications in the social sciences (Vol. 17). ThousandOaks, CA: Sage.

Coffman, D.L., & MacCallum, R.C. (2005). Using parcels to convert pathanalysis models into latent variable models. Multivariate BehavioralResearch, 40(2), 235–259.

Cole, D.A., Ciesla, J.A., & Steiger, J.H. (2007). The insidious effects offailing to include design-driven correlated residuals in latent-variablecovariance structure analysis. Psychological Methods, 12(4), 381–398.


https://www.fatherhood.gov/toolkit/home


Coley, R.L., & Morris, J.E. (2002). Comparing father and mother reportsof father involvement among low-income minority families. Journalof Marriage and Family, 64(4), 982–997.

Costigan, C.L., & Cox, M.J. (2001). Fathers’ participation in family re-search: Is there a self-selection bias? Journal of Family Psychology,15(4), 706–720.

Dalgleish, T., Hill, E., Golden, A.-M.J., Morant, N., & Dunn, B.D. (2011).The structure of past and future lives in depression. Journal of Ab-normal Psychology, 120(1), 1.

De Los Reyes, A., & Kazdin, A.E. (2005). Informant discrepancies in theassessment of childhood psychopathology: A critical review, theo-retical framework, and recommendations for further study. Psycho-logical Bulletin, 131(4), 483–509.

DeVellis, R.F. (2011). Scale development: Theory and applications (3rded.). Thousand Oaks, CA: Sage.

Dion, M.R. (2005). Healthy marriage programs: Learning what works.The Future of Children, 15(2), 139–156.

Downer, J., Campos, R., McWayne, C., & Gartner, T. (2008). Fatherinvolvement and children’s early learning: A critical review of pub-lished empirical work from the past 15 years. Marriage & FamilyReview, 43(1-2), 67–108.

Easterbrooks, M. A., Raskin, M., & McBrian, S.F. (2014). Father involve-ment and toddlers’ behavior regulation: Evidence from a high socialrisk sample. Fathering, 12(1), 71–93.

Ehring, T., Tuschen-Caffier, B., Schnulle, J., Fischer, S., & Gross, J.J.(2010). Emotion regulation and vulnerability to depression: Sponta-neous versus instructed use of emotion suppression and reappraisal.Emotion, 10(4), 563–572.

Eid, M., & Diener, E. (2006). Handbook of multimethod measurement inpsychology. Washington, DC: American Psychological Association.

Eid, M., Geiser, C., & Nussbeck, F.W. (2009). Multitrait-multimethodanalysis in psychotherapy research: New methodological approaches.Psychotherapy Research, 19(4-5), 390–396.

Enders, C.K. (2008). A note on the use of missing auxiliary variablesin full information maximum likelihood-based structural equationmodels. Structural Equation Modeling, 15(3), 434–448.

Enders, C.K. (2010). Applied missing data analysis. New York: GuilfordPress.

Fagan, J., & Lee, Y. (2010). Perceptions and satisfaction with father in-volvement and adolescent mothers’ postpartum depressive symp-toms. Journal of Youth and Adolescence, 39, 1109–1121.

Fitzgerald, H.E., & Bockneck, E.L. (2013). Fathers, children, and therisk-resilience continuum. In N. J. Cabrera & C. S. Tamis-LaMonda(Eds.), Handbook of father involvement: Multidisciplinary perspec-tives (2nd ed., pp 168–185). New York: Routledge.

Furstenberg, F.F. (2014). Fifty years of family change: From consensus tocomplexity. Annals of the American Academy of Political and SocialScience, 654, 12–30.

Geiser, C. (2013). Data analysis with Mplus. New York: Guilford Press.

Geiser, C., Eid, M., & Nussbeck, F.W. (2008). On the meaning of the latentvariables in the CT-C(M-1) model: A comment on Maydeu-Olivares& Coffman (2006). Psychological Methods, 13, 49–57.

Geiser, C., & Lockhart, G. (2012). A comparison of four approaches to ac-count for method effects in latent state–trait analyses. PsychologicalMethods, 17(2), 255.

Gotlib, I.H., & Joormann, J. (2010). Cognition and depression: Currentstatus and future directions. Annual Review of Clinical Psychology,6, 285.

Grigorenko, E.L., Geiser, C., Slobodskaya, H.R., & Francis, D.J. (2010).Cross-informant symptoms from CBCL, TRF, and YSR: Trait andmethod variance in a normative sample of Russian youths. Psycho-logical Assessment, 22(4), 893–911.

Hankin, B.L., Fraley, R.C., & Abela, J.R. (2005). Daily depression andcognitions about stress: Evidence for a traitlike depressogenic cog-nitive style and the prediction of depressive symptoms in a prospec-tive daily diary study. Journal of Personality and Social Psychology,88(4), 673.

Hernandez, D.J., & Brandon, P.D. (2002). Who are the fathers of today?In C.S. Tamis-LeMonda & N. Cabera (Eds.), Handbook of fatherinvolvement: Multidisciplinary perspectives (pp. 33–62). Mahwah,NJ: Lawrence Erlbaum.

Hernandez, D.C., & Coley, R.L. (2007). Measuring father involvementwithin low-income families: Who is a reliable and valid reporter?Parenting: Science and Practice, 7(1), 69–97.

Hewlett, B.S. (1992). Father-child relations: Cultural and biosocial con-texts. New York: de Gruyter.

Hofferth, S.L., & Goldscheider, F. (2010). Family structure and the tran-sition to early parenthood. Demography, 47(2), 415–437.

Howard, K.S., Lefever, J.E.B., Borkowski, J.G., & Whitman, T.L. (2006).Fathers’ influence in the lives of children with adolescent mothers.Journal of Family Psychology, 20(3), 468.

Hu, L., & Bentler, P.M. (1999). Cutoff criteria for fit indexes in covari-ance structure analysis: Conventional criteria versus new alternatives.Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55.

Huberty, T.J., Austin, J.K., Harezlak, J., Dunn, D.W., & Ambrosius, W.T.(2000). Informant agreement in behavior ratings for children withepilepsy. Epilepsy & Behavior, 1(6), 427–435.

Jacobs, Easterbrooks, Mistry, Bumgarner, Fauth, Goldberg, Greenstone,Raskin, Contreras, Coskun, Fosse, Kotake, & Scott (in press). TheMassachusetts Healthy Families Evaluation-2 (MHFE-2): A Ran-domized, Controlled Trial of a Statewide Home Visiting Programfor Young Parents. Final Technical Report to the Children’s Trustof Massachusetts. Medford, MA: Tufts Interdisciplinary EvaluationResearch.

Jensen, P.S., Xenakis, S.N., Davis, H., & Degroot, J. (1988). Child psy-chopathology rating scales and interrater agreement: II. Child andfamily characteristics. Journal of the American Academy of Child &Adolescent Psychiatry, 27(4), 451–461.

Jia, R., & Schoppe-Sullivan, S.J. (2011). Relations between coparent-ing and father involvement in families with preschool-age children.Developmental Psychology, 47(1), 106.

Kalil, A., Ziol-Guest, K.M., & Coley, R.L. (2005). Perceptions of fatherinvolvement patterns in teenage-mother families: Predictors and links



to mothers’ psychological adjustment. Family Relations, 54(2), 197–211.

Kelley, M.L., Smith, T.S., Green, A.P., Berndt, A.E., & Rogers, M.C.(1998). Importance of fathers’ parenting to African-Americantoddler’s social and cognitive development. Infant Behavior & De-velopment, 21(4), 733–744.

Kitterød, R.H., & Lyngstad, J. (2014). She said, he said: Comparing moth-ers’ and fathers’ reports on the non-resident father’s contact with hischildren. Demographic Research, 30, 899–910.

Kline, T. (2005). Psychological testing: A practical approach to designand evaluation. Thousand Oaks, CA: Sage.

Kohler, U., & Kreuter, F. (2012). Data analysis using Stata. College Sta-tion, TX: StataCorp LP.

LaGrange, B., & Cole, D.A. (2008). An expansion of the trait-state-occasion model: Accounting for shared method variance.Structural Equation Modeling: A Multidisciplinary Journal, 15(2),241–271.

Lamb, M.E. (1987). The father’s role: Cross-cultural perspectives. Hills-dale, NJ: Erlbaum.

Lamb, M.E., & Lewis, C. (2010). The development and significance offather–child relationships in two-parent families. In M.E. Lamb (Ed.),The role of the father in child development (5th ed., pp. 94–153).Hoboken, NJ: Wiley.

Lamb, M.E., Pleck, J.H., Charnov, E., & Levine, J.A. (1985). Paternalbehaviour in humans. American Psychologist, 25, 883–894.

Leidy, M., Schofield, T., & Parke, R. (2013). Fathers’ contributions to chil-dren’s social development. In N. J. Cabrera & C. S. Tamis-Lemonda(Eds.), Handbook of father involvement: Multidisciplinary perspec-tives (Vol. 2, pp. 151–167). New York, NY: Routledge.

Lemay, C.A., Cashman, S.B., Elfenbein, D.S., & Felice, M.E. (2010). Aqualitative study of the meaning of fatherhood among young urbanfathers. Public Health Nursing, 27(3), 221–231.

Little, T.D. (2013). Longitudinal structural equation modeling. New York:Guilford Press.

Little, T.D., Cunningham, W.A., Shahar, G., & Widaman, K.F. (2002). Toparcel or not to parcel: Exploring the question, weighing the merits.Structural Equation Modeling, 9(2), 151–173.

Luoma, I., Koivisto, A.-M., & Tamminen, T. (2004). Fathers’ and mothers’perceptions of their child and maternal depressive symptoms. NordicJournal of Psychiatry, 58(3), 205–211.

Malik, N.M., Boris, N.W., Heller, S.S., Harden, B.J., Squires, J., Chazan-Cohen, R. et al. (2007). Risk for maternal depression and child ag-gression in Early Head Start families: A test of ecological models.Infant Mental Health Journal, 28(2), 171–191.

Martin, A., Brazil, A., & Brooks-Gunn, J. (2013). The socioemotional out-comes of young children of teenage mothers by paternal coresidence.Journal of Family Issues, 34(9), 1217–1237.

Martin, A., Ryan, R.M., & Brooks-Gunn, J. (2010). When fathers’ support-iveness matters most: Maternal and paternal parenting and children’sschool readiness. Journal of Family Psychology, 24(2), 145.

McHale, J., Fivaz-Depeursinge, E., Dickstein, S., Robertson, J., & Daley,M. (2008). New evidence for the social embeddedness of infants’early triangular capacities. Family Process, 47(4), 445–463.

McLanahan, S., & Garfinkel, I. (2012). Fragile families. In M. Garrison &E. S. Scott (Eds.), Marriage at the Crossroads: Law, Policy, and theBrave New World of Twenty-First-Century Families (pp. 142–169).

McWayne, C., Downer, J.T., Campos, R., & Harris, R.D. (2013). Father in-volvement during early childhood and its association with children’searly learning: A meta-analysis. Early Education & Development,24(6), 898–922.

Mezulis, A.H., Hyde, J.S., & Clark, R. (2004). Father involvement mod-erates the effect of maternal depression during a child’s infancy onchild behavior problems in kindergarten. Journal of Family Psychol-ogy, 18(4), 575–588.

Mitchell, S.J., See, H.M., Tarkow, A.K., Cabrera, N., McFadden, K.E., &Shannon, J.D. (2007). Conducting studies with fathers: Challengesand opportunities. Applied Developmental Science, 11(4), 239–244.

Mollborn, S., & Morningstar, E. (2009). Investigating the relationshipbetween teenage childbearing and psychological distress using lon-gitudinal evidence. Journal of Health and Social Behavior, 50(3),310–326.

Moses-Kolko, E.L., & Roth, E.K. (2004). Antepartum and postpartumdepression: Healthy mom, healthy baby. Journal of the AmericanMedical Women’s Association, 59(3), 181–191.

Mowbray, C.T., Lewandowski, L., Bybee, D., & Oyserman, D. (2005).Relationship between maternal clinical factors and mother- re-ported child problems. Community Mental Health Journal, 41(6),687–704.

Muthen, L.K., & Muthen, B.O. (1998–2012). Mplus: The comprehen-sive modeling program for applied researchers: User’s guide. LosAngeles: Author.

Naughton, M.J., & Wiklund, I. (1993). A critical review of dimension-specific measures of health-related quality of life in cross-culturalresearch. Quality of Life Research, 2(6), 397–432.

Ordway, M.R. (2011). Depressed mothers as informants on child behavior:Methodological issues. Research in Nursing and Health, 34(6), 520–532.

Pancsofar, N., & Vernon-Feagans, L. (2010). Fathers’ early contributionsto children’s language development in families from low-incomerural communities. Early Childhood Research Quarterly, 25(4), 450–463.

Parke, R.D. (2002). Fathers and families. Handbook of Parenting, 3, 27–73.

Peterson, C., Semmel, A., von Baeyer, C., Abramson, L.Y., Metalsky,G.I., & Seligman, M.E. (1982). The attributional style questionnaire.Cognitive Therapy and Research, 6(3), 287–299.

Pleck, J.H. (2010). Paternal involvement: Revised conceptualization andtheoretical linkages with child outcomes. In M.E. Lamb (Ed.), Therole of the father in child development (5th ed., pp. 67–107). NewYork: Wiley.

Pohl, S., & Steyer, R. (2010). Modeling common traits and method ef-fects in multitrait-multimethod analysis. Multivariate Behavioral Re-search, 45(1), 45–72.

Radloff, L.S. (1977). The CES-D Scale: A self-report depression scale forresearch in the general population. Applied Psychological Measure-ment, 1(3), 385–401.



Radloff, L.S. (1991). The use of the Center for Epidemiologic StudiesDepression Scale in adolescents and young adults. Journal of Youthand Adolescence, 20(2), 149–166.

Radloff, L.S., & Locke, B.Z. (1986). The Community Mental HealthAssessment Survey and the CESD scale. In M. M. Weissman, J.D. Myers & C. E. Ross (Eds.), Community surveys of psychiatricdisorders (pp. 177–189). New Brunswick, NJ: Rutgers UniversityPress.

Reichman, N.E., Teitler, J.O., Garfinkel, I., & McLanahan, S.S. (2001).Fragile families: Sample and design. Children and Youth ServicesReview, 23(4), 303–326.

Roggman, L.A., Bradley, R.H., & Raikes, H.H. (2013). Fathers in familycontexts. In N. C. Cabrera. & C. Tamis-LeMonda (Eds.), Handbookof Father Involvement: Multidisciplinary Perspectives (pp. 186–201).New York: Taylor & Francis. .

Roopnarine, J.L., Fouts, H.N., Lamb, M.E., & Lewis-Elligan, T.Y. (2005).Mothers’ and fathers’ behaviors toward their 3-to 4-month-old in-fants in lower, middle, and upper socioeconomic African Americanfamilies. Developmental Psychology, 41(5), 723.

Roy, K.M., Buckmiller, N., & McDowell, A. (2008). Together but not“together:” Trajectories of relationship suspension for low-incomeunmarried parents. Family Relations, 57(2), 198–210.

Seiffge-Krenke, I., & Kollmar, F. (1998). Discrepancies between moth-ers’ and fathers’ perceptions of sons’ and daughters’ problem be-haviour: A longitudinal analysis of parent-adolescent agreement oninternalising and externalising problem behaviour. Journal of ChildPsychology and Psychiatry, 39, 687–697.

Sharp, L., & Lipsky, M. (2002). Screening for depression across the lifes-pan: A review of measures for use in primary care settings. AmericanFamily Physician, 66(6), 1001–1009.

Sigle-Rushton, W., & McLanahan, S. (2004). Father absence and childwellbeing: a critical review. In P.D. Moynihan, L. Rainwater, & T.Smeeding (Eds.), The Future of the Family (pp. 116–155). New York,NY: Russell Sage Foundation.

Steyer, R., Ferring, D., & Schmitt, M.J. (1992). States and traits in psycho-logical assessment. European Journal of Psychological Assessment,8, 79–98.

Steyer, R., Majcen, A.-M., Schwenkmezger, P., & Buchner, A. (1989).A latent state-trait anxiety model and its application to determineconsistency and specificity coefficients. Anxiety Research, 1(4), 281–299.

Steyer, R., & Schmitt, M.J. (1990). Latent state-trait models in attituderesearch. Quality and Quantity, 24(4), 427–445.

Steyer, R., Schmitt, M., & Eid, M. (1999). Latent state-trait theory andresearch in personality and individual differences. European Journalof Personality, 13, 389–408.

Summers, J.A., Raikes, H., Butler, J., Spicer, P., Pan, B., Shaw, S. et al.(1999). Low-income fathers’ and mothers’ perceptions of the father

role: A qualitative study in four Early Head Start communities. InfantMental Health Journal, 20(3), 291–304.

Surkan, P.J., Gottlieb, B.R., McCormick, M.C., Hunt, A., & Peterson, K.E.(2012). Impact of a health promotion intervention on maternal de-pressive symptoms at 15 months postpartum. Maternal Child HealthJournal, 16(1), 139–148.

Sweeney, P.D., Anderson, K., & Bailey, S. (1986). Attributional style indepression: A meta-analytic review. Journal of Personality and SocialPsychology, 50(5), 974.

Swisher, R.R., & Waller, M.R. (2008). Confining fatherhood: Incarcer-ation and paternal involvement among nonresident White, AfricanAmerican, and Latino Fathers. Journal of Family Issues, 29(8), 1067–1088.

Tamis-LeMonda, C.S., Shannon, J.D., Cabrera, N.J., & Lamb, M.E.(2004). Fathers and mothers at play with their 2- and 3-year-olds:Contributions to language and cognitive development. Child Devel-opment, 75(6), 1806–1820.

Thompson, C., MacLaren, J.E., Harris, A., & Kain, Z. (2009). Brief report:Prediction of children’s preoperative anxiety by mothers and fathers.Journal of Pediatric Psychology, 34(7), 716–721.

Thurber, S., & Sheehan, W.P. (2012). Note on truncated T scores in dis-crepancy studies with the Child Behavior Checklist and Youth SelfReport. Archives of Assessment Psychology, 2(1), 73–80.

Treutler, C.M., & Epkins, C.C. (2003). Are discrepancies among child,mother, and father reports on children’s behavior related to parents’psychological symptoms and aspects of parent–child relationships?Journal of Abnormal Child Psychology, 31(1), 13–27.

Tzilos, G., Zlotnick, C., Raker, C., Kuo, C., & Phipps, M. (2012)Psychosocial factors associated with depression severity in preg-nant adolescents. Archives of Women’s Mental Health, 15,397–401.

van der Toorn, S.L., Huizink, A.C., Utens, E.M., Verhulst, F.C., Ormel, J.,& Ferdinand, R.F. (2010). Maternal depressive symptoms, and notanxiety symptoms, are associated with positive mother–child report-ing discrepancies of internalizing problems in children: A report onthe TRAILS Study. European Child & Adolescent Psychiatry, 19(4),379–388.

Weinberg, M.K., Tronick, E.Z., Beeghly, M., Olson, K.L., Kernan, H., &Riley, J.M. (2001). Subsyndromal depressive symptoms and majordepression in postpartum women. American Journal of Orthopsychi-atry, 71(1), 87–97.

Wong, Y.-L.I. (2000). Measurement properties of the Center for Epidemi-ologic Studies-Depression Scale in a homeless population. Psycho-logical Assessment, 12(1), 69.

Youngstrom, E., Loeber, R., & Stouthamer-Loeber, M. (2000). Patternsand correlates of agreement between parent, teacher, and male ado-lescent ratings of externalizing and internalizing problems. Journalof Consulting and Clinical Psychology, 68(6), 1038.


INFLUENCE OF MOTHER'S DEPRESSION ON HER...

Documents

Transcript of INFLUENCE OF MOTHER'S DEPRESSION ON HER...