The Role of Gatekeeping in Non-Resident Fathers’ Contact with Their Children: Mothers’ and Fathers’ Views

We present two studies addressing maternal gatekeeping in separated families and investigate its association with interparental relationship problems and with father-child contact. The first study included 187 custodial mothers and 160 non-resident or shared parenting fathers. The findings suggest that there are significant links between both parents having a negative view of the other and the father having infrequent contact with his children. Coparenting conflict was found to be associated with both parents being worried about the well-being of their children, but only the fathers’ reports of conflict and child-related worries were shown to be linked to low levels of contact, which suggests that paternal withdrawal rather than protective maternal gatekeeping affects levels of father-child contact. The second study (pairfam) used longitudinal data from 145 mothers on the father’s child support payment history, levels of coparenting conflict and cooperation, and the frequency of father-child contact; and from their children on levels of maternal restrictive gatekeeping. Children’s reports of being pressured to side with their mother were shown to predict less frequent contact with their father one year later, even when controlling for earlier contact. Unexpectedly, fathers who did not provide child support were found to have high levels of contact. Overall, we conclude that the role of gatekeeping seems to be limited.


Introduction
In recent decades, normative expectations regarding fathers' involvement in parenting have changed substantially. Egalitarian gender role attitudes have become more widespread in many countries, endorsing not only women participating in the labor market, but also fathers actively contributing to child rearing (Knight and Brinton 2017;Scarborough et al. 2018). Even though family practices often lag behind these expectations, they have changed. Evidence from the U.S. and Europe suggests that fathers are investing more time in child care, not only in nuclear families, but also after parental separation (Amato et al. 2009;Westphal et al. 2014). Nevertheless, parental separation and divorce still put father-child relationships at substantial risk. In the majority of families with separated parents, the mother is the residential parent who takes primary responsibility for the children's everyday lives. Although joint physical custody or shared parenting has become a more common arrangement among separated parents, most separated fathers are relegated to the role of visiting parent, and thus often have only limited contact with their children.
This paper addresses the issue of non-resident fathers' contact with their children, and seeks to explore several hypotheses that may explain why some fathers manage to maintain frequent contact, while others rarely see their children. The factors affecting separated fathers' involvement have been widely debated, not only because access to fathers' economic, social, and emotional resources is considered important for children's well-being (Bastaits et al. 2012;Coley and Medeiros 2007;Jeynes 2015;King and Sobolewski 2006), but also because issues of visitation are often raised in legal conflicts between separated parents. This is the case in Germany, where our research is conducted. Even though there is considerable instability in couple relationships in Germany, with every third marriage ending in divorce (Statistisches Bundesamt 2018b), the available evidence on separated families is quite limited. Furthermore, as in many other countries, divorce rates in Germany no longer capture levels of family instability, as the rising share of nonmarital births has contributed to increased heterogeneity among separated parents. In 2017, every third childbirth (35 percent) in Germany was to unmarried parents (Statistisches Bundesamt 2018c). While around 80 percent of unmarried parents were cohabiting when their child was born (Langmeyer 2015), these couples were more likely to separate than married parents (Schnor 2012). In 2017, 19 percent of all German families with minor children were single-parent households, and the overwhelming majority of these families were headed by the mother (Statistisches Bundesamt 2018a). Almost half (43 percent) of these single mothers were never married (ibid.). In addition, estimates from surveys suggest that around 13 percent of all households with minor children are stepfamilies, with most being stepfather families (Bundesministerium für Familie, Senioren, Frauen und Jugend 2013).
The involvement of separated fathers in the lives of their children seems to be more limited in Germany than it is in other Western countries. In a cross-national study, Kalmijn (2015) compared 14-year-old students' post-divorce contact and relationship with their father in Germany, the Netherlands, England, and Sweden. The highest share of young people who had no contact with their father was found in Germany (21 percent). Conversely, the findings indicated that the rate of shared physical custody was lowest in Germany (ten percent), closely followed by England (eleven percent), and was highest in Sweden (36 percent). Since the pre-separation division of labor has been shown to affect fathers' post-separation involvement in child rearing (Poortman and van Gaalen 2017), these figures may reflect differences in gendered patterns of family roles. In Germany, the modernized provider model is most prevalent, as the high share of mothers in part-time employment and the substantial gender gap in the number of hours working mothers and fathers spend in employment demonstrate (OECD 2017).
Previous research on determinants of post-separation father-child contact in Germany has followed roughly two lines. From a sociological perspective, structural features that reflect parental resources and commitments have been investigated, including maternal employment, family SES, parents' educational resources, parents' former marital status, the father's legal custody rights, the father's current partnership status, and the children's ages and genders (e.g., Kalmijn 2015;Köppen et al. 2018). From a more psychological perspective, studies that were often inspired by issues raised in legal conflicts between separated parents have focused on the relationship dynamics between parents (e.g., Amendt 2004;Behrend 2010;Blesken 1998;Walper 2006Walper , 2019. In this latter line of research, scholars have investigated not only the father's role identity and approach to coping with conflict, but also the role of the mother as the gatekeeper in the father-child relationship. Earlier debates that focused on extreme forms of maternal gatekeeping suggested that parental alienation is a common process through which the mother involves the child in a close alliance against the father, and thus causes the child to become distanced from the father (Kodjoe and Koeppel 1998). However, more recent research has shown that this description of post-separation family dynamics oversimplifies the processes involved, and neglects considerable variation between cases (Behrend 2010).
Our study seeks to provide a differentiated view of the factors linked to problems in the interparental relationship and their likely outcomes for the father-child relationship. Our focus is on the role of maternal gatekeeping, which has been proposed as a unifying concept that explains the disadvantages in social capital and resources of children whose parents have separated relative to children raised in nuclear families (Austin et al. 2013). In addition to looking at gatekeeping attitudes and behaviors, we will address the mother's and the father's attitudes toward each other and the issues surrounding coparenting, both of which may reflect each parent's resentment of his/her former partner, as well as more objective conditions linked to the father's failure in the role of provider.

Prior Research
In any dual parenting relationship, parents have to decide how to share their responsibilities to their children. Accordingly, engaging in negotiations about the extent and the type of involvement each parent has with the children, which may include making decisions about whether the children require protection from the other parent's behavior, is a natural part of coparenting (Austin et al. 2013). Such decisions are likely to be made more deliberately in separated families than they are in nuclear families, as the parents have to arrange the children's visits and overnight stays with each parent, at least as long as the children are too young to make these arrangements on their own. More importantly, since the often fraught process of separation and divorce does not necessarily facilitate the resolution of prior problems between parents, conflicts about parenting issues and attempts to limit or undermine the other parent's relationship with the child tend to be more common in separated than in nuclear families (Fagan and Barnett 2003;Walper et al. 2005).
As the findings discussed above indicate, the notion of parental gatekeeping is often cited as playing a major role in conflicts between separated parents over parenting time and the non-resident parent's access to the child (Austin et al. 2013). Parental gatekeeping has been defined as encompassing "attitudes and behaviors by either parent that affect the quality of the other parent-child relationship and/or level of involvement with the child" (Austin et al. 2013: 486). It comprises not only attitudes and behaviors that may restrict the other parent's interaction and relationship with the child, but also attitudes and behaviors that may facilitate this relationship (e.g., Altenburger et al. 2018). Restrictive gatekeeping would be evident in efforts to marginalize the other parent, refusals to communicate, and to deliver derogatory messages about the other parent in the presence of the child; while facilitative gatekeeping would be exemplified by efforts to include the other parent in communication and decision-making, and conveying a positive image of the other parent (Austin et al. 2013: 488). Most research that has focused on restrictive gatekeeping has provided evidence that it is linked to fathers being less involved with their children (e.g., Allen and Hawkins 1999;Fagan and Barnett 2003;Stevenson et al. 2014). Furthermore, some of these studies have found evidence that maternal encouragement of the father being involved with his children has positive effects, even after controlling for the quality of the interparental relationship (Fagan and Cherson 2017).
There is considerable overlap between the concepts of gatekeeping and coparenting. Coparenting refers to "the ways that parents and/or parental figures relate to each other in the role of parent" (Feinberg 2003: 96). The term is conceptualized as a multi-dimensional construct that includes features of cooperation, agreement, conflict, and triangulation/undermining (Feinberg 2003;Teubert and Pinquart 2010). The last of these features is at the core of restrictive gatekeeping; i.e., efforts to undermine the other parent in his/her parenting role. Using a typological approach to analyzing coparenting in separated families, Lamela et al. (2016) identified three groups of coparenting relationships: cooperative coparenting (48 percent), highconflict coparenting (13 percent), and undermining coparenting (39 percent). While the undermining coparenting group were found to have levels of agreement and support that were as low as those of the high-conflict group, open conflict was reported less frequently in the undermining group. Interestingly, undermining behavior was similarly evident in the high-conflict group. These results are in line with findings from Germany indicating that undermining behavior is distinct from but related to interparental conflict (Walper et al. 2005).
Interparental conflict has often been cited as a risk factor that threatens a nonresidential father's levels of contact and relationship quality with his children (Coiro and Emery 1998;Walper and Beckh 2006;Walper and Krey 2009); although such effects have not always been found (e.g., Sobolewski and King 2005). There is also evidence that even in married or cohabitating couples, interparental conflict is linked to the father having reduced access to his young children (Hohmann-Marriott 2011). In line with the assumption that interparental conflict triggers restrictive gatekeeping by the mother, findings from a longitudinal study have shown that, over time, the parents having marital problems was linked to later maternal gatekeeping, which, in turn, reduced the level of father-child interaction (Stevenson et al. 2014). However, there is also evidence for reverse effects in intact families, which suggests that the father's involvement may have positive effects on later coparenting (Jia and Schoppe-Sullivan 2011).
While coparenting conflict is often accompanied by undermining or gate-closing behavior, cooperative coparenting can be understood as facilitative gatekeeping or gate-opening behavior. For a separated family, a cooperative coparental relationship may encourage the father to be more involved with his children. In a study that included separated as well as nuclear families, higher levels of coparental coordination were found to be associated with more paternal involvement across family types, whereas low coparental coordination was shown to be associated with less paternal involvement, especially among nonresidential divorced fathers (Finzi-Dottan and Cohen 2016). As might be expected, divorced fathers reported experiencing lower levels of coparental coordination than married fathers. Similar evidence from the Fragile Families Study showed that the father's level of involvement tended to be higher when the coparenting relationship was positive (McClain and DeMaris 2013). Some findings have suggested that in separated families, parental cooperation may play an even more important role than conflict (Sobolewski and King 2005). In this latter study, parental cooperation, but not conflict over child rearing, was found to predict the father's level of contact with his children, as well as his level of responsive fathering and the quality of his relationship with his children; with contact mediating the effects on the quality of his fathering and his relationship with his children.
In seeking to understand the emotional and attitudinal context of maternal gatekeeping, some studies have examined the mother's attitudes regarding the father and her perceptions of his role performance. As has been pointed out, the mother may restrict the father's access to their joint children because she is angry at the father, or because she feels he has opted out of his responsibilities to their children (Greif 1997). Similarly, the mother's continuing hostility toward her ex-spouse has been related to reduced or lost father-child contact (Buchanan et al. 1996), and to the father being driven away (Braver and O'Connell 1998). The mother is particularly likely to harbor feelings of disappointment and resentment if the father is not committed to his role as provider and withholds child support payments. Hence, it might be assumed that a father's access to his children would be restricted only if he was unable or unwilling to provide for them. Findings from the Fragile Families and Child Wellbeing Study supported this assumption in the case of separated families, but pointed to a different effect among fathers in intact families, with these fathers being less involved if their financial contributions to the family were high (Carlson et al. 2017). Other data also show a link between child support payments and fatherchild contact (Amato et al. 2009). While there is only very limited evidence regarding the impact of child support payments on father-child contact in Germany, findings from an online study of separated fathers also suggested that fathers with low economic resources have less access to their children (Amendt 2004). However, as these studies did not address issues of gatekeeping, the question of whether the mother restricts the father's access or the father withdraws because he is failing to provide remains unresolved. Indeed, there are qualitative data suggesting that gatekeeping work -regardless of whether it is gate-opening or gate-closing -can be a dynamic transactional process, rather than a linear and unidirectional process running from the mother to the father (Trinder 2008).
In addition to restrictive and facilitative gatekeeping, protective gatekeeping has been identified as a third type of approach a parent might use to manage the other parent's involvement (Austin 2018). Research has suggested that gate-closing behavior may reflect problems that go beyond the interparental relationship or a parent's feelings of resentment towards his/her former partner. The mother may also limit the father's access if she is worried about the children's well-being (Nixon and Hadfield 2018). Thus, under certain circumstances, the aim of gate-closing can be to protect the children when the other parent's behavior or a given parenting plan is seen as jeopardizing the children's well-being (e.g., Saini et al. 2017). Such concerns may motivate maternal gatekeeping, but they can also lead the father to withdraw in order to avoid exacerbating the children's stress levels and loyalty conflicts. Accordingly, a differentiated view on gatekeeping is needed that focuses not only on issues of parental resentment and conflict, but on parental worries about the children's well-being.

Aims of This Study
Although there is a large body of work on gatekeeping, the evidence on the role of maternal gatekeeping in shaping a father's post-separation access to his children is still limited. In particular, the factors mentioned above have not been jointly considered. Thus, the main aim of our study is to shed light on different features of maternal gatekeeping, as well as the likely predictors of maternal gatekeeping, and the effects of such efforts on the father's contact with his children. Furthermore, we seek to include multiple perspectives that address the mother's, the father's, and the children's views of family dynamics. Our analyses focus on four factors that may restrict a separated father's access to his children. Since our major aim is to explore the role of the interparental dynamic, we will not address cases in which there is no father-child contact (which would cause missing data on relationship problems), but will instead seek to explain situations in which the father-child contact is infrequent.
First, it is assumed that the father's failure to provide (i.e., missing support payments or failing to make payments on time) impedes his access (e.g., Amendt 2004), either because he withdraws in response to being unable to fulfil his financial obligations, or because the mother restricts the children's access to their father in response to his failure to comply with his obligations (provider hypothesis). Second, we expect to find that the mother having a negative view of the father undermines her willingness to facilitate the children's contact with their father, and is thus linked to the children having infrequent contact (resentment hypothesis). We also test whether the father having a negative view of the mother is linked to infrequent contact with the child, based on the assumption that the father's resentment of the mother might lead him to avoid contact. Third, since interparental conflict and coparenting problems have often been identified as potential barriers to father-child contact, we assume that coparenting conflict contributes to infrequent contact, while successful cooperation facilitates contact (coparenting hypothesis). Fourth, our focus is on maternal gatekeeping and parental worries about the children's wellbeing, which may inhibit the father's access to his children (gatekeeping hypothesis). Parental worries are of particular interest, since concerns about children's stress in the context of visitation and interparental problems can be expected to provide a powerful motivation for the mother to restrict father-child contact. At the same time, the father being worried about his children's well-being could lead him to have less contact with his children, as he may avoid contact in order to prevent the children from experiencing stress. Finally, the children's views on their mother's efforts to form an alliance with them against their father (i.e., maternal pressure to take sides) provides relevant information on gatekeeping behavior, since it may be more valid than maternal self-reports.
All of these factors are assumed to be interrelated, as Fig. 9.1 shows. We expect to find that coparenting problems mediate the effects of financial problems, based on the assumption that making insufficient support payments affects father-child contact only if this places stress on the interparental relationship. Similarly, we expect to observe that coparenting problems at least partly mediate the effects of negative attributions on infrequent contact. Furthermore, we assume that protective gatekeeping (which stems from a parent being worried) or restrictive gatekeeping (which places pressure on the children to choose sides) partly mediate the effects of both a negative view of the other parent and coparenting problems on the children's access to their father.

Samples
We used two datasets to test our hypotheses: a small, intensive cross-sectional sample of separated mothers and fathers, most of whom sought counseling, mediation, or parenting training (KiB) and participated in an evaluation study (KiB sample); KiB Sample The KiB sample was drawn from an ongoing intervention study that targets a broad range of separated parents, including those in highly conflicted families. Accordingly, this sample overrepresents conflicted cases (33 percent involved in court litigation 1 ). The majority of the participants were recruited through counseling centers or at municipal institutions, like kindergartens, schools, youth welfare offices, law offices, or family courts. Although this sample is not representative, the higher prevalence of conflict makes it particularly suitable for exploring the role of interparental problems. The data used here come from the pre-test conducted for those parents involved in an intervention (KiB parenting course for separated parents; fathers: 80 percent, mothers: 71 percent) and the smaller control group (fathers: 20 percent; mothers: 29 percent). Since the KiB course does not include former couples in the same parenting group, the maternal and paternal subsamples are largely independent, and comprise a smaller subset of the former sample of couples (22 percent/19 percent of the fathers/mothers). As a prerequisite for the intervention, almost all of the non-resident parents still had contact with their children, but the contact frequency varied between shared parenting and weekly visits to less than one visit per year. Non-resident mothers (n = 18) and resident fathers (n = 22) were excluded due to their low numbers. Similarly, 14 mothers and twelve fathers who had no contact with their children's other parent were not included due to missing information on coparenting conflict. Furthermore, the sample was restricted to parents with a minor child between the ages of three and 17 years. The final sample comprised 160 fathers (non-resident or shared parenting) and 187 mothers (resident or shared parenting) who participated in the written questionnaire assessments. Table 9.1 provides an overview of the sample, separately for fathers and mothers. The parents' reports on the children's ages and genders were based on a target child (randomly selected if the parents had more than one child in the age range three to 17 years). Substantial shares of the fathers and of the mothers reported engaging in shared parenting: 21 percent and 14 percent, respectively. These figures are considerably higher than those suggested by other data for Germany (Kalmijn 2015), and likely reflect high levels of paternal involvement. Furthermore, highly educated parents are overrepresented in this sample, as is typical for the outreach of such interventions. About half of the parents had successfully completed a higher level of general schooling; i.e. they had graduated from the advanced track of schooling required for entry into university or an advanced technical college ("Abitur" or "Fachabitur"). As expected based on other research, the per capita net household income (needs weighted according to the OECD) was substantially higher for the fathers (M = 2766 €, SD = 2094) than for the mothers (M = 1535 €, SD = 821.5).
The average ages of the children did not differ for the fathers (7.2 years, SD = 3.8) and the mothers (7.1 years, SD = 6.7).

Pairfam Sample
The second sample was drawn from the German Family Panel pairfam (Huinink et al. 2011), a three-cohort longitudinal study on family development with annual assessments, which was started in 2008/2009 for three birth cohorts (see www.pairfam.de). The participants were recruited through register data provided by the local administration and personal visits of the interviewer, who conducted the interviews in the participants' homes. The data used here were drawn from waves seven and eight (Brüderl et al. 2017), and were restricted to separated mothers whose child(ren) participated in the child interview (for children aged eight to 15 years) in wave seven. If more than one child in a given family had participated in the child interview, we selected the youngest child for our analyses. In line with the approach used in the KiB, the cases in which there was no contact between the two parents and between the child and his/her non-resident father in wave seven were not included in the analyses, since the core predictors of contact (coparenting, maternal pressure to take sides) were filtered by contact. Omitting cases of parental reunion or recent parental separation (between waves seven and eight), the final sample included N = 145 separated mothers and their youngest child, who was eight to 15 years old. Notes: a The assessment of joint physical custody was based on the parents' estimates of whether the child was spending equal amounts of time at each parent's home in the KiB study, and was based on maternal reports of the number of overnights the child was spending with each parent in the pairfam study (50:50 up to 40:60) b High education: General school qualification for university/university of applied sciences Descriptive information for the pairfam sample is shown in Table 9.1 (right column). Unlike in the KiB results, but in line with other findings for Germany (Walper et al. 2020), only three percent of the mothers in the sample reported engaging in shared parenting, while 97 percent indicated that they had primary residential custody. Compared to the maternal subsample in the KiB study, a smaller share of the mothers in the pairfam sample reported having a high level of education (42 percent with high general school qualification; mean years of education: 12.6, SD = 2.7). Furthermore, the mothers in the pairfam sample had a lower average equivalent household income. Although the average age of the mothers (M = 38.6 years, SD = 4.7) was more than one year younger than it was in the KiB sample, the average age of the children was four years older (11.2 years, SD = 2.0), which suggests that the parents in the pairfam sample were substantially younger when their children were born than the parents in the KiB sample. Several factors may have contributed to this difference, including the smaller share of highly educated mothers in the pairfam sample (as higher education is linked to later childbearing) and the larger share of East German families in the pairfam sample (as the average maternal age at childbirth is lower in East than in West Germany). Of the mothers in the pairfam sample, 32 percent were living with a new partner (remarried or cohabitating), and ten percent were first-or second-generation migrants.

KiB Data
Frequency of Father-Child Contact The frequency of contact between the child and the non-resident father was reported by the resident mother answering the question: "If your child lives with you, how often does the other parent see your child?"; and by the non-resident father answering the question: "If your child does not live with you, how often do you see him/her?" Parents selected a response from a fourpoint continuum ranging from "at least weekly" (coded one; mothers: 58 percent; fathers: 58 percent) across "every two weeks" (coded two; mothers: 25 percent, fathers: 31 percent), "once per month" (coded three; mothers: five percent; fathers: three percent), to "less often than once a month" (coded four; mothers: twelve percent, fathers: eight percent). Cases in which the parents reported engaging in shared parenting but their child's moves between the two parental households were minimal (one father and three mothers), or in which there was a lack of data on contact frequency (nine fathers and nine mothers), were recoded as high contact.
The predictors used in our analyses were consistent with short scales (3-5 items) that had good to satisfactory internal consistency for both parents (Cronbach's Alpha 0.75-0.90). All of the items were answered on a five-point scale (from 1 = never to 5 = very often). Negative Attributions of the other parent's behavior were assessed by five items based on the "Relationship Attribution Measure" (RAM) developed by Fincham and Bradbury (1992). The items indicated attributions of bad intent or dysfunctional behavior to the other parent (e.g., "She/he often hurts me on purpose." or "Most of our disputes are provoked by her/him."). Cronbach's Alpha for this index was 0.87 for fathers and 0.83 for mothers. Coparenting Conflict was indicated by three items based on Ahrons' Coparenting Conflict Scale (Ahrons 1981) (e.g., "Do you and your former spouse have basic differences of opinion about issues related to child rearing?" or "When you and your former spouse discuss parenting issues, how often does this result in an argument?"). Cronbach's Alpha was 0.87 for fathers and 0.88 for mothers. Child-Related Worries were assessed by a three-item indicator developed for this project (e.g., "Are you worried about the mental and/or physical well-being of your child because of the other parent?" or "Do you think that your child sometimes suffers from loyalty conflict?"). Cronbach's Alpha was 0.90 for fathers and 0.75 for mothers. Observations with missing information on any of these items (ranging from 1.3 percent to 13.1 percent) were retained in the models using the Full Information Maximum Likelihood (FIML) method (Arbuckle 1996).

Pairfam Data
Frequency of Father-Child Contact was reported by the mother (question: "How often does the other parent see child x?"). The frequency rating ranged from "daily" (coded one; seven percent) over "several times per week" (coded two; 17 percent), "once per week" (coded three; 15 percent), 1-3 times per month" (coded four; 37 percent), and "less often" (coded five; 19 percent), to "no contact" (coded six; six percent in wave eight; cases without contact in wave seven were excluded in this sample).
In order to test our hypotheses, the following predictors were included. Father's Failure to Provide was indicated by two dummy-coded variables based on information given by the mother on the child support payments made by the father. The first dummy indicator contrasts full payment with partial or no payment. The second dummy indicator contrasts no payment with partial or full payment. The no-payment category also includes cases in which the mother indicated that she was not entitled to child support payments. Although this may have been due to a shared parenting arrangement, the large share of mothers who reported that they were not entitled to receive child support payments (20 percent) suggests that some of the information on the father's inability to provide was false or based on a misinterpretation. Two indicators of coparenting quality were included that were selected from the "Parent Problem Checklist" (Dadds and Powell 1991). Coparenting Conflict was indicated by three items (e.g., "Discussions regarding parenting issues end in fights." Cronbach's Alpha = 0.86). Coparenting Cooperation was measured by a single item ("When there is a problem with your child or children: How often do you and the father of your child try to solve the problem together?"). Maternal Pressure to Take Sides was indicated by a single item ("My mother tries to get me to take sides against my father."). Although pairfam provides six items indicating the pressure to take sides exerted by both parents (see Thönnissen et al. 2019), the internal consistency for the maternal subscale was very weak (three items; Cronbach's Alpha = 0.48). Hence, we chose the item that most closely addressed the targeted maternal behavior. Except for indicators of family structure and contact frequency, missing information was imputed using the Full Information Maximum Likelihood (FIML) method.

Findings from the KiB Data
The KiB data were used to estimate separate path models for maternal and paternal reports. Due to the lack of suitable indicators, these analyses could not address the provider hypothesis. Path models were estimated within a structural equation modelling framework using Maximum Likelihood Estimation (MLE). Indirect effects were calculated using a product-of-coefficients method (MacKinnon et al. 2002) and bootstrap standard errors with 1000 replicates. Mplus statistical software was used. The findings for the mothers and the fathers are shown in Figs. 9.2 and 9.3. In both models, control variables were included if their association with an outcome in the model was statistically significant at p ≤ 0.20. For the mothers' report, the child's age was controlled for effects on coparenting conflict and father-child contact. For the fathers' report, the child's gender was controlled for effects on coparenting conflicts, and the child's age and the father's income were controlled for the effects on the father-child contact. To keep the figure parsimonious, these effects of the control variables are not displayed in Figs. 9.2 and 9.3.
As the figures show, both models have an excellent fit to the data, with many overall similarities, but also a few differences. At the same time, these models provide only partial support for our hypotheses. The model estimated for mothers provides evidence for only one significant predictor of low father-child contact. Supporting the resentment hypothesis, negative attributions were linked to low father-child contact. However, this effect was not found to be very strong (β = 0.21, p < 0.05), yielding a low explained variance for low contact (R 2 = 0.06). As expected, negative attributions were shown to be strongly linked to coparenting conflict (β = 0.42, p < 0.001), and to explain 18 percent of the variance in coparenting conflict, which was, in turn, linked to mothers' worries about the well-being of their children (β = 0.32, p < 0.001). In addition to coparenting conflict, negative attributions substantially predicted maternal worries (β = 0.45, p < .001). Hence, maternal worries could be well explained in this model (R 2 = 0.42). As expected, the model indicated that coparenting conflict partly mediated the effects of negative attributions on maternal worries (indirect effect: b = 0.15, β = 0.13, p < 0.05). However, contrary to our assumptions, neither coparenting conflict nor child-related worries were found to predict low contact. Accordingly, there was no significant indirect The findings for fathers were very similar to those for mothers in terms of the links between negative attributions, coparenting conflict, and fathers' worries regarding the well-being of their children (see Fig. 9.3). In line with the results for mothers, the fathers' negative attributions were found to be strongly linked to coparenting conflict (β = 0.59, p < 0.001) and the fathers' worries about their children (β = 0.56, p < 0.001), with an additional direct link between coparenting conflict and child-related worries (β = 0.30, p < 0.01). However, the fathers' negative attributions regarding the mother were not shown to be a significant predictor of low contact. Only the fathers' worries were found to be linked to low contact (β = 0.46, p < 0.001). Similar to the findings for mothers, the explained variance (including control variables) was lowest for contact (R 2 = 0.16), higher for coparenting conflict (R 2 = 0.37), and highest for paternal worries (R 2 = 0.60).
For fathers, unlike for mothers, all of the indirect paths were significant. In line with our hypotheses regarding mediation, the effects of negative attributions on the fathers' worries about their children were significantly mediated by coparenting conflict (indirect effect b = 0.20, β = 0.17, p < 0.01). Furthermore, the effects of coparenting conflict on infrequent contact were mediated by child-related worries (indirect effect b = 0.11, β = 0.14, p < 0.01). Finally, the indirect link between negative attributions and infrequent contact, mediated by coparenting conflict and paternal worries, proved weak but significant (indirect effect b = 0.06, β = 0.08, p < 0.05). Hence, we found some support for the resentment hypothesis and the interparental conflict hypothesis, although both factors were shown to have only indirect effects on the fathers' levels of contact with their children. More substantially, these findings suggest that the fathers were likely to withdraw if they perceived that their children were stressed or caught in loyalty conflicts.

Findings from the Pairfam Data
While the KiB data allowed us to test the resentment hypothesis, the respective information was missing in the pairfam data. However, as the pairfam included indicators of the fathers' child support payment histories, it allowed us to address the provider hypothesis using the two dummy variables described in section "Pairfam data". Furthermore, we were able to draw on additional information regarding the coparenting relationship and to include coparenting cooperation as a likely resource for more frequent contact. Furthermore, the pairfam data provided us with a more conclusive test of maternal gatekeeping, as the data included reports from the children on maternal pressure to take sides; i.e., the mother's attempts to involve the child in an alliance with her against the father. Finally, the pairfam data allowed us to test our hypotheses longitudinally by predicting father-child contact in wave eight by family dynamics at wave seven (one year earlier), controlling for previous father-child contact.
To shed more light on background factors, we used multiple regression analyses to test the extent to which infrequent father-child contact was linked to demographic factors such as maternal education, child age, and gender. Tables 9.2 and 9.3 present the findings from a series of multiple linear regression analyses based on the sample of separated mothers and their children (n = 146). The predictor variables are shown in rows, and the dependent variables are shown in columns. Based on the model assumptions ( Fig. 9.1), we first analyzed the predictors of the explanatory factors assessed in wave seven: child support payments (dummy-coded as described in section "Indicators"), coparenting with two indicators of conflict and cooperation, and maternal pressure to take sides (based on the children's reports). Table 9.3 reports the findings for the predictors of the outcome; i.e., low frequency of contact with the father in wave eight, using four models. Model 1 (M1) included only demographic background factors and child support payments; Model 2 (M2) added both coparenting indicators; Model 3 (M3) added maternal pressure to take sides, and Model (M4) added infrequent contact with the father in wave seven.
Notes: a Child support payment dummy coded; reference category: irregular or only partial payment; the first column of effects on mediators shows coefficients for both dummy variables as dependent variable Pairfam data (release 8.0.0) from waves 7 and 8, N = 145 separated mothers with child who participated in the child interview; significance: + p < .10, * p < .05, ** p < .01, ** p < .001 As Table 9.2 shows, the demographic variables were largely unrelated to the main predictors of the frequency of contact with the father. Only two out of 15 regression coefficients proved significant with one additional marginal effect. Contrary to the assumption that fathers are often more willing to invest in male children, we found that girls were less likely than boys to be receiving no support payments; i.e., that girls were more likely than boys to be receiving at least some payments. However, the likelihood of receiving full versus no or only some child support payments was not found to differ by the child's age or gender. Negative family dynamics were reported more frequently by younger than older children. The findings indicated that coparenting conflicts declined significantly as the children's ages increased, and that older children reported marginally less maternal pressure to take sides than younger children. No other effects of age were found. Maternal education proved insignificant for any of the main predictors or mediators, but was shown to be negatively linked to the child having infrequent contact with the father; i.e., the children of better educated mothers had more frequent contact with their fathers than the children of less educated mothers (see Table 9.3). This effect was found to be robust for all models predicting child contact in wave eight.
Neither the child's age nor gender had any effect on whether the frequency of the child's contact with the father was reduced (see Table 9.3). The results showed that, in addition to maternal education, child support payments mattered, albeit in an unexpected way. We found that the fathers who were paying no child support were less likely to have infrequent contact than the fathers who were paying some or full child support. Indeed, the findings indicated that while 38 percent of the fathers who were paying no child support saw their children several times per week or daily, only 16 percent of those who provided at least some child support had such frequent access to their children (bivariate analysis, Χ 2 = 13.95, df = 5, p < 0.05). This effect Notes: a Child support payment dummy coded; reference category: irregular or only partial payment; the first column of effects on mediators shows coefficients for both dummy variables as dependent variable Pairfam data (release 8.0.0) from waves 7 and 8, N = 145 separated mothers with child who participated in the child interview; significance: + p < .10, * p < .05, ** p < .01, *** p < .001 remained largely unchanged when introducing the mediators, but was attenuated to marginal significance after controlling for the stability of contact. As was found for the maternal reports in the KiB data, coparenting conflict was not shown to be linked to a lower frequency of contact. However, the analysis found that the parents' cooperation in coparenting had a highly significant effect that was not reduced when the maternal pressure to take sides was also included. As expected, we found that higher levels of coparenting cooperation were linked to more frequent contact between the father and the child one year later. However, this effect disappeared in Model 4, which controls for previous contact. Finally, maternal pressure to take sides showed the expected effect. Although the bivariate link between maternal pressure to take sides and the child's later reduced frequency of contact with his/her father (r = 0.188, p < 0.05) was slightly reduced in the context of the other predictors, it remained significant even when controlling for previous contact (β = 0.143, p < 0.05).
In sum, the pairfam data supported neither the provider hypothesis nor the interparental conflict hypothesis. In fact, our findings on the effects of child support payments were contrary to our assumptions. While coparenting conflict appeared to have no impact on the likelihood of having infrequent contact, our data suggest that coparenting cooperation was significantly linked to later father-child contact, but did not predict changes in contact levels across time when previous contact levels were controlled for. However, some support was found for the maternal gatekeeping hypothesis, since maternal pressure to take sides predicted lower levels of fatherchild contact, even when controlling for previous contact levels.

Discussion
This study addressed different features of maternal gatekeeping in separated families. As we pointed out, there is considerable conceptual overlap between gatekeeping and the quality of coparenting (e.g., Austin et al. 2013;Cannon et al. 2008). While cooperative coparenting is likely to indicate gate-opening behavior, triangulation/undermining coparenting reflects gate-closing behavior. It has also been argued that protective gatekeeping is a special case of restrictive gatekeeping, whereby one parent is seeking to secure the well-being of the children when s/he perceives that the other parent or a particular visitation schedule puts the children at risk. Moreover, it has been suggested that conflict between parents, and coparenting conflict in particular, is a predictor of gate-closing behaviors and attitudes. Using two different datasets, we were able to investigate the extent to which coparenting quality was linked to protective or restrictive gatekeeping, and whether gatekeeping was, in turn, linked to less frequent father-child contact. The first dataset (KiB) allowed us to include information on the mother's and the father's attitudes toward the other parent as likely predictors of coparenting conflict, protective gatekeeping (i.e., each parent's worries about the children's well-being), and the father's access to his children. Our effort to address not only the mother's but the father's worries about the well-being of their children was intended to shed light on an alternative interpretation of infrequent father-child contact: i.e., that it might be driven by protective withdrawal by the father rather than by maternal gatekeeping. The second dataset (pairfam) provided information on the fathers' child support payments, and allowed us to test the extent to which a father's failure to provide was related to negative and positive features of coparenting, maternal gatekeeping, and fatherchild contact. Overall, the findings revealed a more differentiated picture of maternal gatekeeping than the one that is often presented in the literature.
First, our findings from the KiB sample highlighted the role of interparental attitudes, as several previous studies have pointed out (Braver and O'Connell 1998;Buchanan et al. 1996;Greif 1997). For mothers as well as for fathers, having a negative view of the other parent was found to be linked to higher levels of coparenting conflict, more child-related worries, and less frequent father-child contact. However, the effects of these views on father-child contact were shown to be much stronger for mothers than for fathers. The findings indicated that while the father's negative attributions regarding the mother were only indirectly related to infrequent contact, the mother's negative attributions had a direct effect, and proved to be the only significant predictor of infrequent father-child contact. Hence, these findings support the resentment hypothesis, particularly for mothers.
Second, we found only limited support for the maternal gatekeeping hypothesis. According to the KiB data, mothers' child-related worries were not related to infrequent father-child contact. Hence, these data did not confirm our assumption that protective maternal gatekeeping played a significant role in determining the father's access to his children. It should be emphasized that the KiB sample overrepresents highly conflicted cases, in which protective gatekeeping might be particularly likely to occur (e.g., Austin 2018;Trinder 2008). Interestingly, however, we found that the father's, and not the mother's worries about the well-being of the children were linked to reduced contact. This suggests that fathers may withdraw under such circumstances rather than being pushed out. Future research should aim to provide more conclusive evidence based on longitudinal data on the causal links between fathers' child-related worries and their tendency to withdraw. Nevertheless, our findings point to the salience of paternal attitudes, and should be of particular interest for custody evaluators.
Although we found no support for the claim that protective gatekeeping has significant effects, the pairfam data on restrictive maternal gatekeeping were in line with our assumptions. A mother's efforts to get her child involved in an alliance against the other parent (pressure to take sides), as reported by the child, were longitudinally linked to the father having less frequent contact with the child, even when controlling for father-child contact in the previous year. However, it should be noted that this effect was weak, which suggests that maternal gatekeeping is not a powerful tool in determining fathers' access to their children. Only two percent of the variance in father-child contact could be explained by maternal restrictive gatekeeping. At the same time, we have to caution that our indicator of restrictive gatekeeping was based on a single item only. More powerful indicators may yield different findings.
Third, the evidence on the coparenting hypothesis was similarly mixed. Although coparenting conflict was assumed to provide an important predictor of maternal gatekeeping and of the father having reduced access to his children, only the first part of this assumption found some support. In the KiB data, both parents' reports on coparenting problems were found to be linked to being more worried about their children's well-being, but they did not predict low levels of contact. In the pairfam data, coparenting conflict was not shown to be related to maternal restrictive gatekeeping, as reported by the children, or to contact. Hence, the coparenting hypothesis regarding conflict was not supported by our data. However, for coparenting cooperation, the pairfam data suggested that there was a substantial link to more frequent later contact. These findings are in line with other research that suggested that coparenting cooperation or gate-opening plays a more important role than coparenting conflict in contact frequency (Sobolewski and King 2005). At the same time, we have to point out that this link proved insignificant when controlling previous contact; i.e., positive parental cooperation did not appear to drive positive changes in father-child contact. Hence, any causal interpretation of these findings has to await further evidence on possible reciprocal or more long-term links between coparenting and father-child contact.
Fourth, the pairfam data did not support the provider hypothesis. Instead, they suggested that fathers with frequent access to their children were even less likely to pay child support. This finding is unexpected, and does not conform to current German family law, which requires non-resident parents to make financial payments to cover their children's needs, except in cases in which their earnings cover only their own basic needs. However, it is possible that a father who has particularly close contact with his children provides goods and services instead of financial payments to the mother.
Our data clearly have shortcomings. The KiB study does not yet allow for longitudinal analyses, and while the pairfam study enabled us to analyze longitudinal data in a one-year time frame, longer periods with multiple waves would be preferable. Furthermore, our indicators of coparenting cooperation and maternal pressure to take sides in the pairfam study were weak, since we had to rely on single-item indicators. Finally, neither of these datasets provided all indicators of interest. Nevertheless, both sets of analyses, when seen in a conjunction with each other, add to our knowledge of gatekeeping processes. Overall, since only little of the variance in levels of father-child contact could be explained by the predictors analyzed here, our findings suggest that other factors may be more powerful in encouraging or restricting father-child contact. Further research, preferably based on larger samples, should place features of the interparental dynamic in the larger context of the parents' current living conditions, their involvement with new partners, and the legal framework of parental responsibilities that have been found relevant for fathers' involvement with their children (see Köppen et al. 2018).
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.