Mentalizing in first-time fathers: reflective functioning as a mediator between attachment representation and sensitivity

ABSTRACT Sensitivity in parent-child interaction is essential for child development. Since fathers are increasingly involved in childrearing, identifying factors leading to paternal sensitivity is crucial. We examined the relation between attachment representation and reflective functioning (RF) as factors influencing paternal sensitivity in a longitudinal study including N = 40 first-time fathers (M age = 33) and their 6-month-old children. We used the Adult Attachment Interview during pregnancy to assess paternal attachment representation and general RF, the Parental Development Interview to assess fathers’ parental RF, and the Emotional Availability Scale to measure sensitivity at child’s age of 6 month. Data show that secure paternal attachment representation, high general and parental RF are associated with higher levels of paternal sensitivity. Further, parental RF mediates the association between attachment representation and paternal sensitivity. These findings contribute to the identification of a causal interplay in that they suggest an explanatory effect of RF on the association between fathers’ attachment representation, and sensitivity.


Introduction
It is well established that sensitive parenting promotes positive child development, notably better cognitive and language skills, executive functions, as well as fewer behavior problems (Cooke et al., 2022;Malmberg et al., 2016;Rodrigues et al., 2021;Towe-Goodman et al., 2014).Moreover, parental sensitivity is considered to be a key predictor contributing to secure parent-child attachment (Ainsworth et al., 1978;Bakermans-Kranenburg et al., 2003;van Ijzendoorn & De Wolff, 1997;Verhage et al., 2016).Given the importance of sensitive parenting for child development, a central endeavor in developmental science has been to understand which factors lead a parent to act sensitively when interacting with their child.The majority of this literature has focused on the role of mothers only.However, recent evidence indicates that paternal sensitivity also matters for positive child cognitive, socio-emotional and attachment development (e.g.Bakermans-Kranenburg et al., 2003;Cooke et al., 2022;Deneault et al., 2021;Lucassen et al., 2011;Rodrigues et al., 2021;Walter et al., 2019).Therefore, research should also focus on investigating potential mechanisms underpinning paternal sensitivity.
Parental reflective functioning (RF) refers to parents' ability to reflect upon their own mental states and experiences as well as those of their child.Studies have shown that parental RF is associated with sensitive parenting (Luyten et al., 2017;Slade et al., 2005).In addition, research shows that adults' secure state of mind with respect to attachment is related to sensitive caregiving behavior (e.g.Behrens et al., 2016).While these associations have been studied primarily in mothers (e.g.Dollberg, 2021;Slade et al., 2005;Stacks et al., 2014), it is not clear whether or to what extent they are also present in fathers.Therefore, the overall aim of the current study was to investigate whether high RF and a secure state of mind with respect to attachment would predict higher levels of sensitivity in fathers when interacting with their children.As research suggests that the predictive power of coherent reflection upon attachment experiences, actually lies in RF (e.g.Fonagy et al., 1991), the second goal of the study was to investigate whether RF also mediates the association between attachment representations and sensitivity.

Fathers' RF and sensitivity
Recent research has identified parental sensitivity as a key factor contributing to children's development (e.g.Cooke et al., 2022;Rodrigues et al., 2021).Sensitive parenting is defined as the ability to understand a child's attachment needs and signals and to respond to them in a warm, prompt, and appropriate manner (Ainsworth, 1989).Sensitive caregivers provide the opportunity for the child to feel physically and emotionally secure.They are also accepting, monitor the child's behaviors and feelings, and are able to create parentchild interactions that are consistent, harmonious, and synchronized (Fearon & Belsky, 2018).Moreover, sensitive interactions are characterized by common and shared positive affect, pleasure, genuineness, and authenticity (Biringen, 2008).
Importantly, sensitivity implies that caregivers can understand their children's thoughts, needs and feelings to a certain degree and that they can mentalize them.This is why high RF is one of the main factors leading to more sensitive and appropriate responses to infant's signals (Laranjo et al., 2008;van Bakel & Hall, 2020).RF can be described as the overt manifestation or operationalization of mentalizing (see Fonagy et al., 1998;Sharp & Fonagy, 2008;Slade, 2005), so these terms may be used interchangeably (Slade, 2005).It should be noted that the overlap or distinction between mentalization and RF is part of a wider literature inquiring the different developmental components of the mentalization construct (e.g.cognitive, emotional, first order, second order e.g.Choi-Kain & Gunderson, 2008;Wu et al., 2020), although this literature is beyond the scope of the current study.RF refers to a person's ability to reflect upon their own intentional mental states and experiences, as well as on those of others, by thinking about and imagining feelings, desires, and wishes (Luyten et al., 2017;Slade et al., 2005).
Developmental research distinguishes a person's general RF and parental RF, the latter of which referring specifically to the parent's ability as a caregiver to think about themselves and the child in the reflective manner described above.Thus, this ability supports recognizing affective cues, as shown in a study of neural processing of infant cues (Rutherford et al., 2017), which is required for sensitive parental behavior.Indeed, previous research in this field -mainly considering mothers -indicates that parental RF is positively related to parenting quality (Stuhrmann et al., 2022).For example, maternal RF was shown to influence mothers' sensitivity when interacting with their toddlers, which in turn was related to the child's attachment pattern (Stacks et al., 2014).In addition, another study found that mothers with high RF during pregnancy later exhibited more positive behaviors in various interaction tasks with their six months old babies (Smaling et al., 2016).These associations are further supported by meta-analytic evidence showing a moderate association between parental mentalizing and sensitivity (Zeegers et al., 2017).
While high RF is considered a protective factor in parent-child relationships, low RF is considered to be a risk factor for parent-child relationships and child developmental outcomes.Some studies have provided clear evidence, showing that low maternal RF not only increases the risk of maternal insensitivity in interactions with children aged three to 16 months in both healthy and depressed mothers (Ensink et al., 2019;Krink et al., 2018), but also increases the risk of disruptive affective communication (Kelly et al., 2005) and the risk for child behavioral problems (Carlone & Milan, 2021).In addition, children with psychosocial problems such as anxiety disorders, difficulties with emotion regulation, and externalizing behavioral problems are more likely to have mothers with lower parental RF levels (see Camoirano, 2017 for a review).
Some studies, however, failed to establish an association between RF and sensitivity or provided mixed results.When comparing adolescent and adult mothers in interaction with their three months old infants, a study found an association between RF and sensitivity for adult but not for adolescent mothers (Riva Crugnola et al., 2018).Moreover, Perry et al. (2015) did not detect a significant association between either antenatal or postnatal RF and emotional availability of mothers when interacting with their three to six months old babies.Yet, the small sample size (N = 17) may explain the lack of significance, and since only substance-abusing mothers were included, it is difficult to generalize the results to the overall population of parents.Therefore, future research should focus on larger, non-clinical samples.Similarly, Salo et al. (2021) found no or only weak correlations between dimensions of RF and sensitivity when children were 12 months of age.The authors argued that this inconsistent finding may be explained by the methodological limitations of self-report measures, and called for future research to further assess RF and sensitivity using interview and observation-based tools.Therefore, the use of self-reports should ideally be accompanied by other measures, as these methods are argued to be less biased and may thereby allow for more clinically detailed and idiosyncratic information (Salo et al., 2021).Together, self-report and interviews, observations, or experiments can make valuable contributions to the literature.
Besides these methodological issues in parental research a major challenge are gender-specific differences, cautioning against mere extrapolation from one parental gender to the other (Cabrera et al., 2018;Camoirano, 2017;Ruiz et al., 2020;Schoppe-Sullivan & Fagan, 2020).Yet, studies investigating the above associations in fathers are rare and their results are somewhat inconsistent.Planalp et al. (2019) found an association between fathers' mind-mindedness and paternal sensitivity when children were three months old, but these effects were no longer significant when the children were five or seven months of age.While a study examining parental insightfulness (as an operationalization of mentalizing) revealed a positive association with parental attunement for both parents during interactions with their 24-58 months old child diagnosed with autism (DiRenzo et al., 2020), another study with substance-abusing fathers found no association between parental RF and paternal behavior in play situations with their one to seven year-old children (Stover & Coates, 2016).Moreover, other studies found paternal RF to be associated with sensitive fathering during a teaching task with two-year-old children (Buttitta et al., 2019) and triadic interactions between fathers, mothers and their 12 to 38 months old toddlers (León & Olhaberry, 2020).Studies including fathers of children older than five years found no or only weak associations between paternal mentalizing and parenting behavior (Decarli et al., 2023;Gershy & Gray, 2020).While these discrepancies may have various causes, differences in childrens' ages or the use of diverse measures including video analysis, interviews but also self-reports that are accompanied by the limitations mentioned above (Salo et al., 2021), may account for the inconsistent outcomes.

Fathers' attachment representations and sensitivity
In addition to RF, a caregiver's early experiences with their primary caregivers and their capacity to represent and make sense of those experiences are important factors influencing whether a parent will be able to act sensitively toward their child (e.g.Hesse, 2016;van IJzendoorn & Bakermans-Kranenburg, 2019;Verhage et al., 2016).Individuals' mental representations of early childhood experiences are often assessed in the Adult Attachment Interview (AAI; George et al., 1996).The AAI can reveal how a person perceived their early relationships and provide insights into how this affects other relationships.The interview elicits attachment experiences and probes for the speaker's reflection of these.It is designed to activate the attachment system and allows for classification of the verbatim narrative according to both Main's four attachment categories (Reiner et al., 2013) and Fonagy and colleagues' Reflective Functioning Scale (Fonagy et al., 1998).Thus, the assessment of whether an adult presents secure or insecure attachment representations does not depend on a positive or negative nature of childhood experiences.Rather, the analysis of the interview focuses on various linguistic and attentional processes, such as the speaker's attentional flexibility between presenting attachment-related experiences and evaluating their influences, as well as maintaining the principles of a coherent and cooperative conversation.In doing so, the AAI's ability to evaluate ones' state of mind with respect to attachment may predict how individuals will approach other close relationships, including those with partner, friends and offspring (Hesse, 2016).Studies have shown the categories of adult attachment representations are strongly related to an individual's parental sensitivity, which in turn influences their child's attachment pattern (e.g.Behrens et al., 2016;McFarland-Piazza et al., 2012;van IJzendoorn & Bakermans-Kranenburg, 2019;Verhage et al., 2016).Specifically, adults with a secureautonomous (F) state of mind typically engage in a coherent and collaborative discourse and are open to the interview topic, indicating a flexibility of attention that can oscillate between different events and people (Hesse, 2016).With respect to caregiving behavior, these parents show mostly sensitive interactions with their child in a consistent manner that is regardless of the child's arousal and attachment signals (van IJzendoorn & Bakermans-Kranenburg, 2019).Transcripts of adults with a preoccupied (E) state of mind with respect to attachment are generally more incoherent and the discourse is characterised by violations of manner, quantity and/or relevance and responses are sometimes contradictory or inflexible with interviewees held back by particular experiences.Therefore, these adults' memories of attachment-related situations often appear to be unbalanced, and sometimes condemning of themselves or their parents.These adults often describe failings of their parents and sometimes exaggerate in portraying minor incidents (Hesse, 2016).In interaction with their child, these parents tend to respond more sensitively to highly aroused children, but may be insensitive when children do not obviously show distress (van IJzendoorn & Bakermans-Kranenburg, 2019).The discourse of adults with a dismissing (Ds) state of mind regarding attachment is characterised by low coherence, often violating the maxims of quality (lacking consistency and truthfulness) and quantity (lack of memory or refusal to discuss), while the attention is inflexibly directed away from the attachment-focused topics of the interview.These adults usually place a lower value on relationships, they only discuss them briefly and frequently give humorous or generalized, overly positive answers that lack substance.They often describe themselves as strong, self-reliant and independent, while the role of feelings and emotions is minimised or disavowed (Hesse, 2016).When interacting with their children, they are more likely to support their child in stress-free situations, while they may be insensitive when children are highly aroused or distressed (van IJzendoorn & Bakermans-Kranenburg, 2019).
This link between parental attachment representations and sensitivity has been well studied in mothers (Adam et al., 2004;Behrens et al., 2016;Sette et al., 2015), yet questions regarding the influence of a fathers' state of mind with respect to attachment on the quality of father-child interactions have been rarely considered even though attachment theory considers both parents as attachment figures (Fearon & Belsky, 2018).Moreover, existing research findings tend to be inconclusive.For instance, McFarland-Piazza et al. (2012) showed fathers with secure attachment representations to be more sensitive and less hostile during interactions with their eight months old infant compared to fathers with insecure attachment representations.In contrast, Aytuglu and Brown (2021) did not find paternal attachment representations and sensitivity in father-child play with 12 months to be directly associated.However, they detected an indirect association between the constructs, mediated by pleasure in parenting (Aytuglu & Brown, 2021).

The mediating effect of RF on paternal sensitivity
In addition to the link between attachment representations and parental sensitivity, the literature also postulates a link between one's state of mind with respect to his or her own attachment history and the development of mentalizing skills (Camoirano, 2017).When individuals have secure-autonomous attachment representations and can coherently reflect on attachment-related experiences, they are more likely to be able to identify other people as individual psychological agents and to attribute intentional mental states as underlying cause of their behavior (Luyten et al., 2012).In fact, research shows that adults with secure attachment representations have higher RF compared to those with insecure attachment representations, who tend to have lower RF (Bouchard et al., 2008;Huth-Bocks et al., 2014;Nazzaro et al., 2017;Slade et al., 2005).Moreover, recent literature suggests RF to have a mediating function in various associations concerning attachment and parent child interaction.For example, one recent study showed that interest and curiosity about mental states, a sub-dimension of RF measured with the Parental Reflective Functioning Questionnaire (PRFQ), partially mediated the relationship between parents' dismissive status on the AAI and their supportive presence when interacting with their six-year-old child (Kungl et al., 2022).Dollberg (2021) investigated a sample of mothers and their three-month-old children demonstrating that the association between maternal anxious or avoidant attachment experiences and maternal sensitivity could be explained by parental RF.It should be noted, however, that this study investigated attachment experiences in close relationships, that show only little correlation with attachment representations in the AAI (Roisman et al., 2007).Similarly, another study by Nijssens et al. (2018) including both mothers and fathers of children aged 8-13 months identified parents' prementalizing of their children's mental states as a mediator of the relation between the attachment dimensions of anxiety/avoidance and parenting stress (Nijssens et al., 2018).Parenting stress in turn could lead to less sensitivity in interaction with the child (see Booth et al., 2018 for a review).Interestingly, this analysis revealed different associations for mothers and fathers.Whereas for mothers the effects were mainly found with regard to attachment avoidance, for fathers, attachment anxiety played a crucial role.Although there are a few studies suggesting RF as a mediator, to our knowledge no study has yet examined the mediating role of RF between attachment representations and parental sensitivity in fathers.

The present study
Our study addresses the research gap regarding the emergence of sensitive parenting behavior in fathers.We aim to disentangle the role of attachment representations and RF in predicting first-time fathers' sensitivity in a non-clinical sample.In line with the literature reviewed above, we hypothesise that paternal attachment representations and sensitivity in father-child interactions are correlated and that RF functions as a mediator of this association.We postulate that reflective abilities lead to sensitive parenting in the sense that sensitive parents will perceive signals by thinking about the mental states of the child and sensitively meeting the child's needs.We propose that secure attachment representations do not directly lead to increased sensitivity, but rather facilitate the latter through mentalizing abilities.
Previous literature has noted that the relation between RF and sensitivity of fathers has only been studied in children one year of age and older.Therefore, our study extends the knowledge about the predictive power of paternal RF to the infant age range as it includes fathers with their 6 month old children.In addition, a common concern in attachment research is that many studies are based on self-report measures only (e.g.Groh et al., 2017;Salo et al., 2021).Therefore, we selected reliable and valid non-self-report instruments, such as the Adult Attachment Interview (AAI) for assessing paternal attachment representations and general RF, the Parental Development Interview (PDI) to assess parental RF, as well as the Emotional Availability Scales (EAS) to obtain paternal sensitivity by rating a videotaped interaction.To allow for the investigation of potential causal mechanisms underpinning the abovementioned interplay (Fearon & Roisman, 2017), we employed a longitudinal study design to test the following hypotheses: (1) There is a positive correlation between fathers' prenatal general and postnatal parental RF and their sensitivity during interactions with their six-month-old children.
(2) Fathers with secure attachment representations are more sensitive than fathers with insecure attachment representations.
(3) The association between fathers' attachment representations and paternal sensitivity is mediated by paternal RF.

Participants
Our final sample included 40 white, middle-class first-time fathers (M age = 32.8years; range = 21-52) and 40 children (70.0%boys; M age = 6.5 months; range = 5-8).Most fathers (n = 34, 85.0%) reported a minimum of 12 years of education with 70.0% (n = 28) having a university degree.Household income was mostly average (42.5%) or above average (30.0%) according to Austrian standards.Only a few fathers (12.5%) were staying at home with their infants at the time of the postpartum assessment, but 27.5% took at least one month of parental leave after birth.Further details on the sample characteristics are presented in Table 1.At study baseline (T1), the sample initially consisted of N = 45 German-speaking first-time parental couples.The participation rate from the pregnancy to the postpartum assessment was 89% with 11% attrition due to refusal to continue.
A sensitivity analysis confirmed that dropouts did not differ significantly in sociodemographic characteristics except for the number of adverse childhood experiences (ACEs; M sample = 1.2 vs. M drop-out = 2.8; p = .03).Inclusion criteria were: German language proficiency, cohabiting with a spouse, being a first-time parent, and carrying a single embryo to ensure compatibility and reduce disruptive factors.Severe mental health issues (e.g.diagnosis of major depression or psychosis) were exclusion criteria.

Study design
This study follows a longitudinal within-subjects design.Parents were recruited at birth information events of local clinics, by sharing study information flyers with gynecologists and midwives, as well as using social media advertisement.Participation was voluntary and expenses were monetary compensated.This sample is part of a larger longitudinal study 1 (Basic Trust Study) following expectant parents from pregnancy to 18 months postpartum during five timepoints.This current study focuses on study baseline, which was during pregnancy (T1) and the follow-up assessment at about six months after birth (T2).Data were collected between 2019 and 2022 at the Institute for Early Life Care at the Paracelsus Medical University in Salzburg/Austria.At T1, adult attachment representation as well as general RF were assessed with the Adult Attachment Interview (AAI; George et al., 1996).At T2, paternal RF was measured with the Parental Development Interview (PDI-R; Slade et al., 2022).Trained research staff conducted all interviews.Furthermore, 20 minutes of free-play father-child interactions were videotaped in the 360-degree video lab.Demographic information and questionnaires were collected in an online format using Lime Survey (Schmitz, 2020).The ethics committee of the state of Salzburg (415-E/2217/8-2017) approved the study.

Adult Attachment Interview (AAI)
Parental attachment representations were assessed at T1 using the Adult Attachment Interview (AAI; George et al., 1996).The semi-structured interview consists of 20 questions in which individuals report on their childhood experiences and how these may influence their lives today.The interviews lasted between 45 minutes and 2 hours and were audiorecorded.Analyses were based on verbatim transcripts of the audio files.They were systematically scored and classified by a reliable coder, according to the adult attachment classification rating system (Main & Goldwyn, 1994) as secure-autonomous (F), insecuredismissing (Ds), insecure-preoccupied (E), and unresolved (U).Psychometric properties were described as excellent in several studies (Ravitz et al., 2010).In our study, dismissing (25%), preoccupied (2.5%) and unresolved (5%) attachment representations were merged into one single insecure group due to an insufficient number of cases in each category.We also assessed fathers' general RF according to the Reflective Functioning Manual for Application to the Adult Attachment Interview (Fonagy et al., 1998;Steele & Steele, 2008).Scores are provided on a 11-point scale ranging from −1 (representing negative RF) to 9 (exceptional RF).Scores below 3 are indicating low RF, scores between 4 and 6 moderate, and scores equal or greater than 7 high RF.AAIs were coded and double-coded by two independent and reliable coders.Interrater reliability for the RF scale was calculated for 20% of the interviews with intraclass correlation coefficient (ICC) of .73.Both coders were blind to sample characteristics, time points, and research questions.In case of disagreement, consensus ratings were used for the analyses.

Parent Development Interview (PDI)
Parental RF was assessed at T2 using the Parent Development Interview (PDI-R; Slade et al., 2022).Parents were asked standardized questions about their relationship with their child.They were also asked about their feelings and thoughts as well as those of their child in various situations.The interviews lasted between 45 minutes and 1.5 hours and were audiotaped.The evaluation is based on transcripts and was performed by a reliable coder, who established reliability with Arietta Slade and was blinded to all other study variables apart from child age.During the analysis, each of the 15 demand questions of the interview is scored individually and an overall score is assigned.RF scores are provided on a 11-point scale ranging from −1 which represents negative RF to 9 (full RF).Scores below 3 indicate low RF, scores between 4 and 6 moderate and scores equal or greater than 7 indicating high RF (Fonagy et al., 1998).Internal consistency is considered high (Cronbach's alpha = .91),indicating that the score is a good representation of the RF ability of the parent (Sleed et al., 2020).

Emotional Availability Scales (EAS)
Paternal sensitivity was measured with the Emotional Availability Scale, version 4 (EAS; Biringen, 2008).Videotaped play interactions between caregiver and child were systematically analyzed by a reliable rater to assess six dimensions of emotional availability, however, only the scale assessing sensitivity was used in this study.A maximum of seven points can be assigned to that scale whereas lower scores imply lower sensitivity.Authors report acceptable reliability with intra-class correlations (ICCs) ranging from .79 to .92 for the six scales (Biringen et al., 2014).Interrater reliability was calculated for 20% of the videos with intraclass correlation coefficient (ICC) of .69.Both raters established reliability on the measure with Zeynep Biringen and were blinded to all other study variables apart from child age.Disagreements between the two raters on three videos were resolved by using a final rating of an independent master coder.

Data analysis
Descriptive statistics are reported in terms of sample characteristics.To evaluate the associations between general (measured with AAI) as well as parental RF (measured with PDI) and paternal sensitivity (measured with EAS), Pearson-and Spearman correlations and linear regressions were conducted.Independent sample t-tests were used to analyze differences in paternal sensitivity depending on secure vs. insecure paternal attachment representations.To examine the mediating effect of paternal RF on the association between attachment representation and sensitivity, we used Hayes PROCESS version 4.0 model 4 (Hayes, 2022).Effect sizes and 95% confidence intervals were reported.All conducted tests were two-sided, and p-values <0.05 were considered statistically significant.All statistical analyses were performed by use of SPSS 27.MPlus software was used to calculate the model fit of the mediation (Muthén & Muthén, 2017).

Results
Preliminary analyses: No associations between the paternal demographic characteristics (e.g.paternal and child age, gender, or paternal educational attainment) and any of the relevant variables were found (p > .05for all variables).Therefore, they were not included in further analyses.Table 2 shows means, standard deviations, and correlations for all study variables, revealing correlations in the expected direction between sensitivity and general RF (RF_AAI) as well as parental RF (RF_PDI).

H1: Associations between fathers' RF and sensitivity
Linear regression models were applied to examine whether RF predicted sensitivity.Parental RF (measured via the PDI) significantly predicted fathers' sensitivity (F(1, 38) = 25.54;p < .001),with the model explaining 40.2% of the variance (R 2 = .402).A second regression model showed that general RF -prenatally measured with the AAI -was also significantly related to paternal sensitivity (F(1, 38) = 5.57; p = .024),explaining 12.8% of the variance (R 2 = .128).Detailed examination of all components of RF (RF_PDI, RF_AAI) in one multiple regression model reveals RF_PDI being the key predictor of paternal sensitivity (see Table 3).

H2: Association between paternal attachment representations and sensitivity
We compared sensitivity levels between fathers with secure and insecure attachment representations.T-tests revealed that fathers with secure attachment representations (M = 5.2; SD = 1.2) presented significantly higher levels of sensitivity than those with insecure attachment representations (M = 4.4; SD = 0.9; t(38) = −2.23;95% CI[−1.60;−.08], p = .032).A mediation model tested whether associations between attachment representations and paternal sensitivity could be explained by paternal RF (RF_PDI).The overall model fit the data well, χ2(0) = 0.00, p < .001,CFI = 1.00,TLI = 1.00,RMSEA = .00.Full model results and path effects are shown in Figure 1.A total effect (c path) of paternal attachment representation on paternal sensitivity was observed (c = 0.84; p = .023).After entering paternal RF_PDI as a mediator into the model, adult attachment representations predicted the mediator significantly (direct effect path a = 2.00, p < .001),indicating an association between secure attachment representations and higher levels of paternal RF_PDI.Paternal RF_PDI in turn predicted paternal sensitivity significantly (direct effect path b = 0.44, p < .001)and the relation between paternal attachment representation and paternal sensitivity appears to be fully mediated by paternal RF_PDI (indirect effect ab = 0.87, 95% CI[0.35; 1.59]), which is in line with our hypothesis.

Discussion
Given fathers' increased role in childrearing (Bakermans-Kranenburg et al., 2019), the aim of this study was to examine how paternal attachment representations and reflective functioning were related to paternal sensitivity.We postulated high prenatal general and postnatal parental RF to be associated with high sensitivity during fathers' interactions with their 6-month-old child.In addition, we hypothesized that fathers with secure  attachment representations would be more sensitive when interacting with their children and that the association between fathers' attachment representations and paternal sensitivity would be mediated by paternal RF.Our results indicate that, first, both general RF, measured prenatally with the AAI and parental RF, assessed postnatally with the PDI, significantly predicted the degree of paternal sensitivity during interactions with the child at 6 months of age.Second, we found securely attached fathers to be significantly more sensitive with their children than fathers with insecure attachment representations.Furthermore, a mediation analysis revealed that the association between attachment representation and sensitivity is explained by fathers' parental RF.Whereas a significant total effect of attachment representation on sensitivity was detected, the association was no longer significant after adding parental RF as a mediator.Hence, parental RF fully mediates this relationship.
Overall, our findings support the hypothesis that secure attachment representations are important in the development of mentalizing, which in turn can facilitate paternal sensitivity.Although some studies have already explored the association between maternal attachment, RF, and maternal sensitivity, only a few have considered the same questions in fathers, and none of those have used the AAI to assess the paternal attachment representations (Esbjørn et al., 2013;Nijssens et al., 2020).Thus, to our knowledge, this is the first study to disentangle the association between attachment representations, RF, and sensitivity in fathers.Despite a comparatively small sample, we were able to extend the interaction mechanisms already established in mothers to a sample of fathers using non-self-report measures such as interviews and video observation.

Fathers' RF and sensitivity
The significant association between general and parental RF and sensitivity in our data is in line with the associations reported in the literature (e.g.Smaling et al., 2016;Stacks et al., 2014).Since most of the research predominantly focuses on mothers, associations between RF and sensitivity still need to be examined for fathers (e.g.Buttitta et al., 2019;Stacks et al., 2014).Our results contribute to filling this knowledge gap by showing that fathers with high RF interact more sensitively with their children.In line with other studies (e.g.Buttitta et al., 2019;Kelly et al., 2005), we assumed that this association is most likely present because the ability to mentalize allows a father to consider his child as an individual with own desires, feelings, and goals.This, in turn, enables him to recognize the child signals accurately and to respond to them in an appropriate way.
In line with our expectations, general RF measured with the AAI did significantly predict paternal sensitivity.Compared to paternal RF, however, only a smaller amount of variance could be explained by general RF.This may be because the AAI was not originally designed to predict parenting behavior and may therefore measure different facets of RF that are less powerful in predicting sensitive parenting behavior compared with the PDI.The AAI assesses an individual's capacity to truthfully and coherently reflect upon childhood memories and on how early relationship experiences, either positive or negative, with primary caregivers may have influenced one's representations.Although reflective functioning can be derived from this (Fonagy et al., 1998), these aspects of RF may not be as important in determining one's own behavior as a parent.Rather, it might be the ability to reflect properly on the present and the developing relationship between oneself and one's child that is more critical when predicting parental sensitivity (Slade et al., 2005).Indeed, this suggestion for a more direct way of assessing RF in the context of parenting is what originally motivated the development of the construct parental RF in the first place (Sharp & Fonagy, 2008;Slade, 2005).Our results indicate that one's general RF is important for the development of sensitive father-child interactions, but not to the same extent as parental RF.In fact, the multiple regression model that accounts for both components of RF demonstrates that parental RF is the key predictor of paternal sensitivity and general RF does not explain any additional amount of variation.An alternative explanation for this finding could be the timing of the interviews.There was no possibility to capture the influence of interactive experiences with own children on the fathers' states of mind related to attachment experiences, because the AAI was administered during pregnancy.This may limit the predictability of general RF as the presence of a baby may influence parental mentalizing skills.Since babies cannot yet articulate their needs, they rely on a caregiver to imaginatively and intuitively work out what the child may need.This could encourage parents to reflect on their child's inner world (e.g.desires, goals).However, since the PDI, which focuses on the current parent-child relationship, cannot be administered before birth, RF measures with the AAI was used as it is a valid and reliable measure of general RF (Fonagy et al., 1998), which has previously been linked to parenting sensitivity and child attachment (e.g.Ensink et al., 2016;Fonagy et al., 1991).Therefore, our findings are still particularly important when considering the potential for preventive support that early insights can provide.Thus, expectant parents scoring low on the RF-scale applied to the AAI prenatally could receive early, targeted mentalizing training to support sensitive parenting and thereby promote the child's physical and emotional development.

Fathers' attachment representations and sensitivity
As expected, we found an association between attachment representations and paternal sensitivity.Although the literature postulates such associations for both parents, past research has almost exclusively focused on mothers, with only a handful of studies having investigated fathers' attachment representations to this day (Adam et al., 2004;Verhage et al., 2016).The few studies that have included fathers show inconsistent results, even though the same measures of attachment were used (Aytuglu & Brown, 2021;McFarland-Piazza et al., 2012).Our study is consistent with the extensive literature on mothers and the findings of McFarland-Piazza and colleagues, providing evidence that a father's state of mind with respect to early experience with his own parents has a significant influence on the level of sensitivity in interaction with his child.

The mediating effect of parental RF
Despite the relatively small sample size, we found that the association between paternal attachment representations and paternal sensitivity is mediated by parental RF.This result is consistent with other findings that identified parental RF mediating associations between attachment, caregiving, and child outcomes (e.g.Alvarez-Monjarás et al., 2019;Dollberg, 2021;Nijssens et al., 2020;Slade et al., 2005;Smaling et al., 2016).
Our findings support the prevalent assumption that parental mentalizing is closely related to one's attachment representations (Luyten et al., 2020).The literature suggests this is due to the fact that the capacity for RF is first, though not exclusively, formed through early experiences of relating to one's own attachment figures (Fonagy & Target, 1997).While there is a link between mentalizing ability and the coherence of reflection on early relationship experiences, attachment representations are not the determining factor when it comes to paternal sensitivity.In fact, even if shaped by early relationship experiences, the ability to mentalize can also be learned, trained, and further developed later in life (e.g.Letourneau et al., 2020;Luyten et al., 2020;Zayde et al., 2021).
When parents are able to take the child's perspective, they are more likely to respond sensitively to the child's signals in the first place (van Bakel & Hall, 2020;Zeegers et al., 2017).Therefore, in addition to other attachment based interventions, such as Video-Feedback Intervention to Promote Positive Parenting (VIPP) or Attachment and Biobehavioral Catch-Up (ABC), which have already been shown to be effective in supporting parental sensitivity (e.g.Bick & Dozier, 2013;Juffer et al., 2018), promoting effective mentalizing skills as part of family counselling or therapy may be additionally important for achieving higher parenting quality and positive child development outcomes.One such program that already follows this idea is the Minding the Baby TM program.Starting during pregnancy, parents are trained to reflect on their own and the baby's experiences in order to recognize and respond to the baby's needs and to better understand their own feelings as parents.Studies evaluating the effectiveness of the program have shown increases in maternal RF capacity, child-caregiver attachment security (Slade et al., 2020), and long-term positive effects on parenting and child behavior (Londono Tobon et al., 2022).In addition, other studies of interventions targeting parental RF have demonstrated improvements in RF skills as well as positive outcomes in child development (e.g.Gershy et al., 2023;Lo & Wong, 2022;Menashe-Grinberg et al., 2022;Stacks et al., 2022;Suchman et al., 2020).To our knowledge, these programs have only been implemented, tested, and evaluated with mothers.Therefore, our findings of an association between RF and paternal behavior suggest that fathers may also benefit from such programs.

Contributions and limitations
This longitudinal study included different non-self-report measurements such as interviews and observation to assess distinct facets of the variables of interest.No variable was measured by self-report due to the assumption that exclusive use of self-report provides only little to modest information and cannot be compared to the quality of interviews, observations, or a mixed measures approach (e.g.Salo et al., 2021).As such, this study expands our understanding of the factors contributing to paternal sensitive behavior.A limiting factor, however, is still the small sample size of 40 fathers.It should be noted that longitudinal, multi-measure studies typically have smaller sample sizes given the cost and labor associated with this type of study.In addition to these challenges, (expectant) fathers are particularly difficult to recruit, which in turn may explain the smaller sample size.Nevertheless, an a posteriori power analysis of our primary hypothesis revealed that -with a type I error of 5% and a sample size of n = 40 -the study ends up with 99.9% to find a significant relation (slope = 0.431, standard deviation of 1.72 of RF and standard deviation of 1.17 of sensitivity) between parental RF and paternal sensitivity.Therefore, the fact that the expected associations were found even in this relatively small sample highlights the robustness of the detected effects.
Furthermore, the relatively homogeneous sample, in terms of a high socioeconomic status and low-risk, cohabitating, heterosexual volunteering families can be considered as limitation.These characteristics may restrict the generalizability of the results.Yet, even within this relatively well-educated low-risk group, variations in RF were visible.
Moreover, this study did not investigate the influence of mothers, who could potentially moderate the effects identified.Paternal attachment representations might be changed by the (new) relationship and interaction in the father's partnership with the child's mother, which could in turn affect his attachment representations.A mother with high RF may serve as role model for the father.If the mother has high mentalizing capacities, the entire family system may profit from these reflective insights and paternal RF might increase as well.Further research should therefore consider the family as a whole system mutually affecting one another (Cabrera et al., 2018;Cox & Paley, 2003;Lindsey & Caldera, 2006).
Finally, our recruitment strategy of only selecting first-time parents may have affected the results.Although this strategy also represents a strength of the study, as it allows to control for contributing factors on influences on parenting experience and family dynamics (e.g.no siblings yet, no divorce-setting), it limits the generalizability to other family configurations such as single parenthood, multiple siblings, and multiple caregivers.

Future research directions and clinical implications
The data from this study show associations between paternal attachment representations, reflective functioning, and sensitive paternal behavior.Critically, future research should investigate whether the predictive power of attachment representations and RF can be extended to child outcomes.Children in the current sample were only six months old, making it difficult to examine long-term benefits for children in the current design.Further research is necessary to examine long term child outcomes, including the quality of child-father attachment relationships.Furthermore, an investigation with a larger sample is needed, while maintaining the quality of measures at the same level ensuring validity of the data to allow generalization to the overall population of parents.
In a similar vein, future studies should include children of different ages, examining these associations.Research already showed that child age might moderate parenting and child outcomes (Rodrigues et al., 2021), such that higher child age is associated with increased paternal sensitivity in some contexts (e.g.Muzard et al., 2021).In addition, there are concerns about whether gender effects emerge with increasing age.Several studies suggests that the child's gender does not have a significant influence on parental sensitivity in interaction (e.g.Hallers-Haalboom et al., 2014;Rodrigues et al., 2021).However, some studies have found evidence of gender effects on parenting behavior for children over the age of 12 months (e.g.Boeldt et al., 2012;Schoppe-Sullivan et al., 2006;Zvara et al., 2018).Therefore, a replication of the study with older children may also reveal gender effects that were not apparent in this study due to the young age of the children in our sample.
In addition, future research might investigate families from higher risk contexts (e.g.clinical sample, mental health problems, low SES) as mechanisms of action often work differently when families are exposed to multiple stress factors (Camoirano, 2017;Stacks et al., 2014).
Together, these findings have implications for clinical interventions, particularly for fathers with low RF.Early interventions focusing on fathers' parental RF by helping them to understand his child's behavior based on mental states may provide an important preventive function with regard to parental sensitivity, which is in turn a key predictor of a child's socialemotional and cognitive development (Cooke et al., 2022;Rodrigues et al., 2021).Screening fathers' RF already during pregnancy or shortly after birth may allow for preventive efforts that support families as early as possible.Based on this screening, targeted interventions adapted to the needs of the parents can be implemented, which, for example, specifically strengthen the ability to mentalize and therefore increase paternal sensitivity.

Table 1 .
Descriptive statistics of demographics for fathers and children.
Note.M = mean; SD = standard deviation.Child age in months, father's age in years.1Insecureattachment comprises both avoidant and preoccupied fathers. 2 One month unpaid additional leave for fathers after childbirth.

Table 2 .
Descriptive statistics of the study variables with means, standard deviations, and correlations.

Table 3 .
Summary of linear multiple regression to predict paternal sensitivity.