The focus of this paper is on the relationship a woman develops with her unborn child during pregnancy (herewith referred to as maternal-fetal relationships, or MFR). Individual differences in the way mothers conceptualize their relationship with their unborn child have been associated with important outcomes including mental health, well-being, and health practices in pregnancy [1], postpartum parent-infant interaction [2], infant mood [3], and child disorganized attachment, itself a risk factor for psychopathology and poor socio-economic outcome [4],[5]. Thus, understanding MFR appears vital in understanding adaptation in pregnancy and beyond [6],[7], and identifying difficulties with MFR could guide risk-assessment and intervention [1]. But efforts to augment the maternal-fetal relationship have not always been successful [8]-[10] and some urge caution in trying to "interfere" in the maternal-fetal relationship because we know little about its nature and development [1]. The present study attempts to clarify the nature and predictors of MFR.

Research has found MFR to be associated with security, support, and satisfaction in parental and partner relationships, implying that representations of other relationships are important in MFR [11]-[15]. For example, Mikulincer and Florian [16] found that romantic attachment security in adult relationships was associated with quality of MFR and mental health throughout pregnancy. Some studies, though, have found no link between couple or family relationships and MFR [17],[18], and even when associations are found, the underlying mechanisms remain unclear. We suggest that these inconsistencies reflect two core methodological and theoretical issues, which we address in the present study.

First, few MFR studies have used dimensional romantic attachment measures. Some earlier research [16] used attachment-style categories (secure, avoidant, anxious-ambivalent), but it is now accepted that individual differences in adult romantic attachment are best captured along two continuous dimensions of attachment avoidance and attachment anxiety [19],[20]. Those high on attachment avoidance are concerned with independence, uncomfortable with intimacy, and deactivate the attachment system. Those high on attachment anxiety are preoccupied with relationships, concerned about rejection, and hyperactivate the attachment system [20]. Studies have found that attachment avoidance, but not always attachment anxiety, negatively relates to desire to have children before, during, and after pregnancy [21]-[23]. Therefore, the two dimensions of romantic attachment may relate to MFR via different mechanisms, which have been obscured in studies using attachment categories or general relationship security or positivity concepts.

Second, we propose that MFR, though often conceptualized as, and referred to as antenatal 'attachment' [24], is not indicative of the attachment system. A lack of conceptual clarity — attested by widely-varying definitions — can inhibit understanding of maternal-fetal relationships [25]-[27]. Various theoretical definitions and associated measures of the parental-fetal bond have emphasized different elements. For example, Cranley's Maternal-Fetal Attachment Scale [24] measures engagement in behaviors of affiliation and interaction with the fetus. Condon's Maternal Antenatal Attachment Scale [28] considers love for the fetus as the central experience and includes maternal dispositions towards the fetus of knowing, being with, protecting, gratifying needs, and avoiding loss [28]. Overarchingly, protection appears to be a key component in many definitions [26],[29]-[31].

Most of the work looking at the emotional tie between parents and their unborn children is rooted in attachment theory [32], but this theory states that the attachment system involves seeking care from someone who can provide comfort and protection [32],[33], whereas the caregiving system involves attending to need and providing protection [32],[34]. Arguably, the mother's antenatal relationship with her child is not attachment — parents do not (usually) seek care from their unborn children — but may be more akin to the beginnings of caregiving representations [29],[30],[35]. In an adult relationship, attachment security, especially low avoidance, facilitates responsive caregiving in the couple domain [36]-[38], and also the parenting domain [39],[40]. We argue that this same caregiving system may also inform the earliest representation of a mother's relationship with the child.

We aim to examine predictors of MFR, testing the hypothesis that individual differences in romantic attachment indirectly predict MFR via (i.e., are mediated by) the caregiving system. We use attachment and caregiving variables that are both assessed at the romantic level, allowing us to test which element of a couple's relationship impacts most on maternal relationships with the unborn child. We also investigate and control for the influence of other factors shown to be linked with MFR in previous research: pregnancy-specific anxiety [41]-[43], mental health [17], gestational age [27] and parity [17]. In so doing, we intend to provide a more comprehensive account of the interpersonal and intrapersonal antecedents and correlates of MFR.

Previous research has suggested that secure and positive adult romantic relationships are associated with quality of maternal-fetal relationships, but findings have been inconsistent, and the underlying mechanisms have not been clear. We proposed that responsive caregiving is a previously untested predictor of MFR and a key mediator in its link with adult relationships. Supporting this model, we found that the association between individual differences in romantic attachment avoidance and MFR was fully mediated by caregiving responsiveness to partner. This pattern was stable when considering other variables relevant to MFR: mental health, pregnancy anxiety, gestational age, and parity. Mental health was also a significant predictor of MFR, but part of its link was also accounted for by caregiving responsiveness to partner. Thus, our data demonstrate for the first time that mothers' MFR is robustly predicted by responsive caregiving to partner, and support the hypothesis that MFR better reflects the operation of the caregiving system than the care-seeking (i.e., attachment) system.

Extant literature suggests that attachment insecurity may cause difficulties both in parents providing care for their offspring [56] and romantic couples providing care for their partners [57]. But the constraints on effective caregiving may differ depending on the specific attachment insecurity. In particular, we found a significant indirect pathway from romantic attachment avoidance, but not attachment anxiety, to MFR via caregiving. Mikulincer and Shaver [20] propose that caregiving deficits in adults with high avoidance stem from reduced empathic-concern for others' needs. Indeed, high avoidance is associated with deficiencies in sensitive and responsive caregiving, mediated by lack of knowledge about how to provide support, low prosocial orientation, and low relationship commitment and intimacy [36]. It would be fruitful to examine these concepts more thoroughly in the antenatal period to pinpoint directions for providing support or intervention.

We did not find pathways from romantic attachment anxiety to low caregiving responsiveness or MFR. Past findings have often reported negative links between women's attachment anxiety and caregiving variables [36],[55],[58] but not always [38]. Attachment anxiety is thought to inhibit responsive caregiving because the individual focuses on their own distress and attachment needs over the needs of others [36],[58], and may lead to intrusiveness or overinvolvement by "…(coloring) caregiving motives with egoistic desires for acceptance, approval and gratitude" [55] (p.44). It may be that when controlling for their general mental health, highly attachment-anxious women are no longer inhibited by such distress or needs; moreover, they have qualities which could support caregiving, such as comfort with emotional expression and intimacy [55]. Alternatively, features of pregnancy or the antenatal relationship may prevent attachment anxiety from reducing caregiving (and thus MFR). For example, pregnancy involves little overt care-seeking from the fetus, reducing the disruptive effects of another's expressed needs or fear of rejection, especially in relatively low-risk pregnancies (as these were). Future research could further examine whether something about being pregnant may also "spill over" into caregiving representations in the couple relationship, perhaps with altered reciprocity or expectations thereof.

Past research has found that mental health, pregnancy anxiety, gestational age, and parity predict MFR. We uncovered interesting patterns when examining these variables alongside relationship variables. First, poorer mental health related to higher levels of romantic attachment insecurity and pregnancy anxiety. This supports consistent evidence linking attachment anxiety to psychological distress [20], and recent evidence that high-demand conditions - of which pregnancy could be one - also elicit distress in highly-avoidant individuals [59],[60]. But the correlations between attachment insecurity and pregnancy anxiety became non-significant controlling for mental health, implying that insecure women's pregnancy anxiety reflects their overall wellbeing and not relationship-specific anxieties. Moreover, controlling for caregiving responsiveness to partner and mental health, pregnancy anxiety did not predict MFR, indicating that worries about pregnancy do not necessarily inhibit forming a connection with one's unborn child. Overall, the model suggests that pregnancy anxiety and MFR are two separate factors, the former best predicted by mental health, and the latter also predicted by responsive caregiving to partner, but neither best predicted by romantic attachment. Psychological wellbeing, then, is a relevant factor to consider when examining both elements of the pregnancy experience.

Our multiple-group models, consistent with past research, found that MFR was higher at 23 and 33-weeks than 13-weeks gestation, and that both MFR and pregnancy anxiety were higher among women in their first pregnancy than those in subsequent pregnancies [27],[61]. Thus, first-time pregnant women may benefit from help managing their anxieties but appear to form particularly strong antenatal relationships with their child, mitigating against possible consequences. Interestingly, we also found that women in their first (vs. subsequent) pregnancy reported higher mental health and lower romantic attachment avoidance. This pattern was not hypothesized, but is consistent with evidence that the transition to parenthood can involve stressful change, impacting wellbeing and relationship satisfaction [62]. It may also be a function of our sample, in that more avoidant or depressed women may engage relatively less in such online forums in their first pregnancy (or more in subsequent pregnancies). Further research exploring relationship processes in first and subsequent pregnancies is needed. Crucially, our key findings that caregiving to partner and mental health best predict MFR were stable controlling for these additional important factors.

We are mindful of the limitations of our study, which should be addressed in future work. Our sample comprised women registered with an online support provider, and was predominantly middle-class and reasonably well-resourced, indicative of users of this service. The processes we have identified may differ for higher-risk samples, especially those who lack social support, have significant mental health problems, or are experiencing difficult pregnancies. Because of our research questions, we recruited women with partners. This may be one reason that romantic attachment avoidance and anxiety were highly correlated: other research has found higher correlations between attachment dimensions in longer-term couples than other samples [38]. Further research should recruit women without partners; although we linked MFR to responsive caregiving to partner, it should also relate to global caregiving representations (for which new measures have been developed since the present study [63]), or caregiving towards other children. To build on our self-report findings, which focused on cognitive and affective representations of relationship processes, future research may develop behavioral measures, such as emotional facial expression when talking about the baby. Our research was cross-sectional and focused on mothers, albeit in all stages of pregnancy. Future research would benefit from longitudinal and dyadic designs following couples through pregnancy and into the postnatal period to assess whether MFR really does have long term implications for parenting. Such research could also explore other possible paths to MFR, including other characteristics of intimate relationships aside from romantic attachment dimensions.

Our findings support and extend the literature on MFR and transition to parenthood. Our model may go some way to clarifying our understanding of MFR and explaining past inconsistent associations with relationship variables. We suggest that MFR is driven more by individual differences in the caregiving system than the attachment system, alongside psychological functioning. It may therefore be prudent to focus antenatal education and intervention efforts on augmenting caregiving responsiveness, especially for women high in attachment avoidance. We are not claiming that maternal-fetal relationships are synonymous with responsive caregiving: their definition needs further exploration and clarification [64], but our findings provide a direction of focus for elucidating the mechanisms underlying the established associations between couple relationships and parenting [65]-[67]. Thinking about MFR in terms of the caregiving system may help to clarify past findings and to guide future research and interventions.