Great Expectations? Do Mothers’ and Fathers’ Prenatal Thoughts and Feelings about the Infant Predict Parent-Infant Interaction Quality? A Meta-Analytic Review Sarah Foley1 & Claire Hughes Centre for Family Research, University of Cambridge, Free School Lane, Cambridge, CB2 3RQ 1Corresponding author. Email: sf412@cam.ac.uk Abstract Drawing on data gathered from 14 studies involving a total of 1,862 mothers and fathers, this meta-analysis reviews the measures that are used to tap into thoughts and feelings about the unborn infant during pregnancy and examines links between these prenatal measures and parent-child interaction quality. Questionnaire scores for parental-fetal attachment and interview ratings of expectant parents’ representations of their infant showed modest but robust associations with observed parent-child interaction quality. Moderator analyses showed that these associations were significantly stronger for mothers than for fathers. Key lessons for future research include the need for greater consistency in study measures, sample diversity and the examination of associations with child outcomes. Keywords: pregnancy; parenting; observation; representation; parental-fetal attachment Attachment, defined as infants’ adaptive propensity to seek comfort and protection from a familiar caregiver when they are alarmed or unwell, has long been recognized as an important predictor of children’s development and adjustment (Bowlby, 1969; Sroufe, 2005). Using standardized observational paradigms such as Ainsworth’s strange situation (1970), researchers have identified three main profiles of infant attachment: secure (e.g., seeks and accepts comfort), insecure-resistant (e.g., seeks yet resists comfort) or insecure-avoidant (e.g., absence of help-seeking behaviour). Variation in parental sensitivity, defined as the ability to notice, interpret and respond in a timely and appropriate manner to children’s signals (Ainsworth, Bell, & Stayton, 1974), is thought to underpin these contrasts in infant attachment. In contrast, parental displays of frightening or atypical behaviour are thought to lead to a pattern of disorganised attachment in a fourth group of children (e.g., Duschinsky, 2018; van Ijzendoorn, Schuengel, & Bakermans-Kranenburg, 1999). In a further expansion of the field, researchers have also examined parents’ representations of their infants, with a particular focus on two related constructs: (i) mind-mindedness, which refers to parents’ propensity to see their children as individuals with their own minds (Meins, 1997); and (ii) and reflective function, which refers to parents’ ability to reflect upon the internal states that govern their own and their children’s behaviour (Fonagy, Steele, Steele, Moran, & Higgitt, 1991). Importantly, this shift in focus from parent behaviour to parental representations enables predictors of parent-child relationship quality to be examined from as early as the prenatal period. This is an exciting prospect for fostering infant development and adjustment because pregnancy provides a valuable window of opportunity for delivering programmes to support new parents (Sadler et al., 2013). With this goal in mind, researchers have used diverse interviews and questionnaires to examine whether expectant parents’ thoughts and feelings about their unborn infant might contribute to early variation in parental sensitivity. However, the findings from this field have yet to be synthesised. To address this gap in the current study, we first summarise the different interview and questionnaire measures available to tap into expectant parents’ thoughts and feelings about the unborn infant. Next, we outline some of the observational parenting measures used in studies tracking participants across the transition to parenthood. Following this, we aim to provide a quantitative summary of effects from studies adopting similar methodologies and possible moderation effects (Rosenthal, 1984) by conducting a meta-analysis of the strength of associations between prenatal thoughts and feelings about the unborn infant and observed parenting quality in the first years of life. Assessing Expectant Parents’ Thoughts and Feelings about Their Unborn Infants Fonagy and colleagues (1993) found that expectant mothers’ representations of their own caregivers predicted their infants’ attachment security at age 12 and 18 months. Building on this seminal study, other investigators (e.g., Korja et al., 2010) have examined whether parents’ representations of their children (rather than their own caregivers) might provide a more direct predictor of the quality of parent-child interactions and child outcomes. To this end, researchers have developed standardised interviews, such as the Working Model of the Child Interview (WMCI; Zeanah, Benoit, & Barton, 1986) and the Parent Development Interview (PDI; Slade, Grienenberger, Bernbach, Levy, & Locker, 2002). Here we focus on the WMCI as an adapted version of this interview has been used with expectant parents (Benoit, Parker, & Zeanah, 1997). In the WMCI the interviewer asks parents to draw on past and present experiences of their child’s behaviour and personality to describe their thoughts and feelings and expectations of the child in the future. The content of these narratives is then coded according to several distinct dimensions including: richness of parental perceptions, openness to change, intensity of their involvement, coherence, sensitivity during caregiving and acceptance. Researchers also code the predominant affective tone of the representation and the extent to which the narrative is critical or worried. Together, these content- and affect-based coding systems enable researchers to categorise parents’ representations as balanced, disengaged, or distorted, as described below. Parents who demonstrate clear enjoyment of their relationship with the child, coupled with an appreciation of the child as a distinct individual and a recognition that their views of the child may change over time, are classified as having a balanced representation of their child. In contrast, parents who convey a lack of both emotion and content and suggest that they do not see themselves as playing an influential role in their children’s development are classified as having a disengaged representation of their child. Parents who offer incomplete or inconsistent descriptions of the child or the parent-child relationship, use contradictory or confusing statements or show either a narrow focus on their own concerns or idealized descriptions of their children are classified as having a distorted representation of their child. Typically, these three groups map on to the three main infant attachment classifications. That is, infants of parents with balanced representations are typically securely attached, whilst infants of parents with disengaged representations typically show avoidant attachment and infants of parents with distorted representations typically show resistant attachment. Importantly, parent behaviour is thought to underpin this mapping between parental representations and infant attachment. For example, parents with a balanced representation by definition see their infants as individuals with distinct and changeable experiences and so are more likely than other parents to act in a sensitive manner (e.g., by looking for cues about their infants’ intentions and feelings). Conversely, parents who provide disengaged representations of their infants appear disinclined to attend to their infants’ cues and parents who provide distorted representations of their infants find it difficult to read or respond to their infants’ cues (Korja et al., 2010; Schechter et al., 2006; Sokolowski, Hans, Bernstein, & Cox, 2007). In a study that adapted the WMCI for use in pregnancy, Benoit et al. (1997) found mothers’ representations were typically stable over the transition to parenthood. For example, 89% of mothers with balanced prenatal representations also provided balanced representations during postnatal interviews. Likewise, 85% of mothers with unbalanced prenatal representations also showed unbalanced representations at 12 months. Expectant parents’ representations of their unborn infant were also meaningfully related to attachment classification from the strange situation at 12 months (i.e., 91% of mothers with balanced representations had infants who were classified as securely attached). Changes from unbalanced representations during pregnancy to postnatal balanced representations were associated with lower levels of depression, a stable partner relationship and higher family incomes (Vreeswijk, Maas, & van Bakel, 2012). Another research tool that has been used to elicit parents’ representation of children is the Five-Minute Speech Sample (FMSS), a measure that is both much quicker to administer and code than the WMCI and widely used in the clinical literature (Sher-Censor, 2015; Weston, Hawes, & Pasalich, 2017). Traditionally used in adult psychopathology research (Magana et al., 1986), researchers are increasingly using the FMSS to measure parent-child dynamics across a variety of age ranges, including expressed emotion (Magana et al., 1986), parent attributions (Bullock & Dishion, 2004), warmth and criticism (Caspi et al., 2004; Daley, Sonuga-Barke, & Thompson, 2003). Indeed, capitalizing on the potential of the FMSS to substitute for the WMCI, Sher-Censor and Yates (2010) adapted the WMCI coding scheme for use with the FMSS. Weston et al. (2017) examined coding schemes that focus on four different dimensions from the FMSS to assess the strength of their association with observed parent behaviours and parent-child interaction quality. The most widely used of these four dimensions was expressed emotion (k = 12), which assesses the extent to which parental talk is overtly critical, hostile and indicative of an emotionally over-involved relationship. Associations with observed positive and negative parent behaviours ranged in s strength from small to large but were mainly restricted to mid- to late-childhood (Weston et al., 2017). This is a significant gap as in the first few years of life ‘emotional over-involvement’ may be normative rather than pathological (Peris & Miklowitz, 2015). Thus, infancy studies that adopt the FMSS should focus on constructs that are developmentally appropriate to this period. Importantly, the simplicity and open-ended nature of the FMSS means that it can be easily adapted to the prenatal setting and so offers a less time-consuming alternative to the prenatal version of the WMCI. Given the growth in use of the FMSS, we added the parental constructs commonly assessed with the FMSS (noted above) to the search terms used to identify studies to be included in our meta-analysis. Other researchers have applied questionnaire methods to investigate the maternal-fetal relationship, a concept that has deep roots within the psychological literature. Winnicott (1960, p. 165) noted that during pregnancy the expectant mother “shifts some of her sense of self on to the baby that is growing within her” and highlighted the importance of understanding the psychological changes experienced by women during pregnancy. However, this notion that parents develop an emotional tie with their infant during pregnancy is not without controversy, as illustrated by on-going nosological debates (Walsh, Hepper, Bagge, Wadephul, & Jomeen, 2013). In particular the term prenatal ‘attachment’, with its basis in the behavioural caregiving system, has largely been replaced by reference to the parent-fetal relationship (Walsh, 2010). In a methodological review, van den Bergh and Simons (2009) identified the three most widely used questionnaires in this field: (i) the Maternal Foetal Attachment Scale (MFAS; Cranley, 1981); (ii) the Maternal Antenatal Attachment Scale (MAAS; Condon, 1993); and (iii) the Prenatal Attachment Inventory (PAI; Muller, 1993). Adapted paternal versions are also available for each of these questionnaires. While the questionnaires differ slightly from each other in focus it is worth noting that items that consider: (i) expectant parents’ tendency to interact with and ascribe intentions to the fetus are included in both the MAAS and MFAS; (ii) differentiation between self and the fetus are included in both the MAAS and PAI; and (iii) the presence of positive thoughts and feelings are included in both the MFAS and PAI (for a fuller discussion of the similarities and differences between the measures, see van den Bergh & Simons, 2009) Using the MFAS in a cross-sectional study of 252 women, Lindgren (2001) found that, over and above effects of low education, multiparous pregnancy, and higher levels of depression, a lack of maternal-fetal relationship was associated with fewer positive health practices (e.g., giving up smoking, attending prenatal appointments). More generally however, the maternal-fetal relationship is the outcome of interest rather than a predictor of postnatal outcomes. For example, Yarcheski, Mahon, Yarcheski, Hanks and Cannella (2009) identified 72 studies that highlighted 14 predictors of the maternal-fetal relationship. These include social support, gestational age and prenatal testing, and low levels of anxiety, depression, self-esteem and younger maternal age. Interestingly, parity, high-risk status, income, marital status and whether or not the pregnancy was planned were not systematically associated with the maternal-fetal relationship. Indeed, gestational age was the strongest predictor of the maternal-fetal relationship suggesting that the bond develops over time rather than being particularly susceptible to variation in demographic factors. Measuring Parenting in the Postnatal Period Infancy is a time of rapid development in multiple domains and so parents’ ability to tune into and respond appropriately to infant cues is a particularly salient construct to consider in early childhood. Illustrating the centrality of parental sensitivity for infants’ development (beyond simply considering infant attachment security), recent evidence demonstrates associations with variation in children’s memory, problem solving, language and social skills (Mills-Koonce et al., 2015). It is worth noting, however, that parenting is a multifaceted construct (Grusec & Davidov, 2010). Beyond sensitivity, aspects of parenting that appear influential in early childhood include cognitive stimulation (Vallotton, Mastergeorge, Foster, Decker, & Ayoub, 2017) and low levels of criticism and conflict (Hughes & Devine, 2017). Reflecting this diversity of features within parent-child interactions that contribute to child outcomes, Smetana (2017) has called for: (i) a domain specific approach to studying parenting that recognizes the many different roles that parents play; and (ii) recognition of the dimensional nature of variation in parenting quality. Observational methods are widely viewed as the ‘gold standard’ approach to assessing parent-infant interactions, because they offer both greater objectivity and richer detail. Illustrating this point, in a review of the predictive value of diverse parenting measures on child outcomes, Zaslow et al. (2006) found that the strongest predictor of preschoolers’ co-operation and achievement were video-based ratings of maternal support and reduced hostility, with these predictive relations being independent of both family background and questionnaire measures gathered from either mothers or researchers. In our meta-analysis, we therefore focused on studies that gathered detailed observational ratings of parent behaviour, which we examined in relation to both questionnaire- and interview-based measures of expectant parents’ thoughts and feelings of their unborn baby. Are Expectant Parents’ Thoughts and Feelings about Their Unborn Infant Associated with the Quality of Postnatal Parenting? A substantial body of research has identified parental characteristics as a key determinant of parenting quality (Belsky, 1984). Evidence to support this view comes from findings that parenting shows robust associations with variation in mental-health (Murray, Fearon, & Cooper, 2015), personality (Desjardins, Zelenski, & Coplan, 2008) and efficacy (Barnett, Deng, Mills-Koonce, Willoughby, & Cox, 2008). To date however, researchers have rarely considered prenatal thoughts and feelings about the unborn infant as a predictor of parenting. In one exceptional study, Theran, Levendosky, Bogat and Huth-Bocks (2005) found that, among mothers who shifted in the nature of their representation of their infant across the transition to parenthood (either from balanced to unbalanced or from unbalanced to balanced), those who were able to provide a balanced representation in pregnancy displayed higher levels sensitivity and engagement and reduced levels of hostility in their interactions with their infants 12 months. In another exceptional study, de Cock et al. (2016) followed 372 pregnant women across each trimester of pregnancy and through to 8 weeks post-partum and found that, controlling for demographic and postnatal covariates (e.g., age, breastfeeding problems), scores at each prenatal time point on the MAAS predicted variance in self-reported postnatal attachment. As a caveat however, it is worth noting that shared method variance, resulting from the use of the same self-report instrument in both pregnancy and the postnatal period, may have inflated this association in attachment scores across the transition to parenthood. There is currently a limited evidence base and a lack of consensus with regards to the message that prenatal thoughts and feelings about the unborn infant are vital for the postnatal relationship quality. Yet researchers are often not so careful with their language when discussing this association. As just one example, Rossen et al. (2016) open their paper with the claim that “The mothers’ felt bond to her infant is critical to infant health and wellbeing” (pg. 609, emphasis added) and, in their conclusion, argue that “intervening during pregnancy to promote a healthy bond in the postnatal period when feeding, sleeping and other major routines are being established is important.” (pg. 620, emphasis added). Without a clear evidence base for this view, such causal language is, we believe, premature. Moreover, the causal assumptions implicit in this conclusion suggest that mothers’ feelings towards their unborn infant have a direct impact on infants and so should be subject to intervention. To ensure that such conclusions are properly evidence based, the main aim of the current study was to conduct a meta-analysis to establish the strength of associations between expectant parents’ thoughts and feelings about their unborn infant and the quality of their interactions with the infant in the postnatal period. Moderating Effects of Sample Characteristics and Methodological Contrasts Beyond improving the reliability of study findings, meta-analyses also offer a valuable opportunity to establish whether methodological and sample characteristics influence the strength of associations. As outlined above, investigations of associations between prenatal thoughts and feelings about the unborn infant and postnatal parenting have adopted different designs, used distinct methods and drawn on different samples. As a guide to future research in this field, our meta-analysis therefore also assessed whether these between-study differences moderated the association between pre- and postnatal measures. Compared with prenatal questionnaires, prenatal interviews provide greater depth and detail in information and so we expected interview measures to yield stronger associations with postnatal ratings of parent-infant interaction quality. In particular, unlike questionnaires, interviews go beyond simply asking whether parents consider the fetus to be a person and ask parents to think about specific child attributes. Furthermore, interviews provide a more complex understanding of parents’ cognitions, specifically by capturing thoughts about the child, the parent-child relationship and their belief in parents’ ability to shape child development. In contrast, the questionnaires are focused on the present and so do not require parents to time travel. Thus, interviews might be more sensitive than questionnaires at capturing early indices of parents’ tendency to think about future interactions and the infant as a unique individual with specific needs. Arguably, these thoughts may be stronger predictors of parenting than feelings surrounding the experience of pregnancy. We also expected that the magnitude of the association between prenatal thoughts and feelings about the unborn infant and postnatal observed parenting would vary across different measures of parenting. In particular, as parent sensitivity has been a primary focus within attachment theory (van der Voort, Juffer, & Bakermans-Kranenburg, 2014), we expected prenatal measures to show a stronger association with postnatal ratings of sensitivity than ratings of other dimensions, such as conflict. Another between-study contrast that may reduce the strength of the association is the length of the intervening period between prenatal thoughts and feelings about the unborn infant and postnatal parenting. Two lines of evidence support this view. First, the literature on maternal-fetal relationships indicates that the attachment to the unborn infant strengthens across gestation, such that measures gathered early in pregnancy are likely to display weaker associations with observed postnatal parent-infant interaction. Second, the influence of infants on the quality of parent-infant interaction is likely to grow over time (Larsson, Viding, & Rijsdijk, 2008) such that ratings of parent-infant interaction gathered at later ages are expected to show weaker associations with parents’ prenatal thoughts and feelings. For each of these reasons, we hypothesised that the association between prenatal thoughts and feelings and postnatal parent-infant interaction quality would become attenuated as the intervening period increased in length. Parity is another potential moderator of the strength of the association between prenatal thoughts and feelings and postnatal parent-infant interaction quality. However, findings in this respect are inconsistent. While a review of this field showed no systematic association between parity and the strength of questionnaire-rated maternal-fetal relationship (Yarcheski et al., 2009), studies using the WMCI suggest a more nuanced picture in which parity interacts with parent gender. Specifically, whilst parity did not predict the quality of fathers’ prenatal representations (Hall et al., 2014), mothers expecting their first child were less likely than mothers expecting a later-born child to provide disengaged representations (Vreeswijk, Rijk, Maas, & van Bakel, 2015). Given the lack of decisive findings we adopted an exploratory approach to the question of whether parity moderates the strength of the association between prenatal and postnatal measures. Whilst child gender might also be expected to influence the nature of parent-child interactions, findings from a comprehensive meta-analysis challenge this view (Endendijk, Groeneveld, Bakermans-Kranenburg, & Mesman, 2016). Similarly, Arnott and Meins (2008) reported no contrast in mind-mindedness between expectant parents who knew or did not know their infants’ sex. Accordingly, in the current meta-analytic review we did not expect that child gender would moderate the strength of the association between prenatal measures of parenting and postnatal observed parenting. Furthermore, the association between prenatal measures and postnatal observed parenting may also differ by parent gender. Unfortunately, there is insufficient evidence on this point because fathers are under-represented within studies of early parenting, often due to limited availability or a hesitancy to engage with research or health services (Barker, Iles, & Ramchandani, 2017). In an exceptional study to examine both maternal and paternal-fetal relationships, de Cock et al. (2016) found similar associations between parent-fetal attachment and self-reported attachment for fathers and mothers at both 6 months and 24 months. Theoretically however, fathers’ lack of physical connection to the fetus might be expected to attenuate the association between prenatal thoughts and feelings and observed postnatal parent-infant interaction quality (Ives, 2014). We aimed to include data from both expectant mothers and fathers, prioritising data from fathers when studies included both mothers and fathers in order to compensate for the relative scarcity of data on fathers. Finally, it is noted that typically developmental psychologists draw conclusions from Western Educated Industrialised Rich and Democratic (also known as ‘WEIRD’) samples (Henrich, Heine, & Norenzayan, 2010). Variation within these WEIRD samples is often restricted, notably on the basis of income, education, class, ethnicity, religion and ‘risk’. Yet, studies that do recruit more diverse samples highlight that predictors of parenting are often sample-specific. In particular, alongside elevated rates of problems in representing or interacting with infants (Sokolowski et al., 2007), high-risk groups are likely to show greater variability such that stronger associations are expected between prenatal measures of thoughts and feelings about the unborn infant and postnatal observed parenting in samples classified as high-risk. This hypothesis is consistent with findings from two separate meta-analyses regarding postnatal predictors of parenting (Lovejoy, Graczyk, O'Hare, & Neuman, 2000; Paulussen-Hoogeboom, Stams, Hermanns, & Peetsma, 2007). Summary of Main Aims In order to encourage evidence-based decision-making and practice for both parents and clinicians, the current review aimed to integrate the findings from empirical studies of the association between expectant parents’ thoughts and feelings about the unborn infant and their observed postnatal parenting. To this end, the first main study goal was to conduct a meta-analytic investigation of whether expectant parents’ thoughts and feelings about the unborn infant show meaningful associations with observed postnatal parenting. Our second main goal was to establish whether methodological factors (including sample characteristics) moderate the link between prenatal thoughts and feelings about the unborn infant and postnatal parenting quality. Specifically, we hypothesised that this association would be stronger for: (i) interviews compared with questionnaire measures; (ii) ratings of postnatal sensitivity as compared with other parenting dimensions; (iii) studies with a shorter interval between the pre- and postnatal time point; (iv) mothers compared to fathers; and (iv) high-risk groups compared with low-risk samples. Given the absence of decisive data, we adopted an exploratory approach to examining potential moderating effects of parity and child gender. Method Search Strategy A literature search was conducted for studies examining the predictive quality of expectant parents’ thoughts and feelings about the unborn infant and postnatal observed parenting using electronic databases, including Scopus, EBSCOhost (PsychINFO, Child and Adolescent Studies and PsychArticles) and WorldCat (theses), between January 2016 and August 2016. The references of the final papers were also checked and authors of the papers subject to the full-paper screen were contacted to ask if they had (or knew of) any unpublished results, also known as “fugitive literature” (Rosenthal, 1994). In addition, we were active in making personal contact with researchers in the field at four different conferences. The stems from the following key words were used separately: ‘prenatal*’, ‘antenatal*’, ‘pregnan*’, ‘perinatal*, ‘expectant mother’, ‘expectant father’, ‘expectant parent’, and then in conjunction with the following key words: ‘attachment’, ‘attribution’, ‘caregiving’, ‘coherence’, ‘criticism’, ‘discourse’, ‘expressed emotion’, intrusive*’, ‘mentalising’, ‘mind-mindedness’, ‘narrative’, ‘observ*’, ‘parent*’, ‘perception’, ‘representation’, ‘responsiv*’, ‘reflective’ ‘sensitiv*’, ‘speech’, ‘speech sample’, ‘warmth’. The names of known measures were also searched, including ‘Working Model of the Child Interview’, ‘IRMAG-R’, ‘Pregnancy Interview’, ‘Maternal-Foetal Attachment Scale’, ‘Maternal Antenatal Attachment Scale’ and ‘Prenatal Attachment Inventory’. Study Inclusion and Exclusion Criteria The search overall yielded 23,132 articles, excluding duplicates and 8 theses. Given the high volume of references, articles were first screened according to the inclusion criteria on the basis of their title and abstract (please note the inclusion of nursing journals led to a high volume of medical references). Specifically, to be included, studies had to be published in English, empirical and have a pre- and postnatal time point (20% of abstracts were independently double coded with perfect agreement obtained) (see Figure 1). Studies were excluded if they did not include a measure of parent-child interaction quality, for example by focusing on parents’ wellbeing, co-parenting, their own relationship with their parents (e.g., Adult Attachment Interview) or child outcomes (e.g., attachment security, behavioural adjustment). Authors were contacted when studies did not report on the association between all pre- and postnatal measures. However, some authors did not respond (e.g., Chrzan-Dętkoś & Łockiewicz, 2015; Thun-Hohenstein, Wienerroither, Schreuer, Seim, & Wienerroither, 2008; M. Wilson et al., 2000). The final sample consisted of 14 studies. Study Coding The 14 studies were subject to detailed coding and information regarding participant characteristics, study design and effect sizes were extracted (Table 1). Moderators Measures. Table 2 provides a list of the specific questionnaire and interview measures used during pregnancy as well as the observation measures used postnatally. Design. The timing of the prenatal measure (i.e., number of weeks gestation) and postnatal measure (i.e., mean infant age) was recorded to create an interval measure to index the length of time between study time points. Parity. Mother parity was noted and studies were dichotomized as wholly first-time parents (primiparous) or not (both primiparous and multiparous women included in the design). Female to male ratio. We created a continuous variable to reflect the total number of female to male infants in each study. Risk status. A study sample was classified as high-risk if the participants, on average, were either (i) from a low socio-economic background (e.g., low income as defined by the OECD poverty line, and/or low parental education, defined as a below age 18 education level) or (ii) experienced clinical risk (e.g., mental health problems, domestic violence). It was not possible to sub-divide risk for analyses due to the low frequency of each risk type in an already small sample of studies. Planned Analysis Pearson correlation coefficients (r) were used to calculate effect sizes. Studies reporting η2 effect sizes were transformed to r (k = 3) (using Lenhard & Lenhard, 2016 as discussed by; Rosenthal, 1991). Following Ellis’ (2010) guidance, effect sizes were interpreted as modest (.10 to <.30), moderate (>.30 to <.50) and large (>.50). Four of the 14 studies also included fathers (marked with an asterisk in Table 1). In these cases, only data from the fathers were included to maximise the possibility of examining parent gender differences. Some studies reported on numerous dimensions of the postnatal parent interaction quality (Lucassen et al., 2015; Maas, de Cock, Vreeswijk, Vingerhoets, & van Bakel, 2016; McMahon, Camberis, Berry, & Gibson, 2016; Siddiqui & Hägglöf, 2000; Smaling et al., 2016). In these cases, the measure related to sensitivity was used to maximise comparability across studies. In the one study which had multiple postnatal time points (McMahon et al., 2016), the time point in infancy was chosen in order to ensure comparability. Three studies (Fuller, 1990; Pajulo et al., 2012; Theran et al., 2005) did not provide details as to the number of girls and boys included in their study. In these cases, the gender distribution was assumed to be equal. Wilson’s (2016) meta-analysis macros for IBM SPSS version 24 were used to test the two questions: 1. Is there a significant association between measures of expectant parents’ thoughts and feelings about the unborn infant and observed postnatal parenting? 2. What methodological features of individual studies moderate this effect? Following Rosenthal (1984), in order to standardise the distribution of r, each independent effect size was transformed from r to Zr and the analyses were then weighted (w = n – 3) to allow for sample size variation though Pearson’s r (95%CI) are reported for ease of interpretation. Hedge’s (1992) method was used to calculate the mean effect size. To determine the exactness of the overall effect size, 95% confidence intervals were also calculated. Cochran’s Q statistic provides an assessment of homogeneity. A significant Q suggests heterogeneity in the distribution of effect sizes, thus a random-effects model can be used to analyse the effect sizes. Field and Gillett (2010) emphasise that as social science data is drawn from the real world, where heterogeneity is the norm, a random-effects model should be adopted as this model assumes that effect sizes vary randomly between studies. Specifically, the random-effects model assumes two error terms; not only is there error due to sampling from within different study populations, but that additional error is generated as a result of sampling the populations from the population at large. Furthermore, Type 1 error rates are much more likely (5% to 11-28%) when applying a fixed-effects model to heterogeneous data whereas the reverse is not as substantial. Thus, findings are more easily generalizable and, if significant, the heterogeneity within the effect sizes can be explored further rather than be assumed to be the result of error. Three different methods were adopted to establish whether the effect sizes were subject to publication bias. First, a funnel plot of the effect sizes against the sample size was used to explore the data visually. An unbiased sample should produce a graph of data points hanging symmetrically around the mean effect size thus creating the effect of a ‘funnel’ suggesting effect sizes drawn from smaller samples have greater variability. In contrast, it is suggested that an asymmetric funnel plot would indicate publication bias as smaller samples with smaller effect sizes are not published (Light & Pillemer, 1984). Second, Rosenthal’s (1984) fail-safe N, that is the minimum number of null results taken to overturn the observed effect, was calculated to test the effect size resistance to the file-drawer threat. The fail-safe N should exceed a critical value: 5 times the number of effect sizes plus 10 (i.e., 5K + 10) if it is resistant to the file-drawer threat. Whilst valuable and commonly used, Borenstein, Hedges, Higgins and Rothstein (2009) note the problematic emphasis on statistical significance of the fail-safe N. Specifically, that statistical and substantive significance are not synonymous and that the formula used to calculate the fail-safe N assumes effect sizes in the file-drawer to be zero when in reality they could reflect a range of possibilities, including being negative. Simonsohn, Nelson and Simonsohn (2014) also note that study findings may be the result of ‘p-hacking’, which is the selective publishing of significant analyses, and propose the use of p-curves to examine this possibility. The shape of the p-curve (i.e., the distribution of p-values) is determined by the effect and sample size, with only p-values < .05 included. Simonsohn et al. (2014) propose that a genuine effect would produce a right skewed p-curve, a lack of effect would produce a flattened, uniform p-curve and when the effects are the result of p-hacking the p-curve would be skewed to the left. With this is mind, a p-curve was also plotted to help establish whether the research has evidential value. Results Table 1 summarise the characteristics of the 14 samples, that together included 1862 parents (n = 1109 mothers, Mage = 28.69, SD = 3.61 years; n = 753 fathers, Mage = 34.50, SD = .71 years) and their children (Mage = 9.46, SD = 12.52 months). The majority of studies used samples of mothers (k = 10) and studies typically took place in Europe (k = 8) although North America (k = 4), Australia (k = 1) and South America (k = 1) were also represented. The samples were typically recruited from antenatal education classes, hospitals and via leaflets. In line with the internal validity subscale from Downs and Black’s (1998) methodological quality checklist, the reliability of the key prenatal and postnatal measures was used to index study quality (see Table 2). Other indices of bias were examined as potential moderators (e.g., sample size). Typically, the studies included in the meta-analysis were of high quality. It is worth noting that all studies, aside from the thesis (Foley, 2018), were published in peer-review journals. Overall Effect: Do Measures of Expectant Parents’ Thoughts and Feelings about the Unborn Infant Relate to Postnatal Observed Parenting? To calculate the overall strength and significance of the association between prenatal thoughts and feelings about the unborn infant and postnatal parenting, a random-effects model weighted by the inverse variance was used. The weighted mean effect size (r) across the 14 independent correlations was .24, 95%CI [.10, .38] (see the Forest Plot in Figure 2). Although modest, this effect size is significantly greater than zero, Z = 3.40, p < .001 (see Table 3). Publication Bias The funnel plot (see Figure 3) suggests the absence of publication bias, with studies distributed symmetrically about the mean effect size. That said, the small number of effect sizes limits the interpretative value of the graph. A fail-safe N was calculated to address the file-drawer problem, that is, that significant results are disproportionately published, undermining the robustness of the mean effect size. If a result is resistant to the file-drawer threat, the fail-safe N should exceed a critical value (i.e., 5K + 10). The fail-safe N exceeded the critical value of 80, with a total of 823 nonsignificant studies needed to reduce the significance of the mean effect size beyond p > .05. The p-curve for the studies indicating a statistically significant association between prenatal thoughts and feelings about the unborn infant and postnatal parenting was plotted (K = 9). As seen in Figure 4, the p-curve was significantly positively skewed, Z = -7.38, p < .001, suggesting the results are not likely to be the result of selective reporting and as such contain evidential value. Moderator Analyses: Do Methodological Characteristics Moderate the Relationship Between Prenatal Thoughts and Feelings About the Unborn Infant and Postnatal Measures of Parenting? The significant heterogeneity statistic, Q (13) = 94.28, p < .001, and the stem and leaf plot (see Table 1 in supplementary material) highlighted the heterogeneity of effect sizes and so reaffirmed the need to look for potential moderators that might account for variability. Studies were divided into categories according to methodological and sample characteristics. To examine heterogeneity between study findings categorical moderator analyses analogous of ANOVA were conducted. The analysis also provides homogeneity statistics (Q) for both between (Qb) and within groups (Qw). A significant Qb suggests that the effect sizes are significantly different across different categories of the moderator variable whilst a significant Qw suggests that the effect sizes are significantly different from zero. A random-effects model weighted by the inverse variance analogous to ANOVA (D. Wilson, 2016) showed that the strength of the association between prenatal assessments of thoughts and feelings about the unborn infant and postnatal observed parenting varied significantly by parent gender, Qb (1, 12) = 6.32, p = .012 (see Table 3). Specifically, the effect size for mothers was significant, .32, whilst the mean effect size for fathers was nonsignificant, -.01. Further analyses to explore the potential moderator effects of additional methodological features (type of prenatal measure, sensitivity outcome, parity and risk) yielded nonsignificant results confirming the stability of the effect across these groups (see Table 3). Regression Analyses: Do Continuous Study Design Variables Influence the Relationship Between Prenatal Thoughts and Feelings about the Unborn Infant and Postnatal Measures of Parenting? A random-effects linear regression model was run, using inverse variance weights to account for differences in sample size, to examine heterogeneity between study findings on the basis of continuous moderator variables. Heterogeneity in the effect sizes was not confirmed, Q (3, 10) = 4.68, p = .200. The overall model accounted for 25.52% of the variance in effect sizes. None of the characteristics included (study sample size, ratio of girls, or the interval between the two time points) were significant predictors in the model, suggesting the strength of the association was consistent across sample sizes, gender composition of the sample and the length of time between the pre- and postnatal measure. Discussion A focus on pregnancy and maternal antenatal care is a key strategy for achieving the third United Nations (2015) Sustainable Development Goal to ensure healthy lives and promote lifelong wellbeing. For policy-makers interested in the determinants of positive parenting, pregnancy provides a valuable window of opportunity for interventions. The main goal of the current study was to test whether variation in expectant parents’ thoughts and feelings about the unborn infant predicted individual differences in postnatal observational ratings of parent-infant interactions. Existing studies involving expectant parents have typically adopted two approaches: interviews to obtain parents’ representations of the infant and questionnaires to gather information about expectant parents’ thoughts and feelings towards the unborn infant. Our meta-analytic results revealed that each of these methods of assessing prenatal thoughts and feelings showed a modest but significant association with observed postnatal ratings of parent-child interaction quality. While we had hypothesized that the association would be stronger for interview measures than for questionnaires, for sensitivity than other parental outcomes, for studies with a shorter interval between the pre- and postnatal time points, and for high-risk rather than low-risk samples, none of these contrasts were significant. Interestingly, parent gender was the only significant moderator: associations between prenatal thoughts and feelings and postnatal observed interaction quality were stronger for mothers than fathers. Below, we consider each of these findings in turn whilst also highlighting avenues for future research. Different Methods, Similar Results? Whilst it is well recognised that interview-based measures of expectant parents’ representations of their own caregivers during pregnancy predict subsequent parenting and child attachment security (Fonagy et al., 1993), our findings indicate that prenatal interviews about the unborn infant also have predictive utility. An interesting direction for future research using prenatal interviews would therefore be to adopt a dual focus on both parents’ own caregivers and their unborn infants in order to test the interdependence and relative salience of each prenatal representation as a predictor of postnatal parenting. Our review also identified five different interview measures of parents’ prenatal thoughts and feelings of infants, each typically linked to a different coding system. Another promising avenue for future research would therefore be to evaluate the merits of these alternative interview protocols in order to consolidate findings and establish a sufficient evidence base for conducting a systematic comparison of effect sizes. Our meta-analytic results also showed a surprising but reassuring consistency in associations across studies using interviews and questionnaires. That is, questionnaire-based measures of the parent-fetal relationship also predicted the quality of parent-infant interactions. This is interesting; as alluded to before, though similar in many ways, the questionnaires and interviews clearly differ with regards to developmental timing: while questionnaires focus on the present experience of pregnancy, interviews focus on the future child. While it would be premature to equate the parent-fetal relationship with parent-infant attachment, it is clear that this construct warrants further attention (Walsh, 2010). To date, researchers have yet to apply both interviews and questionnaires in the same study. As a result, the relative independence and interplay between the quality of pa rents’ representation of the infant and their concurrent feelings and behaviours towards the fetus as predictors of the quality of postnatal parenting have yet to be established. Given that questionnaires are much more cost-effective than interviews, addressing this methodological gap is another important avenue for future research. Turning to the postnatal period, our findings again indicated similar results for studies that adopted different outcome measures. Specifically, we had anticipated that the effect size would be stronger for studies focused upon parental sensitivity rather than other dimensions of parent-infant interaction quality. The lack of difference in the strength of the association between expectant parents’ thoughts and feelings about the unborn infant and postnatal observational ratings of parent-infant interactions may reflect inconsistencies in the measurement and definition of sensitivity. Not all sensitivity coding schemes follow Ainsworth’s original outline of the construct; notably some (e.g., the CARE Index; Crittenden, 2001) include warmth in their scales. This is important because although warmth is moderately correlated with sensitivity, these constructs emerge from different motivational systems and show both overlapping and independent associations with specific child outcomes, such that researchers are recommended to include separate ratings of parental warmth and sensitivity (Mesman & Emmen, 2013). Furthermore, nearly every study we identified applied a different scheme to code the parent-infant observations (see Table 2) and so it was not possible to determine whether choice of coding scheme moderated the association between prenatal measures of thoughts and feelings about the unborn infant and observed postnatal interaction quality. We therefore recommend that researchers are more consistent in their future choice of parent-infant interaction coding schemes. Counter to our expectations, our findings showed that the strength of the association between pre- and postnatal measures was not stronger for studies with a shorter intervening period. Confirming this, additional moderator analyses, not presented here for reasons of space, showed that the strength of the association did not differ by either trimester or child age. This may reflect the focus on the early postnatal period, as only one study examined parent-child interactions beyond the first 12 months. The addition of longer intervals in future studies may reveal that associations between expectant parents’ thoughts about the infant and dimensions of parent-infant quality becomes attenuated over time as children become more active partners in their relationships. It is also important for future research to establish whether variation in these prenatal constructs is associated with later child outcomes. Informed by recent models (e.g., Grusec and Davidov (2010) that highlight the domain-specific nature of parenting, future research should include several parent and child outcomes in order to establish whether specific profiles of prenatal thoughts and feelings are related to particular parent or child outcomes. Our findings also indicated that the strength of the association between prenatal measures and observed postnatal parent-infant interaction quality was consistent across different sample compositions. In particular, counter to our expectation, the association was not stronger for high-risk samples than for low-risk samples. This lack of moderation effect may, however, simply reflect both the small number of studies of high-risk groups and inconsistencies in the reporting of sample characteristics. For example, some studies did not provide key information on the socio-economic status (SES) of the sample. In these cases it was assumed that there was a lack of SES risk within the sample, as research in this area tends to use middle class samples unless clearly examining a risk variable (e.g., Pajulo et al., 2012). Unfortunately, power restrictions meant that the two risk categories (low SES and clinical risk) needed to be combined, precluding analysis of risk type as a moderating factor. Similarly, while parity also did not moderate the strength of the association, the lack of studies that were focused explicitly on later-born children necessitated a comparison between samples of first-time expectant parents and samples of expectant parents with a mixed experience of parenthood. Other potential moderators that we were unable to assess include: (i) whether or not the pregnancy was planned; and (ii) use of assisted reproductive technologies (ART). Here it is worth noting that women with unintended pregnancies show an increased risk for postnatal depression which subsequently may impact on parent-child interaction quality (Gipson, Koenig, & Hindin, 2008), while expectant mothers and fathers using ART report significantly higher fetal bonding than samples of expectant parents through natural conception (Cairo et al., 2012). We hope that our review will stimulate further research that has greater sample diversity alongside greater consistency in measurement and reporting. Links between Prenatal Thoughts and Postnatal Interactions are Stronger for Mothers than Fathers Consistent with Ives’ (2014) theoretical proposal that expectant fathers’ lack of physical connection to the fetus presents a challenge in developing early close relationships with the infant, empirical evidence indicates that almost half (49%) of paternal prenatal narratives are categorised as disengaged with a further 7% categorised as distorted (Vreeswijk, Maas, Rijk, Braeken, & van Bakel, 2014). In contrast, studies of expectant mothers consistently report that almost two-thirds of narratives can be classified as balanced (Benoit et al., 1997; Vreeswijk et al., 2012). This contrast between expectant mothers’ and fathers’ ability to construct a balanced representation of their unborn infant may explain the stronger association between prenatal thoughts and observed parenting for mothers than for fathers. That said, it is worth noting that methodological factors may also play an important moderating role. In particular studies that have used questionnaires rather than interviews indicate greater stability of responses across the transition to parenthood for fathers than mothers (e.g., de Cock et al., 2016). To date however, only a very limited number of studies have included expectant fathers and so testing whether the type of prenatal assessment moderates the contrast in stability for mothers and fathers requires a larger evidence base. Policy Implications Rather than adding pressure to expectant parents by attempting to identify key characteristics of a psychologically healthy pregnancy, this review aimed to add clarity to the ongoing debate about prenatal determinants of parenting. There is a delicate balance to be struck between ensuring parents are given the information they need to make decisions and restricting parents’ autonomy with regards to their thoughts or behaviours during pregnancy (or beyond). However, the findings from the current meta-analysis are based on a small set of studies and so it is too early to draw any firm conclusions regarding the utility of evaluating expectant parents’ thoughts and feelings or promoting or discouraging specific thoughts and feelings about the unborn infant. For example, the inability to connect with or think about their unborn infant might be a useful coping strategy or defence mechanism in the face of prior loss or infertility (Lee, Mckenzie-Mcharg, & Horsch, 2013). Equally, an emphasis on thinking or feeling a certain way about the unborn infant may be culturally inappropriate or contribute to prenatal anxiety. That said, our meta-analytic results do suggest that future research into how prenatal thoughts and feelings might matter for postnatal parenting is warranted and may offer promising avenues for tailoring early interventions. Conclusions Overall, the current review integrates findings from strands of research that adopt different methods (interviews and questionnaires) to capture variability in expectant parents’ thoughts and feelings regarding their unborn infant. Meta-analytic results revealed a modest but significant and consistent association between these prenatal measures and observed postnatal interaction quality. The only significant moderator of the association was parent gender: associations between prenatal thoughts and feelings and postnatal observed interaction quality were stronger for mothers than fathers. To increase understanding and inform intervention, future research should involve more diverse samples, unpack the role of measurement (in both pre- and postnatal schemes), identify the specificity of the association between prenatal measures of thoughts and feelings about the unborn infant and postnatal parenting and, ultimately, examine whether individual differences in these prenatal measures predict child outcomes. Acknowledgements We would like to thank Amanda Aldercotte for statistical advice along with her assistance with reliability coding and Rory T. Devine, for statistical advice. References Ainsworth, M., & Bell, S. (1970). Attachment, exploration, and separation: illustrated by the behavior of one-year-olds in a strange situation. Child Development, 41, 49-67. doi:10.2307/1127388 Ainsworth, M., Bell, S., & Stayton, D. (1974). Infant-mother attachment and social development. In M. Richards (Ed.), The integration of a child into a social world (pp. 99–135). London: Cambridge University Press. Arnott, B., & Meins, E. (2008). Continuity in mind-mindedness from pregnancy to the first year of life. Infant Behavior and Development, 31, 647-654. doi:10.1016/j.infbeh.2008.07.001 Barker, B., Iles, J., & Ramchandani, P. (2017). Fathers, fathering and child psychopathology. Current Opinion in Psychology, 15, 87-92. doi:10.1016/j.copsyc.2017.02.015 Barnett, M., Deng, M., Mills-Koonce, W., Willoughby, M., & Cox, M. (2008). Interdependence of Parenting of Mothers and Fathers of Infants. Journal of Family Psychology, 22, 561-573. doi:10.1037/0893-3200.22.3.561 Belsky, J. (1984). The determinants of parenting: A process model. Child Development, 55, 83-96. doi:10.2307/1129836 Benoit, D., Parker, K., & Zeanah, C. (1997). Mothers' representations of their infants assessed prenatally: Stability and association with infants' attachment classifications. Journal of Child Psychology and Psychiatry and Allied Disciplines, 38, 307-313. doi:10.1111/j.1469-7610.1997.tb01515.x Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2009). Introduction to Meta-Analysis. Chichester, UK: John Wiley & Sons, Ltd. Bowlby, J. (1969). Attachment and Loss: Vol. 1. Attachment. New York: Basic Books. Bullock, B., & Dishion, T. (2004). Family Affective Attitude Rating Scale (FAARS). Cairo, S., Darwiche, J., Tissot, H., Favez, N., Germond, M., Guex, P., . . . Despland, J. N. (2012). Family interactions in IVF families: Change over the transition to parenthood. Journal of Reproductive and Infant Psychology, 30, 5-20. doi:10.1080/02646838.2012.669830 Caspi, A., Moffitt, T., Morgan, J., Rutter, M., Taylor, A., Kim-Cohen, J., & Polo-Tomas, M. (2004). Maternal expressed emotion predicts children's antisocial behaviour problems: Using monozygotic-twin differences to identify environmental effects on behavioural development. Developmental Psychology, 40, 149-161. doi:10.1037/0012-1649.40.2.149 Chrzan-Dętkoś, M., & Łockiewicz, M. (2015). Maternal romantic attachment, and antenatal and postnatal mother–infant attachment in a sample of Polish women. European Journal of Developmental Psychology, 12, 429-442. doi:10.1080/17405629.2015.1036024 Condon, J. (1993). The assessment of antenatal emotional attachment: Development of a questionnaire instrument. British Journal of Medical Psychology, 66, 167-183. doi:10.1111/j.2044-8341.1993.tb01739.x Cranley, M. (1981). Development of a tool for the measurement of maternal attachment during pregnancy. MS Nursing Research, 30, 281-284. doi:10.1097/00006199-198109000-00008 Crawford, A., & Benoit, D. (2009). Caregivers' disrupted representations of the unborn child predict later infant-caregiver disorganized attachment and disrupted interactions. Infant Mental Health Journal, 30, 124-144. doi:10.1002/imhj.20207 Crittenden, P. (2001). CARE-index manual. Family Relations Institute. Miami, FL. Daley, D., Sonuga-Barke, E., & Thompson, M. (2003). Assessing expressed emotion in mothers of preschool AD/HD children: psychometric properties of a modified speech sample. British Journal of Clinical Psychology, 42, 53–67. doi:10.1348/014466503762842011 de Cock, E., Henrichs, J., Vreeswijk, C., Maas, A., Rijk, C., & van Bakel, H. (2016). Continuous feelings of love? The parental bond from pregnancy to toddlerhood. Journal of Family Psychology, 30, 125-134. doi:10.1037/fam0000138 Desjardins, J., Zelenski, J., & Coplan, R. (2008). An investigation of maternal personality, parenting styles, and subjective well-being. Personality and Individual Differences, 44, 587-597. doi:10.1016/j.paid.2007.09.020 Downs, S., & Black, N. (1998). The feasibility of creating a checklist for the assessment of the methodological quality both of randomized and non-randomized studies of health care interventions. Journal of Epidemiology Community Health, 52, 377-384. doi:10.1136/jech.52.6.377 Duschinsky, R. (2018). Disorganization, fear and attachment: working towards clarification. Infant Mental Health Journal, 39, 17-29. doi:10.1002/imhj.21689 Ellis, P. (2010). The Essential Guide to Effect Sizes: Statistical Power, Meta-Analysis, and the Interpretation of Research Results. Cambridge, UK: Cambridge University Press. Endendijk, J., Groeneveld, M., Bakermans-Kranenburg, M., & Mesman, J. (2016). Gender-Differentiated parenting revisited: Meta-analysis reveals very few differences in parental control of boys and girls. PLoS ONE, 11. doi:10.1371/journal.pone.0159193 Field, A., & Gillett, R. (2010). How to do a meta-analysis. British Journal of Mathematical and Statistical Psychology, 63, 665-694. doi:10.1348/000711010X502733 Foley, S. (2018). Emerging Parental Sensitivity: the Transition to Parenthood Through the Lens of Family Systems Theory. (PhD), University of Cambridge. Fonagy, P., Steele, M., Moran, G., Steele, H., & Higgit, A. (1993). Measuring the ghost in the nursery: An empirical study of the relation between parents' mental representations of childhood experiences and their infants' security of attachment. Journal of the American Psychoanalytic Association, 41, 957-989. doi:10.1177/000306519304100403 Fonagy, P., Steele, M., Steele, H., Moran, G., & Higgitt, A. (1991). The capacity for understanding mental states: The reflective self in parent and child and its significance for security of attachment. Infant Mental Health Journal, 12, 201-218. doi:10.1002/1097-0355(199123)12:3< Fuller, J. (1990). Early patterns of maternal attachment. Health Care for Women International, 11, 433-446. doi:10.1080/07399339009515913 Gipson, J., Koenig, M., & Hindin, M. (2008). The effects of unintended pregnancy on infant, child, and parental health: A review of the literature. Studies in Family Planning, 39, 18-38. doi:10.1111/j.1728-4465.2008.00148.x Grusec, J., & Davidov, M. (2010). Integrating different perspectives on socialization theory and research: A domain-specific approach. Child Development, 81, 687-709. doi:10.1111/j.1467-8624.2010.01426.x Hall, R., De Waard, I., Tooten, A., Hoffenkamp, H., Vingerhoets, A., & van Bakel, H. (2014). From the father's point of view: How father's representations of the infant impact on father-infant interaction and infant development. Early Human Development, 90, 877-883. doi:10.1016/j.earlhumdev.2014.09.010 Hedges, L. (1992). Meta-analysis. Journal of Educational Statistics, 17, 279-296. doi:10.3102/10769986017004279 Henrich, J., Heine, S., & Norenzayan, A. (2010). The weirdest people in the world? Behavioral and Brain Sciences, 33, 61-135. doi:10.1017/S0140525X0999152X Hughes, C., & Devine, R. (2017). For Better or for Worse? Positive and Negative Parentful Influences on Young Children's Executive Function. Child Development. doi:10.1111/cdev.12915 Ives, J. (2014). Men, maternity and moral residue: negotiating the moral demands of the transition to first time fatherhood. Sociology of health & illness, 36, 1003-1019. doi:10.1111/1467-9566.12138 Korja, R., Ahlqvist-Björkroth, S., Savonlahti, E., Stolt, S., Haataja, L., Lapinleimu, H., . . . Lehtonen, L. (2010). Relations between maternal attachment representations and the quality of mother-infant interaction in preterm and full-term infants. Infant Behavior and Development, 33, 330-336. doi:10.1016/j.infbeh.2010.03.010 Larsson, H., Viding, E., & Rijsdijk, F. (2008). Relationships between parental negativity and childhood antisocial behaviour over time: A bidirectional effects model in a longitudinal genetically informative design. Journal of Abnormal Child Psychology, 36, 633-645. doi:10.1007/s10802-007-9151-2 Lee, L., Mckenzie-Mcharg, K., & Horsch, A. (2013). Women's decision making and experience of subsequent pregnancy following stillbirth. Journal of Midwifery and Women's Health, 58, 431-439. doi:10.1111/jmwh.12011 Lenhard, W., & Lenhard, A. (2016). Calculation of Effect Sizes. Retrieved from http://www.psychometrica.de/effect_size.html. Light, R., & Pillemer, D. (1984). Summing up: The science of reviewing research. Cambridge, MA: Harvard University Press. Lindgren, K. (2001). Relationships among maternal-fetal attachment, prenatal depression, and health practices in pregnancy. Research in Nursing and Health, 24, 203-217. doi:10.1002/nur.1023 Lovejoy, M., Graczyk, P., O'Hare, E., & Neuman, G. (2000). Maternal depression and parenting behavior: A meta-analytic review. Clinical Psychology Review, 20, 561-592. doi:10.1016/S0272-7358(98)00100-7 Lucassen, N., Tiemeier, H., Luijk, M., Linting, M., Bakermans-Kranenburg, M. J., van Ijzendoorn, M., . . . Lambregtse-Van den Berg, M. (2015). Expressed Emotion During Pregnancy Predicts Observed Sensitivity of Mothers and Fathers in Early Childhood. Parenting, 15, 158-165. doi:10.1037/a0025855 Maas, A., de Cock, E., Vreeswijk, C., Vingerhoets, A., & van Bakel, H. (2016). A longitudinal study on the maternal–fetal relationship and postnatal maternal sensitivity. Journal of Reproductive and Infant Psychology, 34, 110-121. doi:10.1080/02646838.2015.1112880 Magana, A., Goldtein, M., Karno, M., Miklowitz, D., Jenkins, J., & Fallon, I. (1986). A Brief Method for Assessing Expressed Emotion in Relatives of Psychiatric Patients. Psychiatric Research, 17, 203-212. doi:10.1016/0165-1781(86)90049-1 McMahon, C., Camberis, A.-L., Berry, S., & Gibson, F. (2016). Maternal mind-mindedness: Relations with maternal-fetal attachment and stability in the first two years of life: Findings from an australian prospective study. Infant Mental Health Journal, 37, 17-28. doi:10.1002/imhj.21548 Meins, E. (1997). Security of attachment and the social development of cognition. Hove: Psychology Press. Mesman, J., & Emmen, R. (2013). Mary Ainsworth's legacy: A systematic review of observational instruments measuring parental sensitivity. Attachment and Human Development, 15, 485-506. doi:10.1080/14616734.2013.820900 Mills-Koonce, W., Willoughby, M., Zvara, B., Barnett, M., Gustafsson, H., & Cox, M. J. (2015). Mothers' and fathers' sensitivity and children's cognitive development in low-income, rural families. Journal of Applied Developmental Psychology, 38, 1-10. doi:10.1016/j.appdev.2015.01.001 Muller, M. (1993). Development of the Prenatal Attachment Inventory. Western Journal of Nursing Research, 15, 199-215. doi:10.1177/019394599301500205 Murray, L., Fearon, P., & Cooper, P. (2015). Postnatal depression, mother-infant interactions and child development: prospects for screening and treatment. In J. Milgrom & A. Gemmill (Eds.), Identifying Perinatal Depression and Anxiety: Evidence-based Practice in Screening, Psychosocial Assessment and Management (pp. 139-164). Oxford: Wiley Blackwell. Pajulo, M., Pyykkönen, N., Kalland, M., Sinkkonen, J., Helenius, H., Punamäki, R.-L., & Suchman, N. (2012). Substance-abusing mothers in residential treatment with their babies: Importance of pre- and postnatal maternal reflective functioning. Infant Mental Health Journal, 33, 70-81. doi:10.1002/imhj.20342 Paulussen-Hoogeboom, M., Stams, G., Hermanns, J., & Peetsma, T. (2007). Child negative emotionality and parenting from infancy to preschool: A meta-analytic review. Developmental Psychology, 43, 438-453. doi:10.1037/0012-1649.43.2.438 Peris, T., & Miklowitz, D. (2015). Parental Expressed Emotion and Youth Psychopathology: New Directions for an Old Construct. Child Psychiatry and Human Development, 46, 863-873. doi:10.1007/s10578-014-0526-7 Rosenthal, R. (1984). Meta-analytic Procedures For Social Research (Vol. 6). Beverly Hills: Sage Publications. Rosenthal, R. (1991). Meta-analytic Procedures For Social Research (Vol. 6). Beverly Hills: Sage Publications. Rosenthal, R. (1994). The Fugitive Literature. In H. Cooper & L. Hedges (Eds.), The Handbook of Research Synthesis (pp. 85 - 94s). New York: Russell Sage Foundation. Rossen, L., Hutchinson, D., Wilson, J., Burns, L., Olsson, C., Allsop, S., . . . Mattick, R. (2016). Predictors of postnatal mother-infant bonding: the role of antenatal bonding, maternal substance use and mental health. Archives of women's mental health, 19, 609-622. doi:10.1007/s00737-016-0602-z Sadler, L., Slade, A., Close, N., Webb, D., Simpson, T., Fennie, K., & Mayes, L. (2013). Minding the baby: Enhancing reflectiveness to improve early health and relationship outcomes in an interdisciplinary home-visiting program. Infant Mental Health Journal, 34, 391-405. doi:10.1002/imhj.21406 Schechter, D., Myers, M., Brunelli, S., Coates, S., Zeanah Jr., C., Davies, M., . . . Liebowitz, M. (2006). Traumatized mothers can change their minds about their toddlers: Understanding how a novel use of videofeedback supports positive change of maternal attributions. Infant Mental Health Journal, 27, 429-447. doi:10.1002/imhj.20101 Sher-Censor, E. (2015). Five Minute Speech Sample in developmental research: A review. Developmental Review, 36, 127-155. doi:10.1016/j.dr.2015.01.005 Sher-Censor, E., & Yates, T. (2010). Five minute speech sample narrative coherence coding manual. University of California. Riverside. Siddiqui, A., & Hägglöf, B. (2000). Does maternal prenatal attachment predict postnatal mother-infant interaction? Early Human Development, 59, 13-25. doi:10.1016/S0378-3782(00)00076-1 Simonsohn, U., Nelson, L., & Simmons, J. (2014). P-curve: A key to the file-drawer. Journal of Experimental Psychology: General, 143, 534-547. doi:10.1177/1745691614553988 Slade, A., Grienenberger, J., Bernbach, E., Levy, D., & Locker, A. (2002). The parent development interview and the pregnancy interview: Manuals for scoring. City College of New York and Yale Child Study Center. New Haven, CT. Smaling, H., Huijbregts, S., Suurland, J., van der Heijden, K., Mesman, J., van Goozen, S., & Swaab, H. (2016). Prenatal Reflective Functioning and Accumulated Risk as Predictors of Maternal Interactive Behavior During Free Play, the Still-Face Paradigm, and Two Teaching Tasks. Infancy, 21, 766-784. doi:10.1111/infa.12137 Smetana, J. (2017). Current research on parenting styles, dimensions, and beliefs. Current Opinion in Psychology, 15, 19-25. doi:10.1016/j.copsyc.2017.02.012 Sokolowski, M., Hans, S., Bernstein, V., & Cox, S. (2007). Mothers' representations of their infants and parenting behavior: Associations with personal and social contextual variables in a high-risk sample. Infant Mental Health Journal, 28, 344-365. Sroufe, L. (2005). Attachment and development: A prospective, longitudinal study from birth to adulthood. Attachment and Human Development, 7, 349-367. doi:10.1080/14616730500365928 Tambelli, R., Odorisio, F., & Lucarelli, L. (2014). Prenatal and postnatal maternal representations in nonrisk and at-risk parenting: Exploring the influences on mother-infant feeding interactions. Infant Mental Health Journal, 35, 376-388. doi:10.1002/imhj.21448 Theran, S., Levendosky, A., Bogat, G., & Huth-Bocks, A. (2005). Stability and change in mothers' internal representations of their infants over time. Attachment and Human Development, 7, 253-268. doi:10.1080/14616730500245609 Thun-Hohenstein, L., Wienerroither, C., Schreuer, M., Seim, G., & Wienerroither, H. (2008). Antenatal mental representations about the child and mother-infant interaction at three months post partum. European Child and Adolescent Psychiatry, 17, 9-19. doi:10.1007/s00787-007-0622-3 UN. (2015). General Assembly resolution 70/1: Transforming our world: the 2030 Agenda for Sustainable Development, A/70/L.1. New York: United Nations. Vallotton, C., Mastergeorge, A., Foster, T., Decker, K., & Ayoub, C. (2017). Parenting Supports for Early Vocabulary Development: Specific Effects of Sensitivity and Stimulation through Infancy. Infancy, 22, 78-107. doi:10.1111/infa.12147 van den Bergh, B., & Simons, A. (2009). A review of scales to measure the mother-foetus relationship. Journal of Reproductive and Infant Psychology, 27, 114-126. doi:10.1080/02646830802007480 van der Voort, A., Juffer, F., & Bakermans-Kranenburg, M. (2014). Sensitive parenting is the foundation for secure attachment relationships and positive social-emotional development of children. Journal of Children's Services, 9, 165-176. doi:10.1108/JCS-12-2013-0038 van Ijzendoorn, M., Schuengel, C., & Bakermans-Kranenburg, M. (1999). Disorganized attachment in early childhood: Meta-analysis of precursors, concomitants, and sequelae. Development and Psychopathology, 11, 225-249. doi:10.1017/S0954579499002035 Vreeswijk, C., Maas, A., Rijk, C., Braeken, J., & van Bakel, H. (2014). Stability of fathers' representations of their infants during the transition to parenthood. Attachment and Human Development, 16, 292-306. doi:10.1080/14616734.2014.900095 Vreeswijk, C., Maas, A., & van Bakel, H. (2012). Parental representations: A systematic review of the working model of the child interview. Infant Mental Health Journal, 33, 314-328. doi:10.1002/imhj.20337 Vreeswijk, C., Rijk, C., Maas, A., & van Bakel, H. (2015). Fathers' and mothers' representations of the infant: Associations with prenatal risk factors. Infant Mental Health Journal, 36, 599-612. doi:10.1002/imhj.21541 Walsh, J. (2010). Definitions matter: If maternal-fetal relationships are not attachment, what are they? Archives of women's mental health, 13, 449-451. doi:10.1007/s00737-010-0152-8 Walsh, J., Hepper, E., Bagge, S., Wadephul, F., & Jomeen, J. (2013). Maternal-fetal relationships and psychological health: Emerging research directions. Journal of Reproductive and Infant Psychology, 31, 490-499. doi:10.1080/02646838.2013.834311 Weston, S., Hawes, D., & Pasalich, D. (2017). The Five Minute Speech Sample as a Measure of Parent–Child Dynamics: Evidence from Observational Research. Journal of Child and Family Studies, 26, 118-136. doi:10.1007/s10826-016-0549-8 Wilson, D. (2016). Meta-analysis macros for SAS, SPSS, and Stata. Version 6.1. Retrieved from http://mason.gmu.edu/~dwilsonb/ma.html Wilson, M., White, M., Cobb, B., Curry, R., Greene, D., & Popovich, D. (2000). Family dynamics, parental-fetal attachment and infant temperament. Journal of Advanced Nursing, 31, 204-210. doi:10.1046/j.1365-2648.2000.01245.x Winnicott, D. (1960). The theory of the parent-infant relationship. In L. Caldwell & A. Joce (Eds.), Reading Winnicott. East Sussex: Routledge. Yarcheski, A., Mahon, N., Yarcheski, T., Hanks, M., & Cannella, B. (2009). A meta-analytic study of predictors of maternal-fetal attachment. International Journal of Nursing Studies, 46, 708-715. doi:10.1016/j.ijnurstu.2008.10.013 Zaslow, M., Gallagher, M., Hair, E., Egeland, B., Weinfield, N., Ogawa, J., . . . De Temple, J. (2006). Longitudinal prediction of child outcomes from differing measures of parenting in a low-income sample Developmental Psychology, 42, 27-37. doi:10.1037/0012-1649.42.1.27 Zeanah, C., Benoit, D., & Barton, M. (1986). Working Model of the Child Interview. Brown University Program in Medicine. Providence, RI. Running Head: GREAT EXPECTATIONS? 1 27 GREAT EXPECTATIONS? Table 1 Reviewed Studies, Sample Characteristics, Moderator Variables and Study Effect Sizes Reference Prenatal Postnatal n Parent M age Primip Risk Measure Reliability Child M age (months) Female: Male Outcome Reliability ES Alvarenga et al. (2013) 38 29.00 ✓ Q Questionable 8.00 .00 Sensitivity Good .48 Arnott & Meins (2008)* 17 35.50 Q Not reported 6.00 .50 Mind-mindedness Good .08 Cairo et al. (2012)* 31 34.34 ✓ Q Good 9.04 1.07 Alliance Good .06 Crawford & Benoit (2009) 35 30.29 I Good 12.00 .84 Atypical behaviour Good .64 Foley (2018)* 187 33.84 ✓ I Good 4.00 .82 Sensitivity Good .06 Fuller (1990) 32 29 Q Good .03 1.00 Sensitivity Good .65 Lucassen et al. (2015)* 518 34.30 I Moderate 51.40 1.06 Sensitivity Good -.06 Maas et al. (2016) 273 31.87 Q Moderate 6.08 .96 Sensitivity Moderate .16 McMahon et al. (2016) 132 33.50 ✓ Q Good 7.22 .78 Mind-mindedness Good -.12 Pajulo et al. (2012) 19 25.10 ✓ I Good 4.00 1.00 Sensitivity Not reported .16 Siddiqui & Hägglöf (2000) 100 30.00 Q Good 2.66 1.08 Involvement Good .51 Smaling et al. (2016) 133 22.86 ✓ ✓ I Good 6.02 .85 Sensitivity Good .21 Tambelli et al. (2014) 167 32.60 ✓ ✓ I Good 4.00 .92 Conflict Good .34 Theran et al. (2005) 180 25.00 ✓ I Good 12.00 1.00 Sensitivity Good .20 Note. * = data from fathers and not mothers used in analyses. Primip = Primiparous sample. Q = questionnaire; I = interview. Reliability code based on interpretation of inter-rater agreement or the Cronbach’s α of a questionnaire. ES = effect size. Table 2 Description of Prenatal and Postnatal Measures Time point Category Description k Measures Prenatal Questionnaire Self-reported attachment to fetus 7 Maternal-Foetal Attachment scale (Cranley, 1981) Maternal Antenatal Attachment Scale (Condon, 1993) Prenatal Attachment Inventory (Muller, 1993) Interview Representation of the infant, characteristic of parent 7 Working Model of the Child Interview (Zeanah, Benoit & Barton, 1986) Interview of Maternal Representations During Pregnancy-revised (Ammaniti et al., 1999) Five Minute Speech Sample (Magana, 1986) Pregnancy Interview (Slade et al., 2004) Postnatal Observation Free play 9 Ainsworth Sensitivity Scales (Ainsworth, Bell & Strayon, 1974) Adapted – Bornstein (2008) & Piccinini (2007) NICHD-SECCYD sensitivity scales (Owen, 2002) Meins & Fernyhough (2006; 2010) Adapted – Lewis and Lee-Painter (1974) Adapted – Ainsworth (1971), Lyons-Ruth (1983) & Crittenden (1981) CARE Index (Crittenden, 1981) Mother Infant Coding System (adapted Miller et al, 2002) Structured task 5 Favez (2011) family alliance assessment scale AMBIANCE (Bronfman, 1999) Feeding Scale (Chatoor, 1997) Erickson et al. (1990) Nursing Child Assessment Feeding Scale (Barnard, 1978a) PARENTING BEFORE BIRTH? Table 3 Mean Effects for the Association between Prenatal Thoughts and Feelings about the Unborn Infant and Postnatal Parenting for All Studies and Moderator Analyses Contrast M k z p 95%CI All studies .24** 14 3.40 .000 [.10, .38] Parent Gender Q (1, 12) = 6.32, p = .012 Fathers -.01 4 -.12 .907 [-.24, .21] Mothers .32*** 10 -.4.61 .000 [.19, .46] All 1st time parents Q (1, 12) = 2.78, p = .095 No .34 9 3.56 .011 [.06, .48] Yes .09 5 .71 .644 [-.15, .32] Risk sample Q (1, 12) = .29, p = .591 No .21 9 2.45 .015 [.05, .38] Yes .29 5 2.50 .013 [.06, .51] Prenatal measure Q (1, 12) = .16, p = .689 Questionnaire .27 7 2.54 .011 [.06, .48] Interview .21 7 2.12 .034 [.02, .41] Sensitivity outcome Q (1, 12) =.17, p = .683 No .27 6 2.45 .013 [.06, .49] Yes .22 8 2.37 .018 [.04, .39] Note. CI = confidence intervals. ** p < .01. *** p < .001.