The formation of the bond between mother and child, described in terms of bonding and attachment is a major topic in the field of psychology, since it is an elementary part of a child’s development. Bowlby and Ainsworth led the research into the history of this topic by respectively developing the biological basis of the attachment theory and introducing the affective domain [1,2]. Secure attachment is known to be beneficial in childhood development [3]. There is confusion between maternal-infant bonding and attachment, both being defined in equal terms and used interchangeably in the literature [4]. In general, attachment has a reciprocal aspect [5] or refers to the emotional tie between infant and parent [4]. A principle-based concept analysis of the concept of maternal-infant bonding states that maternalinfant bonding describes ‘maternal feelings and emotions towards the infant’. Operationalization takes place in the affective domain; up until now, only limited evidence exists to show that it also encompasses behavioural or biological components [4].

The development of the bond between mother and child already starts during pregnancy [6] and is known as antenatal, prenatal, maternal-fetal attachment or bonding. Prenatal attachment is defined by Condon and Corkindale [7] as ‘the emotional tie or the bond which normally develops between the pregnant parent and her unborn child’. During the second half of pregnancy in particular, the fetus becomes more human to the mother [2] and, in the third trimester of pregnancy, mothers are able to picture their maternal identity and their baby [8]. They have stronger feelings of attachment to their child when their realization of actual life inside them increases [9]. Prenatal attachment is an abstract concept, representing the affiliative relationship between parent and fetus, containing the core experience of love [2]. Love, as a neurophysiological construct, creates phylogenetic changes in the autonomic nervous system, which promote the establishment of enduring bonds [10]. It is known that mothers associate their (unborn) child with positive feelings, such as love, warmth, joy and happiness [11]. Higher prenatal attachment is associated with better childhood development [12].

The given definition of prenatal attachment is also applicable to the postnatal period and embodies the affective domain. It suggests the direction the relationship takes; from parent to infant, not requiring a response of any kind from the child, which actually makes it the definition of bonding. For this reason, in the current study, the term bonding is used to describe and measure the mother’s feelings and emotions towards her (unborn) baby. However, as the terms bonding and attachment are interwoven and used interchangeably in the literature, in the following part of the introduction, factors associated with both attachment and bonding will be discussed.

Studies investigating the predictors of prenatal bonding often found that maternal depression and bonding are negatively related, showing depression to be a significant predictor of poor prenatal bonding. Moreover, an inverse relationship was found between severity of depression and prenatal bonding, and also, the closer the prenatal bonding, the less symptoms of depression reported by mothers during the final term of pregnancy and postpartum. Poor prenatal bonding seems to predict postpartum anxiety and depression even more than prenatal anxiety or depression. On the other hand, social support and marital relationship are positively related to prenatal bonding. Mothers with a partner report higher levels of prenatal bonding than single mothers and married women with low prenatal bonding scores have low levels of social support outside the partner relationship and high levels of control, domination and criticism within the partner relationship. A meta-analysis showed that social support is the most powerful theoretical predictor studied in relation to prenatal bonding, but that this effect is moderate.

In contrast to studies investigating the correlates of prenatal bonding, there are fewer studies on the correlates of postnatal bonding and a number of factors have been clearly identified to be negatively correlated to postnatal bonding. Maternal postnatal depression has a strong negative association with postnatal bonding. On the other hand, social and partner support were found to be positively correlated to postnatal bonding. Postnatal bonding is known to be the best predictor of bonding later on in life. Moreover, it is assumed that pre- and postnatal bonding is interrelated. This relationship has been investigated in several studies, with results showing modest to moderate correlations; women with low prenatal bonding scores tend to remain in this category for postpartum bonding. However, bonding across pre- and postnatal periods has not been well researched, with bonding typically only being assessed at one time point during pregnancy or after birth. And therefore, the important limitation in all these studies is that, when measuring bonding pre- and postnatally, different instruments are used during pregnancy compared to postpartum. This makes it impossible to examine mean changes in bonding scores over time. In pregnancy, the most frequently used instruments are the Prenatal Attachment Inventory (PAI), Maternal Fetal Attachment Scale (MFAS) and Maternal Antenatal Attachment Scale (MAAS). These all have multifactorial structures and show good reliability, but not for all subscales. Moreover, various studies that used one of the questionnaires found different factor structures within the scales [13]. Several postpartum instruments have been developed, the most common being the Postpartum Bonding Questionnaire (PBQ), Maternal-to-Infant Bonding Scale (MIBS), Parent-to-infant Attachment questionnaire (PAQ), Maternal Attachment Inventory (MAI) and Maternal Postpartum Attachment Questionnaire (MPAS) [13]. All these mentioned that postpartum bonding questionnaires have multifactorial structures with good reliability [13]. However, no factor structure is known with regard to the MIBS.

Apart from the fact that different instruments were used to measure pre- and postnatal bonding, another limitation to these earlier studies is the relatively small sample size investigated in most studies. Therefore, researchers from previous studies point out that more longitudinal research is needed in a larger study population. But firstly, the development of an instrument for measuring bonding both pre- and post-natally is necessary. This could provide more insight into motherchild relationships over time, and could be helpful in detecting mothers with a higher risk of developing problems with bonding, which, in turn, is a predictor of childhood development. According to reviews, early postnatal interventions that enhance bonding, especially those focusing on behaviour, range between marginally successful and quite successful in improving the mother-child relationship. According to the review by Bakermans et al. [14], the best time for improving bonding is at six months after birth or later. However, studies show contradictory results and more research is needed. A review on the effectiveness of prenatal interventions also yielded contradictory results [15] and, to the best of our knowledge, the question of whether early preventive interventions that enhance bonding during pregnancy can be successfully implemented in order to develop an optimal mother-child relationship after birth, has not yet been answered. A single instrument for measuring pre- and postnatal bonding is needed in order eventually to be able to measure the effects of preventive interventions.

Therefore, the aim of the current study was to develop and investigate the psychometric properties of a user-friendly questionnaire for assessing maternal bonding during pregnancy as well as postpartum. We hypothesized that pre-natal and post-natal bonding scores were highly correlated, if measured with the same instrument. Moreover, we also hypothesized that bonding scores correlated negatively with depression and positively with partner support (concurrent validity).

Participants and procedure

Between January 2013 and September 2014, pregnant women who had their first antenatal appointment in the first trimester at the offices one of the 17 participating community midwives in the southeast of the Netherlands, were invited to participate in the HAPPY study (Holistic Approach to Pregnancy and the first Postpartum Year), the design of which has been described in detail elsewhere. During the 19-month recruitment period, 3160 Dutch-speaking Caucasian pregnant women who visited the participating midwives and who met the inclusion criteria (i.e., singleton pregnancy, no diagnosis of severe psychiatric illness or endocrine disorder) were approached. Written informed consent to participate was obtained from 2275 (72%) of the eligible women. These women completed an online questionnaire at several moments during pregnancy and postpartum. We later decided to measure bonding during the HAPPY project, and since this was not included in the original planning, only 1292 women are included in this study, 1163 (90%) of whom returned the questionnaires. During pregnancy, 113 women failed to return complete data, resulting in a final sample of 1050 women for data analysis of whom the characteristics are shown in Table 1. These were similar to those of the total HAPPY cohort of 2275 women (data not shown). The characteristics of the sample participating in this study were similar to those in the 2013 national obstetric register (The Netherlands Perinatal Registry, PRN). However, 64% of the women in the current study were highly educated, which is a higher percentage than that reported by Statistics Netherlands in women in their thirties [16]. Eindhoven, the area where the study was conducted, was nominated in 2011 as the world’s intelligent community of the year [17]. The 1050 participants were randomly divided by SPSS into two subsamples. Data from sample I (n=521) were used to conduct an exploratory factor analysis (EFA) and reliability analysis, while sample II (n=529) was used to perform a confirmatory factor analysis (CFA). The characteristics of these samples are shown in Table 1.

As can be seen, the two samples showed similar characteristics. The women had a mean age of 30, almost all of them were living with a partner and 64% were highly educated. Only with regard to paid jobs, χ² (1, n=1015)=4.7, p=0.031, phi=0.07, did the samples differ. However, a phi of -0.07 showed low effect size with little if any clinical relevance. Both samples met the criteria of four to ten subjects per item, with a minimum of 100 subjects to conduct factor analyses. Of this sample of 1050 women, 774 and 629 completed the questionnaires at eight and 12 months postpartum, respectively. The characteristics of these two postpartum samples were similar to the pregnancy sample (data not shown). These samples were used for test-retest analysis. The study was approved by the Medical Ethics Committee of the Máxima Medical Centre Veldhoven, the Netherlands.

Table 1: Characteristics of two samples of women participating in the HAPPY study (n=1050).

Measures

Pre- and post-natal bonding scale (PPBS)

Based on the literature, we decided to use only positive items describing bonding. It is known that mothers associate their (unborn) child with positive feelings and love is presumed to be the core experience of bonding [2]. Research and clinical practice usually focusses on the lack of positive, warm and caregiving maternal feelings towards the (unborn) baby and it was found that, in particular, low scores for positive statements signal a lack of bonding or a disordered relationship [18]. Moreover, in order to avoid the ‘acceptability influence’ or ‘social desirability’ in mothers' answers, no negative items were included [19]. Our society does not accept negative feelings in early mother-child relationships, which results in women consciously avoiding negative answers to those kinds of statements, thereby making negative items less discriminating and reliable. Therefore, we developed a bonding scale consisting of 14 items describing positive feelings of bonding (Appendix). During pregnancy, the women were asked to complete the following statement: “There are many statements that reflect the feelings a pregnant woman may experience towards her unborn baby, especially when she is aware of the fetal movements. Would you kindly complete the following: During the last four weeks, I could best describe my feeling towards my baby as:...’’ Thereafter, the women were asked to complete the items using a four-point Likert answer scale, ranging from very much (1), much (2), hardly (3), to not at all (4). This 14- item self-rating scale was completed at 32 weeks gestation. Moreover, the women in the sub-sample who completed the entire HAPPY study completed the scale at eight and 12 months' postpartum. During the postpartum assessments and using the same items, the introduction to the questionnaire was adapted into: “We have a number of statements that reflect the feelings a mother may experience towards her baby. Would you kindly complete the following: During the last four weeks, I could best describe my feeling towards my baby as...’’ Total scores ranged from 14 to 56, with higher scores indicating more positive feelings of bonding.

Depressive symptoms

Depressive symptoms were assessed several times during pregnancy and the first year postpartum, including at 32 weeks' gestation and at eight and 12 months post-partum, using the Edinburgh (Postnatal) Depression Scale (EDS). This ten-item questionnaire has previously been validated in the Netherlands for use during the postpartum period and pregnancy [20]. During gestation, cut-offs of 11 during the first trimester and of ten during the second and third trimesters have been described, with a postpartum cut-off of ten. The EDS has been extensively used in perinatal research in over 40 countries, and has shown good psychometric properties. Total scores ranged from 0 to 30, with higher scores indicating more depressive symptoms. The EDS was used for concurrent validity analysis with the bonding scale.

Partner support

The Tilburg Pregnancy Distress Scale (TPDS) has been previously developed and validated for use during pregnancy and contains a domain ‘partner involvement’ [21]. This subscale measures the extent of perceived partner involvement in pregnancy using four items on a fourpoint Likert scale, with higher scores indicating greater involvement. This subscale was used for concurrent validity analysis with the bonding scale.

Baseline characteristics

Several baseline parameters were evaluated at 12 weeks gestation, including demographic, pregnancy-related and psychological characteristics (Table 1).

Statistical methods

Statistical analyses were performed using the Statistical Package for Social Sciences (SPSS version 22, IBM, Chicago, IL, USA). Parallel Analysis was performed using the MonteCarlo PA program. Confirmatory factor analysis was carried out using AMOS (version 18, IBM, Chicago, IL, USA).

Factor analyses

A principal component EFA in sample I was performed using the 14-item bonding scale for testing psychometric properties. Due to the multifactorial structures of existing bonding scales, we used rigorous criteria for multifactorial factor analysis, although the use of positive items only increased the likelihood of the scale having a one-dimensional structure. Since all items related to positive feelings, we expected that all potential factors would be correlated, and chose to use oblimin rotation. A Catell scree plot was used to select factors for retention. Moreover, Parallel Analysis was performed to randomly generate a data matrix with criterion values corresponding to the Eigenvalues from the EFA. Only Eigenvalues that exceeded the corresponding criterion values were retained [22]. As explained by Pallant [22], a subscale of less than three items is not advisable. Internal consistency analyses were conducted using Cronbach’s alpha for the total scale and possible subscales derived from factor analysis. A Cronbach alpha reliability statistic of ≥ 0.70 is considered to be the minimum acceptable criterion for an instrument’s internal reliability. In sample II, CFA was performed on the remaining items in the bonding scale. CFA was used to test the model fit of the factor structures found with EFA, assessing the comparative fit index (CFI), normed fit index (NFI), Tucker-Lewis Index (TLI), and the root mean square error of approximation (RMSEA). Adequate model fit can be assumed with a CFI ≥ 0.80, combined with an NFI ≥ 0.80, a TLI ≥ 0.80, and an RMSEA ≤ 0.05 for a good and ≤ 0.08 for an adequate fit. Finally, EFA was repeated in sample II to verify the factor structure found after CFA.

Concurrent and construct validity

To test for differences in characteristics between the two subsamples, χ² analyses were used for all dichotomous data. Differences in mean scores between samples I and II were analyzed using the t-test (twotailed). Since the two subsamples showed similar characteristics, the two samples were merged to determine concurrent validity. Concurrent validity of the bonding scale was tested by correlating this scale with the EDS and the partner involvement subscale of the TPDS (Pearson’s r correlations, two-tailed). For analysis with the subscale of the TPDS, women without a partner (0.9%) were excluded. Test-retest reliability of the bonding scale was analyzed by correlating pre- and post-natal scores. Differences in mean bonding scores at different measurement times were analyzed using one-way repeated measures ANOVA with post-hoc analysis.

Exploratory and confirmatory factor analysis

Skewness and kurtosis statistics showed that the scores on all 14 items were distributed normally. All assumptions for conducting principal components analysis were met. The Kaiser-Meyer-Oklin value was greater than 0.60 (0.93) and the Bartlett’s test of sphericity value was significant (p<0.001). EFA with oblimin rotation of the 14-item scale in sample I suggested two dimensions with Eigenvalues of 6.4 and 1.2, respectively, 42.9% and 9.1% of variance explained, and 54.5% total explained variance. The Cattel scree plot clearly suggested a one-factor solution, since there was only one factor above the break in the plot and retaining all factors above the break in the plot is recommended [22].

For item loadings, a cut-off score of 0.40 was used and a minimum difference of 0.20 when an item had two loadings. Items 7, 9, 13 and 14 did not discriminate and were omitted. This resulted in a twodimensional scale with Eigenvalues of 4.9 and 1.1, respectively, explaining 60.6% of total variance. The remaining ten items consisted of one factor with seven items and the other factor with only three items. Parallel Analysis showed two components with Eigenvalues exceeding the corresponding criterion values for a randomly generated data matrix of the same sample size (ten variables x 521). When this two-factor structure was tested using CFA in sample II, a poor model fit was found: CFI 0.69, an NFI of 0.70, a TLI of 0.73 and an RMSEA of 0.15 with a lower bound of 0.09. Therefore and according to the Cattel scree plot, in sample I, the EFA was repeated on the ten-item scale using a one-factor solution. This resulted in one dimension with an Eigenvalue of 4.9, explaining 49.4% of variance. This ten-item, onefactor scale was again tested using CFA in sample II, and showed a better but still inadequate model fit (CFI=0.82, NFI=0.86, TLI=0.84 and RMSEA=0.09, lower bound=0.08). However, a closer look at the model estimates revealed that items 3, 4, 6, 11 and 12 showed poor standardized residual co-variances. Therefore, these items were omitted from the model, resulting in a five-item scale with an excellent model fit: a CFI of 0.97, an NFI of 0.98, a TLI of 0.97 and an RMSEA of 0.06, with a lower bound of .03 and Cronbach’s alpha of 0.87. When this five-item model was tested again with an EFA with varimax rotation in sample II, a one-factor structure was found with an Eigenvalue of 3.4, explaining 67.1% of total variance (Table 2).

The items were recoded from 1-4 into 0-3, in order to have a total score range from 0–15. We repeated the EFA and noted similar results at both eight and 12 months post-partum with Eigenvalues of 2.8 and 2.7, respectively, and an explained variance of 56.8% and 53.5%, respectively (Table 2). At eight and 12 months post-partum, Cronbach’s alphas were 0.80 and 79, respectively.

Concurrent and construct validity

Since the characteristics of both subsamples were similar (Table 1), we merged these two samples for concurrent validity analysis (n=1050). Mean scores (SD) and the range of the five-item PPBS, the EDS and the subscale partner involvement of the TPDS are shown in Table 3, including the Pearson correlations between these scales.

Table 3 shows highly significant correlations between bonding scores, with medium effect sizes for prenatal and both postnatal measurement occasions. The postnatal bonding scores at eight and 12 months postpartum also correlated highly significantly with a large effect size (Cohen’s d=2.43). There was a significant inverse correlation between PPBS and EDS scores at 32 weeks gestation as well as at eight and 12 months postpartum with medium effect sizes. Moreover, the PPBS scores and the subscale partner involvement scores of the TPDS were significantly positively correlated with a medium effect size.

Table 2: Final five-item bonding scale with one-factor solution from factor analysis with varimax rotation in 529 (sample II) pregnant women with an appropriate model fit in CFA. (CFI: 0.97, NFI: 0.98, TLI: 0.97, RMSEA: 0.06, lower bound: 0.03).

Table 3: Correlation matrix including mean scores (SD) and range of PPBS, EDS and TPDS scales assessed at each trimester (n=1050).

One-way repeated measures ANOVA were conducted to compare bonding scores at 32 weeks gestation, and at eight and 12 months postpartum. The means and standard deviations are presented in Table 3. There was a significant difference between the mean scores at 32 weeks gestation and at eight and 12 months postpartum, respectively, F(2,578)=119, p<0.001, partial eta squared=0.29, but no significant difference between the scores at eight and 12 months post-partum.

Since there is currently no instrument in existence to measure both pre- and post-natal bonding, this study aimed to develop a user-friendly questionnaire for assessing maternal bonding during pregnancy and postpartum. Our results showed that a five-item bonding scale, the Pre- and Post-natal Bonding Scale (PPBS), had good psychometric properties, a one-factor structure with good internal consistency and excellent model fit in the CFA, for assessing bonding pre- as well as post-natally. With regard to the structure analyses with PPBS, all the assumptions for appropriate factor analyses were met: in both samples I and II, the sample size was large (>10 subjects per item), Cronbach’s alphas were >0.70 and the factor loadings of the retaining items were high (>0.40). CFA showed an excellent model fit. The content of each of the five remaining items on the bonding scale includes the concept of love; a person's feelings that arise, among other things, from natural ties and reveal themselves in warm affection and attachment. This corresponds to the only explanation of bonding to be found in the literature, which states that bonding contains the core experience of love [2]. Thereby, all five items have characteristics of affection; in line with all past research that operationalized bonding only exists in the affection domain. It would seem that the excluded items had different characteristics and measure individual theoretical constructs, such as, for example, patience and optimism.

Our results show significant correlations with medium effect sizes between pre- and postnatal bonding scores, and with a high effect size between both postnatal bonding scores (test-retest properties). Earlier research found a relationship between pre- and postnatal bonding, but in that research, bonding was measured with different instruments and small sample sizes were used. To the best of our knowledge, the current study is the first that has developed an instrument for measuring bonding both pre- and post-natally, using a large sample size. This creates the opportunity to make a reliable comparison between pre and postnatal bonding, by examining the mean changes in bonding scores over time. We found that bonding scores at eight and 12 months postpartum were significantly higher than bonding scores at 32 weeks gestation. However, there was no significant difference between the scores at eight and 12 months' postpartum. The prenatal process of bonding that grows with the increasing ‘realization of actual life inside them’ continues after birth. Possibly, the increasingly active role that the baby plays in the interaction with the mother after birth, and the growing trust in the co-regulation capacities in the mother-baby dyad, stimulate feelings of maternal bonding which then stabilize in the postnatal period.

Previous research has suggested that bonding scores correlate negatively with depression and positively with partner support. Firstly, in the current study, at all measurement times, an identical significant negative correlation between bonding and depression was found with a medium effect size, indicating a relationship between bonding and depression in the expected direction. However, the effect size was modest, confirming that the PPBS and the EDS measure different concepts. The negative association between depression and bonding could be explained by the less emotional availability depressed mothers experience, both during pregnancy and after birth and therefore why they are less able to bond well. Moreover, it is assumed that depression during pregnancy compromises a women’s ability to feel confident in her new role as an expectant mother. Postnatal bonding may also be impaired because depression has selective effects on different aspects of mother-infant interaction [23]. Women suffering from depressive symptoms show less positive interaction overall, with a reduction in affection, responsivity and sensitivity, which is important in infantattachment. Also, during pregnancy, partner support and bonding were positively and significantly correlated, with a medium effect size. Earlier studies found that secure romantic attachment, which often results in higher partner involvement, is related to a better quality of attachment between mother and baby. Therefore, the partner’s reaction and attitude make a powerful contribution to the mother’s adjustment to pregnancy and to her relationship with her unborn child. The association between partner relationship and bonding implies that involvement in relationships is important in enabling the mother to bond with her (unborn) child. The importance of emotional support for the mother was also recognized by Stern [24], who described both the healthy primary preoccupation with forming a bond with the baby and the ability of the mother to gain emotional support, as one of the four basic themes of motherhood. Both the relationship with depression and partner involvement confirm the idea that a woman's emotional state and availability may have a significant impact on her ability to form a healthy bond with her infant.

This study has its strengths and its limitations. Its key strength is that it is the first to measure pre- and post-natal bonding with one and the same instrument, by means of which a large sample size - 1050 women (more than double that of any previous studies) - could be screened. This enabled us to use different samples for EFA and CFA: 521 and 529 women, respectively, in sample I and II. Another strength is the longitudinal use of the EDS to assess the concurrent validity of our newly developed scale, which enabled us to determine the relationship between bonding and depression at different time-points. However, several limitations should be mentioned. In general, in this field of research, it is hard to determine the abstract concept of bonding clearly and unequivocally. A standard definition is lacking, and the interchangeable use of the terms 'bonding' and 'attachment' is widespread. Consequently, no gold standard, such as a diagnostic interview, to measure bonding exists, and thus could not be used in developing the PPBS. Moreover, the current study only included Caucasian women while, in the Netherlands, up to 10-15% of women come from other ethnic groups. This means that the psychometric properties of the scale should be re-evaluated for women of other ethnic origins.

Using the PPBS will provide more insight into the mother-child relationship over time. It can be used as a screening instrument, due to its short duration and user-friendly setup. Such a reliable and validated scale for measuring pre- as well as postnatal bonding, makes it possible to investigate the predictive value of low prenatal bonding scores on postnatal bonding scores and, possibly, on infant attachment. In clinical practice, it can be helpful in detecting mothers with a higher risk of developing bonding problems. As we already know, depression has a negative relationship towards bonding, and it is of great importance to further investigate the relationship between depression, as well as any other psychological problems the mother may have, and bonding over time. Moreover, studies testing interventions to enhance bonding can be conducted, and changes in bonding scores can be measured in the preand postnatal phases. Postnatal interventions that enhance bonding appear to be effective, but more research is needed, also with regard to the best time to improve such bonding [14,25]. Since the review by Mercer et al. [15] only included studies carried out before1991, there is very little current knowledge on the effectiveness of preventive prenatal interventions, such as education on fetal behavior, maternalfetal interactive activity, awareness of fetal activity, and massage. The PPBS can also contribute to the investigation of the effectiveness of these kinds of early preventive interventions for enhancing bonding during pregnancy, with the aim of developing an optimal mother-child relationship after birth.

In conclusion, the five-item Prenatal and Postnatal Bonding Scale (PPBS) is a short user-friendly instrument with good psychometric properties; good factor structure, internal consistency, and concurrent validity. Future research should further elucidate the validity and use of this scale in clinical practice. The PPBS can be used pre- and postnatally for detecting risk groups of women with poor bonding abilities. In future, bonding scores can be related to infant attachment and child development. It may also be possible to implement preventive interventions to enhance bonding, in order to improve a child’s development.