Development of a revised and abbreviated version of the postpartum bonding questionnaire (PBQ‐R): First U.S. validation and association to child outcomes

Abstract

The postpartum bonding questionnaire (PBQ) is a maternal‐reported 25‐item measure of bonding, available in 15 languages, and widely used for clinical and research purposes in the United States (U.S.) and across the globe. Nonetheless, its putative 4‐factor structure initially proposed in 2001 has never generalized or been replicated in other samples, nor has it been studied in U.S. populations. Using a U.S.‐based sample of 610 English‐speaking mothers who completed the PBQ 4 months postpartum—mean 32.51 ± 5.25 years old and 47.5% first‐time mothers—the initial goal of this study was to confirm the 4‐factor/25‐item structure of the PBQ. Aligned with other published studies, our confirmatory factor analysis did not support the 4‐factor/25‐item structure. We then used exploratory factor analysis which supported the creation of a 1‐factor/14‐item abbreviated measure, the PBQ‐R. Unlike previous versions of the PBQ, the PBQ‐R is scored so that higher scores indicate stronger bonding. The validity of the PBQ‐R was supported by its high internal consistency in this sample (w = 0.89), and correlations with maternal depression (ρ = −0.46) and child neurodevelopment (ρ = 0.11 to ρ = 22) and socio‐emotional symptoms (ρ = −0.22 to ρ = −0.33). The new unidimensional shorter PBQ‐R is suitable for use in the U.S. as a measure of general mother‐infant bonding.

1. INTRODUCTION

Promoting the parent/caregiver‐child relationship, or early relational health (ERH), is now recognized as a priority by American and Canadian pediatric societies (Garner et al., 2021; Williams & Canadian Paediatric Society Early Years Task Force, 2023) as an essential promoter of child developmental outcomes (Garner et al., 2021; Shonkoff et al., 2021). Though decades of research consistently affirm the importance of ERH to the health and wellbeing of both parents and children (Atzil et al., 2018; Garner et al., 2021; Le Bas et al., 2020; Shonkoff et al., 2021), we still critically lack a global understanding of ERH phenotypes and their respective role in predicting and explaining ‘for who, when and how’ different child‐oriented life‐course outcomes develop (Dumitriu et al., 2023). To that end, valid and reliable measures that accurately capture distinct components of ERH are essential. Here, we focus on the measurement of parent/caregiver‐reported bonding and its relationship with maternal mental health as well as its capacity to predict child developmental outcomes in the first 3 years of life. Albeit often conflated (Hill & Flanagan, 2020; Wittkowski et al., 2020), “bonding” and “attachment” describe components of parent/caregiver‐child ERH that are, at the core, fundamentally different. Owing its origin to Klaus & Kenell (1982), four decades of research has somewhat led to an agreement that bonding is a mother‐driven concept describing the mother‐to‐infant emotional tie, from pregnancy and beyond birth (Barker et al., 2017; Bicking Kinsey & Hupcey, 2013; Hill & Flanagan, 2020). In contrast, attachment reflects children's internal working models of their caregiver (Cassidy & Shaver, 2016), and is typically assessed using the observer‐based Strange Situation Procedure (Ainsworth et al., 1979).

Bonding provides a critical window into maternal psychological and relational well‐being, with evidence highlighting that bonding difficulties and maternal postpartum depression often co‐occur (Morales et al., 2023; O'Dea et al., 2023). Most importantly, maternal bonding and psychological well‐being each has an undeniable relationship with child socio‐emotional functioning throughout development (Le Bas et al., 2020; Martucci et al., 2021; Prado et al., 2021; Sutherland et al., 2022). Therefore, valid and reliable measures that accurately capture the essence of mother‐to‐infant bonding create opportunities to adopt a relationship‐based perspective to studying child neurodevelopment and socio‐emotional functioning.

The Postpartum Bonding Questionnaire (PBQ) is the most widely used and researched mother‐reported screening instrument for bonding (Lavallée et al., 2022; Wittkowski et al., 2020). The PBQ was originally proposed in English by Brockington et al. (2001) and validated in 2006 (Brockington et al., 2006) for use in the United Kingdom (U.K.). Since its publication in 2001, the PBQ has been widely used for research and clinical purposes across the globe and cited over 1200 times (Brockington et al., 2006, 2001). Speaking to the importance and universal use of the PBQ, the questionnaire has been translated and validated in at least fourteen other languages, including Italian (Busonera et al., 2017), Portuguese (Baldisserotto et al., 2022; Nazaré et al., 2012; Saur et al., 2022), Japanese (Kaneko & Honjo, 2014; Ohashi et al., 2016; Suetsugu et al., 2015), French (Demanche et al., 2022), Spanish (Garcia‐Esteve et al., 2016), Tamil (Vengadavaradan et al., 2019), Dutch (Roxanne et al., 2022), Bengali (Edhborg et al., 2011), Chinese (Siu et al., 2010), German (Gobel et al., 2024; Reck et al., 2006; Steinbrueck et al., 2023), Hebrew (Hakhmigari et al., 2023), Korean (Han & Kim, 2023), Turkish (Dissiz et al., 2024) and Nepali (Pudasainee‐Kapri et al., 2024).

Despite its widespread adoption and otherwise acceptable psychometric properties (Wittkowski et al., 2020), there remains uncertainty over the factor structure, that is, the empirical item grouping that informs subscale scoring (Brockington et al., 2001). The initial factor structure from the 25‐item measure, often called PBQ‐25, originally yielded four factors (Brockington et al., 2006, 2001) defining general bonding disorders (factor 1), severe mother‐infant relationship disorders (factor 2), infant‐focused anxiety (factor 3), and risk of abuse (factor 4). Although the PBQ‐25 is often scored using the four factors listed above, this structure has yet to generalize successfully to other populations (Wittkowski et al., 2020). Other psychometric studies have suggested different factorial structures (Table 1), varying from 1‐factor to 4‐factor structures, and retaining from eight to all 25 items. Further, only one prior study examined the factor structure of the PBQ in an English‐speaking population from the United States (U.S.), reporting on the development of a shortened 10‐item version of the PBQ (Kinsey et al., 2014).

TABLE 1.

Characteristics of PBQ psychometric studies.

Study	Sample size	Country (language)	Infant age in months	Factors (items)	Statistical method	Extraction method
Saur 2022	2207	Brazil (Portuguese)	19.2	1 (8)	EFA	• Unclear
Gobel 2024 DREAM cohort	1968	Germany (German)	2	4 (21)	CFA / EFA Promax	• Scree plot
						• Parallel analysis
Gobel 2024 DREAM cohort	1968	Germany (German)	14	1 (24)	CFA / EFA Promax	• Scree plot
						• Parallel analysis
Kaneko 2016	1786	Japan (Japanese)	3	1 (16)	EFA	• Scree plot
Reck 2006	862	Germany (German)	0.5	1 (16)	PCA	• Scree plot
Garcia‐Esteve 2016	840	Spain (Spanish)	1–1.5	4 (24) and 1 (25)	CFA / EFA Promax	• Scree plot
Edhborg 2011	672	Bangladesh (Bangla)	2–3	4 (24)	PCA Varimax	• Scree plot • Eigenvalues
Current study COMBO cohort	610	U.S (English)	4	1 (14)	CFA / EFA Promax	• Scree plot
						• Eigenvalues
						• Parallel analysis
Baldisserotto 2023	489	Brazil (Portuguese)	1–10	1 (12) and 3 (12)	EFA / CFA Oblimin	• Parallel analysis
Ohashi 2016	392	Japan (Japanese)	1	3 (25)	EFA / CFA Promax	• Scree plot
Steinbrueck 2024	363	Germany (German)	3–6	1 (15)	EFA Varimax	• Eigenvalues
Demanche 2021	353	France (French)	0–6	1 (22)	EFA Varimax	• Scree plot
Han 2023	308	Korea (Korean)	0–12	2 (25)	EFA / CFA Varimax	• Eigenvalues
Hakhmigari 2023	301	Israel (Hebrew)	2	2 (14)	PCA / CFA Oblimin	• Parallel analysis
						• Comparison data etc.
Gobel 2024 MARI cohort	281	Germany (German)	2	4 (21) and 1 (24)	CFA / EFA Promax	• Scree plot
						• Parallel analysis
Roxanne 2022	254	Belgium (Dutch)	1.5–12	1 (21)	PCA	• Scree plot
Dissiz 2024	250	Turkey (Turkish)	2–12	1 (21)	EFA / PCA Varimax	• Scree plot
Vengadavaradan 2019	250	India (Tamil)	1–6	5 (19)	EFA Varimax	• Scree plot
Suetsugu 2015	244	Japan (Japanese)	1	4 (14)	EFA Promax	• Unclear
Nazaré 2012	229	Portugal (Portuguese)	6	1 (12)	CFA Varimax	• Unclear
Wittkowski 2010	132	U.K. (English)	3	3 (22)	CFA / EFA Varimax	• Scree plot
Pudasainee‐Kapri 2024	128	Nepal (Nepali)	1–6	3 (16)	EFA Orthogonal	• Parallel analysis
Busonera 2017	123	Italy (Italian)	3	3 (25)	PCA Oblimin	• Scree plot • Eigenvalues
Brockington 2001	104	U.K. (English)	Unclear	4 (25)	PCA Varimax	• Unclear

Note: Studies are listed from largest (top) to smallest (bottom) sample size.

Abbreviations: CFA: confirmatory factor analysis; EFA: exploratory factor analysis; PCA: principal component analysis.

Herein, we used data from a U.S.‐based cohort of mother‐infant dyads to optimize the use of the PBQ in the U.S. Specifically, we aimed to confirm the original 25‐item/4‐factor structure (PBQ‐25) proposed by Brockington et al. (2006). We then assessed the factor structure that best fits our data and describe a revised version of the PBQ (PBQ‐R). We subsequently used both versions to assess the relationship of maternal bonding with maternal depression in the first 4 months postpartum and maternal bonding at 4 months postpartum with child developmental outcomes throughout the first 3 years of life.

2. METHODS

2.1. Participants

This study is a secondary analysis of data obtained from the COVID‐19 Mother Baby Outcomes (COMBO) Initiative and the Epidemiology of Severe Acute Respiratory Syndrome Coronavirus 2 in Pregnancy and Infancy (ESPI) COMBO (ESPI COMBO) sub‐study. As part of these two parallel studies, we enrolled mother‐infant dyads prenatally or up to 4 months postpartum from 3 U.S. academic hospital centers and followed the health and wellbeing of mothers and infants through both cross‐sectional and longitudinal study designs. Mothers were invited to complete surveys via electronic self‐administration (REDCap) at intake (∼birth), 4, 6, 18, 24, and 36 months postpartum. Assessment timepoints were designed to coincide with ages typically associated with the attainment of social engagement skills (4 months) and developmental milestones (6, 18, 24 and 36 months). In the first 4 months of life, infants rapidly develop the capacity to engage and sustain face‐to‐face interactions (Tronick, 1989), and social interaction skills by 4 months are already predictive of later social functioning (Beebe et al., 2010). Here, we sought to assess maternal bonding at 4 months to reflect mother's feelings toward their baby as they are developing these critical social engagement skills.

Key Findings

• We introduce the PBQ‐Revised (PBQ‐R), a more concise and less burdensome 14‐item, single‐factor measure of postpartum bonding supported by exploratory factor analysis in a large U.S.‐based sample.

• Unlike previous iterations of the PBQ, the PBQ‐R is scored so that higher scores reflect stronger mother‐to‐infant bonding, allowing for a more intuitive interpretation.

• In this sample of U.S.‐based mothers, stronger bonding as measured by the PBQ‐R is associated with lower maternal postpartum depression and better child developmental and socio‐emotional outcomes through 36 months of age.

Statement of Relevance

Valid and reliable measures of mother‐to‐infant bonding create opportunities to adopt a relationship‐based perspective to studying child neurodevelopment and socio‐emotional functioning. The Postpartum Bonding Questionnaire (PBQ) is the most widely used, researched and translated parent/caregiver‐reported screening instrument for bonding, yet there remains ambiguity over its scoring system. Results support the use of a shortened version of the PBQ (PBQ‐R), poised to be of interest to U.S.‐based researchers and clinicians who are seeking a short, low‐burden assessment of maternal postpartum bonding.

Enrollment and study procedures have previously been described (Firestein et al., 2023; Shuffrey et al., 2022). Though mothers were initially enrolled to study the effect of in utero SARS‐CoV‐2 exposure on maternal and child outcomes, published results from the cohort do not support any association so‐far (Ahmed et al., 2025; Firestein et al., 2024, 2023; Shuffrey et al., 2022). Here, we included 690 mothers who completed the PBQ at 4 months postpartum and gave birth between May 2020 and April 2022 at 3 U.S. sites in New York, NY, Birmingham, AL, and Salt Lake City, UT. Study procedures were approved by the Columbia University Irving Medical Center (CUIMC) Institutional Review Board and written informed consent was obtained from all participants.

2.2. Measures

2.2.1. Original 25‐item postpartum bonding questionnaire

At 4 months postpartum, mothers completed the 25‐item Postpartum Bonding Questionnaire (PBQ‐25), in English or Spanish. English respondents completed the PBQ‐25 originally validated in the U.K. by Brockington et al. (2006), and Spanish respondents completed the PBQ‐25 translated and validated in Spain by Garcia‐Esteve et al. (2016). Consistent with the PBQ‐25 administration, items were scored on a 6‐point Likert scale (0‐Always to 5‐Never). It is worth noting that in Brockington et al. (2001), item 8 was “My baby irritates me” and item 10 was “I love my baby to bits”. In Brockington et al. (2006), items 8 and 10 were inverted. Here, we use Brockington et al. (2006)’s item numbering where item 8 is “I love my baby to bits” and item 10 is “My baby irritates me”. The PBQ‐25 comprises of four subscales used to screen for general bonding disorders (12 items [1, 2, 6, 7, 8, 9, 10, 12, 13, 15, 16, 17], cut‐off ≥ 12), severe mother‐infant relationship disorders (7 items [3, 4, 5, 11, 14, 21, 23], cut‐off ≥13), infant‐focused anxiety (4 items [19, 20, 22, 25], cut‐off ≥10), and risk of abuse (2 items [18, 24], cut‐off ≥2) per Brockington et al. (2006). The total cumulative score, ranging from 0 to 125, is traditionally used to screen for general bonding disorders (cut‐off score ≥26), and severe bonding disturbances (cut‐off score ≥40).

2.2.2. Patient health questionnaire‐9

At 4 months postpartum, mothers also completed the 9‐item Patient Health Questionnaire (PHQ‐9), one of the most reliable depression screeners (Kroenke et al., 2001). The PHQ‐9 is a short questionnaire scored on a 4‐point Likert scale (0‐Not at all to 3‐Nearly every day). The cumulative total score ranges from 0 to 27 and serves as an indicator of depression severity (no depression [≤4], mild depressive symptoms [5–9], moderate depressive symptoms [10–14], moderately severe depressive symptoms [15–19], and severe depressive symptom [20–27]). Internal consistency of the PHQ‐9 in our data was good (McDonald's Omega [ω] = 0.88).

2.2.3. Ages and stages questionnaire‐3^rd edition

As part of the COMBO Initiative's 6‐ and 18‐month assessments, mothers reported on their child's development using the developmentally appropriate Ages and Stages Questionnaire‐3^rd Edition (ASQ‐3) form, typically used to screen for delays across 5 domains, that is, communication, gross and fine motor, problem solving, and personal‐social (Squires & Bricker, 2009). The ASQ‐3 is a 30‐item questionnaire rated on a 3‐point Likert scale screening for the achievement of age‐appropriate developmental milestones (0‐not yet, 5‐sometimes, 10‐always). Domain specific cut‐off scores are established by the developers (Squires & Bricker, 2009) to indicate the need for monitoring or presence of developmental delays. On the 6‐month assessment, risk for developmental delays is indicated with a score ≤ 29.65 for communication, ≤ 22.25 for gross motor, ≤ 25.14 for fine motor, ≤ 27.72 for problem solving, and ≤ 25.34 for personal‐social. On the 18‐month assessment, risk for developmental delays is indicated with a score ≤ 13.06 for communication, ≤ 37.38 for gross motor, ≤ 34.32 for fine motor, ≤ 25.74 for problem solving, and ≤ 27.19 for personal‐social. Internal consistency for the 6‐month survey was low for the communication (ω = 0.64) and personal‐social (ω = 0.68) domains, and acceptable on all other domains (ω = 0.75–0.79). On the 18‐month survey, internal consistency was acceptable to good on all domains (ω = 0.74–0.88). Internal consistency on the ASQ in our data is consistent with published data (Squires & Bricker, 2009).

2.2.4. Child behavior checklist

At 24 and 36 months, mothers reported on their child's early psychopathology symptoms using the Child Behavior Checklist (CBCL 1.5–5), a screening tool for behavioral and emotional problems (Achenbach & Rescorla, 2001). The CBCL is a 99‐item questionnaire scored on a 3‐point Likert scale (0‐Not true to 3‐Very true or often true). The items are typically summed to form syndrome scales which are combined into two domain scores (internalizing and externalizing symptoms) and one total problems score (Mansolf et al., 2022). Internal consistency of the CBCL total problems score in our data was excellent (ω = 0.94).

2.2.5. Sociodemographic data

Sociodemographic data (i.e., race, ethnicity, insurance, education, study site) and birth‐related characteristics (i.e., parity, gestational age, mode of delivery, infant sex, maternal age at delivery, breastfeeding) were self‐reported by mothers and/or abstracted from their electronic medical record.

2.3. Statistical analyses

We first examined the demographic characteristics of the sample and the PBQ‐25 scale scores using descriptive statistics. For a comprehensive description of the PBQ‐25 factor scores and to screen for normality of the distributions, we examined medians (inter‐quartile range [IQR]), means (± standard deviation), kurtosis and skewness (Tabachnick & Fidell, 2019), and the Shapiro‐Wilk normality test. The main analyses were limited to mothers who completed the PBQ‐25 in English. Before conducting the below mentioned factorial analyses, consistent with Brockington et al. (2006)’s scoring, 17 of the 25 items were reverse scored (2, 3, 5, 6, 7, 10, 12, 13, 14, 15, 17, 18, 19, 20, 21, 23 and 24), such that higher scores were indicative of worse bonding. Statistical analyses were conducted using the lavaan (version 0.6.19) and psych (version 2.4.12) packages in R (Rosseel, 2012).

The aim of this study was to confirm the 25‐item/4‐factor structure (PBQ‐25) originally proposed by Brockington et al. (2006). We therefore tried to replicate the structure using confirmatory factor analysis (CFA) with maximum likelihood estimation. Global fit indices, that is, chi‐square (χ ² p ≥0.05), comparative fit index (CFI ≥0.90), Tucker‐Lewis index (TLI ≥0.90), root mean square error of approximation (RMSEA ≤ 0.08), standardized root mean residual (SRMR ≤ 0.08), relative chi square (χ²/df ≤ 3), and item loadings (≥0.40), were considered when evaluating model fit (Tabachnick & Fidell, 2019).

When, as described in the Results below, the CFA suggested poor model fit, we created a revised version of the PBQ (PBQ‐R) and evaluated its psychometric properties. We investigated the factor structure of the PBQ in our sample using a series of exploratory factor analyses (EFA) and the minimum residual method for extraction. Prior to running the EFA, and in contrast with the traditional scoring of the PBQ‐25, we reverse scored the positively formulated items (1, 4, 8, 9, 11, 16, 22 and 25; 0‐Never to 5‐Always), such that a higher score indicates a stronger bonding. Though this linear transformation has no impact on within‐variable variance, cross‐variable correlations and derived factors scores, it allows for a more intuitive interpretation of the data, that is, a higher score indicates stronger bonding. Bartlett's test of sphericity (p < 0.05) and the Kaiser‐Meyer‐Olkin (KMO ≤ 0.80) measure of sampling adequacy were used to determine suitability of data for factorializability (Tabachnick & Fidell, 2019). The number of factors to extract was based on scree plot inspection, parallel analysis and conceptual appraisal. Items with factor loadings (λ) ≥0.40 were considered meaningful and were therefore retained (Floyd & Widaman, 1995). We applied rotation (Promax) to increase interpretability of loadings. The internal consistency of the scale and its factors were assessed using McDonald's Omega (ω) coefficients (Dunn et al., 2014). Omega coefficients ≤ 0.69 were considered low, coefficients ranging from 0.70 ≤ ω≤ 0.79 were considered acceptable, 0.80 ≤ ω≤ 0.89 were considered good, and ω≥0.90 were considered excellent (DeVellis, 2017).

To further investigate the validity of our revised version of the PBQ resulting from the EFA (the PBQ‐R) we examined the association between maternal bonding and depression at 4 months postpartum; and between maternal bonding at 4 months postpartum and child developmental outcomes throughout the first 3 years of age. Using Spearman correlations (ρ), we further explored the convergence of the PBQ‐25 and PBQ‐R independently with maternal depression with the PHQ‐9 at 4 months postpartum, child development with the ASQ‐3 at 6 and 18 months, and child behavior problems at 24 to 36 months. Finally, using the full sample combining mothers who completed the PBQ in English and those who completed in Spanish, we evaluated measurement invariance across language of completion of the PBQ‐R (Putnick & Bornstein, 2016). This analysis was conducted in Mplus v.8.6 (Muthén & Muthén, 1998–2017).

3. RESULTS

3.1. Sample characteristics

At 4 months postpartum, 610 mothers across NY (n = 331, 54%), UT (n = 207, 34%), and AL (n = 72, 12%) completed the PBQ in English, and 80 NY‐based mothers completed the PBQ in Spanish (Table 2 for demographic characteristics).

TABLE 2.

Demographic characteristics.

		Full sample	English sub‐sample	Spanish sub‐sample
		N = 690	N = 610	N = 80
Race and ethnicity	n (%)
Asian		35 (5.1)	35 (5.7)	0
Black non‐Hispanic		46 (6.7)	46 (7.5)	0
Hispanic		235 (34.1)	163 (26.7)	72 (90.0)
Native American		3 (0.4)	3 (0.5)	0
Native Hawaiian		1 (0.1)	1 (0.2)	0
White non‐Hispanic		348 (50.4)	347 (56.9)	1 (1.3)
Other combinations		18 (2.6)	11 (1.8)	7 (8.8)
Declined		3 (0.4)	3 (0.5)	0
Missing		1 (0.1)	1 (0.2)	0
Maternal age at birth	M ± SD	32.51 ± 5.25	32.51 ± 5.16	32.57 ± 5.89
Commercial insurance (vs. Medicaid)	n (%)	468 (67.8)	466 (76.4)	2 (2.5)
Primiparous	n (%)	320 (46.3)	290 (47.5)	30 (37.5)
Vaginal delivery (vs. c‐section)	n (%)	444 (64.4)	403 (66.1)	41 (51.3)
Female infant (vs. male or missing n = 1)	n (%)	317 (45.9)	288 (47.2)	29 (36.3)
Infant age at PBQ (in months)	M ± SD	4.71 ± 1.46	4.71 ± 1.44	4.70 ± 1.60

3.2. Distribution of the PBQ‐25

Overall, mothers in this sample reported low levels of bonding disorders; 7.21% (n = 44) of mothers scored in the general bonding disorders range and 2.30% of mothers (n = 14) scored in the severe bonding disturbances range. Data did not follow a Gaussian distribution (factor 1: W = 0.93, p < 0.001; factor 2: W = 0.73, p < 0.001; factor 3: W = 0.90, p < 0.001; factor 4: W = 0.06, p < 0.001; Table 3 for full item‐ and factor‐level distribution). Internal consistency was poor for factor 3 (ω= 0.56), and not estimable for factor 4 due to the factor including less than 3 items.

TABLE 3.

Distribution of the PBQ‐25 completed in English (N = 610).

	Median (min–max)	Mean (SD)	Skewness	Kurtosis	Meeting cut‐off n (%)	Omega
PBQ 1	0 (0–3)	0.28 (0.52)	2.00	4.67
PBQ 2 *	1 (0–5)	0.78 (0.85)	1.12	1.68
PBQ 6 *	0 (0–5)	0.08 (0.36)	6.94	67.61
PBQ 7 *	1 (0–5)	1.19 (1.42)	1.36	1.15
PBQ 8	0 (0–5)	0.18 (0.66)	5.39	32.91
PBQ 9	0 (0–1)	0.04 (0.19)	4.73	20.38
PBQ 10 *	1 (0–4)	0.68 (0.75)	0.75	−0.19
PBQ 12 *	1 (0–4)	0.94 (0.84)	0.74	0.66
PBQ 13 *	0 (0–5)	0.75 (0.91)	1.24	1.56
PBQ 15 *	0 (0–5)	0.09 (0.39)	7.69	82.23
PBQ 16	0 (0–5)	0.19 (0.54)	3.77	19.07
PBQ 17 *	0 (0–2)	0.03 (0.19)	6.96	53.16
F1: General bonding disorders	5 (0–25)	5.21 (3.85)	0.99	1.73	44 (7.2)	0.74
PBQ 3 *	0 (0–2)	0.39 (0.60)	1.28	0.57
PBQ 4	0 (0–3)	0.19 (0.47)	2.72	8.48
PBQ 5 *	0 (0–5)	0.10 (0.39)	5.82	49.53
PBQ 11	0 (0–3)	0.35 (0.57)	1.52	1.82
PBQ 14 *	0 (0–2)	0.19 (0.45)	2.24	4.38
PBQ 21 *	0 (0–3)	0.34 (0.59)	1.60	1.70
PBQ 23 *	0 (0–5)	0.21 (0.56)	3.41	15.77
F2: Severe relationship disorder	1 (0–14)	1.77 (2.40)	1.84	3.81	2 (0.3)	0.84
PBQ 19 *	0 (0–5)	0.72 (0.88)	1.07	0.83
PBQ 20 *	0 (0–2)	0.04 (0.22)	6.06	39.97
PBQ 22	0 (0–5)	0.55 (0.78)	2.13	7.69
PBQ 25	1 (0–5)	1.07 (1.04)	1.31	1.99
F3: Infant‐focused anxiety	2 (0–10)	2.38 (1.99)	0.69	0.11	2 (0.3)	0.56
PBQ 18 *	0 (0–1)	0.00 (0.04)	24.58	603.01
PBQ 24 *	0 (0–2)	0.01 (0.12)	11.74	152.98
F4: Risk of abuse	0 (0–2)	0.01 (0.13)	10.77	129.12	1 (0.2)	NA
Severe bonding disturbances	8 (0–42)	9.38 (7.17)	1.15	1.75	14 (2.3)	0.87

Note: Higher score indicates lower bonding.

^*Reversed scored.

^† Scores at or above cut‐off indicate high‐risk for impairments.

3.3. Confirmatory factor analysis

We first tried to replicate the 25‐item/4‐factor structure initially proposed by Brockington et al. (2006). Although the model converged, the model was ill conditioned, and the associated information matrix couldn't be inverted. Model parameter standard errors could not be estimated, indicating poor fit of this factorial structure in our data (Kline, 2015).

3.4. Exploratory factor analysis

Given that the original 4‐factor structure did not fit these data and therefore failed to generalize, we examined the structure of the PBQ in our sample using a series of EFA. After removing item 18 (‘I have done harmful things to my baby’) because of low measure of sampling adequacy (MSA = 0.55) in the anti‐image correlation matrix (correlation between items after removing common variance), all assumptions for EFA were met (Bartlett's test: p < 0.001; KMO = 0.89). Upon scree plot inspection, there was a sharp drop in slope after the first factor, and the Eigenvalue method also supported 1 factor with one value > 1 (Figure 1). In contrast, the parallel analysis supported a 6‐factor structure (Figure 1). However, the 6‐factor configuration was ruled out because 4 of the 6 factors had 2 or less items with meaningful loadings λ ≥0.40 after rotation (eTable 1).

FIGURE 1.

Factor extraction through scree plot and parallel analysis.

The 1‐factor solution—hereafter the PBQ‐Revised (PBQ‐R)—accounted for 23% of the total variance and was characterized by items describing general mother‐to‐infant bonding feelings. After dropping items with loadings λ < 0.40, 14 items were retained (1, 2, 3, 4, 5, 10, 11, 12, 13, 14, 19, 21, 22, 23; Table 4). The possible score range is 0 to 70, with a higher score indicating stronger bonding. The distribution of the new 14‐item PBQ still did not follow a Gaussian distribution (mean 63.53 ± 5.65; median 65; interquartile range [IQR] 60–68; skewness ‐1.16; kurtosis 1.39; W = 0.88, p < 0.001), but internal consistency was high (ω= 0.89).

TABLE 4.

1‐Factor solution loadings.

	Factor 1
Item 1 * : I feel close to my baby	0.630
Item 2: I wish the old days when I had no babies would come back	0.564
Item 3: I feel distant from my baby	0.635
Item 4 * : I love to cuddle my baby	0.580
Item 5: I regret having this baby	0.478
Item 6: The baby does not seem to be mine	0.311
Item 7: My baby winds me up	0.161
Item 8 * : I love my baby to bits	0.210
Item 9 * : I feel happy when my baby smiles or laughs	0.327
Item 10: My baby irritates me	0.673
Item 11 * : I enjoy playing with my baby	0.625
Item 12: My baby cries too much	0.432
Item 13: I feel trapped as a mother	0.628
Item 14: I feel angry with my baby	0.597
Item 15: I resent my baby	0.334
Item 16 * : My baby is the most beautiful baby in the world	0.244
Item 17: I wish my baby would somehow go away	0.364
Item 18 ‡ : I have done harmful things to my baby	–
Item 19: My baby makes me feel anxious	0.613
Item 20: I am afraid of my baby	0.286
Item 21: My baby annoys me	0.698
Item 22 * : I feel confident when caring for my baby	0.427
Item 23: I feel like the only solution is for someone else to look after my baby	0.440
Item 24: I feel like hurting my baby	0.197
Item 25 * : My baby is easily comforted	0.331

Note: meaningful loadings (≥40) are bolded. Loadings extracted with principal axis factoring.

^*Item was reverse coded.

^‡ Item 18 was removed from the pool of items prior to running factorial analyses.

3.5. Consistency of results across language of administration

We estimated a CFA on item‐level data in the full sample combining English and Spanish responders (n = 690). Results showed that factor loadings were significant (p < 0.001) and λ ≥0.4 except for item 12 (λ = 0.373), and model fit was acceptable (χ²[77] = 485.865, p < 0.001, χ²/df = 6.31, CFI = 0.856, RMSEA = 0.088, SRMR = 0.056) suggesting a 1‐factor structure fit the data when combining language groups. However, we could not establish equivalence of factor loading patterns, that is, weak factorial invariance, across language groups (English/Spanish) because of the lack of variability of some items within the small group of mothers (n = 80) who reported on the PBQ in Spanish (Putnick & Bornstein, 2016).

3.6. Association between maternal postpartum bonding and depression

In the full sample including English and Spanish responders, most mothers reported no or mild depressive symptoms (n = 632; 93%), 6% (n = 42) reported moderate symptoms, and 1% (n = 6) reported severe symptoms. In the full sample including English and Spanish responders, we tested the association between maternal bonding and maternal depression measured with the PHQ‐9 at 4 months postpartum. As expected, both the PBQ‐25 (n = 680, ρ = 0.42, p < 0.001) and the PBQ‐R (n = 680, ρ = −0.46, p < 0.001) correlated significantly with PHQ‐9 scores; supporting the association between lower bonding and greater maternal depressive symptomatology.

3.7. Association between maternal postpartum bonding and child developmental outcomes

In the full sample including English and Spanish responders, we also tested the association between maternal bonding at 4 months postpartum and child developmental outcomes through 36 months of age. Both the PBQ‐25 and the PBQ‐R correlated with domains of child development in similar ways (eTable 2 for full table of correlations). To visually represent the similarity in associations between the PBQ‐25 and the PBQ‐R and these outcomes, we presented scatter plots in Figure 2 in which the x‐axis of the PBQ‐R is reversed. With this inversion, the directionality of the association is the same in the PBQ‐25 and PBQ‐R plots, facilitating interpretation. Higher maternal bonding measured at 4 months postpartum was associated at 6 months with more advanced communication skills (n = 522; PBQ‐25 ρ = −0.21, p < 0.001; PBQ‐R ρ = 0.22, p < 0.001), fine motor skills (n = 522; PBQ‐25 ρ = −0.17, p < 0.001; PBQ‐R ρ = 0.15, p < 0.001), and problem‐solving skills (n = 522; PBQ‐25 ρ = −0.13, p = 0.003; PBQ‐R ρ = 0.11, p = 0.01), as well as personal‐social skills at 18 mo (n = 313; PBQ‐25 ρ = −0.15, p = 0.006; PBQ‐R ρ = 0.15, p = 0.006). Lower bonding at 4 months postpartum was associated at 24 to 36 months with more child internalizing (n = 526; PBQ‐25 ρ = −0.21, p < 0.001; PBQ‐R ρ = −0.22, p < 0.001), externalizing symptoms (n = 526; PBQ‐25 ρ = 0.30, p < 0.001; PBQ‐R ρ = −0.33, p < 0.001) and total problem behaviors (n = 526; PBQ‐25 ρ = 0.29, p < 0.001; PBQ‐R ρ = −0.32, p < 0.001).

FIGURE 2.

Correlation between maternal postpartum bonding and child neurodevelopmental outcomes. Panel A—Original PBQ‐25 correlated with ASQ‐3 communication skills (n = 522, ρ = −0.21, p < 0.001) at 6 months, ASQ‐3 personal‐social skills (n = 312, ρ = −0.15, p = 0.006) at 18 months and CBCL total problems (n = 526, ρ = 0.29, p < 0.001) at 24–36 months. Panel B—PBQ‐Revised (PBQ‐R) correlated with ASQ‐3 communication skill (n = 522, ρ = 0.22, p < 0.001) at 6 months, ASQ‐3 personal‐social skills (n = 313, ρ = 0.15, p = 0.006) at 18 months and CBCL total problems (n = 526, ρ = −0.32, p < 0.001). X‐axis was reversed to visually represent similarity of the association between the PBQ‐25 and PBQ‐R with specified outcomes.

4. DISCUSSION

Early Relational Health (ERH) has widely been accepted as an early predictor of child neurodevelopmental outcomes (Garner et al., 2021; Williams & Canadian Paediatric Society Early Years Task Force, 2023). ERH is generally used as a blanket term for all constructs describing the relational characteristics between parents/caregivers and infants, like bonding, attachment, sensitivity, and emotional connection, regardless of their theoretical origins. Likewise, important heterogeneity exists in patterns of child development across neurodevelopmental (e.g., cognitive, motor, language), and socioemotional domains (e.g., early emerging psychopathology symptoms, prosocial behavior). The characterization and unique contribution of ERH constructs to the promotion and/or buffering of various child developmental outcome domains remains largely unexplored (Dumitriu et al., 2023). Therefore, prior to undertaking mechanistic investigations of key ERH constructs, here we sought to confirm the original factor structure of the PBQ and assess the validity of a revised version of the PBQ using a large sample of U.S.‐based mother‐infant dyads.

In alignment with other PBQ psychometric studies (Garcia‐Esteve et al., 2016; Nazaré et al., 2012; Wittkowski et al., 2020, 2010), the original 25‐item/4‐factor structure (Brockington et al., 2006) could not be confirmed in our sample of English‐speaking U.S.‐based mothers. Alternatively, following EFA, our data supported a single dimension representing general mother‐to‐infant feelings. Here, we argue that the added value of retaining structure beyond one factor seems modest and we suggest that a unidimensional scale with fewer items is more suitable for the measurement of mother‐reported bonding. From a conceptual perspective, bonding has consistently been described as a concept that encompasses all maternal feelings and emotions toward their infant (Barker et al., 2017; Bicking Kinsey & Hupcey, 2013; Hill & Flanagan, 2020).

Thus, for practical purposes and conceptual coherence, we advocate for use of the shorter and revised 1‐factor/14‐item structure (PBQ‐R). Although high variability in factor solutions and configurations of items retained (Figure 3) has been documented across languages, countries, and populations, other studies with adequate sample sizes, that is, 10 participants per item (n > 250), almost all support a single factor solution to screen for general mother‐infant bonding (Demanche et al., 2022; Dissiz et al., 2024; Garcia‐Esteve et al., 2016; Gobel et al., 2024; Kaneko & Honjo, 2014; Roxanne et al., 2022; Saur et al., 2022; Steinbrueck et al., 2023). These striking differences in factor solutions may be a function of methodological variability and quality across studies, for example, sample size, timing of assessment, statistical methods used, but they could also result from the lack of clear conceptual boundaries among different ERH outcomes or from the still poorly understood cultural influences in ERH perceptions and applications.

FIGURE 3.

Overview of PBQ factor items across psychometric studies. Distribution of the original 25 items included in factor structures across studies. Studies listed from largest (top) to smallest (bottom) sample size. Grey squares indicate retained items. Note: Itemization based on Brockington et al. (2006) with item 8 “I love my baby to bits”, and item 10 “My baby irritates me”. Studies using Brockington et al. (2001) itemization (items 8 and 10 interchanged) were switched here for consistency.

In contrast with other studies, we introduced a reversed scoring method of the PBQ, such that higher scores indicate a stronger mother‐to‐infant bond. Such linear transformation allows a more intuitive and strength‐based interpretation of the PBQ‐R score, facilitating and improving the clarity of clinical and research applications. By emphasizing the positive aspects of bonding rather than deficits, this approach aligns with a more strength‐based perspective on ERH.

Leveraging the longitudinal design of the COMBO Initiative and its embedded validated measures, we examined the association between postpartum bonding and maternal and child well‐being and functioning throughout infancy. Consistent with previous studies, we found an association between maternal postpartum bonding and maternal depression. Further, both the PBQ‐25 and PBQ‐R were associated with child neurodevelopment and socioemotional functioning as measured by the ASQ‐3 and the CBCL. This suggests that both the original PBQ‐25 and the shorter PBQ‐R administered in the postpartum period are similarly associated with neurodevelopmental and socioemotional skills through 3 years of age as reported by mothers. Nonetheless, these findings should be extended to evaluate associations between bonding and observational assessments of child development, and assessments reported by other significant caregivers.

Finally, data reported here were collected during the COVID‐19 pandemic. Although many studies have reported significant rises in maternal psychopathology during this period (Adams et al., 2021; Berthelot et al., 2020; Cameron et al., 2020; Connell & Strambler, 2021; Frankel et al., 2021; Grumi et al., 2021; Hessami et al., 2022; Lebel et al., 2020; Morris & Saxbe, 2022; Perzow et al., 2021; Tomfohr‐Madsen et al., 2021; Venta et al., 2021), our rates of maternal depression in our non‐clinical sample are low, that is, 93% reporting none or mild symptoms. These rates are comparable to other PBQ validation studies, reporting rates of maternal depression from 6 (Reck et al., 2006) and 7% (Demanche et al., 2022; Kaneko & Honjo, 2014; Suetsugu et al., 2015) to 10% (Edhborg et al., 2011), and our results are aligned with other pre‐pandemic studies with non‐clinical samples of mothers‐infant dyads (Demanche et al., 2022; Dissiz et al., 2024; Garcia‐Esteve et al., 2016; Gobel et al., 2024; Kaneko & Honjo, 2014; Roxanne et al., 2022; Saur et al., 2022; Steinbrueck et al., 2023). Still, this work should be replicated in a post‐pandemic sample of postpartum mothers in the context of increasing parental psychological strain in the U.S. general population (Office of the Surgeon General, 2024).

5. CONCLUSION AND IMPLICATION FOR CLINICAL PRACTICE AND FURTHER RESEARCH

ERH is an intergenerational intervention target that has strong clinical implications as a crucial driver of optimal child‐oriented life‐course outcomes (Garner et al., 2021) as well as maternal mental health outcomes in the peripartum period (Lavallée et al., 2022; Monk et al., 2022). Several prominent organizations, including ZERO TO THREE (2016), the Alliance for the Advancement of Infant Mental Health (2023) and the World Association for Infant Mental Health (Osofsky et al., 2024) have endorsed the assessment of ERH in early childhood spaces to orient efforts toward the promotion of infant mental health and development. Advancing measurement of mother‐infant bonding therefore has the potential to advance both perinatal, pediatrics and infant mental health clinical practice and research on intergenerational processes relevant to health and wellbeing for parents and their children.

The new unidimensional PBQ‐R has demonstrated good internal consistency and validity and is suitable for use in the U.S. as a measure of general postnatal mother‐infant bonding. In addition to being more concise and less burdensome for postpartum mothers, the performance of the PBQ‐R compares to the PBQ‐25 when looking at the association with maternal depression and measures of child development.

Nonetheless, as research efforts are oriented toward a better understanding of the synergistic relationship between different aspects of ERH, maternal health, and child development and wellbeing, it is clear that a stringent scale development and validation agenda is needed for robust characterization of each component of ERH (Dumitriu et al., 2023; Wittkowski et al., 2020).

CONFLICT OF INTEREST STATEMENT

No competing interests.

Supporting information

Supporting File 1: imhj70052‐sup‐0001‐eTable1.docx

Supporting File 1: imhj70052‐sup‐0001‐eTable2.docx

Lavallée, A. , Warmingham, J. M. , Reimers, M. A. , Curtin, P. , Kyle, M. H. , Austin, J. , Lee, S. , Barker, T. , Hussain, M. , Arduin, E. , Ahmed, I. , Atwood, G. , Ettinger, S. , Smotrich, G. , Turner, J. B. , Fisher, P. W. , Marsh, R. , & Dumitriu, D. (2026). Development of a revised and abbreviated version of the postpartum bonding questionnaire (PBQ‐R): First U.S. validation and association to child outcomes. Infant Mental Health Journal, 47, e70052. 10.1002/imhj.70052

DATA AVAILABILITY STATEMENT

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.