Hormonal Changes in First-Time Human Fathers in Relation to Paternal Investment

Abstract

In many biparental species, males experience hormonal changes across the transition to fatherhood that prepare them for their new caregiving role. In humans, cross-sectional comparisons have revealed hormonal differences between fathers and nonfathers, however it is not clear when hormone levels change in new fathers, how such changes relate to paternal involvement and attachment, or even whether men with particular hormonal profiles are more likely to become fathers. In this longitudinal study, we aimed to elucidate the trajectory of hormonal changes in new fathers with greater temporal resolution than previous studies, and to provide further insight into the direction of causality between hormone levels and paternal behavior by asking whether hormone levels predict behavior at subsequent time points, or vice-versa. We recruited a sample of 51 first-time expecting fathers and measured levels of plasma testosterone, oxytocin, vasopressin and cortisol at four time points spanning 4–5 months gestation through 4 months postnatally, and we compared these changes to those found in a control sample of 57 adult male nonfathers. We also examined the concurrent and lagged relationships between hormone levels and fathers’ self-reported paternal behaviors. From early in the prenatal period, fathers showed lower levels of both testosterone and vasopressin compared with nonfathers, and lower levels of these hormones during the prenatal period predicted greater postnatal paternal investment in the mother and child. Similar to what is known for human mothers, oxytocin levels increased across the gestational period in new fathers, but oxytocin levels were not associated with greater self-reported paternal involvement or attachment. Finally, cortisol levels did not differ between fathers and nonfathers, and were not predictive of paternal involvement or attachment. Our findings raise the possibility that low levels of vasopressin and testosterone found in expecting fathers anticipate a shift in life history strategy toward greater investment in parenting.

Introduction

Fathers care for their offspring in only 5–10% of mammalian species (Clutton-Brock, 1991; Moller, 2003). Among species exhibiting paternal care, the transition to fatherhood can involve changes in levels of hormones such as testosterone (T), oxytocin (OT), vasopressin (AVP), prolactin, estradiol and cortisol (Cort) (Bales and Saltzman, 2016; Storey and Ziegler, 2016). In some cases, these changes are correlated with paternal behaviors. For example, among marmoset monkey fathers, infant carrying is associated with both increased prolactin levels and decreased T levels (Dixson and George, 1982; Nunes et al., 2001).

Many men are involved in raising their offspring, so they too may experience hormonal changes that support their involvement in paternal caregiving. Cross-sectional comparisons of human fathers and nonfathers are consistent with this possibility, as new fathers tend to have lower T levels (Grebe et al., 2019; Meijer et al., 2019), and higher OT levels (Grumi et al., 2021). However, these cross-sectional studies cannot discern whether fatherhood causes hormonal changes from the alternative that men with particular hormonal profiles are more likely to become engaged or invested fathers. Within-subject longitudinal studies that follow men across their transition to fatherhood are more informative in this regard. One such study of men living in the Philippines conclusively demonstrated within-subject decreases in T in men who became partnered fathers across a four year interval (Gettler et al., 2011a). However, that study did not address the specific timing of the hormonal changes. Several subsequent longitudinal studies have not identified significant decreases in T across the perinatal period (Berg and Wynne-Edwards, 2001; Cárdenas et al., 2023; Diaz-Rojas et al., 2023; Perini et al., 2012). Others have documented modest within-subject decreases in T (5–6%) during the perinatal period (Bakermans-Kranenburg et al., 2022; Corpuz and Bugental, 2020; Edelstein et al., 2017), but the observed changes have not matched the magnitude of change in the study of Filipino men (26%) or the magnitude of differences between fathers and non-fathers in cross-sectional comparisons (Berg and Wynne-Edwards, 2001; Gray et al., 2006; Mascaro et al., 2014). These longitudinal studies have included a limited number of pre and/or postnatal measurements, suggesting that additional changes may occur outside of the measured time points.

Discrepancies between cross-sectional and longitudinal studies also exist for OT. One study reported a modest 8% increase across the first 6 months of fatherhood (Gordon et al., 2010), but this magnitude does not match reported cross-sectional differences between fathers and nonfathers on the order of 33% (Mascaro et al., 2014). The Gordon et al. study did not include any prenatal measurements, and may have therefore missed changes that occur earlier in the transition to fatherhood. Another study found no significant change in OT across the perinatal period, but measurements were restricted to late gestation and the early postnatal period (Bakermans-Kranenburg et al., 2022). Although AVP has been implicated in both rodent and non-human primate paternal care (Kozorovitskiy et al., 2006; Wang et al., 1994), one recent study found that AVP levels decreased from the late prenatal to the early postpartum period in new fathers (Bakermans-Kranenburg et al., 2022). Finally, one longitudinal study found that Cort levels are highest just prior to birth in new fathers (Storey et al., 2000), whereas another showed that both T and Cort were more likely to be low in partnered fathers compared with single men (Gettler et al., 2011b).

Variation in hormone levels among new fathers have also been linked with paternal behavior (Grebe et al., 2019). For example, fathers with lower T levels tend to be more involved in caregiving (Gettler et al., 2011a; Mascaro et al., 2013), to report more sympathy for newborn infant crying (Fleming et al., 2002), and to be more responsive to their infants compared with fathers with higher T levels (Weisman et al., 2014). On the other hand, fathers with higher plasma OT and AVP levels tend to engage in more stimulatory behaviors with their infants such as playful touch and moving their infant’s body through space (Apter-Levi et al., 2014; Gordon et al., 2010; Morris et al., 2021). However, these studies cannot establish the causal direction of these associations. That is, they cannot determine whether hormone levels cause paternal behavior or vice-versa. Better insight into causality comes from studies that measure fathers’ change in T from baseline in response to interacting with their infant. Although some studies have shown decreases in paternal T following interactions with their infant, a recent meta-analysis concluded that the overall effect of exposure to child stimuli on paternal T levels is not significant (Meijer et al., 2019).

In this study, we aimed to elucidate the trajectory of hormonal changes in new fathers with greater temporal resolution than previous studies, and to provide further insight into the direction of causality between hormone levels and paternal behavior by asking whether hormone levels at one time point predict behavior at subsequent time points, or vice-versa. Specifically, we recruited a sample of first-time expecting fathers and measured levels of plasma T, OT, AVP and Cort at four time points spanning 4–5 months gestation through 4 months postnatally, and we compared these changes to those found in a control sample of adult male nonfathers. We also examined the concurrent and lagged relationships between hormone levels and fathers’ self-reported paternal behaviors. Our pre-registered hypothesis was that OT and AVP levels will increase, and T levels will decrease, across the transition to fatherhood but not across the equivalent time period in the control sample of non-fathers, and that these changes will be correlated with paternal involvement and attachment. Note that the directionality of the AVP hypothesis was specified before publication of the recent study finding decreases in AVP across the perinatal period in new fathers (Bakermans-Kranenburg et al., 2022). Although Cort was not included in our hypothesis, the above evidence suggests it is expected to increase in fathers just before birth and to decrease postnatally.

Peripheral OT levels have been measured using a variety of different techniques, and concern has been raised that results using different techniques are often uncorrelated. Multiple different assays methods have been employed, including radioimmunoassay (RIA), enzyme immunoassay (EIA), and liquid chromatography-mass spectrometry (LC-MS/MS), and some labs perform an extraction step prior to assay, while others do not. A recent review article argued for the importance of sample extraction to eliminate interfering molecules, and noted the superior sensitivity of LC-MS/MS for measuring lower levels of OT that often fall below the limit of detection of RIA and EIA (Tabak et al., 2023). In this study, we use a newly developed LC-MS/MS method on extracted samples to measure all four hormones of interest.

Material and Methods

Participants

We recruited 51 men between 20 and 45 years of age (M = 32.5, SD = 3.43) whose female partner was pregnant with the man’s first child, along with 57 heterosexual men of the same age range (M = 28.7, SD = 4.80) who were not fathers or expecting fathers, but who hoped to become fathers in the future. All men were partnered and cohabitating with their female partner. Expecting fathers were enrolled between 16 and 20 weeks gestation, except for one participant who could not complete his first study visit until 25 weeks due to a transient health issue. Non-fathers did not have plans to conceive a child within the next year.

Subjects with a history of psychiatric illness (defined as a diagnosed DSM-5 disorder) over the past year, including alcohol use disorder and substance use disorder, were excluded since these may interfere with paternal behavior. We also excluded individuals with conditions that could be associated with abnormal brain function, including a history of seizures or other neurological disorders. Further, we excluded individuals who were currently caring for any children or elderly people in their home.

Participants were included only if they passed the Society of Magnetic Resonance Imaging standardized MRI screening protocol (exclusions for ferrous metal in any part of body, such as pacemakers, cochlear implants, surgical clips or serious medical conditions, claustrophobia).

Study Design

Expecting fathers were recruited to complete five study visits at the following timepoints: visit 1 between 4–5 months gestation, visit 2 at approximately 8 months gestation, visit 3 at approximately 1 month postpartum, visit 4 at approximately 4 months postpartum and visit 5 at approximately 12 months postpartum. Nonfathers completed study visits at similar intervals over 17 months. Blood and saliva samples, height, weight, and questionnaire responses were collected at each study visit. Afterwards, structural and functional MRI scans were also collected at visits 1, 4 and 5 only. Visit 5 data collection is still in progress. The current analysis is limited to hormone and questionnaire data collected at visits 1 – 4. Neuroimaging data will be reported in a separate publication.

Of the 51 fathers that were enrolled, 51 completed visit 1 (mid-gestation), 46 completed visit 2 (late gestation), 49 completed visit 3 (early postnatal), and 49 completed visit 4 (late postnatal). Two fathers dropped out of the study and some additional fathers skipped only visit 2 due to premature birth of their infant. Of the 57 nonfathers that were enrolled, 57 completed visit 1, 51 completed visit 2, 48 completed visit 3, and 48 completed visit 4.

Questionnaires and Anthropometrics

All participants were asked to provide information about their age, race, annual income, years of education, hours per week spent with their partner, and whether they owned a pet because interactions with pets can affect hormone levels (Nagasawa et al., 2015). Participants were also asked to complete the following questionnaires: Dyadic Adjustment Scale, a 32-item questionnaire that assesses an individual’s perceptions of his/her relationship with an intimate partner as a measure of relationship quality (Spanier, 1976); the Pittsburgh Sleep Quality Index, a 19-item questionnaire that evaluates sleep quality with higher scores indicating more acute sleep disturbances (Buysse et al., 1989); and the Edinburgh Postnatal Depression Scale, a 10-item scale that assesses common depressive symptomology (Cox et al., 1987). Finally, height and weight were measured to calculate the body mass index (BMI) for each participant.

For postnatal visits, fathers were asked to complete questionnaires measuring paternal behaviors including the Paternal Postnatal Attachment Scale (PPAS) (Condon et al., 2008) and the Paternal Involvement with Infant Scale (PIWIS) (Singley et al., 2018). These scales were selected based on their previously established reliability in evaluating paternal behaviors and dispositions toward their infants. The PPAS, a measure of fathers’ attachment to their infants, is measured as an average of the three subscales: patience and tolerance, pleasure in interaction, and affection and pride. The PIWIS asks fathers to report how often they take part in 35 different aspects of parenting, using a 7 point scale (1=not at all, 2=rarely, 3=1–2 times per month, 4=a few times per month, 5=a few times per week, 6=once a day, 7=more than once a day) and consists of five subscales. Warmth and attunement measures responsiveness and expressions of love (e.g., smiling at their baby, feeling close to their baby); frustration reflects resentment and jealousy towards caregiving responsibilities (e.g., feeling frustrated when caring for their baby); indirect care measures activities undertaken for the child, but not involving interaction with the child, with the exception of providing economic support (e.g., taking their baby to childcare or medical appointments); control and process responsibility measures fathers’ decision-making and initiative in monitoring infant needs (e.g., making decisions regarding their baby’s well-being, setting their baby’s schedule/activities); and positive engagement measures the amount fathers provide interactive care through activities that potentially promotes child development (e.g., feeding their baby, changing their baby’s diaper). Fathers were also asked how many weeks of paternity leave they took.

Specifics of structural and functional MRI data acquisition will be reported elsewhere.

LC-MS/MS Hormone Assays

Plasma concentrations of Cort, T, OT, and AVP were quantified by liquid chromatography-triple quadrupole tandem mass spectrometry (LC-MS/MS) in the Biomarkers Core Laboratory at the Emory National Primate Research Center. Aprotinin (500 KIU/ml of blood) was added to chilled EDTA tubes before blood collection to prevent the degradation of OT and AVP. All plasma samples were stored at - 80°C until analysis and analyzed in duplicate. All extraction and analytical protocols were performed by an experienced PhD-level chemist blinded to subject history and study information. Before LC-MS/MS analysis, supported liquid extraction (SLE) was used to extract T and Cort from subject plasma samples, while the combination of protein precipitation and solid-phase extraction (SPE) was used to extract OT and AVP. All extraction efficiencies were > 75%. Details of the extraction and LC-MS/MS analytical protocols can be found in Supplemental material. FDA-established acceptance criteria for the calibration curve, quality controls, assay sensitivity, precision, and accuracy for LC-MS/MS assays were followed (Food, 2018). Briefly, mean intra- and inter-assay precisions for all quality control samples were < 15%. Accuracies for all standard samples were ± 16% of their nominal value.

Analysis

Demographic data were compared between fathers and non-fathers using chi-square and two sample t-tests. Questionnaire and anthropometric data were compared between fathers and non-fathers across visits using linear mixed models. Paired t-tests were used to compare paternal behavior between postnatal visits 3 and 4.

To examine changes in OT, AVP, T and Cort levels with the transition to fatherhood, a longitudinal linear mixed effects model was fit that includes group (fathers, non-fathers), visit number (4 time points), and their interaction. The interaction term evaluates if changes across visits differ between fathers and non-fathers. The choice of a linear model was appropriate based on the distribution of the data (Supplementary Figure 1). To assess if the changes in OT, AVP, T and Cort are significantly associated with paternal involvement and attachment during infancy, a longitudinal linear mixed effects model including only fathers was fit. In this fathers-only model, the PIWIS and PPAS scores were entered as time-varying covariates and tested for interactions with visit number to assess whether changes are more pronounced in more involved or attached fathers. The following potential confounding variables were considered for inclusion in the models: Age, BMI, hours per week with partner, time of day for blood draw, sleep quality, relationship satisfaction, pet ownership, SES and race. These variables were included as covariates in regression models when they had significant associations with both fatherhood status and hormone levels.

This study was preregistered at: https://osf.io/c798k

Results

Demographic and Anthropometrics

Expecting fathers (M = 32.5, SD = 3.43) were on average 3.8 years older than non-fathers (M = 28.7, SD = 4.80; t(101.20)=4.71, p<.001; d=0.89), had greater combined income (t(77.64)=5.48, p<.001; d=1.10) and had slightly more years of education (t(106)=2.53, p=.013; d=0.49). Expecting fathers were also more likely to own a pet than non-fathers (86.27% vs. 47.37%, X²(1,108)=18.09, p<0.001, Φ=.41). Expecting fathers and non-fathers did not significantly differ in racial distribution (Fathers: 6 Asian, 4 Black, 4 Hispanic, 1 Multiracial, 36 White, Nonfathers: 10 Asian, 10 Black, 7 Hispanic, 1 Multiracial, 29 White, X²(4,108)=4.83, p=0.31, V=.21) (Table 1).

Table 1.

Demographic and Anthropometric Characteristics of Participants.

	Fathers (N)	Nonfathers (N)	p
Age	32.49 ± 3.43 (51)	28.74 ± 4.80 (57)	<.001
Education (years)	17.47 ± 1.91 (51)	16.153 ± 1.96 (57)	.013
Combined Income	$169,721 ± 87,709 (50)	$91,861 ± 51,876 (56)	<.001
BMI	26.26 ± 3.45 (50)	26.53 ± 4.80 (57)	.74
Dyadic Adjustment	123.29 ± 10.08 (51)	120.36 ± 13.14 (57)	.21
Pittsburgh Sleep Scale	4.27 ± 1.88 (51)	4.18 ± 1.89 (57)	.79
Depression	5.06 ± 3.84 (51)	5.16 ± 3.26 (57)	.89
Hours per week with partner	106.89 ± 25.47 (51)	96.56 ± 24.22 (57)	.03
Pet
Has Pets	86.27% (44)	47.37% (27)	<.001
No Pets	13.73% (7)	58.82% (30)
Race
White	70.59% (36)	50.88% (29)	.15
Asian	11.76% (6)	17.54% (10)
Black	7.84% (4)	17.54% (10)
Hispanic	7.84% (4)	12.28% (7)
Multiracial	1.96% (1)	1.75% (1)

Questionnaires

For the Pittsburg Sleep Quality Scale, there were main effects of both fatherhood status (F(1,107.53)=10.02, p=.002) and visit number (F(3, 229.05)=9.31, p<.001), as well as an interaction between fatherhood status and visit number (F(3, 299.05)=10.14, p<.001). Post hoc tests showed that fathers had worse sleep quality at the early postnatal visit (Fathers: M= 6.43, SD=2.47, Nonfathers: M=4.21, SD=1.97, p<.001, d=1.01) and late postnatal visit (Fathers: M=5.18, SD=1.97, Nonfathers: M=4.21, SD=1.78, p=.013, d=.52), but not the mid-gestation visit (Fathers: M=4.27, SD=1.88, Nonfathers: M=4.18, SD=1.89, p=.79, d=.053) or the late gestation visit (Fathers: M=4.61, SD=2.19, Nonfathers: M=4.29, SD=1.60, p=.41, d=.17) (Supplementary Figure 2). Within non-fathers, sleep quality did not significantly differ between timepoints. On the other hand, fathers had significantly worse sleep at the early postnatal visit (M=6.43, SD=2.47) compared to the mid-gestation visit (M=4.27, SD=1.88, p<.001, d=-.80), late gestation visit (M=4.61, SD=2.19, p<.001, d=-.70), and late postnatal visit (M=5.18, SD=1.97, p<.001, d=-.59). Fathers also had significantly worse sleep at the late postnatal visit compared to the mid-gestation visit (p=.002, d=-.47).

For hours per week spent with their partner, there was a main effect of fatherhood (F(1, 105.06)=10.55, p=.002) but no main effect of visit number (F(3, 209.40)=1.05, p=.37) and no interaction between fatherhood status and visit number (F(3, 209.40)=.78, p=.50). Overall, fathers spent more hours per week with their partners than did non-fathers (Supplementary Figure 3).

For the Dyadic Adjustment Scale, there was a main effect of fatherhood status (F(1,104.02)=4.27, p=.04), where fathers had, on average, greater dyadic adjustment scores than nonfathers (t(106)=2.11, p=.04). However, there was no main effect of visit number (F(3, 209.56)=.87, p=.46) nor was there a significant interaction between fatherhood status and visit number (F(3, 209.56)=1.35, p=.26) (Supplementary Figure 4).

For BMI, there was a main effect of visit number such that BMI increased across visits (F(3, 162.52)=5.53, p=.001). There was no significant effect of fatherhood status (F(1,105.87)=.06, p=.81) and no significant interaction between fatherhood status and visit number (F(3, 162.52)=.84, p=.47) (Supplementary Figure 5).

From the early postnatal to late postnatal visit, fathers reported feeling less frustration with their infant (t(48) = 3.08, p = .003, d=.44), provided more indirect care (t(48) = 2.40, p = .02, d=.34), and became less positively engaged with their infant (t(48) = −4.87, p<.001, d=−.70). There was no significant change between the early postnatal to late postnatal visit in measures of paternal attachment (t(48) = 1.48, p=.14, d=.21), warmth and attunement (t(48) = 1.85, p= .070, d=.26), or control and process responsibility (t(48) = .33, p = .74, d=.047) (Table 2).

Table 2.

Paternal Behavior at Visits 3 and 4 (early and late postnatal) (M +/− 1 SD).

	Visit 3 (N)	Visit 4 (N)	p
Attachment	75.98 ± 8.01 (49)	77.14 ± 7.99 (49)	.14
Warmth and Attunement	6.62 ± 0.44 (49)	6.73 ± 0.39 (49)	.07
Control & Process Responsibility	5.97 ± 0.80 (49)	6.01 ± 0.68 (49)	.74
Frustrations	4.81 ± 1.09 (49)	5.18 ± 1.14 (49)	.003
Indirect Care	2.63 ± 1.13 (49)	3.17 ± 1.48 (49)	.02
Positive Engagement	6.13 ± 0.47 (49)	5.60 ± 0.65 (49)	<.001

Timing of Blood Draws

For the time of participants’ blood draws, there was a main effect of visit number (F(3, 190.61)=3.87, p=.01) in which blood draws occurred earlier in the day for later visits. For example, the average time of blood draw at mid-gestation visits was 1:08 pm and at late postnatal visits was 12:02 pm. There was no main effect of fatherhood status on time of blood draw (F(1, 103.25)=.02, p=.89) and no significant interaction between fatherhood status and visit number (F(3, 190.61)=.46, p=.71) (Supplementary Figure 6). To control for this effect of visit number, time of blood draw was entered into our mixed linear models as described below.

Testosterone

T levels across the four study visits were all significantly positively correlated (Supplementary Table 1), suggesting some interindividual stability in T levels. A linear mixed model revealed no significant effect of fatherhood status (F(1, 100.84) = 3.14, p = .08), visit number (F(1, 233.13) = 1.04, p = .38), or their interaction (F(1, 233.13) = 2.15, p = .10) on plasma T levels. To control for potential confounding variables, the following variables were considered for inclusion in the model: age, BMI, hours per week with partner, time of blood draw, sleep patterns, relationship satisfaction, pet ownership, SES, and race. Of these variables, four were both associated with T levels and related to either fatherhood status, visit number or their interaction: BMI, pet ownership, hours per week with partner and time of blood draw. After including these variables in the model, there was once again no significant effect of fatherhood status (F(1, 96.46) = 2.11, p = .15), visit number (F(3, 238.65) = 1.48, p = .22), or their interaction (F(3, 204.08) = 2.14, p = .10) on plasma T levels. Only time of blood draw and BMI had a significant negative effect in the model (time of blood draw: F(1, 332.82) = 20.29, p < .001; BMI: F(1, 132.38) = 8.37, p = .004).

Among fathers, linear mixed models provided no evidence that more involved or attached fathers had lower T levels or larger declines in T across the four visits. Separate models were specified for the paternal postnatal attachment scale (PPAS) and for all five subscales of the Paternal Involvement with Infants (PIWI) Scale. The only significant interaction was for the Positive Engagement subscale of PIWI at the early postnatal visit, where fathers who were lower in positive engagement showed a larger decrease in T across visits as compared with fathers who were higher in positive engagement (F(3, 128.93) = 2.77, p = .045) (Supplementary Figure 7, Supplementary Table 2).

In subsequent exploratory analyses, separate analyses at each visit showed that fathers had lower T levels than nonfathers at the mid-gestation visit (t(99)=-2.69, p=.011; d=-.52), however differences between fathers and nonfathers did not reach significance at the late gestation visit (t(91)=-.76, p=.45; d=-.16), early postnatal visit (t(93)=-1.28, p=.20; d=-.26) or late postnatal visit (t(92)=-1.61, p=.11; d=-.33) (Figure 1, Supplementary Table 3, Supplementary Figure 8). We also built stepwise regression models to predict T levels at each visit. Fatherhood status was the only significant predictor at the mid-gestation visit (R² = .064, F(1, 99) = 6.73, p = .01, β = −.25). There were no significant predictors at late gestation. However, at both early postnatal and late postnatal visits, both BMI (Early postnatal: t = −2.69, p = .008, β = −.27; Late postnatal: t = −3.03, p = .003, β = −.30) and hours per week with partner (Early postnatal: t = −2.12, p = .037, β = −.21; Late postnatal: t = −2.22, p = .029, β = −.22) were negative predictors of T levels (Early postnatal: R² = .14, F(2, 91) = 7.42, p = .001; Late postnatal: R² = .16, F(2, 90) = 8.65, p < .001; Supplementary Table 4). A follow-up mediation analysis using PROCESS MACRO revealed a significant indirect path from fatherhood status to low T by way of hours per week spent with partner at the early postnatal visit (Indirect effect: −.235, Bootstrapped 95% Confidence Interval [-.533, −.032]) and a marginally significant indirect path at the late postnatal visit (Indirect effect: −.175, Bootstrapped 95% Confidence Interval [-.497, .002]). That is, fathers tended to spend more hours per week with their partner than did nonfathers, and men who spent more hours per week with their partner tended to have lower T levels. This indirect path was not significant at either the mid-gestation visit (Indirect effect: −.036, Bootstrapped 95% Confidence Interval [-.265, .131]) or late gestation visit (Indirect effect: −.008, Bootstrapped 95% Confidence Interval [-.137, .146]).

Figure 1.

Comparison of plasma testosterone levels in fathers and nonfathers across all four study visits. Plasma testosterone levels were measured in fathers and nonfathers at four longitudinal visits that took place at the following approximate times: 4–5 months gestation (visit 1), 8 months gestation (visit 2), 1 month postnatal (visit 3) and 4 months postnatal (visit 4). There was no significant effect of fatherhood status (F(1, 100.84) = 3.14, p = .08), visit number (F(1, 233.13) = 1.04, p = .38), or their interaction (F(1, 233.13) = 2.15, p = .10). However, fathers had lower testosterone levels than nonfathers at visit 1 (t(99)=-2.69, p=.011; d=-.52). Although trending in the same direction, differences between fathers and nonfathers did not reach significance at visit 2 (t(91)=-.76, p=.224; d=-.16), visit 3 (t(93)=-1.28, p=.102; d=-.26) or visit 4 (t(92)=-1.61, p=.056; d=-.33). Error bars represent ± 1 SE.

In further exploratory analyses, we asked whether T levels at any of the four time points predicted paternal attachment, involvement or hours per week spent with partner at any time points. Six significant correlations were observed (Supplementary Table 5). T at mid-gestation negatively predicted hours per week with partner at both the early postnatal visit (r(44)=-.475, p<.001) and late postnatal visit (r(44)=-.373, p=.011). Similarly, late gestation T negatively predicted late postnatal hours per week with partner (r(43)=-.314, p=.036). Thus, prenatal T levels (early and late gestation visits) predicted hours per week spent with partner during the postnatal (early and late postnatal visits), but not the prenatal period (early and late gestation visits). T at the late postnatal visit was also negatively correlated with early postnatal (r(45)=-.546, p<.001) and late postnatal (r(45)=-.422, p=.003) hours per week with partner (Figure 2). Finally, late gestation T negatively predicted early postnatal positive engagement with infants (r(43)=-.425, p=.004).

Figure 2.

Significant correlations between paternal plasma testosterone levels and hours per week spent with their partners at each visit. Correlations coefficients are provided for all significant correlations. *= significant at p<0.05; **= significant at p<0.01.

Oxytocin

In contrast to T, participant’s OT levels were not significantly correlated across study visits (Supplementary Table 6). A linear mixed model revealed no significant overall effect of fatherhood on plasma OT levels (F(1, 81.66)=3.74, p=.057); however there was a significant effect of both visit number (F(3, 196.39)=24.66, p<.001) and a significant interaction between fatherhood status and visit number (F(3, 196.39)=18.94, p<.001). Whereas OT increased from mid-gestation to late gestation visits among expecting fathers, it decreased from mid-gestation to late gestation visits in nonfathers. To control for potential confounding variables, the following variables were considered for inclusion in the model: age, BMI, hours per week with partner, time of blood draw, sleep patterns, relationship satisfaction, pet ownership, SES, and race. Of these variables, four were both associated with OT levels and related to either fatherhood status, visit number, or their interaction: age, hours per week with partner, pooled income, and education. After including these variables in the model, the overall pattern of results was the same. There was again no significant overall effect of fatherhood on plasma OT levels (F(1, 84.64)=3.01, p=.086), and there was again a significant effect of both visit number (F(3, 196.41)=26.90, p<.001) and a significant interaction between fatherhood status and visit number (F(3, 195.49)=19.94, p<.001). In addition, age was positively associated with OT levels (F(1, 87.40)=4.23, p=.04), and hours per week with partner was negatively associated (F(1, 269.48)=11.10, p<.001).

Among fathers, linear mixed models provided little evidence that more involved or attached fathers had higher OT levels or larger increases in OT across the four visits. Separate models were specified for the Paternal Postnatal Attachment Scale (PPAS) and for all five subscales of the Paternal Involvement with Infants (PIWI) Scale. The models revealed two significant interactions: fathers who were higher in Control and Process Responsibility at the late postnatal visit showed less of a decrease in OT from the late gestation visit to the late postnatal visit (F(3, 122.67) = 2.94, p = .036) (Supplementary Figure 9a). Additionally, fathers who scored higher in Frustrations at the early postnatal visit (corresponding to lower feelings of frustration towards their infant) showed a greater increase in OT from the mid-gestation visit to the late postnatal visit (F(3, 105.427) = 2.698, p = .0496) (Supplementary Figure 9b, Supplementary Table 7).

In subsequent exploratory analyses, separate analyses at each visit showed that fathers had lower OT levels than nonfathers at the mid-gestation visit (t(72.05)=-5.09, p<.001, d=-1.01), higher levels than nonfathers at the late gestation visit (t(91)=4.06, p<.001, d=.84), and no significant difference at the early postnatal visit (t(70.10)=-1.88, p=.064, d=-.38) or late postnatal visit (t(90)=.65, p=.52, d=.14) (Figure 3, Supplementary Table 8, Supplementary Figure 10). We also specified regression models to predict plasma OT levels at each visit. At the mid-gestation visit, OT levels were negatively predicted by fatherhood status (t = −4.08, p<.001, β = −.376), education (t = −2.39, p=.019, β = −.216), and hours per week with partner (t = −2.00, p=.048, β = −.181) (R² = .280, F(3, 92) = 11.94). That is, men had higher OT if they were nonfathers, had less education, and spent fewer hours per week with their partner. At the late gestation visit, OT levels were predicted by fatherhood status (t = 3.64, p<.001, β = .375) and age (t = 2.21, p=.03, β = .228) (R² = .274, F(2, 88) = 16.57). OT levels were higher in men who were older and in fathers compared with nonfathers. At the early postnatal visit, only hours per week with partner was negatively associated with OT levels (R² = .103, F(1, 94) = 10.81, p=.001, β=-.321). Finally, none of the variables significantly predicted OT levels at the late postnatal visit (Supplementary Table 9).

Figure 3.

Comparison of plasma oxytocin levels in fathers and nonfathers at all 4 visits. Plasma oxytocin levels were measured in fathers and nonfathers at four longitudinal visits that took place at the following approximate times: 4–5 months gestation (visit 1), 8 months gestation (visit 2), 1 month postnatal (visit 3) and 4 months postnatal (visit 4). There was no significant effect of fatherhood status (F(1, 81.66)=3.74, p=.057), but there was a significant effect of both visit number (F(3, 196.39)=24.66, p<.001) and a significant interaction between fatherhood status and visit number (F(3, 196.39)=18.94, p<.001). Fathers had lower oxytocin levels than nonfathers at visit 1 (t(72.05)=-5.09, p<.001, d=-1.01), higher levels than nonfathers at visit 2 (t(91)=4.06, p<.001, d=.84), and no significant difference at visit 3 (t(70.10)=-1.88, p=.064, d=-.38) or visit 4 (t(90)=.65, p=.52, d=.14) Error bars represent ± 1 SE.

In further exploratory analyses, we asked whether OT levels at any of the four time points predicted paternal attachment, involvement or hours per week with partner at any time point. Three significant correlations emerged (Supplementary Table 10). OT at the late gestation visit negatively predicted late postnatal control and process responsibility (r(45)=-.313, p=.036), early postnatal OT was associated with more frustration at the early postnatal visit (r(45)=-.302, p=.039), and early postnatal OT negatively predicted positive engagement at the late postnatal visit (r(48)=-.300, p=.039).

Cortisol

Cort levels were weakly positively correlated across study visits, with the exception of the mid-gestation and late postnatal visits (Supplementary Table 11). A linear mixed model revealed no significant overall effect of fatherhood on plasma Cort levels (F(1, 105.429)=.008, p=.927). Additionally, there was no significant effect of visit number (F(3, 201.902)=.503, p=.681) and no significant interaction between fatherhood status and visit number (F(3, 201.902)=1.516, p=.212). To control for potential confounding variables, the following variables were considered for inclusion in the model: age, BMI, hours per week with partner, time of blood draw, sleep patterns, relationship satisfaction, pet ownership, SES, and race. Of these variables, two were both associated with Cort levels and related to either fatherhood status, visit number or their interaction: education level and time of blood draw. After including these variables in the model, there was again no significant overall effect of fatherhood on plasma Cort levels (F(1, 102.531)=.002, p=.964), no significant effect of visit number (F(3, 210.982)=1.157, p=.327), and no significant interaction between fatherhood status and visit number (F(3, 213.611)=1.617, p=.186). Time of blood draw was negatively associated with Cort levels (F(1, 320.573)=45.603, p<.001), and education level was not significant (F(1, 100.748)=.649, p=.423).

Among fathers, linear mixed models provided no evidence that more involved or attached fathers had lower Cort levels or larger decreases in Cort across the four visits. Separate models were specified for the Paternal Postnatal Attachment Scale (PPAS) and for all five subscales of the Paternal Involvement with Infants (PIWI) Scale. There were no significant interactions between visit number and paternal behavior measures on Cort levels (Supplementary Table 12).

There was no significant difference in Cort levels between fathers and nonfathers at any of the four visits (Mid-gestation: t(99)=-1.74, p=.085, d=-.35; Late gestation: t(91)=.39, p=.70, d=.081; Early postnatal: t(93)=.46, p=.65, d=.094; Late postnatal: t(92)=1.11, p=.27, d=.23) (Figure 4, Supplementary Table 13, Supplementary Figure 11). We also specified regression models to predict plasma Cort levels at each visit. Fatherhood status was not a significant predictor in any of the regression models for any of the four visits. At visits 1, 2, and 3, time of blood draw was the only significant predictor. Specifically, levels were negatively predicted by time of draw (Mid-gestation: R² = .086, F(1, 99) = 9.276, p=.003, β = −.293; Late gestation: R² = .239, F(1, 91) = 28.552, p<.001, β=-.489; Early postnatal: R² = .192, F(1, 93)=22.12, p<.001, β=-.438). In other words, Cort levels were higher in samples drawn earlier in the day. At the late postnatal visit, Cort levels were only predicted by education (R²=.049, F(1, 92)=4.724, p=.032, β=-.221), where more years of education were associated with lower Cort levels (Supplementary Table 14).

Figure 4.

Comparison of plasma cortisol levels in fathers and nonfathers at all 4 visits. Plasma cortisol levels were measured in fathers and nonfathers at four longitudinal visits that took place at the following approximate times: 4–5 months gestation (visit 1), 8 months gestation (visit 2), 1 month postnatal (visit 3) and 4 months postnatal (visit 4). There was no significant effect of fatherhood (F(1, 105.429)=.008, p=.927), visit number (F(3, 201.902)=.503, p=.681), or their interaction (F(3, 201.902)=1.516, p=.212). There was also no significant difference between fathers and nonfathers at any of the four visits (Visit 1: t(99)=-1.74, p=.085, d=-.35; Visit 2: t(91)=.39, p=.70, d=.081; Visit 3: t(93)=.46, p=.65, d=.094; Visit 4: t(92)=1.11, p=.27, d=.23). Error bars represent ± 1 SE.

In further exploratory analyses, we asked whether Cort levels at any of the four time points predicted paternal attachment, involvement or hours per week with partner. Two significant correlations were observed. Late postnatal Cort was positively correlated with hours per week with partner at the mid-gestation visit (r(47)=.299, p=.041), and early postnatal Cort was positively correlated with early postnatal indirect care (r(45)=.288, p = 0.0499; Supplementary Table 15).

Vasopressin

AVP levels were positively correlated across visits (Supplementary Table 16). A linear mixed model revealed a significant effect of fatherhood status (F(1, 98.72) = 12.31, p < .001) but no significant effect of visit number (F(3, 198.69) = .84, p = .47), or their interaction (F(3, 198.69) = .379, p = .77) on plasma AVP levels. Specifically, fathers had overall lower levels of AVP compared with nonfathers. To control for potential confounding variables, the following variables were considered for inclusion in the model: age, BMI, hours per week with partner, time of blood draw, sleep patterns, relationship satisfaction, pet ownership, SES (income and years of education), and race. Of these variables, four were both associated with AVP levels and related to either fatherhood status, visit number or their interaction: BMI, age, income, and education. After including these variables in the model, there was once again a significant effect of fatherhood status (F(1, 94.90) = 7.48, p = .007), but no effect of visit number (F(3, 194.08) = .86, p = .46) or their interaction (F(3, 193.44) = .25, p = .86) on plasma AVP levels.

Among fathers, separate linear mixed models were specified for the Paternal Postnatal Attachment Scale (PPAS) and for all five subscales of the Paternal Involvement with Infants (PIWI) Scale. There was a significant main effect of the Control and Process Responsibility subscale of PIWI at the early postnatal visit, in which fathers with lower AVP had higher Control and Process Responsibility (F(1, 45.21) = 5.15, p = .028). There was also a significant interaction between visit number and Control and Process Responsibility at the early postnatal visit, in which fathers who were higher in Control and Process Responsibility showed a greater increase in AVP from the late gestation to late postnatal visits (F(3, 93.04) = 4.22, p = .008) (Supplementary Figure 12a). There was also a significant interaction with visit number for the Positive Engagement subscale of PIWI at the early postnatal visit, where fathers who were high in Positive Engagement showed greater increases in AVP from the mid-gestation to late gestation visits (F(3, 84.64) = 3.23, p = .027). There was also a significant main effect and interaction with visit number for the Positive Engagement subscale of PIWI at the late postnatal visit, where fathers who were high in Positive Engagement had lower AVP levels overall (F(1, 44.34) = 17.08, p<.001), and also showed greater increases in AVP from the late gestation to late postnatal visits and from the early postnatal to late postnatal visits (F(3, 92.51 = 3.86, p = .012) (Supplementary Figure 12b, Supplementary Table 17).

In subsequent exploratory analyses, separate analyses at each visit showed that fathers had lower AVP levels than nonfathers at the mid-gestation (t(87.44)=-3.46, p<.001; d=-.68), late gestation (t(79.40)=-2.28, p=.025; d=-.47), and early postnatal visits (t(69.49)=-2.01, p=.048; d=-.41), but not at the late postnatal visits (t(92)=-1.95, p=.055; d=-.40) (Figure 5a, Supplementary Table 18, Supplementary Figure 13). We also specified regression models to predict plasma AVP levels at each visit. At the mid-gestation visit, AVP levels were negatively predicted by fatherhood status (R² = .095, F(1, 99) = 11.47, p=.001, β = −.322). That is, fathers had lower AVP than nonfathers. At the late gestation visit, AVP levels were only predicted by pooled annual income (R² = .092, F(1, 89) = 9.06, p=.003, β = −.304). At the early postnatal visit, both fatherhood status (t = −2,10, p = .039, β = −.204) and BMI (t = 3.09, p = .003, β = .301) were significant predictors of AVP levels (R² = .126, F(2, 92) = 6.65, p=.002). Fathers and men with lower BMI had lower AVP levels. Finally, at the late postnatal visit, AVP levels were negatively predicted by education (R² = .071, F(1, 92) = 8.06, p=.006, β = −.284); Supplementary Table 19)

Figure 5a.

Comparison of plasma vasopressin levels in fathers and nonfathers at all 4 visits. Plasma vasopressin levels were measured in fathers and nonfathers at four longitudinal visits that took place at the following approximate times: 4–5 months gestation (visit 1), 8 months gestation (visit 2), 1 month postnatal (visit 3) and 4 months postnatal (visit 4). There was a significant effect of fatherhood status (F(1, 98.72) = 12.31, p < .001) but no significant effect of visit number (F(3, 198.69) = .84, p = .47), or their interaction (F(3, 198.69) = .379, p = .77). Fathers had lower vasopressin levels than nonfathers at visit 1 (t(87.44)=-3.46, p<.001; d=-.68), visit 2 (t(79.40)=-2.28, p=.025; d=-.47), and visit 3 (t(69.49)=-2.01, p=.048; d=-.41), but not at visit 4 (t(92)=-1.95, p=.055; d=-.40). Error bars represent ± 1 SE. b. Correlation between fathers’ plasma vasopressin levels at Visit 2 and their log-transformed positive engagement scores at Visit 4.

In further exploratory analyses, we asked whether AVP levels at any of the four time points predicted paternal attachment, involvement or hours per week with partner at any time points. Seven significant correlations were observed (Supplementary Table 20). Lower mid-gestation AVP predicted greater positive engagement at the early postnatal visit (r(46)=-.46, p=.001). Lower late gestation AVP predicted greater early postnatal control and process responsibility (r(44)=-.48, p<.001), less early postnatal frustration (r(44)=-.43, p=.003), and greater late postnatal positive engagement (r(45)=-.65, p<.001) (Figure 5b). Lower early postnatal AVP was also associated with greater early postnatal control and process responsibility (r(47)=-.42, p=.003) and greater late postnatal positive engagement (r(48)=-.46, p=.001). Finally, lower late postnatal AVP was associated with more hours per week with partner at the early postnatal visit (r(47)=-.30, p=.042).

Discussion

Demographics

Compared with non-fathers, fathers were older and had higher pooled income. Perhaps, men are more likely to become fathers once they feel financially secure, which becomes more likely as they get older and have been in the work force longer. Fathers were also more likely to own pets compared with non-fathers, possibly because more nurturing men are both more likely to want pets and children or perhaps because they are simply more likely to acquiesce to their partner’s wishes. Another possibility is that couples use pet ownership as a form of preparation, or as a prelude, to having children.

Unsurprisingly, fathers had worse sleep quality than non-fathers in the postnatal period, presumably due to night-time infant feedings and crying. Fathers also spent more time with their partner than did non-fathers. Doing so could be predictive of parenthood, or it could be a reaction to the knowledge that they are expecting. As infancy progressed, fathers became less positively engaged and less frustrated with their infant, and also became more involved in indirect infant care. Indirect care includes behaviors like taking the infant to and from day care. So, this change may reflect many fathers returning to work after a paternity leave, and offloading some direct caregiving responsibilities to other caregivers.

Testosterone

Previous research indicates that human fathers tend to have lower T levels than nonfathers, and that fathers who are more positively engaged with their infants and young children tend to have lower T levels than less positively engaged fathers (Grebe et al., 2019; Meijer et al., 2019). Here we find that although fathers had lower T levels than nonfathers at the mid-gestation visit when their partners were between 4–5 months pregnant, this difference did not persist across the subsequent three visits that extended into the postnatal period, resulting in no overall main effect of fatherhood status on T levels. Contrary to our hypothesis, levels did not decrease across the transition to fatherhood, nor were changes in T across this transition correlated with either self-reported paternal involvement or attachment. Although a number of studies have reported modest decreases in T across the transition to fatherhood (Bakermans-Kranenburg et al., 2022; Corpuz and Bugental, 2020; Edelstein et al., 2017), several other studies have reported no change, as we find here (Berg and Wynne-Edwards, 2001; Cárdenas et al., 2023; Diaz-Rojas et al., 2023; Perini et al., 2012).

Combined with cross-sectional evidence for lower T levels in fathers than nonfathers, our results raise the possibility that rather than T decreasing when men become fathers, men with lower T may be more likely to become fathers. However, a longitudinal study of men living in the Philippines provided evidence to the contrary, as it was single men with higher T who were more likely to subsequently become partnered fathers within the next 4 years (Gettler et al., 2011a). Moreover, after becoming partnered fathers, men experienced an average 33% decline in nighttime salivary T levels (Gettler et al., 2011a). This begs the question of when levels decline across the transition to fatherhood. Our data show that new fathers already have lower T than nonfathers at 4–5 months gestation, implying that declines in T occur even earlier in the transition to fatherhood. Partnered men have lower average T levels than single men (Grebe et al., 2019), so one way of reconciling the above findings is to hypothesize that men who experience larger declines in T upon becoming partnered are more likely to then become fathers. Perhaps such men are psychologically prepared and hormonally primed for fatherhood. Indeed, our data show that expecting fathers with lower T tend to spend more time with their partner during the postnatal period. That is, prenatal paternal T predicts postnatal time spent with the partner, and likely the infant as well, since mother and infant are often in close proximity.

Results from exploratory cross-sectional regression models at each time point are also consistent with this interpretation. We found that at both postnatal visits (visits 3 and 4), men who spent more time with their partner tended to have lower T levels. Mediation analysis further revealed a significant indirect path from fatherhood status to low T by way of hours per week spent with partner. Fathers tended to spend more time with their partners during the postnatal period compared with nonfathers, and spending more time with the partner was associated with lower T levels. We cannot infer the direction of causality from these cross-sectional models. As suggested above, it may be that T levels are predicting time spent with the partner, however it could also be that spending time with their partner lowers T levels. In support of the latter possibility, early postnatal hours per week spent with partner negatively predicted late postnatal T levels. On the other hand, mid-gestation and late gestation T negatively predicted early postnatal and late postnatal hours per week spent with partner. Therefore, causality may be bidirectional and perhaps mutually reinforcing.

Another recent study of first-time, expecting fathers reported significant decreases in T over the course of their partner’s pregnancy. Similar to our findings here, larger prenatal declines in T were associated with greater paternal contributions to household and infant care tasks postpartum (Edelstein et al., 2017).

Previous research has found that men in Western societies tend to experience an increase in abdominal adiposity when they become fathers, and this increase is accounted for by their lower levels of T (Gettler et al., 2017). Increases in abdominal fat should lead to increases in BMI. Although there was no difference in BMI between fathers and nonfathers in our study, cross-sectional models showed that BMI was negatively associated with T at both postnatal visits. This is consistent with evidence that T decreases fat mass (Dandona et al., 2021). However, adipose tissue is also able to convert T to estradiol (Schneider et al., 1979), providing an alternative explanation for the negative association between T and BMI.

Oxytocin

Compared with T, a smaller body of research has compared OT levels between fathers and nonfathers, generally finding higher levels in fathers (Grumi et al., 2021). In contrast, we did not detect an overall main effect of fatherhood status on plasma OT in our study. However, we did detect effects of fatherhood status at two specific time points. Unexpectedly, fathers had significantly lower OT than nonfathers at the mid-gestation visit (29.2% lower in fathers). While the explanation for this difference is unknown, differences in sexual behavior could play a role. OT increases in response to sexual activity and sexual arousal (Carmichael et al., 1987; Murphy et al., 1987), and intercourse frequency tends to decrease during pregnancy (Blumenstock and Barber, 2022), so one possibility is that nonfathers were more sexually active than expecting fathers. T has been reported to increase in men in response to social interactions with women (Flinn et al., 2012; Roney et al., 2007; Roney et al., 2003). Nonfathers had higher levels of T than fathers at the mid-gestation visit so another speculative possibility is that nonfathers found their interaction with our female phlebotomist to be more arousing than fathers did. Higher levels of arousal in nonfathers might translate into higher OT levels. In fact, plasma OT and T were weakly positively correlated at the mid-gestation visit (r(96)=.26, p=.011).

From the mid-gestation to late gestation visits, OT levels increased in expecting fathers (by 30.8%) and decreased in nonfathers (by 16.7%), such that fathers had significantly (15.6%) higher OT than nonfathers at the late gestation visit. The explanation for this difference is also unknown, but one possibility is that OT levels are synchronizing in mothers and fathers, as has been reported for other hormones including Cort, T, prolactin and progesterone (Daneshnia et al., 2024). One study has reported synchronization of couple’s OT levels in the postpartum period (Gordon et al., 2010), but prenatal synchronization has not been investigated to the best of our knowledge. Plasma OT increases about 3–4 fold in expecting mothers over the course of pregnancy (Uvnas-Moberg et al., 2019), suggesting that unmeasured maternal OT levels were likely to be increasing from mid-gestation (4–5 months gestation) to late gestation (at 8 months gestation) in parallel with increases over the same interval in expecting fathers. An interesting parallel is found in the biparental California mouse, in which expecting fathers have higher levels of plasma OT than either virgin males or fathers (Gubernick et al., 1995).

Fathers who were more involved with or attached to their infant did not have higher levels of OT or larger increases in OT across the transition to fatherhood. While two studies have reported associations between paternal stimulatory behavior and endogenous OT levels (Gordon et al., 2010; Morris et al., 2021), a number of studies have not found significant associations between endogenous OT and other aspects of paternal behavior (Gettler et al., 2019; Gordon et al., 2017; Miura et al., 2015). Our negative findings might be due to the lack of any measure paternal stimulatory behavior. Surprisingly, the small number of associations we found between plasma OT levels at each visit and paternal behavior were all negative, with higher OT being associated with less positive paternal behavior. However, there is a considerable risk that these are false positive results since these correlations were all weak and would not survive multiple comparison correction.

Cross-sectional models showed that hours per week spent with partner was negatively associated with plasma OT levels at both the mid-gestation and early postnatal visits. The relationship between OT levels and relationship quality in the published literature is complex (Grebe et al., 2017). Higher levels of OT have been associated with both higher relationship quality (Grewen et al., 2005; Holt-Lunstad et al., 2015; Schneiderman et al., 2012) and with relationship distress (Tabak et al., 2011; Taylor et al., 2010). In the former case, our findings might imply that couples who have high OT levels are secure in their relationship and feel free to spend more time independently. In the latter case, our findings could reflect a compensatory physiological adaptation to motivate couples who are spending less time together to tend to their relationship. Nevertheless, there was no significant association in our data between plasma OT and father’s self-reported relationship quality.

Cortisol

Contrary to expectations (Storey et al., 2000), there was no evidence that Cort levels peaked just before birth in fathers in our study. We sampled expecting fathers at approximately 8 months gestation, so it is possible that levels peaked after that point in closer proximity to the birth. Nor was there evidence that Cort levels decreased in fathers in the postnatal period (Gettler et al., 2011b). Fathers reporting more involvement in indirect caregiving at the early postnatal visit had higher Cort levels overall in comparison with fathers who reported less involvement in indirect care at the early postnatal visit. This finding is of interest given that higher Cort is associated with greater maternal responsiveness (Fleming et al., 1997). Cort’s most fundamental role is to mobilize energy reserves to prepare the body for action (Sapolsky, 1994), so higher levels may reflect motivation to provide care. There was no evidence that changes in Cort across the transition to fatherhood were related to paternal involvement or attachment. The only consistent and reliable predictor of Cort levels in our sample was the time of blood draw, such that levels were higher earlier in the day as expected due to normal diurnal variation (Lacerda et al., 1973).

Vasopressin

AVP levels did not change across the transition to fatherhood, however fathers had consistently lower AVP levels than nonfathers, and fathers with lower AVP levels tended to be more positively engaged with their infants. These findings were unexpected given evidence that AVP signaling is involved with paternal caregiving in some mammals. For example, AVP injections into the lateral septum elicit paternal behavior in male prairie voles (Wang et al., 1994), and AVP-immunoreactive staining in the bed nucleus of the stria terminalis (BNST) predicts paternal behavior in California mice (Bester-Meredith and Marler, 2003). Moreover, marmoset monkey fathers have increased V1a AVP receptor density on neurons in prefrontal cortex compared with non-fathers (Kozorovitskiy et al., 2006). Nevertheless, AVP effects on behavior seem to be highly species-specific (Rigney et al., 2022). In some species, including humans, AVP is implicated in male aggression and the acquisition of dominance (Terranova et al., 2017; Thompson et al., 2004; Thompson et al., 2006), and it is also involved in human male sexual arousal (Murphy et al., 1987). Therefore, similar to T, lower AVP levels in new fathers could potentially reflect a life history strategy shift in the direction of decreased investment in mating effort and male-male competition. There is an interesting parallel to our results in the biparental Titi monkey, where fathers also have both decreased androgens and lower AVP receptor 1a binding in the brain compared with nonfathers (Baxter et al., 2023).

In contrast to our findings here, previous studies reported no difference in AVP levels between nonfathers and either fathers (Gray et al., 2007) or expecting fathers (Cohen-Bendahan et al., 2015). The discrepancy could relate to the sample type, assay type, or stage of fatherhood examined. Whereas we measured AVP in plasma using HPLC in first-time fathers who had infants 4 months of age and younger, Gray et al measured AVP in urine using an EIA in fathers with children who were up to four years of age. Cohen-Bendahan et al measured AVP in expecting fathers between 32 and 34 weeks gestation, which is comparable to our late gestation visit at 8 months gestation, but also measured it in urine with EIA.

Unlike steroid hormones such as T and Cort, the neuropeptides OT and AVP are limited in their ability to cross the blood-brain barrier (Ermisch et al., 1985), which has raised questions about the behavioral relevance of peripheral OT and AVP levels (Churchland and Winkielman, 2012). However, some hypothalamic OT neurons have axon collaterals that project and release OT into both the brain and the periphery (Ross and Young, 2009; Zhang et al., 2021), and concurrent release has been demonstrated in some contexts (Tabak et al., 2023). On the other hand, several studies have failed to find correlations between peripheral and central levels of OT, particularly at baseline (Valstad et al., 2017). Even when peripheral levels do not accurately reflect levels in the brain, they may still have behavioral relevance since peripheral OT may be able to signal the brain indirectly via the vagus nerve or another mechanism (Carter et al., 2020). One limitation of our study is that we conducted a large number of statistical tests without correcting for multiple comparisons. This decision was made to decrease the risk of false negative findings (Fiedler et al., 2012). To mitigate the risk of false positive findings, we emphasized those that were statistically robust and consistent. Nevertheless, it will be important to determine if our findings replicate in subsequent studies. Another limitation is that we did not specifically ask fathers how much time they spent with their infant. We did ask them how much time they spent with their partner, and since the mother and infant are often together in the postnatal period, this is likely positively associated with time spent with infant. However, future studies should specifically ask fathers about time spent with infant. Finally, although we used well-validated scales to measure paternal attachment and involvement, observations of actual paternal behavior may be more objective and less subject to reporting bias (Zahidi, 2019). This could explain the limited number of observed associations between hormonal changes and paternal behavior.

Conclusion

In conclusion, we find that first-time, expecting fathers have lower T and AVP levels than non-fathers from early in gestation, and these low levels predict greater time spent with their partner and greater positive engagement with their infant in the postnatal period. Similar to what is known for human mothers, OT levels increased across the gestational period in new fathers, but OT levels were not related to greater paternal involvement or attachment. Finally, Cort levels did not differ between fathers and nonfathers, and did not predict any of our measures of paternal involvement and attachment. Our findings raise the possibility that low levels of AVP and T found in first-time expecting fathers reflect their anticipation of and preparation for their new parental role.

Data Availability

Data are available at: https://osf.io/u2rdt/files/osfstorage/6761a601ffff088626a33484