Decades of scientific research have shown that childhood adversity of varying degrees of severity and chronicity is associated with changes in brain volume1,2,3. However, generalising from existing studies is challenging due to (1) the wide variability in how adversity is defined; (2) limited evidence on cumulative effects and co-occurrence2,4,5; and (3) challenges in replicability across large cohorts. Despite these challenges, converging evidence suggests that many forms of adversity share common pathways6. Indeed, across adversities, there may be shared aspects related to the biology of stress, and the dysregulation of the stress hormone system has been widely investigated7,8,9. Another shared aspect across many childhood adversities emphasizes the role of parenting and the broader family context10. Here we contribute to the literature on the associations between adversity and brain structure, by focusing on maternal bonding as one crucial aspect of development with demonstrated impact on stress resilience and early neurodevelopmental pathways11,12,13. While it is a challenge to disentangle the maternal relationship from broader adverse aspects of the home environment and beyond, we argue that given its early developmental timeline, maternal bonding may leave individual signatures on the developing brain.

Maternal bonding refers to the emotional tie that a mother develops toward her infant14. One way to assess early maternal bonding is through the Parental Bonding Instrument (PBI)15. The PBI retrospectively measures perceived parenting before age 16 and can be administered after age 16, with demonstrated long-term stability16. Two continuous dimensions are primarily assessed in the PBI (1) maternal care (empathic, sensitive, responsive parenting) and (2) maternal control (intrusive, infantilizing, demanding parenting). The opposite pole of maternal control is captured by PBI factor maternal autonomy, which reflects encouragement of independence, including support for the child’s freedom and decision-making latitude. These dimensions of parenting may support understanding of broader, population-representative developmental mechanisms17.

Certainly, a caregiver’s affection and care are crucial sources of reward-giving during development18,19. By seeking closeness to their mothers, infants carry out goal-directed behaviours to achieve physical and emotional security, which, in turn, rewards them with affection and reassurance20,21,22. Through these early interactions, infants learn and internalize the relationship between rewarding stimuli and appropriate behaviours. A large body of evidence suggests that the dorsal striatum (caudate and putamen) plays an important role in such reward-related learning behaviours23,24. For example, the dorsal striatum is thought to maintain information linking rewards to the appropriate actions25 and causal evidence shows that stimulation of the caudate nucleus directly strengthens reward-related learning26. Striatal associative learning is thought to be optimal during reward anticipation, reinforcing behaviours that have the potential to determine a reward27,28, and supporting habitual and goal-directed behaviours29,30.

Against the background of this literature, several studies support the association between maternal bonding behaviours and the striatum. In a longitudinal cohort study, early maternal care (characterized by affectionate, engaging, and responsive interactions at 3 months) has been shown to shape reward-related brain function in children and may buffer familial risk for psychopathology18. Low maternal care on the PBI has been linked to greater dopamine release following acute stress31 and altered behavioural responsivity to reward after methylphenidate (a dopamine agonist) administration32. Similarly, maternal separation was shown to affect the responsiveness of dopamine neurons in the striatum33. Furthermore, maternal emotional neglect or cold parenting during childhood was shown to affect reward learning21,34 and reward anticipation35 in adolescence or adulthood. At a familial level, healthy collaborative co-parenting was associated with efficient top-down regulation through caudate-prefrontal connectivity36. Ample evidence also links the striatum to other forms of adverse childhood experiences. For example, volume differences in the dorsal striatum (caudate) have been associated with adversity reports identified in childhood trauma questionnaires in adults and adolescents37,38,39,40. Furthermore, reports of harsh parental discipline have been linked to caudate reductions in children41 and increased T2 relaxation time in the right caudate and putamen in young adults42.

In later life, negative mental and cognitive outcomes may reflect lasting effects of early stress on neural systems, driven by chronic glucocorticoid exposure43,44,45,46. Brain regions with prolonged developmental trajectories may be especially vulnerable; notably, the dorsal striatum undergoes substantial changes from childhood into adulthood47,48. Adversity-related alterations, particularly in affective and reward networks, may disrupt typical development and persist long after exposure, potentially remaining observable in later life, with implications for increased dementia risk43,45,48,49,50. These changes in the dorsal striatum may also have effects on psychological well-being. Indeed, high maternal control and low care have been linked to later-life mental health outcomes51,52,53, which in turn relate to alterations in dorsal striatal function and volume in older individuals54,55. Together, this evidence lends support to the idea that dimensions of maternal bonding may impact the structural integrity of the dorsal striatum, leaving lasting signatures in older age. To the best of our knowledge, studies have not yet investigated this association in an older population sample.

In this study, we used a cross-sectional, region-of-interest design to examine associations between perceived maternal care, autonomy, and control (as measured by the PBI) and dorsal striatal volumes (bilateral caudate and putamen) in an older birth cohort. If early maternal interaction shapes reward-related brain development, evidence of its effects in later life may underscore the enduring impact of early caregiving across the lifespan.

Methods

Study sample

Data from the “Insight 46” neuroimaging sub-study of the MRC National Survey of Health and Development (NSHD) 1946 birth cohort was used for this analysis. The NSHD cohort recruited 5362 individuals born during the same week, in March 1946 in England, Scotland, and Wales with follow-up assessments since birth. Members were singleton births from married parents56,57,58. The neuroimaging sub-study was designed to investigate factors impacting late-life brain health, recruiting 502 NSHD members who were selected at random following a study visit at age 60–64 and for whom relevant life course data was available59. This cohort allows examination of potential long-term associations between early life experiences and brain structure in older age. Ethical approval for the NSHD study was provided by Research Ethics Committees in England and Scotland. The neuroimaging sub-study obtained ethical approval by the National Research Ethics Service (NRES) Committee London (REC reference 14/LO/1173). Participants provided written informed consent for their participation56,57,59. The study was conducted in accordance with the Declaration of Helsinki.

Of the 468 participants’ neuroimaging data that was available for this analysis, sixteen were removed due to reported history of neurological events (before the age of 63), existing outlier values in the imaging phenotypes of interest and following QC of the brain extraction result on the T1-weighted images. The final sample included 452 participants and the degree of missingness within each variable of interest is reported in Table 1. Analyses were conducted using listwise deletion, therefore, the total number of participants for each main statistical test is reported below. MRI scanning for the selected participants occurred between mid-2015–early-2018 when participants were 69 to 71 years old (M = 70.17; SD = 0.74; Table 1).

Table 1 Sample characteristics.
Full size table

Maternal bonding

The Parental Bonding Instrument – PBI15 was used to assess perceived maternal attitudes in NSHD responders up to the age of 16 years, based on 24 items and Likert responses on a 1–4 scale: very like-very unlike. The instrument is designed for both mothers and fathers. The current study is concerned with the maternal role in child development and therefore only the maternal scale was investigated.

Psychometric studies of the PBI show good to excellent internal consistency15,60]. While the original study proposed two factors for the PBI, i.e., maternal care and control15, many subsequent large-scale studies supported a three-factor solution that separate between maternal care (e.g.: “Could make me feel better when I was upset”), control (also called “overprotection”, e.g.: “Tried to control everything I did”) and autonomy (also called “non-engagement” or “authoritarianism”, e.g.: “Gave me as much freedom as I wanted”)60,61,62,63,64. In the NSHD sample, factor analyses of PBI data also support the three-factor solution with reference to the following scales: maternal care, control and non-engagement/autonomy61,64. The current analysis was conducted on summed item scores pertaining to these scales following confirmation of previously reported dimensionality within our (smaller) sample (Table S1). “Not applicable” and “Unknown” responses were set to missing and sum scores for each scale were calculated only where all item responses were present. Items were adjusted for consistent directionality within scales, and therefore higher scores suggest that an individual is higher on the characteristic.

The PBI was administered to participants in 1989, when they were 43 years old. Given the retrospective nature of this assessment, recall bias and potential false negative reports cannot be ruled out65,66. However, the PBI was demonstrated to have long-term stability over 20 years in a non-clinical sample, with limited influence of life experiences, depression history and state mood on perceived parenting16. Furthermore, adverse mental health outcomes, have been consistently associated with maternal parenting experiences in adults based on retrospective data from the PBI, suggesting that this questionnaire identifies meaningful early experiences60,62,67.

Adult well-being

Well-being was measured using the Warwick-Edinburgh Mental Well-being Scale (WEMHS)68 via a postal questionnaire sent to participants in 2014 at the age of 68 years, therefore close to the time when scanning occurred57. The scale includes 14 items measuring positive affect experienced during the 2 weeks prior to completion. Likert responses were coded 1–5 (none of the time – all of the time, e.g.: “I’ve been feeling optimistic about the future”), and higher scores suggest greater well-being. The measure has a one-factor dimension, and it showed excellent internal consistency in the NSHD sample when calculated on WEHMS data also collected from participants four-eight years prior53. The computed score made available to researchers as a derived variable was used in this analysis. Here, where more than 4 items were missing, the total score was not calculated.

MRI protocol and structural analysis

Data acquisition is described elsewhere59. In brief: brain imaging was performed using a Biograph mMR 3 Tesla PET/MRI scanner (Siemens Healthcare) to simultaneously acquire both dynamic amyloid PET and MRI data, including high resolution 3D T1-weighted, T2-weighted and FLAIR volumetric scans, among other MRI sequences such as resting state functional MRI. The T1-weighted images used here were obtained using a Magnetization Prepared Rapid Gradient Echo Imaging (MPRAGE) sequence with the following parameters: (MPRAGE): Repetition Time (TR) = 2000 ms; Echo Time (TE) = 2.92 ms; Inversion Time (TI) = 870 ms; flip angle = 8°; voxel size resolution = 1.1 mm3 isotropic. Images underwent manual QC, including issues related to motion, coverage, blurring, contrast.

Data analysis was carried out using FSL tools69. Pre-processing for structural images included the following steps: (i) re-orienting images to the standard (MNI) template, (ii) bias field correction, (iii) brain extraction and (iv) brain tissues segmentation using FMRIB’s Automated Segmentation Tool (FAST) that allows extracting global measures of total grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF)70. Automated segmentation of the left and right caudate and putamen was performed using the FMRIB Integrated Registration and Segmentation Tool (FIRST)71. Derived masks were visually inspected. Resulting tissue-type and subcortical segmentations were normalized to the whole brain volume (brain structure/total brain volume*100).

Statistical analysis

All analyses were conducted on the Dementias Platform UK Data Portal72. The statistical analysis was carried out using Stata Version 18.0 SE (Appendix S1).

First, we determined the number of PBI factors within our analysis sample using Horn’s parallel analysis and the point of inflection in scree plots73, followed by a confirmatory factor analysis (CFA) with a latent variable organization in a structural equation model to determine the factor loadings for each item. The Satorra-Bentler maximum likelihood estimator is considered robust for ordinal-level data, and therefore it was applied here74. Goodness-of-fit was assessed based on the following indices: Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), Tucker Lewis Index (TLI), Standardized Root-Mean Residual (SRMR). Values ≥ 0.95 for TLI and CFI, < 0.07 for RMSEA, and < 0.08 for SRMR are considered to show a good model fit75,76. Cronbach’s alpha assessed internal consistency and average inter-item correlations (AIC). Cronbach’s alpha > 0.70 indicates good internal consistency and AIC is generally expected to be moderate, ranging between 0.15 and 0.5077.

Second, using the resulting factors, sample characteristics were examined by comparing participants grouped by levels of maternal care, control, and autonomy across multiple relevant variables. As per previous approaches31,78, k-means clustering with a set seed was used to group the scores on each maternal sub-scale into three groups: low, medium and high care/autonomy/control, which allowed comparison across demographic and biological variables, and exploration of potential threshold effects. Clusters covered between 20 and 44% of the participant scores (Table 1). Additionally, extreme cases of maternal practices were investigated (10th vs. 90th percentiles), considering that any potential differences if present, would persist after a long follow-up. To ensure robustness, we also ran supplementary regressions using continuous PBI scores. Group differences (k-means clusters) on multiple variables were assessed using one-way ANOVAs, Kruskall-Wallis or chi2 tests, where applicable. There were no significant group differences on age, adult socio-economic status, social class during childhood and early adulthood, birthweight, pubertal growth, genetic risk for Alzheimer’s dementia (ApoE4 carrier status), overall brain volumes, body mass index, and cognition around the time of scanning (Table 1). Sex was significantly different across groups, and together with scores on the Mini Mental State Examination (MMSE), these variables were regressed out of the outcome variables and the resulting standardized residuals were included in the subsequent analyses. De-confounding for MMSE was conducted given the participants’ age as well as evidence that the rate of cognitive decline is larger after 65 years in the general population79. Supplementary analyses controlling for additional confounds are also reported. Furthermore, analyses also explored potential sex differences, given reported sex-specific associations between maternal factors and subcortical structures8. For readability, means and standard deviations are reported for unadjusted dependent variables (brain volumes adjusted for total volume only). Finally, the absolute deviation around the median with a 3.5 threshold was used for outlier exclusion on the brain volume measures prior to de-confounding80,81.

Finally, differences in the bilateral caudate and putamen volumes, as well as well-being, were simultaneously assessed in separate MANOVA’s with each of the three maternal sub-scales as the predictors, using the Wilk’s λ result. The significance level was set at 0.017 (α = 0.05/3). The same threshold was considered for the univariate tests following a significant MANOVA, and any subsequent pairwise comparisons were also Bonferroni corrected within each test. The assumption of equality of covariance matrices was met for all scales (Box’s test: p >.05). Multivariate analyses were followed up by univariate tests (mvreg) and subsequently by Bonferroni corrected pairwise comparisons on the resulting coefficients (pwcompare), as well as by bootstrapped comparisons of means with 1000 replications. Homogeneity of variance was assessed using Levene’s test prior to all univariate analyses (all p >.05). Partial eta-squared (ηp2) was reported to describe effect sizes for both multivariate and univariate tests, and SPSS (IBM, Armonk, NY, USA) was used for this purpose. All analyses were conducted list-wise by modelling the full information datasets (Table 1).

Results

Parental bonding instrument

Horn’s parallel analysis ran 690 iterations and yielded a three-factor solution with adjusted eigenvalues larger than 1 (eigenvalues 1–3: 7.50, 2.64, 1.48) (Table S1; Figure S1). This result was consistent with previous reports61,64 and therefore a CFA was conducted using the same item groupings to assess individual loadings (N = 388 without missing values). Standardized factor loadings ranged from small (0.35: “Let me do the things I liked doing”) to large (0.85: “Let me decide things for myself”) with medium-large average loadings for maternal care, autonomy, and control: 0.67, 0.55, and 0.68, respectively. Goodness-of-fit was adequate: RMSEA = 0.06, CFI = 0.91, TLI = 0.89, SRMR = 0.06, and errors were allowed to correlate for two items (“Gave me as much freedom as I wanted” and “Let me go out as often as I wanted”). Internal consistency was high for care (α = 0.91; AIC range: 0.45-0.50), autonomy (α = 0.80; AIC range: 0.34-0.43), and control (α =. 81; AIC range: 0.44-0.48) (Table S1). As expected, maternal care correlated positively with autonomy (r =.511) and negatively with control (r=-.319). Control also correlated negatively with autonomy (r = −.493).

Study sample characteristics

Table 1 shows relevant sample characteristics, including simple group differences across the three scales (also see Table S2 for overall sample characteristics). Significant group differences between care, autonomy and control groups across all three maternal bonding scales showed that group separation was successful (all p <.001). Well-being was also significantly different across groups and its contribution was assessed inferentially in the subsequent analyses (all p <.0023). Across all other characteristics, there were no significant group differences (all p >.077), except for gender where more females were present in the low autonomy and the high control groups (p <.030).

Multivariate statistical analyses

The left and right caudate and putamen volumes as well as well-being scores were included simultaneously in three separate one-way multivariate analyses to determine the effects of maternal care, autonomy and control on N = 384, 396, and 395, respectively. There was a significant effect of control: F(2,392) = 2.66, p =.003, ηp2 = 0.033. There were no significant effects of care (F(2,381) = 1.78, p =.061, ηp2 = 0.023) or autonomy (F(2,393) = 1.78, p =.061, ηp2 = 0.022).

Follow-up univariate tests on the maternal control scale were therefore conducted. Results were significant for well-being (F(2,392) = 7.52, p <.001, ηp2 = 0.037) and marginally for the right caudate (F(2,392) = 3.90, p =.021, ηp2 = 0.019). First, comparisons of the estimated margins showed that well-being was higher in the low control group (M = 57.46, SD = 7.21) compared to medium (M = 54.85, SD = 7.15) and high (M = 53.69, SD = 7.03), (medium vs. low: p =.011; high vs. low: p =.001) (Tables S3, S4). Robustness of differences was further assessed using bootstrapped comparisons of means, which mirrored these results (and the direction of effects), revealing a significant difference between low and medium (p =.001, 95% CI: 0.12, 0.49) and between high and low control groups (p =.019, 95% CI: − 0.36, − 0.03) on well-being scores (note that CI sign reflects contrast order, directionality remained consistent). Second, for the right caudate, follow-up comparisons were also conducted given the marginal effect, as well as the assumption that any long-term effects in an older population might be small, if detectable. Pairwise differences of the estimated marginal means showed that right caudate volumes were larger in the low (M = 0.25, SD = 0.03) compared to the high (M = 0.24, SD = 0.02) control groups; however, the result remained marginal (p =.023) (Fig. 1). Similarly, simple bootstrapped comparisons confirmed a significant difference between the high and low groups on the right caudate volumes, p =.009, 95% CI [−0.37, − 0.05]. For comprehensiveness, univariate tests and follow-up pairwise comparisons for maternal care and autonomy are also reported in supplementary materials. One-way ANOVAs showed significant differences for well-being only (Table S4; p =.002).

Fig. 1
figure 1

Subcortical brain volumes and well-being scores across three maternal control groups. Note. Table shows subcortical brain volumes adjusted for total brain volume only; Well-being scores are unadjusted. Error bars show standard errors. *p <.02, **p <.01.

Full size image

We also conducted supplementary regression analyses using the continuous PBI scores. Results were largely consistent, with maternal control showing negative associations with the right caudate (β=−0.13, p =.007), left and right putamen (β=−0.12, p =.013; β=−0.13, p =.010) and well-being (β=−0.2, p <.001). Maternal care and autonomy were not associated with striatal volumes (all p >.113), but were positively associated with well-being (both p <.006) (Table S5).

Furthermore, moderation was additionally assessed using multiple MANOVAs and specifying the interaction between each maternal scale separately and well-being (also grouped here using k-means clustering). All interaction terms were non-significant (all p >.873), suggesting that marginal brain volume differences in the right caudate were present independently of well-being status. In addition, no significant correlations were observed between well-being and dorsal striatal volumes (all r <.08) (Table S6). Finally, to assess sex-specific effects, multivariate models were re-run, including interaction terms between sex and each maternal bonding scale (care, autonomy, and control), adjusting for MMSE. None of the interactions reached significance at the adjusted alpha (all p >.031). For completeness, sex-stratified follow-up tests were also conducted, which revealed no consistent sex-specific differences in the effects of maternal bonding on brain volumes. Notably, with reduced power, the previously marginal effect on right caudate volume was non-significant (Tables S7 – S9).

Stratified analyses

Separate MANOVAs assessed bilateral caudate and putamen volumes as well as well-being simultaneously, comparing the 10th and 90th percentile scores on each of the three PBI scales. We found a significant effect of maternal control, N = 104, F(1,102) = 5.25, p <.001, ηp2 = 0.211 and maternal care, N = 91, F(1,89) = 3.02, p =.015, ηp2 = 0.151. There was a non-significant effect of maternal autonomy, N = 111, F(1,109) = 1.39, p =.235, ηp2 = 0.062. Univariate tests showed that participants scored higher on well-being in the high versus low maternal care groups (high: M = 58.02, SD = 7.21; low: M = 52.53, SD = 7.61; F(1,89) = 12.66, p <.001, ηp2 = 0.125). There were no significant differences in brain volumes between maternal care extreme groups (p <.244). Differences in maternal control were found for the right caudate (low: M = 0.25, SD = 0.03; high: M = 0.24, SD = 0.02; F(1,102) = 8.79, p =.004, ηp2 = 0.079), left putamen (low: M = 0.32, SD = 0.03; high: M = 0.31, SD = 0.03; F(1,102) = 6.26, p =.014, ηp2 = 0.058), well-being (low: M = 58.27, SD = 7.18; high: M = 52.41, SD = 6.96; F(1,102) = 16.95, p <.001, ηp2 = 0.142), and marginally for the left caudate (low: M = 0.235, SD = 0.03; high: M = 0.227, SD = 0.02; F(1,102) = 5.79, p =.018, ηp2 = 0.054) and the right putamen (low: M = 0.33, SD = 0.03; high: M = 0.32, SD = 0.03; F(1,102) = 5.39, p =.022, ηp2 = 0.050). Direction of differences mirrored previous results with smaller brain volumes and lower well-being scores in the high control group (Fig. 2). Separate moderation tests also showed that there were no interactions between the extreme PBI groups and clustered well-being (all p >.228). Full statistics for all scales are presented in supplementary materials (Tables S10, S11).

Fig. 2
figure 2

Subcortical brain volumes and well-being scores across two extreme maternal control groups. Note. Table shows subcortical brain volumes adjusted for total brain volume only; Well-being scores are unadjusted. Error bars show standard errors. *p <.02, **p <.01, ***p <.001.

Full size image

Furthermore, additional analyses on outcomes corrected for all relevant variables in Table 1 yielded largely consistent results: no effects for care or autonomy, with persistent maternal control effects for the right caudate (p =.004), while all well-being associations remained significant (see Supplementary Tables S12-S13 for further information and analyses on both k-means and extreme groupings). Similarly, models including sex × (extreme) group interaction terms yielded no significant interactions (all p >.525). However, sex-stratified analyses indicated that females reporting high maternal control had marginally smaller right caudate volumes compared to those with low control (p =.018), with no corresponding effect in males (again, noting small subgroup sizes) (Tables S7-S9).

Discussion

Our results showed that in an older population sample of cognitively healthy men and women, retrospective extreme maternal control was associated with smaller right caudate and left putamen volumes, with marginal associations for the left caudate and right putamen. No significant associations were found between dorsal striatal volumes and maternal care or autonomy, either across the full sample or when comparing extreme groups. Perceived maternal bonding characterised by high care and autonomy alongside low control was associated with higher self-reported well-being both across the full sample and when comparing extreme groups. Supplementary analyses using continuous scores, as well as analyses controlling for additional confounds, yielded largely consistent results (Tables S5, S12-S13).

The maternal control scale of the PBI measures the extent to which a mother was intrusive, infantilizing, overprotective, monitoring, denying of psychological and behavioural autonomy17. Affection from caregivers is a key developmental reward, and early interactions shape children’s motivation to seek proximity, comfort, and security through goal-directed behaviours20,21,22. When such rewards are conditional, inconsistent, or ambiguous, children may be less motivated to pursue proximity and attachment. Disruptions in these early motivational patterns may contribute to alterations in reward-related neurodevelopmental pathways, including structural differences in the caudate and putamen20,21,36,84. This interpretation aligns with neuroimaging evidence linking cold/hostile maternal behaviour to increased caudate activation during reward processing21, and with findings suggesting intergenerational similarities in putamen response to reward85. In line with these findings, maternal control has also been implicated in mental health difficulties, especially when combined with low maternal care. Parenting characterised by high control and low care is associated with increased symptoms of anxiety and depression in both children and adults67,86,87. However, maternal control appears to be a more robust predictor than maternal care for outcomes such as anxiety88,89, mid-life well-being61, eating disorders90, and externalizing disorders62,91,92,93,94. Broader family dynamics, such as coparenting style, have also been shown to influence child externalizing behaviours. For instance, collaborative coparenting predicts lower externalising symptoms in children, whereas undermining coparenting has been linked to increased salivary vasopressin, a hormone implicated in aggression36. Authoritarian parents who prioritise compliance and control17, are more likely to use punitive discipline95, which in turn increases the likelihood of child externalizing behaviours96.

Our results did not identify significant associations between maternal care or autonomy and dorsal striatal volumes. Care and control (the inverse of autonomy on the PBI) are conceptually distinct parenting dimensions with potentially different effects on development97,98,99. Low care has been linked to depression, and control more consistently to anxiety86,87,88. Furthermore, studies on stress regulation suggest distinct physiological profiles related to hypothalamic-pituitary-adrenal (HPA) axis functioning relative to this dichotomy98,99. Whether these physiological differences translate to dorsal striatal structure remains unclear, although cortisol may influence dopamine release31,100. Care fosters exploration and goal-directed behaviour, while low care neglects these behaviours; in contrast, control may promote avoidance and limit stress-adaptation strategies86,98. Finally, the lack of observed associations may reflect the long follow-up interval or involvement of brain regions outside our region-of-interest.

While sex × maternal bonding interactions were not significant, exploratory sex-stratified analyses revealed a marginal association between high maternal control and smaller right caudate volume in females within the extreme groups. Although inconclusive, this warrants further investigation in larger samples. Females may exhibit greater dopamine release to low-intensity stimuli101, and maternal control may predict anxiety more strongly in females88. Notably, striatal changes following early adversity may be more pronounced in males102,103. If present, such sex-dependent differences were likely underpowered in our sample.

Finally, across all PBI dimensions, our findings revealed associations with later-life well-being, aligning with the broader literature linking early parental bonding with psychological adjustment across the lifespan15,51,52,88,89. These results underscore the potential long-term influence of early caregiving experiences on both neurobiological structure and subjective well-being in older adulthood, at a population level, emphasizing the importance of considering early life experiences when assessing later life well-being.

Taken together, our findings suggest that perceived excessive maternal control, reflecting overprotective and autonomy-limiting parenting, may contribute to enduring alterations in reward-related neurodevelopment and well-being across the lifespan. High maternal control may chronically limit children’s opportunities for independent reward-based learning and exploration17. Such chronic stress may disrupt brain development and increase vulnerability to internalizing psychopathology, such as anxiety and depression86,87. Clinically, these findings highlight the importance of autonomy-supportive caregiving that balances structure and guidance with autonomy, consistent with evidence-based interventions targeting parental overcontrol, to reduce child anxiety and build resilience104. Given the observed long-term associations, further research should clarify their developmental trajectories and evaluate whether interventions during critical developmental windows could reduce associated psychological risks.

There are several important limitations to our study. First, while our analysis was conducted on a longitudinal population sample, our outcome variables were measured cross-sectionally, and the parental bonding information was acquired retrospectively. This implies that causality cannot be inferred and recall bias cannot be excluded66,105. Second, although our study matched groups across multiple confounding variables and additionally controlled for other relevant variables, nonetheless confounder bias cannot be ruled out, including bias associated with variables that were unavailable to us (e.g., striatal size at birth)106. Third, while the PBI was administered at age 43, it is important to recognise that the mother-child relationship may evolve dramatically over time, and such changes were not captured in our study. In addition, the 26-28-year gap between the completion of the PBI and the brain measurements introduces the influence of other life factors on brain morphology, such as a variety of health issues or other socio-economic dynamics. These shifts could impact well-being and brain structure in older age, potentially introducing variability in the results that was not accounted for by the PBI measure administered at a single time point. We also acknowledge the potential for healthy survivor bias, as participants imaged in later life may represent a healthier, more socioeconomically advantaged, or resilient subset of the original cohort. This may limit generalisability to the broader aging population, and it is particularly relevant given evidence of accelerated dorsal striatal shrinkage with advancing age107. Finally, analyses using listwise deletion excluded cases and may have introduced bias if data were not missing completely at random.

Conclusion

Our study showed that in a large population sample, extreme retrospective perceptions of maternal control were associated with smaller dorsal striatal volumes and lower well-being at 70 years old. The absence of significant associations between dorsal striatal volumes and maternal care or autonomy highlights the complex relationships between early life experiences and later brain structure17. Clinical guidelines that address internalized parental control patterns, promote emotional autonomy, and consider children’s perceived control may inform mechanistically-tailored psychological treatments.