The neural correlates of grandmaternal caregiving

Abstract

In many societies, grandmothers are important caregivers, and grandmaternal investment is often associated with improved grandchild well-being. Here, we present, to our knowledge, the first study to examine grandmaternal brain function. We recruited 50 grandmothers with at least one biological grandchild between 3 and 12 years old. Brain function was measured with functional magnetic resonance imaging as grandmothers viewed pictures of their grandchild, an unknown child, the same-sex parent of the grandchild, and an unknown adult. Grandmothers also completed questionnaires to measure their degree of involvement with and attachment to their grandchild. After controlling for age and familiarity of stimuli, viewing grandchild pictures activated areas involved with emotional empathy (insula and secondary somatosensory cortex) and movement (motor cortex and supplementary motor area). Grandmothers who more strongly activated areas involved with cognitive empathy (temporo-parietal junction and dorsomedial prefrontal cortex) when viewing pictures of the grandchild desired greater involvement in caring for the grandchild. Finally, compared with results from an earlier study of fathers, grandmothers more strongly activated regions involved with emotional empathy (dorsal anterior cingulate cortex, insula and secondary somatosensory cortex), and motivation (nucleus accumbens, ventral pallidum and caudate nucleus). All in all, our findings suggest that emotional empathy may be a key component of grandmaternal responses to their grandchildren.

1. Introduction

Considerable cross-cultural evidence indicates that humans are cooperative breeders, meaning that offspring are cared for not only by mothers, but also by non-maternal caregivers known as allomothers [1–3]. In many societies, grandmothers are important allomothers, and grandmaternal investment is often associated with improved grandchild health and well-being. For example, among the Hadza hunter–gatherers from Tanzania, grandmothers provision their grandchildren with difficult to extract tubers, and this improves grandchild growth [4]. Grandmothers also provide considerable direct care, such as holding, carrying, cleaning, feeding and supervising infants and children. In rural Pakistan, high grandmaternal involvement in direct caregiving is positively associated with cognitive, fine motor and socioemotional development [5]. More generally, grandmaternal investment has been associated with improved cognitive functioning, improved academic achievement, and decreased emotional and behavioural problems among children [6,7]. In a review of alloparental effects in traditional human societies, the presence of a maternal grandmother improved child survival in 9 of 13 societies [8]. Grandmothers can also increase their daughter's fertility, and hence their own inclusive fitness, by decreasing their daughter's interbirth interval [9]. Findings like these support the hypothesis that the benefits of grandmaternal caregiving selected for an extension of the human female lifespan up to decades beyond the cessation of reproduction [10,11]—a life-history strategy that is unique among primates [12,13]. However, the grandmother hypothesis is not the only theory that has been proposed to explain human female post-reproductive longevity [14].

While grandmothers have probably been an important source of support for parents and their children throughout human evolution, a number of factors have contributed to their increasing importance in modern times. As life expectancy has increased, children are much more likely to have living, healthy grandparents today than they were in the past [15]. In Switzerland, for example, it has been estimated that only 27% of 20 year-olds had a living grandparent in 1900. By the year 2000, that figure had increased to 92% [16]. In addition, as fertility rates have declined in many parts of the world, individual children are receiving more grandmaternal investment because that investment is being partitioned among fewer children [15]. Another factor contributing to increased grandmaternal caregiving in modern societies is the large number of mothers who have entered the workforce [17]. Grandmothers are often important sources of childcare for working parents, especially where state-supported childcare is unavailable [18]. In some parts of the world, parents from rural families are migrating to urban centers for job opportunities, leaving their children behind to be cared for by grandparents [15]. Finally, as rates of non-marital birth and divorce have increased in many parts of the world, single motherhood has become more common and grandmaternal assistance more essential [15].

If grandmaternal investment is the evolutionary explanation for human female post-reproductive longevity, then grandchildren should be a particularly salient stimulus to the brain of post-reproductive women. Furthermore, unlike the offspring of great apes, human children are weaned prior to feeding independence, and depend on allomaternal care for their survival. As such, there has probably been strong selection pressure for children to look and act in ways that allomaternal brains find appealing [1].

In addition to grandmothers, fathers are often important allomothers. In some societies, children are more likely to survive when they have a father present [8,19]. Provisioning is the most fundamental and cross-culturally consistent form of paternal investment [20]; however, fathers may also help with direct caregiving. The ethnographic record reveals considerable cross-cultural variation in the degree of direct paternal caregiving [21]. While fathers in some cultures are minimally involved, human fathers have the potential to be highly involved, attached and nurturing caregivers [22]. Research conducted primarily in modern developed societies indicates that positive paternal engagement is associated with better cognitive, behavioural, social and psychological outcomes for children [23–26].

Here, we present, to our knowledge, the first study to examine grandmaternal brain function in response to grandchild stimuli, and we then compare these results with those from a sample of fathers collected previously.

The neurobiology of parental caregiving has been intensively studied in laboratory rats. Numan [27] argues that rats have separate neural systems motivating approach and avoidance of offspring, and that parental behaviour emerges when approach motivation exceeds avoidance motivation. The medial preoptic area (MPOA) is a critical node that both activates the mesolimbic dopamine (DA) (approach) system and inhibits an avoidance circuit that projects from the medial amygdala to the periaqueductal grey of the midbrain. Pregnancy-related hormones like oxytocin (OT), prolactin and oestrogen augment MPOA function so that maternal behaviour emerges at parturition. OT also facilitates maternal behaviour through actions on the mesolimbic DA system (ventral tegmental area (VTA)) and nucleus accumbens (NA) to facilitate DA release in NA, which suppresses NA inhibition of ventral pallidum (VP) neurons, allowing the VP to become responsive to infant stimuli [28]. This same model may also apply to allomaternal caregiving [29,30]. For example, MPOA lesions inhibit paternal behaviour in California mice [31], and optogenetic stimulation of MPOA galanin neurons facilitates paternal behaviour in laboratory mice [32]. Additionally, allomaternal care in female prairie voles depends on OT acting in the NA [33].

Although neuroimaging studies suggest that human parents also make use of this subcortical parental motivation circuit, they also recruit cortical regions involved with understanding others' facial expressions (inferior frontal gyrus (IFG)), others' feelings (anterior insula (AI), anterior cingulate cortex (ACC)) and others' thoughts and beliefs (dorsomedial prefrontal cortex (dmPFC) and temporo-parietal junction (TPJ)) [34–36]. Human fathers seem to recruit a similar set of brain regions as human mothers, consistent with the idea of a global parental caregiving system that is mainly consistent across parents [30,37,38]. For example, one study showed increased activation in the VTA when fathers viewed pictures of their young children, and the magnitude of that activation was positively correlated with the degree to which fathers were involved with instrumental care of their child [39]. In other research, intranasal OT administration was found to both increase paternal sensitivity [40,41] and to augment fathers' neural response to pictures of their child within areas involved with reward, motivation and empathy [42]. Given these findings for fathers, we hypothesized that grandmothers would recruit the same global parental caregiving system as mothers and fathers when presented with grandchild stimuli.

This hypothesis gave rise to the following specific prediction.

Prediction 1: when viewing photographs of their grandchildren, as compared with photographs from unknown children and known and unknown adults, grandmothers will activate brain areas involved in maternal motivation (MPOA, midbrain, NA, caudate nucleus, VP), emotional empathy (AI and dorsal anterior cingulate cortex (dACC)), and cognitive empathy (IFG, dmPFC and TPJ).

Given previous evidence that parental involvement and attachment can modulate brain activation in response to child stimuli [38,39,43], we also predicted that grandmaternal involvement and attachment would modulate brain activation.

Prediction 2: grandmaternal contact (actual and desired), care, warmth and financial investment will be positively correlated with grandmaternal neural responses to viewing pictures of their grandchild within the brain areas involved in maternal motivation (MPOA, midbrain, NA, caudate nucleus, VP), emotional empathy (AI and dACC) and cognitive empathy (IFG, dmPFC and TPJ).

These hypotheses and predictions as well as our analysis plan were pre-registered at: Open Science Framework (https://osf.io/cd8rw).

2. Material and methods

(a) . Participants

Participants were 50 women under 90 years of age (59.26 ± 7.80) who had at least one biological grandchild between 3 and 12 years of age. If grandmothers had more than one grandchild in this age range, they were asked to choose the one grandchild they felt emotionally closest to as the focus for this study. The 50 focal grandchildren averaged 6.98 years (±2.81) of age. Twenty three grandchildren were female. We also recruited each of the grandchildren's same-sex parent to provide stimuli for our functional magnetic resonance imaging (fMRI) experiment (see below). If the same-sex parent was not available, we instead used a same-sex relative of a similar age as the parent (e.g. if a boy's father was unavailable, we might use his uncle).

Recruitment and inclusion/exclusion criteria are described in the electronic supplementary material.

(b) . Stimuli collection

Parents of the selected grandchild provided front-facing, headshot photographs of the grandchild and the grandchild's same-sex parent, with both happy and neutral facial expressions.

(c) . Data collection procedures

(i) . Questionnaires

Grandmothers completed 10 questionnaires online via Research Electronic Data Capture (REDCap). Several of these measured grandmaternal involvement with and attachment to the grandchild. The Amended Parental Responsibility Scale is a 14 item scale measuring perceived parenting responsibility [44,45], that has been modified to compare primary responsibilities between grandmother and parent. Respondents designate who has the primary responsibility for a given task using a 5 point Likert scale ranging from 1 (almost always completed by grandparent) to 5 (almost always completed by parent). The Amended Positive Affect Index is a 10 item scale [46] that has been modified to measure the grandmother's assessment of the degree of positive feelings between the grandmother and grandchild. It measures understanding, fairness, trust, respect, and affection within the relationship. This scale was initially created to reflect a parent–child relationship. The Amended Supportive Engagement Behaviours Index is a modified version of a 10 item subscale of the Parent Behaviour Inventory [47]. These questions measure grandmaternal warmth and involve behaviours demonstrating acceptance through affection, shared activities, and emotional and instrumental support to a child. Several other scales were included to measure variables that might impact grandmaternal caregiving and grandmaternal brain function. The PHQ-9 is a 9 item depression section of the Patient Health Questionnaire that asks respondents how often they experience a set of problems which correlate to the criteria for depression as described in the DSM-IV [48]. The Pain Disability Index is a 7 category instrument designed to measure the impact that pain has on one's ability to participate in daily life activities [49]. The Financial Strain Questionnaire is a self-reported measure of financial stress of the respondent [50]. The Amended UCLA Loneliness Scale is a measure of self-reported loneliness, where respondents rate the frequency that they experience different aspects of loneliness [51]. This scale used three questions out of the 20 on the original UCLA Loneliness Scale. The Brief Pain Inventory includes two questions asking respondents to rate their pain on a scale of 1–10 on average and in that moment. These questions provide a subjective assessment of pain levels [52]. Cronbach's alpha for these scales ranged between 0.85 and 0.95 in our study (electronic supplementary material, table S1).

(ii) . Neuroimaging acquisition

After completing the interview, grandmothers were scheduled for an fMRI scan at the Facility for Research and Education in Neuroscience (FERN) on the Emory University campus. Upon arrival, grandmothers completed an MRI safety screening form and provided a passive drool saliva sample for future genotyping analyses that are not included in this manuscript.

Subjects were positioned in the Siemens Trio 3 T MRI scanner. Subjects lay motionless in a supine position in the scanner with padded head restraint to minimize head movement during scanning. Each scanning session began with a 15 s localizer, followed by a 5 min T1-weighted MPRAGE anatomical scan (TR = 1900 ms, TE = 2.27 ms, matrix = 256 × 256, FOV = 250 mm, slice thickness = 1.00 mm, gap = 0 mm). After collecting the anatomical scan, functional scans without contrast were acquired using an EPI sequence with the following parameters: TR = 1200 ms, TE = 30 ms, matrix = 74 × 74, FOV = 220 mm, slice thickness = 3.0 mm, gap = 0 mm, 54 axial slices. TE was minimally decreased from the typical value (32 ms) in order to reduce magnetic susceptibility artefact in the orbitofrontal region.

During the functional scan, participants viewed images of (i) one of their grandchildren, (ii) a sex, race and age-matched unknown child, (iii) the same-sex parent of the grandchild shown, (iv) a sex, race and age-matched unknown adult to the parent, and (v) pictures of a series of objects (kitchen utensils). At the beginning of the task, participants were instructed to, ‘Please observe each picture and try to share the emotions of the person you see’. Participants were also asked to press a button each time they were prompted in writing, in order to verify participant attention during the functional imaging acquisition. Each stimulus was presented for 5 s, with a variable inter-trial interval of between 1 and 4 s. Stimuli were presented in the following order: same-sex parent neutral expression, unknown adult happy expression, unknown grandchild (UGC) neutral expression, own grandchild (OGC) happy expression, object 1, object 2, object 3, object 4, same-sex parent happy expression, unknown adult neutral expression, UGC happy expression, OGC neutral expression, object 5, object 6, object 7, object 8. This sequence was repeated five times for a total duration of 10 min and 13 s.

(d) . Data analysis

Neuroimaging analysis was conducted with the Oxford Center for Functional Magnetic Resonance Imaging of the Brain's software library (FSL, http://www.fmrib.ox.ac.uk/fsl/). The preprocessing pipeline is described in the electronic supplementary material.

A general linear model (GLM) was defined for each subject that models the neural response to the following five regressors: OGC, familiar same-sex parent (FP), unknown sex-matched, age-matched, and race-matched child (UGC), unknown sex-matched, age-matched and race-matched adult (UA), and objects (OBJ) as control stimuli. Each regressor was convolved with a double gamma haemodynamic response function. We specified the following two contrasts: (i) (OGC–UGC) and (ii) (OGC–UGC) –(FP–UA). The individual-level GLM was implemented using FILM (FMRIB's improved linear model).

At the group level (i.e. second-level analysis), the outputs from the individual-level contrasts were analysed with FLAME1 (i.e. FMRIB's local analysis for mixed effects) in FSL [53,54].

Prediction 1: voxel-wise, one sample t-tests were used to compare each of the main contrasts with zero.

Prediction 2: for each of the main contrasts, beta values for each of the five grandmaternal caregiving covariates listed above were compared with zero using a one sample t-test. This tests for an association between the contrast value and the covariates (i.e. whether the contrast is moderated by the covariate).

Our analyses included both voxel-wise analysis defined by regions-of-interest (ROIs) and exploratory whole brain approaches.

(e) . Regions-of-interest analyses

ROI analyses were focused on 10 regions: MPOA, midbrain (substantia nigra and VTA), NA, caudate nucleus, VP, dACC, AI, dmPFC, TPJ, and IFG. ROI definitions are provided in the electronic supplementary material. All 10 ROIs are shown in the electronic supplementary material, figure S1.

Z-statistic (Gaussianized t) images were thresholded at p < 0.05 corrected for multiple comparisons across all ROI voxels based on Gaussian random field theory (i.e. a small volume correction). Only activations including five or more voxels were reported.

To explore the possibility that activations to the same-sex parent were stronger for biological children versus in-laws, a two-sample t-test was used to compare activation between the two groups, across those ROI voxels that were active for the contrast same-sex parent > grandchildren (FP > OGC).

(f) . Whole brain analyses

Voxelwise, whole brain exploratory analyses were thresholded using clusters determined by Z > 3.1 (voxelwise 1-tailed p < 0.001), and a familywise error (FWE)-corrected cluster significance threshold of p < 0.05 was applied to the suprathreshold clusters [55].

3. Results

(a) . Grandmother characteristics (electronic supplementary material, table S2)

The self-reported racial distribution of grandmothers was 26 White, 21 Black, 2 Hispanic and 1 Biracial. Seventeen grandmothers were working full time, nine were working part time and 24 were either retired or not working. Annual income ranged widely from $13 200 to $225 000 (M = $69 984, s.d. = $49 406). In general, grandmothers were highly involved with their grandchildren, spending an average of 23.7 h per week with them (s.d. = 25.6). Ten grandmothers lived with their grandchildren (recorded as 12 × 7 = 84 h per week), and three of these 10 were custodial grandmothers (i.e. primary caregivers). On average, grandmothers spent 3% (s.d. = 3.0%) of their annual income on their focal grandchild (range = 0–17%). According to the Amended Parental Responsibility Scale, grandmothers were typically less involved in instrumental caregiving than parents were (M = 25.1, s.d. = 11.6; scale ranges from 12 to 60, 36 is equal involvement), and their degree of involvement closely matched their preferred level of involvement (M = 25.4, s.d. = 7.2). Grandmothers also reported very high levels of positive affect and supportive engagement towards their grandchild (Positive Affect Scale: 0 min, 40 max; M = 37.0, s.d. = 4.2; Supportive Engagement Scale: 0 min, 50 max; M = 47.4, s.d. = 4.5). Only six grandmothers (12%) showed more than minimal depressive symptomology, with four exhibiting mild depression and two showing moderate depression. Overall, grandmothers reported little to no financial strain (M = 1.5, s.d. = 0.8; 1 = not at all, 4 = a lot). They reported only rarely feeling lonely or isolated (M = 2.0, s.d. = 0.88; 1 = never, 4 = often). Finally, grandmothers in our study had relatively little disability owing to pain (M = 1.1, s.d. = 1.5; 0 = no disability, 10 = worst disability). Correlations among caregiving, demographic and background variables are presented in the electronic supplementary material, table S3. In summary, the grandmothers in our study were mentally and physically healthy, and highly positively involved with their grandchildren.

(b) . Neuroimaging results

(i) . Prediction 1

The contrast between viewing pictures of the OGC and an unknown child of the same age, race and sex as the grandchild (OGC > UGC) yielded widespread activation in areas previously implicated in parental caregiving, including bilateral IFG, bilateral insula, bilateral striatum, thalamus, midbrain, supplementary motor cortex, and anterior and posterior cingulate cortex (figure 1; electronic supplementary material, table S4). ROI analysis confirmed activation in all predicted brain areas, except that the left MPOA activation did not exceed our five-voxel spatial extent threshold (electronic supplementary material, table S5). The OGC and the UGC stimuli differ in terms of their familiarity to the grandmother. To determine if grandchildren activated participant's brains after controlling for both age and familiarity of the stimuli, we examined the contrast [(OGC-UGC)-(FP-UA)]. This contrast yielded widespread activation in left IFG, bilateral insula, bilateral opercular cortex (including the secondary somatosensory cortex, S2), motor cortex and supplementary motor area (SMA) (figure 2; electronic supplementary material, table S6). This pattern of activation overlapped extensively with the Neuroquery [56] brain map for the search term ‘empathy’ (figure 2, red). In analyses of predicted ROIs, activation was detected in the left caudate nucleus, bilateral AI and left IFG (electronic supplementary material, table S5). The reverse contrast which examined activation specific to the familiar parent, after controlling for age and familiarity of the stimulus, revealed activation across bilateral dorsolateral prefrontal cortex (dlPFC), posterior cingulate, precuneus, superior parietal lobe and visual cortex (figure 2; electronic supplementary material, table S6). ROI analyses revealed activation in several of our predicted parental brain areas, including bilateral MPOA, bilateral midbrain, left NA, bilateral VP, right dACC, right AI, right TPJ and bilateral IFG (electronic supplementary material, table S5). In interpreting this result, it is important to note that 64% (32 out of 50) of familiar parents were the grandmothers' biological child.

Figure 1.

Grandmaternal brain activation. Z-statistic image for the contrast grandchild (OGC) > UGC is shown in yellow-orange. Regions in yellow-orange show greater activation to the OGC. Results are thresholded using clusters determined by Z > 3.1 (voxelwise 1-tailed p < 0.001), and a FWE-corrected cluster significance threshold of p < 0.05 was applied to the suprathreshold clusters. (Online version in colour.)

Figure 2.

Grandmaternal brain activation, controlled for age and familiarity of the stimulus. Z-statistic image for the contrast (OGC-UKG)-(FP-UA) is shown in yellow (positive activation) and blue (negative activation). The Neuroquery brain map for the search term ‘empathy’ is shown in red. Results are thresholded using clusters determined by Z > 3.1 (voxelwise 1-tailed p < 0.001), and a familywise error (FWE)-corrected cluster significance threshold of p < 0.05 was applied to the suprathreshold clusters. (Online version in colour.)

To determine if neural responses to biological children differed from the response to in-laws, we identified parental brain ROIs in which there was stronger activation to the familiar parent compared with the grandchild (FP–OGC). This was true of bilateral MPOA, bilateral midbrain, bilateral NA, bilateral VP and bilateral IFG. For each of these ROIs, the average response across all activated voxels did not differ between biological children and in-laws (all p > 0.05).

(ii) . Prediction 2

To determine if grandmaternal brain function was modulated by their degree of attachment and involvement with the grandchild, we tested whether the above contrasts were moderated by scores on the Amended Parental Responsibility Scale, the Positive Affect Index, and the Supportive Engagement Scale, or by the proportion of annual income spent on the grandchild (electronic supplementary material, table S2). There was no significant association between grandmaternal involvement in instrumental caregiving, as measured by the Amended Parental Responsibility Scale, and grandmaternal neural responses. However, we modified this scale to also ask grandmothers how involved they would ‘like to be’ with their grandchild. Grandmothers who scored higher on this aspirational measure had stronger activation within the left TPJ for the contrast OGC > UGC (electronic supplementary material, table S7). Stronger left TPJ activation was also found for the contrast [(OGC-UGC) > (FP-UA)] (figure 3), and for this contrast grandmothers who aspired to be highly involved also had stronger activation in the left dACC and dmPFC (electronic supplementary material, table S8). Warmth and positive affect towards the grandchild, as measured in both the Supportive Engagement Scale and the Positive Affect Index, did not significantly modulate grandmaternal brain function. Nor did the number of hours per week the grandmother reported spending with her grandchild. However, grandmothers who spent a greater proportion of their annual income on their grandchild had stronger activation in the left and right MPOA for the contrast OGC > UGC (electronic supplementary material, table S9). This result was driven by a single grandmother who spent a very large proportion of her income on the grandchild, and the result became non-significant when she was omitted.

Figure 3.

Moderation of grandmaternal brain function by desired involvement with instrumental caregiving. The region of the left TPJ where activation [(OGC-UGC)>(FP-UA)] is correlated with desired involvement with instrumental caregiving is shown in yellow (a). Results are thresholded using clusters determined by Z > 3.1 (voxelwise 1-tailed p < 0.001), and a FWE-corrected cluster significance threshold of p < 0.05 was applied to the suprathreshold clusters. Data from the peak voxel within the left TPJ are plotted (b). (Online version in colour.)

(c) . Comparison with fathers

To determine the specificity of these grandmaternal neural responses, we compared these results with those from a sample of fathers we studied previously using a similar paradigm [42]. For the contrast between the father's own child and an unknown child, activation was observed in the caudate nucleus, substantia nigra, thalamus, medial prefrontal cortex, posterior cingulate cortex and visual cortex. ROI analysis revealed activation in only a subset of the areas that were activated in grandmothers for the equivalent contrast, including bilateral midbrain, left caudate nucleus and bilateral VP (electronic supplementary material, table S10). To explore this difference more systematically, we pooled results from n = 30 fathers and an equivalent number of grandmothers, and we statistically compared activations for the same contrast between the two groups. For the contrast between own child (for fathers) or OGC (for grandmothers) and an unknown child (for fathers) or an UGC (for grandmothers), grandmothers showed stronger activation than fathers in several areas, including: medial orbitofrontal cortex (mOFC), right precentral gyrus, right insula, bilateral NA, bilateral S2, subgenual and dACC. Fathers show stronger activation in visual cortex, cerebellum and left dorsolateral prefrontal cortex (figure 4 and table 1). ROI analysis showed stronger activation in grandmothers than fathers in bilateral NA, right VP and right dACC, whereas fathers showed stronger activation in left TPJ (electronic supplementary material, table S11).

Figure 4.

Grandmaternal versus paternal brain activation. Regions where grandmothers have stronger activation than fathers for the contrast own child (or grandchild)—unknown child (or grandchild) are shown in yellow. Regions where fathers have stronger activation are shown in blue. Results are thresholded using clusters determined by Z > 3.1 (voxelwise 1-tailed p < 0.001), and a FWE-corrected cluster significance threshold of p < 0.05 was applied to the suprathreshold clusters. Bottom: plot of contrast values within the peak voxel of the dACC for fathers and grandmothers. (Online version in colour.)

Table 1.

Activated brain regions.

region of interest	MNI coordination of local maxima (mm)			local maxima z	number of voxels
	x	y	z
own child–unknown child (for fathers) or OGC–UGC (for grandmothers)
positive direction (grandmothers > fathers)
right opercular cortex, extending into right post central gyrus	66	−12	16	4.91	1164
left mOFC extending into left subgenual anterior cingulate cortex and left NA	−18	26	−14	5.75	862
dorsal anterior cingulate cortex	0	12	26	5.0	318
left pre central gyrus, extending into post central gyrus, and opercular cortex	−54	−6	22	4.11	195
left post central gyrus, extending into left precentral gyrus	−62	−12	36	3.9	170
left post central gyrus	−64	−20	28	4.36	169
negative direction (fathers > grandmothers)
bilateral visual cortex	4	−84	0	4.82	979
left cerebellum	−16	−78	−22	4.22	175
left dlPFC	−46	56	0	4.98	160

4. Discussion

Based on our own review of the literature on parental brain function [34,35,37], along with Numan's recent synthesis of research in animals and humans [57], we identified 10 different brain regions that are involved in important aspects of parental behaviour, including parental motivation, emotional empathy and cognitive empathy. We predicted that grandmothers would activate each of these regions when viewing pictures of their grandchildren. This prediction was supported insofar as grandmothers activated all 10 regions for the contrast between viewing grandchildren and unknown children of the same age. This finding supports the notion of a global parental caregiving system that is used by mothers and allomothers alike. Other studies also support this conclusion. For example, Abraham and colleagues showed that primary caregiving mothers, secondary caregiving fathers and primary caregiving fathers all similarly activated several of these parental brain regions, including the ACC, IFG, insula, and VTA, when viewing videos of themselves interacting with their infant [38].

Among the 10 activated regions was the MPOA. Studies with rodents have established the MPOA as a critical hub in parental brain circuitry [32,57,58], but the MPOA has rarely been reported in human neuroimaging studies of parental brain function, perhaps because of its relatively small size. Yet, we were able to define an approximate MPOA ROI within the human hypothalamus and demonstrate activation within it for grandmothers. We encourage future studies to adopt similar methodology to determine if this is a reproducible finding in human mothers and allomothers.

For several parental brain ROIs, grandmothers showed significantly stronger activation to the same-sex parent than the grandchild. In most cases, the same-sex parent was the grandmother's own biological child, and this could explain why parental brain areas are activating to the same-sex parent. To explore this possibility, we compared grandmaternal activation to biological children with that of in-laws for the ROI voxels that were more active for same-sex parents than grandchildren. Contrary to expectations, there was no difference between the two groups (all p > 0.05). On the one hand, this may call into question the specific role of these brain regions in parenting. On the other hand, it is possible that grandmothers come to develop parental orientations towards their son-in-law or daughter-in-law. In general, grandmothers reported quite positive relationships with their in-laws. Grandmothers rated the average quality of the relationship as 6.2 (s.d. = 3.4) on a 10 point scale (0 = poor and 10 = excellent). Beyond these parental brain ROIs, whole brain analyses additionally showed stronger activation to same-sex parents versus grandchildren within a number of other areas, including most prominently the precuneus. The role of the precuneus in theory of mind [59] might indicate greater engagement of cognitive empathy for the grandmother's adult child or in-law, as compared with her grandchild. On the other hand, areas involved in emotional empathy, such as the insula and secondary somatosensory cortex [56,60], were more strongly recruited by grandchild stimuli (figure 2).

Previously, we showed that paternal involvement in instrumental caregiving was positively correlated with the paternal midbrain response to viewing pictures of their child [39]. Accordingly, we reasoned that grandmaternal brain activation would also be modulated by their degree of attachment or involvement with the grandchild. However, we found only limited evidence for this. Only one of our five measures yielded significant correlations with grandmaternal brain activation. The extent to which grandmothers wished to be involved in caring for their grandchild was positively correlated with activation within the left TPJ. The left TPJ is involved in mental state attribution and perspective-taking [61–63]. Thus, it may be that grandmothers who have more cognitive empathy for their grandchild wish to be more involved in caring for them. Interestingly, adult caregivers of dementia patients who have more cognitive empathy report less subjective stress and depression [64,65]. This measure of desired involvement in caregiving may be a purer reflection of caregiving motivation because circumstances beyond the grandmothers' control (e.g. not living in the same city) may impact their actual involvement in caregiving.

In addition to grandmothers, fathers can also be important allomothers. Although not as reliably as grandmothers, fathers appear to keep children alive in many pre-industrial societies [8], and perhaps even in some modern developed societies [19]. Paternal provisioning of children is common across human cultures and probably facilitated mothers' ability to shorten their interbirth intervals and increase their fertility in comparison with other great apes [3,66,67]. Using a previously collected dataset from human fathers [42], we found that although fathers activated several of our parental brain regions when viewing pictures of their children, they also did not activate several others. While fathers activated a number of subcortical regions involved in parental motivation such as the midbrain, caudate nucleus and VP, in contrast to grandmothers, they did not activate cortical areas involved with emotional empathy, such as the AI and dACC. Nor did they activate other key empathy-related areas that were activated in grandmothers, such as the IFG and TPJ. They did, however, show activation in the dmPFC, which has been linked with cognitive empathy (e.g. theory of mind) [59]. A direct statistical comparison between fathers and grandmothers showed that not only did grandmothers have stronger activation in many of these areas linked with empathy (e.g. dACC, insula, S2; figure 4), they also showed stronger activation in subcortical motivational areas such as the NA, VP and caudate nucleus, along with the mOFC, a known target of the mesolimbic DA system. Thus, overall, grandmothers exhibited a pattern of activation suggestive of greater empathy and stronger parental motivation than that seen in fathers. It should be noted that both grandmothers and fathers were instructed to empathize with the people in the pictures they were viewing, so it remains to be seen if grandmothers would engage these brains regions to the same extent in the absence of explicit instructions.

It would be premature to conclude from our study alone that grandmothers are more empathic and motivated allomothers than fathers are. It is possible that our sample of grandmothers or fathers were not representative. For example, our grandmothers were remarkably mentally and physically healthy, and they were also highly positively engaged with their grandchildren. It may be that women who are healthy and who have enjoyed their experience as a grandmother are more likely to participate in a research study on grandmothers. However, there may be a similar bias at work for fathers. It may also be that other types of child stimuli (e.g. laughter, crying, play behaviour) would yield equal or stronger activation in parental brain regions in fathers as compared with grandmothers. Moreover, there is apt to be significant variation among fathers such that some fathers are as, or more, empathic and motivated than the average grandmother. Indeed, overlapping distributions can be seen clearly in our own fMRI data, in which some fathers have dACC activation that exceeds the average response in grandmothers (figure 4). Finally, while the task paradigm we used for fathers was similar to that we used for grandmothers, there were some notable differences that could have influenced our comparison of the two groups. In the fatherhood study [42], each picture stimulus was presented for 3 s as compared with 5 s in the grandmother study. In addition to the happy and neutral facial expression stimuli used for grandmothers, fathers also viewed pictures of their children with sad facial expressions. Fathers viewed many more pictures of the child than grandmothers did (48 versus 10), but this would be expected to bias our findings in the opposite direction of what was found (i.e. stronger activation in grandmothers). Finally, fathers viewed their children in stimuli blocks that followed blocks with pictures of unknown children and adults, whereas grandmothers viewed their grandchild's picture mixed in sequence with other stimuli.

This study suggests a number of potential directions for future research. Functional neuroimaging provides one window into the neurobiology of grandmaternal caregiving. However, future studies might examine neuroendocrine correlates of grandmaternal care as has been done in mothers and fathers [68,69]. For example, one study found higher levels of urinary vasopressin in grandmothers compared with a well-matched control sample of non-grandmothers [70]. Human allomothers are not limited to fathers and grandmothers [1,2]. Future studies might focus on allomaternal brain function in other relatives such as older siblings, aunts, uncles and grandfathers. However, there may be meaningful differences among allomothers. For example, fathers have come to occupy a special role in many modern families [71]. It would also be intriguing to know if the quality of care provided by unrelated caregivers such as babysitters, nannies and day care workers could be predicted by the extent to which they are able to engage parental brain circuitry in response to stimuli from the child they are caring for. Future studies that compare activation between various types of caregivers, as we did here with grandmothers and fathers, should scan all caregivers using an identical task paradigm so that differences can be confidently attributed to the carer variable. Finally, as mentioned above, future studies might also explore allomaternal responses to other types of child stimuli such as laughter, crying, speech, play, etc.

This study has certain limitations. Our sample of grandmothers was mentally and physically very healthy, highly positively engaged with their grandchildren, and experienced little to no financial strain. As such, our results may not generalize to samples of grandmothers who are less healthy, less positively engaged with their grandchildren, and experiencing greater economic challenges. Also, despite acquiring a relatively large sample of grandmothers as a strategy for protecting against false positive findings [72], we examined multiple contrasts across multiple ROIs. Therefore, despite statistically correcting for comparisons across voxels, our results may still be vulnerable to false positives. This is less likely to be true for the whole brain analyses of main effects, where results were very robust (e.g. figures 1 and 2). Further confirmatory research will be important to verify our other findings.

In conclusion, we find that when grandmothers view photos of their grandchild, they particularly activate brain regions involved in emotional empathy, and regions involved in movement, motor preparation and motor simulation. Furthermore, in comparison with fathers, grandmothers more strongly activate brain regions involved with emotional empathy, as well as brain regions involved with reward and motivation.

Data accessibility

Data are publicly available at: https://osf.io/6vxa4/?view_only=b0270d7d95f44bdd874897e9efd6311b [73].

Authors' contributions

J.K.R.: conceptualization, resources, supervision, writing—original draft, writing—review and editing; A.G.: data curation, formal analysis, writing—review and editing; M.L.: formal analysis, supervision, writing—review and editing. All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

The authors have no competing interests to declare.

Funding

This work was supported in part by NIH grant no. P50MH100023.