The experience of being a mother involves functional and structural brain changes that support the establishment of sensitive caregiver responses to infants (Kim, Strathearn, & Swain, Reference Kim, Strathearn and Swain2016). Mothers, compared with non-mothers, have been shown to exhibit a specific right prefrontal response when discriminating infant facial expressions (Nishitani, Doi, Koyama, & Shinohara, Reference Nishitani, Doi, Koyama and Shinohara2011). Mothers also respond to their own versus unfamiliar infants’ cry sounds and faces with a more intense activation in regions involved in viso-emotional processing, empathy, or emotion regulation (Kim et al., Reference Kim, Strathearn and Swain2016; Rocchetti et al., Reference Rocchetti, Radua, Paloyelis, Xenaki, Frascarelli, Caverzasi and Fusar-Poli2014). A higher activation is also found when mothers are responding to an emotional face-matching task during the late than in the early postpartum period, which may reflect long-lasting adjustments in emotional empathy-related areas such as the insula and the middle and inferior frontal gyrus (IFG; Gingnell et al., Reference Gingnell, Bannbers, Moes, Engman, Sylvén, Skalkidou and Sundström-Poromaa2015). Brain changes before and after pregnancy are associated with pronounced and long-lasting gray matter (GM) volume reductions in empathy areas such as the superior temporal sulcus, anterior cingulate cortex, and middle and inferior frontal gyrus, among others (Hoekzema et al., Reference Hoekzema, Barba-Müller, Pozzobon, Picado, Lucco, García-García and Vilarroya2017). Moreover, some of these areas also undergo structural increases in GM during the postpartum period, suggesting maternal brain adaptations to the child's evolving needs (Barba-Müller, Craddock, Carmona, & Hoekzema, Reference Barba-Müller, Craddock, Carmona and Hoekzema2019). From a complementary perspective, the current study examined possible volume differences in those mothers exhibiting a drastic disregard of their own child's needs compared with those showing sensitive caregiving as well as whether these volumetric differences are functionally related to observed mother–child interactive behavior. Finding this association may further illustrate how critical the empathy-related areas are for appropriate mother–child bonding interactions.

An example of extremely insensitive caregiving is maternal neglect that consists of the mothers’ failure to provide for the child food, clothing, shelter, medical care, supervision, or emotional support. It is the most common and severe form of child maltreatment that puts the child's safety at risk (Petersen, Joseph, & Feit, Reference Petersen, Joseph and Feit2014; Stoltenborgh, Bakermans-Kranenburg, & van IJzendoorn, Reference Stoltenborgh, Bakermans-Kranenburg and van IJzendoorn2013). Negligence also disrupts the establishment of a child's secure attachment and healthy psychosocial development (Weinfield, Sroufe, Egeland, & Carlson, Reference Weinfield, Sroufe, Egeland, Carlson, Cassidy and Shaver2008), and it entails negative and cumulative behavioral and neurobiological alterations for the offspring (see a review by Teicher, Samson, Anderson, & Ohashi, Reference Teicher, Samson, Anderson and Ohashi2016). Studying the neurobiological basis of maternal neglect could help in the tuning of intervention efforts directed at reducing risk and promoting infant health.

Our first objective was to examine GM and WM volume differences between neglectful (NM) and non-neglectful control (CM) mothers who are sociodemographically similar. In searching for brain differences between NM and CM, we expected that at least part of the volumetric differences would overlap with empathy-related areas. Those areas undergo functional and structural adaptations to sensitive caregiving in normal mothering (Kim et al., Reference Kim, Strathearn and Swain2016). Parental empathy, defined as the appropriate perception, understanding, and experience of an infant's emotional states, is one of the crucial abilities for providing caring responses to one's own infant's needs that is lower in NM (León et al., Reference León, Rodrigo, Quiñones, Hernández, Lage, Padrón and Bobes2014; Rodrigo et al., Reference Rodrigo, León, Quiñones, Lage, Byrne and Bobes2011). The lower empathic skills that have been observed in NM seem to be tightly coupled with their lower brain reactivity to infant cues. Compared with CM, NM showed a generic attenuated response to infant and adult crying faces in viso-limbic areas such as the bilateral lingual, bilateral cerebellum, bilateral fusiform gyrus, right hippocampus, parahippocampal gyrus, and right amygdala (León et al., Reference León, Rodrigo, El-Deredy, Modroño, Hernández-Cabrera and Quiñones2019). Lesser activations in NM than in CM were also specifically shown to infant crying faces in the left, middle, frontal, and anterior cingulate areas, underscoring their difficulties with responding to the infant's emotional cues (León et al., Reference León, Rodrigo, El-Deredy, Modroño, Hernández-Cabrera and Quiñones2019). In this study, brain differences were assessed using voxel-based morphometry (Ashburner & Friston, Reference Ashburner and Friston2000), an approach used to quantify structural brain properties for the investigation of volume differences in brain anatomy.

With respect to the GM volume, studies have shown that adults with higher scores on the emotional Empathic Concern (EC) scale of the Interpersonal Reactivity Index (Davis, Reference Davis1980), defined as the embodied simulation of and sympathy with others’ emotions, showed greater GM volume in the anterior insula (Eres, Decety, Louis, & Molenberghs, Reference Eres, Decety, Louis and Molenberghs2015; Mutschler, Reinbold, Wankerl, Seifritz, & Ball, Reference Mutschler, Reinbold, Wankerl, Seifritz and Ball2013) and the IFG (Banissy, Kanai, Walsh, & Rees, Reference Banissy, Kanai, Walsh and Rees2012) than those with lower scores. In turn, adults with higher scores on the cognitive the Perspective Taking (PT) scale of the Interpersonal Reactivity Index, defined as the mental-state understanding and perspective taking from others, showed greater GM volume in the middle cingulate cortex and the adjacent dorsomedial prefrontal cortex than those with lower scores (Eres et al., Reference Eres, Decety, Louis and Molenberghs2015). Importantly, GM volume increases in sensitive mothers with perceived higher parental care in childhood were also found in the superior/middle frontal areas, orbitofrontal gyrus, superior temporal gyrus, and fusiform gyrus (Kim et al., Reference Kim, Leckman, Mayes, Newman, Feldman and Swain2010). Given these findings, we expected to find some volumetric differences between NM and CM in empathy-related areas, given the severe disregard of the child's needs and the lower scores in empathic concern that have been observed in NM (León et al., Reference León, Rodrigo, Quiñones, Hernández, Lage, Padrón and Bobes2014; Rodrigo et al., Reference Rodrigo, León, Quiñones, Lage, Byrne and Bobes2011).

Structural differences in WM volume are also expected between NM and CM. White matter disruption has been reported in postpartum depression in the fronto-thalamic circuit and in interhemispheric connectivity (Silver et al., Reference Silver, Moore, Villamarin, Jaitly, Hall, Rothschild and Deligiannidis2018). In turn, a higher EC score in adults has been shown to be positively associated with greater WM structural integrity in the tracts linking areas involved in affective processing of infant faces such as the inferior longitudinal fasciculus and the inferior fronto-occipital fasciculus (IFOF; Parkinson & Wheatley, Reference Parkinson and Wheatley2012). Given that NM compared with CM showed a volume reduction in the inferior longitudinal fasciculus and IFOF in a diffusion tensor imaging study (Rodrigo et al., Reference Rodrigo, León, Góngora, Hernández, Byrne and Bobes2016), we also expected a smaller WM volume in NM than in CM in those areas traversed by these tracts.

As a second objective, we examined the functional effects of GM and WM volume differences between NM and CM on Emotional Availability (EA), as a proxy for the quality of mother–child bonding interactions. A play task that measures the ability to read and respond appropriately to each other's emotional signals in mother–child dyads has been used to assess EA (Biringen, Reference Biringen2000; Biringen, Derscheid, Vliegen, Closson, & Easterbrooks, Reference Biringen, Derscheid, Vliegen, Closson and Easterbrooks2014). Emotional availability is also predictive of the mothers’ reported child attachment (Altenhofen, Clyman, Little, Baker, & Biringen, Reference Altenhofen, Clyman, Little, Baker and Biringen2013). A previous study had shown lower scores in NM than in CM in EA associated with a lesser volume in the inferior longitudinal fasciculus and IFOF (Rodrigo et al., Reference Rodrigo, León, Góngora, Hernández, Byrne and Bobes2016). Therefore, we expected to find a positive association between smaller GM and WM volumes and lower EA scores in NM compared with CM.

Once the possible effect of volumetric differences on maternal sensitivity between NM and CM had been tested, we wanted to go a step further in determining the possible role of trait emotional and cognitive empathy in that relationship. A higher score on EC has been associated with the intention to provide needed care to the child (Lin & McFatter, Reference Lin and McFatter2012) and the expression of maternal warmth and positive affect (Stern, Borelli, & Smiley, Reference Stern, Borelli and Smiley2015). In turn, a higher score on PT has been related to increases in one's own infant's pupil dilation in response to others’ emotional displays, an early precursor of empathy (Upshaw, Kaiser, & Sommerville, Reference Upshaw, Kaiser and Sommerville2015). Therefore, if volumetric differences are found in areas corresponding either to the emotional or cognitive empathy circuits, we expected that introducing EC or PT, respectively, as mediators would increase the possibilities of finding an association between volumetric differences and EA.

Altogether, this study can provide evidence of the volumetric alterations underlying neglectful mothering and the role played by the emotional and cognitive empathy-related brain areas in maternal caregiving.

Methods

Participants

Forty-eight mothers (25 NM and 23 CM) voluntarily participated in the experiment. They were all recruited through the same primary health center in Tenerife, Spain. Written consent was obtained from all of the participants and the Ethics Committee of the University of La Laguna approved the study's protocol. General inclusion criteria were being the biological mother of a child who was under three years old who had not been placed in foster care at any point in their history and had not been born prematurely or suffered perinatal or postnatal medical complications according to the pediatricians’ reports. Specific inclusion criteria for the neglectful mother group were a substantiated case of neglect registered in the last 12 months by Child Protective Services and complying with the indicators of the Maltreatment Classification System for severe neglect (Barnett, Manly, & Cicchetti, Reference Barnett, Manly and Cicchetti1993) according to the pediatrician of the primary health center in charge of the case. Thus, these mothers scored positively on physical neglect (inadequate food, hygiene, clothing, and medical care), lack of supervision (child is left alone or in the care of an unreliable caregiver), and educational neglect (lack of cognitive and socioemotional stimulation and lack of attention to the child's education). Inclusion criteria for the control group were negative scores in all of the Maltreatment Classification System neglect indicators, also according to the pediatrician, and the absence of Child Protective Services or Preventive Services records for the family. As for the sociodemographic profile of mothers, which was reported by the social worker, they were all in their early 30s; they had a similar number of children and mean age of the target child; the NM were more likely than CM to live in one-parent families and to receive financial assistance, indicating an overload in their caregiver task and financial difficulties; and the groups shared similar, low socioeconomic backgrounds (Table 1). According to the neglect risk profile rated by the social workers on the presence or absence of risk indicators for neglect, most mothers in the neglectful group had a history of childhood maltreatment or neglect (of the mother when she was a child). They also scored positively in poor household management, disregard of the child's needs, and rigid/inconsistent discipline norms [see Appendix 1 for details on the risk profile measures].

Table 1. Sociodemographic and neglect risk profile in Neglectful and Control groups

Note: Group comparisons with mean scores were performed with t tests, while those with percentage values were performed with the chi-Square (χ²) statistic. *p < .05 ***p < .001.

Behavioral and personality measures

The Mini International Neuropsychiatric Interview

(M.I.N.I. 6.0, Spanish version; Ferrando, Bobes, Gibert, Soto, & Soto, Reference Ferrando, Bobes, Gibert, Soto and Soto2000). All of the participants completed the M.I.N.I., which includes 15 major psychiatric disorders (Table 2). The two groups mainly differed on five psychopathological variables (marked in italics) that survived the Bonferroni correction, which were evaluated by using a principal component analysis. The results yielded a one factor solution: “Psychiatric Disorders,” with moderate intercorrelations among the five variables, KMO = 0.68, eigenvalue = 2.74, with explained variance of 55%, and the coefficient scores in Psychiatric Disorders being higher in neglectful (M = 0.63, SD = 1.02) than in control mothers (M = −0.69, SD = 0.21); t (46) = 6.3, p = .000; δ = 1.82. None of the mothers in either group were being medicated for psychiatric disorders at the time of testing. The coefficient score for Psychiatric Disorders (PD) was used as a regressor in the statistical parametric mapping (SPM) model. Given that the dichotomized variables (Group for mothers and PD) were certainly related (r = 0.67), we previously ruled out their potential multicollinearity before including each in the model as a covariate. Research shows that the magnitude of such a correlation is not a reliable indicator when there are problems with collinearity between two variables (Belsley, Reference Belsley1991). Therefore, we tested whether our use of “psychiatric disorders” as a covariate in the SPM model with Group had the potential for problematic multicollinearity. We calculated the variance inflation factor, the tolerance, and the condition number, associating each predictor variable with the group as well as the shared variance of the psychiatric variables with the group [See Appendix 2, Table A1]. Once evidence of noncollinearity was obtained, PD was included as a covariate in the SPM model to control as much as possible for its effect on brain volumetric differences. All of the analyses were performed with R (R Core Team, 2019).

Table 2. Psychopathological conditions stratified by group

Note: Variables that are shown in italics were submitted to principal component analysis. *p ≤ 05 **p ≤ .01 ***p ≤ .001

Emotional Availability (EA)

This variable was measured in the context of mother–child free play using the EA Scale: Infancy to Early Childhood Version (Easterbrooks & Biringen, Reference Easterbrooks and Biringen2005). This scale operationalizes parental and child behavior on six subscales that were factorized into one factor, EA, by using principal component analysis, given the high intercorrelations among the scales. Two external observers who were blind to the mothers’ grouping made the ratings from the videos, and the inter-rater reliability of the ratings in each scale was adequate [see Appendix 3 and Table A2 for the scales and testing]. The EA factor score was lower in the neglectful dyads (M = −0.61, SD = 0.92) than in control dyads (M = 0.66, SD = 0.55); t (46) = −5.75, p  = 1.66, and it was used as a dependent variable.

Interpersonal Reactivity Index

(IRI; Spanish version; Pérez-Albéniz, De Paúl, Etxeberría, Montes, & Torres, Reference Pérez-Albéniz, De Paúl, Etxeberría, Montes and Torres2003). Only the Empathic Concern (EC) and the Perspective Taking (PT) scales were used. The EC scale assesses the respondents’ feelings of warmth, compassion, and concern for others (emotional empathy), whereas the PT scale describes the tendency to spontaneously adopt others’ points of view in everyday life (cognitive empathy). The scores on EC and PT were lower in NM (M = 26, SD = 3.59; M = 23.56, SD = 4.35, respectively) than in CM (M = 28.08, SD = 3.56; M = 26.21, SD = 3.74, respectively), and t (46) = −2.01, p = .05; δ = 0.58; t(46) = −2.25; p = .05; δ = 0.65, respectively.

Procedure

After the selection of the sample, social workers reported on the sociodemographic and neglect risk profile and asked mothers for permission to be contacted by phone. Those mothers who gave permission were contacted by our collaborator and were informed about the general goal of the study (to participate in a study about mother–child relationships avoiding the use of the term neglect in any case) and the procedure to be followed upon their acceptance. Then, the collaborator picked them up at their homes at their convenience to bring them to the scanning session at the hospital where they gave their informed written consent and passed the MRI sequence under a resting state condition without stimuli being presented. In a second session carried out at their homes, the same collaborator collected the mothers’ response to the questionnaire, gave a gift to the child, and video recorded the mother–child play interaction. At the end of the session, the mothers were given a monetary compensation (100 euros).

MRI Processing and Analysis

Structural Image Acquisition

High-resolution T1-weighted MPRAGE anatomical volumes were acquired on a General Electric 3T scanner located at the university hospital's magnetic resonance service for biomedical research at the University of La Laguna. A total of 196 contiguous 1mm sagittal slices were acquired with the following parameters: repetition time = 8.716 ms, echo time = 1.736 ms, field of view = 256 × 256 mm², in-plane resolution = 1 mm × 1 mm, flip angle = 12.

Voxel-Based Morphometry Processing

T1 images were preprocessed using the Voxel-Based Morphometry toolbox (http://dbm.neuro.uni-jena.de/vbm.html) and the SPM8 software package. The images were corrected for bias-field inhomogeneity, and the tissue was then classified into gray and white matter and cerebrospinal fluid and registered to standard space using high-dimensional DARTEL normalization (Ashburner, Reference Ashburner2007). The segmentation approach that was used is based on an adaptive maximum a posteriori technique, which does not need a priori information about tissue probabilities (Rajapakse, Giedd, & Rapoport, Reference Rajapakse, Giedd and Rapoport1997). This procedure was further refined by accounting for partial volume effects and by applying a hidden Markov random field model, which incorporates spatial prior information of the adjacent voxels into the segmentation estimation (Tohka, Zijdenbos, & Evans, Reference Tohka, Zijdenbos and Evans2004). To measure regional differences in absolute GM and WM volumes in the obtained volumetric segmentations, the warped images were modulated. All of the normalized modulated images were smoothed with a filter of a 10-mm Gaussian kernel. An additional quality check based on the sample homogeneity was conducted.

Statistical Analyses of GM and WM volumes

General linear model analyses (i.e., in SPM, full factorial design (Gläscher & Gitelman, Reference Gläscher and Gitelman2008) were performed using the individual gray/white matter volumetric segmentations as dependent variables and including Group (Control vs. Neglectful mothers) as a between-subject factor. The models included one regressor described above as Psychiatric Disorders. The age of the mothers (mean centered) was also included as a nuisance covariate.

The resulting statistical parametric maps were thresholded at a peak level of p < 0.001 (uncorrected), adjusting the cluster spatial extent to capture only those clusters corrected for multiple comparisons using the whole-brain or the small-volume FWE corrections (p < 0.05). The second-level inferences were tested using a threshold of p < 0.001 (uncorrected), with a voxel extent adjusted such that only those peaks or clusters with a p-value corrected for multiple comparisons using family-wise error (Nichols & Hayasaka, Reference Nichols and Hayasaka2003) were considered to be significant. All local maxima were reported as Montreal Neurological Institute coordinates. Gray matter regions were labeled with reference to the Automated Anatomical Labeling atlas (AAL; Tzourio-Mazoyer et al., Reference Tzourio-Mazoyer, Landeau, Papathanassiou, Crivello, Etard, Delcroix and Joliot2002).

Statistical Analyses of Brain-Behavior Associations

As a first step, regression analysis was performed with the whole sample to test the hypothesis of whether the presumable group volumetric differences in GM and WM were associated with the mother–child emotional availability in the play task (EA). Based on the results of the second level whole-brain analysis previously described, the volume per region of interest per subject was extracted. Five GM and two WM variables were used. To examine possible Group interaction effects on EA, we first performed a regression analysis in which Group was included with the volumetric variables. Maltreatment status was also included due to its potential association with EA. Because Group did not show any interaction effects with volumetric measures and Maltreatment effects on EA disappeared due to the high redundancy with Group (17 out 25 in NM were maltreated), Group was removed from the regression. Therefore, the regression analysis that is reported includes Maltreatment status and the volumetric variables in the whole sample to determine their effects on EA.

As a second step, we performed mediation analyses with those volumetric variables not exhibiting a direct association with EA in the reported regression model to further test whether they were related to EA through the mediation of EC and PT. For each mediation model with EC/PT, each of the aforementioned brain areas acted as the independent variable, EC/PT as a mediator, and EA as the dependent variable.

Results

Volumetric differences in GM and WM in neglectful as compared with control mothers

The voxel-based morphometry analysis showed three distributed clusters with a pattern of smaller GM volume in NM than in CM. Each of these clusters spanned several contiguous anatomical regions in both hemispheres (Table 3 and Figure 1). Cluster 1 included a large midline region comprising the anterior/middle cingulate (ACC/MCC) cortex with extension to both hemispheres, superiorly to the left precentral gyrus, the left superior frontal gyrus, and the most inferior part of the right supplementary motor area. Cluster 2 involved a broad region from the pars triangularis within the IFG to the adjacent part of the middle frontal gyrus. Cluster 3 was circumscribed to the posterior part of the right insula. In addition, the voxel-based morphometry analysis showed one posterior cluster exhibiting the opposite pattern, that is, greater GM volume in NM than in CM. As this cluster comprised a broad region, it was anatomically divided into two distinct areas, fusiform and cerebellum, to test their effects separately.

Figure 1. GM volume alterations in neglectful mothers in empathy-related regions. Smaller GM volumes in NM as compared to CM were found in the bilateral anterior/middle cingulate cortex, right posterior insula, and right inferior frontal gyrus. A reversed pattern of greater GM volume in NM as compared to CM was found in the right fusiform and the right cerebellum.

Table 3. Gray and white matter volumetric differences showing cingulate, frontal, and insula smaller GM volume clusters (NM < CM); one posterior (cerebellum and fusiform) greater GM volume cluster (NM > CM); and one frontal smaller WM volume cluster (NM < CM).

Note: Whole-brain voxel corrections were applied. All of the reported local maxima belong to a significant cluster (p-value FWE corrected < 0.05) with the exception of the probability value signaled with an asterisk, indicating that this cluster was significant after a small volume correction (cluster-level p-value (FWE-corr) = 0.001; peak-level p-value (FWE-corr) = 0.019). NM: Neglectful mothers; CM: Control mothers; Sup: Superior; Ant: Anterior; Inf: Inferior; Mid: Middle; Supp: Supplementary; Tri: Triangular; R: Right; L: Left.

The WM analyses showed two frontal bilateral clusters of smaller WM volume in NM than in CM, traversed by the association fiber tracts sensitive to individual differences in empathic concern (See Table 3 and Figure 2). For visualization purposes, the significant WM clusters were used to build the 3D regions of interest that are represented in Figure 2 (below). The surrounding pathways were reconstructed using the Q-Space diffeomorphic reconstruction (Yeh & Tseng, Reference Yeh and Tseng2011) in Montreal Neurological Institute space implemented in DTI Studio (Jiang, Van Zijl, Kim, Pearlson, & Mori, Reference Jiang, Van Zijl, Kim, Pearlson and Mori2006). Thus, bilaterally, clusters comprised the IFOF, the corpus callosum, and the anterior corona radiata. The left WM cluster also included the anterior thalamic radiation (Oishi, Faria, Van Zijl, & Mori, Reference Oishi, Faria, Van Zijl and Mori2010). The smaller WM volume in the frontal clusters was located adjacent to the smaller GM volume in the two frontal clusters (Figure 3).

Figure 2. WM volume alterations in neglectful mothers in frontal regions. (a and b) Smaller WM volumes in NM as compared to CM were found in two frontal bilateral clusters, and (c) 3D individual tracking representation (see dotted ovals) shows that the right frontal cluster is traversed by the inferior fronto-occipital fasciculus (IFOF), the corpus callosum, and the anterior corona radiata (ACR), whereas the left frontal cluster is traversed by the anterior thalamic radiation (ATR).

Figure 3. Overlapping of WM and GM volume frontal alterations in neglectful mothers. The smaller WM volumes in the bilateral frontal clusters were located adjacent to the two reduced GM volumes in the bilateral frontal clusters. Notice that only letter (c) in Figure 2 is in italic and should be ordinary found like (a) and (b) and like the ones in Figure 4.

Figure 4. Regression and mediation models of brain differential areas in neglectful mothers on emotional availability (EA). The regression model shows (a) a negative relationship between greater GM volume in Fusiform_R, and (b) a positive relationship between greater WM_R volume in frontal cluster and EA. Mediation models with trait empathy show (c) that emotional concern significantly mediated the positive relationship between IFG and Insula volumetric measures and EA, with no direct effects; and (d) that emotional concern significantly mediated the negative relationship between cerebellum volumetric measures and EA, with no direct effects.

Effects of differences in GM and WM volumes on Emotional Availability

For the regression analyses, the maltreatment classification (21 maltreated and 27 non-maltreated mothers), the five GM areas (ACC/MCC, right inferior frontal gyrus, Insula, Fusiform_R, and Cerebellum) and the two WM areas (WM_L and WM_R frontal areas) showing volumetric differences in the previous section were used as predictors of EA. The results showed as significant regressors the volumetric measures in the GM of Fusiform_R, in the WM of Frontal_R, and Maltreatment status, F (3,44) = 7.98, p < 0.001, explaining a moderate proportion of the variance in EA, R ² = 0.352; AdjR ² = 0.308. Greater volume in frontal WM_R was associated with higher scores in EA, Estimate = 3.73; t (44) = 2.08, p = .04, whereas greater GM volume in Fusiform_R, Estimate = −3.35; t (44) = −1.94, p = .05, and belonging to the maltreatment group, Estimate = −0.69; t (44) = −2.74, p = .01, were associated with lower scores in EA (Figure 4a & b).

Empathic Concern and Perspective Taking as mediators of the effects of GM and WM volumes on Emotional Availability

We performed mediation analyses (Tingley, Yamamoto, Hirose, Keele, & Imai, Reference Tingley, Yamamoto, Hirose, Keele and Imai2014) to further test whether those five areas not directly linked to EA in the regression model (ACC/MCC, IFG_R, Insula_R, Cerebellum, and WM_L) would be related through the mediation of EC and PT. Three of the five mediating models showed significant relations. Using a bootstrap resampling procedure, their parameters fell outside the confidence intervals, indicating that the results are not likely to be random. The analysis indicated that EC mediated in the positive relationship that was observed for both the IFG (Average Causal Mediation Effects, ACME = 4.03, p = 0.001) and the Insula (ACME = 3.39, p = 0.03) on EA (Figure 4c) and in the negative relationship that was observed for the Cerebellum (ACME = −2.56, p = 0.04) on EA (Figure 4d). Neither mediated nor direct significant effects were found for EC in relation to ACC/MCC and WM_L and EA, and no effect was found for PT for any of the models.