Potential Neural Mediators of Mom Power Parenting Intervention Effects on Maternal Intersubjectivity and Stress Resilience

Abstract

Stress resilience in parenting depends on the parent's capacity to understand subjective experiences in self and child, namely intersubjectivity, which is intimately related to mimicking other's affective expressions (i. e., mirroring). Stress can worsen parenting by potentiating problems that can impair intersubjectivity, e.g., problems of "over-mentalizing" (misattribution of the child's behaviors) and "under-coupling" (inadequate child-oriented mirroring). Previously we have developed Mom Power (MP) parenting intervention to promote maternal intersubjectivity and reduce parenting stress. This study aimed to elucidate neural mechanisms underlying the effects of MP with a novel Child Face Mirroring Task (CFMT) in functional magnetic-resonance-imaging settings. In CFMT, the participants responded to own and other's child's facial pictures in three task conditions: (1) empathic mirroring (Join), (2) non-mirroring observing (Observe), and (3) voluntary responding (React). In each condition, each child's neutral, ambiguous, distressed, and joyful expressions were repeatedly displayed. We examined the CFMT-related neural responses in a sample of healthy mothers (n = 45) in Study 1, and MP effects on CFMT with a pre-intervention (T1) and post-intervention (T2) design in two groups, MP (n = 19) and Control (n = 17), in Study 2. We found that, from T1 to T2, MP (vs. Control) decreased parenting stress, decreased dorsomedial prefrontal cortex (dmPFC) during own-child-specific voluntary responding (React to Own vs. Other's Child), and increased activity in the frontoparietal cortices, midbrain, nucleus accumbens, and amygdala during own-child-specific empathic mirroring (Join vs. Observe of Own vs. Other's Child). We identified that MP effects on parenting stress were potentially mediated by T1-to-T2 changes in: (1) the left superior-temporal-gyrus differential responses in the contrast of Join vs. Observe of own (vs. other's) child, (2) the dmPFC-PAG (periaqueductal gray) differential functional connectivity in the same contrast, and (3) the left amygdala differential responses in the contrast of Join vs. Observe of own (vs. other's) child's joyful vs. distressed expressions. We discussed these results in support of the notion that MP reduces parenting stress via changing neural activities related to the problems of "over-mentalizing" and "under-coupling." Additionally, we discussed theoretical relationships between parenting stress and intersubjectivity in a novel dyadic active inference framework in a two-agent system to guide future research.

Keywords: Bayesian active inference; PAG = periaqueductal gray; amygdala; dorsomedial prefrontal cortex; empathy; intersubjectivity; parenting intervention; parenting stress.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Ho, Muzik, Rosenblum, Morelen, Nakamura and Swain.

Figures

Figure 1

The design of Child Face Mirroring Task. Note that the task order in this figure did not represent the actual order. To protect the privacy, the pictures used in the task are not included here. However, examples of the task stimuli can be found in (71).

Figure 2

Whole brain results in the reference sample of healthy mothers (n = 45) from Study 1: Brain regions that were relatively activated (in hot color) or deactivated (in cool color) in pairwise Task contrasts of Observe vs. Rest (A), React vs. Rest (B), Join vs. Rest (C), React vs. Observe (D), Join vs. Observe (E), and Join vs. React (F).

Figure 3

Key results in the reference sample of healthy mothers (n = 45) from Study 1: Brain regions that were relatively activated (in hot color) or deactivated (in cool color) in pairwise Task contrasts of Join vs. Observe (A), React vs. Observe (B), and Join vs. React (C). The dmPFC was inhibited in Join vs. Observe and Join vs. React, with the bar charts for each Task's mean (±s.e.) separately (D). pMCC, posterior middle cingulate cortex; SMA, supplemental motor area; dmPFC, dorsomedial prefrontal cortex; Insula/FO, insula/frontal operculum; L. AMY, left amygdala; OFC, orbitofrontal cortex; FIO, frontal inferior orbital. *p < 0.05.

Figure 4

From T1 to T2 in the clinical study sample (Study 2), MP, relative to Control, showed greater inhibition in the dmPFC during React to Own vs. Other Child, with the bar charts for each Task's mean (±s.e.) separately. *p < 0.05.

Figure 5

The left amygdala's MMR(all) [Join[Own vs. Other Child] vs. Observe[Own vs. Other Child]] differential responses was activated in Joyful vs. Distressed expression, while it was inhibited in the Distressed expression in the reference sample, with the bar charts of each expression's mean (±s.e.) separately (A). From T1 to T2 in the clinical study sample, MP, relative to Control, showed greater activation in the Joyful expression in the Join[Own vs. Other Child] vs. Observe[Own vs. Other Child], with the bar charts for each expression's mean (±s.e.) separately (B). *p < 0.05.

Figure 6

From T1 to T2 in the clinical study sample, MP, relative to Control, showed greater differential responses of MMR(all) [Join[Own vs. Other Child] vs. Observe[Own vs. Other Child]] in the left frontoparietal regions (A), midbrain (B), left nucleus accumbens (NAc) (C), left and right amygdala (AMY) (D,E), with the bar charts of each region's mean (±s.e.) (F). *p < 0.05.

Figure 7

From T1 to T2 in the clinical study sample, the T2-T1 differences in parenting stress index (dPSI) were negatively associated with the concomitant increases in the MMR(all) [Join[Own vs. Other Child] vs. Observe[Own vs. Other Child]] differential responses in the left superior temporal gyrus (STG) (A), right STG/insula (B), cerebellum (C), and hypothalamus (D), each with the dPSI depicted on the x-axis, against the T2-T1 difference in the region's differential response on the y-axis, in the scatter plots. The Pearson's correlation r scores and p-values are embedded in the plots. The bar charts of each region's mean (±s.e.) are depicted in (E). *p < 0.05.

Figure 8

From T1 to T2 in the clinical study sample, the T2-T1 differences in parenting stress index (dPSI) were positively and negatively associated with the concomitant increases in the MMR(all) [Join[Own vs. Other Child] vs. Observe[Own vs. Other Child]] differential functional connectivity [MMR(all)-dependent PPI] between the dmPFC and PAG (A) and that between the dmPFC and NAc (B), respectively, each with the dPSI depicted on the x-axis, against the T2-T1 difference in the region's differential response on the y-axis, in the scatter plots. The Pearson's correlation r scores and p-values are embedded in the plots. The MP vs. Control differed in the MMR(all)-dependent PPI between dmPFC and PAG, but not that between dmPFC and NAc, with the bar charts of each region's mean (±s.e.) depicted in (C). *p < 0.05.

Figure 9

From T1 to T2 in the clinical study sample, the T2-T1 differences in parenting stress index (dPSI) were negatively associated with the concomitant increases in the MMR(j-d) [Join[Own vs. Other Child's Joyful vs. Distressed] vs. Observe[Own vs. Other Child's Joyful vs. Distressed]] differential responses in the left amygdala (A) and right NAc (B), but they were positively associated with that in the PAG (C), each with the dPSI depicted on the x-axis, against the T2-T1 difference in the region's differential response on the y-axis, in the scatter plots. The Pearson's correlation r scores and p-values are embedded in the plots. The MP vs. Control differed in the MMR(j-d) in the left amygdala, but not the right NAc and PAG, with the bar charts of each region's mean (±s.e.) depicted in (D). *p < 0.05.

Figure 10

Scatter plots of Study 2 T2-T1 changes in PSI (x-axis) and T2-T1 differential responses in the left amygdala (y-axis) in the contrasts of MMR(joy) (A), MMR(dis) (B), MMR(amb) (C), and MMR(neu) (D). The T2-T1 left amygdala MMR(all) responses were increased in MP but decreased in Control group (E). *p < 0.05.

Figure 11

The single-mediator model for each of the three mediators: M1 = T2–T1 differences in the MMR(all) in the left STG (A), M2 = T2–T1 differences in MMR(all)-dependent PPI between dmPFC and PAG, (B) and M3 = T2–T1 differences in the MMR(j-d) in the left amygdala showed that each mediator significantly mediated the MP effects on reducing parenting stress from T1 to T2 (C). The age of Own Child was used as a covariate in all mediation models. See Table 3 for the statistical results of these three single-mediator models.

Figure 12

A Bayesian active inference framework for a single-agent system (A), a strongly coupled dyadic system (B), and an under-coupled dyadic system (C). In (A), an agent and its environments form a single-agent system, depicted as a four-node hierarchical network. E, a node representing events from environments at the bottom level; S, a node representing the agent's sensation; A, a node representing the agent's action; M, a node representing the agent's internal model. The S and A nodes are positioned at the intermediate level and the M node is positioned at the top level. The prediction error, defined as the difference between the data in the S node and the prediction in the A node, is bounded by free energy. When the free energy is minimized by the M node, the agent can reliably predict the environments, and thus the adaptation of the agent to the environments is optimized. In (B), a strong coupling between two agents is formed when Agents 1 and 2 are coupled by their S's and A's nodes, wherein A1 causes S2 and A2 causes S1. Due to the coupling, each agent's prediction errors are also coupled and thus the adaptation is optimized when the collective free energy is minimized. In an optimal state, M1 and M2 will be highly consistent with one another, indicating a high level of intersubjectivity. In (C), under-coupling ensues when Agent 1 discards Agent 2's M2 and S2 and instead only focuses on Agent 2's behaviors A2 in relation to Agent 1's S1 and A1. Due to the under-coupling, Agent 1 tends to misattribute the causes of Agent 2's behaviors.