Neural Correlates of Altered Emotional Responsivity to Infant Stimuli in Mothers who Use Substances

Abstract

Mothers who use substances during pregnancy and postpartum may have altered maternal behavior towards their infants, which can have negative consequences on infant social-emotional development. Since maternal substance use has been associated with difficulties in recognizing and responding to infant emotional expressions, investigating mothers’ subjective responses to emotional infant stimuli may provide insight into the neural and psychological processes underlying these differences in maternal behavior. In this study, 39 mothers who used substances during the perinatal period and 42 mothers who did not underwent functional magnetic resonance imaging while viewing infant faces and hearing infant cries. Afterwards, they rated the emotional intensity they thought each infant felt (‘think’-rating), and how intensely they felt in response to each infant stimulus (‘feel’-rating). Mothers who used substances had lower ‘feel’-ratings of infant stimuli compared to mothers who did not. Brain regions implicated in affective processing (e.g., insula, inferior frontal gyrus) were less active in response to infant stimuli, and activity in these brain regions statistically predicted maternal substance-use status. Interestingly, ‘think’-ratings and activation in brain regions related to cognitive processing (e.g., medial prefrontal cortex) were comparable between the two groups of mothers. Taken together, these results suggest specific neural and psychological processes related to emotional responsivity to infant stimuli may reflect differences in maternal affective processing and may contribute to differences in maternal behavior in mothers who use substances compared to mothers who do not. The findings suggest potential neural targets for increasing maternal emotional responsivity and improving child outcomes.

Introduction

Maternal substance use (SU) during the perinatal period remains a major public health concern (Forray and Foster, 2015; McCance-Katz, 2019). Prenatal exposure to substances during pregnancy can directly affect the fetus and may disrupt fetal development (Louw, 2018; Richardson et al., 1999; Srikartika and O’Leary, 2015). Additionally, some mothers with SU disorders exhibit differences during postnatal behavioral interactions with their infant, such as difficulties soothing their distressed infant and recognizing their infant’s emotional responses, which may have lasting consequences for the child’s socioemotional and cognitive development (Cataldo et al., 2019; Flykt et al., 2021; Mayes and Truman, 2002; Rutherford et al., 2011; Rutherford et al., 2021; Rutherford and Mayes, 2019). One important mechanism through which mothers appropriately respond to infants is via empathy. Empathy, the ability to understand the experience of another’s feelings, is an important contributor to successful social interactions (Blair, 2005; Decety et al., 2016). Indeed, studies on empathy in parents have found correlations between empathy and positive child outcomes (Meng et al., 2020; Stern et al., 2015).

Infants communicate through vocalizations and facial expressions, which convey important information that elicits attention and affection from caregivers. In return, appropriate, sensitive, and contingent responses by mothers are crucial for healthy infant socioemotional development, facilitating attachment and strengthening the mother-infant relationship (Lewis and Feiring, 1989; Pederson et al., 1990; Stams et al., 2002). Measuring neural and psychological responses to infant affective stimuli (i.e., infant cries and faces) may thus provide insight into maternal behavior. Indeed, some studies have identified correlations between psychological (Leerkes, 2010) and neural (Kim et al., 2020) responsivity to infant stimuli and maternal behavior such as maternal sensitivity during mother-infant interactions.

Studies investigating subjective responses to infant stimuli often utilize both or either of the following questions for mothers to rate: “how sad/happy do you think the baby feels” (referred to in this manuscript as ‘think’-rating), and/or “how sad/happy do you feel in response to this image/sound” (referred to in this manuscript as ‘feel’-rating) (Strathearn et al., 2008). For instance, mothers had significantly higher ‘think’-ratings than non-mothers when viewing videos of infants laughing (Bjertrup et al., 2021), while mothers exposed childhood maltreatment had lower ‘think’- and ‘feel’-ratings than mothers who were not exposed (Olsavsky et al., 2019). In Rutherford et al. (2020), which included an earlier analysis of data in this manuscript, the authors reported the emotional intensity mothers thought that the infants felt (i.e., ‘think’-ratings), and did not find an effect of any (past and current/perinatal) maternal SU on ratings. In this paper, for completeness, we present subjective report data regarding both ‘think’- and ‘feel’-ratings. To our knowledge, this distinction between ‘think’- and ‘feel’-ratings has not been reported in the context of maternal SU comparing mothers who use substances versus mothers who do not.

These ‘think’- and ‘feel’-ratings may arguably be conceptualized within an empathy framework. Empathy can be subdivided into cognitive and affective domains (Davis, 2018; Deutsch and Madle, 1975; Gladstein, 1983; Wispé, 1986), and distinct brain regions have been associated with each (Decety, 2011; Shamay-Tsoory, 2011). Cognitive empathy is a top-down system that supports inferences of the mental states of others, similar to ‘think’-ratings in which mothers need to infer emotional intensities of infants’ experiences. The medial prefrontal cortex (mPFC), superior temporal sulcus (STS), and temporo-parietal junction (TPJ) have often been associated with cognitive empathy and perspective-taking (Cerniglia et al., 2019; Corradi-Dell’Acqua et al., 2020; Decety, 2011; Eres et al., 2015; Frith and Frith, 2006; Preckel et al., 2018; Schurz et al., 2014), though these brain regions are sometimes also associated with affective empathy (de Waal and Preston, 2017; Knight et al., 2019; Miller et al., 2020). Conversely, affective empathy is the ability/tendency to share in the emotional experiences of others, similar to ‘feel’-ratings, in which mothers rate the emotional intensity of how they feel when perceiving infant stimuli. This process is more commonly associated with the amygdala, inferior frontal gyrus (IFG), and insula (Corradi-Dell’Acqua et al., 2020; Eres et al., 2015; Leigh et al., 2013; Lockwood, 2016; Molenberghs et al., 2016; Novak et al., 2022; Shamay-Tsoory et al., 2009), though these brain regions are sometimes also associated with cognitive empathy (Cerniglia et al., 2019; Hooker et al., 2010; Shamay-Tsoory et al., 2009). Indeed, studies have identified differences in neural responses to infant stimuli between mothers who do versus do not use substances (Kim et al., 2017; Landi et al., 2011; Lowell et al., 2020), including some of these brain regions associated with cognitive and affective empathy.

Studies of people with SU disorders versus those without have found seemingly conflicting results regarding whether specifically affective empathy (Ferrari et al., 2014; Martinotti et al., 2009; Maurage et al., 2011) or both cognitive and affective empathy (Le Berre, 2019; Massey et al., 2018; Nachane et al., 2021) differ between groups. However, few studies have been conducted in the context of maternal SU and maternal empathy, i.e., the ability/tendency to recognize an infant’s feelings (cognitive empathy) and feel an infant’s emotions (affective empathy), which may be particularly relevant to understanding how SU potentially influences mother-infant relationships (Abraham et al., 2018; Ojha, 2021).

Here we studied maternal SU and responsivity to infant stimuli by investigating ways in which mothers using substances may differ from mothers not using substances, considering empathy as a possible framework for understanding potential differences. We utilized subjective ratings of infant stimuli to probe psychological differences between the two groups of mothers and performed a region-of-interest (ROI) analysis focusing on brain regions associated with cognitive and affective empathy to identify potential neural differences. The specific questions we sought to address in this study were: 1) are there differences in subjective responses to infant stimuli as a function of perinatal maternal SU?; 2) are these differences in subjective responses specific to SU or are they due to group differences in clinical affective (e.g., depression) or demographic measures?; and 3) are any SU group differences in subjective responses reflected in differences in neural activation in specific brain regions? Based on studies of individuals with SU disorders, we hypothesized that mothers who used substances during the perinatal period would have lower subjective responses related to both ‘think’- and ‘feel’-ratings of infant stimuli and that this would be independent of clinical and demographic measures. We also hypothesized that there would be corresponding altered neural functioning in brain regions previously implicated in cognitive and affective empathy and that neural activity in these brain regions in response to infant stimuli would be able to predict maternal SU status.

Material & Methods

Participants

This is a re-analysis of a previously published dataset (Rutherford et al., 2020). Yale School of Medicine’s Human Investigation Committee approved all procedures prior to recruitment. Mothers of infants were recruited from the local community as part of a larger study on parenting and addiction (Rutherford et al., 2020) through flyers placed in local community areas and women’s health clinics. All mothers were compensated $80 for completing the magnetic resonance imaging (MRI) visit. Mothers were included in the pregnancy and/or postpartum SU group if they met any of the following criteria: any SU (including alcohol) on 3 or more occasions during pregnancy, near-daily or daily use of substances during postpartum, and/or heavy alcohol use (3 or more drinks) on at least three occasions a month during the postpartum period. Mothers who did not use substances or used substances but did not meet these criteria were classified as non-substance-using (non-SU) (Supplementary Table 1). This emphasis on perinatal SU differs from the categorization of SU status in the original analysis (Rutherford et al., 2020), where mothers with a history of SU were also included in the SU category. Participants also completed self-report measures of depression (Beck Depression Inventory (Beck et al., 1996)) and anxiety (State Trait Anxiety Inventory (Spielberger, 1983)).

Infant stimuli

Infant face and cry stimuli were presented pseudorandomly using E-Prime 1.2. Infant face stimuli consisted of 20 unique happy and 20 unique sad faces; for each type of face, half of the images were of the mother’s own infant, the other half were unknown infants. Example stimuli are presented in Fig. 1a. Own and unknown infant face stimuli were matched on race and affect intensity, with the latter coded by research staff trained to reliability (Kim et al., 2017). Own and unknown cry stimuli were matched according to cry intensity.

Fig. 1. Subjective ratings of infant stimuli.

a) Examples of infant stimuli presented to mothers. b) ‘Feel’-ratings of all infant stimuli except own infant happy faces were significantly lower for SU mothers as compared to non-SU mothers. c) ‘Think’-ratings of infant stimuli were not significantly different between SU and non-SU mothers. O: own infant (solid bars), U: unknown infant (striped bars). n = 41 non-SU, 36 SU mothers. *: p < 0.05, **: p < 0.01.

Infant faces were centrally presented in color on a grey background and infant cries were presented through headphones with a blank visual display. Each stimulus was presented for 2s with an inter-stimulus interval jitter of 2–11s, consistent with other studies (e.g., Strathearn et al., 2008). For each of the 6 stimuli (own/unknown infant, cry, sad face, happy face), 6 trials were presented within each functional MRI (fMRI) run, and there were 7 runs of data collected, leading to a total of 42 trials per stimulus type. Each run lasted approximately 5 minutes, for a total task duration of approximately 40 minutes. Participants were instructed to attend to the infant stimuli during scanning.

Subjective ratings of infant stimuli

After scanning, the stimuli were presented again and participants were asked to rate each stimulus based on ‘how happy/sad do you think the baby was feeling’ (‘think’-rating) and ‘how happy/sad did this picture/sound make you feel’ (‘feel’-rating) on a Likert scale of 1–9 (least-most). They were also asked to identify whether the stimulus presented was of their own infant or an unknown infant. Four participants’ rating data (1 non-SU mother, 3 SU mothers) were excluded because they failed to identify their infant’s face on >90% of the trials and/or performed below chance levels at identifying their infant’s cries. There were no significant differences between non-SU and SU mothers at identifying their own infant faces or cries before or after exclusion of these participants. The final sample comprised 41 non-SU and 36 SU mothers. Due to missing data on clinical variables, analyses that examined potential effects of mood (depression, anxiety) variables on associations between maternal SU and ratings had a sample size of 33 non-SU and 26 SU mothers before missing data were imputed.

MRI acquisition

Data were acquired with a Siemens Trio 3T MRI system employing a standard 12-channel head coil. Functional data were collected with a gradient echo, echoplanar sequence: repetition time (TR) = 2000ms, echo time (TE) = 30ms, flip angle = 80°, field of view 220×220mm, matrix = 64×64, slice thickness = 4mm, and 32 slices.

Univariate fMRI analyses

As reported previously in Rutherford et al. (2020), data pre-processing and subject-level and group-level statistics were completed in Statistical Parametric Mapping (SPM12) (http://fil.ion.ucl.ac.uk/spm/). Functional images were aligned prior to normalization to Montreal Neurological Institute (MNI) space. Functional runs where participant motion was in excess of 3 mm or degrees were excluded. To be included in the analysis, participants needed at least 4 functional runs passing motion exclusion criteria. Data were then smoothed using a 6mm full-width-half-maximum Gaussian kernel. Nine non-SU and 9 SU mothers were excluded for excessive motion, difficulty pre-processing, or missing data; the final sample with imaging data consisted of 33 non-SU and 30 SU mothers. Of this sample, clinical affective data (depression and trait anxiety) from 28 non-SU and 21 SU mothers was available.

For first-level statistical analyses, we constructed mass-univariate general linear models for each participant. Face and cry stimuli onsets were convolved with the hemodynamic response function and modelled with temporal derivatives. High-pass filtering (128s) was applied to all models, and motion parameters from realignment were included as regressors of no interest in the final analyses.

We performed an ROI analysis focusing on brain regions associated with cognitive and affective empathy to identify potential neural differences between the two groups of mothers, in contrast with the whole-brain analysis in Rutherford et al. (2020). Six brain regions were chosen a priori for analyses given their involvement in empathy. Binary masks for the STS ROI were obtained from Ekert et al. (2021) and TPJ and mPFC were obtained from Wang et al. (2020) (https://identifiers.org/neurovault.collection:6262). Amygdala and insula masks were obtained from the Automated Anatomical Labelling (AAL) atlas. The IFG mask was derived from the Talairach Daemon database using WFU_PickAtlas (RRID:SCR_007378), though the IFG mask from AAL yielded similar results. Mass univariate analysis in SPM were performed for second-level statistical analyses. One-sample t-tests were performed to generate activation maps. Six contrasts were created, one for each stimulus type, to compare activation in response to each stimulus relative to baseline. The ROI signal extractor Marsbar toolbox (http://marsbar.sourceforge.net) was used to extract mean parameter estimates for each ROI of every participant. Beta-weights were extracted from each ROI for each contrast irrespective of their significance in the SPM contrast.

Multi-voxel pattern analysis (MVPA)

In addition to the traditional univariate fMRI analysis mentioned above, we also utilized MVPA, which uses brain activity to detect differences between groups of people with increased sensitivity (Norman et al., 2006). It can be used to quantify how accurately patterns of activity in each ROI predict participant characteristics (e.g., their SU status) (Norman et al., 2006). These analyses were performed using PRoNTo (Schrouff et al., 2013). For classification analyses to predict maternal SU status from brain activity, beta images generated by SPM (one for each stimulus, run, and participant) were used as inputs. We included betas from participants who had at least four runs; the final sample included 33 non-SU and 30 SU mothers.

Statistical analyses

Statistical analyses were performed in SPSS v28 and graphs were plotted using GraphPad Prism 9. Chi-square, t-tests, or Mann-Whitney tests were used to determine differences in demographic characteristics between the two groups of mothers. A repeated-measures four-way ANOVA (non-SU/SU, own/unknown infant, cry/sad face/happy face, think/feel rating) was performed to quantify differences in subjective ratings. This analysis was run with and without clinical measures that were found to be different between the two groups of mothers (i.e., depression, trait anxiety) as covariates. Since we had missing clinical data, we utilized the fully conditional specification (FCS) method with 20 imputations for imputing missing clinical data (Graham et al., 2007) to perform exploratory moderation analyses were performed to identify potential moderating effects of clinical measures on the relationship between SU and infant stimuli ratings. Pearson’s or Spearman’s correlations were calculated between clinical and demographic measures and subjective ratings, depending on data normality. Next, a repeated-measures four-way ANOVA (non-SU/SU, ROI, own/unknown infant, cry/sad face/happy face) was conducted to identify differences in neural responses to infant stimuli between the two groups of mothers. Finally, for the MVPA analysis, statistical significance was determined using leave-one-subject-out cross-validation and permutation tests (1000 permutations), and Bonferroni correction was applied to control for multiple comparisons (i.e., a p-value of 0.05 / 6 ROIs = 0.0083). For all figures and tables, *: p < 0.05, **: p < 0.01. All error bars indicate 95% confidence intervals.

Results

Participant characteristics

Sample demographics and clinical features are presented in Table 1. On average, participants in this study were mothers who were about 30 years old with, on average, 8-month-old infants (range: 4.8–14.7 months). Mothers with SU and without (non-SU) did not differ on demographic measures, including age, education level, or parity (non-SU: 37.5% primiparous, SU: 41.7% primiparous, X²(1) = 0.1, p = 0.8). The groups were also comparable on infant demographics, including infant age, sex, and proportion of infants born preterm (SU: 15.8%, non-SU: 9.4%, X²(1) = 0.5, p = 0.5). SU mothers had higher trait-anxiety scores than non-SU mothers, though the mean scores for both groups were within the ‘no to low anxiety’ range. SU mothers had higher scores on all SU measures except caffeine use. Among SU mothers, tobacco was the most used substance (66.7% and 74.4% of mothers using during pregnancy and postpartum, respectively), followed by alcohol (30.8% and 25.6% of mothers using during pregnancy and postpartum, respectively), cannabis, and cocaine (Supplementary Table S1).

Table 1. Participant characteristics.

non-SU: non-substance-using, SU: substance-using mothers. SD: standard deviation. The “days born early” variable was defined relative to due date. BDI: Beck Depression Inventory. STAI: State-Trait Anxiety Inventory. FTND: Fagerstrom Test for Nicotine Dependence. n = 31–36 non-SU, 24–30 SU mothers.

	non-SU	SU	p
Demographics (mother)
Age in years, mean (SD)	28.5 (5.1)	29.5 (5.3)	0.47
Income (<$30k/year), n (%)	21 (61.8)	22 (73.3)	0.42
Education in years, mean (SD)	13.9 (2.8)	13.1 (2.0)	0.21
Race/Ethnicity (non-Caucasian), n (%)	19 (54.2)	16 (55.2)	0.64
African American	8	10
Caucasian	16	13
Hispanic/Latino	6	4
Other / no answer	5	2
Number of children, mean (SD)	2.1 (1.2)	2.1 (1.6)	0.99
Demographics (infant)
Age in months, mean (SD)	8.0 (2.0)	7.7 (1.4)	0.53
Gender, female, n (%)	18 (51.4)	15 (51.7)	0.98
Days born early, mean (SD)	6.0 (14.9)	5.7 (12.9)	0.66
Clinical measures
BDI, mean (SD):	6.8 (6.7)	10.6 (9.7)	0.06
STAI-state, mean (SD):	29.3 (8.2)	32.7 (9.7)	0.13
STAI-trait, mean (SD):	31.1 (8.4)	38.6 (10.2)	0.002
SU measures
Total use past 30 days, mean days (SD)	1.6 (5.1)	8.0 (11.6)	<0.001
History tobacco, n (%):	12 (28.6)	33 (82.5)	<0.001
FTND, mean (SD):	0.10 (0.40)	2.40 (2.21)	<0.001
Caffeine use, mean (SD):	2.3 (0.9)	2.5 (0.9)	0.69

Subjective ratings of infant stimuli

Mothers were asked to provide ‘think’- and ‘feel’-ratings in response to three types of infant stimuli (cries, sad faces, happy faces) of their own and unknown infants (Fig. 1a). Consistent with previous findings (Kim et al., 2017; Rutherford et al., 2020), ratings of one’s own infant were higher than of unknown infants (main effect of infant identity: F_1,75 = 77.7, p < 0.001). In general, ‘think’-ratings were higher than ‘feel’-ratings (main effect of rating question: F_1,75 = 58.0, p < 0.001), indicating that both groups of mothers felt less intensely than the intensity they thought the infant was feeling. There was also a significant four-way interaction: F_2,150 = 3.11, p = 0.047. Post-hoc tests revealed that ‘feel’-ratings were lower for SU mothers than non-SU mothers; this was true for all infant stimuli except own infant happy faces (Fig. 1b). However, both groups of mothers had comparable ‘think’-ratings for all six infant stimuli (Fig. 1c).

Since there were some differences in clinical measures between the two groups of mothers, we included them as covariates. The four-way interaction was no longer significant when conducting an ANCOVA with depression and trait-anxiety scores as covariates (F_2,110 = 2.57, p = 0.082). However, this may be due to the lower sample size (n = 59 versus n = 77), since clinical data were only available for a subset of participants. To further consider mood differences between the two groups of mothers, we performed exploratory moderation analyses on the imputed data to determine whether psychiatric factors contributed to differences in ‘feel’- ratings. For infant stimuli that had significant differences in ‘feel’-rating between non-SU and SU mothers, depression and anxiety scores were not significant moderators of associations between maternal SU status and ‘feel’-ratings (Supplementary Table S2), suggesting that maternal SU status statistically predicted ‘feel’-ratings regardless and potentially independently of maternal affective variables. However, there was a significant interaction between trait-anxiety scores and SU status with respect to ‘feel’-ratings to own infant happy faces that suggested that trait anxiety may negatively influence maternal emotional responses to own infant happy faces in mothers without SU as compared to those with SU (Supplementary Table S2).

Additionally, there were no significant correlations between any clinical measures (i.e., depression, trait anxiety) and any subjective ratings at an exploratory uncorrected p-value of 0.05, with the exception of trait anxiety and ‘think’-ratings (Spearman’s rho = −0.28, p = 0.03, n = 59) and ‘feel’-ratings (Spearman’s rho = −0.28, p = 0.03, n = 59) related to happy faces of mothers’ own infants. However, neither of these ratings differed between non-SU and SU mothers. As additional exploratory analyses, we evaluated associations between demographic variables and subjective ratings. There was no association between parity or infant sex and subjective ratings, as determined by repeated-measures ANOVAs. There were also no correlations between infant age and any of the subjective ratings, though there was a positive correlation between mother age and ‘think’-ratings of own (rho = 0.31, p = 0.02, n = 56) and unknown infant cries (rho = 0.28, p = 0.04, n = 56). However, since the two groups of mothers were not significantly different in age, age was not included as a covariate.

Infant stimulus-evoked brain activations

We next considered differences in neural responses between the two groups of mothers. We quantified BOLD activity in six ROIs (Fig. 2a) in response to each of the six types of infant stimuli and performed a repeated-measures four-way ANOVA (non-SU/SU, ROI, own/unknown infant, cry/sad face/happy face). In general, there was stronger brain activation in response to infant cries than faces (main effect of infant stimulus: F_2,122 = 6.9, p = 0.001), and to own infant than unknown infant stimuli (main effect of infant identity: F_1,61 = 16.0, p < 0.001). There was also a significant interaction between maternal SU status and ROI: F_5,305 = 2.27, p = 0.047. This interaction remained significant after controlling for depression and trait-anxiety scores (F_5,225 = 2.39, p = 0.039). Specifically, SU mothers demonstrated less activation to infant stimuli than non-SU mothers in brain regions previously implicated in affective empathy, such as the IFG and insula; amygdala responses were numerically lower but statistically similar in SU versus non-SU mothers (p = 0.055) (Fig. 2b). There were no differences in brain regions implicated in cognitive empathy (Fig. 2b).

Fig. 2. Infant cue-evoked brain activation.

a) Brain regions selected a priori for ROI analyses. Yellow: amygdala, orange: inferior frontal gyrus (IFG), red: insula; green: medial prefrontal cortex (mPFC), blue: superior temporal sulcus (STS), purple: temporo-parietal junction (TPJ). b) Average blood-oxygen-level-dependent (BOLD) activation in response to all infant stimuli. SU mothers exhibited significantly lower IFG and insula activation as compared to non-SU mothers. n = 33 non-SU (light bars), 30 SU (dark bars) mothers. *: p < 0.05, **: p < 0.01.

Independently, we used MVPA to quantify patterns of neural activity in each brain region in response to each infant stimulus to predict maternal SU status. Insular responses to unknown infant cries and own infant sad faces, and IFG responses to own infant sad faces, statistically predicted maternal SU status (Supplementary Table S3). None of the brain regions implicated in cognitive empathy predicted maternal SU status. These data suggest that brain activity in regions linked to affective empathy differs between SU and non-SU mothers.

Discussion

The current study investigated neural and subjective responses to infant stimuli and differences between mothers who used substances during the prenatal and/or postpartum period and mothers who did not. We identified differences in ratings of emotional responses (‘feel’- ratings) and corresponding differences in neural responses to infant stimuli in brain regions implicated in affective empathy. Notably, we did not identify any differences in ratings of infant emotional intensity (‘think’-ratings), nor neural differences in brain regions implicated in cognitive empathy between the two groups of mothers. This work adds to the limited literature on the neuroscience of maternal SU and its links with maternal responses to infants, which provides insight into potential mechanisms for how maternal SU relates to maternal brain function and behavior.

Maternal SU and emotional responses to infant stimuli

Our behavioral results on emotional responses to infant stimuli (i.e., lower ‘feel’-ratings in mothers who use substances) are in line with several studies that have identified lower affective empathy in people who use substances (Carlyle et al., 2020; Martinotti et al., 2009; Massey et al., 2018; Maurage et al., 2011; Nachane et al., 2021; Preller et al., 2014). Notably, in non-SU mothers, those with greater empathic emotion in response to videos of their infant’s distress evidence more sensitivity to distress during mother-infant interactions (Leerkes, 2010), suggesting affective empathy towards infant stimuli may be an important component of sensitive maternal behavior. In the current study, the observed lower levels of ‘feel’-ratings to infant stimuli (particularly non-happy ones) and links to specific regional brain activations in mothers using substances suggest less sensitive maternal responsivity and may in part explain increased rates of child maltreatment associated with parental SU (Canfield et al., 2017; Wong et al., 2021). That said, it is interesting to note that affective ratings of own infant happy faces were not different between the two groups of mothers with and without SU, suggesting a potential protective effect of positive emotional expressions in the context of maternal SU. However, it is also important to note that trait anxiety may negatively impact affective responses to positive infant stimuli, especially in mothers without SU. Future work should determine associations between neural responses to varying infant emotions and associations with maternal behavior in mothers with and without SU.

Our neural results replicate and extend prior findings in an independent sample (Landi et al., 2011) identifying lower amygdala, IFG, and insula responses to unknown infant stimuli in mothers who use substances. Since there was no significant interaction between maternal SU status and infant identity, our results suggest that lower IFG and insula responses likely also occur in response to own infant stimuli. These results are in line with other studies which have identified positive associations between amygdala, IFG, and insula responses to own infant’s cries and maternal sensitivity (Musser et al., 2012; Wan et al., 2014), and suggest that lower activation in these brain regions in response to infant stimuli in SU mothers may be associated with lower maternal sensitivity.

Maternal SU and recognition of infant emotional intensity

We did not observe any differences between the two groups of mothers in ratings of infant emotional intensity (‘think’-ratings) or cognitive-empathy-related brain regions. This was surprising since a previous study found lower mPFC responses to unknown infant happy faces in mothers who used substances compared to mothers who did not (Landi et al., 2011). The difference in results might reflect differences in experimental design: in the prior study (Landi et al., 2011), participants were only exposed to unknown (not own) infant stimuli and also viewed neutral infant faces during the scan. Speculatively, these differences could have led to different mental states (e.g., less emotional engagement due to unfamiliar stimuli) and brain activity during the task, thus leading to differences in results.

This lack of difference may also be because the group of mothers using substances in this study had relatively mild SU. We infer this from multiple demographic measures (Table 1). First, although there is no clinical cut-off for the number of days of SU as measured on the ASI-Lite, eight out of 30 days (<30% of the days) on average may not constitute a significant disruption to daily functioning. Second, our nicotine-dependence scores were in the ‘low/moderate’ range, whereas a study that identified alterations in PFC grey-matter volume has nicotine-dependence scores twice as high (Zhong et al., 2016). Third, SU during pregnancy is associated with premature birth (Louw, 2018; Richardson et al., 1999; Srikartika and O’Leary, 2015), which was not observed in our SU population. Our results suggest that mild SU may be associated with impairments in affective processing of infant stimuli, while cognitive processing is altered in more severe instances of SU.

Associations between other variables and subjective responses to infant stimuli

Some studies have found differences in emotional responsivity in mothers dependent on variables other than substance use such as depression and anxiety, most commonly a blunting of neural and behavioral responses to infant stimuli (Pechtel et al., 2013). We did not find any associations between depression or trait anxiety scores and subjective ratings that were different between the two groups of mothers in this study. We believe this may be because the depression and anxiety scores of mothers in our sample were relatively low. Future studies could explore this with a wider range of depression and anxiety scores. With regard to demographic variables, we did not find an association between parity and subjective ratings. There are indeed studies that suggest differences in emotional responsivity between primiparous and multiparous women. For example, primiparous mothers exhibited greater neural responses to infant facial expressions than multiparous mothers, perhaps indicative of increased saliency of infant stimuli to first-time mothers (Bunderson et al., 2020; Maupin et al., 2019). However, these were event-related potential (ERP) studies and neither included behavioral ratings, so perhaps these differences in neural responses to infant stimuli are associated with psychological processes other than empathy or related to the different signal sources.

Limitations and Future Directions

The cross-sectional nature of the experimental design prevents causal inferences regarding potential impact of maternal SU on neural responses to, and subjective ratings of, infant stimuli. While associations exist between maternal SU and neural and psychological responses to infant stimuli, maternal SU may not causally underlie the findings. Other factors that co-occur with or precede SU may contribute. For example, early life trauma is a common trigger for SU in women and can influence maternal brain and behavior (Kim et al., 2014; Levy et al., 2019; Strathearn et al., 2019), and maternal exposure to childhood maltreatment has been associated with lower ‘think’- and ‘feel’-ratings (Olsavsky et al., 2019). Longitudinal studies will allow better disambiguation of the role of maternal SU in shaping maternal behavior alongside other potential risk factors.

Another limitation is that this study did not incorporate validated measures of maternal empathy, such as parental empathy questionnaires (Gonzalez and Rodriguez, 2021; Kilpatrick, 2005; Stern et al., 2015) or measures of empathic maternal behavior (e.g., mother-infant interactions during free play), which would be necessary for drawing conclusions regarding maternal SU and maternal empathy. Future studies are necessary to examine directly how such measures may correlate with the infant-stimulus subjective ratings described in this manuscript.

Another limitation is that analyses did not distinguish between types of substances consumed by mothers, given the small sample size. Since there are common and separable neural alterations in SU disorders (Klugah-Brown et al., 2020), future studies should identify which specific substances used during the perinatal period are associated with lower maternal subjective responsivity and lower neural activation in related brain regions. Interestingly, studies of individuals who use specific substances (e.g., alcohol, stimulants) have generally identified impairments (as opposed to enhancements) in affective empathy (Massey et al., 2018), suggesting that a common neural mechanism may underlie such differences.

Conclusions

By identifying similarities and differences in neural and psychological correlates of subjective responses to infant stimuli in mothers using substances, this work provides novel information for understanding maternal behavior in mothers who use substances by directly connecting brain regions implicated previously in cognitive and affective processing and subjective ratings of infant stimuli, through the lens of maternal empathy. While studies have begun to investigate neural and psychological correlates of parental empathy using infant-specific measures (Abraham et al., 2018), to our knowledge, this is the first work in the context of maternal SU. Our findings suggest neural targets in mothers using substances during the perinatal period that may promote affective empathy towards their infants. Some potential interventions to increase affective empathy include loving-kindness meditation (Leppma and Young, 2016) and administration of intranasal oxytocin (Geng et al., 2018). Future longitudinal work should track the neural and behavioral effects of such interventions with regard to improving maternal emotional responsivity towards infants and subsequent offspring outcomes.