Maternal Perinatal Anxiety and Neural Responding to Infant Affective Signals: Insights, Challenges, and a Road Map for Neuroimaging Research

Abstract

Anxiety symptoms are common among women during pregnancy and the postpartum period, potentially having detrimental effects on both mother and child’s well-being. Perinatal maternal anxiety interferes with a core facet of adaptive caregiving: mothers’ sensitive responsiveness to infant affective communicative ‘cues.’ This review summarizes the current research on the neural correlates of maternal processing of infant cues in the presence of perinatal anxiety, outlines its limitations, and offers next steps to advance future research. Functional neuroimaging studies examining the neural circuitry involved in, and electrophysiological studies examining the temporal dynamics of, processing infant cues during pregnancy and postpartum are reviewed. Studies have generally indicated mixed findings, although emerging themes suggest that anxiety may be implicated in several stages of processing infant cues— detection, interpretation, and reaction— contingent upon cue valence. Limitations include inconsistent designs, lack of differentiation between anxiety and depression symptoms, and limited consideration of parenting-specific (versus domain-general) anxiety. Future studies should incorporate longitudinal investigation of multiple levels of analysis spanning neural, cognitive, and observed aspects of sensitive caregiving.

1. Introduction

The perinatal period has been characterized as a period of psychological vulnerability (e.g., Saxbe et al., 2018). Anxiety is one of the most prevalent mental health conditions pregnant and postpartum women experience (Uguz et al., 2019), manifesting in excessive, difficult-to-control worries and disturbing thoughts (Goodman et al., 2016; Howard and Khalifeh, 2020; Pawluski et al., 2017). Maternal anxiety has been associated with adverse outcomes for mothers, the mother-child relationship, and children’s development and well-being (e.g., Korja et al., 2017; Lawrence et al., 2019; Martini et al., 2015; Micco et al., 2009; Skouteris et al., 2009). Specifically, high levels of anxiety have been established as a risk factor for compromised maternal sensitive responsiveness and attunement towards infant cues (e.g., Hakanen et al., 2019; Stevenson-Hinde et al., 2013), which are signals such as facial expressions and cries that inform on the infant’s needs. Sensitive responsiveness is considered a basic building block of optimal mother-child relationships (e.g., Ainsworth et al., 1974; Feeney and Woodhouse, 2016; Posada et al., 2016), highlighting the importance of understanding how anxiety impacts maternal processing of infant cues. Neuroscientific examination of maternal responses to infant cues and their modulation by anxiety can provide mechanistic insights into the impact of anxiety on sensitive caregiving, and thereby on children’s development. The aim of this review is to (1) critically summarize the current knowledge on anxiety and neural correlates of maternal processing of infant cues in pregnancy and postpartum; (2) detail the limitations and challenges in this growing body of research; (3) identify gaps in the field’s understanding of the mechanisms by which perinatal anxiety may modulate parenting and the emerging mother-child relationship; and (4) offer a road map for future research on maternal anxiety and processing infant cues.

1.1. Maternal Anxiety in Pregnancy and Postpartum, Parenting, and Child Development

Heightened anxiety is a common experience among women during the perinatal period. While estimates vary widely, 15-30% of pregnant women and 4-26% of postpartum women are thought to have clinically significant levels of anxiety (e.g., Dennis et al., 2017; Fairbrother et al., 2016; Heron et al., 2004; Matthey et al., 2003; Reck et al., 2009; Uguz et al., 2019; Wenzel et al., 2005). According to a recent meta-analysis by Dennis and colleagues (2017), 22.9% of pregnant women report having elevated symptoms of anxiety and 15.2% meet diagnostic criteria for anxiety disorders based on clinical interviews; after the first month postpartum, anxiety estimates appear to stabilize, with 15% of women reporting high anxiety symptoms and 9.3-9.9% of postpartum women meeting diagnostic criteria for anxiety disorders based on clinical interviews during the first year postpartum. That is, maternal anxiety may be at least as common as depression during the perinatal period, thought to affect 15-20% of women perinatally (e.g., Ashley et al., 2016; Reck et al., 2009; Shorey et al., 2018), and more prevalent than rates of anxiety in the general population (Dennis et al., 2017; Remes et al., 2016; Steel et al., 2014; Uguz et al., 2019). Taken together, the perinatal period may confer increased risk for anxiety (Dennis et al., 2017; Uguz et al., 2019). Yet, while maternal (postpartum) depression has been widely discussed in society and the research literature, maternal anxiety has historically received less attention (Matthey et al., 2003).

Although some level of anxiety may be adaptive during the perinatal period for creating a nurturing environment (Leckman et al., 2004), perinatal anxiety has been linked with a multitude of adverse outcomes for children, ranging from increased risk for congenital malformations and obstetrical complications (Anniverno et al., 2013; Grigoriadis et al., 2018; Hansen et al., 2000), to cognitive and socioemotional risk persisting into adolescence (Rogers et al., 2020). Specifically, maternal perinatal anxiety has been associated with lower child self-regulation (Korja et al., 2017), changes in children’s brain structure and function across primarily frontal, temporal, and limbic brain regions (Adamson et al., 2018; Van den Bergh et al., 2020), and increased levels of emotional and behavioral problems, including anxiety (Davis and Sandman, 2012; Lawrence et al., 2019; Micco et al., 2009). Yet, despite its high prevalence and known associations with adverse outcomes, maternal anxiety has remained underdiagnosed and undertreated (Goodman and Tyer-Viola, 2010; Matthey et al., 2003).

The potentially detrimental effects of maternal anxiety on the well-being of two generations underscore the importance of understanding the mechanisms by which anxiety may impinge early caregiving. Although maternal anxiety may enact its long-term effects on child development through multifactorial mechanisms, with contributions from genetics, epigenetics, hormonal alterations, the uterine environment, and the postnatal environment (Glover et al., 2018; Van den Bergh et al., 2005; 2020), one key area of interest has been maternal anxiety’s impact on parenting and the mother-infant relationship. Maternal anxiety during pregnancy and/or infancy has been linked with increased rates of insecure attachment (Scher and Mayseless, 2000; Stevenson-Hinde et al., 2011), decreased maternal sensitivity (Booth et al., 2018; Mertesacker et al., 2004; Nicol-Harper et al., 2007; Stevenson-Hinde et al., 2013), atypical, less contingent synchrony patterns (Beebe et al., 2011; Feldman, 2007; Moore et al., 2016), and increased intrusiveness (Feldman et al., 2009; Hakanen et al., 2019; Kaitz and Maytal, 2005; Kaitz et al., 2010). Together, these findings characterize maternal anxiety as a risk factor for compromised maternal sensitive responsiveness. That is, maternal anxiety may be implicated in abnormal processing of, and responding to, infant cues, thereby leading to the disruption of mother-infant interactions, potentially having negative effects on child outcomes. However, it should also be noted that some of the findings relating maternal anxiety and sensitive responsive caregiving are inconsistent, and may depend on the specific aspects of sensitive responsiveness or mother-infant interaction context (Feldman et al., 2009; Ginsburg et al., 2006; Grant et al., 2010; Nath et al., 2019; Reck et al., 2018). Neurocognitive measures of maternal processing of infant cues can help bridge this gap by breaking down the internal processes comprising maternal sensitive responsiveness, thereby providing more mechanistic insight into the impact of anxiety on postpartum caregiving.

1.2. The Present Review’s Framework: Leveraging Neuroimaging to Understand the Impact of Perinatal Maternal Anxiety on Sensitive Responsiveness to Infant Cues

The aim of the current review is to advance the field’s understanding of the mechanisms by which perinatal anxiety may modulate sensitive responsiveness by summarizing the findings, challenges, and limitations in the current research on the impact of maternal anxiety on the neural correlates of maternal processing of infant cues, and offering future directions to advance the study of maternal anxiety. In early mother-infant relationships, parenting behavior is often elicited by infant cues; the connections between the infant’s cue and the mother’s behavioral response to the cue is bridged by several neurocognitive mechanisms taking place in the mother’s mind (Figure 1, red box)— mental processes (and their neural substrates) connecting between the mother’s senses and her thoughts and actions, including perception, attention, and higher-order decision-making processes (c.f., Bishop, 2007). Specifically, drawing from Ainsworth and colleagues’ (1974) seminal definition of maternal sensitivity, several stages of mental processing can be implicated in producing behavioral sensitive responsiveness to the infant’s cues: When the infant displays a cue, the mother is required to detect the presentation of a cue and interpret the cue’s meaning to evaluate her infant’s current needs. Therefore, two general sequential stages of processing are required to take place: (1) detection of the infant’s cue, which includes early perceptual and cognitive processes such as perception of the cue and initial attention allocation, which may be dependent on cue saliency (Figure 1A); and (2) interpretation of the infant’s needs based on the cue, requiring more advanced cognitive processes involved in decision making, such as sustained attention and elaborative processing, stimulus evaluation, and response selection, as well as high-order processes such as mentalizing (Figure 1B). Detection and interpretative processes relating to infant cues may be modulated by (3) emotional reactivity in response to the cues and by (4) the mother’s ability to regulate her thoughts, behaviors, and feelings (Bridgett et al., 2015; Leerkes, 2010; Figure 1C). In light of the mother’s understanding of these cues, she can respond to her infant (Ainsworth et al., 1974; Figure 1D); the infant then reacts according to whether and how the mother’s response has met their needs by producing another communicative cue, resulting in continuous transactions in the dyad, such that the infant’s and the mother’s continue to elicit reactions in one another (Sameroff, 2009; Figure 1E). Inadequate maternal responsivity to infant cues could result from difficulties in one or more of these stages of processing: (1) detection of infant cues, (2) cognitive interpretation of cues, (3) emotional responding to cues, or (4) self-regulation.

Figure 1.

A conceptual depiction of the stages of processing infant cues, detection (panel A), and interpretation of infant cues (panel B), and the processes modulating them (panel C). These stages lead to parental behavior, which then elicits subsequent infant cues (D) within a continuous mother-infant transactional interaction (E). These stages of processing provide specific mechanisms to assess with neuroimaging methodologies to understand sensitive-responsive caregiving in general, and its modulation by maternal anxiety in specific.

By breaking down the different stages of processing infant cues as depicted in Figure 1, and examining their neural correlates, more mechanistic insights may be gained in understanding how anxiety can impact maternal sensitive responsiveness. Recent research suggests that women undergo structural and functional neural reorganization geared toward facilitating parenting and bonding during the perinatal period (Barba-Müller et al., 2019; Hoekzema et al., 2017; Kim et al., 2010). These changes raise the possibility that concurrent anxiety may interfere with neural processes that prepare women for motherhood during pregnancy and the initial months postpartum (Barba-Müller et al., 2019; Cárdenas et al., 2019), rendering the study of the impact of anxiety on the prenatal and postnatal maternal brain during the transition to parenthood especially important for understanding the emergence of positive mother-infant relationships. More specifically, there are several pathways through which anxiety during pregnancy and the first months postpartum may continue to shape subsequent mother-infant relationships, including (1) continuity in anxiety symptomology from pregnancy to postpartum (Heron et al., 2004; Martini et al., 2015; Skouteris et al., 2009), and/or (2) early modulation of neural responses to infant cues starting already in pregnancy or the immediate postnatal period. Therefore, gaining knowledge on the impact of maternal anxiety on maternal processing of infant cues across the transition to motherhood can further illuminate the emergence of mechanisms implicated in maternal sensitive responsiveness, as well as inform early identification and prevention efforts.

Maternal brain research has generally employed two main neuroimaging methodologies to examine the impact of anxiety on neural responses to infant cues. The first, functional magnetic resonance imaging (fMRI), can help identify how anxiety may impact specific neural circuitry involved in caregiving (Barrett and Fleming, 2011; Paul et al., 2019). fMRI tracks oxygen levels in blood flow, measured via the blood oxygenation level-dependent (BOLD) signal, under the assumption that more active regions receive more blood flow; BOLD signal has a high spatial resolution of several millimeters and is effective in detecting neural systems implicated in processing infant cues across both cortical and subcortical areas, but has poorer temporal resolution given the relatively slow blood flow response (Ogawa et al., 1998). The second neuroimaging approach, electrophysiology, which include electroencephalography (EEG) and the event-related potential (ERP) methods, can inform on the temporal dynamics of processing infant cues, that is, on the neurocognitive stages of processing as they evolve in time, before, during, and after stimulus (e.g., infant cue) presentation (Maupin et al., 2015; Papousek, 2000; Squire and Stein, 2003). ERPs represent summed activity from populations of cortical pyramidal neurons time-locked onto stimulus presentation with millisecond temporal resolution suitable for disentangling stages of cortical processing (Luck, 2014), including processing infant cues. That is, while fMRI is useful for understanding the neural circuitry involved in maternal processing of infant cues and functional relation between brain areas, EEG is useful for understanding how infant cue processing evolves in the time domain given its millisecond temporal resolution. A third relevant neuroimaging approach that has been incorporated in the study of parental processing of infant cues is functional near-infrared spectroscopy (fNIRS), which tracks changes in cerebral blood flow through optical changes in light attenuation (Pinti et al., 2020). This method focuses on functional changes in activation, and can be situated between fMRI and EEG in terms of temporal and spatial resolution, such that fNIRS has higher temporal resolution than fMRI (but lower than EEG), and higher spatial resolution than EEG (but lower than fMRI; Pinti et al., 2020). Typically, across neuroimaging approaches, maternal neural responses to infant cries or images of infant faces are examined by contrasting neural activity in different conditions, most commonly based on infant familiarity (own versus unknown infant), affective valence (e.g., cry intensities, happy versus distressed facial expressions), or infant versus non-infant stimuli (e.g., adult faces; Kuzava et al., 2020; Maupin et al., 2015; Paul et al., 2019; Young et al., 2017).

This review summarizes the current knowledge on the neural correlates of maternal anxiety during processing of affective stimuli to provide a lens through which to understand potential neurocognitive mechanisms in maternal prenatal and postpartum anxiety’s impact on sensitive responsiveness (Figure 1), parenting, and ultimately child development. Past reviews concerning the neural basis of maternal psychopathology have mainly considered depression, or depression and anxiety together (Bjertrup et al., 2019), and have largely focused on fMRI or animal models (Kim et al., 2016; Pawluski et al., 2017; Pechtel et al., 2013); the current review contributes to this literature by focusing on maternal anxiety and incorporating fMRI and EEG/ERP research as complementary approaches to better understand the impact of maternal anxiety on processing infant cues. A specific focus is given to studies examining correlates of processing infant-related stimuli during the prenatal and postpartum (i.e., perinatal) periods, as periods in which caregiving practices are especially important for children’s development (Gee, 2020). The review starts with examination of functional studies, mostly employing fMRI, examining the link between maternal anxiety and neural circuitry implicated in parenting. Next, ERP studies focusing on associations between anxiety and the temporal dynamics of processing infant cues are reviewed. Afterwards, findings from these two methodological lines of inquiry are summarized and integrated, followed by consideration of the limitations in the current body of research. Lastly, open questions and avenues for future research are proposed as a road map for advancing the field’s understanding of the impact of maternal anxiety on processing infant cues and maternal sensitive responsiveness.

2. Review of the Present Literature on Maternal Anxiety and Processing Infant Cues

2.1. Maternal Anxiety and the Neural Circuitry Involved in Processing Infant Cues: Functional Neuroimaging Studies

Several studies have used functional neuroimaging methods to better understand the impact of maternal anxiety on the functioning of neural circuits involved in parenting. An accumulating body of research implicates three general neural circuits in parenting behaviors: reward and motivation (e.g., amygdala and striatum), self-regulation (e.g., medial prefrontal cortex and anterior cingulate cortex), and empathy/mentalizing networks (e.g., temporal-parietal junction and precuneus; for reviews, see Barrett and Fleming, 2011; Feldman, 2015; Gholampour et al., 2020; Kim, 2016; Pereira and Ferreira, 2016; Swain et al., 2014b). Research on maternal anxiety has mainly focused on reward and motivation (e.g., amygdala) and self-regulation neural circuits (e.g., prefrontal cortex; PFC). Most of the studies reviewed in this section have employed fMRI in samples of postpartum women; in one study (Roos et al., 2011), fNIRS was utilized to examine PFC activity during pregnancy. Of note, much of the initial work in this area has been in studies focused on depression that also included measures of anxiety, and few studies exist studying the specific impact of anxiety.

Concerning reward and motivation circuits, mothers generally evidence increased amygdala reactivity to stimuli of their own infant compared to unfamiliar infants (Barrett et al., 2012; Strathearn and Kim, 2013), suggesting that the amygdala is implicated in socioemotional responding to motivationally-relevant infant cues (e.g., Barrett and Fleming, 2011). Only two studies have addressed the link between maternal anxiety and amygdala dysregulation. Barrett and colleagues (2012) examined the relation between emotional distress and amygdala reactivity to infant cues among mothers three months postpartum. Higher levels of self-reported anxiety, parenting stress, and depression were correlated with lower basolateral amygdala reactivity to photographs of mothers’ own (versus an unfamiliar) happy infant. Further, mothers with decreased basolateral amygdala reactivity to their own happy infant also reported having less positive attachment feelings towards their child (Barrett et al., 2012). The basolateral amygdala is implicated in reward pathways and reinforcement learning (Pessoa, 2010), suggesting that for mothers with high emotional distress, positive infant cues may be less salient, rewarding, or motivating compared to mothers with low emotional distress. Wonch and colleagues (2016) examined mothers’ amygdala dysregulation as manifested in functional connectivity (measured via seed-based psychophysiological interaction [PPI] analysis) between this area and the insula, which is also considered part of fear and anxiety circuits (e.g., Shin and Liberzon, 2010). Their sample included mothers diagnosed with postpartum depression and a control group of non-depressed mothers, who participated at two to five months postpartum. Across both groups, higher levels of anxiety and depression were related to decreased functional connectivity between the amygdala and the insula while mothers viewed photographs of their own and unfamiliar smiling infants, presumably reflecting reduced processing of social and emotional stimuli. Together, these studies suggest decreased detection of, and emotional reactions to, positive infant cues in the context of maternal anxiety.

Studies have also examined the relations between maternal anxiety and areas involved in self-regulation, focusing primarily on the PFC. Roos and colleagues (2011) reported an association between prenatal maternal anxiety and PFC reactivity to pictures of fearful adult faces (compared to a baseline rest condition), as measured using fNIRS among expectant mothers followed across the first, second, and third trimesters of their pregnancy. Throughout pregnancy, state anxiety was related to increased right and left PFC activation, while trait anxiety was related to increased dorsolateral activation in the left PFC and decreased activation in the ventral-left, medial, and right PFC. Interestingly, stronger PFC response to fearful faces was also related to decreased selective attention to threat, suggesting altered fear processing patterns during pregnancy. Prefrontal reactivity to adult affective stimuli has also been linked with postpartum anxiety in two fMRI studies (Gingnell et al., 2015; Noriuchi et al., 2008). Gingnell and colleagues (2015) examined emotional processing of adult faces portraying negative affect (anger, fear) in a group of mothers who completed a face-matching and a non-social perceptual matching task within two days from delivery and again at 4-6 months postpartum, in addition to a control group of non-postpartum women. Postpartum women exhibited higher reactivity to affective stimuli in the inferior frontal gyrus (IFG) and the insula, two regions involved in emotional reactivity and regulation, compared to the control group. Among postpartum women, reactivity in the IFG and insula was correlated with anxiety levels two days following birth and with depression levels at 4-6 months postpartum. In addition, Noriuchi and colleagues (2008) reported that mothers of 16-month-old infants who had stronger right orbitofrontal reactivity to video clips of their own versus other infants also reported higher intensity of anxious feelings while watching their infant’s videos. Taken together, despite heterogeneity in regions examined, these studies implicate increased activity of frontal cortical regions related to self-regulation in the presence of maternal anxiety, possibly reflecting selection of strategies to regulate the mother’s own anxiety symptoms (e.g., via cortico-limbic pathways), the mother’s emotional reaction to the cues, or the infant’s distress (Noriuchi et al., 2008; Roos et al., 2011).

The above-mentioned studies have all examined domain-general anxiety, that is, not in specific relation to parenting or parenthood. To the authors’ knowledge, only two fMRI studies have examined functional neural correlates of processing infant cues while considering parenting-specific anxiety, concerns and worries related to caregiving context (e.g., worrying about one’s competence as a parent, concerns about the child’s development and health), rather than general anxiety symptomology. Kim and colleagues (2015) studied the associations between anxious and intrusive parenting-related thoughts and the neural correlates of infant cry processing (own infant and unknown infant) in a sample of mothers one month postpartum. High parenting-related anxious and intrusive thoughts were associated with increased activity in the motor cortex, premotor cortex, fusiform gyrus, and hippocampus while mothers listened to their own infant’s cry. This pattern was thought to reflect increased stress and action-oriented neural responses. Elevated levels of anxious thoughts were also associated with decreased substantia nigra reactivity (an area implicated in reward processing) to own infant cry, which was further related to lower infant socio-emotional competence during the second year of life. That is, decreased maternal reward response to infant cues early postpartum may be linked to later infant development.

Advancing the investigation of parenting-specific anxiety, Rutherford and colleagues (2019) have utilized connectome-based predictive modeling to explore whole-brain patterns of functional connectivity in maternal processing of infant affective cues and anxiety. Mothers were presented with affective infant faces and cries at two and eight months postpartum and reported on their domain-general trait anxiety and on parenting-specific anxiety concerning bonding with their infants. At two months postpartum, elevated parenting-specific anxiety, but not general trait anxiety, was related to higher functional connectivity between the cerebellar and motor-sensory-auditory networks and between the motor-sensory-auditory and frontoparietal networks, with decreased connectivity between the cerebellar and frontoparietal networks. Furthermore, reduced connectivity between two and eight months postpartum predicted a decrease in maternal parenting anxiety. These findings implicate two networks in parenting-specific anxiety: (1) hyperactivity to infant cues between networks associated with action and perception, suggesting increased saliency of infant cues, and (2) hypoactivity between networks associated with self-regulation and control, suggesting difficulties in regulating the response to these infant cues.

Most studies on the role of maternal anxiety in functional correlates of processing infant cues have not incorporated maternal behavior. However, Guo and colleagues (2018) focused on how maternal anxiety may modulate associations between neural responses to infant cues and maternal observed behavior in a sample of mothers of 4 months old infants. Their results revealed that maternal anxiety severity moderated the association between observed caregiving and amygdala-right posterior superior temporal sulcus (RpSTS) functional connectivity when listening to cry sounds. Specifically, a whole-brain PPI analysis indicated that amygdala-RpSTS connectivity was related to more positive mother-infant interactions only among highly anxious mothers. The RpSTS is considered important for the perception of social cues (Van Overwalle and Baetens, 2009), and connectivity between the STS and the amygdala is thought to play a role in inhibiting automatic defense responses (Pessoa et al., 2002). It is possible that enhanced amygdala-RpSTS functional connectivity may play a compensatory role, enabling mothers to maintain a positive parenting style despite elevated anxiety.

In sum, functional neuroimaging studies suggest that maternal anxiety may be related to decreased responsivity in brain regions implicated in saliency and motivation when mothers view positive cues (Barrett et al., 2012; Kim et al., 2015; Wonch et al., 2016) and increased activation in brain regions related to self-regulation in response to negative stimuli (Gingnell et al., 2015; Noriuchi et al., 2008; Roos et al., 2011). However, it is important to note that patterns of results are mixed, with one study examining postpartum parenting-specific anxiety indicating a contrasting pattern showing increased activity in saliency networks and decreased activity in regulation networks when several types of cues (i.e., infant faces and vocalizations) and valence values are included (Rutherford et al., 2019). This mixed pattern raises questions concerning the specificity of findings to maternal anxiety as it relates to parenting, and to perception of infant, rather than adult, affective cues.

2.2. Maternal Anxiety and Temporal Dynamics of Processing Infant Cues: Electrophysiological Studies

A handful of studies have used electrophysiological methods to investigate how maternal perinatal anxiety may affect the neural processing of infant cues. Although functional neuroimaging studies provide vital information on neural circuitry involved in infant cue processing, they do not provide high resolution with respect to the timing and stages of these processes. The speed with which neurocognitive processes unfold is essential for contingent interactions, and indeed the timely identification and response to infant cues is inherent to optimal early caregiving in general and the definition of sensitive responsiveness in specific (Ainsworth et al., 1974; Feldman, 2012). Given the rapid, often intuitive, nature of caregiving and mother-infant interactions (Papousek, 2000), the sensitivity of ERPs to the temporal dynamics of processing infant cues (i.e., the timing of neural responsivity to infant cues) renders this technique especially effective in studying caregiving (Bornstein et al., 2013; Maupin et al., 2015; Rutherford and Mayes, 2011; Young et al., 2017).

Studies on ERP correlates of parenting indicate preferential processing of more emotionally-laden stimuli, and especially of infant distress expressions (Kuzava et al., 2020; Maupin et al., 2015; Young et al., 2017). Examining the responsivity of specific ERP components can provide insight into the stages of processing infant cues (Figure 1). Earlier ERP components are generally thought to reflect initial detection and perceptual processing of stimuli (Eimer and Holmes, 2007; Luck, 2014); enhanced processing of infant distress cues has been observed in the N170 component (~170 ms post-stimulus onset; e.g., Kuzava et al., 2020; Peltola et al., 2014; Proverbio et al., 2006), which represents structural encoding of faces (Eimer, 2011) and indicates early detection of infant distress. Later ERP components generally represent sustained allocation of attention and elaborative cognitive processing and thus can inform the interpretative processing of infant cues (Eimer and Holmes, 2007; Hajcak et al., 2010; Luck, 2014). Notably, increased amplitudes in response to infant distress versus neutral or happy expressions are consistently evident in the P300 and the late positive potential (LPP; ~400-600 ms post-stimulus onset; Kuzava et al., 2020; Malak et al., 2015; Peltola et al., 2014; Rutherford et al., 2017), suggesting increased processing of, and sustained attention to, infant distress.

The few studies examining how maternal anxiety is related to ERP correlates of processing infant cues suggest that high levels of anxiety symptoms are especially related to the LPP component, particularly in response to neutral infant faces. Malak and colleagues (2015) had postpartum women (mean infant age of 8 months) passively view photographs of neutral and distress infant faces. For all mothers, the LPP amplitude was higher when viewing distress than neutral infant faces. However, in response to neutral infant faces, higher levels of self-reported anxiety were correlated with increased LPP amplitude. Of note, there was no association between maternal anxiety and the LPP elicited by distress infant faces. Rutherford and colleagues (2017) extended this work and reported a similar association in a sample of women in their third trimester of pregnancy, evidencing a link between prenatal maternal anxiety and LPP amplitude in response to neutral infant faces, controlling for maternal depression. Further, this link was specific to neutral infant faces— expectant mothers’ anxiety levels were unrelated to neutral adult faces or to neutral non-social (house) images (Rutherford et al., 2017). These findings suggest that maternal anxiety may impact processing of infant cues before birth, representing a potential neurobiological marker for prenatal vulnerability. An association between anxiety, LPP, and neutral child faces was also reported in a sample of mothers of school-age children who performed a Go/NoGo task (Kungl et al., 2020), suggesting that this anxiety-LPP link may persist from pregnancy throughout parenthood. Furthermore, consistent with the notion that the relation between maternal anxiety and LPP is specific to neutral infant and child cues, a recent study failed to find an association between anxiety in pregnancy or at four months postpartum and LPP response to neutral or negatively valenced images of objects and scenes, unrelated to parenting (Brooker et al., 2020). Together, these studies implicate prenatal and postnatal maternal anxiety with an enhanced LPP response to neutral infant faces, reflecting late-stage processing likely involved in the interpretation of infant cues rather than their detection.

Although the current body of research on maternal anxiety and ERPs is limited, it is also consistent in pointing towards the LPP as a promising direction for understanding how maternal anxiety may shape the transition to parenthood. Outside the context of parenting, the LPP has been interpreted as reflecting sustained attention to motivationally relevant emotional stimuli (e.g., Hajcak et al., 2010). In the case of maternal anxiety, the LPP appears sensitive to neutral infant cues, which can be considered ambiguous as they provide neither clearly-negative nor clearly-positive affective or physiological signals (by definition). In the general population, anxiety has been linked with negativity biases, denoting a tendency to perceive or interpret neutral and/or ambiguous information as negative or threatening (e.g., Fox, 2002; MacLeod and Cohen, 1993; van Bockstaele et al., 2014; Yoon and Zinbarg, 2007). Therefore, the sensitivity of the LPP to neutral infant cues may reflect a tendency to interpret these cues as negative (Malak et al., 2015; Rutherford et al., 2017), or to invest more effort in trying to decipher their emotional valence. An enhanced prenatal LPP response to neutral infant faces has also been linked to mothers’ tendency to misattribute their infant’s internal states or interpret them as excessively negative or hostile at three months postpartum (Rutherford et al., 2018), providing additional support for the notion that atypical LPP responsivity to neutral infant cues may reflect a disruption in maternal interpretation of these cues’ meaning. Interestingly, a recent study investigating fluctuations in maternal and infant arousal during interactions found that while nonanxious mothers were mostly reactive to high infant arousal, anxious mothers were also reactive to more subtle changes in the infant’s arousal levels (Smith et al., 2021), supporting the notion of a negativity bias in increased responses to infant neutral or low-distress cues in the presence maternal anxiety. The specificity of the link between maternal anxiety and the LPP response to neutral infant faces, but not adult faces or non-social stimuli, indicates that in the case of maternal anxiety, this effect is likely to be specific to processing of motivationally-relevant parenting-specific stimuli (Rutherford et al., 2017).

2.3. Interim Summary: Insights on the Associations Between Maternal Perinatal Anxiety and Processing Infant Cues

Maternal anxiety affects a considerable number of mothers in the perinatal period and predicts later maternal mood disorders (Heron et al., 2004; Martini et al., 2015; Skouteris et al., 2009) and child anxiety disorders (Lawrence et al., 2019; Micco et al., 2009), highlighting the impact of maternal anxiety on the well-being of two generations. However, more work is needed in understanding how maternal anxiety may impact child developmental outcomes. Given that maternal anxiety confers risk for compromised sensitive responsiveness in early caregiving (e.g., Nicol-Harper et al., 2007; Stevenson-Hinde et al., 2013), focusing on this period may be particularly valuable with view to breaking down the different stages of processing infant cues via neuroimaging methods (Figure 1). In this section, studies examining the associations between maternal neural responses to infant cues and anxiety symptoms were reviewed. In the majority of reviewed studies, anxiety levels were measured on continuous scales with self-report assessments.

As reviewed here, fMRI studies generally suggest that both hyporeactivity to positive cues in motivation/saliency neural circuits (e.g., amygdala) and hyperreactivity to negative cues in regulatory neural circuits (e.g., prefrontal cortex) are related to maternal anxiety, while ERP literature points towards hyperreactivity to, and sustained processing of, neutral infant cues. In considering the stages of processing infant cues as outlined in Figure 1, the reviewed findings suggest that maternal anxiety may exert a differential impact on infant cue processing depending on the cue’s valence: (1) lower saliency of positive cues may reflect under-detection or lower emotional reactivity; (2) increased late-stage processing of neutral cues may indicate interpretation of these cues towards the negative; and (3) emotional reactions to negative cues may require more effortful regulation. However, it should also be made clear that different findings in fMRI studies were not completely in line with one another; specifically, a study employing a range of stimuli in varying valence values showed an opposite pattern that suggested decreased regulation and increased saliency of infant stimuli in the presence of parenting-specific anxiety (Rutherford et al., 2019).

ERP and fMRI methodologies tap into different aspects of neurocognitive processing, and thus complement one another. While fMRI has high spatial and low temporal resolution (Ogawa et al., 1998), ERPs have high temporal resolution but are spatially limited to the cortex (Luck, 2014). It has been suggested that EEG taps into fast neural transmission channels whereas fMRI mostly taps slow channels (Hari and Parkkonen, 2015), such that rapid processing may not be captured by fMRI (Bornstein et al., 2013) whereas processing in subcortical areas may not be captured by EEG. Therefore, the reviewed body of work may imply that maternal anxiety differentially impacts either the time scale of, or the transmission pathways involved in, processing infant cues, depending on emotional valence. This conclusion is premature, however, as there are several methodological sources of variability that hinder the comparability between fMRI and ERP studies that may give rise to some of the scattered patterns observed in this review. Therefore, it is necessary to critically consider the limitations of the research discussed.

3. Limitations and Challenges in Studying Perinatal Maternal Anxiety and Processing Infant Cues

3.1. Methodological Issues in Studying Maternal Neural Processing of Infant Cues in the Presence of Anxiety

3.1.1. Challenges in leveraging the complementary information provided by fMRI and ERP methodologies.

An important limitation in the current literature is that variability in the stimuli sets employed hinders the ability to systematically compare between the studies reviewed here. For instance, the fMRI studies reporting associations between maternal anxiety and amygdala activation only used positive infant facial expressions (Barrett et al., 2012; Wonch et al., 2016), while most studies concerning anxiety and prefrontal activity only used negative adult facial expressions (Gingnell et al., 2015; Roos et al., 2011). An opposing pattern of results emerged in a study that aggregated several types of infant cues (faces and cries) and valence values (happy, sad, and neutral; Rutherford et al., 2019). Consequently, it is currently unclear whether maternal anxiety is characterized by decreased amygdala reactivity only to negative cues, or also to neutral infant cues; similarly, given evidence of differential impact of maternal anxiety on processing of infant compared to adult affective expressions (Rutherford et al., 2017), it is unclear whether one can extrapolate from increased prefrontal reactivity in maternal anxiety in response to negative adult expressions to how anxiety may impact processing infant cues, and thereby maternal sensitivity towards her infant. With respect to ERP research, studies that reported on the association between maternal anxiety and the LPP response to neutral infant faces did not include happy stimuli (Malak et al., 2015; Rutherford et al., 2017), which may be required to ascertain that the sensitivity of the LPP to neutral cues indeed reflects a negativity bias. The absence of consistent comparisons between valence values of infant affective cues or the inclusion of only adult face stimuli limits comparability between studies and a more general understanding of the specificity of maternal anxiety in shaping how mothers process different types of infant cues. In addition, most reviewed studies have included fairly small samples that limit the ability to detect reliable correlations between anxiety and neural measures.

Taken together, these methodological discrepancies and limitations point to the critical importance of creating more standardized, carefully controlled, and well-powered protocols in the assessment of maternal brain responses to infant cues within and across neuroimaging modalities. To be able to benefit from the complementary information that functional and electrophysiological methodologies can potentially provide, it is imperative for future research to employ equivalent experimental settings (i.e., conditions and tasks). Only then could the field discern whether variability in clinical symptomatology (in this context, maternal anxiety) differentially impacts (1) different types of infant cues, and/or (2) different levels of neural circuitry and time scales of processing infant cues.

3.1.2. Timing of assessments during the transition to parenthood.

Studies have also varied with respect to the timing of assessment during the perinatal period. Importantly, only two studies have examined prenatal anxiety and processing infant cues (Roos et al., 2011; Rutherford et al., 2017). The fMRI studies examining maternal anxiety and the neural circuitry involved in processing infant cues have focused on the postpartum period, likely related to the reluctance to scan women during pregnancy. By using fNIRS, Roos and colleagues (2011) were able to uniquely examine prenatal PFC activity at three time points across pregnancy. Their results indicated similar associations between anxiety and PFC reactivity to negative stimuli as patterns evidenced in postpartum fMRI studies (Gingnell et al., 2015; Noriuchi et al., 2008). However, given that fNIRS is limited to assessing cortical activity and has lower spatial resolution, the relation between prenatal anxiety and subcortical areas, such as those involved in reward and saliency circuits, remains unclear.

There is research to suggest that anxiety (as well as depression) trajectories and their relations with postpartum caregiving may change throughout pregnancy and postpartum (Hakanen et al., 2019; Heron et al., 2004), and that neurobiological changes occurring at different stages of pregnancy may have specific effects on postpartum maternal behavior (Toepfer et al., 2019). A recent study indicated that maternal prenatal LPP response to affective non-infant stimuli during the second, but not the third, trimester of pregnancy predicted infant temperamental reactivity (Brooker et al., 2020), suggesting that the timing of processes may also have an impact on infant emotion regulation. The association between maternal anxiety and the LPP to neutral infant faces appears consistent over the transition to parenthood, as indicated in samples of women in the third trimester of their pregnancy (Rutherford et al., 2017) and in the first year postpartum (Malak et al., 2015), as well as among mothers of school-age children (Kungl et al., 2020). However, the longitudinal trajectories of anxiety and LPP to infant faces from early in pregnancy to postpartum, and their associations with caregiving, have not been longitudinally examined. Therefore, further longitudinal work into changes in neural responding to infant cues and maternal anxiety from pregnancy to postpartum is needed.

3.2. Measuring and Conceptualizing Anxiety in the Perinatal Period

3.2.1. Disentangling anxiety and depression in the perinatal period.

Another important limitation to consider is that the reviewed studies vary in their consideration of maternal anxiety in relation to maternal depression. Some of the fMRI studies have not isolated the effect of maternal anxiety, but rather examined depression and anxiety together, or have focused on samples with clinical levels of depression (Barrett et al., 2012; Wonch et al., 2016). Thus, it is difficult to discern whether neural responding to infant cues is shaped by the shared commonality between these two comorbid conditions, or whether anxiety or depression are uniquely predictive of neural responding. For example, decreased amygdala reactivity to positive stimuli has also been observed in studies of maternal depression (e.g., Moses-Kolko et al., 2010), perhaps suggesting a more general blunted reactivity to emotional stimuli in maternal depression rather than maternal anxiety-specific effects.

In contrast, ERP findings relating maternal anxiety to the LPP response to neutral infant expressions remain when controlling for depression symptoms (Rutherford et al., 2017), supporting the notion that this pattern could be unique to maternal anxiety. Thus, while maternal depression may be related to more attenuated responses to emotional infant signals (e.g., Laurent and Ablow, 2012; Moses-Kolko et al., 2014; Rutherford et al., 2016), maternal anxiety may be related to higher engagement with potentially negative infant cues. This notion resonates with the conceptualization of maternal anxiety as a condition of heightened parental engagement in infant-related preoccupation, and perinatal depression as a condition of low parental preoccupation or disengagement (Swain et al., 2014b). Consequently, some findings, such as decreased amygdala reactivity to positive infant stimuli in both anxiety and depression (Barrett et al., 2012; Wonch et al., 2016), may reflect aspects of comorbid depression and anxiety, rather than anxiety in isolation.

Future studies should strive to disentangle the variance associated with anxiety and depression, especially as they may have differential associations with sensitive responsive caregiving (Booth et al., 2018). When examining continuous levels of anxiety symptoms (as in most of the current literature), one possible solution is to measure anxiety and depression using separate instruments and conduct correlational analyses while statistically controlling for depression symptoms as a covariate (e.g., as in Rutherford et al., 2017). Of note, when choosing this approach and selecting instruments, one should also keep in mind that some common measures of postpartum depression, such as the Edinburgh Postnatal Depression Scale (EPDS), also include items specifically addressing anxiety (Pawluski et al., 2017). Another approach is to examine group differences between mothers and pregnant women meeting diagnostic criteria for either an anxiety disorder, major depression, or both (as well as control, non-clinical group; see Feldman et al., 2009).

3.2.2. Domain-general vs. parenting- or pregnancy-related anxiety.

A more general limitation concerns the conceptualization of anxiety in parenting. Most of the studies examining maternal anxiety have focused on general-domain symptoms of anxiety, regardless of whether they arise in the parenting domain or in other contexts, or whether the content of worries and concerns is specific to parenting or not. The small number of studies examining postpartum parenting-related anxiety in relation to neural responses to infant cues (Kim et al., 2015; Rutherford et al., 2019) also raises the need to systematically contrast domain-general anxiety symptoms against parenting-specific anxious thoughts or concerns.

There are several potential limitations to focusing solely on general anxiety measures (e.g., such as the Beck Anxiety Inventory [BAI], Beck et al., 1988; or the State-Trait Anxiety Inventory [STAI], Spielberger et al., 1970). First, such questionnaires include symptoms that may not adequately capture more typical variability in anxiety in the perinatal period, for example by asking about sleep quality (Davies et al., 2021). Second, whereas general anxiety measures tap into worrying about negative outcomes more globally, a specific source of anxiety during this period concerns parenting and child-related thoughts and concerns, not captured by these scales (Huizink et al., 2004). Measures of prenatal anxiety (e.g., the Pregnancy-Related Anxiety Questionnaire-Revised [PRAQ-R]; Huizink et al., 2004) and postpartum anxiety (e.g., the Postpartum Specific Anxiety Scale [PSAS]; Fallon et al., 2016) tap into worries related to the child’s development and health, the parent’s competence, and the experience of pregnancy and seem to capture different variance in parenting experiences than variance related to general anxiety (Fallon et al., 2021; Huizink et al., 2004). To further elucidate mechanisms of anxiety in the perinatal period, studies should incorporate both domain-general and domain-specific aspects of anxiety in early parenting, and systematically examine their relative and combined contributions to caregiving in general, and processing infant cues in specific.

3.2.3. Variance in perinatal anxiety: Beyond linear correlations.

Relatedly, a further challenge in studying processing of infant cues in the context of perinatal maternal anxiety is that some levels of parenting- and pregnancy-specific anxiety may be adaptive as part of the “normative” transition to parenthood. In his seminal work, Winnicott (1984[1956]) described a stage of primary parental preoccupation, characterized as “almost an illness,” where new parents focus their attention on their infant in a manner that may resemble a state of heightened infant-focused anxiety. Although moderate levels of parental preoccupation during the third trimester and the first months postpartum are likely adaptive for creating a nurturing environment, excessive and persistent parental preoccupation may characterize anxiety (Swain et al., 2014b).

To the authors’ knowledge, studies to date have mainly explored linear associations between anxiety and neural responding to infant cues; non-linear (mainly quadratic) associations should be further explored to examine whether moderate levels of anxiety are adaptive for caregiving. Furthermore, as current studies have mainly relied on self-report scales in assessment of anxiety, discerning between clinical and non-clinical levels of anxiety based on diagnostic interviews, as well as assessing the onset of anxiety disorders (e.g., whether anxiety symptoms were present before pregnancy, started during pregnancy, or emerged postnatally) would facilitate the field’s understanding of how clinically-significant levels of anxiety may modulate neural processing of infant cues and sensitive responsiveness. In addition, the distinction between domain-general and parenting-specific anxiety may be particularly relevant for detecting deviations from neural responses related to the “normative” course of primary parental preoccupation, as this notion is essentially focused on parenting- and infant-related anxious concerns (Kim et al., 2016; Swain et al., 2011; Swain et al., 2014b).

To address this limitation, future studies should consider incorporating both domain-general and parenting-specific anxiety measures to elucidate the impact of anxiety on neural processing of infant cues by: (1) examining quadratic associations between neural measures and anxiety measures; (2) disentangling the contributions of domain-general and parenting-specific anxiety; and (3) identifying profiles of combinations between general anxiety, parenting-specific anxiety, and neural responses to infant cues longitudinally. These aims require considerably larger sample sizes than those employed in past studies, as well as recruitment of samples in which large variability in maternal anxiety (including exceptionally low levels of anxiety) may be captured.

4. Future Directions in the Study of Perinatal Maternal Anxiety and Processing Infant Cues: A Road Map

The study of how maternal anxiety modulates the neurocognitive basis of maternal processing of infant cues is still in its infancy. Several methodological and conceptual avenues for future research could help advance the field’s knowledge on maternal anxiety and how it may impact parenting and child development, and better inform evidence-based interventions. Several future directions are proposed in the following passages and are summarized in Figure 2: (1) systematic consideration of multiple levels of explanation, forging links between neural, cognitive, and behavioral levels of analysis; (2) more extensive utilization of advancement in neuroimaging techniques; (3) longitudinal tracking of symptomology and neural processing of cues across the transition to parenthood; and (4) research on paternal anxiety.

Figure 2.

A summary of open questions and future directions in studying the impact of parental anxiety on processing infant cues and sensitive responsive caregiving in the perinatal period. Notes. MVPA = Multivoxel pattern analysis. RSA = representational similarity analysis.

4.1. Levels of Analysis in the Study of Perinatal Maternal Anxiety and Neural Correlates of Processing Infant Cues

As discussed, maternal sensitive responsiveness can be conceptualized as the output of a series of processes (Figure 1). Analyzing maternal anxiety in the context of these processes can provide insight into how, and in what stages, maternal anxiety (and psychopathology more generally) may impair adaptive parenting behaviors. This conceptual framework resonates with social-cognitive neuroscience perspectives (e.g., Marr, 1982; Ochsner and Gross, 2008), wherein multiple levels of analysis—behavioral, cognitive, and neural— can be employed to understand human behavior. Social-cognitive neuroscience largely focuses on drawing connections between levels of analysis by using them to inform or constrain one another, or to contrast competing theories (Krakauer et al., 2017; Turner et al., 2017). This perspective is highly useful in gaining insights into the modulating effects of psychopathology, especially in developmental contexts (Cicchetti and Posner, 2005). In conceptualizing early maternal caregiving, these three levels of analysis correspond to connections between maternal behavior during mother-infant interactions, cognitive and affective processing of infant cues, and their neural instantiations.

Echoing general trends in cognitive neuroscience (Krakauer et al., 2017; Niv, 2020), the bulk of research on maternal anxiety and the neural correlates of processing infant cues has overlooked links between maternal brain and maternal behavior and cognition. To the authors’ knowledge, only one study has related neural data and observed maternal behavior in the context of maternal anxiety (Guo et al., 2018). Illuminating questions are therefore left unanswered— such as whether neural responses elicited by infant cues mediate the relations between maternal anxiety and compromised sensitive responsiveness. Considering the importance of temporal dynamics in maternal processing of infant cues and behavioral responsivity to these cues, future studies should especially focus on aspects of parenting that depend on timely and appropriate understanding of, and responsiveness to, infant cues, such as mind-mindedness (Meins, 2013), sensitivity (Ainsworth et al., 1974), contingency (Beebe et al., 2010), non-intrusiveness (Biringen et al., 2014), and affect synchrony (Feldman, 2012).

Further, only a handful of studies have incorporated cognitive or affective measures in their examination of maternal anxiety (Kim et al., 2015; Roos et al., 2011). Because different levels of analysis tap into different information, leaving out any level would limit the field’s understanding of maternal caregiving (c.f. Cacioppo et al., 2010) and increase susceptibility to reliance on reverse inference in interpreting studies’ results (Krakauer et al., 2017; Poldrack, 2006). While neuroimaging provides information that cognitive-behavioral measures may miss, measures tapping into cognitive and affective processes can shed light on the mechanisms connecting neural processing and behavior. For example, to bolster interpretations that increased prefrontal reactivity among anxious mothers reflects the selection of self-regulation strategies (Noriuchi et al., 2008; Roos et al., 2011), it would be necessary to incorporate additional validating measures, such as emotion regulation choices (Shafir et al., 2016).

Considering the robust association between LPP response to neutral infant (and child) cues and maternal anxiety (Kungl et al., 2020; Malak et al., 2015; Rutherford et al., 2017), further investigation of this link through multiple levels of analysis constitutes a particularly promising future direction. Cognitive-behavioral measures of negative interpretation bias can be incorporated to test the hypothesis that the LPP elicited by neutral cues marks a tendency to interpret neutral cues as more negative in maternal anxiety. This negativity bias may manifest in the perinatal period in frequent misinterpretation of neutral infant expressions as if they convey distress states; at the level of maternal behavior, an increased saliency of distress in the dyad would likely elicit inadequate responses to the infant, noncontingent to the signals the baby is displaying. The infant’s reaction, in turn, may be incompatible with the mother’s anticipations, further validating her anxiety and leading to a vicious cycle whereby the dyad is continuously dysregulated. These aspects of mother-infant interactions can be tested by incorporating measures of sensitivity, non-contingency, intrusiveness, and affect synchrony. Over time, the infant may consequently develop hypersensitivity to potentially negative stimuli, manifesting in fearful temperament, and later in behavioral inhibition and child negativity bias (as may be suggested by the association between prenatal maternal anxiety and child LPP; van den Heuvel et al., 2018). As this example illustrates, including multiple levels of analysis within the same (longitudinal) sample can be used to promote more complete mechanistic accounts of the impact of anxiety on maternal sensitivity and mechanisms in the intergenerational transmission of anxiety.

4.2. Further Utilization of Neuroimaging Techniques in Studying Maternal Anxiety and Processing Infant Cues

The reviewed body of work contributes to the field’s understanding of how anxiety may impact mothers’ processing of infant cues. However, only a narrow range of available techniques for analyzing neuroimaging data has been employed to date, and more widespread implementations of additional tools could potentially provide valuable information on untouched upon questions.

4.2.1. Multivoxel pattern analysis (MVPA).

Most fMRI studies in this review have focused on univariate analyses based on general linear models analyzing BOLD signals from single voxels or averaged across a region. MVPA of fMRI data has been gaining popularity in social-cognitive neuroscience (Popal et al., 2020; Weaverdyck et al., 2020), and opens a wide array of additional questions. MVPA enables detection of distributed patterns of activity across voxels which may not be detected by univariate analysis focused on peak or mean activity of the whole region (Haxby et al., 2001). Through training of supervised machine learning models, MVPA can be used to decode the content of mental representations. For example, one could use these patterns to examine whether a mother currently maintains a representation of a distressed or a happy infant cue, or her own versus an unknown infant (c.f., Goesaert and Op de Beeck, 2013). Representational similarity analyses (RSA; Kriegeskorte and Kievit, 2013) can be used to examine the extent to which patterns are similar across conditions, and can incorporate other neural (e.g., EEG) or behavioral (e.g., task performance) measures (Popal et al., 2020). These methods can be leveraged to understand the effect of anxiety on mothers’ representations of infant cues. For example, to examine negativity bias, one could test the similarity between neutral and distressed conditions as a function of anxiety; negativity bias would manifest in higher similarity between the two conditions or lack of ability to differentiate between them during decoding. Examination of these analyses across brain regions can also inform on stage of processing and whether representational structural changes in different neural circuits (c.f., Weaverdyck et al., 2020).

4.2.2. Electrophysiological correlates of processing infant cues.

Only three studies have examined electrophysiological correlates of infant and child cue processing in maternal anxiety; the consistency of their findings associating maternal anxiety with the LPP (Kungl et al., 2020; Malak et al., 2015; Rutherford et al., 2017) indicates that this method can provide valuable and reliable insights in studying maternal anxiety. As highlighted earlier, EEG methods are especially advantageous for understanding dynamic processes and for teasing apart stages of processing (Luck, 2014), such as the sequential stages of parental responsiveness to infant cues (Figure 1, panels A and B). Furthermore, because some aspects of early caregiving may happen intuitively (Papousek, 2000), some judgments may occur so rapidly that they may not be detectable by fMRI or self-report measures (Bornstein et al., 2013; Rutherford and Mayes, 2011). More widespread electrophysiological investigation of neural responses elicited by infant cues in the context of maternal anxiety may provide valuable knowledge on temporal dynamics of infant cue processing.

All the electrophysiological studies included in this review have focused solely on ERPs. Extensive research on EEG frequency-band oscillations has established these dynamics as measures of coordinated brain activity reflecting affective and cognitive processing as they develop in time (e.g., Fries, 2005; Klimesch, 2012; Knyazev, 2007). Additionally, EEG coherence within frequency bands and phase synchronization between different bands, tapping into coupling between oscillations, can complement fMRI connectivity studies by providing insight into how connectivity unfolds in the time domain (Sakkalis, 2011; Uhlhaas and Singer, 2006). In spite of evidence for differential EEG patterns in response to affective stimuli in anxiety and depression in the general population (e.g., Imperatori et al., 2019; Knyazev et al., 2016), these aspects of EEG data have not been examined in relation to maternal anxiety to date, and could help clarify aspects of mothers’ attention allocation and regulation.

4.3. Longitudinal Investigation of the Associations between Anxiety Symptoms and Neural Processing of Infant Cues Across the Transition to Parenthood

As mentioned, the perinatal period confers risk for maternal psychopathology (Saxbe et al., 2018), and anxiety in particular (e.g., Dennis et al., 2017; Fairbrother et al., 2016; Uguz et al., 2019), raising the possibility that neural changes occurring during this period may interact with maternal anxiety in shaping the maternal brain. Although there is some evidence to suggest structural and functional neural changes occurring during pregnancy and early postpartum (Barba-Müller et al., 2019; Cárdenas et al., 2019; Hoekzema et al., 2017), research on neural changes during pregnancy is scarce. Little is known on the extent to which variability in neural changes is implicated in maternal and child mental health and behavior (Cárdenas et al., 2019), and the specific role of anxiety in these neural changes has barely been addressed (Roos et al., 2011). Therefore, many open questions remain to be examined: How is maternal anxiety related to changes in the maternal brain occurring throughout pregnancy into the first months postpartum? Do pre-existing tendencies or psychopathologies beyond anxiety, such as self-criticism or negativity biases, interact with anxiety in modulating neural reorganization? Can specific patterns that distinguish normative maternal preoccupation from maladaptive anxiety be discerned already in early stages of pregnancy?

To better understand both normative and maladaptive processes during this period of psychological vulnerability and neural plasticity, future studies should longitudinally follow women from early in pregnancy to the postpartum period to better explore the covariance between trajectories of neural changes and anxiety symptoms throughout the transition to parenthood. Importantly, neural markers that may indicate increased risk for subsequent maternal anxiety and/or compromised sensitivity early in pregnancy, or even prior to conception, could help in devising means to identify women at risk and provide treatment before or during neural reorganization. The LPP is highlighted as a promising potential neural marker of prenatal risk, as maternal anxiety appears to be related to LPP responsivity as early as the third trimester of pregnancy (Rutherford et al., 2017) and prenatal LPP responsivity predicts postpartum attributions concerning the baby’s intentions (Rutherford et al., 2018). Thus, the emergence of the LPP’s responsivity to infant cues in earlier stages of pregnancy should be further explored longitudinally.

4.4. Paternal Anxiety and Parenting

This review has focused on maternal anxiety in pregnancy and postpartum as a period of increased risk for women during the formation of early mother-infant bonding. Although the transition to parenthood also confers risk for increased anxiety symptoms for fathers (Hughes et al., 2020; Leach et al., 2016), to the authors’ knowledge, studies have yet to examine how paternal anxiety may modulate fathers’ neural responses to infant cues. An emerging body of research on the paternal brain implicates generally similar neural circuits in parenting among mothers and fathers (Swain et al., 2014a), although this similarity may depend on fathers’ caregiving roles or experiences (Abraham et al., 2014). Given evidence for sex differences in emotional distress trajectories in early parenthood (Hughes et al., 2020) and differential contribution of maternal and paternal anxiety and worries to parenting and child anxiety (Möller et al., 2015; Pereira et al., 2014), it is likely that at least some sex differences emerge also at the neural level. Research on neural correlates of fathers’ processing of infant cues and anxiety during the transition to fatherhood would improve the field’s understanding of how this risk may impact paternal sensitivity. Moreover, a better understanding of the paternal transition would also provide a context for understanding maternal changes during the perinatal period, for example by contrasting biological (e.g., hormonal, fetal environment) versus environmental factors that may impact anxiety during pregnancy and infant neurodevelopment. Relatedly, future studies should also incorporate assessments of anxiety and neural responses to infant cues in nontraditional family structures, including in same-sex couples, co-parenting relationships, and surrogacy (Abraham et al., 2014; Imrie and Golombok, 2020).

4.5. Clinical Implications

Although more work is needed to clarify how maternal anxiety impacts mothers’ processing of infant cues, some clinical considerations can be derived from this review. First, neural correlates of processing infant cues can be useful dependent measures in studies examining outcomes of interventions and preventive programs among mothers with high levels of anxiety (c.f., Bernard et al., 2015; Swain et al., 2017). Two features of neural measures of infant cue processing render them especially effective in this context: (1) they are less susceptible to subjectivity and social desirability biases in comparison to self-report measures, and (2) they can help tease apart stages of processing infant cues, and thereby also inform on mechanisms of change in evaluating the therapeutic effects of intervention programs in non-clinical (Kolijn et al., 2020) as well as clinical or at-risk samples (Bernard et al., 2015). This review highlights the LPP response to neutral cues, amygdala response to positive cues, and PFC response to negative cues as potential neural dependent measures for evaluating the efficacy of interventions and preventive programs among mothers with anxiety, and their underlying mechanisms. A recent study on ERP changes following a parenting intervention administered to a non-clinical sample of parents of preschool/school-age children failed to find an intervention-related change in the LPP (Kolijn et al., 2020); however, we suggest applying a similar approach in a sample of mothers exhibiting higher levels of anxiety and focusing on the LPP response to neutral infant/child faces could prove important in understanding modulation of anxiety-LPP associations by parenting-based interventions.

Second, a promising avenue for future research and clinical practice concerns the application of EEG/ERP as screening tools in clinical settings, given its scalability for large samples. Neural markers that may indicate increased risk for subsequent maternal anxiety and/or compromised sensitivity early in pregnancy, or even prior to conception, could help in devising means to identify women at risk and provide early treatment, while incurring minimal risk. As emphasized, the LPP component shows promise as a biomarker of maternal anxiety during pregnancy (Rutherford et al., 2017) and in predicting subsequent postpartum parental attributes (Rutherford et al., 2018), and its utility as a screening tool for risk during pregnancy should be further investigated.

Finally, future translational research can build on this review to offer additional targets for intervention and prevention. The conclusions of this review suggest that maternal anxiety may impact maternal processing of infant cues in multiple stages depending on cues’ emotional valence. Therefore, interventions targeted at bolstering maternal sensitivity in this population may benefit from fine-tuning mothers’ responses to different types of infant cues based on context. Specifically, working with mothers on modifying their interpretations of ambiguous cues as less negative may help mothers reduce the saliency of negative emotionality in the dyad. At the same time, clinicians can work with mothers to better identify and respond to positive emotions in the dyad to increase their saliency and reward value. Notwithstanding, a cautionary note should be made as more research on how neural biomarkers are related to observed behavior and bonding among mothers with high levels of anxiety is required.

5. Conclusion

In summary, an understanding of how maternal anxiety may shape the neural correlates of responding to infant cues is limited, despite the high prevalence rates of perinatal anxiety amongst women. The reviewed literature suggests a complex pattern wherein anxiety is likely implicated in multiple stages of detection, interpretation, and reaction to infant cues. The functional neuroimaging literature overall revealed lowered reactivity to positive cues in motivation and saliency circuits and increased involvement of regulatory circuits when processing negative cues, whereas the electrophysiological literature overall indicates an extended elaborative processing of neutral infant facial cues suggesting a negativity bias in interpretation of ambiguous cues. However, several methodological inconsistencies between studies hinder the generalizability of these conclusions and highlight the importance of adopting standardized and well-controlled paradigms in the study of maternal processing of infant cues, including comprehensive inclusion of valence conditions, distinguishing the effects of anxiety symptoms from those of depression symptoms (which may differentially impact maternal processing of infant cues), and differentiating between domain-general and parenting-specific anxiety symptoms. Several paths of expansion for this body of literature are promising, with the possibility of systematic links between neural, cognitive, and behavioral levels of analysis, further utilization of neuroimaging techniques, longitudinal work, and consideration of paternal anxiety. Considering the vast evidence of the associations between maternal anxiety and negative outcomes for mothers and children, the possibility of abetting such outcomes holds promise for clinical prevention and intervention.

Highlights.

• Anxiety symptoms are prevalent among women in the prenatal and postnatal periods

• This review focuses on how anxiety modulates neural responses to infant cues

• Anxiety modulates detection, interpretation, or regulation, depending on valence

• Limitations include inconsistent methodologies and consideration of depression

• Future work should incorporate cognition and behavior for multiple analysis levels