THE IMPACT OF DEPRESSION ON MATERNAL RESPONSES TO INFANT FACES IN PREGNANCY

ABSTRACT Research has suggested that prenatal depression may be associated with disrupted maternal responses to infant stimuli, with depressed pregnant women not showing the bias toward distressed infants as that observed in nondepressed pregnant women. The current study examined the effects of depression on self‐ reported responses to infant stimuli, in early pregnancy. Women with clinical depression (n = 38), and nondepressed women (n = 67) were recruited from a wider cognitive behavioral therapy trial. They completed Maternal Response Scales in which they were presented with images of distressed, neutral, and happy infant faces, with no time limit. The women rated their responses to these images along three dimensions—wanting to comfort, wanting to turn away, and feelings of anxiety—using Likert scales via a computerized task. There was evidence that women with depression in pregnancy showed different responses than did women without depression. Women with depression were substantially more likely to be in the highest quartile for ratings of wanting to turn away, odds (OR) ratio = 4.15, 95% confidence intervals (CIs) = 1.63–10.5, p = .003, and also were substantially less likely to be in the highest quartile for wanting to comfort a distressed infant face, OR = 0.22, 95% CIs = 0.09–0.54, p < .001. Findings are consistent with there being both a heightened avoidant and a reduced comforting response toward distressed infants in depressed pregnant women, providing some support that depression disrupts maternal preparations at a conscious level.


Keywords
The effects of maternal depression on child outcomes are well-documented, including its impact on child language development, cognitive functioning, and ongoing risk of psychosocial and emotional difficulties (Goodman et al., 2011). Although it is unknown exactly how depression has an impact on child outcomes possible mechanisms include: shared heritability, mediation through stressful life events or mediation through aspects of parenting such as exposure to negative maternal cognitions, affect and behaviours (Goodman et al., 2011;Goodman & Gotlib, 1999). A growing body of research has begun to examine the influence of maternal responsiveness on mother-child interactions, the importance of how attentive to infant-related stimuli the mother is, and how motivated she is to respond. Subsequently, a number of studies have found that postnatal depression is associated with disrupted mother-infant interactions (Murray, Fiori-Cowley, Hooper, & Cooper, 1996) through low maternal responsiveness (defined as a low attraction or motivation toward anything infant-related (Pearson, 2010).
Although research to date has focused on maternal responsiveness postnatally, there is increasing evidence from both animal and human studies to suggest that maternal responsiveness develops before birth and across pregnancy (Fleming, Steiner, & Corter, 1997;Maestripieri & Zehr, 1998). Studies also have indicated that sensitivity toward emotional faces and attentional biases toward infant stimuli develop across pregnancy (Pearson, Cooper, Penton-Voak, Lightman, & Evans, 2010;Pearson, Lightman, & Evans, 2009. Importantly, the maternal responsiveness and motherchild attachment observed in pregnancy appears to be predictive of maternal sensitivity postnatally, indicating that the patterns observed in pregnancy may be indicators of later maternal responses (Pearson et al., 2011;Kim, Park, & Shin, 2006;. A number of research studies have linked maternal perinatal depression to insecure or disorganized mother-infant attachments (Campbell et al., 2004;Wan & Green, 2009). Although research looking at the developmental trajectories of such children is still at an early stage, it may be that attachment relationships mediate the transmission of "disorders" such as depression (Wan & Green, 2009). Evidence has suggested that maternal sensitivity or responsiveness is associated with attachment status (Ainsworth, 1979;Bigelow et al., 2010;Egeland & Farber, 1984;Isabella & Belsky, 1991;Meins, Fernyhough, Fradley, & Tuckey, 2001). It appears, then, that maternal responsiveness to infant needs is linked to, or predictive of, attachment security (Mills-Koonce, Gariepy, Sutton, & Cox, 2008).
Maternal responsiveness is typically measured by either recording observed behavior or by measuring cognitive processing of related stimuli that are salient, such as infant faces. Evidence from observational studies have found differences in how the mother cradles her infant ) as well as the amount of maternal eye gaze, maternal language, the amount of gesturing, and the level of coordinated attention between mother and child Campbell et al., 2007). Research also has indicated that women who have antenatal depression, as compared to nondepressed women, are less sensitive when playing with their baby (Murray et al., 1996).
In research using cognitive-processing measures, infants' faces are often used because they are considered to be "salient" stimuli for mothers. These are used to measure maternal sensitivity, as infant facial expression is an important nonverbal signal from the infant to the mother. Evidence also has suggested that the processing of infant faces is distinct from that of adult faces because infants have a different facial configuration (Lorenz, 1971;Stein et al., 2010), meaning that the measure is specific rather than just capturing changes in responses to faces in general.
When applying this approach to measuring maternal responsiveness during pregnancy, Pearson, Cooper et al., 2010 found that depressed pregnant women did not show an attentional bias toward distressed as compared with nondistressed infant faces whereas nondepressed pregnant women did show this bias. The results suggested that women with depression had a bias in the opposite direction (i.e., away from distressed infants), although the confidence intervals included no effect. It is not clear whether there is a true difference that would been seen only in a larger sample, and if so, whether this difference is due to an increased speed of the depressed women to disengage from distressed faces or due to a slower speed to disengage from nondistressed faces. This study focused on early cognitive processing, and previous studies of early cognitive processing have produced inconsistent results. The length of time that a stimulus is presented is thought to influence its emotional valence, with increased presentation time being associated with increased emotionality . For example, in a design utilizing later cognitive processing, Stein et al. (2010) found that when infant facial stimuli were presented for 200 ms as opposed to 100 ms, mothers rated the infant emotions as more extreme. Women with postnatal depression also rated "negative" infant faces more "negatively" than did nondepressed women. Although this study has indicated that mothers with depression may rate negative infant stimuli more negatively, the mother's own emotional responses to the infant facial stimuli are not clear. Processing of infant stimuli in the context of perinatal depression is complex, and it is possible that although these mothers may judge stimuli as negative, their own emotional responses may not be affected, such as wanting to comfort a distressed infant.
The current study aimed to examine self-reported emotional responses to infant stimuli from early to midpregnancy (6-18 weeks at recruitment) in women with and without depression. Looking at women's self-reported responses to stimuli early in pregnancy may help to develop an understanding of mothers' interpretation of, and behavior toward, infant emotion. The study aimed to investigate whether the responses of depressed women were most consistent with a lack of reward (not wanting to comfort) or with an avoidant response (wanting to turn away). To control for the fact that a woman may not want to comfort because they are experiencing anxiety (avoidant response), measures of anxiety also were included to assess whether comforting responses were independent of anxiety responses. We hypothesized that pregnant women with depression would report heightened avoidance (turning away) and a reduced comforting response to the presentation of infant faces. Following previous evidence, we also hypothesized that this effect would be the strongest for distressed infant faces.

Participants
A total of 176 pregnant women were referred to the study by community midwives in the North Bristol region at their booking appointments. They were initially recruited for a wider study aiming to validate a depression screening instrument, which also included a pilot cognitive behavioral therapy (CBT) trial. Women were invited to participate in the current study if they were able to fluently read English and were less than 18 weeks pregnant so that they did not exceed the eligibility window of 19 weeks by the time they had completed the study. They were excluded if midwives were aware of any current psychotic illness. A total of 379 responses were received from the midwives, including 71 (19%) who declined to take part in the study at booking. Of those agreeing  (7) 4 (4) 4 (11) 0 (0) No Data for Occupation 23 (13) 16 (15)  5 (13) 12 (18) Ethnicity, n (%) White Caucasian 161 (91) 105 (100) 38 (100) 67 (100) Other 15 (9) 0 (0) 0 (0) 0 (0) to take part, 88 (29%) screened positive, and 220 (71%) negative for depression. Eighty women were excluded, and 52 declined after researcher contact (see Table 1), leaving 176 women. Only a subsample of the women who were originally recruited took part in the current study due to time limits. All participants gave informed consent. Of the 105 women who completed the Maternal Response Scales (MRS) designed for this study, 38 women were identified as depressed using the Clinical Interview Schedule-Revised (CIS-R;  according to International Statistical Classification of Diseases and Related Health Problems, 10th Edition (ICD-10; World Health Organzation, 1992) diagnostic criteria (n = 38), and 67 women were identified as having no diagnosis of depression (n = 67). In the final sample, mean age of participants was 31 years, and mean weeks of gestation was 13. Primiparous women represented 47% of the sample. At the time of assessment, only 8 women were currently taking antidepressant medication, and 8 had stopped medication due to pregnancy. Full sample demographics are included in Table 1.

Procedure
Between May 2010 and February 2011, women who were referred to the study by midwives during routine practice were visited at home by a researcher. The women were then asked to complete the CIS-R , the Edinburgh Postnatal Depression Scale (EPDS; Cox, Holden, & Sagovsky, 1987), a questionnaire regarding demographic and pregnancy information, an attention bias task not reported here (for a full description and previous findings, see Pearson, Cooper et al., 2010), and the MRS. The women always were asked to complete the attention bias task before the MRS task, in case the conscious processing involved in the MRS task influenced women's responses in the attention bias task, which involves early cognitive processing. Ethical approval was given by the Southmead National Health Service Research Ethics Committee and North Bristol Trust, as part of a wider CBT trial (09/H0102/75).

Measures
The CIS-R . The CIS-R is widely used with community samples and is a computerized, self-administered interview that generates severity scores on a scale of 0 () to 4 () for 14 classified neurotic symptoms encompassing both anxiety and depression . These symptoms then cluster according to ICD-10 criteria. The interview is fully standardized and is as reliable as the Composite International Diagnostic Interview (CIDI; World Health Organization, 1992) (Robins, Wing, & Wittchen et al., 1998). It also is as reliable when conducted by a lay or trained interviewer . The CIS-R uses a total symptom score of 12 or above to classify individuals as having a clinically significant mental health disorder. Because some of the CIS-R symptoms classifying a diagnosis of depression also can be problematic indicators in pregnancy because they involve physical symptoms, the EPDS also was completed, which does not include these.
The EPDS. The EPDS (Cox et al., 1987) is a self-report questionnaire that screens for postnatal depression symptoms by avoiding the use of physical prognostic indicators. The EPDS is rapid to use and has high reliability and validity for both postnatal and antenatal women (Cox et al., 1996). The EPDS uses a symptom score from 9 to 12 to classify mild to moderate depressive symptoms and over 13 to classify major depressive symptoms (Cox et al., 1987).
The MRS. A measure of self-reported emotional response to infant facial expressions was designed and developed for this study. Ten pregnant or recently pregnant women were consulted as to relevant emotional language to use on the measure, and the scales were piloted for understanding of language and ease of use. Previous use of similar scales in the research literature  was examined, and the scales were amended after piloting to be on a scale from 1 (low) to 8 (high) to be a more sensitive measure that is able to detect subtle variations in emotional self-report.
Administering the MRS involved presenting distressed, neutral, and happy infant face images (seven of each). These images were randomized and presented in e-Prime Professional 2.0 software (Psychology Software Tools, Pittsburgh, PA), with no time limit, to allow for the stimuli to be consciously processed. Conscious processing in this kind of task is thought to take effect after 500 ms (R.P. Cooper, 2010). Mean time taken in this study was 197,536 ms (for full reaction times, see Table 2). Under each picture, one of three MRS Likert scales was presented (see Figure 1): I want to comfort, I want to turn away, and I feel anxious. Each image was presented three times, in random order, each time with one of the scales presented underneath the pictures (anxiety, comforting, turning away). Participants were asked to press the number key that felt the most relevant to them for each picture and scale. To reduce social desirability effects, participants were encouraged to press the most meaningful response for them and were told that no answer was right or wrong.
The infant faces used in this study were those developed by Pearson, Cooper et al. (2010). The pictures were cropped, converted into gray scale, and matched to adult faces (Ekman & Friesen, 1971) for brightness positioning and contrast. Pictures included distressed, neutral, and happy faces ( Figure 2).These images were piloted on nonpregnant women to test for emotional valence, and participants were forced to make a decision regarding whether the infant was distressed, neutral, or happy. Mean accuracy for this was 90 to 100%. These images have been validated in previous studies examining perinatal attentional processing (Pearson, Cooper et al., 2010), and were used with full permission from the authors.

Analyses
The analyses for this study followed three stages. Following from our hypotheses, Stage 1 examined the relationship between depression and the primary responses of turning away and wanting to comfort, for distressed infant faces. Stage 2 examined these   For all analyses, the independent variable was depression ("depressed/nondepressed" according to CIS-R classification), with the dependent variable being maternal response (The MRS). These analyses were conducted using ordinal logistic regressions, as the dependent variable (MRS responses) was converted into an ordinal variable because it was not normally distributed and could not be normalized using transformations. As such, the MRS data were grouped into quantiles to derive an ordinal variable. The data fit most evenly into quartiles, with close-to-equal frequencies. The quartiles and their score ranges are shown in Table 3. Single predictor unadjusted ordinal logistic models were initially run on the data. The models were then adjusted for parity and age.
All analyses were repeated using the EPDS. As results were comparable, we primarily present results us-ing the CIS-R, which is the more in-depth and diagnostic instrument.
Ordinal logistic regression models examine the odds ratio (OR) of an individual scoring at one level of an ordinal variable compared to the lowest level. Generalized rather than simple ordinal logistic regressions were conducted, as the proportionality of odds assumption was violated. The generalized ordinal logistic regression models calculated the OR for being in each quartile of the MRS (compared to the lowest quartile) according to whether women were depressed.
Primary analyses investigated the relationship between depression and responses of wanting to comfort and wanting to turn away. Our primary analyses focused on these responses to distressed infant faces because previous evidence has suggested that depression appears to particularly affect responses to distressed infant stimuli (Pearson, Cooper et al., 2010). Turning away and anxiety responses were correlated (see correlation matrix, Table 4); however, we also investigated anxiety responses in secondary analyses.

Impact of Depression on Turning Away to Distressed Faces
The ordinal logistic regression model, adjusted for age and parity, provided evidence that depressed women were substantially more likely to be in the highest (vs. the lowest) quartile for re-porting "wanting to turn away," as compared to women without depression, OR = 4.15, p = .003, CI = 1.63-10.50. Depression did not influence the likelihood of being in the second or third quartile, as compared to the lowest quartile. The percentage of depressed and nondepressed women in Quartiles 1 and 4 is shown in Figure 3. Response patterns were  reversed in depressed and nondepressed groups of women (see Table 5).

Impact of Depression on Wanting to Comfort Distressed Faces
The ordinal logistic regression model provided evidence that depressed women were significantly less likely to be in the highest (vs. lowest) category for wanting to comfort, OR = 0.22, p < .001, CI = 0.09-0.54. Depression did not influence the likelihood of being in Quartile 2 or 3, as compared to the lowest quartile. The percentage of depressed and nondepressed women in the lowest and highest quartiles (Figure 4) again shows a reversed trend for depressed and nondepressed women.

Neutral Faces
The same response pattern for depressed and nondepressed women was observed for turning away to neutral faces, with depressed women being significantly more likely to be in the highest versus the lowest quartile for turning away, OR = 4.46, p = .04, CI = 1.59-12.49. However, the lack of a comforting response toward stimuli was found to be specific to distressed faces for women with depression, as there was a nonsignificant effect of depression on wanting to comfort neutral faces, OR = 0.72, p = .5, CI = 0.28-1.87.

Happy Faces
There was a significant effect of depression on wanting to turn away from happy infant stimuli, with depressed women being more likely to be in the highest quartile for turning away, OR = 4.14, p = .003, CI 1.63-10.51, as compared to nondepressed women. However, there was no effect of depression on the likelihood of wanting to comfort happy faces, OR = 0.61, p = .33, CI = 0.23-1.64.

Independent Effect of Depression on Distress
These models were run again in a combined model including responses toward distressed, neutral, and happy faces. This allowed us to explore whether there were any independent effects of depression on responses to distress specifically, over and above general responses toward infant faces. This model indicated that the effect of depression on women's desire to comfort a distressed infant remained following adjustments for their responses toward happy and neutral faces, OR = 0.16, p = .003, CI = 0.05-0.53. However, there was not an independent effect of depression on women's desire to turn away from a distressed infant; the effect diminished following adjustment for responses to neutral and happy faces, OR = 2.61, p = .091, CI = 0.86-7.92.

DISCUSSION
To our knowledge, this is the first study investigating selfreported responses to infant stimuli in prenatally depressed women.
The study indicates that prenatally depressed women, in earlymidpregnancy (8-19 weeks) already report an altered response pattern to infant stimuli, as compared to women who are not depressed. Early in pregnancy, depressed women reported that they were much more likely to want to turn away from an infant when shown a picture of an infant face displaying any emotion and much less likely to want to comfort the infant when shown distressed infant faces specifically. This effect was independent of parity and maternal age.
Investigating responses in pregnancy is important, as there is evidence to suggest that pregnancy may be important in the development of emotional and behavioral responses to infants in preparation for the maternal role, and that depression may disrupt this. For example, research has suggested that depression in pregnancy is predictive of maternal ratings of child affect at 6 months' postpartum (Huot, 2004). Previous studies also have examined early attentional processing of infant stimuli, which is potentially relevant to maternal responsiveness (Pearson, Cooper et al., 2010). In women with perinatal depression, later conscious processing of infant faces were examined, and showed that women with perinatal depression rate infant faces more extremely than do women without depression . The study reported here develops this further by investigating women's reports of their own responses to infant faces. One of the strengths of this study is its sample size: 105 women early in their pregnancies. Women with depression were overrepresented in the study (n = 38), as compared to the normal pregnant population, due to midwives preferentially referring women with depression to the study. Data on medication were available for only 83 of the 105 women; however, an even number of depressed and nondepressed women were currently taking or had, because of pregnancy, stopped taking antidepressant medication, so this small number is unlikely to have had an influence on the results. However, the nature of the recruitment strategy also allowed a large amount of demographic information to be collected, allowing for potentially confounding variables such as age and parity to be adjusted for in statistical analysis.

Potential Mechanisms: The Understimulation or the Overstimulation Hypothesis
Two main hypotheses could explain the difference in responses to infant stimuli found in women with prenatal depression. The first hypothesis is that maternal responses may be hypersensitive and avoidant in women with depression. This is supported by studies that have indicated that individuals with depression show hypersensitivity to, followed by avoidance of, punishment and negative stimuli (Elliott, Sahakian, Herrod, Robbins, & Paykel, 1997;Eshel & Roiser, 2010). There is evidence that depressed women give more attention to and are more easily able to recognize distressed faces as well as misinterpret neutral or happy faces as distressed (Gur et al., 1992). The second hypothesis is that women with depression may be experiencing an understimulation of maternal responses due to impaired reward responses . For example, there is evidence that depressed woman show reduced reward processing. They show anhedonia (the lack of a positive bias) as well as the reduced ability to learn a delayed reward association (Pizzagalli et al., 2009). In this way, it is possible that depression in pregnancy disrupts the learning of positive reward association toward infant faces. Functional magnetic resonance imaging studies also have suggested that women with depression show less brain activation in neural areas involved in reward processing in response to their own infant's cry than do women without depression (Laurent & Ablow, 2011).
This study aimed to explore further whether women with prenatal depression were more likely to report reduced reward responses (not wanting to comfort) or heightened avoidant responses (wanting to turn away) when presented with infant faces. Findings indicate that women with prenatal depression have a reduced response of both types; that is, both reduced reward responses and heightened avoidant responses to infant distressed faces, as compared to comparison nondepressed women.
It could be argued, however, that depressed women will show an altered response to almost any stimuli, and not specifically to infant stimuli. Importantly, within this study, although women showed both an increased avoidant and a lowered reward response to the distressed infant pictures, the reduced comforting responses were observed only for distressed infant faces, and not for neutral and happy faces. The strongest observed effect was reduced comforting responses, only seen in depressed women for distressed infant faces, indicating that women with depression during pregnancy specifically have a reduced desire to comfort distressed infants and are not simply unresponsive to infant faces in general. This can be contrasted to the heightened turning-away response, which was seen in depressed women for all faces (distressed, neutral, and happy), indicating that women with prenatal depression were more likely to report heightened wanting to turn away from all of the infant stimuli regardless of level of distress. The fact that the depressed women in this study showed the strongest response to distressed infant faces antenatally is potentially highly significant for behavioral responses toward infants postpartum. These results provide further support to the idea that some aspects of maternal responsiveness may develop antenatally. Pregnancy provides a potential opportunity to intervene to improve outcome for mothers and their offspring.

Limitations
The scales used in this research were developed specifically for the study and are a relatively simple measure of emotional selfreport. Although they have face validity, they have not been tested for construct validity; therefore, it cannot be assumed that a selfreport of not wanting to comfort indicates a lack of reward. In addition, as no further variables after parity and age were included in the analysis, it remains possible that other contextual factors such as marital conflict could influence MRS scores independently of depression. Although anxiety was accounted for using the MRS, diagnosis of anxiety using the CIS-R was not adjusted for, which also may have influenced MRS scores. This topic would be useful to further research.
Limitations to this study also include the fact that emotional self-report is prone to bias. For example, although not wanting to comfort was the strongest effect in this study, it is possible that women with depression felt that it was acceptable for them to report that they do not want to comfort but not acceptable to say that they want to turn away. Further limitations include that the infant images used in this design were not the women's own children. It could be argued that the women in this study's responses to unknown infant faces cannot be generalized to their responses and behaviors that they would show later in response to their own infants.

Conclusion
This study indicates that early in pregnancy, women with depression are more likely to report wanting to turn away from all infant faces, whatever the emotion displayed. They also are specifically less likely to report wanting to comfort if the infant face appears distressed. The fact that this response was present so early in pregnancy has implications for the ways in which depressed mothers may interact with their infants behaviorally before and after birth. As this study sampled only women in the first and second trimesters of pregnancy, it also would be interesting to study women in their third trimester to investigate the development of these conscious maternal preparations.