The influence of paternal and
maternal depressive and
anxious symptoms on infant
emotion processing.
An eye-tracking experiment.
Masterscriptie Orthopedagogiek Pedagogische en Onderwijskundige Wetenschappen Universiteit van Amsterdam M. van der Vliet Supervisor: Drs. E. Aktar Tweede beoordelaar: Dr. W. de Vente Amsterdam, april 2014
1 Table of contents Abstract ...2 Samenvatting ...3 Introduction ...4 Method ... 11 Participants ... 11 Instruments ... 11 Procedure ... 13
Preparing the data for the analysis ... 13
Analysis ... 14
Results ... 16
Descriptives ... 16
Analysis of a basic model with the main effect of time: fixed order data ... 16
Testing the effect of emotion and the presence of a negativity bias: fixed order data ... 18
The influence of parental depression on emotion processing: fixed order data ... 19
The influence of parental anxiety on emotion processing: fixed order data ... 20
Analysis of a basic model with the main effect of time: random order data ... 21
Testing the effect of emotion and the presence of a negativity bias: random order data ... 21
The influence of parental depression on emotion processing: random order data ... 24
The influence of parental anxiety on emotion processing: random order data ... 24
Discussion ... 26
2
The influence of paternal and maternal depressive and anxious symptoms on infant emotion processing. An eye-tracking experiment.
Abstract
Infants’ reactivity towards emotion expressions was suggested to be stronger towards negative than positive and neutral expressions (Vaish et al., 2008), an effect called negativity bias. Parental depression and anxiety seem to affect emotion processing, but research is scarce and mainly concerns mothers. This research aimed to test these processes and included both parents. The experiment included 63 14-month-old infants, from the general population. Their reactivity was measured by an eye-tracker, where bigger pupil sizes suggest more arousal. The infants watched photos of a neutral, happy, fearful, angry, and sad facial expression five times. Both parents filled in questionnaires about anxious and depressive symptoms. Infants showed more reactivity towards the happy expression than the neutral one. Results on the negative emotions were inconclusive. Depressive and anxious symptoms in fathers predicted less reactivity to each emotion. Maternal symptoms of anxiety predicted higher reactivity to each emotion. Her depressive symptoms predicted lower reactivity towards anger and higher reactivity towards happiness and sadness. So, even mild symptoms affect infant development and influences are different for fathers and mothers.
3
De invloed van symptomen van depressie en angst bij vaders en moeders op de emotieverwerking van baby’s. Een eye-tracking experiment.
Samenvatting
Baby’s reacties op uitgedrukte emoties zouden sterker zijn richting negatieve dan positieve of neutrale uitdrukkingen (Vaish et al., 2008), een effect genaamd negativity bias. Depressie en angst van ouders lijken de emotieverwerking te beïnvloeden, maar onderzoek is schaars en vooral gericht op moeders. Het doel van het onderzoek was deze processen te bekijken bij beide ouders. Het experiment betrof 63 baby’s van 14 maanden uit de algemene populatie. Hun reactiviteit werd gemeten door een eye-tracker, waarbij grotere pupillen een hogere alertheid impliceren. De baby’s bekeken foto’s van een neutrale, blije, bange, boze en verdrietige gezichtsuitdrukking vijf keer. Beide ouders vulden vragenlijsten over angstige en depressieve symptomen in. Baby’s lieten een sterkere reactie zien richting de blije uitdrukking dan de neutrale. Resultaten van de negatieve emoties waren inconsistent. Vaders’ angstige en depressieve symptomen voorspelden minder sterke reacties op elke emotie. Moeders’ angstige symptomen voorspelden sterkere reacties op elke emotie. Moeders’ depressieve symptomen voorspelden minder sterke reacties op boosheid en sterkere reacties op blijdschap en verdriet. Dus zelfs milde symptomen beïnvloeden de ontwikkeling, op verschillende wijzen voor vaders en moeders.
4
Introduction
From the moment they are born, children are surrounded by information about their social and physical environment. One source of this information is the expression of emotions by people around the child. Although their visual acuity does not reach an adult level until they are approximately eight months old (Gross, 2011), infants as young as two months old can discriminate happy faces from sad or angry faces (Nelson & Horowitz, 1983). In the second half of the first year, infants start to perceive basic emotions as organized wholes. They are able to distinguish intense and mild expressions of different emotions (LaFreniere, 2012). Besides, they respond to sad or fearful faces differently than to happy or surprised faces, even if different people express the emotions in slightly varying ways (LaFreniere, 2012). After this, the ability to process facial emotions continues to develop into adolescence (De Haan, Belsky, Reid, Volein, & Johnson, 2004).
The early ability to recognize emotions is adaptive, because it can help infants to decide how to react to uncertain events. Infants between ten and twelve months old start using social referencing to determine what to do when confronted with novelty (Gross, 2011; LaFreniere, 2012; Vaish, Grossman, & Woodward, 2008). That is, when infants come across a new and ambiguous situation, they look at their parents to decide how to react. If parents give a positive facial expression, children are more likely to approach the situation, while if parents give a negative facial expression, children will avoid it. Social referencing enables infants to regulate their emotions (LaFreniere, 2012) and to learn from others about their environment (Vaish et al., 2008). This sociocognitive skill does not only aid basic survival, but is also involved in the transmission of culture (Vaish et al., 2008).
Building on the results of social referencing studies, Vaish, Grossman, and Woodward (2008) proposed a theory on the early development of emotional processing. According to this theory, in the first months of their lives, most infants see positive faces only or most of the time (Malatesta & Haviland, 1982). The consequence might be that the evaluative neutral point of infants becomes skewed towards a positive evaluation of the situation, instead of being an evaluation between positive and negative (Vaish et al., 2008). In the second half of the first year, infants start to move around by themselves. This independent locomotion,
starting with crawling and walking, brings the infant close to household hazards, such as stairs, hot beverages, sharp objects, and fire places. At this point in development, most parents seem to start showing more negative emotions, such as fear and anger, towards their children to regulate their behavior (Campos et al., 2000; Vaish et al., 2008). At approximately the same
5 time, infants start to use social referencing. This combination of events might be the reason for the development of a negativity bias in the months around an infant’s first birthday (Vaish et al., 2008). This bias causes children and adults to show a stronger response to negative than to neutral or positive stimuli. For example, in studies that compared the influence of positive, neutral en negative conditions, infants behaved in the same way when facing positive and neutral cues. Negative cues, on the other hand, did affect or change the infants’ behavior or attention (reviewed in Vaish et al., 2008). The negativity bias seems to serve an evolutionary adaptive purpose: it helps people to explore their environment safely, while appropriately avoiding harmful situations (Vaish et al., 2008).
Recent research indicates the presence of a negativity bias in processing facial expressions of emotions (Hunnius, De Wit, Vrins & Von Hofsten, 2011). Both adults and 4- and 7-month-old infants avoided looking at the inner features of a face for a longer time, when facing threat related emotions compared to neutral, positive, or sad emotions. That is, the infants looked less at the area including the eyes, nose and mouth when confronted with an angry or a fearful facial expression.
Recently, there has been an increase in the use of pupil dilation as a measure of physical arousal caused by cognitive or emotion processing. The results are mixed in their support of the negativity bias theory. Both Bradley and colleagues (2008) and Partala and Surakka (2003) found that students’ pupil sizes increased when emotional stimuli were presented compared to neutral stimuli. They used visual and auditory stimuli, respectively. However, they did not find a difference in arousal measured by pupil size between positive and negative emotional stimuli. Hepach and Westermann (2013), on the other hand, found more pupil dilation in 10- and 14-month-old infants while viewing an angry face compared to a happy one. Besides, both age groups showed a higher arousal when an angry actor
performed an incongruent positive act instead of a negative act. But only 14-month-olds showed a heightened arousal when a happy actor performed a negative act. Hepach and Westermann view this development in sensitivity for unexpected negativity as an indication of the development of a negativity bias between 10 and 14 months of age.
Vaish and colleagues (2008) describe several hypotheses about the development and functions of the negativity bias. One of them is that, since young infants are more used to positive than negative emotions, infants are more affected by negative emotional expressions because they are unfamiliar and therefore unexpected. This idea builds on the observation of a predominantly positive environment in the first months of an infant’s life. Another hypothesis
6 is that the negativity bias increases the chances for an infant’s survival. As infants explore their environment, a strong reaction to negative cues is necessary to stay out of situations that may be harmful or otherwise dangerous (Vaish et al., 2008).
Neurological research supports the importance of experiences on emotional processing. Very young infants might be naturally biased to process biologically salient, species specific signals displayed in faces (Hunnius et al., 2011). At the point in development that infants develop independent locomotion and start to explore the broader environment, there is also “a functional maturation of a neural circuitry that involves emotion-related systems” (Leppänen & Nelson, 2012, p. 202). Such maturation and specialization of cortical areas is caused by the selective loss of synapses and neurons, which is determined by experiential factors (Grossman, 2010). Thus, the individual-specific experiences with emotional expressions shape the neural possibilities to process emotions (Hunnius et al., 2011).
For some infants, the environmental influences on emotion processing might be atypical or different. This is the case when one or both parents are depressed or anxious. Depressed mothers were found to interact differently with their infants than non-depressed mothers did. They interacted less with their infant, were less sensitive and responsive, they touched, talked to or smiled at their infants less often and shared less moments of joint attention (Gross, 2011). They showed more negative and less positive emotional expressions towards their children, too (LaFreniere, 2012). Some depressed mothers became physically intrusive toward their children, instead of withdrawn (Gross, 2011).
Growing up with a mother who is not very positive and responsive due to depression, might affect an infant’s development in several ways. Infants with ‘normal’ parents become distressed when their parent suddenly stops showing emotional expressions in an experiment called the still-face paradigm. Infants of depressed mothers, however, showed the same behavior and emotional expressions during spontaneous and still-face episodes (LaFreniere, 2012; Gross, 2011). In experiments using EEG measures, infants of depressed mothers had a different pattern of brain activity while processing emotional expressions than infants without a depressed parent (De Haan et al., 2004; Gross, 2011). Infants adapt to the depression in their environment by showing more “depressed” behavior themselves (Gross, 2011). They vocalize less and show more gaze aversion, more protest behavior, a lower activity level and a more negative interaction style than infants of non-depressed mothers (LaFreniere, 2012). Other, non-depressed, adults are affected by this behavior in turn. They show less optimal interactive behavior towards these “depressed” infants, even if they are unaware of the maternal
7 depression (Gross, 2011; LaFreniere, 2012). Thus, most likely, these children are exposed to an environment showing more negative affect and less positive affect.
On the long term, children of depressed mothers have a higher risk than other infants to develop internalizing or externalizing problems or psychopathology. Besides, infants of depressed mothers show more negative and less positive affect and behavior. The effect sizes indicating these risks are stronger for young children than for older children with a depressed parent (Goodman et al., 2011).
Little is known about the influence of maternal or paternal anxiety on the emotional development of children (Bögels & Phares, 2008; Weinberg, Beeghly, Olson, & Tronick, 2008), even though parental anxiety seems to have strong effects on development (Weinberg & Tronick, 1998; Weinberg et al., 2008). The research that exists mainly concerns parents with post traumatic stress disorder (PTSD) or panic disorder (PD). Children of parents with PD seem to be more sensitive for scary things, such as evocative faces or air puffs, than children of parents with a major depressive disorder or no psychiatric disorder (Pine et al., 2005). Behavioral inhibition is more prevalent in children of parents with PD, too, which is a risk factor for developing anxiety disorders (Weinberg & Tronick, 1998). Infants of mothers with PD show a higher level of cortisol (Pine et al, 2005), a hormone associated with
psychological stressors (Dickerson & Kemeny, 2004).
Compared to children of parents with no anxiety or depressive disorder, children of parents with PD and PTSD have a higher risk to develop an anxiety disorder themselves (Pine et al., 2005; Bögels & Phares, 2008; Weinberg & Tronick, 1998). An anxious rearing style is associated with the severity of internalizing symptoms in children, especially with anxiety symptoms (Roelofs, Meesters, Ter Huurne, Bamelis & Muris, 2006). When watching photos of evocative faces, parental PD, more than the child’s own anxious symptoms, predicted greater attention allocation and higher levels of fear in the children (Pine et al, 2005).
Weinberg and Tronic (1998) found that infants of mothers who were in treatment for PD or a major depressive disorder, were more sad, angry, fussy and crying in the still-face experiment or when a stranger arrived, than control infants. However, in a replication study with a larger sample (Weinberg et al., 2008), no significant differences were found between the groups. They suggest that it may not be the disorder itself but the mothers interactive behavior and the symptoms she expresses, that influence the infant’s development (Weinberg et al., 2008). So, there is evidence suggesting anxiety in parents influences infants’ emotional development. However, research is scarce and inconclusive about what this influence precisely is.
8 Vaish and colleagues (2008) suggest that the depression of a parent may influence the development of emotional processing as described in their theory about the negativity bias. They do not describe any hypothesis about the influence of other psychiatric disorders, such as anxiety disorders. Because infants of depressed parents might not be surrounded by as many positive emotions in the first months of their lives as children of non-depressed parents, their psychological reference point might not become that much positively skewed. As a consequence, they might not develop a negativity bias, which might result in a maladaptive development. According to Vaish and colleagues, the possible lack of a negativity bias in infants of depressed mothers might dramatically influence children’s development even beyond the emotional domain. If infants lack a positively skewed evaluation, they might be preoccupied with negative events more often than infants of non-depressed mothers (Vaish et al., 2008). And after the negative event has passed, they might not minimize the negative state to restore a mildly positive state (Frederickson & Levenson, 1998; Taylor, 1991; Vaish et al., 2008). This might interfere with the formation and maintenance of social bonds. It might result in a lower sense of well-being or in a depression in the children themselves (Taylor, 1991). Learning and behavior might be influenced, because the critical signal function of the negativity bias will not be served (Vaish et al., 2008). So, understanding the role of parental depression in the development of a negativity bias provides information about infants’ emotional development, but has implications for other domains as well.
Maternal emotional disposition affects infants’ processing of facial expressions of emotion in typical populations as well, as De Haan and colleagues (2004) suggest. The 7-month-olds in their study generally looked longer at fearful than at happy faces. Children of mothers with a high score on positive affect showed this looking pattern, while infants whose mother scored low on positive affect did not. However, this difference approached but did not reach significance. Grossman (2010) supports the statement that maternal expressions of emotion play a prominent role in infants’ developing understanding of emotions. He reviews studies that show that 3-month-olds can differentiate their mothers’ emotional expressions, but not a stranger’s expressions (e.g. Kahana-Kalman & Walker-Andrews, 2001). He suggests this may be the case because the mother’s expressions of emotion are more likely to have consequences that affect the infant.
A note should be made on the literature described above. The majority of research about infant development is conducted with mothers instead of fathers or parents in general. Gross (2011) suggests that the relative neglect of fathers in research is because fathers would
9 be less involved with or around their infant than mothers. This explanation seems valid in the Netherlands. In 81,9% of the Dutch families that consist of father, mother and child(ren), fathers work more hours per week than mothers (Korvorst & Traag, 2010). Besides, 74% of Dutch mothers of 1-year-old infants work less than 27 hours per week (Centraal Bureau voor de Statistiek, 2013). However, this difference in physical presence does not mean that fathers are not present at all or that fathers do not influence their children’s lives and development. To illustrate the findings confirming this idea, Wilson and Durbin (2010) found that there is no significant difference in the size of the correlation between parental depression and parenting behavior for fathers and mothers. Besides, Bögels and Phares (2008) found that fathers might influence their child’s development in a different, sometimes more significant, way than do mothers.
New research is needed to further test the theory offered by Vaish and colleagues (2008). More specifically, it is unclear if and to what extent the negativity bias exists for different facial expressions of negative emotions (i.e. sadness, fear, and anger). Besides, it is unclear if the amount of positivity and negativity in the infants’ environment does
significantly influence the development of a negativity bias in emotion processing. Answering this question has important implications. It might reveal early risk- and protective factors for infants of depressed or anxious parents. Knowing whether infants of depressed or anxious parents are at risk for developmental problems, and being able to identify this risk by
signaling the absence of a negativity bias, might help to intervene early and prevent problems. To be able to judge the infant’s situation accurately, it is important to know the influence of maternal versus paternal depressive symptoms. Therefore, the central question in this research is: Do fathers’ and mothers’ depressive or anxious symptoms influence their infants’
processing of facial expressions of happy, fearful, angry, and sad emotions?
To specify the different elements that are necessary to answer the central question, five sub-questions have been formulated:
1. Do 14-month-old infants show a negativity bias in emotion processing, that is, do infants show more pupil dilation while watching faces with negative emotion expressions than while watching faces with positive and neutral emotion expressions?
2. Do higher levels of exposure to parents with depressive symptoms alter infants’ pupil responses to negative emotions?
10 infants’ pupil responses to emotion?
4. Do higher levels of exposure to parents with anxious symptoms alter infants’ pupil responses to negative emotions?
5. Is there a difference in the influence of maternal and paternal anxious symptoms on infants’ pupil responses to emotion?
Depressive symptoms means that parents show a negative or seemingly neutral mood, low levels of energy and concentration, little decisiveness, little positive affect, or feelings of worthlessness or despair (Gross, 2011). Anxious symptoms means that parents show a fear of a specific object, animal or situation, of social evaluative situations, of places with many people, of separation from their loved ones, or show other symptoms described in the DSM-IV-TR1 anxiety disorders section (Bögels & Van Melick, 2004; American Psychiatric
Association, 2007). Depressive and anxious symptoms were measured on a continuous scale. Hypotheses are that a negativity bias is present in 14-month-old infants, because negative emotions serve an important signal function for them. Based on the studies by Hepach and Westermann (2013), Bradley et al (2008), and Partala and Surakka (2003), a negativity bias is hypothesized to be a state of heightened arousal shown by bigger pupil sizes when negative emotions are expressed.
It is hypothesized that both maternal and paternal depressive symptoms result in a reduced development of the negativity bias in infants, compared to infants with positive parents. So, infants with at least one parent with depressive symptoms will show less pupil dilation while looking at faces with negative, compared to neutral and positive, emotion expressions than infants of parents without those symptoms.
Whether parental anxiety has a similar or different influence on emotion processing compared to depression, whether there is a difference in the negativity bias for different emotions, and whether there is a difference in the influence of maternal and paternal depressive or anxious symptoms on infant emotion processing, remains unclear in the literature. Therefore, these relations will be explored.
11
Method Participants
Participants in this study were 63 infants, who were 14 months old (M = 14 months; 8 days,
SD = 20 days). The personal information of three infants were missing. Of the remaining 60
infants, 62% were boys and 38% were girls. 58% of the infants was able to walk
independently. Mothers were between 26 and 45 years old (M = 34.4, SD = 4.2), fathers were between 28 and 54 years old (M = 37.9, SD = 5.5). 75% of the mothers finished university, 17% finished higher professional education (HBO), 3% finished secondary professional education (MBO), and 5% something else. For the fathers, 56% finished university, 25% finished HBO, 5% finished MBO, 10.5% finished high school, and 3.5% something else. The sample included relatively more high educated parents than the overall Dutch society and no parents in their early twenties. The distribution of work approached the average Dutch family: 63% of mothers was working part-time, 19% was working full-time, 10% was a housewife, the others were involuntarily not working. For the fathers, 63% was working full-time, 26% was working part-time, 5% was unemployed, and 5% did something else. There were no stay-at-home dads in the sample.
To recruit participants for this study, the birth records of the municipality of
Amsterdam have been consulted. When the babies were six months old, the parents received a flyer asking to participate in research on the development of children. From the database of families who responded, parents of 13- and 14-month-olds were randomly chosen and called with an explanation of the study and the request to participate. The parents provided informed consent prior to participation and received a 10 euro compensation for their participation.
Instruments
The eye-tracker and stimuli. The pupil size data were collected with a Tobii T120
eye-tracker (Tobii Technology, 2011). The eye-tracker looked like a flat-screen monitor attached to an adjustable arm. The resolution of this 17 inch monitor was 1280x1024 pixels. On the lower side of the monitor was an integrated eye tracker. For each eye, 120 pupil size data points were collected per second. With the adjustable arm the screen could be placed exactly at 60 centimeters straight in front of the child. This was necessary for the proper recognition of the child’s eyes by the eye-tracker.
The stimuli presented to the infants were black-and-white photos of a woman looking neutral and expressing basic emotions: happy, sad, angry and fear. The photos were validated
12 by students of the University of Amsterdam. The students had to label the emotions expressed in the photo’s and rated the intensity and naturality of the expressions on a 7-point scale. The percentages for correctly labeling the emotions were 97.5% for the happy face, 95% for the angry face, 95% for the fearful face and 97.5% for the sad face. They thought, as is predicted in adult psychology (e.g. Peeters & Czapinski, 1990; Huang & Luo, 2006; Öhman & Mineka, 2001), that the negative emotions were more intense than the positive one2. Previous research has shown that infants show an advantage in processing female faces (Hunnius et al., 2001) rather than male faces. In this study, facial expressions were from a female stranger.
At the start of the experiment, each photo was shown once in an order that is the same for each infant: neutral, happy, fear, angry, sad respectively. This part of the experiment was referred to as ‘the fixed order part’. It was followed immediately by the random order part, in which each photo was shown four times in a random order. A small drawn duck that was shaking from left to right and was making a sound was added to the stimuli, in order to draw the infant’s attention to the screen between the display of each photo.
Depressive symptoms. Beck’s Depression Inventory (BDI; Beck, Steer & Brown,
1993) is a 21 item questionnaire about adult depression. Both parents indicated the amount of depressive symptoms they experienced, by answering questions using a four point Likert-scale. An example of a question in the English version is: “Pessimism: (0) I am not
discouraged about my future. (1) I feel more discouraged about my future than I used to be. (2) I do not expect things to work out for me. (3) I feel my future is hopeless and will only get worse.” The Dutch translation of the BDI by Van der Does (2002) is used in this study. The internal consistence of the scale was high: Chronbach’s α = .90 for fathers and α = .79 for mothers. Previous research shows that the test-retest reliability was high, too (r = .82). The BDI’s correlations with other scales measuring depression were .81, .79, and .85, indicating a good convergent validity. Persons who were diagnosed with a depression scored significantly higher than people without a depression (COTAN, 2005).
Symptoms of anxiety. The Screen of Child Anxiety and Related Disorders, Adult
version (SCARED-A; Bögels & Van Mellick, 2004) is a 71 item questionnaire about parental anxiety. Both parents indicated the amount of anxiety symptoms they experience, by
answering questions using a three point Likert-scale. The questionnaire is Dutch. An example of a question, translated into English, is: “I am afraid of an animal that most people do not fear.” The internal consistency of this scale was high: Chronbach’s α = .86 for fathers and α = .90 for mothers.
13
Procedure
The experiment took place at the language lab of the University of Amsterdam. The infant and one parent were welcomed in an office with a play area. Here, the researcher talked with the parent about the objectives and procedures of the research, the possibility to quit, and any questions the parent might have. The parent signed informed consent. Then the researcher played with the infant for a while. When the infant felt comfortable, the experiment started in the lab next to the office.
In the lab, the infant was placed in a car seat on a table. The eye-tracker was placed at 60 centimeters in front of the child’s face. The parent could sit behind the child in the same room, so they remained close. The researcher controlled the computers linked to the eye-tracker from a different room. There was a window between the two rooms and a microphone enabled communication. The light in the lab was low and there was a black curtain behind the eye-tracker, to prevent visual distraction and to ensure a good measurement by the eye-tracker. Before the real experiment stared, the gaze of each infant was calibrated. A five-point calibration was used: a square, with moving blocks inside, was successively shown in each corner and at the centre of the screen. The calibration was finished when the eye-tracker has captured the infant’s gaze at each calibration point. Then, the photos of the stranger woman with a neutral, happy, surprised, sad, fearful, or angry facial expression were shown to the infant. The photos were shown in the order described previously. Before each subsequent photo, the wiggling duck was shown until the infant’s gaze was at the centre of the screen.
When the experiment was finished, the accompanying parent received an envelope with the questionnaires about their symptoms and their background information. Both parents were asked to fill in these questionnaires at home and to send them back to the university.
Preparing the data for the analysis
The data collected by the eye-tracker and the questionnaires, were transformed into the following variables. The display of an emotion on the screen was transformed from a nominal variable into dummy variables. Each emotion was recoded as individual dummy variable with neutral as reference. The time an infant looked at a picture was measured in milliseconds from the moment an infant first looked at the picture until it stopped looking at that same picture. There was a maximum of 3000 ms for each picture. The time is aggregated by 50 ms via averaging.
14 milliseconds. It was assumed that the infant would have continued looking at the screen during this brief time span, and a technical problem caused the missing data. If there was no data for more than 500 milliseconds, it might have been caused by the infant looking away from the screen. Because of this, these moments were excluded from the analysis. Then the mean size of the left and right eye at each time point was calculated. Finally, the time points that had the same value for time after rounding by 50 ms were taken together per picture, and the mean of the pupil sizes at these time points was calculated. The pupil sizes of the infants during the display of the duck or a black screen were not included in the analysis.
Infants may have shown a variation in their pupil sizes while looking at a picture, regardless of the content of that picture. For example, when a grayscale picture was displayed after the brief display of a black screen, there was a sudden increase in illumination. The brightness of this screen would cause the pupils of the child to constrict at first, before showing the response to the emotion displayed (Bradley et al., 2008). To reduce the interference of such influences with the outcomes of the analysis, pupil responses were baselined to the neutral face by subtracting the response to the neutral face from the response to the face that was shown at the corresponding time point. The resulting variable showed whether pupils became bigger or smaller while looking at an emotion compared to neutral, and was used as an index of pupil reactivity in the current study.
Maternal and paternal depressive and anxious symptoms were calculated by averaging parents’ scores on Beck’s Depression Inventory and the SCARED-A, respectively. If more than half of the items on a scale were missing, no mean score was calculated. The final score on the BDI could be between 0 and 3, the final score on the SCARED-A could be between 0 and 2.
So, the variables used in the analysis are: The difference in pupil size watching emotions compared to neutral faces, a dummy for each emotion, and depressive symptoms and anxious symptoms for both fathers and mothers. Data from the fixed order and random order part of the experiment were analyzed separately. For the random order part, the mean scores of the four displays of each picture were calculated.
Analysis
The dataset resulting from the preparation described above, had the following structure. For each separate picture, time-series data were provided including 60 measurements of pupil size. As a consequence, subsequent data points were probably more related to each other than to
15 other data points. These time-series data were available for 60 different infants. The pupil sizes within each subject were expected to be more related than the pupil sizes between subjects, which made the data hierarchical.
To be able to keep the time series structure and the hierarchical structure in the data, the analysis were performed using multilevel linear models. In these models, infant subject number was the grouping variable, with random intercepts and slopes. Besides, the
autocorrelation between the subsequent measurements of pupil reactivity was taken into account, by using a first-order autoregressive covariance matrix for time in the model. This was entered in the model as a random effect. So, the hierarchy of the model contains two levels. The measurements at the different time points and the questionnaire data were level one variables, infant subject number was the level two variable. More details on the models tested are described in the results chapter.
The fit of the basic models was compared using χ2 of the change, which was calculated by subtracting the -2 log likelihood (-2LL) information criteria. Whether the change was significant was decided using the χ2 distribution table. The significance of the interactions were evaluated based on t-tests, at a .05 significance level in a two-tailed test.
16
Results Descriptives
This section provides information about the variables used in the analysis, demographic information about the participants was provided in the Methods chapter. The prevalence of symptoms of depression and anxiety in parents is displayed in Table 1. Depressive symptoms of fathers and mothers in the sample did not significantly differ from each other, t (53) = .78,
p = .44. Depressive symptoms within families were not correlated, r (53) = .15, p = .296. The
same applied to the difference in symptoms of anxiety in fathers and mothers, t (53) = -1.58, p = .121, and the correlation of symptoms of anxiety within families, r (53) = -.01, p = .926. The pupil size differences compared to neutral were normally distributed, with
skewness of .69 (SE = .02) and kurtosis of .43 (SE = .04). The time infants looked at a picture was also normally distributed, with skewness of .20 (SE = .02) and kurtosis of -1.07 (SE = .04). The anxious symptoms of parents were normally distributed, with skewness of 1.36 (SE = .33) and kurtosis of 1.98 (SE = .64) for fathers, and skewness of 1.00 (SE = .32) and kurtosis of .51 (SE = .63) for mothers. Depressive symptoms of mothers were normally distributed, with skewness of .68 (SE = .32) and kurtosis of -.06 (SE = .63). Depressive symptoms of fathers were non-normally distributed, with skewness of 1.59 (SE = .32) and kurtosis of .276 (SE = .63). There were two outliers here, fathers who scored higher than 1 on the BDI questionnaire. They were not excluded from the analysis.
Table 1. The presence of symptoms of depression and anxiety in the sample.
Fathers Mothers
Variable M (SD) Min.-Max. M (SD) Min.-Max.
Depressive symptoms
BDI mean score (0-3) .29 (.29) .00 - 1.24 .26 (.20) .00 - .81 BDI total score (0-63) 6.03 (6.02) 0 - 26 5.49 (4.11) 0 – 17 Anxious symptoms
SCARED-A mean score (0-2) .19 (.14) .01 - .61 .23 (.15) .03 - .66 SCARED-A total score (0-142) 13.09 (9.83) 1 – 43 16.56 (10.67) 2 – 47 Note. Parentheses in the first column contain the minimum and maximum score that are possible for the variable. The Min.-Max. columns show the minimum and maximum scores of the parents in the sample.
Analysis of a basic model with the main effect of time: fixed order data
For this analysis the pupil size difference was used as the dependent variable. First, a main effects model was tested including subject as a grouping variable, time as a fixed linear effect and the first order auto regression (AR1) of time as an effect with random intercepts and
17 slopes. The fixed linear effect of time was not significant, B = -.00, SE = .02, p = .999. The fixed intercept was significant, B = .23, SE = 02, p < .001. The fit of the model to the data was -2LL = 212.96, AIC = 222.96, BIC = 260.93, df = 5.
The shape of the graphs in Figure 1 indicates there might have been a quadratic effect of time rather than a linear one. Adding time2 showed a significant fixed effect (B = .00, SE = .00, p < .001) and a significant improvement of the model, χ2(1) = 63.69, p < .01. So, this effect was included in the model.
Figure 1. The mean pupil sizes of infants while looking at the different photos, baselined by
neutral. The left graph displays the fixed order data, the right graph displays the random order data.
Table 2. Model including emotion to predict the pupil size in the fixed order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept -.04 .02 -.08 -.00 .041 Time .00 .02 -.04 .04 .998 Time2 .00 .00 .00 .00 <.001 Happy .21 .01 .20 .22 <.001 Fear .27 .01 .26 .28 <.001 Angry .36 .01 .35 .36 <.001 Sad .41 .01 .40 .42 <.001 Random intercept +
time (AR1 diagonal) .02 .00 .02 .03 <.001
Residual .03 .00 .03 .03 <.001
18 Table 3. Comparing papillary reactivity towards negative and neutral emotional expressions
to happy facial expressions, fixed order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept .17 .02 .13 .21 .041 Time .00 .02 -.04 .04 .998 Time2 .00 .00 .00 .00 <.001 Neutral -.21 .01 -.22 -.20 <.001 Fear .06 .01 .05 .07 <.001 Angry .14 .01 .13 .15 <.001 Sad .20 .01 .19 .21 <.001 Random intercept +
time (AR1 diagonal) .02 .00 .02 .03 <.001
Residual .03 .00 .03 .03 <.001
Note. The dependent variable is the difference in pupil size compared to neutral emotion display. The reference of the dummy variables of emotion is happy.
Testing the effect of emotion and the presence of a negativity bias: fixed order data
The main effect of emotion was significant in this part of the data (p < .001 for all emotions). Watching a photo of an emotional expression resulted in more pupil dilation than watching a neutral expression, see Table 2 and the left graph in Figure 1. The inclusion of emotion significantly improved the model, χ2(4) = 6953.02, p <.01.
To test negativity bias in infants’ pupil responses to emotion expressions, the same model was repeated with happy instead of neutral as a reference. If there was a negativity bias in infants’ pupil responses, then infants would show more pupil dilation while watching faces with negative emotion expressions than while watching faces with positive and neutral emotion expressions. So, to accept this hypothesis, (1) the pupil sizes while watching happy and neutral faces should not have differed from each other and, (2) the pupil sizes while watching negative emotions should have been significantly bigger than while watching the happy face. Results displayed in Table 2 and 3 did not support the first hypothesis since the pupil reactivity while watching a happy face was significantly higher then while watching a neutral face. The second hypothesis was confirmed by the results in Table 3. Fearful, angry and sad faces did cause successively more pupil reactivity than happy faces.
19 Table 4. The influence of parental depression on emotion processing, fixed order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept -.01 .04 -.09 .01 .686 Time .00 .02 -.04 .04 .998 Time2 .00 .00 .00 .00 <.001 Happy .21 .01 .20 .23 <.001 Fear .33 .01 .32 .35 <.001 Angry .44 .01 .42 .46 <.001 Sad .41 .01 .39 .43 <.001 Paternal depression -.08 .07 -.21 .05 .213 Maternal depression -.03 .10 -.23 .17 .747 Happy*Paternal depression -.16 .02 -.19 -.13 <.001 Fear*Paternal depression -.26 .02 -.29 -.23 <.001 Angry*Paternal depression -.21 .02 -.24 -.18 <.001 Sad*Paternal depression -.13 .02 -.16 -.09 <.001 Happy*Maternal depression .12 .03 .07 .17 <.001 Fear*Maternal depression -.00 .03 -.06 .05 .880 Angry*Maternal depression -.16 .03 -.22 -.11 <.001 Sad*Maternal depression .06 .03 .00 .12 .035
Random intercept + time
(AR1 diagonal) .02 .00 .01 .02 <.001
Residual .03 .00 .03 .03 <.001
Note. The dependent variable is the difference in pupil size compared to neutral emotion display.
The influence of parental depression on emotion processing: fixed order data
New variables were added to the model described in Table 2, which is from now on referred to as the ‘basic emotion model’. The main effect of maternal depression was not significant, B = .02, SE = .11, p = .845. After adding this main effect, interactions of depression and
emotion were added to the model. Maternal depression did predict increased reactivity in infants pupil responses to happy faces (B = .08, SE = .03, p = .004) and decreased reactivity towards angry faces (B = -.17, SE = .03, p < .001). It did not significantly predict infants’ reactivity while watching a fearful or sad expression (p = .392 for fear*maternal depression and p = .313 for sad*maternal depression).
Depression of fathers did have a significant main effect on pupil size, B = -.24, SE = .07, p < .001. Adding interactions to the model showed that the decrease in reactivity due to
20 paternal depression depended on the emotion infants watched: B = -.15, SE = .02, p < .001 for happy*paternal depression; B = -.26, SE = .02, p < .001 for fear*paternal depression; B = -.22,
SE = .02, p < .001 for angry*paternal depression; and B = -.12, SE = .02, p < .001 for sad*
paternal depression.
Then the main effects and the interactions between emotion and parental depression for both fathers and mothers were added in one model. The result was a model in which all interactions, except for fear*maternal depression (p = .880), were significant predictors of pupil reactivity. This is displayed in Table 4. Paternal depression predicted lower levels of arousal while watching emotional expressions. Maternal depression predicted higher levels of arousal while watching happy or sad faces, and lower levels of arousal while watching angry faces, compared to infants with no depressive mother.
Table 5. The influence of parental anxiety on emotion processing, fixed order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept -.03 .04 -.12 .05 .430 Time .00 .02 -.04 .04 .998 Time2 .00 .00 .00 .00 <.001 Happy .21 .01 .19 .23 <.001 Fear .24 .01 .22 .26 <.001 Angry .30 .01 .27 .32 <.001 Sad .32 .01 .30 .34 <.001 Paternal anxiety -.14 .13 -.41 .12 .279 Maternal anxiety .07 .12 -.16 .31 .542 Happy*Paternal anxiety -.28 .03 -.34 -.21 <.001 Fear*Paternal anxiety -.42 .03 -.48 -.35 <.001 Angry*Paternal anxiety -.40 .04 -.47 -.32 <.001 Sad*Paternal anxiety -.29 .04 -.36 -.21 <.001 Happy*Maternal anxiety .16 .03 .10 .23 <.001 Fear*Maternal anxiety .41 .03 .35 .47 <.001 Angry*Maternal anxiety .49 .03 .43 .55 <.001 Sad*Maternal anxiety .54 .03 .47 .61 <.001
Random intercept + time
(AR1 diagonal) .02 .00 .01 .02 <.001
Residual .03 .00 .03 .03 <.001
Note. The dependent variable is the difference in pupil size compared to neutral emotion display.
The influence of parental anxiety on emotion processing: fixed order data
The analyses of the influence of parental anxiety were performed in a way similar to the analyses about depression. Maternal anxiety was a significant main effect in predicting pupil
21 reactivity (B = .39, SE = .14, p = .005), predicting more arousal in children of mothers who were anxious. Adding the interactions of maternal anxiety and emotion to this model, showed that infants of anxious mothers were significantly more aroused by each emotion (B = .17, SE = .03, p < .001, for happy; B = .41, SE = .03, p < .001, for fear; B = .49, SE = .03, p < .001, for angry; and B = .53, SE = .03, p < .001, for sad).
When the main effect of paternal anxiety on pupil size was added to the basic emotion model, this was significantly predicting a decrease in infant arousal, B = -.43, SE = .14, p = .003. Adding the interactions of paternal anxiety and emotion to this model, showed that all interactions significantly predicted a decrease in infants’ pupil reactivity (B = -.29, SE = .04, p < .001, for happy; B = -.44, SE = .04, p < .001, for fear; B = -.41, SE = .04, p < .001, for angry; and B = -.31, SE = .04, p < .001, for sad).
After creating a model in which the main effects of anxiety and the interactions between the emotion dummies and anxiety of both parents were added, all eight interactions were still significant. Maternal anxiety was still increasing infants’ reactivity towards the emotions and paternal anxiety was still decreasing it. This model is displayed in Table 5.
Analysis of a basic model with the main effect of time: random order data
The analyses of the random order data were similar to the analyses of the fixed order data. The fixed linear effect of time was not significant, B = -.000, SE = .01, p = .996. The fixed intercept was significant, B = .02, SE = .01, p =.002. The fit of the model to the data was -2LL = -13454.77, AIC = -13444.77, BIC = -13406.90, df= 5.
Adding time2 showed a significant fixed effect (B = .00, SE = .00, p < .001) and a significant improvement of the model, χ2(1) = 48.22, p <.01. So, this was included in the model.
Testing the effect of emotion and the presence of a negativity bias: random order data
The model including the display of the different emotions, made a better prediction of the infants’ pupil sizes than the model without the emotions, χ2
(4) = 816.69, p < .01. Watching a happy face predicted more arousal in infants than watching a neutral face (p < .001).
Watching a fearful or angry face predicted less arousal than a neutral face (p < .001). Watching a sad face predicted no significant difference in arousal compared to watching a neutral face (p = .230). This is displayed in Table 6 and the right graph in Figure 1. The results of this part of the data rejected both assumptions of the negativity bias,
22 which are described in the fixed order section. Contrary to the first hypothesis, results in Table 6 and 7 show that watching a happy face predicted a significant increase in arousal compared to watching a neutral face. In contrast to the second hypothesis, results in Table 7 show significantly less pupil reactivity during the display of negative emotions compared to the positive one. Besides, Table 6 shows that watching a negative face resulted in the same (for sad) or less arousal than watching a neutral face.
Table 6. Model including emotion to predict the pupil size in the random order data.
Parameter B
Standard Error of B
95% Confidence Interval
p
Lower limit Upper limit
Intercept .02 .01 .01 .04 .007 Time .00 .01 -.01 .01 1.000 Time2 .00 .00 .00 .00 <.001 Happy .02 .00 .01 .03 <.001 Fear -.08 .00 -.08 -.07 <.001 Angry -.02 .00 -.03 -.01 <.001 Sad .00 .00 -.00 .01 .230 Random intercept +
time (AR1 diagonal) .00 .00 .00 .00 <.001
Residual .02 .00 .02 .02 <.001
Note. The dependent variable is the difference in pupil size compared to neutral emotion display.
Table 7. Comparing papillary reactivity towards negative and neutral emotional expressions
to happy facial expressions, random order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept .04 .01 .03 .06 <.001 Time2 .00 .00 .00 .00 <.001 Neutral -.02 .00 -.03 -.01 <.001 Fear -.10 .00 -.11 -.09 <.001 Angry -.04 .00 -.05 -.03 <.001 Sad -.02 .00 -.02 -.01 <.001 Random intercept +
time (AR1 diagonal) .00 .00 .00 .00 <.001
Residual .02 .00 .02 .02 <.001
Note. The dependent variable is the difference in pupil size compared to the neutral emotion display. The reference of the dummy variables of emotion is happy.
23 Table 8. The effect of maternal depression on emotion processing, random order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept .02 .01 -.01 .04 .223 Time .00 .01 -.01 .01 .999 Time2 .00 .00 .00 .00 <.001 Happy -.00 .01 -.02 .01 .467 Fear -.09 .01 -.10 -.08 <.001 Angry -.05 .01 -.06 -.03 <.001 Sad .00 .01 -.01 .02 .522 Maternal depression .04 .04 -.05 .12 .368 Happy*Maternal depression .07 .02 .03 .11 .001 Fear*Maternal depression .04 .02 .00 .08 .038 Angry*Maternal depression .12 .02 .08 .16 <.001 Sad*Maternal depression -.04 .02 -.08 .00 .066
Random intercept + time
(AR1 diagonal) .00 .00 .00 .00 <.001
Residual .02 .00 .02 .02 <.001
Note. The dependent variable is the difference in pupil size compared to neutral emotion display.
Table 9. The effect of paternal depression on emotion processing, random order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept .04 .01 .02 .07 .001 Time .00 .01 -.02 .02 .998 Time2 .00 .00 .00 .00 <.001 Happy -.01 .01 -.02 .00 .191 Fear -.10 .01 -.11 -.09 <.001 Angry -.05 .01 -.06 -.04 <.001 Sad -.02 .01 -.03 -.01 .001 Paternal depression -.04 .03 -.09 .02 .209 Happy*Paternal depression .06 .01 .04 .09 <.001 Fear*Paternal depression .07 .01 .04 .09 <.001 Angry*Paternal depression .08 .01 .06 .11 <.001 Sad*Paternal depression .05 .01 .02 .08 <.001
Random intercept + time
(AR1 diagonal) .00 .00 .00 .00 <.001
Residual .02 .00 .02 .02 <.001
24
The influence of parental depression on emotion processing: random order data
To analyze the influence of parental depression on infant emotion processing in this part of the data, new variables were added to the basic emotion model in Table 6. There was no significant main effect of maternal depression on infants’ pupil reactivity during the
experiment, B = .08, SE = .04, p = .054. Then, the interactions between maternal depression and emotion were added. Maternal depression significantly interacted with watching a happy, fearful, or angry face, but not a sad face (see Table 8). Infants of mothers with depressive symptoms were more reactive towards these emotions than the other infants.
Depression of fathers was no significant main effect, either, B = .02, SE = .03, p = .554. The interactions between paternal depression and emotion were all significant (see Table 9), predicting more arousal in infants with depressive fathers than in infants with non-depressive fathers.
Putting the interactions of fathers’ and mothers’ depression with emotion in the same model, strongly decreased the significance of the predicted influence of maternal depression. Only the interaction between maternal depression and the display of the happy face remained significant (B = .08, SE = .02, p = .001). The interactions with watching a fearful and angry expression were no longer significant (p = .133 and p = .073, respectively) and the interaction with watching a sad expression became even less significant (p = .280). The interactions between paternal depression and emotion did not change a lot (B = .06, SE = .01, p < .001, for happy; B = .07, SE = .01, p < .001, for fear; B = .08, SE = .01, p < .001, for angry; and B = .06,
SE = .01, p < .001, for sad). The presence of depressive symptoms in fathers did not explain
the decrease in the significance of the predictions made by maternal depression. As the symptoms within families were not correlated, depressive symptoms in fathers did not predict the part of the variance maternal depression would have predicted. Due to this, analyzing depression of fathers and mothers separately seemed to result in a better prediction of their infants emotion processing than putting them in the same model.
The influence of parental anxiety on emotion processing: random order data
The main effect of maternal anxiety on infants’ pupil reactivity while watching facial expressions, was not significant, B = -.01, SE = .053, p = .857. When including the interactions of maternal anxiety and emotion, maternal anxiety predicted a significant decrease in pupil reactivity during the display of fearful faces (B = -.06, SE = .03, p = .031) and a significant increase in reactivity during the display of angry faces (B = .06, SE = .03, p = .016). The other interactions were not significant (for happy*maternal anxiety, p = .828; for
25 sad*maternal anxiety, p = .460).
The main effect of paternal anxiety on infants’ pupil reactivity, was not significant either, B = -.02, SE = .06, p = .726. The interactions of paternal anxiety with the different emotions significantly predicted more pupil reactivity towards happiness, fear and anger (B = .07, SE = .03, p = .020, for happy; B = .15, SE = .03, p < .001, for fear; B =.08, SE = .03, p = .007, for angry) and less reactivity towards sadness (B = -.12, SE = .03, p < .001).
When the interactions of both maternal and paternal anxiety with the emotions were added to the basic emotion model, the same interactions as the ones above were significant and the directions of their influences were similar, see Table 10. So, infants of fathers with anxious symptoms, compared to infants of fathers with no anxious symptoms, showed more arousal while watching a happy, fearful or angry face and less arousal while watching a sad face. Infants of anxious mothers showed more arousal while watching an angry face and less arousal while watching a fearful face. The arousal caused by watching a happy or sad face was similar for infants of mothers with and without anxious symptoms.
Table 10. The effect of parental anxiety on emotion processing, random order data.
Parameter B Standard Error
95% Confidence Interval
p
Lower limit Upper limit
Intercept .04 .02 .00 .08 .040 Time .00 .01 -.01 .01 .999 Time2 .00 .00 .00 .00 <.001 Happy -.00 .01 -.02 .02 .925 Fear -.09 .01 -.11 -.07 <.001 Angry -.05 .01 -.07 -.03 <.001 Sad .01 .01 -.00 .03 .115 Paternal anxiety -.06 .06 -.18 .06 .345 Maternal anxiety -.01 .05 -.11 .10 .905 Happy*Paternal anxiety .07 .03 .01 .13 .020 Fear*Paternal anxiety .15 .03 .09 .21 <.001 Angry*Paternal anxiety .08 .03 .03 .14 .005 Sad*Paternal anxiety -.12 .03 -.18 -.06 <.001 Happy*Maternal anxiety .00 .03 -.05 .05 .999 Fear*Maternal anxiety -.06 .03 -.11 -.01 .018 Angry*Maternal anxiety .06 .03 .01 .11 .021 Sad*Maternal anxiety .01 .03 -.04 .06 .637
Random intercept + time
(AR1 diagonal) .00 .00 .00 .00 <.001
Residual .02 .00 .02 .02 <.001
26
Discussion
To assess whether fathers’ and mothers’ depressive or anxious symptoms influenced their infants’ processing of facial expressions of happy, fearful angry, and sad emotions, several sub questions have been analyzed. First, the presence of a negativity bias in 14-month-old infants was addressed. The hypothesis, based on the theory by Vaish and colleagues (2008) was that infants would show more pupil dilation while watching faces with negative emotion expressions than while watching faces with positive and neutral emotion expressions.
Results showed that the display of emotion expressions affected infants pupil dilation. Watching a photo of a happy face consistently resulted in more arousal than watching a neutral face. This finding rejected the statement of Vaish and colleagues that, when neutral was added as a condition in research on emotion, a positive condition no longer significantly affected infants’ reactions. The result that the positive condition caused a stronger reaction in infants than the neutral condition, was consistent with research by Bradley and colleagues (2008), Geangu, Hauf, Bhardwaj, and Bentz (2011), and Partalla and Surakka (2003). The results were inconsistent about the influence of negative emotions on infants’ reactivity. In the first part of the experiment, all infants watched each emotion once, in the order neutral, happy, fear, angry, sad. In this fixed order part, infants showed more pupil dilation towards the negative emotions compared to the happy and neutral expressions. In the second part of the experiment, all photos were shown four times in a random order. In this random order part, infants showed less pupil dilation towards the negative expressions
compared to the happy expressions. The arousal caused by watching a sad face was similar to watching a neutral face. Watching an angry or a fearful face caused even less arousal than watching a neutral face. This finding in the random order data is contradictive to the results in most previous research, which suggest stronger responses towards negative than positive or neutral expressions (e.g. Vaish et al, 2008; Bradley et al., 2008; Partala & Surakka, 2003; Hepach & Westermann, 2013; Geangu et al., 2011). Hunnius and colleagues (2011) did find that reactions towards sad and neutral faces were similar. However, reactions to positive faces did not differ from these two conditions. Besides, their research showed reduced looking towards the eyes of angry and fearful faces. They interpreted this result as an avoidant pattern towards these threat related expressions, suggesting increased instead of reduced arousal. Two types of effects might have influenced the reactivity of the infants towards the negative emotions. On the one hand, there is habituation. This means that, when a stimulus has been presented several times, infants gradually lose their interest in that stimulus. But,
27 when that stimulus is replaced by a different one, the infant regains it interest and pays close visual attention to this new stimulus (Gross, 2011). In the present experiment, each infant viewed each photo 5 times, resulting in the presentation of 25 photos. During the experiment, infants watched negative emotions 60 percent of the time, neutral emotions 20 percent of the time and happy emotions 20 percent of the time. None of the emotions were shown for the first time during the random order part, as they were all shown once in the fixed order part. So, the low levels of arousal caused by negative emotions in the random order part of the
experiment, might be caused by a gradual decrease in attention and arousal. A display of the happy face was relatively new to the infants. This might have caused new interest and as a consequence more arousal.
On the other hand, the order of presentation of the photo’s in the first, fixed order, part of the experiment, might have influenced the infants’ responses in this part. Perhaps the growth in arousal towards positive and then negative emotions compared to neutral, was not or not fully caused by the emotional valence of the display. It might have been caused by the fact that each emotion expression was a new stimulus. Or the emotional valence of the previous emotions might have played a role. Previous research is inconclusive on this last matter though, as Schupp, Schmälzle, Flaisch, Weike and Hamm (2013) describe. Some authors found that an emotionally congruent context caused a stronger response towards a stimulus, others found this caused a weaker response. And, contrary to their own expectations, Schupp and colleagues found that the emotional valence of the context did not influence students responses towards a specific emotional stimulus, much (Shupp et al., 2013). So, based on the present experiment and previous research, one cannot conclude whether the results of the fixed order part or the random order part provide the most reliable description of infants’ processing of negative emotions. New research that systematically presents the emotions in different orders, should clarify this. What can be concluded is that watching emotion expressions significantly alters infants arousal. Besides, watching a happy face causes more arousal than watching a neutral face, and watching a happy face causes a different amount of arousal in infants than watching a negative facial expression.
The second and third sub question concerned the influence of parental depression on emotion processing. The question was whether depressive symptoms in parents alter infants’ responses to negative emotions. A lower pupil reactivity in infants of parents with depressive symptoms was hypothesized based on previous research, which was mainly conducted with mothers. So, the other question was whether maternal and paternal depression had similar
28 effects on infant reactivity, or not.
Maternal and paternal depressive symptoms were both significant predictors of infant pupil reactivity during the display of emotion expression. The predictions made by maternal depression were slightly different from the ones made by paternal depression. Paternal depression seemed to be a stronger predictor of infant reactivity than maternal depression. In the fixed order part of the data, paternal depressive symptoms significantly predicted infants’ reactions towards all emotions, while maternal depressive symptoms did not predict the reactions to a fearful face. When analyzing the random order part of the data, maternal depression was a significant predictor when analyzed separately. But when paternal depression was included in the same model, maternal depression was hardly significant. Then, what did they predict? Depressive symptoms in fathers predicted lower pupil reactivity towards all emotions their children watched during the fixed order part of the experiment, compared to infants who’s fathers did not show depressive symptoms. For depressive symptoms in mothers, the same pattern was applicable during the display of an angry expression. However, during the display of happy and sad faces, infants of mothers with depressive symptoms showed stronger reactivity than infants of mothers without those symptoms.
In the random order part of the data, depressive symptoms in fathers predicted an increase in infant reactivity towards emotion expressions. Maternal depressive symptoms predicted more reactivity in infants towards happy, fearful and angry faces. It did not predict a change in reactivity towards sad faces. However, the reliability of this result can be discussed. All infants viewed the photos in a different, random order. In order to identify differences between individuals in a reliable way, it is important that the context of the measurement is the same for every individual. That means that not only the stimuli and the location should be identical, but also the order of presentation (Carlson & Moses, 2001). If the order of
presentation would have affected infants’ pupil reactivity, a difference between individuals might have been a difference between presentation orders. In the fixed order part of the experiment, all infants watched the photos in the same order. So, the measured effect of depressive symptoms is more reliable there.
The pattern of reduced reactivity towards emotion expressions in infants of parents with depressive symptoms, is consistent with previous research and theories. The pattern could be explained by the “depressive” interaction style and activity level infants of
29 more neutral affect themselves and a lower activity level, they would be less aroused by the display of emotions than other infants. This does not explain the more differentiated pattern of reactions predicted by maternal depression, though. Vaish and colleagues (2008) suggest that infants of depressed parents develop a psychological reference point that is not as positively skewed as the reference point of other infants. The consequence would be that the infants show relatively little arousal towards negative emotions compared to positive emotions or neutral faces. The reduced levels of arousal towards negative expressions predicted by
paternal depressive symptoms and the reduced levels of arousal towards the angry expression predicted by maternal depression, could be explained by this theory. Besides, the relatively high arousal caused by the happy expression in infants of mothers with depressive symptoms, could be explained by the theory that this expression is less common to these children (Vaish et al., 2008). Both theories do not explain why infants of mothers with depressive symptoms, were more aroused by watching a sad face, though. Perhaps a sad face, as a symptom of
depression, was an signal for the infants of, for example, reduced responsivity in their mothers. The fourth and fifth sub question concerned the influence of anxious symptoms in parents on infant emotion processing. The questions were whether parental anxious symptoms altered infants reactions towards negative emotions, and whether there was a difference between fathers’ and mothers’ symptoms. These questions were explorative.
Fathers’ and mothers’ anxious symptoms turned out to have a different effect on infants reactivity towards emotions. For the fixed order part of the experiment, the influences of fathers and mothers were the opposite of each other. Infants of anxious fathers were less responsive towards each emotion expression compared to infants of fathers without symptoms of anxiety. Anxious symptoms in mothers, on the other hand, predicted more reactivity
towards emotions in their infants.
In the random order part of the experiment, the influences of parental anxiety were more differentiated. Infants of fathers with anxious symptoms, compared to infants of fathers with no anxious symptoms, showed more pupil dilation while watching a happy, fearful or angry face and less pupil dilation while watching a sad face. Infants of mothers with
symptoms of anxiety showed more pupil dilation while watching an angry face and less pupil dilation while watching a fearful face. The arousal caused by watching a happy or sad face was similar for infants of mothers with and without anxious symptoms. But, as has been explained for depressive symptoms, the results on parental anxiety of the fixed order part of the experiment were more reliable.
30 Bögels and Phares (2008) also described that fathers’ and mothers’ symptoms of anxiety could influence their child’s emotion processing in different ways. They wrote that some evidence showed that social anxiety in fathers, but less so in mothers, made the
relationship between father and child less warm and more conflictual. Relationships which are conflictual and not very warm have been described for infants with depressive parents, too. In their case, the theory predicts a lower activity level and lower attention towards negative emotions, than is seen in infants with no depressive parent (Vaish et al, 2008; Gross, 2011; LaFreniere, 2012). This lower reactivity was also observed in the current experiment for children with anxious fathers. A different finding Bögels and Phares described was that children of fathers, but not mothers, with Post Traumatic Stress Disorder, were more likely to develop this disorder, too. This suggests that they were more responsive to scary things, which contradicts the lower responsiveness predicted by paternal anxiety in the current experiment. This might be due to a low prevalence of PTSD symptoms in fathers in the current sample. Pine and colleagues (2005) found that infants of anxious parents were more responsive towards evocative faces and showed higher levels of cortisol, a hormone
associated with psychological stressors. The current experiment showed a similar effect for mothers with symptoms of anxiety. Unfortunately, Pine and colleagues did not describe the male-female ratio in their sample of anxious parents, due to which it is unclear how valid this comparison is.
A difficulty in interpreting the results of this research and deciding the implications for clinical practice, is the difference in the results between the fixed order and random order parts of the experiment. As is described above, it is not clear whether this difference is due to habituation or an effect based on the order of the presentation of stimuli. Previous research is more in support of the findings in the fixed order part of the data, but it is not conclusive. New research that systematically compares the effects of emotions in different presentation
sequences, should clarify to what extend order of presentation, habituation and emotion itself affect infants pupil reactivity during the display of emotion expressions.
A different point of discussion is the measurement of parental depression and anxiety. The presence of symptoms of depression and anxiety were assumed to affect the behavior of parents towards their infants. A questionnaire about the parental expressiveness of emotions was filled in by the parents to check this assumption. But, as the questionnaire turned out to be not reliable enough, this expressiveness could not be considered. Although an effect of depression and anxiety was found, it might be caused partially by a hereditary predisposition
