Infants’ Pupil Responses to Others’ Emotions:
How Maternal Negative Emotionality can shape Infants’ Attention for Expressions Mae Nuijs
Abstract
Background: Previous research suggests that infants of 12 months and older give more attention to fearful than positive expressions, referred to as negativity bias. The first aim of this study was to investigate whether infants also show a negativity bias in their pupil responses to angry and sad expressions (i.e., larger responses to negative than positive or neutral expressions). The second aim was to investigate whether this bias is related to the degree mothers experience negative emotions. The specificity of this relationship was also investigated; maternal anxiety or depression might be more related to infants’ responses to fearful or sad expressions, respectively, than angry expressions. It was also investigated whether the amount of time mothers spend with their infants affect the degree to which maternal negative emotions affect infants’ negativity bias. Method: We measured pupil responses of 14 month-old infants (M = 14.7, N = 103) to happy, neutral, fearful, angry, and sad expressions. Results: Infants’ pupil responses to negative expressions were not larger than responses to positive or neutral expressions (i.e., no negativity bias). The difference in infants’ pupil responses to negative and positive expressions was related to the degree to which their mothers experienced negative emotions. The more mothers experienced negative emotions, the larger infants’ pupil responses were to negative versus happy expressions, which was a non-specific relationship. The time mothers spend with their infants did not affect this relationship. Conclusion: Although we did not observe a negativity bias in infants’ pupil responses to expressions, a higher level of maternal negative emotionality was related to an increased negativity bias in infants’ pupil responses. These findings support the growing body of evidence suggesting that maternal negative emotionality affects how infants process emotions.
1. Introduction
Reading expressions of others is essential for survival since it enables humans to detect emotional states of others and thereby provides cues on how to behave (Grossmann, 2010). In comparison with adults, infants are even more dependent on using expressions of others to guide their behaviour as they enter most situations for the first time. From an evolutionary perspective, it might be adaptive for infants to give more attention to negative than positive or neutral expressions because negative expressions can signal danger and thereby can help infants to avoid harmful situations (Vaish, Grossmann, & Woodward, 2008). In line with this idea, a recent review on infants’ emotion processing suggested that, with the start of
locomotion (approximately 12 months and older), infants attend more to negative than positive facets of their environment, referred to as negativity bias (Vaish et al., 2008). In particular, it was suggested that infants give more attention to negative than positive expressions.
Consistent with the notion of a negativity bias, both behavioural (e.g., looking times) and event-related potential (ERP) studies provided evidence for a negativity bias when examining infants’ attention for expressions. Behavioural studies suggested infants looked longer at fearful than happy expressions when these expressions were presented simultaneously
(Kotsoni, de Haan, & Johnson, 2001; Peltola, Leppänen, Mäki, & Hietanen, 2009). However, when fearful and happy expressions were presented separately, one of the three studies in this regard (Nelson & Dolgin, 1985), did not find enhanced attention to fearful expressions (de Haan, Belsky, Reid, Volein, & Johnson, 2004; Ludemann & Nelson, 1988). ERP studies also suggested infants show enhanced attention to fearful expressions: Infants showed a larger negative central (Nc), which is thought to reflect attention (Richards, 2003), to fearful than happy expressions (de Haan et al., 2004; Peltola et al., 2009). In sum, both behavioural and
ERP studies suggest infants show a negativity bias for fearful expressions, but possibly only when expressions are presented simultaneously.
Although several studies suggested that infants show a negativity bias for fearful
expressions, it remains unclear whether this negativity bias is specific to fearful expressions or generalizes to other negative expressions. In particular, it remains unclear whether infants’ differential attention to fearful expressions reflects enhanced attention to negative versus positive expressions, or enhanced attention to fearful versus positive expressions. Previous studies on infants’ attention for other negative expressions found contradicting results. For example, infants had a similar amount of fixations on pictures of sad and happy expressions (Field, Pickens, Fox, Gonzalez, & Nawrocki, 1998). Moreover, again contrary to a negativity bias, infants looked longer at happy than angry expressions (Grossmann, Striano, &
Friederici, 2007). Examining infants’ brain responses resulted in a different conclusion: Infants showed a larger Nc to angry than happy expressions, suggesting a negativity bias in infants’ brain responses. The difference in infants’ looking behaviour and brain responses can possibly be explained by the fact that infants already showed an adult-like response to angry expressions. That is, similar to highly anxious adults, infants detect angry expressions as a threatening signal (reflected by the enhanced brain response), but then look away to avoid the threatening expression (i.e., a vigilant-avoidant pattern; e.g., Rohner, 2002). Together, these findings suggest that a negativity bias for fearful expressions might not generalize to other negative expressions. However, a negativity bias for angry expressions might exists when infants’ brain responses instead of looking behaviour are measured.
Given that infants’ emotional development is suggested to be heavily dependent on individual experiences (e.g., Leppänen & Nelson, 2009), some studies investigated whether infants’ attention for expressions is shaped by experiences in the rearing environment (e.g., Field et al., 1998; de Haan et al., 2004). In particular, these studies examined whether the
degree to which mothers experience negative emotions, assuming this relates to their daily expressions, is related to infants’ attention for expressions. For example, one study showed that infants of depressed mothers, who typically have a flat affect (Field, 1992), showed a decreased sensitivity to sad expressions: They looked less at sad faces than infants of non-depressed mothers (Field et al., 1998). In turn, infants of mothers, who express a lot of positive emotions, showed a heightened sensitivity to fearful expressions: Infants looked longer at and had larger brain responses to fearful versus happy expressions (de Haan et al., 2004). In sum, these studies suggest that infants show either a heightened or decreased
sensitivity to negative expressions dependent on the degree to which their mothers experience negative and positive emotions.
However, it is unclear whether the degree to which mothers experience negative emotions is related to infants’ attention for both negative and positive expressions (i.e., infants’
negativity bias). Infants’ attention to negative and positive expressions is of particular interest because atypical responses to both negative (e.g., attentional biases in anxiety disorders) and positive expressions (e.g., slower detection of happy expressions in depressive disorders) make individuals more vulnerable to develop emotion-related disorders (Mathews &
MacLeod, 2005; Surguladze et al., 2004). Hence, if the degree to which mothers experience negative emotions appears to be related to infants’ negativity bias, mothers who experience a lot of negative emotions might make their infants more vulnerable to develop emotion-related disorders. In order to investigate whether the mother’s negative emotions increase infants’
risk to develop emotion-related disorders later in life, the relationship between the degree to which mothers experience negative emotions and infants’ negativity bias needs to be
established first.
Moreover, three other questions remain unanswered about how the degree to which mothers experience negative emotions relates to infants’ negativity bias. First, no study yet
investigated how maternal anxiety is related to infants’ negativity bias. Given that maternal anxiety puts children at increased risk of developing emotional problems (e.g., O’ Connor, Heron, Golding, Beveridge, & Glover, 2002), it is important to investigate whether maternal anxiety affects infants’ negativity bias early in emotional development. Second, it is unclear whether the amount of time mothers spend with infants, make infants more or less affected by the mother’s negative emotions. When mothers experience a high degree of negative
emotions, they are suggested to be emotionally unavailable (i.e., affectively unresponsive to the infant; Field, 1994). Emotional unavailability of mothers can result in emotional
dysregulation of infants and is suggested to be more detrimental for infants than physical unavailability (Field, 1994). Therefore, infants might be more affected by the mothers’ negative emotions when mothers spend a lot of time with their infants. Third, it is unclear whether the relationship between the mother’s negative emotions and infants’ negativity bias is specific. For example, maternal depression might be more related to infants’ attention for sad expressions than fearful expressions since children of depressed mothers also show biased processing for sad faces only (e.g., Gibb, Benas, Grassia, & McGeay, 2009). Together, these findings suggest that it is unclear whether the degree to which mothers experience negative emotions, including anxiety, is related to infants’ negativity bias and whether this relationship is specific and moderated by the time mothers spend with their infants.
Previous studies measured infants’ attention for expressions with a variety behavioural and physiological indices (e.g., heart rate variability or ERPs). However, behavioural
measures are prone to subjectivity because they are often coded by observers. In addition, technical procedures associated with certain physiological measures (e.g., ERPs) can cause infants distress which might interfere with reliably measuring infants’ attention to
expressions. Pupil responses are a promising index of attention for expressions because they are involuntary, unbiased, temporally sensitive, and can be easily recorded without causing
extra distress (Karatekin, 2007). Furthermore, pupil responses are part of the sympathetic nervous system which prepares us for the ‘fight or flight’ response (Bradley, Miccoli, Escrig, & Lang, 2008; Karatekin, 2007). Given the close relationship of the sympathetic nervous system and emotion-related brain circuits (e.g., amygdala), pupil responses can be thought of as an index of attention for expressions (Laeng, Sirois, & Gredebäck, 2012).
Recently, two studies investigated infants’ emotion processing by measuring their pupil responses to expressions of peers (Geangu, Hauf, Bhardwaj, & Bentz, 2011) and strangers (Gredebäck, Ericksson, Schmitow, Laeng, & Stenberg, 2012). Consistent with the negativity bias, infants had larger pupil responses to fearful than neutral expressions of strangers. However, this difference was only found in infants cared for by both parents (Gredebäck et al., 2012). In addition, infants had larger pupil responses to videos of peers experiencing distress versus peers experiencing happiness or a neutral state (Geangu et al., 2011). However, videos of distress were always presented after neutral or happy videos. Hence, infants might only show an increased response to negative expressions when they are preceded by positive or neutral expressions. Thus, the order of the expression might explain the strength of the response more than the valence of the expressions.Furthermore, contrary to studies on infants’ emotion processing, adult studies suggested pupil responses are larger for both negative and positive expressions compared with neutral expressions, suggesting a different definition of the negativity bias (Bradley et al., 2008; van Steenbergen, Band, & Hommel, 2011). Together, these findings suggest that it remains unclear whether infants show a negativity bias in their pupil responses to expressions, how this negativity bias is defined, and whether this bias is due to order effects.
In summary, two key questions about infants’ attention for expressions remain unanswered: (1) It remains unclear whether infants show a negativity bias in their pupil responses to fearful and other negative expressions and (2) whether this bias is related to the
degree to which mothers experience negative emotions, including anxiety. Furthermore, two sub-questions remain unanswered about the relationship between infants’ negativity bias and the degree to which mothers experience negative emotions. That is, it remains unclear whether this relationship is specific and whether it is moderated by the time mothers spend with infants. This study aims to address these questions by measuring pupil responses of 14-month-old infants to happy, neutral, fearful, angry, and sad expressions, which are presented in a random order to control for the effect the order of the expressions might have on the pupil responses. The age of 14 months was chosen because Dutch infants start to walk around this age (van der Sluijs, Sakkers, & Bronswijk, 2009) and this has been suggested to be
accompanied by an increased interest in expressions of the caregiver (Campos et al., 2000).In addition, mothers (from a non-clinical population) were asked to report the degree to which they experienced negative emotions and how many days they took care of their infants.
Based on the previous findings of physiological studies (e.g., de Haan et al., 2004) and the preliminary findings of pupil response studies (e.g., Geangu et al., 2011), we expected infants to show a negativity bias for all negative expressions (Hypothesis 1). In other words, we expected infants to have larger pupil responses to negative expressions compared with positive or neutral expressions. Based on the finding that maternal depression was related to infants’ decreased sensitivity to sad expressions (Field et al., 1992), we expected that the more mothers experienced negative emotions, the smaller infants’ pupil responses would be to negative expressions compared with positive or neutral expressions (i.e., a decreased negativity bias; Hypothesis 2, see Figure 1). Given that emotional unavailability of mothers affects infants’ emotion processing (Field, 1994), we expected that the more time mothers spend with their infants, the more the mothers’ negative emotions affect infants’ negativity bias (Hypothesis 3). Lastly, we expected maternal anxiety or depression to be more strongly related to either a decreased negativity bias for fearful or sad expressions, respectively, than
angry expressions (i.e., specificity; Hypothesis 4, see Figure 2). In order to test whether different definitions of the negativity bias (i.e., negative vs. positive or negative vs. neutral) result in different conclusions about the negativity bias, the hypotheses were tested with both definitions.
Figure 2. Path analysis of hypothesis 4 with type of experiment, age, and gender as covariates (see section 2.5.).
Figure 1. Hypothetical model of hypotheses 2 and 3. The type of experiment, age, and gender are modelled as covariates (see section 2.5.). The indicators of maternal negative emotionality are not displayed in the model.
2. Method
2.1. Participants
The final sample consisted of 103 14-month-old infants (57 boys, M = 14.62, SD = 0.76, and 46 girls, M = 14.76, SD= 0.72). An additional 15 infants were tested but not included in the final sample because they did not attend sufficiently to the expressions (see data reduction for details). The participants were recruited by sending information letters via the municipality to all registered parents with infants of 14 to18 months old and by placing flyers in childcare centers. Infants were excluded if they suffered from visual or neurological abnormalities. Parents gave written informed consent for their infant to participate in the study. Most parents that participated in this study also participated in other research projects about language development but data of these projects were not used in this study.
2.2. Materials 2.2.1. Expressions
The stimuli were black-and-white pictures (1280 x 1240 pixels) of an adult woman showing one the following five expressions: happy, fearful, sad, angry, and neutral. The pictures were standardized in luminance and size, similar to previous studies (e.g., de Haan et al., 2004; Gredebäck et al., 2012). Prior to the experiment, 36 adults were asked to label the pictures. The ratings confirmed that the majority of the adults perceived the expressions as intended (happy: 94.4%; fearful: 88.9%; sad: 97.2%; angry: 94.4%; neutral was not validated because it cannot be labelled).
This study examined data of two similar experiments. Both experiments consisted of 4 blocks with 5 trials per block. In each block, 5 expressions were presented in a random order with two restrictions: (1) Expressions did not appear twice consecutively and (2) all
experiment, all blocks started with a neutral expression. Every trial started with an attention grabbing animation of a moving bird who made a sound (500 milliseconds), followed by a blank screen (500 milliseconds), followed by an expression (3000 milliseconds in the first experiment, 2000 milliseconds in the second experiment), and the trial ended with a blank screen (500 milliseconds). The duration of one trial was 3.5 to 4.5 seconds and the duration of 4 blocks was 70 to 90 seconds.
2.2.2. Pupil responses
Pupil responses were measured in millimeters for every 8.33 milliseconds with a Tobii T120 eye-tracker, recording with a frequency of 119.5‐121.3 Hz. The infant’s gaze was
calibrated before the experiment with a 5-point calibration procedure. The eye-tracker
measurements were completed in a dimly illuminated room with no other objects present that could distract the infant’s attention.
2.2.3. Maternal Negative Emotionality
Maternal negative emotionality consisted of three types of feelings: Depressive feelings, anxious feelings, and other aversive mood states referred to as negative affect. Depressive feelings were assessed with the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). The BDI measures depression severity with 21 items on 4-point severity scales, with higher scores indicating higher severity. The BDI has a good internal consistency (α = .81) and test-retest reliability (correlation coefficients > .60; Beck, Steer, & Carbin, 1988). The concurrent and construct validity of the BDI is also high (Beck et al., 1988). Anxious feelings were assessed with the Screening for Anxiety Related Emotional Disorders – Adults, which is an adult version of the SCARED-C (SCARED-A, Bögels & van Melick, 2004). The SCARED-A measures the frequency of anxiety symptoms with 71 items,
on a 3-point scale ranging from 0 (Almost never) to 2 (Often). The validity and reliability of the SCARED-A is not yet investigated, but the SCARED-C has a good internal consistency (α =.74 to .93), test-retest reliability (correlation coefficients =.70 to .90), and discriminative validity (Birmaher et al., 1997). Negative affect was assessed with the Negative Affect subscale of the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988). The Negative Affect subscale of the PANAS measures the frequency of six negative states with 21 items, on a 5-point scale ranging from 1 (Very slightly or not at all) to 5 (Extremely). The Negative Affect subscale has a good internal consistency (α = .84 to .87), test-retest reliability (correlation coefficient = .71), and a good construct validity in clinical and non-clinical samples (Crawford & Henry, 2004; Watson et al., 1988).
2.2.4. Daily care
Mothers were asked to indicate for every morning and afternoon of the week who took care of the infant in general: the mother, father, both parents, or someone else (e.g., nanny). The total score was the number of days mothers spend with their infants.
2.3. Procedure
Before the start of the experiment, infants got time to play with toys to get adjusted to the new environment. After the experiment was explained to the parent(s), the infant was placed in a car seat in front of the eye-tracker and the parent was seated behind or beside the car seat. If the infant refused to sit in the car seat, the infant could sit on the lap of the parent. The eye-tracker screen was placed at approximately 60 centimeters from the infant’s face. After calibration, four tasks were presented on the eye-tracker screen with a small break of approximately 3 minutes between each task. This study only used the data of one of the four tasks (see section 2.2.1.). The total duration of the instructions and experiment was half an
hour to one hour. The parent received 10 € as compensation for the visit and mothers were asked to fill in the questionnaires. The study was approved by the ethics committee of the University of Amsterdam.
2.4. Data reduction
The data reduction procedure was similar to previous studies that examined infants’ pupil responses (e.g., Geangu et al., 2011; Gredebäck et al., 2012). In order to examine pupil responses to the expressions, we defined an area of interest on the eye-tracker screen by specifying the x and y coordinates which covered the face. If samples of one eye were missing, samples of the other eye were used for interpolation. Missing samples for both eyes were extrapolated linearly from previous valid values. These valid values were calculated by the average pupil responses of both eyes that were within 4 standard deviations of infants’ own pupil distributions. Missing samples were only replaced if they lasted shorter than 500 milliseconds because it was assumed this was caused by errors of the eye-tracker. Longer sequences of missing samples were not replaced because it was assumed that infants were not looking at the screen. A baseline correction was performed on the pupil responses by
subtracting the average responses to blank pictures across trials from the pupil responses to the expressions. Pupil responses from both eyes were averaged across trials and blocks for all the five expressions separately. Trial means were only used when there were pupil samples for at least 500 milliseconds of the stimulus presentation. Infants with pupil samples of less than 500 milliseconds for all the five expressions across trials were removed from the analyses.
2.5. Statistical analyses
Multilevel modeling was used (MLM) to investigate whether infants showed a
negativity bias for all negative expressions (Hypothesis 1). The negativity bias was defined as the difference in infants’ pupil responses between negative (sad, fearful, and angry
expressions) and positive (i.e., happy expressions; reference category 1) or neutral expressions (reference category 2). MLM was used because the data have a multilevel structure as pupil responses to the expressions are nested in infants. MLM was preferred over a repeated-measures analysis because participants with missing values could still be included in the analysis. MLM was implemented through SPSS MIXED MODELS, version 20.
Structural equation modeling (SEM) was used to investigate whether maternal
negative emotionality was related to infants’ negativity bias and whether this relationship was moderated by the time mothers spend with their infants (Figure 1; Hypothesis 2 and 3). SEM was preferred over a regression analysis because it can handle missing values and can account for measurement error (Kline, 2011). First, the measurement model was tested with maternal negative emotionality as latent factor and maternal anxiety, depression, and negative affect as indicators of this factor. Second, the fit of the structural model was tested with maternal negative emotionality as independent variable (exogenous variable) and infants’ negativity bias as dependent variable (endogenous variable). The negativity bias was calculated by two difference scores: (1) The difference in mean pupil responses to negative minus positive expressions and (2) negative minus neutral expressions. The time mothers spend with their infants was added as moderator of the relationship between maternal negative emotionality and infants’ negativity bias. SEM was implemented through the lavaan and semTools packages in R, version 3.0.2 (Pornprasertmanit et al., 2013; R Core Team, 2013; Rosseel, 2012).
A path analysis was performed to investigate whether maternal anxiety or depression would be more related to either a decreased negativity bias to fearful or sad expressions, respectively, than angry expressions (Hypothesis 4). The negativity bias was calculated by subtracting the mean pupil response to positive expressions from the mean pupil response to either fearful, sad, or angry expressions. The positive expression was chosen as a reference category because only with positive as reference category, the relationship between maternal negative emotionality and infants’ negativity bias was significant (see Results). The path analysis was implemented through R (R Core Team, 2013).
Data of two similar experiments were used. In order to control for the differences between the two experiments (see section 2.2.1.), type of experiment was added as a covariate in both analyses. In order to explore whether the effect of the expressions or the effect of maternal negative emotionality on infants’ negativity bias was different for boys and girls or was related to infants’ age, gender and age were added as covariates both analyses.
3. Results
The data were inspected for missing values and outliers. Missing values were detected in both the pupil and maternal negative emotionality scores: angry expression (6.1%), fearful expression (6.1%), happy expression (3%), neutral expression (3%), sad expression (6.1%), depression score (BDI; 3%), anxiety score (SCARED; 3%), and the moderator time spent with the infant (7%). Little’s MCAR test suggested that these observations were missing completely at random, χ2 = 63.13, df = 72, p = .763. Before conducting the analyses to test the relationship between maternal negative emotionality and infants’ negativity bias, 4 of the 103 participants were deleted because all the maternal negative emotionality scores were missing. Six outliers were detected with the outlier labeling rule (Hoaglin & Iglewicz, 1987; Tukey,
1977). The extreme scores of the outliers were replaced with the next most extreme scores that were within the normal range specified by the labeling rule.
3.1. Negativity Bias
In order to investigate whether infants showed a negativity bias (Hypothesis 1), MLM was performed with pupil responses to the expressions as dependent variable and type of expression (i.e., happy, neutral, sad, angry, or sad) as first-level predictor. The covariates gender, age, and the type of experiment were added as main effects and interaction effects with type of expression. In order to examine whether pupil responses to the expressions differed significantly, pairwise comparisons were performed with first happy and second the neutral expression as reference category.
The predictor ‘type of expression’ was initially entered as a random effect based on the hypothesis that there might be individual differences in the effect of the type of expression on pupil responses. The model failed to converge and therefore all predictors were entered as fixed effects. As shown in Table 1, the full model with all predictors fit significantly better than one in which only the intercept was included. Thus, the model with predictors improved the model beyond the model that only considered variability in pupil responses between infants. As shown in Table 1, the model with random intercept fit significantly better than the model without random intercept. Thus, there was significant variability in pupil responses across infants. Therefore, the model with the random intercept was used as final model. Table 1 summarizes the three models with the log likelihood difference, the difference in degrees of freedom and the chi-square difference.
In this final model, the pupil responses to all expressions were not significantly different from each other, F(4 , 388.60) = 1.15, p = .333 (see Table 2 for mean pupil responses). Therefore, pairwise comparisons were not examined. The predictor type of
expression did not interact significantly with the covariates: Type of experiment (F(4, 388.07) = 2.04, p = .088), gender (F(1, 388.79) = 0.24, p = .918), or age (F(4, 387.89) = 1.39, p = .235).
Table 1
Comparison of Multilevel Models.
Model -2 Log Likelihood df* χ2 Difference test p
Intercepts only -454.86 2
Model with random intercept -492.67 7 M1-M2 = 37.81 < .001 Model without random intercept -113.09 1 M2-M3 = 379.58 < .001 * Difference in degrees of freedom between the two models.
Note. Smaller values of the -2 Log Likelihood refer to better fit.
Table 2
Infants’ Mean (SDs) Pupil Responses in Millimeters to the Expressions Expression Mean pupil response (SD)
Angry .26 (.23) Happy -.23 (.21) Fearful -.26 (.25) Sad -.25 (.21) Neutral -.24 (.24)
Note. Means are negative because of the baseline correction with blank pictures.
Table 3 shows the regression coefficients, test statistics, p-values, and confidence intervals for the fixed and random effects of the final model. Given that the mean pupil response to sad expressions was used as a reference category for the other expressions (SPSS automatically chooses the last level of a predictor as reference category), the intercept refers to the mean pupil response to sad expressions across infants. As shown in Table 3, the intercept is not significantly different from zero. The regression coefficients refer to the increase in mean pupil response to a certain expression from the overall mean pupil response to sad expressions. Table 3 shows that none of the expressions significantly predict pupil responses. There was significant variability across infants in pupil responses and a statistically significant residual, indicating there is room for improvement of the model. These findings
indicate that, contrary to a negativity bias, there was no difference in pupil responses between the expressions while controlling for the effect of the type of experiment, gender, and age.
Table 3
Results of the Final Two-Level Model
Random effects at Level 1 (Pupil Responses) and Level 2 (Individual Differences)
95% Confidence Parameter Standard p Interval Effect estimate error Wald Z (one-sided) Lower Upper Intercept 0.01 0.00 13.91 < .001 0.01 0.01 Residual 0.03 0.01 6.61 < .001 0.02 0.05 Fixed effects
95% Confidence Regression Standard Interval Expression Coefficient error t ratio df p Lower Upper Intercept 0.35 0.58 0.60 160.22 .546 -0.79 1.49 Angry 0.05 0.42 0.13 388.93 .900 -0.78 0.89 Fearful -0.45 0.42 - 1.05 388.94 .293 -1.28 0.39 Happy 0.13 0.42 0.31 388.93 .757 0.70 0.96 Neutral 0.48 0.42 1.15 388.97 .253 -0.35 1.32 Note. The intercept refers to the mean pupil response to sad expressions because pupil responses to sad expressions were used as reference category. The mean pupil response to sad expressions differs from the mean of sad expressions in Table 1 because the multilevel model accounts for missing values.
3.2. Maternal Negative Emotionality and Infants’ Negativity Bias
In order to investigate whether maternal negative emotionality was related to infants’ negativity bias and whether this relationship was moderated by the time mothers spend with their infants, SEM analyses were performed. The data were inspected for violations of the assumptions of SEM. The data were not multivariate normal: Mardia’s test statistic suggested significant skewness = 15.04 (χ2 = 223.05, p < .001) and kurtosis = 59.73 (Z= 13.94, p < .001). The maximum likelihood estimation robust (MLR) estimator was chosen because it produces robust ‘Huber-White’ standard errors and scaled test statistics when data are non-normal and incomplete (Arminger, Gerhard, & Schoenberg, 1989; Huber, 1972).
3.2.1. Maternal Negative Emotionality: Measurement Model
In order to test the structural model (Figure 2), the measurement model was tested first with maternal negative emotionality as latent factor and maternal anxiety, depression, and negative affect as indicators of this factor. The variance of the latent factor was fixed to unity. It appeared that only the indicator maternal depression had an acceptable level of reliability (see Table 3 for estimates and reliability of the indicators). Since the primary interest is in the parameter estimates and not the fit of the whole model, the low reliability of the indicators was sufficient.
Table 3
Standardized Factor Loadings, Standard Errors, Test Statistics and the Model-Based Reliability of Indicators of the Measurement Model
Standardized factor Standard Model-based loadings error Z value p value reliability Maternal Depression 0.89 0.76 5.06 < .001 0.79 Maternal Anxiety 0.58 1.69 3.75 < .001 0.33 Maternal Negative Affect 0.35 0.69 2.71 < .01 0.13
3.2.2. Structural Equation Model
In order to test whether maternal negative emotionality was related to infants’
negativity bias (Hypothesis 2), the fit of the structural model (Figure 1) was tested twice with reference categories ‘happy’ and ‘neutral’. As shown in Figure 1, the moderator was modelled as a latent factor with interactions terms between ‘maternal negative emotionality’ and ‘time spent with the infant’ as indicators. These interaction terms were computed by double-mean centering, which is recommended when the data are non-normal (Lin, Wen, Marsh, & Lin, 2010). The variances of the latent factors were fixed to unity and the exogenous variables were allowed to covary.
3.2.2.1. Reference Category: Happy
The fit of the model was good with a non-significant chi-square of χ2 (28) = 29.37, p = .394. The other approximate fit indices also suggested good model fit (GFI = 0.99, RMSEA = 0.02, AGFI = 0.99, NNFI = 0.98, PNFI = 0.43, CFI = 0.99, and IFI = 0.85). As presented in Figure 3, maternal negative emotionality was significantly related to infants’ negativity bias with happy as reference category, standardized coefficient = 0.37, p < .05. This positive coefficient indicates that a higher level of maternal negative emotionality was related to larger pupil responses to negative expressions compared with happy expressions. The type of
experiment, gender, age, and the time spent with the infant were not significantly related to infants’ negativity bias. In line with the second hypothesis, maternal negative emotionality was related to infants’ negativity bias while controlling for the effect of the type of
experiment, gender, and age. However, contrary to the second hypothesis, a higher level of maternal negative emotionality was related to an increased negativity bias instead of decreased negativity bias.
The time mothers spend with their infants did not significantly moderate the
relationship between maternal negative emotionality and infants’ negativity bias, standardized coefficient = -0.33, p = .051. Contrary to the third hypothesis, these findings indicate that the amount of time mothers spend with their infants did not affect the degree to which the
mothers’ negative emotions affect infants’ negativity bias. However, given the medium effect size of the coefficient (see Cohen, 1992) and low power to detect a model with a good fit (i.e., 23% chance to detect a model with good fit according to the not-close-fit hypothesis with H0:
spent with the infant’ might be non -significant due to a lack of power.
Given that the model with happy as reference category resulted in a significant
relationship between maternal negative emotionality and infants’ negativity bias, the residuals of this model and the close-fit/poor-fit hypotheses were further inspected. Table 5 shows the matrix of correlation residuals of this model. Correlation residuals with absolute values > .10 suggest that the model does not explain the data very well (Kline, 2011). As shown in Table 5, the residuals between maternal negative affect and the other variables were especially large. This can be explained by the low reliability of maternal negative affect (α = .13) which is below the number at which it is recommended to delete an indicator from further analyses (α < .30; Joreskog, 1993). In order to examine whether maternal negative affect might be related to infants’ negativity bias directly, an extra path was added to the model, but this path was non-significant. Given that the primary interest of this study was in the parameter estimates
Figure 3. Results of the structural model with happy as reference category. The standardized coefficients and standardized residuals variances (arrows pointing to the indicators) are depicted. Covariances between exogenous variables are allowed but not displayed in the figure.
and not the fit of the whole model, the indicator ‘maternal negative affect’ was not removed from the model. The value of RMSEA is .06 (Confidence interval: 0 - .08) which means that the close-fit hypothesis is not rejected (p = .291) based on the lower bound value of 0. The poor-fit hypothesis can be rejected because the upper bound value of .08 does not exceed .10 (p = .177). These results suggest that the significant relationship between maternal negative emotionality and infants’ negativity bias can be reliably interpreted.
Table 5
Correlation Residuals for the Structural Model with the Happy Expression as Reference Category
Note. Correlations of .10 or higher are displayed in bold.
3.2.2.2. Reference Category: Neutral
The fit of the model was tested again with ‘neutral’ as reference category. The fit of the second model was good with a non-significant chi-square, χ2 (28) = 28.51, p = .438. The other approximate fit indices also suggested good model fit (GFI = 0.99, RMSEA = 0.01, AGFI = 0.99, NNFI = 0.99, PNFI = 0.43, CFI = 0.99, and IFI = 0.85). Maternal negative emotionality was not significantly related to infants’ negativity bias with neutral as reference category, standardized coefficient = -0.08, p = .664. The time mothers spend with their infants did not significantly moderate the relationship between maternal negative emotionality and
infants’ negativity bias, standardized coefficient = -0.02, p = .900. The type of experiment, gender, and age were not significantly related to infants’ negativity bias. Contrary to the second hypothesis, these findings indicate that there was no relationship between maternal negative emotionality and infants’ negativity bias when the negativity bias was defined as the difference in pupil responses between negative and neutral expressions.
3.2.2.3. Exploratory Models
An exploratory model was tested to investigate whether maternal negative
emotionality affected not only infants’ negativity bias but also infants’ pupil responses to expressions in general. In this exploratory model, the negativity bias was replaced by pupil responses to the five expressions separately. The fit of the model was poor with a significant chi-square, χ2 (41) = 61.19, p = .02. The other approximate fit indices suggested acceptable model fit (GFI = 0.99, RMSEA = 0.08, AGFI = 0.99, NNFI = 0.88, PNFI = 0.36, CFI = 0.95, and IFI = 0.89). Since the fit of the model was poor, the parameter estimates could not be interpreted reliably. A second exploratory model was tested to investigate whether maternal negative emotionality affected negative expressions in general. In this second exploratory model, the negativity bias was replaced by the mean pupil response to negative expressions (mean of sad, angry, and fearful expressions). The fit of the model was good with a non-significant chi-square, χ2 (25) = 31.81, p = .164. The other approximate fit indices also suggested good model fit (GFI = 0.99, RMSEA= 0.05, AGFI = 0.99, NNFI = 0.88, PNFI = 0.39, CFI = 0.95, and IFI = 0.83). The second exploratory model revealed that maternal negative emotionality was not significantly related to pupil responses to negative expressions, standardized coefficient = -0.08, p = .664. These findings indicate that maternal negative emotionality was not related to infants’ pupil responses to negative expressions alone.
3.3. Specificity of the Negativity Bias
In order to examine whether maternal anxiety or depression was more related to either a decreased negativity bias to fearful or sad expressions, respectively, than angry expressions (Hypothesis 4), a path analysis was tested. Table 4 shows that maternal anxiety and
depression were not significantly related to the differences scores of pupil responses to angry, fearful, and sad expressions. These findings indicate that, contrary to the fourth hypothesis, the negativity bias was not specific: maternal anxiety and depression were not more related to infants’ negativity bias to fearful and sad expressions versus angry expressions.
Table 4
Results of the Path Analysis with the Predictors Maternal Anxiety and Maternal Depression. Maternal Anxiety
Effect Estimate Standard Error Z –value p – value Angry 0.06 0.23 0.26 .794
Fearful 0.24 0.29 0.83 .408 Sad -0.07 0.19 -0.37 .711
Maternal Depression
Effect Estimate Standard Error Z –value p – value Angry -0.40 0.61 -0.66 .507
Fearful -0.41 0.50 -0.81 .417 Sad -0.53 0.47 -1.12 .261
4. Discussion
This study investigated whether infants’ pupil responses to expressions showed a negativity bias and whether these responses were related to the 1) degree to which mothers experience negative emotions and 2) the amount of time mothers spend with their infants. Finally, the specificity of this relationship was also tested (e.g., maternal anxiety might be more related to infants’ pupil responses to fearful expressions than other expressions). The results demonstrated that infants did not show a negativity bias in their pupil responses to expressions. In other words, infants’ pupil responses to negative expressions were not larger compared with positive or neutral expressions. Consistent with our hypothesis, the degree to
which mothers experienced negative emotions affected infants’ negativity bias. However, a higher level of maternal negative emotionality was related to an increased instead of
decreased negativity bias. In addition, the mother’s negative emotions were only related to the difference in infants’ pupil responses between negative and positive expressions and not to infants’ responses to negative expressions alone. Furthermore, the relationship between
maternal negative emotionality and infants’ negativity bias was not specific. That is, maternal anxiety was not more strongly related to infants’ pupil responses to fearful expressions than sad or angry expressions. The degree to which maternal negative emotionality affected infants’ negativity bias was not dependent on the amount of time mothers spend with their infants.
Contrary to previous studies examining infants’ responses to expressions (e.g., de Haan et al., 2004; Peltola et al., 2009), this study did not provide support for a negativity bias in infants’ pupil responses to either fearful, sad, or angry expressions. One possible
explanation is that whether infants show a negativity bias in response to expressions is dependent on how infants’ responses are measured. Most studies that reported a negativity bias in infants’ responses to expressions, examined infants’ brain responses (i.e., ERPs) to expressions and since infants’ pupil responses did not show a negativity bias, brain and pupil responses may tap different aspects of attention (see Hong, Walz, & Sajda, 2014). In line with this idea, previous studies suggested different measures of responses to expressions can result in different conclusions about the negativity bias (e.g., Bradley, Miccoli, Escrig, & Lang, 2008). Another possible explanation is that infants might looked away more often when negative expressions were presented compared to positive or neutral expressions and hence enhanced pupil responses to negative expressions could not be recorded while infants’ brain responses possibly would have shown a negativity bias. In line with this idea, mean pupil responses to negative expressions were missing more often than mean pupil responses to
positive and neutral expressions (6.1% vs. 3%). This was possibly due to the fact infants more often avoided looking at the negative expressions compared with positive or neutral
expressions (see Grossmann et al., 2007). In order to investigate whether the contrasting findings are due to infants’ averted gaze when seeing negative expressions, future research should test infants’ gazes, pupil, and brain responses to expressions simultaneously.
However, these two explanations are not sufficient since two previous studies did find a negativity bias when examining infants’ pupil responses to expressions (Geangu et al., 2011; Gredebäck et al., 2012). One possible explanation is that infants only show a negativity bias when negative expressions are directed at objects or persons (signalling that these objects and persons might be harmful for the infant) instead of when expressions are directly gazing at the infants. In line with this idea, one study suggested infants’ brain responses are more
pronounced when fearful expressions are directed at objects compared with fearful expressions directly gazing at the infant (Hoehl, Palumbo, Heinisch, & Striano, 2008). A second possible explanation is that the static expressions used in this study have a low
ecological validity and thus are not realistic enough to elicit pupil responses similar to infants’ pupil responses to expressions in daily life. In line with this idea, some studies suggest
dynamic faces are easier to interpret and elicit higher physiological responses than static expressions (e.g., Alves, 2013; Ambadar, Schooler, & Cohn, 2005). However, the majority of studies on infants’ emotion processing used static expressions and were able to detect
differences in infants’ responses to expressions (e.g., de Haan et al., 2004; Gredebäck et al., 2012). Moreover, other studies suggested that dynamic expressions do not improve emotion recognition compared with static expressions and hence static expressions are probably realistic enough to elicit responses (e.g., Fiorentini & Viviani, 2011). A third explanation is that the relatively short duration of the expressions (3 s) compared to other studies (e.g., 50 s in Geangu et al., 2011; 5 s in Gredebäck et al., 2012) was not long enough to elicit strong
responses and might only allowed limited processing. Consistent with this idea, one study found differences in infants’ pupil responses only within certain time windows (Geangu et al., 2011). Thus, we might have found a negativity bias in later time windows if the expressions were presented longer. Another possible explanation is that the negativity bias found in the study of Geangu et al. (2011) was due to the fixed order since this study used a random instead of fixed order.
Consistent with previous studies, the degree to which mothers experienced negative emotions affected infants’ attention for expressions (e.g., Field et al., 1998). Moreover, this study extended previous research by suggesting that in addition to maternal depression, maternal anxiety, and negative affect are related to infants’ pupil responses to expressions. However, contrary to studies on maternal depression, the degree to which mothers
experienced negative emotions was related to a heightened instead of a decreased sensitivity for negative expressions (Field et al., 1998; Field, Diego, Hernandez-Reif, 2009). These conflicting results can possibly be explained by the fact that previous studies investigated infants’ brain responses instead of pupil responses (Field et al., 1998) and tested younger
infants (3 months and younger; Field et al., 1998; Field et al., 2009). Consistent with the notion that age plays a role, older children (between 8 and 14 years old) of depressed and anxious mothers also displayed a heightened sensitivity in their pupil responses to negative expressions compared to children of non-depressed/anxious mothers (Burkhouse, Siegle, & Gibb, 2014). Together, these findings seem to suggest that infants of mothers who experience a high level of negative emotions show atypical responses to expressions but how these atypical responses are defined might be dependent on age. In order to confirm this idea, future research can compare pupil responses to expressions of children of different age groups in clinical and non-clinical populations.
Contrary to our third hypothesis, the amount of time mothers spend with their infants did not affect the degree to which the mothers’ negative emotions affected infants’ responses to expressions, although this might be due to a lack of power. One possible explanation is that whether the amount of time mothers spend with their infants affects infants’ pupil responses is also dependent on the amount of time fathers spend with infants. For example, the mother’s negative emotions might affect infants’ responses to expressions to a lesser extent when fathers express a lot of positive emotions to infants. Hence, the amount of time fathers spend with their infants and the degree to which fathers experience negative emotions might play a role in determining how much the mother’s negative emotions influence infants’ responses to expressions. In line with this idea, one study suggested that how infants respond to
expressions is dependent on whether infants are cared for by either parents or mainly one parent (Gredebäck et al., 2012). Future research can address this by measuring the degree to which fathers experience negative emotions and how much time they spend with their infant. Another possible explanation is that the amount of time mothers spend with infants is a weak indicator of the degree to which mothers express negative emotions to their infants. For example, mothers might spend a lot of time with their infant (e.g., being in the same house), but interact only little with their infant when they experience a high level of negative
emotions because they might want to hide these negative emotions. In order to more reliably assess the mothers’ expressions to the infant, future research can possibly observe mothers while interacting with their infants and let trained observers code the amount of negative and positive expressions mothers express to their infants.
Although previous studies suggested maternal anxiety and depression were specifically related to infants’ responses to angry and sad expressions, respectively (e.g., Burkhouse et al., 2014; Field et al., 1998), this study did not provide evidence for a specific relationship between maternal negative emotionality and infants’ negativity bias. One possible
explanation is that mothers from previous studies suffered from a clinical level of negative emotions and thus expressed more negative expressions to their infants typical to their disorder. However, there is a high within-category heterogeneity between individuals suffering from the same disorder (i.e., individuals with the same diagnosis can have very different complaints; Krueger & Bezdjian, 2009). Hence, the fact that the mothers suffer from a higher level of anxiety and depression does not per se mean that they also express more negative expressions specific to their disorder. Therefore, future studies can possibly test whether a model with separate predictors of maternal negative emotionality versus a model with one latent factor of maternal negative emotionality as predictor, similar to this study, is a better predictor of infants’ responses to expressions.
There are several limitations of this study that need to be addressed. First, the study was based on the assumption that the mother’s negative emotions would influence infants’ responses to expressions through exposure to the mother’s negative expressions. However, infants’ pupil responses could be related to the mother’s negative emotions because of shared genes as much as through interactions with the mother. Future research can address this issue by using a behavioural genetic design. Second, the correlational design of this study does not allow drawing causal conclusions. In other words, infants’ responses to expressions might also affect how frequently mothers respond to infants with negative emotions. Future research can address this issue by using a prospective design or by manipulating the degree to which infants are exposed to negative emotions (e.g., investigate whether interventions effective in reducing depressive and anxious symptoms of mothers change infants’ responses to
expressions). Third, this study only measured infants’ emotion processing with pupillometry. Given that different measures of infants’ responses to expressions can result in different conclusions, future studies should combine different measures of infants’ responses to
as previously described, this study had only low power to detect a model with good fit. Therefore, future studies should replicate the tested models with larger sample sizes to investigate whether, with a sufficient level of power, the amount of time mothers spend time with their infants might play a role in determining the degree to which the mother’s negative emotions affect infants’ negativity bias.
Although these issues limit the conclusions that can be drawn from this study, this study extends previous research in two important ways. First, this study was first to suggest that infants of 14 months old, with mothers who experience a high level of negative emotions, show a heightened instead of decreased sensitivity to negative expressions in their pupil responses, similar to older children (Burkhouse et al., 2014). Second, unlike previous studies (e.g, Burkhouse et al., 2014; Field et al., 1998), this study suggested maternal negative
emotionality is not only related to infants’ emotion processing in clinical populations but also in non-clinical populations. Given that infants’ pupil responses to expressions seem to be related to maternal negative emotionality in both clinical and non-clinical populations, pupil responses may serve as a biomarker of risk for developing emotion-related disorders later in life. In order to confirm that atypical pupil responses to expressions are a risk factor for developing emotion-related disorders, future research should investigate whether atypical pupil responses in infancy predict the onset of emotion-related disorders in adulthood.
In summary, the current findings add to the growing body of evidence suggesting that maternal negative emotionality is related to infants’ emotion processing in both clinical and non-clinical populations. However, this relationship was not found to be specific or to be affected by the time mothers spend with their infants. Furthermore, the results suggested infants did not show a negativity bias in their pupil responses. Although future research is needed to delineate the genetic vs. exposure effects and to investigate whether infants’ pupil responses predict the onset of emotion-related disorders later in life, infants’ pupil responses
seem to be a promising bio marker of risk for developing emotion-related disorders and can possibly be used as screening tool for preventive interventions for mothers who suffer from high levels of negative emotions.
References
Alves, N. T. (2013). Recognition of static and dynamic facial expressions: a study review. Estudos de Psicologia (Natal), 18, 125-130.
Ambadar, Z., Schooler, J. W., & Cohn, J. F. (2005). Deciphering the enigmatic face the importance of facial dynamics in interpreting subtle facial expressions. Psychological
Science, 16, 403-410. doi: 10.1111/j.0956-7976.2005.01548.x
Arminger, G., & Schoenberg, R. J. (1989). Pseudo maximum likelihood estimation and a test for misspecification in mean and covariance structure models. Psychometrika, 54.3, 409-425. doi: 10.1007/BF02294626
Beck, A. T., Steer, R. A., & Carbin, M. G. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Psychology Review,
8, 77-100. doi: 10.1016/0272-7358(88)90050-5
Beck, A. T., Ward, C., & Mendelson, M. (1961). Beck depression inventory (BDI). Archive of
General Psychiatry, 4, 561-571.
Bögels, S. M., & Melick, M. V. (2004). The relationship between child-report, parent self- report, and partner report of perceived parental rearing behaviors and anxiety in children and parents. Personality and Individual Differences, 37, 1583-1596. doi:0.1016/j.paid.2004.02.014
Bradley, M. M., Miccoli, L., Escrig, M. A., & Lang, P. J. (2008). The pupil as a measure of emotional arousal and autonomic activation. Psychophysiology, 45, 602-607.doi: 10.1111/j.1469-8986.2008.00654.x
in children of depressed and anxious mothers. Journal of Child Psychology and
Psychiatry, 55, 1009-1016. doi:10.1111/jcpp.12225
Campos, J. J., Anderson, D. I., Barbu‐Roth, M. A., Hubbard, E. M., Hertenstein, M. J., &
Witherington, D. (2000). Travel broadens the mind. Infancy, 1, 149-219. doi: 10.1207/S15327078IN0102_1
Cohen, J. (1992). A power primer. Psychological bulletin, 112, 155-159. doi: 0.1037/0033- 2909.112.1.155
de Haan, M., Belsky, J., Reid, V., Volein, A., & Johnson, M. H. (2004). Maternal personality and infants’ neural and visual responsivity to facial expressions of emotion. Journal of
Child Psychology and Psychiatry, 45, 1209–1218. doi:
10.1111/j.1469-7610.2004.00320
Field, T. (1994). The effects of mother's physical and emotional unavailability on emotion regulation. Monographs of the Society for Research in Child Development, 59, 208-227. doi: 10.1111/j.1540-5834.1994.tb01286.x
Field, T., Diego, M., & Hernandez-Reif, M. (2009). Depressed mothers’ infants are less responsive to faces and voices. Infant Behavior and Development, 32, 239-244. doi:10.1016/j.infbeh.2009.03.005
Field, T., Pickens, J., Fox, N. A., Gonzalez, J., & Nawrocki, T. (1998). Facial expression and EEG responses to happy and sad faces/voices by 3‐month‐old infants of depressed
mothers. British Journal of Developmental Psychology, 16, 485-494. doi: 10.1111/j.1540-5834.1994.tb01286
Fiorentini, C., & Viviani, P. (2011). Is there a dynamic advantage for facial expressions? Journal of Vision, 11, 1-15. doi:10.1167/11.3.17
response to others' positive and negative emotions. Public Library of Science One, 6, e27132. doi: 10.1371/journal.pone.0027132
Gibb, B.E., Benas, J.S., Grassia, M., & McGeay, J.E. (2009). Children’s attentional biases and 5-HTTLPR genotype: Potential mechanisms linking mother and child depression.
Journal of Clinical Child and Adolescent Psychology, 38, 415–426. doi:10.1080
/15374410902851705
Gredebäck, G., Eriksson, M., Schmitow, C., Laeng, B., & Stenberg, G. (2012). Individual differences in face processing: Infants’ scanning patterns and pupil dilations are influenced by the distribution of parental leave. Infancy, 17, 79-101. doi: 10.1111/ j.1532-7078.2011.00091.x
Grossmann, T. (2010). The development of emotion perception in face and voice during infancy. Restorative Neurology and Neuroscience, 28, 219-236. doi: 10.3233/RNN-2010-04999
Grossmann, T., Striano, T., & Friederici, A. D. (2007). Developmental changes in infants’ processing of happy and angry facial expressions: A neurobehavioral study. Brain &
Cognition, 64, 30-41. doi: 10.1016/j.bandc.2006.10.002
Hoaglin, D. C., & Iglewicz, B. (1987). Fine-tuning some resistant rules for outlier labeling. Journal of the American Statistical Association, 82, 1147-1149. doi: 10.1080/01621459.1987.10478551
Hoehl, S., Palumbo, L., Heinisch, C. & Striano, T. (2008). Infants’ attention is biased by emotional expressions and eye gaze direction. Neuroreport, 19, 579–582. doi: 10.1097/WNR.0b013e3282f97897
Hong, L., Walz, J. M., & Sajda, P. (2014). Your eyes give you away: Prestimulus changes in pupil diameter correlate with poststimulus task-related EEG dynamics. Public Library
Huber, P. J. (1972). The 1972 wald lecture robust statistics: A review. The Annals of
Mathematical Statistics, 43, 1041-1067.
Joreskog, K. (1993). Testing structural equation models. In K. A. Bollen, & J. S. Long (Eds.),
Testing structural equation models (pp. 294–316). Newbury Park, CA: Sage
Publications.
Karatekin, C. (2007). Eye tracking studies of normative and atypical development.
Developmental Review, 27, 283–348. doi: 10.1016/j.dr.2007.06.006
Kline, R. B. (2011). Principles and practice of structural equation modeling. (Third edition). The Guilford Press.
Kotsoni, E., de Haan, M., & Johnson, M.H. (2001). Categorical perception of facial
expressions by 7-month-old infants. Perception, 30, 1115–1125. doi: 10.1068/p3155 Krueger, R. F., & Bezdjian, S. (2009). Enhancing research and treatment of mental disorders
with dimensional concepts: toward DSM‐V and ICD‐11. World Psychiatry, 8, 3-6.
Laeng, B., Sirois, S., & Gredebäck, G. (2012). Pupillometry: A window to the preconscious?
Perspectives on Psychological Science, 7, 18-27. doi: 10.1177/1745691611427305
Lin, G. C., Wen, Z., Marsh, H. W., & Lin, H. S. (2010). Structural equation models of latent interactions: Clarification of orthogonalizing and double-mean-centering
strategies. Structural Equation Modeling, 17, 374-391. doi: 0.1080/10705511 .2010.488999
Ludemann, P. M., & Nelson, C. A. (1988). Categorical representation of facial expressions by 7-month-old infants. Developmental Psychology, 24, 492-501.
Mathews, A., & MacLeod, C. (2005). Cognitive vulnerability to emotional disorders. Annual
Review of Clinical Psychology, 1, 167-195. doi: 10.1146/annurev.clinpsy.1
.102803.143916
by seven-month-old infants. Child Development, 56, 58-61.
O'Connor, T. G., Heron, J., Golding, J., Beveridge, M., & Glover, V. (2002). Maternal antenatal anxiety and children's behavioural/emotional problems at 4 years: Report from the Avon Longitudinal Study of Parents and Children. The British Journal of
Psychiatry, 180, 502-508. doi: 10.1192/bjp.180.6.502
Peltola, M. J., Leppänen, J. M., Mäki, S., & Hietanen, J. K. (2009). Emergence of enhanced attention to fearful faces between 5 and 7 months of age. Social Cognitive and
Affective Neuroscience, 4, 134-142. doi: 10.1093/scan/nsn046
Pornprasertmanit, S., Miller, P., Schoemann, A., Rosseel, Y., Quick, C., & Pornprasertmanit, M. S. (2013). Package ‘semTools’.
Preacher, K. J., & Coffman, D. L. (2006). Computing power and minimum sample size for RMSEA [Computer software]. Available from http://quantpsy.org/.
R Core Team (2013). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Richards, J.E. (2003). Attention affects the recognition of briefly presented visual stimuli in infants: An ERP study. Developmental Science, 6, 312–328. doi: 10.1111/1467-7687.00287
Rohner, J. C. (2002). The time-course of visual threat processing: High trait anxious individuals eventually avert their gaze from angry faces. Cognition & Emotion, 16, 837-844. doi: 10.1080/02699930143000572 ‘
Rosseel, Y. (2012). lavaan: An R package for structural equation modeling and more Version 0.5-12 (BETA).
van Steenbergen, H., Band, G. P., & Hommel, B. (2011). Threat but not arousal narrows attention: Evidence from pupil dilation and saccade control. Frontiers in
Surguladze, S. A., Young, A. W., Senior, C., Brébion, G., Travis, M. J., & Phillips, M. L. (2004). Recognition accuracy and response bias to happy and sad facial expressions in patients with major depression. Neuropsychology,18, 212-218. doi: 10.1037/0894-4105.18.2.212
Vaish, A., Grossmann, T., & Woodward, A. (2008). Not all emotions are created equal: The negativity bias in social-emotional development. Psychological Bulletin, 134, 383. doi: 10.1037/0033-2909.134.3.383
van der Sluijs, J. A., Sakkers, R. J. B., & Bronswijk, J. A. H. M. (2009). Algemeen. In Kinderorthopedie (pp. 11-16). Bohn Stafleu van Loghum.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality
