Is behavioral Inhibition at the age of 2,5 related to anxious behavior at the age of 4,5 in social referencing situations?

Is Behavioral Inhibition at the age of 2,5

related to Anxious Behavior at the age of

4,5 in Social Referencing Situations?

Master Thesis

Inessa Ros

5934729

First assessor: Maartje Raijmakers

Second assessor: Susan Bögels

Word count: 7865

Table of contents

Discussion and Conclusion p. 18

Major findings p. 23

Abstract

Background: The temperamental trait of Behavioral Inhibition (BI), defined by withdrawn/inhibited

behavior towards unfamiliar situations, plays an important role in the development of anxiety in children,

alone and together with interpersonal factors such as parenting behaviors. Interpersonal anxious behaviors

can be divided in fearful and avoidant behavior. This study investigates whether BI at the age of 2.5 year is prospectively associated with fearful/avoidant behavior at the age of 4.5, and whether this link is affected by the nature of the SR situations (social or nonsocial) and by the gender of the parent present during these situations.

Method: BI at the age of 2.5 (N = 111) was measured with an adapted version of the Laboratory

Temperament Assessment Battery (Lab-TAB; Goldsmith & Rothbart, 1996). Fearful/avoidant behavior was measured during Social Referencing (SR) tasks at the age of 4.5. During the social task the children-parent dyads were expected to interact with a stranger, whereas during the nonsocial task the children-parent dyads were confronted with a robot. All children experienced these SR situations with their mother and father.

Results: Higher levels of BI at the age of 2.5 were linked to more avoidance at the age of 4.5 during SR

tasks. Early BI and later fear showed interaction with the kind of SR task; during the social SR tasks children with high BI showed more fearful behavior and during the nonsocial SR tasks children with low BI showed less fearful behavior. The link between early BI and later avoidance did not differ across SR situations. Lastly there were differences in level of fear/avoidance per task; Children show more avoidance behavior during the nonsocial SR task and more fearful behavior during the social SR task.

Conclusions: This research has shown that BI in toddlerhood is prospectively related to more avoidance and

behavior in early childhood in SR situations. Next to that it is found that children with high toddler BI display higher levels of fear during nonsocial situations, and children with low toddler BI show lower levels of fear during nonsocial situations. Lastly the sort of task influences the level of fear and avoidance.

Keywords: Social referencing, Behavioral Inhibition, Development, Fear, Anxiety, Avoidance, Children,

Word count abstract: 346 Abbreviations:

Social Referencing (SR) Behavioral Inhibition (BI)

Laboratory Temperament Assessment Battery (Lab-TAB; Goldsmith & Rothbart, 1996)

Introduction

In the development of anxious behavior of children, interpersonal factors such as fearful and avoidant behaviors during social referencing (SR) situations, and child characteristics, such as behavioral inhibition (BI; Kagan, Reznick, Clarke, Snidman, Garcia-Coll, 1984), seem to play a significant part. BI is a temperamental characteristic, encompassing elevated stress responses, social withdrawal, hyper-vigilance towards threat, and avoidant coping styles towards new or unfamiliar objects, situations, or persons (White, Helfinstein, & Fox, 2010). BI is generally assumed to be a unitary construct, and is assessed by combining ratings of specific behaviors towards novel social and nonsocial stimuli (Dyson, Klein, Olino, Dougherty, & Durbin, 2011). BI may constitute a predisposition for anxious psychopathology (White et al., 2010) and this vulnerability in children with BI is especially apparent in the case of social anxiety (Clauss & Blackford, 2012). For example, Schwartz, Snidman, and Kagan (1999) found that high BI in toddlerhood predisposed adolescents to social anxiety, while in the case of an uninhibited temperament this association ceased to exist. Thus when children did have an uninhibited temperament, they did not show this predisposition to social anxiety. Muris, van Brakel, Arntz and Schouten (2011) also found in their

children aged 5 to 8 years. These findings show that children with an inhibited temperament are more likely to show socially anxious behaviors and to develop (social) anxiety disorders later in life, while children with uninhibited temperament are not.

Next to the heightening effects of BI on later anxious behaviors, a classification of an inhibited temperament during infancy shows greater continuity across development compared to a classification of uninhibited temperament. (Kagan, Reznick & Snidman, 1987). BI is highly unstable from infancy to toddlerhood in normal development. However, children who are at the high end (top 10 to 15 %) continuously remain to show inhibited behavior (Kagan, Reznick & Snidman, 1989). Kagan & Moss (1962) found as well that high BI showed an exceptional high level of continuity from toddlerhood trough adulthood, even compared to other behavioral tendencies shown in the first three years of life. One of the explanations for this stability of high BI is the provoking effect it has on particular parenting dimensions like over-involvement, overprotection, lack of autonomy promotion or encouragement, and aversive or negative behavior. These parental behaviors are part of the overall parental control dimension (McLeod, Wood & Weisz, 2007). Overprotection and over-involvement are for instance discouragement of the independent attempts to explore and restricting the child’s own actions. These behaviors can also encompass a lack of autonomy promotion or encouragement (Rubin, Burgess & Hastings, 2002). Overprotection and over-involvement or intrusiveness can increase the child’s wariness or fearfulness during peer interaction (Rubin, Stewart & Coplan, 1995). Anxious rearing patterns overall can specifically heighten social anxiety symptoms in children (Muris, et al., 2011). On the other hand, children who are dispositionally fearful or wary may trigger overprotective or over-involvement behavior from parents during (potentially) emotion-arousing situations, such as unfamiliar situations (Rubin et al., 2002). These examples show that these parenting styles are both initiated by the child’s BI, and play a part in maintaining the child’s inhibitive behavior (Murray et al, 2009). Thus, in the development of anxiety, moderately stable BI plays an important role. But this relationship is not independent of interpersonal factors such as parenting behaviors; there is a bidirectional association between these

constructs. There is evidence that the development of anxiety in children might be partially funded by parental behaviors (Aktar, Majdandzic, de Vente, & Bögels, 2013).

In the current study, the effect of BI on fearful and avoidant behaviors of children during social referencing (SR) situations is examined. In these situations parents are active parties in the sense that they actively engage together with their children during these situations. Children use SR when they use the emotional and behavioral expressions of their parents to determine how to respond towards ambiguous or novel situations (Feinman, 1982; Feinman, Roberts, Hsieh, Sawyer, & Swanson, 1992). SR has been shown to play a significant role in children’s anxious behavior. Gerull and Rapee (2002) showed for example that mothers’ expressions of fear lead to a fearful and avoidant expression in infants aged 15-20 months towards potentially fear-provoking objects. In addition, de Rosnay, Cooper, Tsigaras, and Murray (2006) found that 12-14 months old infants’ expression of fear towards a stranger was subsequently determined by maternal expressions of either anxious or non-anxious social responses towards that same stranger. Murray et al. (2008) found this same effect of modeling of fearful behaviors in a group of mothers with a social phobia and their babies aged 10-14 months. They also found an interactive effect of BI on parenting behaviors and its effect on children’s anxious behaviors. Until this date most research to fearful and avoidant behavior during SR situations focused on children in the end of infancy, beginning

toddlerhood (age range 10-20 months old), and not on later stages of life. From 6-18 months SR behavior is supposed to be most salient (Emde, 1992). Although parents’ influence possibly

changes as children grow older, it is still likely that parents play a significant role in the behavior of children in later stages of life (Aktar, Majdandzic, de Vente, & Bögels, 2014). Woodruff-Borden, Morrow, Bourland and Cambron (2010) found for example that the interaction-style of anxious parents is different with their children, aged 6-12 years, compared to non-anxious parents. The anxious parents tend to show less agreeing and praising, and seem to ignore their children more during a task. This implies that anxious parents show different behaviors compared to non-anxious parents, which can have negative effects on the behavior of their children. Children of anxious

parents might be less skilled in coping with stressful life experiences (Woodruff-Borden et al., 2010). Thus, although SR behavior tends to be most salient at the age of 6-18 months old, the effect of parenting behaviors seems to be salient in older aged children as well. However, this is yet to be investigated in SR situations.

Moreover, the majority of the studies use social SR situations to measure the level of infants’ fear/avoidance during SR situations. Aktar et al. (2013) measured the level of fear and avoidance of babies aged 12 months during both social and nonsocial SR situations and found that the level of fear/avoidance differs between these different SR situations. Babies tended to display more fear and less avoidance in the nonsocial SR situation compared to the social SR situations. In a follow-up of this sample in toddlerhood by Aktar et al. (2014) showed that toddlers display more fearful/avoidant behavior during a nonsocial SR situation compared to a social SR situation. Thus, babies and toddlers tend to display a difference in level of fear/avoidance behavior during social and nonsocial situations. Whether this differential association is present in early childhood is yet to be investigated.

The relationship between BI and fear/avoidance in SR situations is also mostly investigated with social stimuli, while nonsocial stimuli might give innovative results too. Aktar et al. (2013) examined this relationship between BI and fearful behavior in babies aged 12 months in SR situations with both social and nonsocial stimuli. They found that BI was a predictor of

fearful/avoidant behavior during both social and nonsocial SR situations. In addition, Aktar et al. (2014) investigated the effect of BI measured at the age of 12 months on fearful and avoidant behavior of toddlers aged 2.5 years during both social and nonsocial SR situations. They found that BI at the age of 12 months was prospectively related to higher levels of fearful and avoidant

behavior at the age of 2.5 years in both social and nonsocial SR situations. However this effect was only found when mothers were present during the SR situations. Thus the relationship between BI and fearful and avoidant SR behaviors, is commonly measured in social situations, while the

relationship seems to be present in both social and nonsocial SR situations. However, this possible relationship is not yet examined in children aged 4.5 years.

Interestingly, studies that focus on the relation between BI and fearful and avoidant behaviors in SR situations mostly use child-mother dyads (e.g. De Rosnay et al., 2006; Murray et al., 2008), while the possible differential effect of fathers, compared to mothers, during SR might be especially interesting in the case of anxious behaviors. Bögels and Perotti (2011) attribute a unique role to fathers in the development of anxiety. They suggest that, from an evolutionary perspective, fathers are specialized in external protection (e.g. external world confrontations), while mothers are specialized in internal protection (e.g. providing food/comfort). Therefore the information signaled by fathers with respect to potential threats may instinctively be more influential on children

compared to information signaled by mothers in that same situation. In support of this idea, Bögels, Stevens, and Majdandzic (2011) found that in high BI children, aged 8-12 years old, fathers had more influence on their children’s level of fear, while in moderate or low BI children, mothers were more influential on the level of fear. This shows that when a child is anxious and feels the need for protection, thus when he/she has a high BI, the child seems to focus on the signals of the father. While when the need is less apparent, when the child has a moderate or low BI, the child focuses on the signals of the mother. In addition, the earlier finding of Aktar et al. (2014) revealing that BI at 12 months is only predictive of fearful/avoidant behavior at the age of 2.5 with mothers, indicates a differential effect of temperament on child behavior during SR situations with mothers compared to fathers. These results might show that paternal presence can stimulate a toddler to approach

unfamiliar objects, persons or situations. This is consistent with fathers’ assumed role of stimulating the child to experiment with experiences outside the family, relevant for the socialization of the child (Bögels & Phares, 2008). To summarize, there is some evidence supporting the idea that mothers and fathers have different roles, which can be dependent on the level of BI. However, whether this effect of gender is still apparent when BI is measured in toddlerhood and the fearful/avoidant behavior is measured at 4.5 is yet to be examined.

Overall, earlier research has shown that high BI is moderately stable from toddlerhood on and can have a heightening effect on anxious behaviors later in life. The stability of an inhibited temperament can for instance be explained by the provoking and maintaining effects high BI can have on particular parenting dimensions. Earlier research on the relationship between BI and fearful/avoidant behaviors in SR situations tends to focus on children in the age range on 10-20 months old, while the effect of parental behaviors in later life seems to be salient as well. Also the majority of SR study tends to focus on social SR situations, while the relationship between BI and fear/avoidance seems to be relevant in nonsocial SR situations as well. However this is yet only investigated in infants and toddlers. Finally, SR studies predominantly focus on mother-child dyads, while the differential effects of paternal behaviors in the case of anxiety might be particularly interesting. This study is the first in the developmental field of research that investigates the

possible relation between BI at the age of 2.5 and fearful/avoidant behaviors at the age of 4.5 in SR situations. In addition it investigates whether this possible relation is differentially affected by the presence of fathers or mothers, and/or by the nature of SR situations, which are either social or nonsocial.

The purpose of the present study is to investigate whether the level of fear and avoidance behaviors, measured at the age of 4.5, is affected by the level of BI, measured at the age of 2.5. Based on earlier findings it is hypothesized that children with high BI at the age of 2.5 would display more fearful and avoidant behaviors at the age of 4.5 compared to children with low BI. Second it is examined whether the possible relation between BI at the age of 2.5 and fearful and avoidant behaviors at the age of 4.5 is affected by the kind of the SR situations, which are either social or nonsocial. Based on earlier findings it is hypothesized that the association between BI and later fearful and avoidant behavior will hold across social and non-social SR situations. Third it is examined whether the association between early BI at the age of 2.5 and fearful and avoidant behavior at the age of 4.5 is affected by the gender of the parent present during the SR situation. Based on earlier findings it is hypothesized that fathers and mothers display a differential effect on

the relation between BI and later fearful and avoidant behavior. On an exploratory basis, it is investigated whether children tend to display different levels of fearful/avoidant during a social compared to a nonsocial situation. It is also explored whether this same possible differential effect is present with fathers compared to mothers.

This study is relevant for better knowledge about BI in toddlerhood and its possible prospective relationship with children's’ fearful and avoidant behavior at a later age. It’s also relevant for better knowledge about the possible differences between fathers’ and mothers’ effects on the child’s anxious behavior in social and nonsocial situations.

Method Participants

The sample consisted of 111 couples with their first-born children. In the children’s sample 53 were boys and 58 were girls. These families are part of a larger ongoing longitudinal study at the

University of Amsterdam on emotional and social development from the prenatal period to middle childhood. The measures of the larger (ongoing) longitudinal study include a prenatal measurement (not used in this study), a 4-month measurement (not use in this study), a 12-month measurement (Aktar et al., 2013, not used in this study), a 2.5 years-old measurement (Aktar et al., 2014), used for the BI in this study, and this 4.5 years-old measurement. The participating parents read and signed and informed consent prior to their participation. The age and relevant socio-demographic characteristics of the sample are shown in Table 1.

Table 1

Age and Socio-demographic variables of the sample

Mothers Fathers Girls Boys

Mean age (SD) 36.05 (4.20) 39.06 (5.42) 4.47 (0.16) 4.45 (.0.06)

Mean educational level (SD) * 7.12 (1.17) 6.65 (1.52) Mean professional level (SD) ** 8.77 (2.10) 8.30 (2.50)

Current working status (%) HK 4.7 1.9

FT 11.3 65.1

PT 75.5 29.2

Note. HK: Housekeeper, FT: Fulltime, PT: parttime. *Mean educational level range: 1 (primary education) to 8 (university). **Mean professional level range: 1 (manual labour which requires no education) to 11 (labour which requires a university degree).

Materials and Procedure

Each parent visited the lab separately when his/her child was about 2.5 and about 4.5 years old. During the visit at the age of 2.5 the children and their parents completed the set of BI-tasks. During both visits at the age of 4.5 the children and their parents completed both the social and the

nonsocial SR task. The order of the parental visits was randomized. In 48 % of the cases, the child visited the lab first with their mother and in 52 % of the cases the child visited the lab first with their father.

BI-measurement.

The toddlers’ level of BI was assessed at the age of 2.5 years-old measurements by the Lab-TAB (Goldsmith & Rothbart, 1996). During the 8 BI-measurements the toddlers were confronted with unfamiliar objects; a mechanic stunt-car and three unknown sorts of toys, two unfamiliar situations; two versions of a risk room, and two unfamiliar persons; a female stranger and a clown. The

mechanic stunt-car was regulated by the test-leader during the assessment. This car makes noises during the visits and rides along a standardized road. The three sorts of unknown toys are a dinosaur, a bug and a parrot. The risk room consists of unfamiliar combination of gymnastic toys

such as a tunnel, a mattress, and a balance beam. During the unfamiliar person situations the toddlers were expected to meet an unknown person and a clown. During the mother visit they completed the stunt-car-task, the stranger-task and a random version (A or B) of the riskroom. During the father visist they completed the clown-task and a random version (A or B) of the riskroom. During a house visit they completed the unfamiliar toys-tasks. During all of the

measurements the parents were present, but they sat behind the child and were instructed to stay as neutral as possible during these tasks. The order of the BI tasks was the same for all children. The BI-tasks were recorded and coded afterwards by trained coders. During each BI-task several kinds of toddler behaviors were coded such as the proximity to the parent, the latency of the first reaction, latency to touch the unknown objects/toys, startle response, facial, vocal, verbal and bodily fear and attempt to escape (see Goldsmith & Rothbart, 1996). The scores of all BI-measurements are

standardized for each task, and these scores are averaged and standardized into one final BI score. The inter-observer reliability for the children’s BI-variables was measured by calculating the intraclass correlations (ICC’s) between the coders in 19.96% of the recorders that were double-coded. The mean ICC for the BI-measurements was .91.

SR measurement.

During both the social and the nonsocial SR task the parents were actively present. During both tasks the child and their mother/father were in the Family-lab of the UvA. During the nonsocial SR task the children and their parents were confronted with a robot. This robot was about 40 cm’s high and is blue/red eyed and was colored white/silver with black. This task was divided in different time-intervals. During the first interval the parents were instructed to be silent and let their children react to the robot naturally without interruption. During the second interval the parents were

instructed to talk to their children about the robot for about one minute. During the third interval the parents were instructed to motivate their children to approach the robot and (if they dared) touch it for about one minute. During all these intervals the robot walked around in the room and made

noises. After these intervals the testleader came in with a toybed in which robot was put to sleep to end the task. First she set it off and then she putted it in the bed. She tried to motivate the child to put the blanket on the robot and, when possible, give it a nightkiss or sing a sleeping song for the robot.

During the social SR tasks the children and their parents were supposed to interact with a stranger. This stranger was trained to behave in a reliable and relaxed manner towards the parent and the child. In this situation the parent was seated in front of the stranger and the child was seated behind a table next to the parent. At the start of the task a female stranger entered the room. First the parent talked to the stranger for about two minutes. In the interaction the stranger asked

standardized questions about the parents’ personality. Next the child was asked to swap seats with the parent and converse with the stranger for about one minute. In this interaction the stranger asked standardized questions to the child about their preferences in playful situations or at school.

Afterwards the stranger thanked the parent and the child for the conversation and left the room. The level of fear and avoidance was measured during the social and nonsocial SR situations. Both situations were videotaped and coded by trained coders. The coding was done according to an adaptation of the protocol, which was originally used in the study of Murray et al. (2008). It was adapted for different age ranges and tasks for the study of Aktar et al. (2013) and Aktar et al. (2014) and this study. For the coding, each situation was divided in specific time intervals. The specific behaviors for every child/parent were coded on a 5-point scale, which captures the intensity and persistence of any given expressions. The final scores of each behavior will be the average of all time intervals. The coders were unaware of the child's BI score. For this research one pair of observers were trained to code the children’s behavior. 23% of the data was double-coded to measure the inter-observer reliability. The averages of double-coded recordings were used in the final data.

Child fear was captured in facial fear using the AFFEX scoring system (Affective

open mouth (where teeth can be shown) and lips. Second it was based on bodily manifestations of fear, such as freezing (Murray et al., 2008). Fidgeting may also be an indication of bodily distress like fear (Jabson, Venkatraman, & Dishion, 2003). Lastly vocal fear may be signaled. This can be shown for example in vocal expressions of fear, crying faces, and vocal expressions of distress.

The avoidant behavior was captured in children’s attempts to avoid contact with the stranger/robot. The avoidance behavior was defined by the same behavioral dispositions

defined/used by Ainsworth, Blehar, Waters, and Wall (1978). Avoidance was captured in active attempts to avoid the stranger/robot either by physically withdrawing or turning away or trough gaze avoidance (Murray et al., 2008).

The inter-observer reliability was measured by calculating the ICC’s between the two coders in the 23% double coded files. The mean ICC for the children-variables was .87. The ICC for the social tasks for fear was .76 and for avoidance was .95. The ICC for the nonsocial tasks for fear was .92 and for avoidance was .85.

Statistical Analysis

All hypothesis were tested with fixed-effects multilevel regression models consisting of a first child-level; the toddler BI-level which is a continuous variable, and a second level; the task-characteristics which can be either social/nonsocial and either with father/mother. This model is conducted for child fear and avoidance during the SR tasks. Task and parent gender were entered as dichotomous dummy variables in this model. For BI the standardized continuous score was used. All variables used in this model were sufficiently normally distributed; skewness and kurtosis were within the range of -2 and +2. The residuals, which were checked after the main analysis, were also normally distributed. The estimation method was maximum likelihood. The multilevel models were analyzed using IBM SPSS Statistics 20 (IBM).

A power analysis was conducted with Power Analysis in Two-level designs (PinT; Snijders & Bosker, 1993). The data from Aktar et al. (2014) and Aktar et al. (2013) was used as reference

variables because their research setup is comparable to mine. For the two-level design with one variable on the first level, two variables on the second level, and four groups in total an ideal sample size was calculated to be more then 59 per group. However, Maas and Hox (2005) found in their simulation study that only a sample size of 50 or less on the second level in multilevel analysis can lead to biased estimations. Therefore the power in my research was considered sufficient.

Results

In the study of Aktar et al. (2013) and Murray et al. (2008) avoidance and fear were treated as separate variables, because in Murray et al. (2008) there was no significant association between them and they have shown different patterns of results. In this study first the raw associations were estimated by calculating Pearson’s correlations r to check whether fear and avoidance were distinct outcome variables. There were very small associations between fear and avoidance: r = .16, p < .05. Given this findings and similar findings of Murray et al. (2008) and Aktar et al. (2013) avoidance and fear were treated as distinct outcome measures.

Next it was checked which covariance structure was the best fit for this data by using a Likelihood Ratio Test (LRT). Hereby the fit of a model with an Unstructured covariance structure (UN), which is the most complex, was compared to other more simple covariance structures (Compound Symmetry, Autoregressive-1) by comparing their -2 Log Likelihood (-2LL) on a χ²-distribution. The main effects fear- and avoidance-models were compared. For the fear-model the UN covariance structure was the best fit. It had a -2LL of 474.772, Akaike’s Information Criterion (AIC) was: 504.772 and Schwarz’s Bayesian Criterion (BIC) was: 566.074. For the avoidance-model the UN covariance structure was also the best fit as well: -2LL: 570.298, AIC: 600.298, BIC: 661.600.

Third it was checked whether the theoretical relevant interactions improved the fit of the model. The LRT was used to compare the several models and check the best fit. Every theoretical relevant interaction was included one by one and then it was checked whether their fit was better,

compared to a model with only main effects. Hereby the fit of the model with the interaction was compared to the main-effects-model by comparing their -2LL on a χ²-distribution. The fear-model with only main effects had a -2LL of: 474.772, AIC was: 504.772 and BIC was: 566.074. For the fear-model the inclusion of the interaction between Parent Gender and BI did not improve the fit of the model significant (-2LL: 474.306, AIC: 506.306, BIC: 571.694). Therefore it was left out of the final fear-model. The interaction between Task and BI did improve the fit of the model significant (-2LL: 460.750, AIC: 492.750, BIC: 558.139). Therefore it was included the final fear-model. Thus the final fear-model is with main effects of BI, task and parent gender and the interaction between task and BI.

The avoidance-model with only main effects had a -2LL of: 570.298, AIC was: 600.298 and BIC was: 661.600. For the avoidance-model both interactions between Parent Gender and BI (-2LL: 570.166, AIC: 602.166, BIC: 667.554) and Task and BI (-2LL: 468.435, AIC: 600.435, BIC:

665.823) did not improve the fit of the model significant. Therefore they were left out of the final avoidance-model. Thus the final avoidance-model is with main effects of BI, Task and Parent Gender and no interactions.

For the final fear-model the parameter estimates, standard errors and p-values per variable are shown in Table 2.

Table 2

Standardized parameters, standard errors and p-values for the fear-model

Variable β SE p

Fear model

Intercept 2.62 .06 .000*

Toddler BI -.29 .14 .040*

Parent Gender .02 .03 .542

Toddler BI*Task .45 .12 .000*

Note. Significant effects p < .05 are indicated with an asterisk.

The fear-model in Table 2 reveals that Toddler BI was significantly related to the level of fear at the age of 4.5, t (111.692) = -2.082, p < .05. This result showed that toddler BI and fear were significantly negatively associated: The higher the toddlers’ BI, the lower the fear-level at the age of 4.5. Visual inspection of this association revealed that the significant negative link is stemming from outliers at the higher end of the distribution of the standardized z-scores of toddler BI and fear. Once the outlying cases were removed, the association between fear and toddler BI ceased to reach significance. The other associations remained the same: Parent gender and fear were not

significantly related and task and fear and the interaction between toddler BI, task and fear were significantly related. We therefore do not interpret the results about fear and toddler BI further. The effect of parent gender on fear failed to reach significance. Children’s level of fear did not

significantly differ between fathers and mothers. Task and fear are significantly negatively

associated: The children show higher levels of fear during the social stranger task compared to the nonsocial robot task, their mean level of fear is shown in Table 3.

Table 3

Means and standard deviations (SD) of level of fear during the social and nonsocial task

Variable Mean SD

Social Task

Fear 2.39 .32

Nonsocial Task

Lastly it has become clear that there was a significant interaction between task and toddler BI in the fear-model; the effect of toddler BI on fear was different during the different sorts of tasks.

Figure 1. The interaction effect between BI (standardized scores on the x-axis) and level of fear

(standardized scores on the y-axis) per task. The blue line displays the distribution of fear and BI in the nonsocial task. The green line displays the distribution of fear and BI in the social task. The BI-range is from -1.01 (low) to 1.36 (high). Children with high BI showed more fearful behavior and children with low BI showed less fearful behavior during a nonsocial SR task, compared to a social SR task.

Figure 1 revealed that children with high BI at the age of 2.5 tended to show higher levels of fear at the age of 4.5 during the nonsocial robot task compared to the social stranger task. Next to that it was shown that children with low BI at the age of 2.5 displayed lower levels of fear at the age of 4.5 during the nonsocial robot task compared to the social stranger task. Overall it seemed that the level of toddler BI and its effects on the level of fear at the age of 4.5 was of more influence in the nonsocial task compared to the social task.

For the final avoidance-model the parameter estimates, standard errors and p-values per variable are shown in table 4.

Table 4

Standardized parameters, standard errors and p-values for the avoidance-model

Variable β SE p Avoidance model Intercept .72 .09 .000* Toddler BI .19 .07 .005* Task .72 .05 .000* Parent Gender .03 .03 .369

The avoidance-model in Table 4 revealed that toddler BI was significant related to the level of fear at the age of 4.5, t (111.692) = 2.837, p < .05. This result showed that toddler BI and

avoidance were significantly positively associated: The higher the toddlers’ BI, the higher the level of avoidance at the age of 4.5. In contrast, the effect of parent gender on avoidance failed to reach significance. It had no differential effect whether the father or the mother was present during the SR tasks. Task and fear were significantly positively associated. The children showed higher levels of avoidance during the nonsocial robot task compared to the social stranger task, see Table 5. Table 5

Means and standard deviations (SD) of BI and avoidance during the social and nonsocial task.

Variable Mean SD

Social Task

Avoidance 1.46 .46

Nonsocial Task

Avoidance 2.18 .54

To summarize, the results reveal that BI at toddlerhood was significantly positively associated with later avoidance. A higher BI in toddlerhood was related to a higher level of avoidance at the age of 4.5 during the SR tasks with either the mother or the father present. In the fear-model the association between toddler BI and fear was influenced by the sort of task; children with high BI tended to show higher levels of fear during the nonsocial robot task compared to the social stranger task, while children with low BI displayed lower levels of fear during the nonsocial robot task compared to the social stranger task. Overall it seemed that the level of toddler BI and its effects on level of fear at the age of 4.5 was of more influence in the nonsocial task compared to the social task. The possible influence of task on the association between toddler BI and avoidance and

the possible influence of parent gender on toddler BI and fear and avoidance could not be shown, because these interactions did not improve the model fit and thus were not significant.

The exploratory analyses revealed that children displayed higher levels of fear during the social stranger task compared to the nonsocial robot task. Next to that it was found that children displayed higher levels of avoidance during the nonsocial robot task compared to the social stranger task. Lastly, in both the fear- and the avoidance model parent gender failed to reach significance; during the tasks the level of fear or avoidance was equal when the father was present compared to when the mother was present.

Discussion and Conclusion

This research has shown that children who had a higher level of inhibited temperament in

toddlerhood displayed more avoidance at the age of 4.5 during SR situations compared to children who had a lower level of inhibited temperament in toddlerhood. Next to that it was found that children with a higher level of inhibited temperament in toddlerhood showed more fear at the age of 4.5 during a SR situation with a nonsocial stimulus, in this case the robot, compared to a SR

situation with a social stimulus, in this case the stranger. This was the case while children with a lower level of inhibited temperament in toddlerhood displayed less fear at the age of 4.5 during the SR situation with a nonsocial stimulus compared to the SR situation with a social stimulus. Overall it seemed that the level of inhibited temperament was of more influence during the SR situation with the nonsocial stimulus compared to the SR situation with the social stimulus. The sort of SR situation did not have a differential effect on the relation between level of inhibited temperament during toddlerhood and level of avoidance at the age of 4.5. There was also no difference in level of fear and avoidance at the age of 4.5 and its relation with level of inhibited temperament in

toddlerhood during the SR situations when the father was present compared to when the mother was present. However, there was a difference in level of fear and avoidance during the SR situations: Children displayed more fearful behavior during the SR situation with the social stimulus compared

to the SR situation with the nonsocial stimulus, while children showed more avoidance during the SR situation with the nonsocial stimulus compared to the SR situation with the social stimulus. Lastly there was no difference in level of fear or avoidance when the father was present during the SR situations compared to when the mother was present during the SR situations.

The major finding of this research is that BI in toddlerhood, measured at the age of 2.5, seemed to be prospectively associated with later avoidance at the age of 4.5. Children with a higher level of BI in toddlerhood showed more avoidance in early childhood during SR situations

compared to children with a lower level of BI in toddlerhood. This was in line with the expectations and earlier findings. Likewise for example, Muris, van Brakel, Arntz and Schouten (2011) also found in their longitudinal study that high BI was a specific risk factor for developing social anxiety symptoms, like avoidance, in children aged 5 to 8 years. Thus, children who behaved inhibited towards unfamiliar objects, situations or persons in toddlerhood when their parents were not

actively present, showed a tendency to avoid later unfamiliar objects, situations and persons in early childhood when their parents were actively present.

The second major finding was that the level BI in toddlerhood was more strongly related to the level of fear of children aged 4.5 during the SR situation with the nonsocial stimulus compared to the SR situation with the social stimulus. Thus when a child was behaviorally inhibited during his/her toddlerhood, this tended to heighten his/her fearful behavior during nonsocial situations in early childhood, while this influence of BI was less during social situations. This finding was

remarkable because earlier findings (e.g. Claus & Blackford, 2012, Schwartz et al., 1999, and Muris et al., 2011) have shown that BI was significantly related to anxiety symptoms in the social field. These findings suggest that BI would be of reasonable influence in social situations. One reasonable explanation for this contrary finding is that children display more socially desirable behavior

towards a stranger (being the social situation), compared to a robot (being the nonsocial situation). At the age of 4.5 they might have felt the urge to behave socially as ’normal’ as possible towards a stranger because that might be expected of them by that age, while they did not feel this urge

towards a robot. It can be the case that children therefore displayed less fearful behavior during social situations in this research, compared to nonsocial situations. Thus, it could be that other factors play a significant role during the social situation, compared to the nonsocial situation, which could have influenced the effect of temperament on earful behavior. However, this needs to be investigated in future research.

One other explanation for this unexpected finding could be that the level of avoidance and fear were divided constructs in this research. It might be that if children avoid rather than confront the unfamiliar object/person, they are less likely to experience fear because they increase the distance between them and the unfamiliar/fearful stimulus. In addition, the coding scheme of avoidance might rule out fearful behavior by definition. This is because when the coders saw extreme avoidance, fear was more difficult or even impossible to see by definition: when a child turns his/her back towards a robot or a person, it was clear avoidance, but you could not code fear by definition because the coders could not see the face/body language of the child for example. Thus, by the way the coding scheme was defined the extreme avoidance behavior could rule out the fear behavior until a certain level. This might have influenced the association between BI and fear. However, in the less extreme cases of avoidance (e.g. gaze avoidance, putting legs upwards) fear could still be coded, so this effect could only happen in the more extreme cases of avoidance. It could be interesting in future research to measure the parent behaviors and define fear/avoidance as a single construct (e.g. anxiety behaviors overall), as did Aktar et al. (2014).

The finding that a child’s fearful and avoidant behavior during SR situations was

independent of which parent (father or mother) was present during the SR situation was contrary to the expectations. The fact that the interaction between parental gender, BI and fearful/avoidant behaviors did not improve the model fit is probably due to insignificant results. Thus, contrary to expectations, this study did not find any differential effect of parental gender on BI and its

prospective relation with fearful/avoidant behaviors. Earlier findings assumed a differential role and effect of fathers and mothers on their children’s behavior (e.g. Bögels & Phares, 2008). And earlier

findings also presumed a role of BI in this differential effect; Children with high BI tend to focus on their fathers’ signals, while children with low BI tend to focus on their mothers’ signals (Bögels et al., 2011). The unexpected findings in this study can be explained by the fact that the children in this sample had a relatively average level of BI in toddlerhood and therefore little variation in their level of BI. In Table 4 it is shown that the mean level of BI of the sample is .018. The BI in

toddlerhood ranges from -1.01 (low) to 1.36 (high), thus .018 is relatively medium. Thus it could be that because of their overall average and little variation in level of BI in toddlerhood the children experienced less apparent need for one of their parents at all because they did not experience extreme temperamental anxiety overall. In addition, Kagan et al. (1987) have proposed that high BI-children are ’qualitatively’ different from average BI-BI-children. Therefore they stated that

identifying the extreme groups, and measure behavior of only high BI children for example, was the best way to do research with this kind of sample and the way to explore BI-related behaviors. Thus, the non-significant effect of parental gender on BI and/or level of fear/avoidance can be explained by the limited variation around the average level of BI in this sample. In line with these results, Aktar et al (2013) did not find a differential effect of parental gender on the level of fear/avoidance with babies during SR situations as well.

The exploratory finding that the children tended to display more fear during the social situation can be explained by the setup of the SR situations. During the social situation the children were expected to converse with a stranger. Hereby the stranger displayed social attention and was an active initiator of a conversation and was therefore more difficult to avoid compared to the robot. Thus the children were actively motivated by the stranger to talk to them. During the nonsocial situation the children were motivated by their parents to approach the robot, but if they did not approach, or were avoiding the robot, there were no active consequences, except for the possible reaction of the parent. While during the social situation, the stranger would be even more pursuing to talk to the child (until a certain limit). The children might have experienced less liberty to converse to the stranger, compared to approaching/being relaxed around the robot, because an

unknown person initiated the contact instead of known person like the parent. Therefore they might have experienced more fear during this social situation compared to the nonsocial situation, because they might have felt more urge to talk to the stranger, which can have heightened their level of fear because they might have do something which can they found scary. Therefore the level fear could be higher during social situations compared to nonsocial situations.

The exploratory finding that the children tended to show more avoidance during the nonsocial situation compared to the social situation can also be explained by the setup of the SR situations. During the nonsocial situation the children were free to move around the room and thus to avoid the robot actively. This was made explicit during the instruction of the task. While during the social situation they were expected to take a seat facing the stranger/at a table. Thus their active movements were limited during this situation. They could move around the room if they wanted, or refuse to sit on the chair opposite to the stranger, but it was not expected or made explicit that they were allowed to do so. Even though the avoidance behavior was coded not only in active physical avoidance, but also in gaze avoidance/turning away from the stranger, the children may have felt less freedom to do so during the social situation compared to the nonsocial situation. And, as stated earlier, extreme avoidance might have ruled out the coding of fear during the nonsocial SR

situation. Thus, the setup and instructions during the nonsocial and social situation may have influenced the level of avoidance during these different situations.

An interesting direction for future research might be to assess BI not as unitary construct, as is done in this study and earlier studies (e.g. Aktar et al. 2013, Aktar et al., 2014, Murray et al., 2008), but to assess it by defining a social part of the behaviors and a nonsocial part of the BI-behaviors. Dyson et al. (2011) found for example that BI in social contexts (in this study the unfamiliar persons: stranger or clown) and BI in nonsocial contexts (in this study the objects and situations) were uncorrelated constructs. Social and nonsocial BI even exhibited distinct patterns of relations with parental reports of anxiety behaviors and temperament. These results suggest that BI is a heterogeneous construct. An interesting future direction might be to measure whether social BI

is related to fearful/avoidant behaviors specifically in social SR situations and nonsocial BI is related to fearful/avoidant behaviors during nonsocial SR situations and not vice versa. If so, this might also partially explain negative association of toddler BI and later fearful behavior. If the BI construct consists of two parts, future research might find the expected positive associations if BI is divided accordingly.

The study has some limitations and therefore the findings should be interpreted accordingly. One limitation in that the findings in this study are correlational; this study is longitudinal by nature, but not experimental. Therefore no causal relations could be found between the measured constructs in this study. Future research might include an exmperimental setup, which can make causal

inferences about the relation between BI and later fear/avoidance for example. One other limitation is that other constructs might have influenced the level of BI at the age of 2.5 or the level of

fear/avoidance at the age of 4.5, which are not taken into account in this study. Ladd and Pettit (2002; cited in Murray et al., 2008) found for example that socialization practices such as whether the child is regularly exposed to unknown persons (e.g. by attending day-care) might play an important role in the measured constructs in this research. Future research might take these, and possible other important factors, into account in measuring the relation between BI in toddlerhood and fearful/avoidant behaviors in early childhood.

To summarize, children who behaved inhibited in toddlerhood avoided unfamiliar situations with persons as well as objects more in early childhood compared to children who behaved

uninhibited in toddlerhood. Children with an inhibited temperament in toddlerhood showed more fearful behavior at the age of 4.5 and children with an uninhibited temperament in toddlerhood showed less fearful behavior at the age of 4.5 during a nonsocial SR situation, compared to a social SR situation. The nature of the situations influenced the level of fear and avoidance overall;

children at the age of 4.5 showed higher levels of fear during SR situations with a social stimulus, while during the SR situation with a nonsocial stimulus avoidance behavior was more observed. Overall BI seemed to be prospectively associated with important social and nonsocial avoidance

behaviors that can play an important role in children’s development. Parental and environmental support and encouragement could be very important for inhibited children to avoid negative future consequences because of their temperament.

Major Findings

• Toddler’s BI was positively associated with avoidance in early childhood; children with high BI at the age of 2.5 tended to show higher levels of avoidance at the age of 4.5 and vice versa.

• Toddler’s BI was negatively associated with fear in early childhood; children with high BI at the age of 2.5 tended to show lower levels of later fear at the age of 4.5 and vice versa, however these results seemed to be the consequence of leverage and therefore were not interpreted accordingly.

• There seemed to be no differential effect of the parent present on the association between BI and fearful/avoidant behavior.

• The association between BI in toddlerhood and fearful behavior in early childhood was dependent of the type of SR situation: Children with high BI at the age of 2.5 showed more fearful behavior at the age of 4.5 and children with low BI at the age of 2.5 showed less fearful behavior at the age of 4.5 during a nonsocial SR situation, compared to a social SR situation.

• Children showed overall more avoidance during a nonsocial SR situation and they show more fearful behavior during a social SR situation.

• There seemed to be no differential effect of the gender of the parent present on the level of fear and avoidance during the SR situations.

Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A

psychological study of the Strange Situation. Hillsdale, NJ: Erlbaum.

Aktar, E., Majdandzic, M., de Vente, W., & Bögels, S. M. (2013). The interplay between expressed parental anxiety and infant behavioural inhibition predicts infant avoidance in a social referencing paradigm. Journal of Child Psychology and Psychiatry, 54, 144–156.

Aktar, E., Majdandzic, M., de Vente, W., & Bögels, S. M. (2014). Parental social anxiety disorder prospectively predicts toddlers’ fear/avoidance in a social referencing paradigm. Journal of

Child Psychology and Psychiatry, 55, 77-87.

Belsky, J., & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to environmental influences. Psychological Bulletin, 135, 885–908.

Bögels, S. M., & Perotti, E. C. (2011). Does father know best? A formal model of the paternal influence on childhood social anxiety. Journal of Child and Family Studies, 20, 171–181. Bogels, S. M., & Phares, V. (2008). Fathers’ role in the etiology, prevention and treatment of child

anxiety: A review and a new model. Clinical Psychology Review, 28, 539–558.

Bögels, S. M., Stevens, J., & Majdandzic, M. (2011). Parenting and social anxiety: fathers’ versus mothers’ influence in their child’s anxiety in ambiguous social situations. Journal of Child

Psychology and Psychiatry, 52, 599-606.

Clauss, J. A., & Blackford, J. U. (2012). Behavioral inhibition and risk for developing social anxiety disorder: a meta-analytic study. Journal of the American Academy of Child and

Adolescent Psychiatry, 51, 1066–1075.

Di Nardo, P.A., Brown, T.A., & Barlow, D.H. (1994). Anxiety disorders interview schedule for

DSM-IV: Lifetime version (ADIS-IV-L). San Antonio, TX: Psychological Corporation.

Dyson, M. W., Klein, D. N., Olino, T. M., Dougherty, L. R., & Durbin, E. C. (2011). Social and nonsocial behavioral inhibition in preschool-age children: Differential associations with parent-reports of temperament and anxiety. Journal of Child Psychiatry and Human

Emde, R.N. (1992). Social referencing research: Uncertainty, self and search for meaning. In S. Feinman (Ed.), Social referencing and the social construction of reality in infancy (pp. 79– 95). New York: Plenum Press.

Feinman, S. (1982). Social referencing in infancy. Merrill-Palmer Quarterly, 28, 445–470.

Feinman, S., Roberts, D., Hsieh, K., Sawyer, D., & Swanson, D. (1992). A critical review of social referencing in infancy. In S. Feinman (Ed.), Social referencing and the social construction

of reality in infancy (pp. 4–15). New York: Plenum Press.

Gerull F. C., Rapee R.M. (2002). Mother knows best: effects of maternal modelling on the

acquisition of fear and avoidance behavior in toddlers. Behavior Research and Therapy 40, 279–287.

Goldsmith, H. H., & Rothbart, M. K. (1996). The laboratory temperament assessment battery,

locomotor version (manual). Technical Manual, Department of psychology, University of

Wisconsin, Madison, WI, USA.

Ingram, R.E., & Luxton, D. (2005). Vulnerability-stress models. In B. L. Hankin & J. R. Z. Abela (Eds.), Development of psychopathology: A vulnerability-stress perspective (pp. 32– 46). New York: Sage.

Izard. C. (1982). A system for identifying affect expressions by holistic judgements (AFFEX). Newark, DE: Instructional Resources Center.

Jabson, J., Venkatraman, S., & Dishion, T. (2003). The simple affect coding system. Unpublished, available from Child & Family Center, 195 West 12th Ave., Eugene, OR.

Kagan, J., & Moss, H. A. (1962). Birth to maturity: A study in psychological development. London: Wiley.

Kagan, J., Reznick, J. S., & Snidman, N. (1987). The physiology and psychology of behavioral inhibition in children. Child Development, 58, 1459-1473.

Kagan, J., Reznick, J. S., & Snidman, N. (1989). Issues in the study of temperament. In G. A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 133-144). London: Wiley.

Kagan, J., Reznick, J. S., Clarke, C., Snidman, N., & Garcia-Coll, C. (1984). Behavioral inhibition to the unfamiliar. Child Development, 55, 2212-2225.

McLeod, B. D., Wood, J. J., & Weisz, J. R. (2007). Examining the association between parenting and childhood anxiety: A meta-analysis. Clinical Psychology Review, 27, 155-172.

Maas, C. M. J. & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling. Methodology,

1, 86-92.

Muris, P., van Brakel, A. M. L., Arntz, A., & Schouten, E. (2011). Behavioral inhibition as a risk factor for the development of childhood anxiety disorders: An longitudinal study. Journal of

Child and Family Studies, 20, 157-170.

Murray, L., Creswell, C., & Cooper, P.J. (2009). The development of anxiety disorders in childhood: An integrative review. Psychological Medicine, 39, 1413–1423.

Murray, L., De Rosnay, M., Pearson, J., Bergeron, C., Schofield, E., Royal-Lawson, M., & Cooper, P. J. (2008). Intergenerational transmission of social anxiety: The role of social referencing processes in infancy. Child Development, 79, 1049–1064.

Rosenbaum, J. F., Biederman, J., Bolduc-Murphy, E. A., Fara-one, S. V., Chaloff, J., Hirshfeld, D. R., & Kagan, J. (1993). Behavioural inhibition in childhood: A risk factor for anxiety disorders. Harvard Review of Psychiatry, 1, 2–16.

de Rosnay, M., Cooper, P. J., Tsigaras, N., & Murray, L. (2006). Transmission of social anxiety from mother to infant: An experimental study using a social referencing paradigm.

Behaviour Research and Therapy, 44, 1165–1175.

Rubin, K. H., Burgess, K. B., & Hastings, P. D. (2002). Stability and social-behavioral consequences of toddlers’ inhibited temperament and parenting behaviors. Child

Rubin, K. H., Stewart, S. L., & Coplan, R. J. (1995). Social withdrawal in childhood: conceptual and empirical perspectives. In T. H. Ollendick & R. J. Prinz (Eds.), Advances in Clinical

Psychology (pp. 157-196). New York: Plenum.

Schwartz, C.E., Snidman N.,Kagan J. (1999). Adolescent social anxiety as an outcome of inhibited temperament in childhood. Journal of American Academic Child Adolescence Psychiatry

38, 1008–1015.

Snijders, T. A. B. & Bosker, R. J. (1993). Standard errors and sample sizes in two-level research.

Journal of Educational Statistics, 18, 237-260.

White, L. K., Helfinstein, S. M., & Fox, N. A. (2010). Temperamental factors associated with the acquisition of information processing biases and anxiety. In J. A. Hadwin, & A. P. Field (Eds.), Information processing biases and anxiety: A developmental perspective (pp. 233-252). West- Sussex: Wiley-Blackwell.

Woordruff-Borden, J., Morrow, C. Bourland, S., & Cambron, S. (2002). The behavior of anxious parents: examining mechanisms of transmission of anxiety from parent to child. Journal of

Clinical Child and Adolescent Psychology, 31, 364-374.

Zuckerman, M. (1999). Vulnerability to psychopathology: A biosocial model. Washington, DC: American Psychological Association.