Observer-Rated Environmental Sensitivity Moderates Children’s Response to Parenting Quality in Early Childhood

Abstract

According to several developmental theories some children are more sensitive to the quality of their environment than others, but most supporting empirical evidence is based on relatively distal markers of hypothesized sensitivity. This study provides evidence for the validity of behaviorally observed Environmental Sensitivity as a moderator of parenting effects on children’s early development in a sample of 292 children (Mage = 3.74; SD = 0.26) and their mothers. Sensitivity was coded using a newly developed observational measure for the specific and objective assessment of Environmental Sensitivity, the Highly Sensitive Child-Rating System (HSC-RS). HSC-RS factorial structure, associations with temperament traits, and interactions with parenting quality in the prediction of socio-emotional child outcomes are reported. Findings supported a one-factor solution. Observed sensitivity was relatively distinct from observed temperament and interacted with both low and high parenting quality in the development of behavior problems and social competence at ages three and six.

Parental care is one of the strongest, most robust and consistent predictors of child development, shaping social, emotional and cognitive development (Bernier, Calkins, & Bell, 2016; Fay-Stammbach, Hawes, & Meredith, 2014; Groh, Fearon, IJzendoorn, Bakermans-Kranenburg, & Roisman, 2017). However, several concepts inspired by evolutionary theory (Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007; Belsky & Pluess, 2009; Boyce & Ellis, 2005) suggest that children may be differentially susceptible to developmental effects of the environment, including parenting quality, with some being more sensitive and responsive than others to both negative and positive parenting practices. A vast array of behavioral, physiological, and genetic variables has been identified in empirical studies as markers of such sensitivity to environmental influences (Belsky & Pluess, 2009; Pluess & Bartley, 2015). Among these, Negative Affect and Difficult Temperament in infancy represent the most extensively studied susceptibility traits with strong empirical support for their moderating effects regarding the influence of parenting quality on children’s development (Belsky & Pluess, 2009; Slagt, Dubas, Dekovic, & van Aken, 2016). However, these measures tend to focus exclusively on negative emotions and easily observed associated behaviors such as crying or fearfulness—usually based on caregiver reports—and may therefore not fully capture the hypothesized construct of Environmental Sensitivity (Pluess, 2015; Pluess et al., 2018). Recently, a new self-report measure, the Highly Sensitive Child (HSC) scale (Pluess et al., 2018), which has also been applied in a parent-report format (Slagt, Dubas, van Aken, Ellis, & Dekovic, 2018), has proven helpful in capturing different levels of Environmental Sensitivity in children, allowing assessment of their response to both negative and positive environments, such as their response to the quality of parenting (Scrimin, Osler, Pozzoli, & Moscardino, 2018; Slagt et al., 2018) and school-based psychological programs (Nocentini, Menesini, & Pluess, 2018; Pluess & Boniwell, 2015). The aim of the current study was to overcome established limitations of questionnaire measures by developing an observational rating system of Environmental Sensitivity that is suitable for pre-schoolers, and then testing whether such observationally rated sensitivity predicts individual differences in response to both negative and positive aspects of parenting.

The Environmental Sensitivity Framework

According to computational simulation (Wolf, van Doorn, & Weissing, 2008), and empirical studies (Bakermans-Kranenburg & van IJzendoorn, 2011; Belsky & Pluess, 2011; Lionetti et al., 2018; Pluess et al., 2018), individuals vary in their sensitivity to environmental factors with a minority of the population, between 20% to 30%, being particularly sensitive to the quality of their environment. Since the early 1990s, several researchers have investigated such individual differences in sensitivity to environmental influences and have recently been integrated into the broader framework of Environmental Sensitivity (Pluess, 2015). Traditionally, one of the first concepts for the study of individual differences in response to environmental factors has been the Diathesis-Stress model (Monroe & Simons, 1991), according to which increased sensitivity constitutes a vulnerability, making children disproportionately susceptible to adverse rearing conditions. In more recent years, new concepts have emerged proposing that more sensitive children may not only be more negatively affected by adverse environmental conditions, but also benefit disproportionately from the influence of positive environments, as formulated in the concepts of Differential Susceptibility (Belsky & Pluess, 2009), Biological Sensitivity to Context (Boyce & Ellis, 2005), and Sensory Processing Sensitivity (Aron & Aron, 1997). Differential Susceptibility specifically posits that highly sensitive children would show higher sensitivity to both negative and positive environments, based on the evolutionary argument that low and high developmental plasticity represent alternative strategies maintained by natural selection (Belsky, Hsieh, & Crnic, 1998; Belsky & Pluess, 2009). Biological Sensitivity to Context focuses on differences in physiological reactivity (e.g. arterial pressure, cortisol reactivity) to environmental stimuli (Boyce & Ellis, 2005), and, in accordance with the notion of conditional adaptation, proposes that heightened sensitivity can develop in response to either very negative or very positive environments (Del Giudice, Ellis, & Shirtcliff, 2011; Ellis & Boyce, 2011). Finally, Sensory Processing Sensitivity posits that individual differences in Environmental Sensitivity are due to deeper processing of sensory input (Acevedo, Aron, Pospos, & Jessen, 2018; Aron & Aron, 1997). Importantly, all three models propose that some individuals are more sensitive to both the negative effects of adversity as well as the positive effects of supportive experiences. Whereas the “dark side” of such Differential Susceptibility is represented by the Diathesis-Stress model, its “bright” side (i.e. positive outcomes in response to nurturing and positive environments) is captured by the more recently proposed model of Vantage Sensitivity (Pluess & Belsky, 2013), which describes the propensity of some individuals to benefit disproportionately from positive environmental conditions, such as intervention programs (de Villiers, Lionetti, & Pluess, 2018; Pluess & Belsky, 2013). Although the different theories feature slightly different perspectives on sensitivity to environmental influences, Differential Susceptibility (together with Biological Sensitivity to Context and Sensory Processing Sensitivity), as well as Diathesis-Stress and Vantage Sensitivity can all be integrated into a broader framework of Environmental Sensitivity (Pluess, 2015), given that all describe that individuals differ in their sensitivity to environmental quality, either to negative, positive, or both (Belsky & Pluess, 2009; Belsky & van IJzendoorn, 2017; Boyce & Ellis, 2005; Lionetti et al., 2018; Pluess et al., 2018; Wolf et al., 2008).

Measures of Environmental Sensitivity

Temperament

The two temperament traits most extensively studied as markers of heightened sensitivity to parenting in infancy and toddlerhood are Negative Emotionality and Difficult Temperament (Belsky & Pluess, 2009). With negative affect as one of the key characteristics, both Negative Emotionality and Difficult Temperament are generally considered vulnerability factors from a perspective of Diathesis-Stress, but recent empirical evidence appears to point towards a for-better-and-for-worse effect. For example, Difficult Temperament in infancy has been shown to moderate the impact of both high and low parental sensitivity on various developmental outcomes, including behavior problems and cognitive competence at age three (Poehlmann et al., 2011), teacher-rated behavior problems at first grade (Bradley & Corwyn, 2008), reading and social competence in fifth grade (Pluess & Belsky, 2010), and socioemotional functioning in sixth grade (Pluess & Belsky, 2010). Empirical support for the moderating effect of Negative Emotionality to both negative and positive environmental contexts, however, appears to be restricted to infancy and toddlerhood (Slagt, Dubas, et al., 2016). From middle childhood onwards, more negatively emotional children tend to become exclusively vulnerable to negative experiences in line with a Diathesis-Stress perspective (Kopala-Sibley et al., 2016; Slagt, Dubas, et al., 2016; Slagt, Semon Dubas, & Aken, 2016). One explanation for this phenomenon may be that sensitive infants, initially indicated by high Negative Emotionality, that are reared in low-quality environments develop into children that are disproportionately more vulnerable to the negative effects of adversity with reduced sensitivity to positive experiences. Sensitive infants raised by emotionally available and supportive parents, on the other hand, may be less likely to express Negative Emotionality when they are older (Perry et al., 2013). Hence, Negative Emotionality may not be the best suited phenotype as a marker of sensitivity to both positive and negative environmental influences across development.

Sensory Processing Sensitivity

A promising phenotype that has been shown to capture sensitivity to both negative and positive stimuli in both children and adults is the individual trait of Sensory Processing Sensitivity (Aron, 2002; Aron, Aron, & Jagiellowicz, 2012; Pluess & Boniwell, 2015; Slagt et al., 2018). The concept of Sensory Processing Sensitivity has originally been developed to explain individual differences in adults regarding sensory sensitivity and the extent to which information is processed (Aron & Aron, 1997). Individuals scoring high on Sensory Processing Sensitivity appear to be particularly sensitive to subtle stimuli, they tend to be generally more aware of their environment, are more easily disturbed by even mildly negative environmental inputs, but also more likely to appreciate positive exposures (Aron et al., 2012). Based on Sensory Processing Sensitivity literature, this in-depth processing of sensory stimuli may represent a core mechanism underlying Environmental Sensitivity.

Sensory Processing Sensitivity can be assessed in adults with the self-report Highly Sensitive Person (HSP) scale (1997) and in school-aged children with the Highly Sensitive Child (HSC) scale (Pluess et al., 2018), which has also been shown to perform well as an adapted parent-rated version for younger children (Slagt et al., 2018). Both the HSC and HSP scales have been found to capture sensitivity to both negative and positive environmental aspects, in accordance with the Environmental Sensitivity framework (Greven et al., 2019). Although a parent-report sensitivity measure for children has initially been developed (Aron, 2002) and recently tested for its psychometric proprieties (Boterberg & Warreyn, 2016), this measure has not yet been empirically validated as capturing individual differences in Environmental Sensitivity.

The HSC scale includes 12 items which capture sensitivity as manifested in an increased appreciation for positive environmental stimuli and subtleties (e.g., some music can make me really happy; I notice when small things have changed in my environment, reflecting the Aesthetic Sensitivity subscale of the measure), in a stronger feeling of being overwhelmed when exposed to potentially adverse experiences (e.g., I am annoyed when people try to get me to do too many things at once, reflecting the Ease of Excitation subscale), and in lower sensory threshold (e.g., Loud noises make me uncomfortable, reflecting the Low Sensory Threshold subscale). According to empirical studies, items included in the HSC scale, load on the three mentioned subscales, as well as a general unitary sensitivity factor (Pluess et al., 2018). Importantly, the scale has been found to reflect sensitivity to various contextual factors such as intervention programs and parenting quality (Pluess & Boniwell, 2015; Scrimin et al., 2018; Slagt et al., 2018). As recently summarized in a meta-analysis, the sensitivity trait measured with the HSC/HSP scale is relatively distinct from other common individual traits although it is correlated with several traits, most strongly with Neuroticism/Behavior Inhibition (r = .42) (Lionetti et al., 2019). Importantly, established temperament traits such as Negative Affect, Positive Affect, Effortful Control, Behavior Inhibition and Behavior Activation were found to explain in total 31% of the variability in the HSC score in children and the Big Five personality traits explained 15% of the variance of HSC in adolescents (Pluess et al., 2018). The detected overlap between sensitivity and other temperament traits may be due to sensitivity aspects that are reflected in various established temperament traits. For example, highly sensitive children get overstimulated more easily because they process information more deeply, which may result in more crying than their less sensitive peers (Aron, 2002). Hence, sensitivity is captured to some degree by the temperament traits of Negative Emotionality and Difficult Temperament, which comprises intense negative mood, high withdrawal in response to new situations and slow adaptation to changes (Thomas, Chess, Birch, Hertzig, & Korn, 1963). Highly sensitive children also need more time when approaching a new situation, which may manifest in more cautious behavior similar to the one described by the temperament trait of Behavioral Inhibition. However, although there are similarities between established temperament traits and sensitivity, the temperament traits that have been associated with Environmental Sensitivity in empirical studies tend to be biased by a focus on the “dark side” of sensitivity and often fail to capture positive aspects of sensitivity (i.e., Vantage Sensitivity).

The Highly Sensitive Child Rating-System

In the current paper we introduce a newly developed observer-rated measure for the assessment of sensitivity in children, the Highly Sensitive Child Rating-System (HSC-RS).¹ The HSC-RS, designed to assess sensitivity in children aged 3 to 5 years, is rooted within the broad perspective of the Environmental Sensitivity framework and drawing from the theory of Sensory Processing Sensitivity (Aron et al., 2012; Pluess, 2015). For the development of the new measure, we aimed to operationalize Environmental Sensitivity at the behavioral level by focusing on heightened sensory sensitivity and deeper processing of the environment.

The HSC rating system, whose development is described in the procedure section of this paper, is made up of 10 rating scales (Table 1) that code global behaviors associated with sensitivity. Each scale is in a Likert format ranging from 1 to 7, with higher scores reflecting higher levels of Environmental Sensitivity. The HSC rating system has been developed for the application to episodes from the Laboratory Temperament Assessment Battery, the Lab-TAB (Goldsmith, Reilly, Lemery, Longley, & Prescott, 1999). The Lab-TAB is a laboratory based procedure consisting of a series of episodes (33 in total) developed to elicit specific behavioral reactions which allow coding for a range of established temperament traits. The Lab-TAB scoring system combines the sum of the facial, bodily, and vocal manifestation of a specific behavior across each channel (Durbin, Hayden, Klein, & Olino, 2007), with global ratings of the behavioral trait using all the relevant behaviors during specific episodes (Dyson, Olino, Dunn, Goldsmith, & Klein, 2012). According to exploratory factor analysis of 12 laboratory episodes (11 adopted from the Lab-TAB and one episode designed afterwards) that tap into key temperament traits, five latent factors of temperament have been found in preschool children: Assertiveness/Sociability, Positive Affect/Interest, Dysphoria, Fear/Inhibition, and Constraint versus Impulsivity (Dyson et al., 2012). Importantly, from a perspective of Environmental Sensitivity, high levels of anger and sadness (which loaded on the Dysphoria factor) and fear (which loaded on the Fear/Inhibition factor) at age three, have been found to predict depressive symptoms and anxiety in children exposed to a hurricane at age nine (Kopala-Sibley et al., 2016). However, although the Lab-TAB allows coding for basic temperament traits which, to some extent, have been found to moderate the effects of environmental exposures, none of these traits specifically capture Environmental Sensitivity.

Table 1.

The Highly Sensitive Child Rating System (HSC-RS): LabTAB episodes and scales. For details on LabTAB episodes and on the rationale behind each HSC-RS scale please refer to the Appendix

LabTAB episodes	Scales
Risk room	1. Pause to check before exploring a new environment:  Assessing the degree to which the child shows cautious behavior when approaching the environment in combination with curiosity and interest
	2. Cautious and collaborative attitude towards the experimenter:  Evaluating the degree of the child’s cautious yet collaborative attitude in relation to a relatively new and unfamiliar adult
Tower of patience	3. Attending to experimenter’s directions:  Assessing the degree of the child’s reactive attitude rather than proactive attitude (i.e., the child listening to and not interrupting the experimenter’s directions)
Stranger approach	4. Compliance with the experimenter’s request:  Similar to scale 3
	5. Fearfulness in response to the stranger’s entrance:  Assessing fearfulness when encountering new situations, in this case, the entrance of a unfamiliar adult
Exploring new objects	6. Hesitancy paired with curiosity:  As scale 1
Pop-up snakes	7. Positive response/overexcitement:  Aims to capture positive emotional reactivity in response to a positive environmental experience
Transparent box	8. Attention to toys’ detailed features:  Measures the degree to which the child is investigating the toys taken out from the box by the experimenter
	9. Careful perseverance when trying to open the box:  Assessing the child’s degree of persistency in opening the box, keeping an adequate and respectful approach, due to increased attention for details. See also scale 8
Impossibly perfect children	10. Preference for (and commitment to drawing) beautiful circles:  Evaluates the commitment to draw beautiful circles (i.e., appreciation of aesthetics)

To the best of our knowledge, the HSC-RS is the first attempt to measure Environmental Sensitivity directly and observationally. Being based on the Lab-TAB procedure, a well-tested and widely used measure, the HSC-RS allows to use existing data to code for Environmental Sensitivity. In contrast to self-report and parent-rated measures of sensitivity, the HSC-RS allows for the assessment of sensitivity during early developmental periods when self-report measures can’t be applied, and it also permits the assessment of sensitivity in children when parents are not available or unable to accurately rate their child (as, for example, in the context of parental clinical disorders). Furthermore, the combination of observational sensitivity measures with both parent-report and child-report questionnaires (Boterberg & Warreyn, 2016; Pluess et al., 2018; Slagt et al., 2018) will allow multi-method and multi-informant approaches for the assessment of sensitivity in future studies, which will contribute to a more thorough understanding of children’s Environmental Sensitivity as well as its development across the life course.

The Current Study

The main objective of the current study was to create an observational measure for the assessment of different levels of Environmental Sensitivity in 3–5 year-old children. In order to do so we coded a selection of existing Laboratory Temperament Assessment Battery (Lab-TAB) videos (Goldsmith et al., 1999) from three-year-old children who participated in the Stony Brook Temperament Study (Klein & Finsaas, 2017). The goal was to create a psychometrically strong observational rating-system that captures sensitivity in a more specific way than established temperament traits. Though within the Sensory Processing Sensitivity theory sensitivity is considered a unitary trait (Aron et al., 2012), consequent psychometric analyses of self-report and parent-report scales suggest the existence of several factors (Greven et al., 2019). However, rather than trying to recreate these factors we aimed to capture the most relevant sensitivity behaviors that could be observed in the available selection of Lab-TAB episodes.

Concerning associations with observed temperament from the Lab-TAB, we expected the HSC-RS to positively correlate with Fear/Inhibition and Dysphoria, given the relatively strong and consistently reported association between HSC/HSP self-report questionnaires and Neuroticism/Behavior Inhibition (Greven et al., 2019). Additionally, because sensitive individuals process information more deeply and have a more cautious approach when approaching new environments (Aron et al., 2012), we anticipated a positive association between HSC-RS and Constraint, which captures behavioural regulation and impulse control. Finally, given that Positive Affect/Interest and Assertiveness/Sociability Lab-TAB factors fall under the broad umbrella of extraversion, we expected both to correlate negatively with our rating system.

Based on the hypothesis that sensitivity would be associated with a stronger response of children to both low and high parenting quality, we also tested in the same data set, in order to validate the new measure, whether observationally assessed sensitivity moderates the effects of parenting at age three on behavior problems and social competence at three and six years. Based on existing empirical findings (Belsky et al., 2007; Belsky & Pluess, 2009; Pluess et al., 2018), we expected that highly sensitive children would be more sensitive to the impact of both low and high parenting quality compared to their low-sensitive peers. Furthermore, we expected moderating results to remain stable after controlling for the observed temperament trait of Negative Affect, which has previously been reported to moderate the impact of negative events on negative outcomes in children from the same study (Kopala-Sibley et al., 2016).

Method

Participants

Participants of the current study were drawn from the longitudinal Stony Brook Temperament Study (SBTS), an ongoing study in the USA, aiming at investigating temperamental and environmental antecedents of various emotional disorders (Klein & Finsaas, 2017). The original dataset included 559 children with no significant medical conditions or developmental disabilities, resident in Long Island, NY, and who lived with at least one English-speaking biological parent (Dyson et al., 2012). In the current study we considered a random subset of 292 children (54% male, Mage = 3.7, SD = 0.26) who took part in the Lab-TAB procedure at age three. According to eta squared effect sizes, children included in the current analysis did not differ from the rest of the sample in terms of the age at assessment (η² =.00), nor in terms of observed temperamental traits coded with the Lab-TAB (Assertiveness/Sociability, (η² =.00); Dysphoria (η² =.00); Fear (η² =.00); Positive Affect/Interest, (η² =.00); and Constraint, (η² =.01). Due to attrition over time, the sample at age six included only 226 children (53% male). However, according to comparison between the 226 included and the 66 lost cases, we did not find any significant differences between the two groups regarding any of the outcome and predictor variables at age 3 (η² =.00 for Externalizing Behavior Problems; η² =.01 for Internalizing Behavior Problems and Social Competence and η² =.00 for Authoritative, Authoritarian, and Permissive parenting).

Procedure

To develop the rating system, we first defined, at a descriptive level, the prototypical behavioral profile of a highly sensitive child. In order to do so, we drew on the Sensory Processing Sensitivity literature according to which sensory sensitivity and deeper processing of environmental information are defined as key features of sensitivity (Aron, 2002). Informed by empirical findings of significant associations between self-reported sensitivity, Behavior Inhibition and Negative and Positive Affect (Pluess, 2015; Pluess et al., 2018), we expected that sensitive children would show longer latency before approaching new objects due to requiring more time to process the object and a stronger emotional reaction in response to both negative and positive environmental stimuli. Due to the hypothesized deeper processing of environmental information, we also expected higher initial inhibition in combination with greater interest for new stimuli. Regarding interpersonal relationships, we expected that sensitive children’s openness toward new stimuli would lead to a more cautious yet cooperative behavior in the relationship with an unfamiliar and friendly adult figure (i.e., the experimenter), and a more fearful reaction in response to strangers. Second, from this first broad characterization of the highly sensitive child profile, we moved toward the identification of specific behaviors that could be observed and coded to assess an individual’s degree of Environmental Sensitivity. In order to identify observable markers of sensitivity in the Lab-TAB procedure, we selected an initial sample of 10 cases from the Stony Brook Temperament Study, applying a stratified random procedure in order to achieve a subsample that represented the full observed range of Lab-TAB temperaments traits in the total sample (Dyson et al., 2012; Goldsmith et al., 1999). We then identified seven episodes out of the 12 available ones from the Stony Brook Temperament Study as suitable for the assessment of the proposed sensitivity behaviors and developed a set of 14 scales with Likert-scale options ranging from 1 to 7 to rate these behaviors. The 14 scales were applied to the seven identified episodes of the 10 selected children. A first manual was drafted and then adapted based on the further coding of a selection of additional 50 cases. Exploratory statistical analyses across the 60 coded cases suggested that four out of the 14 scales were highly correlated with other scales in the rating system (Pearson’s r ranging from .81 to .91). Hence, we reduced the final number of scales to 10 (summarized in Table 1 and in the appendix), by excluding these 4 scales. We applied the 10-scale rating system to an additional set of randomly selected 101 cases, reaching a sample of 161 cases which we used to test the psychometric proprieties with an exploratory approach and estimate inter-rater agreement with an independent trained coder. Inter-rater agreement based on 59 cases (37% of the sample) was highly satisfactory for the overall mean score (ICC = .91 [.85 −.94]), and ranged from ICC = .57 for episode two to .89 for episode three (see supplementary information). Finally, we added 131 randomly selected additional cases, reaching a final sample of 292 cases which we applied a confirmatory factor analysis to, estimated associatiExons with temperament, and compared individual-environment interaction models to explore the newly developed measure’s ability to capture increased sensitivity to the environment.

Institutional Review Board approval for the current study was obtained from Stony Brook University (study name: Observations of Active and Inactive Children, protocol number: 88933–35).

Measures

Child Temperament

Children’s temperament was coded with an adapted version of the Laboratory Temperament Assessment Battery (Lab-TAB) (Goldsmith et al., 1999) when children were aged three. In the Stony Brook Temperament Study, the Lab-TAB procedure lasted approximately 2 hours and included a standardized set of 11 episodes adopted from the Lab-TAB, and one episode designed specifically for the SBTS. Five empirically derived Lab-TAB dimensions of temperament (Assertiveness/Sociability, Dysphoria, Fear/Inhibition, Positive affect/Interest and Constraint; see Dyson et al. (2012)), together with a Negative Affect scale, which represents a composite of fear, sadness and anger, were included in the analysis. Psychometric evidence in support of the five factor structure of temperament in this dataset are reported more in detail elsewhere, together with a more detailed description of Lab-TAB episodes (Dyson et al., 2012).

Parenting

Parenting was assessed when children were aged three using the Parenting Styles and Dimensions Questionnaire (PSDQ) (Robinson, Mandleco, Olsen, & Hart, 2001), a parent-report measure widely used in large studies and across many cultures (Padilla-Walker, Carlo, & Nielson, 2015; Xu, Farver, & Zhang, 2009). The questionnaire measures the following three parenting styles on a 5-point Likert scale ranging from 1 (never) to 5 (always): authoritative (based on 15 items capturing emotional support and rule setting/reasoning, e.g., responsive to child’s feelings and need; emphasizes the reasons for rules), authoritarian (based on 12 items capturing low support, high coercion and hostility, e.g., scolds and criticizes to make child improve; grabs child when being disobedient), and permissive (based on 5 items reflecting an indulgent caring attitude, low in rule setting, e.g., finds it difficult to discipline child; spoils child). The PSDQ is characterized by good psychometric properties (Olivari, Tagliabue, & Confalonieri, 2013), predicts behavioural problems and social competences in children (Kim et al., 2012; Rinaldi & Howe, 2012), and has been found to interact with children’s physiological reactivity in predicting children’s behavioral and socio-emotional outcomes (Miller et al., 2013). Internal consistency (Cronbach’s alpha) based on data from 414 mothers who completed the questionnaire in the SBTS when children were aged three was .82 for authoritative .74 for authoritarian, and .74 for permissive parenting.

Child Outcomes

The Child Behavior Checklist (CBCL) 1½−5 (Achenbach, 2009) was used to assess externalizing and internalizing behavior problems at the age of 3 years, and the CBCL 4–18 at age six. Reliability for internalizing and externalizing behaviour problems were .84 (36 items) and .91 (24 items) at age three, and .86 (32 items) and .88 (35 items) at age six, respectively. In order to assess social competence, mothers were interviewed when children were aged three with semi-structured questions from the Vineland Adaptive Behavior Screener socialization subscale (Sparrow, Balla, Cicchetti, Harrison, & Doll, 1984), a interview conducted by a trained research assistant and focused on evaluating the level of children’s social skills required for everyday life (e.g., interpersonal interactions and sensitivity, manners, responsibility). Internal consistency for the 15-item scale was .51. When children were aged six, mothers reported on children’s social competence using a 7-item questionnaire developed by Eisenberg and associates (1993) which includes items assessing social relationship with peers and the general level of functioning and social competences in every-day life. Reliability was satisfactory with .78.

Data Analysis

First, we investigated the psychometric properties of the HSC-RS in the initial exploratory sample of 161 children with a preliminary Exploratory Factor Analysis. Then, we tested the factor structure with a Confirmatory Factor Analysis in the total sample of 292 children. In order to evaluate goodness of fit of the solution that emerged from the exploratory phase, we used the Tucker Lewis index (TLI), the comparative fit index (CFI), the root mean square error of approximation (RMSEA) and the standardized root mean square residuals (SRMR). CFI and TLI values of > .95 and > .97, respectively, are considered as acceptable and good fit (Schermelleh-Engel, Moosbrugger, & Müller, 2003). For RMSEA, values lower than .05 are considered a good fit and values ranging from .05 and .08 an adequate fit. For SRMR, values less than .08 are considered to reflect good fit (Schermelleh-Engel et al., 2003).

Second, to explore the association between the HSC-RS factor score and temperament traits, in the total sample of 292 children, we computed bivariate correlations between our rating system and the five temperament factors, and explored the percentage of variance of the HSC-RS factor score explained by temperament applying a multiple regression model including the five factors of temperament (Dyson et al., 2012) as predictors of the HSC-RS. Third, we explored the interaction between HSC-RS and parenting in the prediction of behavior problems and social competence at ages three and six testing and comparing a main effects model (with parenting variables and gender as predictors) with a series of interaction models that included HSC-RS in interaction with the different parenting dimension. In order to test whether inclusion of the interaction term significantly improved model fit, we used the Akaike weights (w_i) criterion, which provides a measure of the strength of fit for each model, and represents the probability that the same model would be predicted in new data (McElreath, 2016; Vandekerckhove, Matzke, & Wagenmakers, 2015; Wagenmakers & Farrell, 2004). Akaike weights range from 0 to 1, and the higher the value, the better the model is at describing data accurately. We then explored parameters and associated p-values for the model that fit data best and compared children low and high in sensitivity (i.e., top and bottom 25%) for models yielding significant interactions. More specifically, we calculated the cross-over interaction point and associated confidence intervals and contrasted, for each outcome, strong and weak patterns of Differential Susceptibility, Diathesis-Stress, and Vantage Sensitivity, applying a comparative modelling approach based on a re-parametrized regression equation model to contrast low and high sensitive children (Belsky, Pluess, & Widaman, 2013; Widaman et al., 2012). The “strong” version of each interaction model assumes that low-sensitive children are not affected at all by the quality of the environment (i.e., slope = 0), whereas the “weak” version assumes that low-sensitive children are affected by the environment, though to a lesser extent than high-sensitive children.

Finally, because Negative Emotionality and Difficult Temperament emerged consistently as markers of infants’ susceptibility to parenting in previous studies (Slagt, Dubas, et al., 2016) and because children in the SBTS that are high in negative affect have been shown to be more vulnerable to the negative effects of disaster stress (Kopala-Sibley et al., 2016), we also tested whether detected interaction effects with HSC-RS would remain stable after inclusion of the interaction term between Negative Affect and parenting, to exclude the possibility that HSC-RS interaction effects simply reflect the moderating properties of Negative Affect.

Bivariate associations among all variables are provided in the supplementary material document. All analyses were performed using the statistical software R and, specifically, lavaan package for fitting linear models (Rosseel, 2012), and ggplot2 for graphical representations (Wickham, 2009). The syntax in R for fitting and comparing models of Environmental Sensitivity is reported in the supplementary materials section.

Results HSC-RS Factorial structure

Initial Exploratory Factor Analysis (EFA) on the subsample of 161 cases suggested that a one latent factor solution fit the data best. Confirmatory Factor Analysis (CFA) in the total sample further supported a one-factor solution to fit data reasonably well according to all indices except for RMSEA, which fell slightly below the desired value (N = 288, CFI = .934, TLI = .907, RMSEA = .09, SRMR = .047). The EFA plot with eigenvalues, and CFA item loadings, are provided in the supplementary information document. The HSC factor score correlated at .98 with the HSC items mean.

Bivariate Associations with Age, Gender and Temperament

Bivariate correlations between the HSC-RS factor score, age, gender, and the five Lab-TAB temperament factors (data for all measures were available for N = 282) are reported in Table 2. The HSC-RS factor score correlated positively and significantly with gender (r = .24; p < .001, 0 = male, 1 = female) but not with age (r = .03, p = .52). Pertaining to temperament, HSC-RS correlated significantly and negatively with Dysphoria (r = −.25, p <.001, Positive Affect/Interest (r = −.21, p < .001) and Assertiveness/Sociability (r = −.48, p < .001), and positively and significantly with Constraint (r = .52, p < .001). Results from the multiple regression analysis showed that Assertiveness/Sociability, Dysphoria, Fear/Inhibition, Positive Affect/Interest and Constraint together explained 50% of the variance of the HSC factor score.

Table 2.

Bivariate associations between the HSC-RS factor score, sex, age, and the five temperament traits

	HSC-RS	Sex	Age	Assertiveness	Dysphoria	Fear/inhibition	Positive affect/Interest
HSC-RS	--
Sex	.24**	--
Age	.03	.01	--
Assertiveness/Sociability	−.48**	.09	.04	--
Dysphoria	−.25**	−.13*	<.01	.05	--
Fear/inhibition	.27**	.14	−.11	−.21	.24**	--
Positive affect/Interest	−.21**	.12	.20**	.47**	−.08	−.15*	--
Constraint	.52**	.30**	.15	−.12	−.34**	<.01	.02

^**p ≤ .001

^*p ≤.05

Interaction between HSC-RS and Parenting

Externalizing and Internalizing Behavior Problems

The model including the interaction between permissive parenting and HSC-RS received the strongest support (see Table 3) in predicting externalizing and internalizing behavior problems at age three (Akaike weights are reported in Table 3; standardized parameters and associated p values of interaction terms were, respectively, β = .42, p = .05 for the externalizing domain and β = .45, p =.03 for the internalizing domain (data for all measures were available for N = 252). Similarly, the interaction of permissive parenting and HSC-RS predicted internalizing behavior problems at age six (Akaike weights are reported in Table 3; β = .48, p = .05, N = 208). However, for externalizing problems at age six, it was the main effects model that fit the data best.

Table 3.

HSC-RS factor score X parenting in predicting behaviour problems and social competence. Identification of interaction patterns based on Akaike weights

	Externalizing Problems	Internalizing Problems	Social Competence
Age three
Main effect model	.22	.19	.10
Authoritative parenting X HSC-RS	.09	.08	.81
Authoritarian parenting X HSC-RS	.09	.07	.05
Permissive parenting X HSC-RS	.59	.66	.04
Age six
Main effect model	.44	.23	.23
Authoritative parenting X HSC-RS	.18	.10	.49
Authoritarian parenting X HSC-RS	.16	.11	.08
Permissive parenting X HSC-RS	.22	.55	.20

Note: In bold are models that received the best support for each outcome at age three and six

Social Competence

In contrast to behavior problems, the model including the interaction between authoritative parenting and HSC-RS consistently outperformed the main effects model in the prediction of social competence, based on Akaike weights comparison (Table 3) both at age three (β = 1.24, p = .01, N = 252) and six (β = 1.07, p = .06, N = 208).

Follow-up Analysis: Identification of Environmental Sensitivity Patterns

In order to follow-up significant interaction effects, we contrasted different patterns of Environmental Sensitivity based on a re-parametrized equation approach by comparing extreme groups: highly sensitive children (scoring above the 75^th percentile of the HSC-RS factor score, n = 73), and low sensitive children (scoring below the 25^th percentile, n =72). Akaike weights resulting from the model comparison are reported in Table 4, and a graphical representation of interaction effects (i.e., simple slopes) for high and low sensitive children is provided in Figures 1 to 5. Equation parameters for the reparametrized equation are reported in the supplementary information document.

Table 4.

Follow-up analyses: Identification of Environmental Sensitivity Patterns based on Akaike weights

	HSC-RS-X-Permissive on externalizing, age three	HSC-RS-X-Permissive on internalizing, age three	HSC-RS-X- Permissive on internalizing, age six	HSC-RS-X-Authoritative on social competence, age three	HSC-RS-X - Authoritative on social competence, age six
Differential Susceptibility
Strong	.62	.20	.20	.19	.31
Weak	.28	.07	.12	.08	.13
Diathesis Stress
Strong	.00	.52	.48	.00	.01
Weak	.00	.20	.20	.00	.07
Vantage Sensitivity
Strong	.05	.00	.00	.52	.08
Weak	.04	.00	.00	.19	.30

Note: the solutions with the best fit are marked in bold

Figure 1.

Simple slopes between permissive parenting and externalizing behavior problems at age three for children scoring low (β = −.16, p = .54) and high (β = .93, p < .001) on the Highly Sensitive Child-Rating System (HSC-RS).

Figure 5.

Simple slopes between authoritative parenting and social competence at age six for children scoring low (β = −.05, p = .44) and high (β = .18, p < .001) on the Highly Sensitive Child-Rating System (HSC-RS).

For externalizing behavior problems at age three, the pattern receiving strongest support was strong Differential Susceptibility according to Akaike weights (.62, see Table 4). More specifically, the estimated cross-over point (C) for HSC-RS group-X-permissive parenting (C =13.08, SE = 1.47) fell close to the mean of permissive parenting (M = 11.13, SD = 3.34). Furthermore, the 95% confidence interval for the estimated C (95% CI ([10.20 – 15.96]) was within the range of observed permissive parenting (range = 6 – 21), suggesting a for-better-and-for- worse interaction pattern consistent with Differential Susceptibility (see Figure 1).

For internalizing behavior problems at age three, the model reflecting a strong Diathesis-Stress pattern received the strongest support (.43, Table 4). The estimated cross-over point of the interaction between HSC-RS group and permissive parenting (C =6.95, SE = 2.15) fell far from the mean of permissive parenting (M = 11.13, SD = 3.34) and the lower end of the 95% CI (95% CI [2.73 – 11.15]) was outside of the observed range (range = 6 to 21), in line with a vulnerability pattern (see Figure 2). Consistent with results that emerged at age three, the strong Diathesis-Stress pattern also received the best support at age six (Table 4). The lower 95% CI of the estimated cross-over point for HSC-RS group-X-permissive parenting (C = 7.59, SE = 1.86, CI [3.95 – 11.23]) at age six fell also outside of the lower range of the permissive parenting, suggesting again a vulnerability pattern (see Figure 3).

Figure 2.

Simple slopes between permissive parenting and internalizing behavior problems at age three for children scoring low (β = .02, p = .90) and high (β = .65, p = .01) on the Highly Sensitive Child-Rating System (HSC-RS).

Figure 3.

Simple slopes between permissive parenting and internalizing behavior problems at age six for children scoring low (β = −.14, p = .16) and high (β = .57, p = .03) on the Highly Sensitive Child-Rating System (HSC-RS).

For social competence at age three, the strong Vantage Sensitivity pattern received the strongest support. The estimated cross-over point of the interaction between HSC-RS groups and authoritative parenting (C = 45.69, SE = 9.19, 95%) fell far from the mean of authoritative parenting (M = 61.73, SD =6.70), and the lower end of the 95% CI (CI = [27.68 – 63.70]) was outside of the observed range (range = 43 – 75), in line with a Vantage Sensitivity pattern (Figure 4). At age six, a weak Vantage Sensitivity pattern and a strong Differential Susceptibility pattern had comparable fit (Table 4). However, the lower CI of the estimated cross-over point (C =54.58, SE =6.48, 95% CI = [41.88 –67.28]) fell very close to the lower range of authoritative parenting, suggesting that a Vantage Sensitivity effect is more plausible (see Figure 5).

Figure 4.

Simple slopes between authoritative parenting and social competence at age three for children scoring low (β =−.03, p = .69) and high (β = .14, p = .05) on the Highly Sensitive Child-Rating System (HSC-RS).

Sensitivity Analysis

When Negative Affect and its interaction term were included in the interaction model with permissive parenting in the prediction of children’s externalizing behaviour problems at age three, of children’s internalizing behavior problems at ages three and six, and in the interaction model with authoritative parenting in the prediction of social competence at ages three and six, results for the detected interactions between HSC-RS and parenting remained overall stable and Negative Affect did not interact with parenting in the prediction of the outcome variables (parameters and associated p values are provided in the supplementary materials document). Hence, HSC-RS moderated the effects of permissive parenting on negative outcomes (i.e., externalizing behavior problems at age three, and internalizing) at age three and six), and of authoritative parenting on positive outcomes (i.e., social competence at age three and six), independent of the influence of Negative Affect.

Discussion

In the current paper we aimed to investigate the psychometric properties and moderating effects of a newly developed observational measure of children’s sensitivity to environmental factors. This new measure, consisting of a set of scales applied to existing episodes of the Lab-TAB, was found to capture individual differences in children’s Environmental Sensitivity, with children scoring high showing higher sensitivity to both low and high parenting quality regarding the development of both negative and positive outcomes across ages three and six. This result is in line with theories of Environmental Sensitivity, according to which some children show an enhanced sensitivity to the both low and high quality of their rearing environment. Traditionally, Difficult Temperament and Negative Affect have been identified as phenotypic markers of such sensitivity in infancy and toddlerhood (Slagt, Dubas, et al., 2016). Recently, a self-report measure aimed at capturing Environmental Sensitivity more specifically has been developed for children aged 8 years and older (Pluess et al., 2018), with empirical evidence confirming its ability to capture individual differences in sensitivity to both negative and positive parenting practices (Slagt et al., 2018) as well as to intervention programs (Nocentini et al., 2018; Pluess & Boniwell, 2015). However, existing measures of Environmental Sensitivity are mostly based on self- or parent-report. This study represents the first attempt at investigating sensitivity at an observational level with a measure specifically developed to capture sensitivity in 3–5 year-old children. Using data from the Stony Brook Temperament Study (Klein & Finsaas, 2017), we recoded a series of video-taped laboratory episodes, previously used for the assessment of child temperament, by applying a newly developed rating-system designed to capture sensitivity at an observational level, the Highly Sensitive Child – Rating System (HSC-RS). After investigating the newly developed measure’s psychometric proprieties and its association with other temperament traits, we tested its validity by exploring the interaction with parenting followed by a model comparison approach (Belsky et al., 2013), in which we explored different patterns of Environmental Sensitivity (Pluess, 2015).

Analyses regarding the psychometric properties of the HSC-RS showed that the scale is robust and reflects a one-factor solution. The detected single sensitivity factor correlated significantly with other temperament traits, but was not fully captured by any of those nor by their combined effect, suggesting that the observed sensitivity dimension represents a unique and relatively distinct trait. More specifically, as expected, the HSC-RS correlated negatively with Positive Affect/Interest and Assertiveness/Sociability Lab-TAB factors, both capturing aspects of Extraversion, and positively with Constraint and Fear/Inhibition. Contrary to our expectations, no positive association was found with Dysphoria, though sensitivity has been frequently reported to correlate with Neuroticism in middle childhood/adolescence and in adult samples (Lionetti et al., 2019). One explanation could be that the Lab-TAB Dysphoria factor includes sadness as well as anger. While it is reasonable to expect an association between sadness and sensitivity as assessed with the HSC-RS, there is no theoretical or empirical reason to expect that sensitive children would display more anger than less sensitive children.

Regression models with HSC-RS as continuous score were run to test whether the HSC-RS does indeed measure differences in Environmental Sensitivity. According to results, the HSC-RS moderated the impact of several parenting dimensions on both negative and positive child outcomes. According to follow-up analyses with simple slopes, children scoring high (i.e., above the 75% percentile of the HSC-RS) were more affected by the influence of parenting quality compared to children scoring low (below the 25% percentile). More specifically, children in the high sensitive group were more sensitive to the negative influence of high levels of permissive parenting in relation to externalizing behavior problems at age three (in line with a Differential Susceptibility pattern of findings) and internalizing behavior problems at age three and six (reflecting a Diathesis-Stress pattern), as well as to positive effects of high levels of authoritative parenting in the prediction of social competence at ages three and six (consistent with Vantage Sensitivity). On the other hand, children scoring low in sensitivity were generally less sensitive to the influence of parenting, appearing resilient when faced with high levels of permissive parenting but at the same time not benefiting from high levels of authoritative parenting. Importantly, when we controlled for the moderating role of Negative Affect, an established marker of sensitivity reported in a large number of studies (Belsky & Pluess, 2009; Slagt, Dubas, et al., 2016), findings remained stable, suggesting that our observational sensitivity measure captured more than the susceptibility commonly reflected in Negative Emotionality. The absence of any significant interaction with the authoritarian parenting dimension may suggest that high levels of this parenting practice may impact on all children’s adjustment equally, irrespective of children’s differences in sensitivity. Alternatively, the sample was low-risk and characterized by low levels of authoritarian parenting scale which makes the detection of interaction effects in relation to this parenting style rather difficult.

In sum, we found empirical support that the newly developed observer-rated sensitivity measure moderated the effects of parenting quality on children’s outcomes, and that some children were indeed more sensitive to the effects of high levels of both low and high quality parenting on both negative and positive developmental outcomes at ages three and six. Furthermore, the new measure captured a unique trait of Environmental Sensitivity which did not completely overlap with other more established temperament traits. These findings not only validate the newly developed measure of Environmental Sensitivity, but also confirm that individual differences in sensitivity can be assessed observationally in young children. Additionally, the finding that children scoring high on this scale are more negatively affected by high levels of negative parenting as well as more positively influenced by high levels of positive parenting adds further support for the existence of individual differences in Environmental Sensitivity (Belsky & Pluess, 2009, 2013; Pluess & Boniwell, 2015; Slagt et al., 2018)

Strengths and Limitations

The findings reported in this paper are characterized by several methodological strengths, including the use of an observer-rated measure of temperament, independent coding of temperament and sensitivity (i.e., different raters for different outcomes), a longitudinal design, and application of recently developed statistical procedures to competitively compare various models of Environmental Sensitivity (including, for the first time, Vantage Sensitivity). Finally, given that our observational rating-system for sensitivity is based on the widely known and used Lab-TAB procedure (Goldsmith et al., 1999), which has been applied in many other studies, researchers in possession of Lab-TAB data will be able to recode their videos, after being trained on the HSC-RS, in order to obtain a measure of Environmental Sensitivity on the behavioral level without having to collect new data.

However, results should also be considered in light of several methodological limitations. Most importantly, parenting quality and child outcomes were based exclusively on parent report. Further research should consider the inclusion of observational measures for both the quality of the environment (i.e., parenting) and children’s developmental outcomes (Slagt, Dubas, et al., 2016). Furthermore, the sample included children from US-based families with a predominately middle-class background. Future studies should test whether the same set of scales is applicable to other more disadvantaged populations from different cultural backgrounds. Lastly, our observational measure of sensitivity has been developed based on 3-year-old children and we do not yet know whether it could be applied equally well across a wider age range.

Future Directions

Future research should continue to investigate whether observer-rated levels of Environmental Sensitivity moderate the impact of other environmental influences, such as education or peer influence. Another future direction pertains to the longitudinal evaluation of the stability and change in individual sensitivity levels. Given that our observer-rated measure of Environmental Sensitivity shares a common theoretical background with existing sensitivity measures for school-children, adolescents (Pluess et al., 2018), and adults (Aron & Aron, 1997), it is now possible to investigate sensitivity longitudinally, something which has not been done yet. Finally, it is important to expand the study of individual differences in Environmental Sensitivity across different cultures.

Conclusion

Developmental theories and empirical evidence (Belsky & Pluess, 2009; Boyce & Ellis, 2005; Del Giudice, Hinnant, Ellis, & El-Sheikh, 2012) suggest that children are not equally susceptible to the influence of environmental quality, with some more vulnerable in the face of adversity but also more likely to flourish when exposed to positive environmental conditions, as a function of individual differences in sensitivity. Such Environmental Sensitivity is an important individual characteristic that is different from other common temperament and personality traits, as reported in children and adults (Lionetti et al., 2018; Pluess et al., 2018). The current study suggests that it is possible to measure Environmental Sensitivity at the behavioral level in young children using the Highly Sensitive Child – Rating System (HSC-RS), which is psychometrically robust, does not completely overlap with other sensitivity-related temperament traits, and significantly moderates both negative and positive aspects of parenting in the prediction of developmental outcomes.