Environmental Sensitivity in Children: Development of the Highly Sensitive Child Scale and Identification of Sensitivity Groups Mic

Environmental Sensitivity in Children: Development of the Highly Sensitive Child Scale and Identification of Sensitivity Groups

Michael Pluess, Elham Assary, and Francesca Lionetti

Queen Mary University of London

Kathryn J. Lester University of Sussex

Eva Krapohl King’s College London

Elaine N. Aron and Arthur Aron Stony Brook University

A large number of studies document that children differ in the degree they are shaped by their developmental context with some being more sensitive to environmental influences than others. Multiple theories suggest that Environmental Sensitivity is a common trait predicting the response to negative as well as positive exposures. However, most research to date has relied on more or less proximal markers of Environmental Sensitivity. In this paper we introduce a new questionnaire—the Highly Sensitive Child (HSC) scale—as a promising self-report measure of Environmental Sensitivity. After describing the development of the short 12-item HSC scale for children and adolescents, we report on the psychometric properties of the scale, including confirmatory factor analysis and test–retest reliability. After considering bivariate and multivariate associations with well-established temperament and personality traits, we apply Latent Class Analysis to test for the existence of hypothesized sensitivity groups. Analyses are conducted across 5 studies featuring 4 different U.K.-based samples ranging in age from 8–19 years and with a total sample size of N � 3,581. Results suggest the 12-item HSC scale is a psychometrically robust measure that performs well in both children and adolescents. Besides being relatively independent from other common traits, the Latent Class Analysis suggests that there are 3 distinct groups with different levels of Environmental Sensitivity—low (approx. 25–35%), medium (approx. 41–47%), and high (20–35%). Finally, we provide exploratory cut-off scores for the categorization of children into these different groups which may be useful for both researchers and practitioners.

Keywords: differential susceptibility, Environmental Sensitivity, personality, temperament

Supplemental materials: http://dx.doi.org/10.1037/dev0000406.supp

Children’s development is shaped by many factors, including various aspects of the environment in which they grow up (e.g., child care, see Belsky, Vandell, et al., 2007; socioeconomic status and parenting, see Bornstein & Bradley, 2014). One of the reasons for the often significant impact environmental fac- tors have on developmental outcomes is children’s ability to register and process specific characteristics of their develop- mental context (Pluess, 2015). This capacity for Environmental

Sensitivity enables them to respond and adapt to the challenges and opportunities associated with particular environmental con- ditions. Although, at first glance, one may expect that all children should have a similar ability to adapt to the develop- mental context, given the fundamental importance of adaptation for successful development, a large number of empirical studies suggest that children differ substantially in Environmental Sen- sitivity, with some being more and some less affected by

This article was published Online First September 21, 2017. Michael Pluess, Elham Assary, and Francesca Lionetti, Department of

Biological and Experimental Psychology, School of Biological and Chem- ical Sciences, Queen Mary University of London; Kathryn J. Lester, School of Psychology, University of Sussex; Eva Krapohl, Social, Genetic, & Developmental Psychiatry, King’s College London; Elaine N. Aron and Arthur Aron, Department of Psychology, Stony Brook University.

We express our gratitude to Anna M.T. Bosman, Joep Bakker, and Sietske Walda, for the initial development of the 38 child sensitivity items as well as to Thalia Eley and Hannah Brown for their role in collecting data for the test–retest reliability study. We also thank the TEDS team for inclusion of the HSC scale in their data collection and gratefully acknowl-

edge the ongoing contribution of the participants in TEDS and their families. TEDS is supported by a program grant to Robert Plomin from the U.K. Medical Research Council [MR/M021475/1; and previously G0901245], with additional support from the US National Institutes of Health [HD044454; HD059215; AG046903] and the European Commis- sion [602768]. Francesca Lionetti is supported with a grant of the European Commission [H2020-MSCA-IF-2015-704283].

Correspondence concerning this article should be addressed to Michael Pluess, Department of Biological and Experimental Psychology, School of Biological and Chemical Sciences, Queen Mary University of London, Mile End Road, London E1 4NS, United Kingdom. E-mail: m.pluess@ qmul.ac.uk

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Developmental Psychology © 2017 American Psychological Association 2018, Vol. 54, No. 1, 51–70 0012-1649/18/$12.00 http://dx.doi.org/10.1037/dev0000406

51

contextual factors (Belsky & Pluess, 2009, 2013; Ellis & Boyce, 2011; Obradovic & Boyce, 2009).

In this paper we address three empirical objectives related to the measurement of Environmental Sensitivity in children and adoles- cents. First, we provide extensive information on the development and psychometric properties of an Environmental Sensitivity self-report measure, the child version of the Highly Sensitive Person scale (HSP scale; Aron & Aron, 1997). Second, we examine bivariate and mul- tivariate associations between this new measure of child Environmen- tal Sensitivity and established measures of temperament and person- ality. Third, we investigate the distribution of Environmental Sensitivity in the sample population to test for the existence of groups with different degrees of sensitivity as proposed by several theories and suggest exploratory cut-off scores for these different groups. These objectives are addressed across five studies featuring four different U.K.-based samples with children ranging in age from 8–19 years and a total sample size of N � 3,581.

Individual Differences in Environmental Sensitivity

Environmental Sensitivity, defined as the ability to register and process external stimuli (Pluess, 2015), is one of the most basic individual characteristics and observable across most species. Without this ability, an organism would not be able to perceive, evaluate, and respond to various environmental conditions, whether these are of physical or psychosocial nature, and whether they are negative or positive (i.e., whether they threaten or promote the development, survival, and reproductive success of the individual). Although adap- tation is relevant for all people, empirical studies suggest that indi- viduals differ substantially in their degree of Environmental Sensitiv- ity (for review, see Belsky & Pluess, 2009, 2013) with more and less sensitive types coexisting in the same population (Wolf, van Doorn, & Weissing, 2008). Differences in how people approach, respond and interact with their immediate environment are also reflected in con- cepts of temperament and personality. Although the various temper- ament theories differ significantly from each other, one thing they seem to have in common is that they all suggest that some individuals appear more reactive to contextual factors than others, with more environmentally sensitive individuals described as, for example, in- hibited/reactive (Kagan, 1989). A growing number of empirical stud- ies provide evidence that temperament traits do indeed predict differ- ences in Environmental Sensitivity (for a meta-analysis, see Slagt, Dubas, Dekovic, & van Aken, 2016). For example, Pluess and Belsky (2010) found that infant temperament rated by mothers when children were 6 months old predicted children’s sensitivity to the parenting quality they experienced during the first 4.5 years of life. Children with a more difficult temperament were both more negatively affected by low parenting quality and more positively by high parenting quality vis-à-vis teacher-rated social skills at age 11 years compared with children with a less difficult temperament (for a reanalysis applying more stringent methodology, see Roisman et al., 2012). Similarly, Kim and Kochanska (2012) reported that negative emo- tionality assessed at 7 months was associated with increased sensitiv- ity to both low and high mother-child mutuality at 15 months regard- ing the development of self-regulation at 25 months. More negatively emotional infants had the lowest self-regulation when mother–child mutuality was low and the highest self-regulation when mutuality was high whereas low negatively emotional children were generally less affected by differences in mother-child mutuality. Although a large

number of studies suggest that difficult temperament is associated with heightened sensitivity to the environment, it is important to acknowledge that it remains to be determined which component of the typically multidimensional concept of “Difficult Temperament” re- flects such sensitivity. Furthermore, “Difficult Temperament” is often assessed with different measures, which makes comparison between studies challenging. However, according to a recent meta-analysis of temperament-parenting interactions, it may be negative emotionality, rather than surgency or effortful control, that predicts sensitivity to parenting (Slagt et al., 2016).

Several of the Big Five personality traits have also been shown to reflect individual differences in Environmental Sensitivity. For example, low extraversion—or introversion—has been associated with higher sensitivity to both high and low parental overreactivity in the prediction of aggression in adolescence (De Haan, Prinzie, & Deković, 2010). Not surprisingly, childhood neuroticism—or irritability/negative emotionality—has repeatedly been shown to increase the response to environmental influences, albeit mostly negative ones, including exposure to violence in adolescence (Ho et al., 2013) and stressful life events in adulthood (van Os & Jones, 1999). Finally, openness to experiences has recently been associ- ated with increased parental Environmental Sensitivity to both low and high perceived social support (Slagt, Dubas, Denissen, Dek- ović, & van Aken, 2015).

Gray’s (1981, 1982) personality theory, which originally pro- posed that individual differences in response to reward and pun- ishment are driven by two distinct biological systems, can also be considered from a perspective of Environmental Sensitivity: Whereas the Behavioral Inhibition System (BIS) captures sensitiv- ity to threatening stimuli, the Behavioral Activation System (BAS) describes sensitivity to rewarding (i.e., positive) experiences.1

Several experimental studies provide evidence that BIS and BAS do indeed predict specific sensitivity to either negative or positive environmental influences. For example, BIS has been found to predict the negative emotional response to unpleasantly loud noises (Heponiemi, Keltikangas-Jarvinen, Puttonen, & Ravaja, 2003) and higher negative reactivity to negative life events (Gable, Reis, & Elliot, 2000). BAS, on the other hand, has been associated with positive emotional responsivity to anticipated monetary re- ward (Carver & White, 1994) as well as stronger brain activation in response to appetitive food pictures (Beaver et al., 2006).

Concepts for Individual Differences in Environmental Sensitivity

There are several theoretical frameworks for variability in En- vironmental Sensitivity that have emerged since the mid to late 1990s with the three most prominent being Sensory-Processing Sensitivity (Aron & Aron, 1997; Aron, Aron, & Jagiellowicz, 2012), Differential Susceptibility Theory (Belsky, 1997b, 2005;

1 It is important to acknowledge that Gray revised his original theory (see McNaughton & Gray, 2000). In brief, BIS is now thought to produce alert interest and a pause in activity that allows for the processing of conflicting information, a balancing of or negotiation between the urge to approach and satisfy needs (i.e., BAS), and the urge to stop and consider risks, costs, or how best to make use of an opportunity. In the case of threat, a third strategy of fight, flight, or freeze is suggested. However, popular measures of BIS-BAS (i.e., Carver & White, 1994) have been developed earlier and do not reflect that conceptual change.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

52 PLUESS ET AL.

Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007; Belsky & Pluess, 2009, 2013), and Biological Sensitivity to Context (Boyce & Ellis, 2005; Ellis & Boyce, 2008). Each of the three concepts provides unique and important theoretical insights re- garding individual differences in general Environmental Sensitiv- ity—discussed in more detail elsewhere (Ellis, Boyce, Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2011; Pluess, 2015). However, the significant and distinctive contribution shared across all three frameworks is the notion that sensitive individuals differ not only in their response to environmental adversity (e.g., child maltreatment, stressful life events, poverty etc.)—as the traditional Diathesis-Stress model would imply—but also in response to positive supportive aspects of the environment (e.g., sensitive parenting, social support etc.). This new aspect of variability in sensitivity to positive experiences has recently been extracted from the more general models of Environmental Sensitivity and further developed into the framework of Vantage Sensitivity (Pluess, 2017; Pluess & Belsky, 2013). According to the concept of Van- tage Sensitivity, people differ fundamentally in their response to positive environmental influences and exposures as a function of

inherent characteristics with some being more sensitive and some being more resistant to the beneficial effects of positive experi- ences, including psychological interventions (e.g., Albert et al., 2015).

Recently, these different concepts have been integrated into an overarching metaframework of Environmental Sensitivity (see Fig- ure 1 for an illustration) according to which people differ in their sensitivity to environmental influences with some being more and some less affected by negative and/or positive exposures (Pluess, 2015).

Measuring Environmental Sensitivity

Most evidence for individual differences in Environmental Sensitivity is based on research reporting cross-over interactions between some contextual measure (e.g., parenting quality) and a wide range of individual traits that can be categorized into gene- tic (e.g., 5-HTTLPR; van IJzendoorn, Belsky, & Bakermans- Kranenburg, 2012), physiological (e.g., cortisol reactivity; Obra- dovic, Bush, Stamperdahl, Adler, & Boyce, 2010) and behavioral/

Figure 1. Illustration of the different models describing individual differences in Environmental Sensitivity: Diathesis-Stress (A) describes variability in response to adverse exposures, and Vantage Sensitivity (B) variability in response to supportive exposures, whereas the remaining three models Sensory Processing Sensitivity (C), Differential Susceptibility (D), and Biological Sensitivity to Context (E) describe individual differences in response to both negative and positive experiences. Consequently, Models C, D, and E reflect the combination of Models A and B. Adapted from Figure 1 in “Individual Differences in Environmental Sensi- tivity,” by M. Pluess, 2015, Child Development Perspectives, 9, pp. 138–143. Copyright 2015 by Wiley. Adapted with permission.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

53ENVIRONMENTAL SENSITIVITY

psychological sensitivity factors (e.g., negative emotionality; Kim & Kochanska, 2012) in the prediction of some behavioral outcome measure (e.g., social skills). Although these sensitivity factors may represent important markers of Environmental Sensitivity at dif- ferent levels of analysis—some more proximal than others—none of them describe and capture the hypothesized phenotypic trait of Environmental Sensitivity directly. In fact, to our knowledge ex- plicit phenotypic measures of Environmental Sensitivity are cur- rently not available with the exception of the Highly Sensitive Person scale (Aron & Aron, 1997), a 27-item questionnaire de- signed to measure Sensory Processing Sensitivity in adults (but some measures may capture important aspects of Environmental Sensitivity, e.g., Orienting Sensitivity measured with the Adult Temperament Questionnaire, see Evans & Rothbart, 2007, 2008). According to Aron (1996; Aron & Aron, 1997) Sensory Process- ing Sensitivity (SPS) is a relatively stable personality trait that reflects an individual’s sensitivity to environmental influences and manifests itself in (a) greater awareness of sensory stimulation, (b) behavioral inhibition as described by McNaughton and Gray (2000) rather than Carver and White (1994) or Gray’s earlier theory (1981, 1982), (c) deeper cognitive processing of environ- mental stimuli, and (d) higher emotional and physiological reac- tivity (for an extensive review, see Aron et al., 2012). The Highly Sensitive Person scale (HSP scale; Aron & Aron, 1997) aims at capturing these cognitive and behavioral components of sensitivity and appears to succeed at doing so, most notably in fMRI studies reporting deeper or more elaborate cognitive processing in indi- viduals with higher HSP scores (Acevedo et al., 2014; Jagiellowicz et al., 2011), as well as behavioral studies. For example, Aron, Aron, and Davies (2005) were able to demonstrate consistent associations between HSP scores and heightened sensitivity to contextual factors in a series of studies, including an experimental one in which undergraduates completed a cognitive task. Students were randomly assigned to a situation that either implied they were doing much better or much worse than the peers sitting around them. Participants with high scores on the HSP Scale reported more negative affect than others after the task if they were led to believe they had done worse than others, but the least negative affect in the condition where they were led to believe they had done better. Those scoring low, on the other hand, did not differ significantly in negative affect regardless of condition, suggesting they were generally less affected by the experimental manipula- tion. More recently, Booth, Standage, and Fox (2015) tested in a cross-sectional study whether SPS assessed with the HSP Scale in adulthood moderated the effects of retrospectively reported child- hood experiences on adult life satisfaction. A significant interac- tion emerged suggesting that those scoring high were more nega- tively affected by negative childhood experiences than those scoring low.

In contrast to other common personality traits, SPS has been suggested to follow a dichotomous rather than a normal distribu- tion with about 20% of the general population falling into a highly sensitive category and about 80% into a less sensitive category (Aron et al., 2012; for an unpublished taxometric analysis of the HSP scale, see Borries, 2012). Interestingly, the proposition that a minority of the population is more sensitive to environmental influences is consistent with empirical findings on the distribution of temperament traits found to reflect heightened Environmental Sensitivity to both negative and positive aspects of the early

environment. For example, a taxometric analysis of the distribu- tion of Infant Reactivity or Behavioral Inhibition (Kagan, Reznick, & Snidman, 1987) suggested that such reactivity is distributed categorically rather than continuously, with about 10% of children being characterized with especially high reactivity (Woodward, Lenzenweger, Kagan, Snidman, & Arcus, 2000). Intriguingly, several of the candidate gene variants that have been repeatedly associated with increased Environmental Sensitivity to negative as well as positive exposures (Belsky et al., 2009; Belsky & Pluess, 2009, 2013) have a comparable frequency. For example, 18.4% of a large Dutch sample were homozygous for the 5-HTTLPR short allele (Pluess et al., 2011) which has been associated with in- creased sensitivity to both negative and positive influences (van IJzendoorn et al., 2012). The proposition that there might be two distinctive sensitivity patterns has been described in the popular Orchid-Dandelion metaphor (Ellis & Boyce, 2011) according to which Orchids represent the minority in the population who are generally more sensitive (i.e., they do exceptionally well in ideal conditions and exceptionally badly in poor ones) and Dandelions the majority who are generally less sensitive to environmental quality (i.e., they are resilient and can grow anywhere). However, although widely observed individual differences in Environmental Sensitivity may reflect the existence of different sensitivity groups with high sensitivity characterizing a minority of the general population, this hypothesis has not been tested empirically in children yet.

Although Aron and Aron (1997) originally hypothesized that the 27 items of the HSP scale would reflect a single factor of Envi- ronmental Sensitivity, other studies have found that a three factor structure was a better fit for the data (Smolewska, McCabe, & Woody, 2006). The three factors that typically emerge are (1) Aesthetic Sensitivity (AES), capturing aesthetic awareness (e.g., being deeply moved by arts and music); (2) Low Sensory Thresh- old (LST), which reflects unpleasant sensory arousal to external stimuli (e.g., reaction to bright lights and loud noises); and (3) Ease of Excitation (EOE), which refers to being easily over- whelmed by external and internal demands (e.g., negative response to having a lot going on, to being hungry). Smolewska et al. (2006) investigated correlations between the HSP scale and personality measures in adults, including the Big Five personality traits and BIS/BAS scales by Carver and White (1994), and found that the HSP total score was significantly and positively correlated with neuroticism (r � .45) and openness (r � .19), as well as both BIS (r � .32) and BAS (r � .16 for the reward-responsiveness sub- scale). When investigating associations with the three HSP sub- scales, they found that while neuroticism and BIS were correlated with all three factors, openness had a significant association only with Aesthetic Sensitivity (r � .37), Low Sensory Threshold with lower extraversion (r � �.12), and Ease of Excitation and Aes- thetic Sensitivity with the BAS Reward-Responsiveness scale (r � .19 and r � .18, respectively; for similar findings, see Gerstenberg, 2012). At first sight this correlation pattern appears to suggest that Aesthetic Sensitivity may reflect Environmental Sensitivity to more positive experiences, whereas Ease of Excitation and Low Sensory Threshold reflect sensitivity to more negative experiences. Important to note is also that the three subscales tend to be correlated with each other, suggesting that there may exist a general sensitivity factor (Lionetti et al., 2017).

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

54 PLUESS ET AL.

In summary, the HSP scale represents a promising self-report measure of Environmental Sensitivity in adults. However, there are currently no self-report versions of the scale for use with children and adolescents (but for the first evaluation of a parent- rated child scale, see Boterberg & Warreyn, 2016). In this paper we address this gap, across five studies, presenting a new brief child version of the HSP scale and investigating associations with common personality and temperament variables as well as testing for the existence of different sensitivity groups. More specifically, in Study 1 we describe the creation of a 12-item child HSP scale drawing on a pool of 38 child sensitivity items in a sample of 12-year old children. In Study 2, we test the psychometric prop- erties of the new 12-item scale in an independent sample of 11-year olds. In Study 3 we report test–retest reliability of the 12-item scale in a different sample of 10-year old children. In Study 4 we apply the same scale to a large sample of adolescents at age 17 years. Finally, in Study 5, we report findings of latent class analyses across the different samples to test for the existence of hypothesized sensitivity groups in childhood and adolescence and provide exploratory cut-off scores that can be used to approx- imately categorize children and adolescents into the identified sensitivity groups.

Study 1

The main objective of Study 1 was to create a short and psychometrically robust Highly Sensitive Child (HSC) scale draw- ing on 38 existing sensitivity items for children, which have been adapted from the 27 items included in the adult HSP scale. Besides being brief and psychometrically sound, the self-report measure should be suitable for children and adolescents and reflect the same factor structure as the adult version. Once the HSC scale was created, it was then tested for its psychometric properties as well as for its associations with related constructs of behavioral inhibi- tion and activation, temperament, and affect.

Method

Participants. The sample included 334 children (251 girls and 83 boys) with a mean age of 12.06 years (range � 11–14 years;

SD � 0.67) recruited from two secondary schools in East London, United Kingdom (one of the school was a girls-only school which explains the higher proportion of girls in this particular sample). The sample was ethnically diverse with 55.4% of Asian, 15.9% of African/Caribbean, 8.1% of White/European, 2.1% of Middle Eastern, and 18.6% of mixed ethnicity.

Procedure and development of scale. Children were asked to complete all questionnaires on a computer at school during class. To create a short and psychometrically robust HSC scale that is comparable in content and structure to the adult scale, the factor structure of the adult scale was consulted. As reported by Smolewska et al. (2006) a three factor structure seemed to fit the data collected with the adult HSP scale best with 12 items loading on the factor “Ease of Excitation,” 7 items on “Aesthetic Sensi- tivity,” and 6 items on “Low Sensory Threshold” (two items did not load clearly on any of the three factors and were excluded). To create a HSC scale that is comparable to the HSP scale, we first selected among the remaining 25 HSP items from Smolewska et al.’s (2006) factor analysis those that (a) had a factor loading of �.5 and (b) could be easily adjusted for the use with children. Twelve items met these criteria. Then, we conducted a Principal Component Analysis (PCA), constrained to three components (given that the HSP scale reflects three factors) across a pool of 38 sensitivity items for children (HSC-38, provided in supplementary information) that have been developed based on the 27 HSP items for adults, to test whether the HSC items would reflect similar factor loadings as the corresponding adult HSP items as reported by Smolewska et al. (2006). The final 12-item HSC scale included 5 Ease of Excitation items, 4 Aesthetic Sensitivity items, and 3 Low Sensory Threshold items (see Table 1 for a list of the specific items).

Measures. Children completed 38 items from an unpublished sensitivity scale (HSC-38, see supplementary information), which has been developed initially to measure Sensory Processing Sen- sitivity in Dutch school-age children (Walda, 2007). The 38 items aim at capturing the same information as the adult HSP scale (Aron & Aron, 1997). Items such as “When someone is sad, that makes me feel sad too,” “I find it unpleasant to have a lot going on at once,” and “When I am hungry, I get in a bad mood” were rated

Table 1 HSC Rotated Component Matrix (Study 1)

Items

Factor

1 (EOE) 2 (AES) 3 (LST)

1. I find it unpleasant to have a lot going on at once .53 .07 .15 2. Some music can make me really happy .04 .79 �.02 3. I love nice tastes .18 .83 .00 4. Loud noises make me feel uncomfortable .35 .02 .67 5. I am annoyed when people try to get me to do too many things at once .71 .26 �.02 6. I notice it when small things have changed in my environment .29 .44 .03 7. I get nervous when I have to do a lot in little time .66 .26 .23 8. I love nice smells .13 .79 .24 9. I don’t like watching TV programs that have a lot of violence in them .05 .04 .66

10. I don’t like loud noises .10 .06 .86 11. I don’t like it when things change in my life .48 .22 .45 12. When someone observes me, I get nervous. This makes me perform worse

than normal .70 .00 .14

Note. EOE � Ease of Excitation; AES � Aesthetic Sensitivity; LST � Low Sensory Threshold. Bold values indicate the factor that the item loads strongest on.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

55ENVIRONMENTAL SENSITIVITY

by children on a scale from 1 � not at all to 7 � extremely, with higher scores indicating higher levels of sensitivity. The internal reliability of the 38 items was good with Cronbach’s alpha � .92.

Behavioral inhibition and activation was measured with the 24-item Behavioral Inhibition and Behavioral Activation Scales (BIS-BAS; Carver & White, 1994). The Behavioral Inhibition Scale (BIS) is based on 7 items (e.g., “Criticism or scolding hurts me quite a bit,” “I worry about making mistakes”) whereas the Behavioral Activation Scale (BAS) features three subscales (i.e., “Reward Responsiveness,” “Drive,” and “Fun Seeking”). For the current study, all 17 BAS items (e.g., “It would excite me to win a contest,” “I’m always willing to try something new if I think it will be fun”) were pooled into one scale. BIS-BAS items are rated on a Likert scale ranging from 1 � very false to 4 � very true. Higher scores indicate higher levels of behavioral inhibition (BIS) and activation (BAS). In the current sample internal consistencies of BIS and BAS were � � .80 and � � .91, respectively.

Temperament was measured with the 65-item Early Adolescent Temperament Questionnaire—Revised (EATQR; Capaldi & Roth- bart, 1992) which assesses 12 aspects of temperament (i.e., Acti- vation Control, Affiliation, Attention, Fear, Frustration, High- Intensity Pleasure, Inhibitory Control, Perceptual Sensitivity, Pleasure Sensitivity, Depressive Mood, Aggression, and Shyness). Items (e.g., “I feel shy about meeting new people,” “I feel pretty happy most of the day,” “When I am angry, I throw or break things”) are rated on a 5-point Likert scale, ranging from 1 � almost always untrue of you, to 5 � almost always true of you. For the current study, we combined these subscales—as recommended by others (Putnam, Ellis, & Rothbart, 2001; Snyder et al., 2015)— into three superordinate dimensions of temperament: (1) Effortful Control (EC; based on Attention, Activation Control, and Inhibi-

tory Control), (2) Negative Emotionality (NE; based on Fear, Frustration and Shyness), and (3) Positive Emotionality (PE; based on Surgency, Pleasure Sensitivity, Perceptual Sensitivity and Af- filiation). Higher scores on each subscale indicate higher levels on that temperament dimension. Internal consistencies of the scales were acceptable with � � .86 for EC, � � .69 for NE, and � � .84 for PE.

Positive and negative affect were measured with the child ver- sion of the Positive and Negative Affect Scales (PANAS; Laurent et al., 1999). The Positive Affect (PA) scale includes 12 items (e.g., “Interested,” “Excited”) and the Negative Affect (NA) scale 15 items (e.g., “Upset,” “Guilty”). All items are rated on Likert scale, ranging from 1 � not at all to 5 � almost every day. Higher scores indicate higher state levels of positive or negative affect. The internal consistency of the PANAS was good with � � .92 for PA and � � .93 for NA.

Data analysis. To create the HSC scale, we conducted Prin- ciple Component Analyses (PCA) on the 38 sensitivity items (applying Varimax rotation with Kaiser normalization). For the first PCA the number of components was defined by Eigenval- ues �.1 and in a second analysis we constrained the model to three components, informed by the 3-factor structure of the adult HSP scale (Smolewska et al., 2006). We then selected 12 items, out of the 38 items, that were most similar to the highest loading items of the adult HSP scale as reported by Smolewska et al. (2006). The PCA was then repeated with the 12 selected items to verify whether items would load on the specific component they had been selected for. Next, we applied Confirmatory Factor Analyses (CFA) to the 12-item scale to test two competitive models (see Figure 2 for an illustration of the difference between the two models): (1) a 3-factor model with five items in factor 1 (Ease

Figure 2. Graphical illustration of (A) 3-factor model: EOE, LST and AES factors; (B) bifactor model: EOE, LST, and AES factors plus a HSC general factor. EOE � Ease of Excitation; AES � Aesthetic Sensitivity; LST � Low Sensory Threshold.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

56 PLUESS ET AL.

of Excitation), four items in factor 2 (Aesthetic Sensitivity) and three items in factor 3 (Low Sensitivity Threshold); and (2) a bifactor model which includes a shared general factor in addi- tion to the three separate factors based on recent findings which suggest that the adult HSP scale fits a bifactor model better than a 3-factor model (Lionetti et al., 2017). To test the bifactor model, one of the factor loadings in the general factor and one of the loadings in each of the domain specific factors were set to 1 (F. F. Chen, West, & Sousa, 2006). The robust maximum likelihood was used as estimation method. Two relative fit indices were considered for the evaluation of goodness of fit for each model: the Tucker Lewis index (TLI) and the comparative fit index (CFI), both of which perform well with small and large samples (the �2 statistic is extremely sensitive to sample size and not well suited for the current analysis). CFI and TLI values of � .95 and � .97, respectively, were considered as acceptable and good fit (Schermelleh-Engel, Moosbrugger, & Müller, 2003). The root mean square error of approximation (RMSEA) and the standardized root mean square residuals (SRMR) were also used. For RMSEA, values � .05 were considered as a good fit and values ranging from .05 and .08 as an adequate fit. For SRMR, values less than .08 were considered to reflect good fit (Schermelleh-Engel et al., 2003). The 3-factor and bifactor models were compared according to three criteria: (1) qualita- tive evaluation of the fit indices of each model; (2) the CFI criterion according to which the null hypothesis of no differ- ences between the two competing models should not be rejected if the difference in the CFIs between two nested models is smaller than |0.01| (Cheung & Rensvold, 2002); and (3) the scaled �2 difference test according to which the null hypothesis (i.e., no differences between the two competing models) should not be rejected if the associated p value is greater than .05 (Satorra, 2000) with lower �2 reflecting better model fit.

Internal consistency of the HSC scale was measured with Cron- bach’s alpha. A one-way ANOVA was conducted to test for ethnic differences in HSC and an independent samples t test to investigate gender differences. We then tested bivariate correlations between the mean of the 38 child sensitivity items, the mean of our newly created 12-item HSC scale and its subscales as well as behavioral inhibition and activation, temperament, and affect. Furthermore, we ran multivariate regression models to investigate convergent validity and to estimate how much of the variance in HSC was explained by related measures, including all HSC scales simulta- neously as dependent variables in the same model and, thus, taking the interdependence among variables into account. Finally, we tested divergent validity of the HSC scale with the heterotrait– monotrait (HTMT) ratio of correlations (Henseler, Ringle, & Sarstedt, 2015). The HTMT ratio represents the average of the correlations of items across different constructs (e.g., HSC, BIS, PA etc.) relative to the average of the correlations of items within the same construct (e.g., the 12 HSC items). HTMT ratio values that are equal or lower than .85 indicate that divergent validity is confirmed.

The level of significance for all analyses was set at � � .05. Analyses were conducted using R software (Rosseels, 2016; semTools Contributors, 2016) and SPSS version 20 (IBMCorp., 2011).

Results

Principal component and confirmatory factor analyses. Principal component analysis (PCA) of the HSC-38 resulted in nine principle components that accounted for 61% of the cumula- tive variance. However, the scree plot pointed toward a three- component solution. After constraining the PCA to three principle components, 40% of the variance was explained (see supplemen- tary information for detailed results). PCA of the 12 selected items suggested that the three principle components explained 55% of the cumulative variance. Table 1 shows the 12 selected items and their loadings on the three principal components, reflecting the same three factors as reported with the adult HSP scale (Smolewska et al., 2006).

The Confirmatory Factor Analysis (CFA) of the 3-factor model showed acceptable model fit with �2 � 106.84, df � 51, p � .001; RMSEA � .06, 90% CI [.05, .08]; CFI/TLI � .907/.880; SRMR � .06. Similar model fit indices emerged for the bifactor model (�2 � 94.804, df � 46, p � .001; RMSEA � .06, 90%, CIs [.05, .08]; CFI/TLI � .919/.884 SRMR � .06). However, although the two models showed comparable fit indices the CFI difference (CFI [DIFF] � .012) and the scaled �2 difference, �2[DIFF] � 11.8, df � 5, p � .04 between them suggests that the bifactor model is the better fitting solution (more details of these analyses are provided in the supplementary information document).

Descriptive statistics and internal reliability. Table 2 shows the mean values and standard deviations for the mean of the 38 child sensitivity items (HSC-38), the HSC total scale, the three HSC factors (Ease of Excitation, Aesthetic Sensitivity, and Low Sensory Threshold), and all other measures used in this study. The HSC scale showed adequate internal consistency with � � .79, 90% CIs [.75, .82]. HSC subscales showed acceptable but lower internal consistencies, which was to be expected considering the low item numbers in each subscale, with � � .71, CIs [.65, .76] for Ease of Excitation, � � .73, CIs [67–78] for Aesthetic Sensitivity, and � � .66, CIs [.58, .72] for Low Sensory Threshold. There were no significant differences in HSC as a function of ethnicity (F(51) � 1.21, p � .45). A small gender difference was observed, with females (M � 4.41, SD � .93) scoring significantly higher than males (M � 4.07, SD � 1.08) with t(283) � �2.55, p � .05.

Bivariate correlations. Bivariate associations between all variables are reported in Table 3. Most importantly, the mean of the12-item HSC scale is highly correlated with the mean of the 38 child HSP items (r � .93). BIS and BAS are correlated with HSC and the three subscales except for Low Sensory Threshold which was not associated with BAS. Regarding temperament, Effortful Control, Negative and Positive Emotionality were correlated with HSC and all subscales except for Low Sensory Threshold, which was not correlated with Positive Emotionality. Finally, Positive Affect was positively correlated with Aesthetic Sensitivity (r � .41) and Negative Affect with Ease of Excitation (r � .16) and Low Sensory Threshold (r � .13). (Bivariate correlations between the EQTAR subscales and HSC are provided in supplementary information).

Multivariate regression. The first model, which included BIS, BAS, EC, PE, NE, PA, and NA as predictor variables of HSC explained 34% of the variance. The second model with the three subscales as outcomes explained 30% of the variance of Ease of Excitation, 35% of Aesthetic Sensitivity, and 17% of Low Sensory

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

57ENVIRONMENTAL SENSITIVITY

Threshold. Standardized parameter estimates and associated p values are reported in Table 4.

Divergent validity. Heterotrait-monotrait (HTMT) ratio of correlations values for each pair of measures ranged from .14 for Ease of Excitation-PA to .67 for Ease of Excitation-BIS, suggest- ing that divergent validity was established. Furthermore, associa- tions among the HSC total score and subscales Ease of Excitation, Low Sensory Threshold and Aesthetic Sensitivity were consis- tently higher than associations between HSC and other measures (detailed HTMT results are provided in the supplementary infor- mation document).

Discussion

According to Study 1, the mean of the 12-item Highly Sensitive Child scale was strongly associated with the mean of the 38 child sensitivity items but reflected the identical 3-factor structure as the adult scale. Importantly, confirmatory factor analyses suggested that although the measure consists of three distinct subscales, these subscales also load on a general factor of sensitivity. Hence, the total mean score of the scale can be used to indicate Environmental Sensitivity even though the three subscales appear to capture different components of sensitivity. For example, Aesthetic Sen-

Table 2 Means and Standard Deviations of All Measures (Study 1, 2, 3, and 4)

Measure Study 1 Study 2

Study 3

Study 4Session 1 Session 2

HSC-38 4.15 (.90) — — — — HSC 4.33 (.98) 4.68 (.93) 4.01 (.86) 4.04 (.84) 3.98 (.96) HSC-EOE 4.13 (1.18) 4.59 (1.21) 3.70 (1.26) 3.67 (1.14) 3.81 (1.37) HSC-AES 5.15 (1.23) 5.56 (1.08) 5.15 (1.02) 5.23 (.91) 5.16 (1.00) HSC-LST 3.58 (1.53) 3.67 (1.68) 3.01 (1.32) 3.10 (1.29) 2.70 (1.38) BIS 18.88 (4.04) 19.66 (3.58) — — — BAS 37.36 (7.51) 39.11 (6.68) — — — EC 3.14 (.60) 3.30 (.57) — — — NE 3.00 (.58) 3.06 (.62) — — — PE 3.09 (.54) 3.26 (.52) — — — PA 44.54 (9.95) — — — — NA 27.70 (10.7) — — — — Neuroticism — — — — 15.97 (4.37) Extraversion — — — — 21.75 (3.92) Openness — — — — 21.70 (3.66) Agreeableness — — — — 21.94 (3.52) Conscientiousness — — — — 22.41 (3.65)

Note. HSC-38 � Mean of 38 Highly Sensitive Child Items; HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; BIS � Behavioral Inhibition System; BAS � Behavioral Activation System; EC � Effortful Control; NE � Negative Emotionality; PE � Positive Emotionality; PA � Positive Affect; NA � Negative Affect.

Table 3 Bivariate Correlations (Study 1)

Measure 1 2 3 4 5 6 7 8 9 10 11 12 13

1. HSC-38 — 2. HSC .93�� — 3. HSC-EOE .80�� .86�� — 4. HSC-AES .68�� .71�� .43�� — 5. HSC-LST .63�� .69�� .44�� .18�� — 6. BAS .42�� .41�� .31�� .50�� .11 — 7. BIS .55�� .55�� .49�� .38�� .36�� .62�� — 8. PE .29�� .27�� .17�� .37�� .08 .40�� .32�� — 9. NE .38�� .37�� .36�� .19�� .26�� .21�� .40�� .61�� —

10. EC .29�� .27�� .18�� .29�� .15� .39�� .33�� .82�� .71�� — 11. PA .16�� .14� �.01 .41�� �.06 .38�� .14� .34�� .08 .33�� — 12. NA .15� .09 .16�� �.09 .13� �.08 .10 .04 .19�� �.02 �.38�� — 13. Age �.10 �.10 �.04 �.17�� �.02 �.18�� �.19�� �.18�� �.12� �.21�� �.15�� .30�� — 14. Gender .18�� .15� .10 .10 .15� .06 .19�� .09 .13� .10 �.08 .08 �.01

Note. HSC-38 � Mean of 38 Highly Sensitive Child Items; HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; BIS � Behavioral Inhibition System; BAS � Behavioral Activation System; EC � Effortful Control; NE � Negative Emotionality; PE � Positive Emotionality; PA � Positive Affect; NA � Negative Affect; Gender: 1 � male, 2 � female. Bold values indicate statistically significant results. � p � .05. �� p � .01.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

58 PLUESS ET AL.

sitivity seems to capture sensitivity to more positive aspect of the environment, indicated by correlations with the behavioral activa- tion system (BAS) and positive emotionality and affect, whereas Ease of Excitation and Low Sensory Threshold tend to reflect sensitivity to more negative contextual factors, as shown in cor- relations with the behavioral inhibition system (BIS) as well as negative emotionality and affect. This may also explain why the total score was associated with both negative and positive emo- tionality. Finally, multivariate regression analyses provided evi- dence that Environmental Sensitivity as measured with the HSC scale does not simply reflect effects of well-known temperament traits and affect. Divergent validity was further supported by heterotrait-monotrait ratio of correlations analysis.

Study 2

To replicate the findings of Study 1, the same psychometric properties and associations with temperament, behavioral inhibi- tion, and activation were investigated in an independent sample.

Method

Participants. The sample included 258 children (113 girls and 145 boys) from a secondary school in East London, United King- dom. Children were on average 11.17 years old (range � 11–12 years, SD � .38) and were of ethnically diverse backgrounds: White (20.9%), African/Caribbean (20.2%), Asian (34.9%), Mid- dle Eastern (4%) and mixed-ethnicity (23.3%).

Procedure and measures. Children completed all measures on a computer during regular class at school. To measure Envi- ronmental Sensitivity, the 12-item HSC was used rather than the 38 child sensitivity items. In addition, children also reported on behavior inhibition and activation with the BIS-BAS (Carver & White, 1994) and on temperament with the EATQR (Capaldi & Rothbart, 1992). Measures were used exactly the same way as described in Study 1. However, positive and negative affect (PA- NAS) were not measured in this sample.

Data analysis. The same methods and statistical analyses were applied as described in detail in Study 1.

Results

Confirmatory factor analysis. The confirmatory factor anal- ysis on the 12 items showed good model fit for the 3-factor model

(�2 � 63.019, df � 51, p � .12; RMSEA � .03, 90% CIs [.00, .05]; CFI/TLI � .968/.959; SRMR � .05). For the bifactor model, the negative variance of one statistically nonsignificant Ease of Excitation item was fixed to 0 (F. F. Chen et al., 2006). The results of the bifactor model were satisfactory: �2 � 48.73, df � 46, p � .48; RMSEA � .01, 90% CIs [.00, .04]; CFI/TLI � .995/.994; SRMR � .04. The 3-factor and bifactor models showed compa- rable fit indices with slightly stronger support for the bifactor model. The CFI difference was significant and equal to .027— confirmed by a significant scaled �2 difference, �2[DIFF] � 13.1, df � 4, p � .01—and, thus, supporting the use of both the HSC total score as well as the individual Ease of Excitation, Aesthetic Sensitivity and Low Sensory Threshold subscales.

Descriptive statistics and internal reliability. Table 2 shows the mean scores and standard deviations for HSC, the three HSC subscales and all other measures used in this sample. The HSC scale showed acceptable internal consistency with a Cronbach’s alpha of .72, 90% CIs [.66, .77] whereas the HSC subscales had slightly lower internal consistencies with � � .66, 90% CIs [.59, .72] for Ease of Excitation, � � .62, 90% CIs [.54, .69] for Aesthetic Sensitivity, and � � .63, CIs [.54, .70] for Low Sensory Threshold. Consistent with Study 1 there were no significant differences in HSC as function of ethnicity (F(48) � 1.27, p � .13) but the gender difference was only marginally significant (t(245) � �1.93, p � .06).

Bivariate correlations. Similar to Study 1, all HSC scales were positively correlated with both BIS and BAS except for Low Sensory Threshold which was not associated with BAS (see Table 5). The strongest associations with BIS/BAS emerged between Ease of Excitation and BIS, and between Aesthetic Sensitivity and the BAS (r � .29 and r � .35, respectively). Regarding temperament, Effortful Control, Neg- ative and Positive Emotionality were associated with all HSC scales. However, the correlation between Ease of Excitation and Negative Emotionality and between Aesthetic Sensitivity and Positive Emotionality stood out (r � .49 and r � .50, respec- tively). (Bivariate correlations between the EQTAR subscales and HSC are provided in supplementary information).

Multivariate regression. The multivariate regression models included BIS, BAS, EC, PE and NE as predictor variables of HSC and subscales. The model predicting HSC explained 26% of the variance and the model predicting the subscales explained 26% of

Table 4 Multivariate Regression (Study 1)

Measure

HSC HSC-EOE HSC-AES HSC-LST

� z p � z p � z p � z p

BAS .13 1.73 .08 .14 1.72 .09 .26 3.56 <.01 �.11 �1.29 .20 BIS .38 5.36 <.01 .33 4.31 <.01 .16 2.37 .02 .37 4.39 <.01 PE .01 .09 .93 �.06 �.57 .57 .26 3.29 <.01 �.187 �1.38 .17 NE .24 3.24 <.01 .34 4.19 <.01 .01 .19 .85 .16 1.57 .12 EC �.12 �1.20 .23 �.18 �1.76 .08 �.18 �1.89 .06 .12 .89 .38 PA .10 1.53 .13 �.02 �.22 .83 .28 3.86 <.01 �.01 �.22 .83 NA .09 1.64 .10 .10 1.52 .13 .04 .70 .48 .07 1.21 .23

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; BIS � Behavioral Inhibition System; BAS � Behavioral Activation System; EC � Effortful Control; NE � Negative Emotionality; PE � Positive Emotionality. Two models were run, the first including the HSC total score as the only dependent variable and the second model with EOE, AES, and LST simultaneously included as dependent variables. Bold values indicate statistically significant results.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

59ENVIRONMENTAL SENSITIVITY

the variance of Ease of Excitation, 26% of Aesthetic Sensitivity, and 15% of Low Sensory Threshold (see Table 6).

Divergent validity. HTMT values for each pair of constructs ranged from .12 for Low Sensory Threshold-BAS to .71 for Aesthetic Sensitivity-PE and, hence, confirm divergent validity. Associations between the HSC total score and its subscales were consistently higher than association with the other measures (see supplementary information document for HTMT results).

Discussion

The findings of Study 2 confirm the bifactor structure of the HSC measure, suggesting that the total scale reflects three com- ponents whose items also load on a general sensitivity factor. The bivariate correlations provide further suggestive evidence that Aesthetic Sensitivity may reflect sensitivity to more positive en- vironmental aspects, whereas Ease of Excitation (and Low Sensory Threshold) seems to capture sensitivity to more negative contex- tual factors (with the total HSC score correlating again with both negative and positive emotionality, see discussion in Study 1). According to the regression results the different temperament traits fail to account for the majority of the variance of HSC, suggesting that Environmental Sensitivity is not fully explained or captured by existing concepts as confirmed in the heterotrait-monotrait ratio of correlations findings.

Study 3

Study 3 aimed at investigating test–retest reliability of the created 12-item HSC measure in an independent child sample.

Method

Participants. Data for this study were obtained from the Pic- tures and Words Study (PAWS). PAWS is a longitudinal study of information processing and mood featuring a sample of 155 chil- dren (Brown et al., 2014). Data were collected across three data waves with children recruited from two primary schools in Lon- don. For the current study, data were collected during the third wave resulting in a sample of 104 children (59 girls and 45 boys) with a mean age of 9.82 years (range � 8–11 years, SD � .45). Eighty-one percent of the sample identified as white.

Procedure and measures. The original study included sev- eral psychological measures of information processing and mood. For the current study, only data from the 12-item HSC scale collected at the third wave of data collection were used. The third wave of data collection comprised two data collection sessions scheduled to take place approximately two-three weeks apart (mean interval � 15 days, range 9–22 days, SD � 2.46). Children were seen individually in a quiet classroom and completed a computerized version of the HSC scale at both sessions (via

Table 5 Bivariate Correlations (Study 2)

Measure 1 2 3 4 5 6 7 8 9 10

1. HSC — 2. HSC-EOE .83�� — 3. HSC-AES .61�� .32�� — 4. HSC-LST .69�� .37�� .11 — 5. BAS .25�� .23�� .35�� �.01 — 6. BIS .32�� .29�� .24�� .15� .66�� — 7. PE .41�� .28�� .50�� .15� .59�� .44�� — 8. NE .50�� .49�� .25�� .31�� .37�� .50�� .39�� — 9. EC .48�� .40�� .43�� .23�� .61�� .55� .67�� .59�� —

10. Age .09 .05 .10 .07 .03 .02 �.08 �.12 �.06 — 11. Gender .12 .06 .02 .19�� .10 .12 .10 .22�� .05 .02

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; BIS � Behavioral Inhibition System; BAS � Behavioral Activation System; EC � Effortful Control; NE � Negative Emotionality; PE � Positive Emotionality; Gender: 1 � male; 2 � female. Bold values indicate statistically significant results. � p � .05. �� p � .01.

Table 6 Multivariate Regression (Study 2)

Measure

HSC HSC-EOE HSC-AES HSC-LST

� z p � z p � z p � z p

BAS �.16 �1.94 .05 �.06 �.63 .53 .07 .84 .40 �.34 �3.71 <.01 BIS .04 .48 .63 .02 .28 .78 �.07 �1.50 .29 .12 1.31 .19 PE .24 2.69 .01 .05 .51 .61 .40 5.24 <.01 .13 1.28 .20 NE .30 4.08 <.01 .39 4.37 <.01 �.03 �.41 .69 .23 2.57 .01 EC .19 .08 .08 .14 1.21 .23 .15 1.42 .16 .14 1.15 .25

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; BIS � Behavioral Inhibition System; BAS � Behavioral Activation System; EC � Effortful Control. Two models were run, the first including the HSC total score as the only dependent variable and the second model with EOE, AES, and LST simultaneously included as dependent variables. Bold values indicate statistically significant results.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

60 PLUESS ET AL.

EPrime 2.0) with responses made using the computer keyboard. Items were presented onscreen but also read aloud to ensure comprehension.

Data analysis. Internal reliability of the 12-item HSC scale was examined with Cronbach’s alpha and test–retest reliability was calculated by correlating scores for HSC and the three sub- scales from Session 1 with scores of repeated measurement at Session 2. A test–retest reliability of .70 or higher was considered adequate (McCrae, Kurtz, Yamagata, & Terracciano, 2011).

Results

Descriptive statistics and internal reliability. Mean scores and standard deviations for the HSC sum score and the three subscales are provided in Table 2, separately for each of the two data collection sessions. The HSC scale showed acceptable inter- nal consistency with � � .71 and .74 for Session 1 and Session 2, respectively. The subscales showed lower internal consistencies with � � .73/.69 for Ease of Excitation, � � .49/.46 for Aesthetic Sensitivity, and � � .49/.55 for Low Sensory Threshold.

Test–retest reliability. Test–retest reliability estimates for the HSC score (r � .68) and the subscales (r � .57-.78) are presented in Table 7 and were acceptable. Furthermore, estimates remained stable when the interval between data collection sessions was partialed out.

Discussion

Findings of Study 3 confirm the internal consistency found in Studies 1 and 2 and suggest that test–retest reliability of the HSC scale is acceptable in a sample of children whose ages range from 8–11 years. Although there is substantial stability across measure- ments, mean scores do show some variability over time, which suggests that the measure may pick up measurement error or short-term changes in self-reported sensitivity. It is conceivable that stability would be higher at older age, which remains to be tested.

Study 4

In Study 4 the performance of the developed 12-item HSC scale was tested in a large sample of adolescents followed by exploring associations with the Big Five personality traits.

Method

Participants. Data for Study 4 were obtained from a subset of the Twins Early Development Study (TEDS), a large epidemiolog-

ical study of more than 16,000 twin pairs born in England and Wales in 1994, 1995, and 1996. TEDS includes extensive data on various aspects of development, including cognitive abilities, per- sonality, behavior, educational achievement and family environ- ment, collected at regular intervals from a sample that is represen- tative of the U.K. population (Kovas et al., 2007). Data and recruitment procedures are reported in detail elsewhere (Haworth, Davis, & Plomin, 2013). Data on the 12-item HSC scale were collected for 2,945 individuals when twins were approximately 17 years old. Data on the Big-Five personality traits were available for a subset of the same sample (1,174 individuals). Participants with severe medical disorders, history of perinatal complications, or unknown zygosity were excluded from the analyses (n � 77). Furthermore, only data from one sibling per twin pair were in- cluded (random selection) to account for relatedness between individuals in this particular sample. The final HSC sample in- cluded 1,431 adolescents (595 males, 836 females), with a mean age of 17.06 (range � 15–19 years, SD � .88) on return of the HSC questionnaires. The ethnicity of the majority (93%) of the sample was identified as Caucasian.

Procedure and measures. Data for the measures used in the current analysis were obtained by self-report via online or paper questionnaires.

Environmental Sensitivity was measured with the 12-item HSC scale. Big-Five personality traits Agreeableness, Extraversion, Neuroticism, Openness to Experience and Conscientiousness were measured with the 30 item Five Factor Model Rating Form (Mullins-Sweatt, Jamerson, Samuel, Olson, & Widiger, 2006). Items (e.g., “fearful, apprehensive versus relaxed, unconcerned, cool,” “strange, odd, peculiar, creative versus pragmatic, rigid.”) were rated on a Likert scale ranging from 1 � low to 5 � high. Higher scores indicate higher levels of the personality trait. Inter- nal reliability consistency of the scale was acceptable with � � .73 for Neuroticism, � � .70 for Extraversion, � � .65 for Openness, � � .65 for Agreeableness, and � � .75 for Conscientiousness.

Data analysis. The factor structure (confirmatory factor anal- ysis) and internal reliability of the HSC scale was examined by applying the same methodological approach as in Studies 1 and 2. Associations between HSC, HSC subscales and the Big-Five per- sonality traits were investigated with bivariate correlations. Fur- thermore, multivariate regression and heterotrait-monotrait ratio of correlations analysis were applied to investigate divergent validity, following the same procedures adopted in Studies 1 and 2.

Results

Confirmatory factor analysis. The 3-factor model (Ease of Excitation, Aesthetic Sensitivity, Low Sensory Threshold) yielded good model fit (�2 � 323.88, df � 51, p � .001; RMSEA � .06, 90% CIs [.06, .07], CFI/TLI � .935/.91; SRMR � .05). The bifactor model also fit the data well (�2 � 286.53, df � 46, p � .001, RMSEA � .06, 90% CIs [.05, .07], CFI/TLI � 945/921, SRMR � .70). The two models showed comparable fit indices with slightly stronger support for the bifactor model. The CFIs difference was trivial (equal to .01) though in the presence of a significant scaled �2 difference, �2[DIFF] � 47.2, df � 5, p � .001.

Descriptive statistics and internal reliability. Mean scores and standard deviations for HSC, the three HSC subscales, and the Big-Five personality traits are presented in Table 2. Females (M �

Table 7 Test–Re-Test Reliability of HSC Across 15 Days (Study 3)

Measure r

HSC .68��

HSC-EOE .66��

HSC-AES .57��

HSC-LST .78��

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Exci- tation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold. �� p � .01.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

61ENVIRONMENTAL SENSITIVITY

4.13, SD � .96) scored significantly higher than males (M � 3.78, SD � .92) with t(1429) � 6.81, p � .001. Internal consistency was good for the HSC total scale (� � .82) and acceptable for the subscales (Ease of Excitation with � � .81; Aesthetic Sensitivity with � � .65; Low Sensory Threshold with � � .71).

Bivariate correlations. Unadjusted associations between HSC and the Big-Five personality traits are presented in Table 8. HSC was positively associated with Neuroticism (r � .31) and Openness (r � .18) and negatively with Extraversion (r � �.18) but did not correlate with Agreeableness and Conscientious. Al- though Ease of Excitation and Low Sensory Threshold correlated with Neuroticism (r � .38, and r � .22, respectively) and Extra- version (r � �.28 and r � �.22, respectively), Aesthetic Sensi- tivity was not associated with Neuroticism but correlated posi- tively with Extraversion (r � .20), Openness (r � .25), and Conscientiousness (r � .16).

Multivariate regression. The multivariate regression model with the five personality traits as predictor variables explained 14% of the variance of HSC. A second model with the HSC subscales as outcome variables explained 17% of the variance of Ease of Excitation, 10% of Aesthetic Sensitivity, and 14% of Low Sensory Threshold (see Table 9 for the standardized parameter estimates).

Divergent validity. HTMT values ranged from .12 for Low Sensory Threshold–Conscientiousness to .48 for Ease of Excitation- Neuroticism providing evidence of divergent validity. Similar to the previous studies reported in this paper, associations among the HSC total score and subscales Ease of Excitation, Low Sensory Threshold and Aesthetic Sensitivity were consistently higher than associations with other measures (detailed results are provided in the supplemen- tary information document).

Discussion

The 12-item HSC scale performed just as well with 15–19 years old adolescents as with 8–12 years old children. The observation that a bifactor model fit the data best further confirms that the scale reflects both a general sensitivity factor and three separate sensi- tivity components. Bivariate correlations also provide additional evidence that the subscales capture different aspects of sensitivity with Aesthetic Sensitivity reflecting Openness and to a lesser

degree Conscientiousness, while Ease of Excitation and Low Sen- sory Threshold are associated with higher Neuroticism and lower Extraversion. Future studies should investigate these correlations further by considering associations with the different facets of the identified personality traits. However, in the current study all five personality traits accounted for no more than 14% of the variance of HSC and the .85 HTMT criterion was always met, suggesting that HSC is not well captured with common personality traits and that divergent validity is established.

Study 5

The aim of Study 5 was to explore whether there exist different sensitivity groups as suggested by theory (e.g., Aron et al., 2012; Boyce & Ellis, 2005) and empirical studies (e.g., Wolf et al., 2008). Although Environmental Sensitivity—like many other per- sonality traits—is a continuous and normal distributed dimension (see supplementary information for the distribution of HSC in all the samples included in this paper), people may fall into different sensitivity categories. For example, Boyce and Ellis (2005) sug- gested that there are two kinds of children: “Orchids,” who are more sensitive to their environment, requiring particularly support- ive contexts to thrive, and “Dandelions,” who are less sensitive and do well in most environments. The general understanding is that about 20–30% of the population fall into the highly sensitive Orchid-category and 70–80% into the less sensitive Dandelion- category (e.g., Aron et al., 2012; Boyce & Ellis, 2005). However, this proposition, although very popular, has not yet been tested empirically with HSC data. In the current study we applied Latent Class Analysis—a data-driven and hypothesis-free approach—to the combined samples of Studies 1 and 2 (children) as well as to the sample of Study 4 (adolescents) to investigate, for the first time, the existence of two or more sensitivity groups in children and adolescents. A recent similar analysis in multiple adult sam- ples featuring the 27-item HSP scale, yielded a three- rather than a two-class solution with 31% of the sample population falling into a high sensitive group, about 40% into a medium sensitive group, and the remaining 29% into a low sensitive groups (Lionetti et al., 2017). In keeping with the Orchid–Dandelion metaphor, individ- uals belonging to the medium sensitive group have been described

Table 8 Bivariate Correlations (Study 4)

Measure 1 2 3 4 5 6 7 8 9 10

1. HSC — 2. HSC-EOE .89�� — 3. HSC-AES .58�� .29�� — 4. HSC-LST .74�� .54�� .18�� — 5. Neuroticism .31�� .38�� �.00 .22�� — 6. Extraversion �.18�� �.27�� .20�� �.22�� �.36�� — 7. Openness .18�� .05 .25�� .17�� �.05 .27�� — 8. Agreeableness .03 �.03 .04 .08 �.21�� .19�� .25�� — 9. Conscientiousness �.01 �.13�� .16�� .03 �.19� .29�� .09� .26�� —

10. Age .02 .01 .07�� �.01 �.01� .05 .04 .04 �.02 — 11. Gender .18�� .15�� .07�� .18�� .22�� �.04 .08 .12�� .08 .03

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold; Gender: 1 � male; 2 � female. Bold values indicate statistically significant results. � p � .05. �� p � .01.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

62 PLUESS ET AL.

as “Tulips,” who are less sensitive than Orchids but more sensitive than Dandelions.

In addition to testing for the existence of different sensitivity groups, Study 5 also aimed at exploring whether it would be possible to identify preliminary cut-off scores that could be used to determine the specific sensitivity group individual children and adolescents fall into based on their HSC scores.

Method

Participants. Study 5 made use of the samples used in studies 1, 2, and 4. The samples from Studies 1 and 2 were combined into a large child sample with N � 592. For the adolescent sample we used the twin sample from Study 4 which included one randomly selected sibling from each twin pair (sample A, n � 1,470). To replicate findings in adolescents, we reran the same analysis on the other half of the sample made up of the nonselected twin pair sibling (sample B, n � 1,473).

Procedures and measures. All participants provided data on the same 12-item HSC scale (see Study 1 for more details). These 12 items were the basis for the Latent Class Analysis.

Data analysis. To test for the existence of different sensitivity groups we performed a series of Latent Class Analyses (LCA) on the HSC scale, testing models with 1 to 6 classes. The optimal number of classes was determined based on the following criteria: (a) Akaike’s Information Criterion (AIC), (b) Bayesian Informa- tion Criterion (BIC), (c) Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-A), and (d) Entropy. AIC and BIC are comparative indices, the lower the values the better the model. The LMR-A compares the fit of the specified class solution to a model with one fewer class. A significant p value suggests that the specified model provides a better fit to the data than the more parsimonious model. Entropy refers to the confidence with which individuals can be clearly categorized into the different classes, with values approach- ing 1 indicative of a clear delineation of class membership (Ny- lund, Asparouhov, & Muthén, 2007). Once the optimal number of sensitivity classes was determined, based on the criteria outlined above as well as in consideration of theory and parsimony, we investigated the distribution and overlap between the different sensitivity classes to identify exploratory cut-off scores for chil- dren and adolescents. Sensitivity (i.e., probability of correctly identifying all individuals that belong to a particular group) and specificity (i.e., probability of correctly identifying those individ- uals that do not belong to the particular group) for these cut-off

scores were estimated by comparing agreement between LCA class membership and the categorization based on the proposed cut-off scores. For children, the agreement between LCA class membership and cut-off categorization was estimated within the same sample. For adolescents, the agreement was tested using two samples by applying cut-offs based on the sample A LCA results in sample B.

Latent Class Analysis was performed using Mplus version 7 (Muthén & Muthén, 1998–2015) and program R (Pastore, 2016) was used for visual inspection of class distributions.

Results

Latent class analysis for child sample. Model fit indices are reported in Table 10. The one-class model had the highest AIC and

Table 9 Multivariate Regression (Study 4)

Measure

HSC HSC-EOE HSC-AES HSC-LST

� z p � z p � z p � z p

Neuroticism .28 6.39 <.01 .31 6.83 <.01 .07 1.44 .15 .18 3.88 <.01 Extraversion �.15 �3.33 <.01 �.17 �3.86 <.01 .14 2.96 <.01 �.25 �5.29 <.01 Openness .19 4.31 <.01 .07 1.45 .15 .22 4.62 <.01 .21 5.21 <.01 Agreeableness .04 .87 .39 .05 1.11 .27 �.06 �1.26 .21 .07 1.70 .09 Conscientiousness .04 1.03 .30 �.06 �1.34 .18 .12 2.77 <.01 .10 2.33 .02

Note. HSC � Highly Sensitive Child Scale; HSC-EOE � Ease of Excitation; HSC-AES � Aesthetic Sensitivity; HSC-LST � Low Sensitivity Threshold. Two models were run, the first including the HSC total score as the only dependent variable and the second model with EOE, AES, and LST simultaneously included as dependent variables. Bold values indicate statistically significant results.

Table 10 Latent Class Analysis (Study 5)

Model AIC BIC LMR-A (p) Entropy

Children

1 class 25727.97 25830.56 2 classes 25086.17 25244.34 659.71 (.016)� .77 3 classes 24682.61 24896.35 410.49 (<.001)�� .85 4 classes 24535.97 24805.29 170.55 (.159) .82 5 classes 24344.16 24669.04 215.18 (.196) .86 6 classes 24259.14 24639.59 109.67 (325) .84

Adolescents (subsample A)

1 class 67497.81 67624.84 2 classes 64639.20 64835.04 2854.50 (�.001)�� .82 3 classes 63703.86 63968.51 951.30 (<.001)�� .80 4 classes 63141.62 63475.08 582.10 (.003)�� .80 5 classes 62718.95 63121.22 443.98 (.002)�� .82 6 classes 62465.54 62936.62 276.49 (.314) .80

Adolescents (subsample B)

1 class 67030.28 67157.36 2 classes 64286.91 64482.82 2740.47 (�.001)�� .81 3 classes 63352.08 63616.83 950.81 (<.001)�� .81 4 classes 62873.54 63207.13 499.27 (.07) .82 5 classes 62542.41 62944.83 353.40 (.06) .81 6 classes 62353.56 62824.82 212.61 (.56) .82

Note. Subsample A refers to data used in Study 4; Subsample B refers to the other half of the TEDS sample described in Study 4. Bold values indicate best fitting models. � p � .05. �� p � .01.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

63ENVIRONMENTAL SENSITIVITY

BIC values (25727.97 and 25830.56, respectively). The 2-class model had the lowest entropy (.77), but was significantly better than the baseline model with one class according to LMR-A (p � .02). The three-class model yielded a significant LMR-A (p � .001), entropy increased to .85, and BIC and AIC values decreased substantially (24682.61 and 24896.35, respectively), suggesting that the three-class model fit the data significantly better than the two-class model. Models with four, five or six classes were ex- plored but all rejected because none of them had a significantly better fit than the three-class model.

According to the best fitting three-class LCA model, 24.67% of children belong to a low sensitive group, 41.24% of children to a medium sensitive group, and 34.08% to a high sensitive group. Means and standard deviations of HSC and subscales for each of the three groups are reported in Table 11.

Latent class analysis with adolescent sample. Results of the different models are reported in Table 10. For sample A (same as in Study 4) the one-class model yielded the highest AIC and BIC values (67497.81 and 67157.04, respectively). The two-class model had a better fit to data when compared with the one-class model, but it was the three-class model that fitted data best, presenting lower AIC and BIC scores compared with the two-class model. LMR-A results also confirmed the three-class solution as significantly better than the two-class model (p � .001) and entropy was satisfactory with .80. Models with four, five, and six classes were also explored. LMR-A results suggested an improve-

ment in models with four and five classes while entropy remained constant. The four-class solution identified an additional class with 13.2% of the sample characterized with particularly low HSC scores while the three groups identified in the three-class model (low, medium, high) remained largely unchanged. The five-class solution identified one additional medium class, between medium and high groups, on top of the four-class model with 11.9% of the sample. However, the three initial classes (low, medium, high) remained. Considering these findings in light of the results of the child sample and in combination with the parsimony principle in selecting the best number of classes (Masyn, 2013), the three-class solution was identified as the best candidate.

To explore the robustness of the three-class solution further, we repeated the LCA in the other half of the TEDS sample (sample B with n � 1,473). Again, the two-class model was significantly better than the one-class baseline model. The three-class model had a significantly better fit that the two-class model, manifested in lower AIC and BIC scores compared with the two-class model. In contrast to findings with sample A, data from sample B sug- gested that models with four, five, and six classes did not fit the data better than the three-class model better (see Table 10).

The three-class solution for both adolescent samples was similar to the one that emerged for the child sample: 34.90–34.98% of adolescents belonged to a low sensitive group, 41.04–46.90% to a medium sensitive group, and 21.20–23.97% to a highly sensitive group. For means and standard deviations of HSC and subscales for each of the three classes see Table 11.

Exploratory cut-off scores for child sensitivity groups. Intersection points between the distributions of HSC scores for the three sensitivity groups are presented in Figure 3. The overlap of distributions suggest 4.17 and 4.75 as the intersection points for the low and high sensitivity group, respectively, resulting in the following exploratory cut-off scores: � 4.17 for the low-sensitive HSC group, � 4.17 and � 4.75 for the medium-sensitive, and �4.75 for the high-sensitive group. Applying these cut-off scores to the sample and comparing the resulting categorization with the LCA classification, we obtained a sensitivity of .51 (i.e., 51% of children were correctly categorized as members of the specific sensitivity group) and specificity of .78 (i.e., 78% of children were correctly identified as not being part of the specific sensitivity group) for the low-sensitive versus medium-sensitive group and a sensitivity of .77 and specificity of .72 for the medium-sensitive versus high-sensitive groups.

Exploratory cut-off scores for adolescent sensitivity groups. Given that the adolescent sample was based on twin pairs which were randomly divided into two subsamples, we were able to determine cut-off criteria in subsample A and then apply them to subsample B. Intersection points between the three groups in sample A were 3.64 between low and medium sensitive groups and 4.65 between medium and high sensitive groups (see Figure 4), resulting in the following cut-off scores: � 3.64 for the low- sensitive HSC group, � 3.64 and � 4.65 for the medium-sensitive, and �4.65 for the high-sensitive group. Applied to sample B, satisfactory sensitivity and specificity values emerged with .88 and .92, respectively, for the classification between low-sensitive and medium-sensitive individuals, and .69 and .86 for medium- sensitive versus high-sensitive ones.

Table 11 Descriptives for the Three Latent Classes (Study 5)

Groups 1 2 3

Children

Frequency 24.67% 41.24% 34.08%

Mean (SD) Mean (SD) Mean (SD)

HSC 3.68 (.80) 4.24 (.67) 5.39 (.63) HSC-EOE 3.68 (1.02) 4.06 (1.13) 5.17 (.97) HSC-AES 3.87 (.84) 5.74 (.85) 5.91 (.78) HSC-LST 3.42 (1.29) 2.54 (1.12) 5.07 (1.14)

Adolescents (subsample A)

Frequency 34.98% 41.04% 23.97%

Mean (SD) Mean (SD) Mean (SD)

HSC 3.00 (.51) 4.22 (.45) 5.06 (.63) HSC-EOE 2.38 (.72) 4.33 (.79) 5.07 (.97) HSC-AES 4.70 (1.12) 5.41 (.77) 5.40 (.85) HSC-LST 1.76 (.77) 2.65 (.88) 4.56 (.92)

Adolescents (subsample B)

Frequency 34.90% 46.90% 21.20%

Mean (SD) Mean (SD) Mean (SD)

HSC 3.02 (.50) 4.20 (.46) 5.08 (.60) HSC-EOE 2.36 (.69) 4.28 (.76) 5.13 (.96) HSC-AES 4.82 (1.06) 5.32 (.81) 5.48 (.91) HSC-LST 1.70 (.72) 2.55 (.90) 4.46 (.91)

Note. Subsample A refers to data used in Study 4; Subsample B refers to the other half of the TEDS sample described in Study 4.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

64 PLUESS ET AL.

Discussion

Consistent with theory, the Latent Class Analyses confirmed the existence of a highly sensitive group making up 20–35% of the population. However, the best fitting models suggested three rather than two distinct sensitivity groups, a new finding which none of the current theories on Environmental Sensitivity predicted. Be- sides the highly sensitive group (20–35%), there was also a medium sensitivity group (approx. 41–47%) and a low sensitive group (approx. 25–35%). The three group solution emerged con- sistently across all three samples. Importantly, these LCA results are very similar to the three-class solution that emerged recently when the same analysis was conducted in adult samples with the 27-item HSP scale (Lionetti et al., 2017), suggesting that there are not only Orchid- and Dandelion- (Boyce & Ellis, 2005) but also Tulip-children. The exploratory cut-off scores for the categoriza- tion of individuals into the three different sensitivity groups were characterized by moderate to weak sensitivity and specificity, performing better in adolescents than children. While the cut-off scores between medium and highly sensitive individuals were similar for children and adolescents, the cut-off scores between medium and low sensitive groups differed as a function of age with a higher cut-off score found in children. Some of this difference might be explained by the observation that the overlap between medium and low sensitive groups was substantially higher in the child compared with the adolescent sample. This suggests that it may be more difficult to differentiate between low and medium sensitive children at age 11 compared with adolescents at 17. However, given that measurement invariance of the HSC scale has not been tested and demonstrated yet, this interpretation has to be considered preliminary at this stage. Taking results from adult samples into account (Lionetti et al., 2017), we propose a general average cut-off of 3.8 between the low and medium sensitivity groups and a general average cut-off of 4.7 between the medium and high sensitivity groups. However, in the absence of validation studies confirming that these three groups capture qualitative dif- ferences, these exploratory cut-off scores should only be used as rough indicators of an individual’s sensitivity group membership when considering their position on the continuous HSC/HSP scales which range from 1 to 7. Given that HSC mean scores may

vary between cultures, which is yet to be investigated, it may be more helpful to divide a sample into bottom 30% (i.e., low sensi- tive group) and top 30% (i.e., highly sensitive group) with the remaining 40% making up the medium sensitive individuals, to create the three identified sensitivity groups. Once a sample has been divided into the three groups by applying the proposed 30/40/30 split approach, it is then also possible to determine the specific cut-off scores between these groups. Importantly, the total score of the HSC scale appears to be normally distributed which suggests that sensitivity exists on a continuum. Hence, it may be most appropriate to consider sensitivity as a continuous dimension along which people can be categorized into three different groups. Further research should aim at investigating this continuous- versus-categorical nature of sensitivity and test whether and how the three detected groups differ qualitatively from each other. In addition, future work should validate whether group membership based on the proposed preliminary cut-off scores predicts behav- ioral differences in sensitivity to environmental influences.

General Discussion

A growing number of empirical studies provide evidence for the theoretical proposition that children differ in their Environmental Sensitivity, with some being more affected by the quality of their environment than others (Belsky & Pluess, 2009; Boyce & Ellis, 2005; Ellis et al., 2011; Pluess, 2015). The first objective of the current study was to investigate the psychometric properties of a new self-report measure of Environmental Sensitivity for children and adolescents—the Highly Sensitive Child (HSC) scale. The second aim was to test associations between the HSC scale and well-established temperament and personality traits. The third objective aimed at investigating whether children and adolescents can be categorized into distinct groups that differ in their Envi- ronmental Sensitivity.

Psychometric Properties and Construct Validity of the HSC Scale

Findings of the current study suggest that it is possible to assess Environmental Sensitivity with a 12-item questionnaire in children as young as 8 years. Consistent with a recent confirmatory factor

Figure 3. The distributions of the HSC mean score for each of three sensitivity groups in the child sample with cut-off scores for the low, medium, and high sensitivity groups.

Figure 4. The distributions of the HSC mean score for each of three sensitivity groups in the adolescent subsample A with cut-off scores for the low, medium, and high sensitivity groups.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

65ENVIRONMENTAL SENSITIVITY

analysis of the adult HSP scale (Lionetti et al., 2017), the HSC scale seems to fit a bifactor model which includes the three established factors but also a general sensitivity factor across all 12 items. Hence, although the scale captures different components of Environmental Sensitivity, it does also reflect a general trait of Environmental Sensitivity.

The observed associations with temperament and personality traits provide more insight into the three sensitivity components of the measure. Whereas Ease of Excitation and Low Sensory Threshold seem to be more strongly associated with traits that reflect sensitivity to negative environmental factors (e.g., BIS, Negative Emotionality, Negative Affect, and Neuroticism), Aes- thetic Sensitivity correlates with measures that may confer sensi- tivity to more positive experiences (e.g., BAS, Positive Emo- tionality, Extraversion, Openness, Conscientiousness). The co- occurrence of sensitivity to negative and positive environmental influences may also explain the finding that the total scale corre- lates with both BIS and BAS as well as both negative and positive emotionality. This interpretation fits well with the literature on the different theoretical models of Environmental Sensitivity (Pluess, 2015). Whereas Diathesis-Stress (Monroe & Simons, 1991; Zuck- erman, 1999) describes primarily individual differences in vulner- ability to adverse exposures, Vantage Sensitivity (Pluess, 2017; Pluess & Belsky, 2013) refers to interindividual variability in the propensity to benefit from positive experiences. Differential Sus- ceptibility (Belsky, 1997a, 2005; Belsky, Bakermans-Kranenburg, et al., 2007; Belsky & Pluess, 2009, 2013; Ellis et al., 2011), on the other hand, describes the combination of both Diathesis-Stress and Vantage Sensitivity with susceptible individuals being more af- fected by both negative as well as positive environmental influ- ences as a function of general sensitivity factors (e.g., genes, physiological reactivity, personality traits). Applied to the HSC measure, the total score of the scale may capture general sensitivity as described in the Differential Susceptibility model combining both Diathesis-Stress (i.e., sensitivity to adversity as measured with Ease of Excitation and Low Sensory Threshold subscales) and Vantage Sensitivity (i.e., sensitivity to positive experiences as reflected in the Aesthetic Sensitivity subscale). A recent twin study provides additional support for categorizing the three HSC com- ponents into sensitivity to negative (Ease of Excitation/Low Sen- sory Threshold) and positive (Aesthetic Sensitivity) environmental influences based on the finding that Ease of Excitation and Low Sensory Thresholds are genetically more similar to each other than to Aesthetic Sensitivity (Assary, Zavos, Krapohl, Keers, & Pluess, 2017). However, although this interpretation may seem reasonable in light of the discussed theoretical models and observed empirical findings, it has to be acknowledged that the adult HSP scale was originally developed to capture a unidimensional construct of Sensory Processing Sensitivity rather that different sensitivity components (Aron & Aron, 1997). The three factors—Ease of Excitation, Low Sensory Threshold, and Aesthetic Sensitivity—emerged in subse- quent studies conducted by other research groups (Booth et al., 2015; Liss, Mailloux, & Erchull, 2008; Smolewska et al., 2006; Sobocko & Zelenski, 2015) and do not represent a priori designed subscales. Hence, it is important to be cautious when interpreting the meaning of the typically observed three-factor structure (and in particular when trying to use the subscales separately).

Existence of Sensitivity Groups

Theoretical reasoning (Aron & Aron, 1997; Aron et al., 2012; Belsky, 1997b) and accompanying empirical research (Pluess, 2017; Pluess & Belsky, 2015; Wolf et al., 2008) suggest that people fall into different sensitivity groups with about 10–35% of the population considered to be highly sensitive. Given that the majority of existing research on Environmental Sensitivity appears to focus on this more sensitive group it is not surprising that much less is known about the less sensitive 65–90%. Hence, the finding of the current study that there appear to be three rather than two distinct categories of Environmental Sensitivity is of great impor- tance. Although we found that a highly sensitive group made up 20–35% of the sample, our analyses suggested that the less sen- sitive individuals can be categorized into two distinct groups rather than one: a medium sensitive group representing approx. 41–47% of the population and a low sensitive group (approx. 25–35%). Importantly, this three group solution emerged consistently across multiple and independent samples in childhood, adolescence, as well as adulthood (Lionetti et al., 2017). These findings provide empirical evidence for the proposition that most people are sensi- tive to their environment but to different degrees (Pluess, 2015). While we know a fair bit about the highly sensitive group (Aron & Aron, 1997; Aron et al., 2012) our understanding of the medium and particularly the low sensitive group is very limited. It is conceivable that the medium sensitive group is simply somewhat less sensitive than the highly sensitive group. The low sensitive group, on the other hand, may capture those that are particularly resilient to adverse conditions but also less able to benefit from positive exposures (i.e., showing Vantage Resistance; Pluess & Belsky, 2013).The existence of three groups is certainly reconcil- able with classic findings in research on infant temperament. For example, Kagan (1997) found that about 20% of 4-month-old infants were highly reactive (i.e., behavioral inhibition) to envi- ronmental stimulation whereas about 40% showed low reactivity (i.e., behavioral disinhibition) with the remaining 40% not clearly fitting either group. Our LCA findings suggest that the undefined 40% may represent the medium sensitive individuals (i.e., Tulips), which are distinct from the 20% highly sensitive (i.e., high reactive or Orchids) and the 40% low sensitive children (i.e., low reactive or Dandelions). Future research should replicate the three group structure and investigate characteristics associated with these three sensitivity groups in more detail (e.g., developmental history, personality and temperament differences, genetic differences etc.). A further point to be investigated is whether the proportions of the three groups change over time. The current findings suggest that in middle childhood more children fall into the high sensitive and less children into the low sensitive group compared with adolescence. This may indicate that younger children are more sensitive to their environment than adolescents or adults (Lionetti et al., 2017). However, longitudinal research is needed to investigate the devel- opment and stability of sensitivity over the life course to reject the alternative hypothesis that these differences are simply attributable to the scale performing differently at the different ages.

Cut-Off Scores for Sensitivity Groups

According to the current study, the detected cut-off scores should be used with caution when trying to categorize individual children and adolescents into the three detected sensitivity groups.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

66 PLUESS ET AL.

Although the cut-off scores were comparable between children and adolescents, they seemed to work slightly better for adolescents than for children and were slightly better at differentiating the highly sensitive individuals from medium sensitive ones than distinguishing the low sensitive from the medium sensitive chil- dren. One reason for this difference may be that the scale was developed to measure the high end rather than the low end of the sensitivity spectrum. Additional studies are required to test and confirm the validity and usefulness of the exploratory cut-off scores. The proposed general cut-off scores (i.e., 3.8 and 4.7), based on all available results from child, adolescent and adult samples, should only be used as rough indicators of sensitivity group membership, and only in addition to considering the con- tinuous mean score. As an alternative approach we propose to divide a sample into top and bottom 30% (i.e., high and low sensitivity, respectively) with the remaining 40% making up the medium sensitivity group.

Empirical Evidence for the Moderating Effects of the HSC Scale

Although the HSC scale appears to be a promising and psycho- metrically sound phenotypic marker of Environmental Sensitivity, it remains to be determined whether it does indeed predict indi- vidual differences in response to environmental influences as theory suggests. Recently, several findings emerged providing first empirical evidence for the validity of the HSC scale as a measure of Environmental Sensitivity. For example, HSC was found to predict treatment response related to a universal school-based resilience-promoting intervention (Pluess, Boniwell, Hefferon, & Tunariu, 2017) in a sample of 166 eleven-year-old girls in London, United Kingdom, with those scoring in the top 25% of HSC benefitting from the intervention regarding the reduction of de- pression symptoms whereas girls in the bottom 25% of HSC completely failed to do so (Pluess & Boniwell, 2015). Similarly, HSC moderated the effects of a school-based antibullying inter- vention in a large randomized controlled trial involving 2,042 children from 13 schools in Italy (Nocentini, Menesini, & Pluess, 2017). Although the intervention was effective across the whole sample, treatment effects were moderated by HSC and gender, with boys scoring high on HSC benefitting from the effects of the intervention regarding the reduction of self-reported victimization and internalizing symptoms. In contrast, boys scoring low on HSC did not respond to the intervention at all. Environmental Sensitiv- ity measured with the HSC scale has also been found to play a significant role among juvenile offenders in the United States as reported by Donley, Fine, Simmons, Pluess, and Cauffman (2016). The longitudinal study on reoffending behaviors featured a sample of 1,216 male adolescents aged 13–17 years who have been arrested for low-level crimes. The juvenile offenders completed the HSC scale and were interviewed repeatedly across 1.5 years on the quality of their home environment and reoffending behaviors. Adolescents living in adverse home environments were on average more likely to reoffend than those living in more supportive home environments, but HSC significantly moderated the effect of the home environment on the risk for reoffending. Consistent with a hypothesis of Environmental Sensitivity, more sensitive individu- als benefited more from positive home environments compared with the less sensitive adolescents. Focusing on natural variation in

parenting quality, Slagt, Dubas, van Aken, Ellis, and Deković (2016) investigated whether parent-rated HSC moderated the ef- fects of both negative and positive parent-reported parenting prac- tices on the development of teacher-rated externalizing and pro- social behavior in a longitudinal study involving 264 four- to seven-year-old Dutch children and their mothers. The 12-item HSC scale was adapted for the use as parent-rated measure of children’s sensitivity. Several significant interaction effects emerged. Most notably, HSC moderated effects of changes in negative and positive parenting between assessment points in the prediction of teacher reported externalizing behavior problems: Children rated high on the HSC scale had fewer problems if positive parenting increased and negative parenting decreased, but also more problems when positive parenting decreased and nega- tive parenting increased. Children with low scores on HSC, on the other hand, were not affected by changes in parenting quality.

The findings from these four studies not only validate the HSC scale as a measure of Environmental Sensitivity to both negative and positive environmental influences but also emphasize the importance of considering individual differences in Environmental Sensitivity in different fields, from developmental to clinical psy- chology (Pluess, 2015).

Strengths and Limitations

The five original studies reported in this paper are characterized by significant strengths, including large samples, replication of results, and the application of sophisticated statistical procedures, but the findings should be considered in light of methodological limitations. Most importantly, all data are based on self-report. Future research should aim at identifying more objective markers of child Environmental Sensitivity. Furthermore, all data were provided by children and adolescents residing in the United King- dom. Although some of the included samples were highly diverse, future studies should test whether similar findings emerge in other populations. Furthermore, although the HSC scale has been de- signed to reflect the same factor structure as the adult HSP scale, measurement invariance between child and adult samples has not been established yet.

Conclusion

Environmental Sensitivity is an important individual character- istic that is related to, but largely distinct from, other common temperament and personality traits. The current study suggests that it is possible to measure Environmental Sensitivity reliably in children and adolescents with the Highly Sensitive Child scale, a 12-item self-report measure with good psychometric properties. Furthermore, recent studies featuring samples from four different countries confirm the validity of the HSC scale by providing empirical evidence that HSC reflects individual differences in response to a wide range of environmental influences (Donley et al., 2016; Nocentini et al., 2017; Pluess & Boniwell, 2015; Slagt et al., 2016).

Future research should continue to investigate the hypothesized moderating function of Environmental Sensitivity regarding the effects of various environmental factors (e.g., parenting quality, education etc.) and psychological intervention. Of particular inter- est are differences between the three sensitivity groups as well as

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

67ENVIRONMENTAL SENSITIVITY

the development over the life course. To be able to do so, it will be necessary to develop measures of Environmental Sensitivity for younger children, including infants. Future work should also aim at identifying the specific psychological and biological mecha- nisms underlying individual differences in Environmental Sensi- tivity, including neuroimaging studies (for fMRI studies on the adult HSP scale, see Acevedo et al., 2014; Jagiellowicz et al., 2011) as well as quantitative behavioral genetics (Assary et al., 2017) and molecular genetics studies (C. Chen et al., 2011; Keers et al., 2016). Finally, it is important to investigate differences in Environmental Sensitivity across different cultures.

In conclusion, children and adolescents differ substantially in their sensitivity to environmental influences. Such differences in Environmental Sensitivity can be measured in children and ado- lescents with a short and simple yet psychometrically robust self- report measure—the Highly Sensitive Child (HSC) scale. Most children and adolescents appear to fall into one of three sensitivity groups: About 30% of children are characterized by high sensitiv- ity, 40% by medium sensitivity, and the remaining 30% by low sensitivity.

References

Acevedo, B. P., Aron, E. N., Aron, A., Sangster, M. D., Collins, N., & Brown, L. L. (2014). The highly sensitive brain: An fMRI study of sensory processing sensitivity and response to others’ emotions. Brain and Behavior, 4, 580–594. http://dx.doi.org/10.1002/brb3.242

Albert, D., Belsky, D. W., Crowley, D. M., Latendresse, S. J., Aliev, F., Riley, B., . . . the Conduct Problems Prevention Research Group. (2015). Can genetics predict response to complex behavioral interventions? Evidence from a genetic analysis of the Fast Track Randomized Control Trial. Journal of Policy Analysis and Management, 34, 497–518. http:// dx.doi.org/10.1002/pam.21811

Aron, E. N. (1996). The highly sensitive person: How to thrive when the world overwhelms you (rev. ed.). New York, NY: Broadway Books.

Aron, E. N., & Aron, A. (1997). Sensory-processing sensitivity and its relation to introversion and emotionality. Journal of Personality and Social Psychology, 73, 345–368. http://dx.doi.org/10.1037/0022-3514 .73.2.345

Aron, E. N., Aron, A., & Davies, K. M. (2005). Adult shyness: The interaction of temperamental sensitivity and an adverse childhood envi- ronment. Personality and Social Psychology Bulletin, 31, 181–197. http://dx.doi.org/10.1177/0146167204271419

Aron, E. N., Aron, A., & Jagiellowicz, J. (2012). Sensory processing sensitivity: A review in the light of the evolution of biological respon- sivity. Personality and Social Psychology Review, 16, 262–282. http:// dx.doi.org/10.1177/1088868311434213

Assary, E., Zavos, H. M., Krapohl, E., Keers, R., & Pluess, M. (2017). Sensitivity to environmental influences is a heritable trait. Manuscript submitted for publication.

Beaver, J. D., Lawrence, A. D., van Ditzhuijzen, J., Davis, M. H., Woods, A., & Calder, A. J. (2006). Individual differences in reward drive predict neural responses to images of food. The Journal of Neuroscience, 26, 5160–5166. http://dx.doi.org/10.1523/JNEUROSCI.0350-06.2006

Belsky, J. (1997a). Theory testing, effect-size evaluation, and differential susceptibility to rearing influence: The case of mothering and attach- ment. Child Development, 68, 598 – 600. http://dx.doi.org/10.2307/ 1132110

Belsky, J. (1997b). Variation in susceptibility to rearing influences: An evolutionary argument. Psychological Inquiry, 8, 182–186. http://dx.doi .org/10.1207/s15327965pli0803_3

Belsky, J. (2005). Differential susceptibility to rearing influences: An

evolutionary hypothesis and some evidence. In B. Ellis & D. Bjorklund (Eds.), Origins of the social mind: Evolutionary Psychology and Child Development (pp. 139–163). New York, NY: Guilford Press.

Belsky, J., Bakermans-Kranenburg, M. J., & van IJzendoorn, M. H. (2007). For better and for worse: Differential susceptibility to environmental influences. Current Directions in Psychological Science, 16, 300–304. http://dx.doi.org/10.1111/j.1467-8721.2007.00525.x

Belsky, J., Jonassaint, C., Pluess, M., Stanton, M., Brummett, B., & Williams, R. (2009). Vulnerability genes or plasticity genes? Molecular Psychiatry, 14, 746–754. http://dx.doi.org/10.1038/mp.2009.44

Belsky, J., & Pluess, M. (2009). Beyond diathesis stress: Differential susceptibility to environmental influences. Psychological Bulletin, 135, 885–908. http://dx.doi.org/10.1037/a0017376

Belsky, J., & Pluess, M. (2013). Beyond risk, resilience, and dysregulation: Phenotypic plasticity and human development. Development and Psy- chopathology, 25, 1243–1261. http://dx.doi.org/10.1017/S095457941 300059X

Belsky, J., Vandell, D. L., Burchinal, M., Clarke-Stewart, K. A., McCart- ney, K., & Owen, M. T., & the NICHD Early Child Care Research Network. (2007). Are there long-term effects of early child care? Child Development, 78, 681–701. http://dx.doi.org/10.1111/j.1467-8624.2007 .01021.x

Booth, C., Standage, H., & Fox, E. (2015). Sensory-processing sensitivity moderates the association between childhood experiences and adult life satisfaction. Personality and Individual Differences, 87, 24–29. http:// dx.doi.org/10.1016/j.paid.2015.07.020

Bornstein, M. H., & Bradley, R. H. (2014). Socioeconomic status, parent- ing, and child development. London, UK: Routledge.

Borries, F. (2012). Do the “highly sensitive” exist? A taxonometric inves- tigation of the personality construct sensory processing sensitivity (Un- published Doctoral thesis). University of Bielefeld, Germany.

Boterberg, S., & Warreyn, P. (2016). Making sense of it all: The impact of sensory processing sensitivity on daily functioning of children. Person- ality and Individual Differences, 92, 80–86. http://dx.doi.org/10.1016/j .paid.2015.12.022

Boyce, W. T., & Ellis, B. J. (2005). Biological sensitivity to context: I. An evolutionary-developmental theory of the origins and functions of stress reactivity. Development and Psychopathology, 17, 271–301. http://dx .doi.org/10.1017/S0954579405050145

Brown, H. M., Eley, T. C., Broeren, S., Macleod, C., Rinck, M., Hadwin, J. A., & Lester, K. J. (2014). Psychometric properties of reaction time based experimental paradigms measuring anxiety-related information- processing biases in children. Journal of Anxiety Disorders, 28, 97–107. http://dx.doi.org/10.1016/j.janxdis.2013.11.004

Capaldi, D. M., & Rothbart, M. K. (1992). Development and validation of an early adolescent temperament measure. The Journal of Early Ado- lescence, 12, 153–173. http://dx.doi.org/10.1177/027243169201200 2002

Carver, C. S., & White, T. L. (1994). Behavioral-Inhibition, Behavioral Activation, and Affective Responses to Impending Reward and Punish- ment—the Bis Bas Scales. Journal of Personality and Social Psychol- ogy, 67, 319–333. http://dx.doi.org/10.1037/0022-3514.67.2.319

Chen, C., Chen, C., Moyzis, R., Stern, H., He, Q., Li, H., . . . Dong, Q. (2011). Contributions of dopamine-related genes and environmental factors to highly sensitive personality: A multi-step neuronal system- level approach. PLoS ONE, 6, e21636. http://dx.doi.org/10.1371/journal .pone.0021636

Chen, F. F., West, S. G., & Sousa, K. H. (2006). A comparison of bifactor and second-order models of quality of life. Multivariate Behavioral Research, 41, 189 –225. http://dx.doi.org/10.1207/s15327906mbr41 02_5

Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Mod- eling, 9, 233–255. http://dx.doi.org/10.1207/S15328007SEM0902_5

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

68 PLUESS ET AL.

de Haan, A. D., Prinzie, P., & Deković, M. (2010). How and why children change in aggression and delinquency from childhood to adolescence: Moderation of overreactive parenting by child personality. Journal of Child Psychology and Psychiatry, 51, 725–733. http://dx.doi.org/10 .1111/j.1469-7610.2009.02192.x

Donley, S., Fine, A., Simmons, C., Pluess, M., & Cauffman, E. (2016). Environmental sensitivity among juvenile offenders: Do some juvenile offenders benefit from positive home environments more than others?

Ellis, B. J., & Boyce, W. T. (2008). Biological sensitivity to context. Current Directions in Psychological Science, 17, 183–187.

Ellis, B. J., & Boyce, W. T. (2011). Differential susceptibility to the environment: Toward an understanding of sensitivity to developmental experiences and context. Development and Psychopathology, 23, 1–5. http://dx.doi.org/10.1017/S095457941000060X

Ellis, B. J., Boyce, W. T., Belsky, J., Bakermans-Kranenburg, M. J., & van Ijzendoorn, M. H. (2011). Differential susceptibility to the environment: An evolutionary—Neurodevelopmental theory. Development and Psy- chopathology, 23, 7–28. http://dx.doi.org/10.1017/S0954579410000611

Evans, D. E., & Rothbart, M. K. (2007). Developing a model for adult temperament. Journal of Research in Personality, 41, 868–888. http:// dx.doi.org/10.1016/j.jrp.2006.11.002

Evans, D. E., & Rothbart, M. K. (2008). Temperamental sensitivity: Two constructs or one? Personality and Individual Differences, 44, 108–118. http://dx.doi.org/10.1016/j.paid.2007.07.016

Gable, S. L., Reis, H. T., & Elliot, A. J. (2000). Behavioral activation and inhibition in everyday life. Journal of Personality and Social Psychol- ogy, 78, 1135–1149. http://dx.doi.org/10.1037/0022-3514.78.6.1135

Gerstenberg, F. X. R. (2012). Sensory-processing sensitivity predicts per- formance on a visual search task followed by an increase in perceived stress. Personality and Individual Differences, 53, 496–500. http://dx .doi.org/10.1016/j.paid.2012.04.019

Gray, J. A. (1981). A critique of Eysenck’s theory of personality. In H. J. Eysenck (Ed.), A model for personality (pp. 246–276). Berlin, Germany: Springer-Verlag. http://dx.doi.org/10.1007/978-3-642-67783-0_8

Gray, J. A. (1982). The neuropsychology of anxiety: An enquiry into the functions of the septo-hippocampal system. New York, NY: Oxford University Press.

Haworth, C. M., Davis, O. S., & Plomin, R. (2013). Twins Early Devel- opment Study (TEDS): A genetically sensitive investigation of cognitive and behavioral development from childhood to young adulthood. Twin Research and Human Genetics, 16, 117–125. http://dx.doi.org/10.1017/ thg.2012.91

Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43, 115–135. http://dx.doi.org/10.1007/s11747-014-0403-8

Heponiemi, T., Keltikangas-Jarvinen, L., Puttonen, S., & Ravaja, N. (2003). BIS/BAS sensitivity and self-rated affects during experimentally induced stress. Personality and Individual Differences, 34, 943–957. http://dx.doi.org/10.1016/S0191-8869(02)00079-X

Ho, M. Y., Cheung, F. M., You, J., Kam, C., Zhang, X., & Kliewer, W. (2013). The moderating role of emotional stability in the relationship between exposure to violence and anxiety and depression. Personality and Individual Differences, 55, 634–639. http://dx.doi.org/10.1016/j .paid.2013.05.004

IBMCorp. (2011). IBM SPSS Statistics for Windows, version 20.0. Ar- monk, NY: IBM Corp.

Jagiellowicz, J., Xu, X., Aron, A., Aron, E., Cao, G., Feng, T., & Weng, X. (2011). The trait of sensory processing sensitivity and neural responses to changes in visual scenes. Social Cognitive and Affective Neurosci- ence, 6, 38–47. http://dx.doi.org/10.1093/scan/nsq001

Kagan, J. (1989). Temperamental contributions to social behavior. Amer- ican Psychologist, 44, 668–674. http://dx.doi.org/10.1037/0003-066X .44.4.668

Kagan, J. (1997). Temperament and the reactions to unfamiliarity. Child Development, 68, 139–143. http://dx.doi.org/10.2307/1131931

Kagan, J., Reznick, J. S., & Snidman, N. (1987). The physiology and psychology of behavioral inhibition in children. Child Development, 58, 1459–1473. http://dx.doi.org/10.2307/1130685

Keers, R., Coleman, J. R., Lester, K. J., Roberts, S., Breen, G., Thastum, M., . . . Eley, T. C. (2016). A genome-wide test of the differential susceptibility hypothesis reveals a genetic predictor of differential re- sponse to psychological treatments for child anxiety disorders. Psycho- therapy and Psychosomatics, 85, 146–158. http://dx.doi.org/10.1159/ 000444023

Kim, S., & Kochanska, G. (2012). Child temperament moderates effects of parent-child mutuality on self-regulation: A relationship-based path for emotionally negative infants. Child Development, 83, 1275–1289. http:// dx.doi.org/10.1111/j.1467-8624.2012.01778.x

Kovas, Y., Haworth, C. M., Dale, P. S., Plomin, R., Weinberg, R. A., Thomson, J. M., & Fischer, K. W. (2007). The genetic and environmen- tal origins of learning abilities and disabilities in the early school years (Monographs of the Society for Research in Child Development). Lon- don, UK: Blackwell.

Laurent, J., Catanzaro, S. J., Joiner, T. E., Jr., Rudolph, K. D., Potter, K. I., Lambert, S., . . . Gathright, T. (1999). A measure of positive and negative affect for children: Scale development and preliminary valida- tion. Psychological Assessment, 11, 326–338. http://dx.doi.org/10.1037/ 1040-3590.11.3.326

Lionetti, F., Aron, A., Aron, E., Burns, G. L., Jagiellowicz, J., & Pluess, M. (2017). Dandelions, tulips, and orchids: Evidence for the existence of low, medium, and high sensitive individuals in the general population. Manuscript submitted for publication.

Liss, M., Mailloux, J., & Erchull, M. J. (2008). The relationships between sensory processing sensitivity, alexithymia, autism, depression, and anx- iety. Personality and Individual Differences, 45, 255–259. http://dx.doi .org/10.1016/j.paid.2008.04.009

Masyn, K. E. (2013). Latent class analysis and finite mixture modeling. In T. D. Little (Ed.), The Oxford handbook of quantitative methods in psychology (Vol. 2, pp. 551–611). London, UK: Oxford University Press.

McCrae, R. R., Kurtz, J. E., Yamagata, S., & Terracciano, A. (2011). Internal consistency, retest reliability, and their implications for person- ality scale validity. Personality and Social Psychology Review, 15, 28–50. http://dx.doi.org/10.1177/1088868310366253

McNaughton, N., & Gray, J. A. (2000). Anxiolytic action on the behav- ioural inhibition system implies multiple types of arousal contribute to anxiety. Journal of Affective Disorders, 61, 161–176. http://dx.doi.org/ 10.1016/S0165-0327(00)00344-X

Monroe, S. M., & Simons, A. D. (1991). Diathesis-stress theories in the context of life stress research: Implications for the depressive disorders. Psychological Bulletin, 110, 406–425. http://dx.doi.org/10.1037/0033- 2909.110.3.406

Mullins-Sweatt, S. N., Jamerson, J. E., Samuel, D. B., Olson, D. R., & Widiger, T. A. (2006). Psychometric properties of an abbreviated in- strument of the five-factor model. Assessment, 13, 119–137. http://dx .doi.org/10.1177/1073191106286748

Muthén, L. K., & Muthén, B. O. (1998–2015). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

Nocentini, A., Menesini, E., & Pluess, M. (2017). Environmental sensitiv- ity predicts treatment response to anti-bullying intervention: Evidence of vantage sensitivity. Manuscript submitted for publication.

Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535–569. http://dx.doi.org/10.1080/10705510701575396

Obradović, J., & Boyce, W. T. (2009). Individual differences in behavioral, physiological, and genetic sensitivities to contexts: Implications for

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

69ENVIRONMENTAL SENSITIVITY

development and adaptation. Developmental Neuroscience, 31, 300– 308. http://dx.doi.org/10.1159/000216541

Obradović, J., Bush, N. R., Stamperdahl, J., Adler, N. E., & Boyce, W. T. (2010). Biological sensitivity to context: The interactive effects of stress reactivity and family adversity on socioemotional behavior and school readiness. Child Development, 81, 270–289. http://dx.doi.org/10.1111/j .1467-8624.2009.01394.x

Pastore, M. (2016). Overlapping: Estimation of overlapping in empirical distribution. Retrieved from https://cran.r-project.org/web/packages/ overlapping/index.html

Pluess, M. (2015). Individual differences in environmental sensitivity. Child Development Perspectives, 9, 138–143. http://dx.doi.org/10.1111/ cdep.12120

Pluess, M. (2017). Vantage sensitivity: Environmental sensitivity to posi- tive experiences as a function of genetic differences. Journal of Person- ality, 85, 38–50. http://dx.doi.org/10.1111/jopy.12218

Pluess, M., & Belsky, J. (2010). Differential susceptibility to parenting and quality child care. Developmental Psychology, 46, 379–390. http://dx .doi.org/10.1037/a0015203

Pluess, M., & Belsky, J. (2013). Vantage sensitivity: Individual differences in response to positive experiences. Psychological Bulletin, 139, 901– 916. http://dx.doi.org/10.1037/a0030196

Pluess, M., & Belsky, J. (2015). Vantage sensitivity: Genetic susceptibility to effects of positive experiences. In M. Pluess (Ed.), Genetics of psychological well-being (pp. 193–210). Oxford, UK: Oxford University Press. http://dx.doi.org/10.1093/acprof:oso/9780199686674.003.0012

Pluess, M., & Boniwell, I. (2015). Sensory-processing sensitivity predicts treatment response to a school-based depression prevention program: Evidence of vantage sensitivity. Personality and Individual Differences, 82, 40–45. http://dx.doi.org/10.1016/j.paid.2015.03.011

Pluess, M., Boniwell, I., Hefferon, K., & Tunariu, A. (2017). Preliminary evaluation of a school-based resilience-promoting intervention in a high-risk population: Application of an exploratory two-cohort treat- ment/control design. PLoS ONE, 12(5), e0177191. http://dx.doi.org/10 .1371/journal.pone.0177191

Pluess, M., Velders, F. P., Belsky, J., van IJzendoorn, M. H., Bakermans- Kranenburg, M. J., Jaddoe, V. W., . . . Tiemeier, H. (2011). Serotonin transporter polymorphism moderates effects of prenatal maternal anxiety on infant negative emotionality. Biological Psychiatry, 69, 520–525. http://dx.doi.org/10.1016/j.biopsych.2010.10.006

Putnam, S. P., Ellis, L. K., & Rothbart, M. K. (2001). The structure of temperament from infancy through adolescence. In A. Eliasz & A. Angleitner (Eds.), Advances in research on temperament (pp. 165–182). Lengerich, Germany: Pabst Science.

Roisman, G. I., Newman, D. A., Fraley, R. C., Haltigan, J. D., Groh, A. M., & Haydon, K. C. (2012). Distinguishing differential susceptibility from diathesis-stress: Recommendations for evaluating interaction effects. Development and Psychopathology, 24, 389–409. http://dx.doi.org/10 .1017/S0954579412000065

Rosseels, Y. (2016). Package lavaan. Retrieved from http://cran.r-project .org/web/packages/lavaan/lavaan.pdf

Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample analysis of moment structures. In R. D. H. Heijmans, D. S. G. Pollock, & A. Satorrs (Eds.), Innovations in multivariate statistical analysis (pp. 233–247). New York, NY: Springer. http://dx.doi.org/10.1007/978-1- 4615-4603-0_17

Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of structural equation models: Tests of significance and descrip-

tive goodness-of-fit measures. Methods of Psychological Research, 8, 23–74.

semTools Contributors. (2016). semTools: Useful tools for structural equa- tion modeling. Retrieved from http://cran.r-project.org/package� semTools

Slagt, M., Dubas, J. S., Deković, M., & van Aken, M. A. (2016). Differ- ences in sensitivity to parenting depending on child temperament: A meta-analysis. Psychological Bulletin, 142, 1068–1110. http://dx.doi .org/10.1037/bul0000061

Slagt, M., Dubas, J. S., Denissen, J. J., Deković, M., & van Aken, M. A. (2015). Personality traits as potential susceptibility markers: Differential susceptibility to support among parents. Journal of Personality, 83, 155–166. http://dx.doi.org/10.1111/jopy.12091

Slagt, M., Dubas, J. S., van Aken, M. A. G., Ellis, B., & Deković, M. (2016). Sensory processing sensitivity as a marker of differential sus- ceptibility to parenting. Manuscript submitted for publication.

Smolewska, K. A., McCabe, S. B., & Woody, E. Z. (2006). A psychomet- ric evaluation of the Highly Sensitive Person Scale: The components of sensory-processing sensitivity and their relation to the BIS/BAS and “Big Five.” Personality and Individual Differences, 40, 1269–1279. http://dx.doi.org/10.1016/j.paid.2005.09.022

Snyder, H. R., Gulley, L. D., Bijttebier, P., Hartman, C. A., Oldehinkel, A. J., Mezulis, A., . . . Hankin, B. L. (2015). Adolescent emotionality and effortful control: Core latent constructs and links to psychopathol- ogy and functioning. Journal of Personality and Social Psychology, 109, 1132–1149. http://dx.doi.org/10.1037/pspp0000047

Sobocko, K., & Zelenski, J. M. (2015). Trait sensory-processing sensitivity and subjective well-being: Distinctive associations for different aspects of sensitivity. Personality and Individual Differences, 83, 44–49. http:// dx.doi.org/10.1016/j.paid.2015.03.045

van IJzendoorn, M. H., Belsky, J., & Bakermans-Kranenburg, M. J. (2012). Serotonin transporter genotype 5HTTLPR as a marker of differential susceptibility? A meta-analysis of child and adolescent gene-by- environment studies. Translational Psychiatry, 2, e147. http://dx.doi .org/10.1038/tp.2012.73

Van Os, J., & Jones, P. B. (1999). Early risk factors and adult person— Environment relationships in affective disorder. Psychological Medi- cine, 29, 1055–1067. http://dx.doi.org/10.1017/S0033291799001026

Walda, S. A. E. (2007). Hoogsensitiviteit bij kinderen in het basisonderwijs [High sensitivity in children from regular education]. Unpublished Mas- ter’s thesis. Radboud University, Nijmegen, the Netherlands. Retrieved from http://www.annabosman.eu/documents/SietskeWalda2007.pdf

Wolf, M., van Doorn, G. S., & Weissing, F. J. (2008). Evolutionary emergence of responsive and unresponsive personalities. PNAS Pro- ceedings of the National Academy of Sciences of the United States of America, 105, 15825–15830. http://dx.doi.org/10.1073/pnas.080547 3105

Woodward, S. A., Lenzenweger, M. F., Kagan, J., Snidman, N., & Arcus, D. (2000). Taxonic structure of infant reactivity: Evidence from a taxometric perspective. Psychological Science, 11, 296–301. http://dx .doi.org/10.1111/1467-9280.00259

Zuckerman, M. (1999). Vulnerability to psychopathology: A biosocial model. Washington, DC: American Psychological Association. http://dx .doi.org/10.1037/10316-000

Received August 30, 2016 Revision received June 14, 2017

Accepted June 26, 2017 �

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

70 PLUESS ET AL.