ti

en

basilityup

nd npoor

adverse outcomes on child wellbeing in center child care. The regulation of noise levels in child carecenters is needed to provide optimal child well-being.

can benge frop tom, whe, & V

(Groeneveld, Vermeer, Linting & Van IJzendoorn, 2010). In thisstudy we examine noise levels in center child care and their asso-ciations with quality of care and children's emotional well-being.

Noise is a central component of the environmental chaos theorythat originates from family research and states that environmental

ds, music, and ac-l sources of noise

g is reected ine, and the lack ofctivities (Matheny,haos and crowdingunfavorable out-

comes, for instance with less parental talk, more negative parent-child interactions, more child social withdrawal and aggression,more child helplessness, less optimal child cognitive development,and more stress for adults and children (Evans, 2006; Evans &Wachs, 2010).

Investigations relevant to child development focused on theeffects of noise on children in the home environment (e.g. Babisch,Schulz, Seiwert, & Conrad, 2012; Evans, 2006), teachers and ado-lescents in schools (e.g. Enmarker & Boman, 2004), and children in

* Corresponding author.

Contents lists availab

Journal of Environm

w.e

Journal of Environmental Psychology 42 (2015) 190e201E-mail address: vermeer@fsw.leidenuniv.nl (H.J. Vermeer).countries implying that a large proportion of children may be dailyexposed to high noise levels. In this study, we investigate howchildren's emotional well-being is associated with noise levels incenter-based child care. A recent study by Linting, Groeneveld,Vermeer and Van IJzendoorn (2013) showed that in home-basedchild care there is a threshold for noise beyond which lower childwell-being can be observed. At regular child care centers, noiselevels are generally higher than in home-based child care settings

transport) or social (e.g. chatter, classroom sountivities). In this study, both social and non-sociawill be considered.

Environmental chaos in the home settinhouseholds with high levels of crowding, noisfamily routines, order, and regular planning of aWachs, Ludwig,& Phillips, 1995). High levels of cin the family setting have been associated withCenter-based child care for children under ve is an importantrearing and care environment for millions of children in Western

of child well-being. In addition, the sources of noise can be cate-gorized as non-social (e.g. noise from trafc, roads, aircraft, andChild social-emotional well-beingQuality of careEnvironmental chaosNonlinear analysis

1. Introduction

Noise levels in child care centersthese levels have been reported to ralevels of adult normal conversation, ucomparable to high way noise andhearing damage (Manlove, Frankhttp://dx.doi.org/10.1016/j.jenvp.2015.05.0030272-4944/ 2015 Elsevier Ltd. All rights reserved. 2015 Elsevier Ltd. All rights reserved.

overwhelming. In fact,m 45 dB, comparable toore than 90 dB, which isn sustained, can causeernon-Feagans, 2001).

chaos is detrimental for child development (Evans &Wachs, 2010).Noise can be dened according to intensity (i.e. low or high averagenoise levels), variability (i.e. the differences in peaks and lows), andduration: occasional versus chronic noise (Enmarker & Boman,2004; Kjellberg, Landstrom, Tesarz, Soderberg, & kerlund, 1996).Linting et al. (2013) showed that not only noise intensity but alsonoise variability beyond certain levels is a predictor for lower levelsCenter child careNoiseKeywords:regression analysis showed that more hours in care, higher child age and higher general child carequality were related to higher levels of well-being. Noise, a major aspect of environmental chaos, hasNoise in center-based child care: Associachild emotional wellbeing

C.D. Werner, M. Linting, H.J. Vermeer*, M.H. Van IJzCentre for Child and Family Studies, Leiden University, The Netherlands

a r t i c l e i n f o

Article history:Received 8 April 2014Received in revised form26 April 2015Accepted 10 May 2015Available online 11 May 2015

a b s t r a c t

Children attending center-noise levels, noise variab(N 64) involving childrenmaximum levels of noise aconrmed the threshold hyand below 65 dbA, and f

journal homepage: wwons with quality of care and

doorn

ed child care are daily exposed to high noise levels. Associations between, caregiving quality and child well-being were investigated in centersto four years (N 245; M 34.50 months). We examined minimum andoise variability for optimal child well-being. Nonlinear regression analysisthesis: optimal child well-being was observed for noise levels over 60 dbAnoise variability over 6.69 dbA and below 7.44 dbA. Linear multilevel

le at ScienceDirect

ental Psychology

lsevier .com/locate/ jep

available and willing to participate during the study period ofapproximately six months. We distinguished between qualied

nmeprimary schools (for reviews see Evans, 2006; Shield & Dockrell,2003). Fewer studies addressed the effects of noise in child caresettings, targeting toddlers in child care (Corapci, 2010; Groeneveldet al., 2010; Hambrick-Dixon, 1986; Linting et al., 2013; Manloveet al., 2001; McAllister, Granqvist, Sjolander, & Sundberg, 2009) orchild care professionals (Lindstrom, Persson Waye, Sodersten,McAllister, & Ternstrom, 2011; Sala et al., 2002). This is an impor-tant area of study because noise levels might be related to quality ofcare and are hypothesized to have a negative impact on childdevelopment if noise is going beyond a tolerable level.

The effects of noise encompass three major domains: health,cognition, and psychological well-being. In the health domainrather consistent adverse outcomes have been shown for noiselevels and physiological stress indicated by higher blood pressureand higher levels of stress hormones in children and adults (for areview see Evans, 2006). Studies in the domain of cognition showedadverse outcomes of noise on children's memory and attention(Evans, 2006; Klatte, Bergstrom, & Lachmann, 2013; Shield &Dockrell, 2003), psychomotor performance, reading skills, andcentral information processing (Evans, 2006; Klatte et al., 2013).Studies considering the third domain, psychological well-being,showed negative associations between noise and parentalcommunication and responsiveness (Evans, 2006), emotional well-being (Stansfeld et al., 2009), and positive associations withannoyance (Babisch et al., 2012; Enmarker & Boman, 2004; Haineset al., 2001; Maxwell, 2010), lack of motivation, helplessness, lack ofpatience, and aggression (Evans, 2006). It should be stressed thatndings were mixed and that child participants in the studiesvaried in age. More importantly, most studies regarded non-socialnoise, which may have different effects on children than socialnoise. It can be argued that social noise may have lower averagelevels and lower maximum peaks than noise from aircrafts orsubways passing. The former may be more controllable. Althoughunpredictable and uncontrollable noise are generally perceived asmore annoying, Enmarker and Boman (2004) found that socialnoise from chatter in classrooms was perceived as more annoyingthan non-social noise.

Some researchers showed that negative associations betweendevelopmental outcomes and noise were stronger for older chil-dren, because older children may be more aware of the noise andtherefore experience it as more intrusive or distracting thanyounger children (see Evans, 2006; Eysel-Gosepath, Daut, Pinger,Lehmacher, & Erren, 2012). Younger children, on the other hand,may be more vulnerable to negative environmental inuences andmay therefore more negatively affected by higher noise levels(Evans, 2006; Eysel-Gosepath et al., 2012). Apart from age, indi-vidual differences in sensitivity to noise may explain why somechildren may be more affected than others (Enmarker & Boman,2004; Maxwell & Evans, 2000).

Corapci (2010) already pointed out the importance of studies tochild care noise and chaos for child cognitive development. How-ever, to our knowledge only three studies have been conducted tonoise in relation to child emotional well-being in the child caresetting, two with a correlational design (Groeneveld et al., 2010;Linting et al., 2013) and one case study (Kishimoto, 2012).

1.1. Aim of the study

The aim of this study is to examine associations between noiseand child emotional well-being in center child care. In our studyaverage noise levels and noise variability are main predictors forthe outcome variable child emotional well-being. Group size andchild-caregiver ratio are taken into account as additional indicatorsof environmental chaos, in particular of crowding. The associations

C.D. Werner et al. / Journal of Envirobetween noise and quality of care, caregiver working experience incaregivers and caregivers-in-training. Qualied caregivers areformally employed by the organizations and have formal re-sponsibility for the children in the group, whereas caregivers-in-training are not formally employed, but working at the child carecenters under supervision as part of their vocational training. Fromeach participating group, one qualied caregiver who met the in-clusion criteria was randomly selected.

Selection of target children per group was based on parentalconsent and their attendance on the day scheduled for assessment.If more than four children with consent were present at theassessment day, selection was done randomly. There was a mini-child care, caregiverworking hours per week, child gender, age, andhours in care are explored.

We specically investigate whether a threshold of noise can befound beyond which lower child well-being can be observed incenter-based child care. We expect that the threshold hypothesis ofLinting et al. (2013) that was found to be conrmed in home-basedchild care also applies to center child care, implying that beyondcertain noise levels noise and noise variability are associated withlower levels of child emotional well-being; over and above asso-ciations with child, caregiver, and child care characteristics.

2. Method

2.1. Recruitment

This study is part of a larger investigation into the effectivenessof a video-feedback intervention aimed at professional caregiversin center-based child care. Here, we only present the pretest data.Participants in this study were children attending center-basedchild care. For recruitment we targeted child care centers in anurban area in the western part of the Netherlands. Letters of invi-tation were sent to180 child care organizations. Initially, 91 centersfrom 35 organizations agreed to participate. From the 91 centersthat initially agreed to participate by letter or by phone, 66 actuallystarted with the rst assessment. The other 25 centers showednon-response (27%), rather than dropping out after assessment.The non-response centers did not differ signicantly from partici-pating centers in terms of demographic region, neighborhood orcenter size. The most common reason for centers not to take part inour studywas the lack of parental consent for at least three childrenin the group who attended the child care center on the same day(n 13 centers). Other reasons for not participating were a lack ofinterest from the managers (n 3) or unwillingness of professionalcaregivers to participate (n 6). From the 66 centers that agreed totake part, two centers withdrew from the study directly after therst assessment; the managers did not allow us to use the obser-vational data in our study. Given the variety of participating orga-nizations and spread in demographic region our sample reectedthe composition of Dutch child care centers in the western part ofthe Netherlands.

To avoid selection bias, one group per center was randomlyselected for participation. Toddler groups (for children 2e4 years)and mixed-age groups (for children 0e4 years) could be included.Furthermore, groups within centers were eligible for participationif parents of at least three children attending the group on the sameday of the week provided written consent. Three centers thatagreed to participate only had eligible groups of young infants(0e18 months), so they were excluded from the study.

Professional caregivers from eligible groups had to be workingfor a minimum of two days (16 h) per week in a xed group and be

ntal Psychology 42 (2015) 190e201 191mum of three and a maximum of four target children per group.

2.2. Participants

In total, 245 children and 64 caregivers from 64 centers wereincluded for analysis. Approximately 49% of the children were boys(n 121). The children had a mean age of 34.50 months (SD 7.78)at the time of the assessment and they attended toddler groups(67%) and mixed age groups (33%). For more descriptive statistics,see Table 1. The majority of caregivers (n 43) had a degree invocational training. A minority (n 8) had nished a highereducational degree on a bachelor's or master's level. Few caregivers(n 3) had low educational levels, i.e. only primary school or highschool degree. Educational level was not reported by 10 caregivers(16%). The majority of caregivers (65%) were born in the

pleasure, self-condence, and relaxation. Scores were based onthree video fragments of ve minutes each of the target child atchild care. Every two and a half minutes a score was registered,resulting in six intervals.

Well-being scores are presented on a seven-point scale, rangingfrom 1 very lowwell-being (signals of discomfort are clearly present,e.g. crying, screaming) to 7 very high well-being (signals of comfortare clearly present, e.g. enjoyment, smiling). Scores were aggre-gated across the time periods. Bivariate correlations between thethree sessions for observed child well-being ranged from .28 to .33andwere all signicant on the p< .01 level. Themean change scoresbetween the three sessions ranged from .56 to .59. Eight observerswere trained to reliably assess child well-being. All observers met

C.D. Werner et al. / Journal of Environmental Psychology 42 (2015) 190e201192Netherlands. The other caregivers were born in Surinam (n 4), theDutch Antilles (n 2), Turkey (n 1), Cape Verde (n 1); or thiswas not reported (n 14).

2.3. Procedure

The child care centers were visited for assessment once from9:00 AM until 1:00 PM. Afterward, questionnaires were sent to theprofessional caregivers and the center managers. During the visit, ashortened version of the Early Childhood Environment RatingScale-Revised (ECERS-R; Harms, Clifford, & Cryer, 1998; Vermeer,Werner, Linting, Van IJzendoorn, & Groeneveld, 2012) was admin-istered. The professional caregiver and four target children wererecorded with a digital video camera at three predetermined timepoints (at 9.30 AM, 10.30 AM and 11.30 AM) during regular activ-ities on this same morning. In three one-hour sessions, each of thethree or four target children was recorded with a camera for veminutes and the caregiver was recorded for ten minutes. The targetchildren and the caregiver were recorded successively, so thatfragments did not overlap. During these recordings noise levelswere registered for 30 min. All recorded fragments were rated af-terward on caregiver sensitivity and child emotional well-being.

To obtain independence in ratings, researchers who adminis-tered the ECERS-R did not code video material from that particularsetting. In addition, caregiver sensitivity and child emotional well-being were coded by different researchers. Caregivers reported thatthe morning of observation was representative in terms of activ-ities, number of children, and child behavior in the group.

2.4. Measures

2.4.1. Child emotional well-beingChildren's emotional well-being was measured with the Well-

being Scale, developed and validated by the Dutch Consortiumfor Child care Research (NCKO; De Kruif et al., 2007). This scalecontains several indicators of the child's well-being, such as

Table 1Descriptive statistics (prior to winsorizing and imputation of missing data).

N

Child level Child age (months) 227Hours in care per week 180Child well-being 242

Group level Caregiver age (yrs) 53Caregiver work hours per week 54Caregivers years of experience 55Quality of care (ECERS-R) 64Caregiver sensitivity 64Noise level average (dbA) 61Noise variability 61Group size 63

Child-caregiver ratio 63the criteria of reliability: mean intraclass correlation (ICC; two-waymixed, absolute agreement) was .79 (ranging from .71 to .80). In-ternal consistency (Cronbach's alpha) across the six intervals was.70.

2.4.2. Noise levels and variabilityData Logger Sound Level Meters (type CEN) were used to mea-

sure noise levels in decibels at the child care centers. We used dB(A) weighting which simulates the response of the human ear.Noise recording was conducted in parallel to the three pre-determined videotaped sessions of 30 min; the sound level meterwas placed in the room or playground where the caregiver andchildren were at the time of the observations. If the caregiver andchildren moved to another area during the observations (e.g. frominside to outside playground), the observer placed the sound levelmeter in the new area. Noise levels were automatically recordedevery second, and recordings were averaged across the threeobservation sessions to obtain a nal score. Noise variability wascomputed as the average standard deviation of noise levels acrossthe three observation sessions (Linting et al., 2013).

For noise levels, bivariate correlations between the three ses-sions ranged between .26 and .40 and were signicant on thep < .05 level. For noise variability, bivariate correlations betweenthe three sessions ranged from .23 to .40 and were signicant onthe p < .05 level, except for the non-signicant correlation of .23between the rst and the third session. Because the videotapedsessions ran parallel to the video-fragments, we were able todistinguish between different types of sources of noise.

2.4.3. Sources of noiseSources of noise were coded for all child care groups, using three

10-min videotaped observations of the caregiver. By using thecaregiver observations rather than child observations, sources ofnoise at the group level may be more accurately identied becausethe scope of the observations included the whole group setting,rather than just one target child. The observation scheme

M SD Min Max

34.50 7.78 9.00 47.0027.91 9.55 8.00 55.004.50 .48 3.00 6.33

32.09 8.63 22.00 52.0029.42 5.74 16.00 40.007.70 6.95 .60 35.004.00 .78 1.89 5.394.84 .91 2.33 6.33

61.69 3.30 54.95 70.837.39 .81 6.10 9.68

12.14 3.11 4.00 18.00

5.41 1.76 2.00 13.00

nmedeveloped by Linting et al. (2013) for home-based child care formedthe basis for our coding with some minor adaptations for centerchild care. For instance, it was highly unlikely that pets weresources of noise in center child care, because of regulations.

Five main categories for sources of noise were distinguished:outdoor noise, background noise, children's noise, adults' noise,and noise from the handling of toys. The coding form and in-structions for coding are presented in Appendix A. Moreover, takinginto account centers with multiple groups, we further denedbackground noise as sounds that originated from inside the targetgroup (e.g. background media for music from a CD player, andtelephone ringing) versus sounds from outside the target group(background:human conversation and background:other), forinstance when caregivers and parents of children from othergroups were talking in the corridor, or children of other groupswere playing in an area nearby the target group.

We further dened sounds originating from children and adultsas positive vocalization (talking, laughing, singing), negative vocali-zations (for children: crying, screaming, ghting; for caregivers:scolding, shouting), neutral sounds (clapping, coughing, drinking),and sounds of moving objects (plates, chairs, toys, or kitchen tools).For the children's sounds, we added a separate category of childrenmoving around, because the running, jumping and climbing couldbe a separate source of noise. Finally, there was a category forsounds originating from the handling of different types of toys. Foreach of the ve main categories there was also an option for codersto add an item beyond the other denitions (other). In addition, wecoded whether the observation was inside or outside the buildingfor the majority of the time.

For each variable, coders reported on a four-point scale whetherthe described source of noise was 0 absent, 1 occasionally present(less than 25% of the time), 2 often present (between 25 and 75% ofthe time), or 3 constantly present (more than 75% of the time)during the 10-min observation. In total, 20 variables for sources ofnoise were rated for three sessions per child care center. Fourcoders were trained to be reliable on the coding instrument byusing a training set of 14 observations from ve centers.

Four coders were trained to be reliable on the coding instrumentby using a training set of 14 observations from ve centers. Inter-rater reliability of this training set was established to a criterion of75% of absolute agreement of the four coders with the rst author.The mean percentage of agreement for the training set was 81%(range 78%e85%). For the remaining DVDs interrater reliability wasestablished on the basis of double coding of 10% of the DVDs (sixsessions for each coder) by the rst author; mean percentage ofagreement for these 18 sessions was 78% (range 76%e82%).

2.4.4. Caregiver sensitivityCaregiver sensitivity was coded for three video recorded frag-

ments during regular child care activities. The scale to code care-giver sensitivity was developed and validated by the NetherlandsConsortium for Child care Research (De Kruif et al., 2007;Helmerhorst, Riksen-Walraven, Vermeer, Fukkink, & Tavecchio,2014). The Caregiver Interaction Prole (CIP; Helmerhorst et al.,2014) is a group rating scale based on scales developed to mea-sure sensitivity in the parent-child context (Ainsworth, Bell, &Stayton, 1974; Erickson, Sroufe, & Egeland, 1985).

Caregiver sensitivity refers to the degree to which the caregiverprovides adequate and sufcient emotional support to all childrenin her care who need it, during stressful and non-stressful situa-tions. It also refers to the level to which a caregiver is able toadequately divide attention among the children, showing interestin the children's activities and acknowledging their needs, emo-tions and competences.

C.D. Werner et al. / Journal of EnviroSensitivity ratings are presented on a seven-point scale, rangingfrom 1 very low to 7 very high. A caregiver scoring high on the scaleis very much involved with the children, and responds promptlyand adequately to the signals of all children in her care, by takingthe children's perspective. A caregiver scoring lowon this scalemayshow either emotional distance or indifference towards the chil-dren, or she may be uninvolved with the children, because ofadministrative or cleaning tasks in the group and thus missing thesignals of the children. Two independent observers were trained tobe reliable coders, using the NCKO reliability set (Helmerhorst et al.,2014). Intraclass correlation (ICC) for both coders was .75 (two-waymixed, absolute agreement).

Bivariate correlations between the three sessions of intra-caregiver sensitivity ranged from .24 to .59 and were all signi-cant on the p < .01 level. Themean change scores between the threesessions of sensitivity ranged from .86 to 1.02. Approximately 75%of the video recordings were coded by the two observers. Theremaining 25% were coded by the third author of this paper, whowas involved in developing the scale.

2.4.5. General child care qualityTo asses general child care quality a shortened version of the

ECERS-R (Harms et al., 1998; Vermeer et al., 2012) was adminis-tered in all participating groups. The ECERS-R is a revised version ofthe original instrument and has been used extensively acrosscountries. It contains seven subscales with 43 items and hasdemonstrated its validity and reliability (Perlman, Zellman, & Le,2004). Training and administration of the full ECERS-R are quitetime consuming; therefore in the current study a shortened versionwith 18 items was used. Psychometric properties of the shortenedversion of the ECERS-R are satisfactory (Vermeer et al., 2012). TheECERS-R item scores are presented on a seven-point scale so thatthe nal score for general child care quality is computed as theaverage of 18 items, ranging from 1 inadequate quality to 7 excellentquality.

In the reported study, internal consistency of this instrumentwas adequate, with Cronbach's alpha of .79. Seven observers weretrained by the third author of this paper to be reliable on theshortened ECERS-R. The training encompassed reviewing and dis-cussing the items and eld observations. Interrater reliability wasestablished to a criterion of 80% agreement within one rating pointfor three consecutive observations for all observers. The meanpercentage of agreement for these three observations was 90%(range 87%e92%).

2.4.6. Observed group size and caregiver-child ratioThe number of children and caregivers present during the three

observation sessions was registered by the observers. Group sizerefers to the total number of people in the room (both children,qualied caregivers and caregivers-in-training). The child-caregiver ratio was calculated as the number of children dividedby the number of qualied caregivers in the room.

2.4.7. Demographic informationCenter managers provided background information on the child

care centers through a questionnaire. In a background question-naire for the caregivers, information was gathered on their age,level of education, birth country, years of working experience inchild care, and working hours per week. They were also asked toreport the date of birth and number of hours of child care atten-dance per week for each of the target children in their care.

2.5. Data analysis

2.5.1. Data inspection

ntal Psychology 42 (2015) 190e201 193Data were collected on the group level (N 64) and the child

level (N 245). The dataset was inspected for normality of distri-butions and outliers. On the group level, the scores for caregiveryears of experience and child-caregiver ratio were not normallydistributed. For caregiver years of experience, one outlier accoun-tedmainly for skewness of the distribution.We applied a proceduresimilar to winsorizing: we replaced the inuential outlier with avalue closer to the distribution, keeping the rank order of theobserved scores. After dealing with the outliers, all predictors andoutcome variables were normally distributed.

Percentages of missing data ranged from 0% to 17% for variableson the group level. Caregiver questionnaires were not returned byten caregivers (16%), leading to missing data on caregiver age, yearsof experience and working hours per week for these subjects.Regarding data on the child level, percentages of missing dataranged between 1% for observed well-being and 27% for hours incare.

To obtain a complete dataset prior to the analyses, multipleimputation was performed (ten times) (Goldstein & Woodhouse,1996; Van Buuren, 2011) including available variables in the dataset on the child level and the group level, using predictive meanmatching to impute missing data (Little, 1988; Rubin, 1986) andtaking the nested structure of the data into account. Finally, thepooled imputed dataset (N 245) was used for subsequent

multilevel analysis. Before applying multilevel regression analysis,all predictor variables were centered by using the average score foreach imputed dataset.

2.5.2. Multilevel analyses

The sample consisted of children (N 245) who were nestedwithin child care groups (N 64). This dependency among childrenwas taken into account by performing multilevel analysis, or hier-archical linear modeling (HLM), using Mixed models in SPSS 19.0. Atwo-level random intercept model was used to predict well-beingon the individual child level (level 1) from child characteristics.Differences between the groups on child well-being were investi-gated with variables on the group level (level 2). Full maximumlikelihood was used for estimating the model parameters. Toinvestigate model t we used 2 log likelihood (-2LL) values.However, imputed datasets do not provide pooled values of -2LL.Therefore, we used the imputed dataset with the highest e 2LL inthe rst model (Model 0) to evaluate model tness (see Table 3).

2.5.3. Nonlinear data analysisWe used categorical regression analysis (CATREG; Meulman,

Heiser & IBM SPSS, 2012) in SPSS 21.0 to examine possible

Table 2Pearson correlations between characteristics of the children, caregivers and child care groups before and after imputationc (N 245).

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

1 Child gendera e .10 .00 .12 .03 .04 .09 .13 .07 .03 .02 .032 Child age .10 e .03 .28 .10 .11 .07 .04 .11 .18 .11 .043 Child hours in care .04 .04 e .16 .03 .10 .12 .08 .13 .03 .07 .144 Child well-being .12 .29 .20 e .18 .03 .18 .17 .06 .06 .06 .055 Caregiver experience (yrs) .04 .10 .05 .21 e .12 .12 .02 .08 .15 .01 .056 Caregiver working hours .03 .11 .11 .03 .14 e .25 .21 .02 .16 .12 .047 Quality of child care .09 .08 .16 .14 .13 .28 e .44 .15 .07 .01 .148 Caregiver sensitivity .13 .02 .11 .17 .00 .22 .44 e .08 .30 .16 .199 Group size .07 .13 .18 .04 .14 .10 .15 .08 e .35 .28 .2510 Ratiob .03 .19 .03 .07 .14 .16 .07 .30 .36 e .22 .1211 Noise level (dbA) .01 .10 .12 .07 .01 .10 .00 .14 .27 . 22 e .2812 Noise variability .02 .03 .16 .05 .04 .03 .15 .18 .24 .14 .28 ea Boy 0, girl 1.

latio

l we

nt (S

)

))

)

C.D. Werner et al. / Journal of Environmental Psychology 42 (2015) 190e201194b Number of children per caregiver.c Bivariate correlations before imputation are presented under the diagonal, corre

Table 3Results of two-level hierarchical linear regression analysis to predict child emotiona

Parameter Model 0 Model 1 Model 2

Coefcient (SE) Coefcient (SE) Coefcie

Fixed effectsIntercept 4.50 (.04) 4.46 (.05) 4.46 (.05Level 1 (child)Gender .08 (.06) .08 (.06)Age .01* (.00) .01* (.00Hours in care .01* (.00Level 2 (group)Caregiver years of experienceCaregiver working hoursQuality of care (ECERS-R)Caregiver sensitivityGroup sizeCaregiver-child ratioNoise levelsNoise variabilityRandom parametersIntercept (variance) .05* (.02) .04* (.02) .04* (.022 Log likelihooda 332.12 316.22 300.05*p < .05.a We selected the imputed dataset with the highest value on this variable in Model 0ns after imputation above the diagonal; values .30 are presented in bold.

ll-being (N 245).

Model 3 Model 4 Model 5 Model 6

E) Coefcient (SE) Coefcient (SE) Coefcient (SE) Coefcient (SE)

4.50 (.04) 4.50 (.03) 4.50 (.04) 4.50 (.03)

.01* (.00) .02* (.00) .01* (.00) .02* (.00)

.01* (.00) .01* (.00) .01 (.00) .01* (.00)

.01 (.01).00 (.01)

.12* (.05) .11* (.05)

.03.(04) .03 (.05).02 (.01) .01 (.01).01 (.03) .01 (.03).00 (.01) .00 (.01).04 (.05) .01 (.05)

.04* (.02) .03* (.01) .04* (.02) .03 (.01)300.81 294.19 299.50 292.82. The relative changes in -2LL were comparable for all imputed datasets.

(CATPCA; Linting & Van der Kooij, 2012) in SPSS 21.0 to examine

nmemultivariate relations between the ratings for sources of noise. Incontrast to linear principal component analysis, nonlinear PCA al-lows researchers to handle nonlinear relationships between vari-ables and to specify analysis levels separately for each variable, sothat these can be in accordance with the measurement level of thevariables (Linting, Meulman, Groenen, & Van der Kooij, 2007a,2007b).

Sources of noise were measured at an ordinal measurementlevel, as these variables were scored on four-point Likert scales. Foreach of the 20 sources of noise variables, we aggregated scores overthe three 10-min observations. The aggregated scores were thenrounded off to the original four-point scale, so that child carecenters' average scores were categorized as 0, 1, 2, or 3 for eachvariable. Next, we examined whether numerical, (spline) ordinal or(spline) nominal scaling were the best tting transformations foreach of the variables. We nally chose to use ordinal trans-formation, taking into account the quantication plots and theoriginal four-point ordinal rating scale. In Appendix B the pro-cedures for nonlinear PCA are described in more detail.

In order to relate the solution of the nonlinear PCA to theaverage noise levels and average noise variability of the centers, wecreated three categories for these variables: centers fell in thecategory of low noise (65.00 dbA, n 16), andfor variability centers could be of low variability (8.00dbA, n 11). These cut-off points were used in noise studies before(Belojevic, Jakovljevic, Stojanov, Paunovic, & Ilic, 2008; Manloveet al., 2001). These categorized variables of noise were used inthe CATPCA only and plotted as supplementary variables in thenonlinear relations between child well-being and noise character-istics, similar to Linting et al. (2013). Nonlinear regression is analternative to linear regression, developed for performing ordinaryleast squares regression on possibly nonlinearly related variables(Van der Kooij, 2007). The method is quite similar to using linearregression with transformed data (e.g. log transformation). Anadvantage of CATREG is that an optimal transformation is per-formed, that is, a transformation that best reects the relation be-tween the transformed predictor and the response, given particularrestrictions imposed by the researcher.

We applied a spline nominal analysis level to our predictorvariables average noise level and noise variability, which meansthat the transformation of these variables follows a smooth curvethat may go up and down with the original order of the values. Weused the SPSS default settings, estimating a quadratic curve in threedata intervals. To ensure interpretability of the results, the responsevariable well-being was analyzed numerically (without trans-formation). The reported p-values from the CATREG models arebased on 50 bootstraps.

Currently, nonlinear regression in SPSS does not allow for hi-erarchically structured (multilevel) data, nor does it allow poolingof results across imputed data sets. Therefore, analyses could onlybe performed on the group level (N 64) and for one (randomlyselected) imputation. We aggregated well-being scores for theselected imputed dataset by computing the mean across childrenwithin the same child care group. The nonlinear regression analysiswas repeated for all other imputed datasets and similar resultswere yielded. We therefore present the results for one randomlyselected imputed dataset only.

2.5.4. Nonlinear principal component analysisWe used nonlinear (categorical) principal component analysis

C.D. Werner et al. / Journal of Envirosolution dened by the sources of noise.3. Results

3.1. Descriptive statistics

Means, standard deviations and the range of scores for variableson the child level and the group level are provided in Table 1.Bivariate Pearson correlations between background characteristicsand outcome variables are presented in Table 2. Given the multi-level nature of the data, bivariate Pearson correlations are inter-preted in an explorative way. Correlations with a value of .30 orhigher are discussed here, because this value corresponds toapproximately 10% explained variance which makes it meaningfulfor interpretation. Following this guideline, larger group sizes wereassociated with more children per caregiver. Higher general childcare quality was associated with higher caregiver sensitivity levels.Finally, larger group sizes were associated with higher caregiversensitivity.

3.2. Multilevel analyses

The intra class correlation for well-being (calculated on a modelwith just a random intercept and no predictors) was .22, indicatingthat multilevel analyses are indeed preferable beyond regularlinear regression. As a comparison, in educational research, intra-class correlations of around .10 are quite common (Hox, 2010;Twisk, 2006). We hierarchically added terms to the model, rston level 1 and then on level 2. Results of the Models 1 to 6 arepresented in Table 3. The multilevel analyses showed that morehours in care, higher child age and higher general quality in thegroup were related to higher levels of well-being. The other pre-dictors in the model did not signicantly contribute to the predic-tion of well-being.

3.2.1. ModeratorsThe interaction of child age with hours in care was not signi-

cant (b .00, SE .00, t .32, p .75). As a next step, we testedcross-level interactions. None of these interactions turned out to besignicant (see Appendix C, Table C1).

3.3. Nonlinear analyses

To test nonlinear relations between noise characteristics andwell-being, we used CATREG on the aggregated data. The child caregroups (N 64) had a total mean score for aggregated well-being of4.50 (SD . 32), ranging from 3.79 to 5.25. The linear model inCATREG (no transformation, numerical level of analysis for allvariables) only including average noise and noise variabilityexplained 3% of the variance in well-being (R2 .03, p .43). Themodel with a spline ordinal transformation level for the noisevariables did not show notable improvement (R2 .05, p .24). Themodel allowing for nonmonotonic nonlinear relations, however,showed much improvement over the linear and ordinal models(R2 .24, df 8, F 2.12, p .049) and well-being could besignicantly predicted by average noise levels (b .36, p < .001)and noise variability (b .33, p < .001). Transformation plots ofthe nonlinear regression analyses are displayed for average noiselevels (Fig. 1) and noise variability (Fig. 2).

3.3.1. Average noise levelsThe interpretation of the positive relation between transformed

noise and well-being (b .36, p < .001) becomes clear in Fig. 1. Therst part of the plot shows that as noise levels increased, child well-being levels (slightly) decreased. After a certain noise level (about60.63 dbA) was reached, the association changed: well-being

ntal Psychology 42 (2015) 190e201 195increased with an increase of noise. Finally, beyond a noise level

nmeC.D. Werner et al. / Journal of Enviro196of about 65.14 dbA, an increase in noise level was related to adecrease in child well-being, again. There seemed to be a minimumamount of noise required to reach optimal well-being levels, but wealso found a maximum. If noise levels fell below the rst or abovethe latter threshold, child well-being decreased.

3.3.2. Noise variabilityFor noise variability (b .33, p < .001) the pattern was com-

parable to the pattern for average noise levels (see Fig. 2). Well-being decreased with increasing noise variability for the lowerrange of scores (approximately from 6.10 dbA to 6.69 dbA). Then,there was an increase of well-being up to a certain level of noisevariability (around 7.44 dbA), after which there was a decrease inwell-being with an increase of noise variability. In the last part ofthe gure (beyond 8.23 dbA), an increase of noise variability seemsto correspondwith an increase of well-being, again. However, given

Fig. 1. Plot of nonlinear regression analysis (spline nominal transformation) for child

Fig. 2. Plot of nonlinear regression analysis (spline nominal transformation) for chintal Psychology 42 (2015) 190e201the fact that only a relatively small number of observations (n 9)are plotted in this part, this nding is less reliable.

3.3.3. Nonlinear model with covariatesIn an additional analysis, we entered covariates child hours in

care and child age as numerical variables to the model to seewhether noise indicators would remain signicant predictors (seeTable 4). Child age proved to be a signicant predictor (b .44,p < .001), indicating that older children had higher levels of well-being. The nonlinear model with covariates added 16% explainedvariance in well-being (R2 .40, p < .01). Average noise levels(b .39, p < .001) and noise variability remained signicant pre-dictors (b .31, p < .001) in this model.

3.3.4. Nonlinear principal component analysis for sources of noiseFor this analysis we used the categorized variables including the

emotional well-being predicted by average noise levels. Note: b .36, p

low, middle, and high categories for noise levels and noise vari-ability. We rst explored a two-dimensional solution in CATPCA.There were no missing data and no outliers in the object plot(N 64). The variance accounted for (VAF) by the two-dimensionalsolution was 25%, with eigenvalues of 2.51 (VAF 12%) for the rstdimension and 2.39 (VAF 12%) for the second dimension. As thescree plot of the saved transformed variables in a linear PCAshowed an elbow curve after the second dimension and as thethird dimension did not contribute much in terms of interpretation(Linting et al., 2007a), we decided to staywith the two-dimensional

Table 4Standardized regression coefcients from CATREG for outcome variable childemotional well-being.

Predictor Model A(R2 .24, p .05)

Model B(R2 .40, p < .01)

b Bs SE p b Bs SE p

Average noise levelsa .36 .10

nmeshould be viewed as an important aspect of the child care envi-ronment on its own. We conclude that noise is an important aspectin the chaos literature, but also that it is an element that should beconsidered separately from organizational chaos, a notion that hasbeen put forward by Matheny et al. (1995).

Group size and child-caregiver ratio are still important factors totake into account. In our study, these variables were positivelyassociated with one another and child-caregiver ratio was associ-ated with higher caregiver sensitivity. In previous research, morechildren per caregiver have been associated with more negativecaregiver-child interaction (De Schipper, Riksen-Walraven, &Geurts, 2006). To explain this difference in ndings, we point outthat in our study unstructured situations were used as the contextto measure caregiver sensitivity, as opposed to structured playsituations of De Schipper et al. (2006). Furthermore, a threshold ofchild-caregiver leading to sub-optimal levels of child well-beinghas not yet been established empirically. Ratios of as many asseven children to one caregiver may not result in more negativechild-caregiver interaction for children aged three years (DeSchipper et al., 2006) and in our study the child-caregiver ratioranged from 2:1 to 13:1 and was on average 5.41:1 (SD 1.76). Itshould be noted that we did not measure crowding as the numberof children per square feet of available space which would be amore valid indicator of this component of chaos.

The effect sizes for noise levels and noise variability as pre-dictors of child well-being in the nonlinear analysis were sub-stantial, given the high percentage of explained variance and themoderate to large values of the regression coefcients. This in-dicates that noise should be acknowledged as an important factorin child care contributing to child emotional well-being. However,generalizability of the ndings is dependent on replications of thecurrent study in other settings and countries. Importantly, datawould be more valid if collected on several occasions across a one-month period instead of being derived from an extensive but singlemorning visit. Noise levels at child care centers during the morningmay differ from noise levels during the afternoon. Therefore itwould be interesting to investigate noise levels throughout the dayin future studies.

4.1.1. Child, caregiver, and care characteristicsThe average scores for observed emotional well-being indicate

that children in our sample were generally feeling quite well intheir child care center. Child emotional well-being could be pre-dicted from general child care quality, but, surprisingly, not fromcaregiver sensitivity. One explanation is that caregiver sensitivityshowed variation across time and was only modest in stability. Toaccount for the modest stability and the context-specic nature ofcaregiver sensitivity, multiple measurements were included.

Insufcient variance in caregiver sensitivity may explain whywe were unable to nd a signicant relation with well-being:caregivers scored relatively high. In contrast, there was substan-tial variance in general quality of care, which allowed us to ndsignicant relations with well-being. Still, the associations betweencaregiver sensitivity and quality of care in our study imply thatlower child care quality, predictive of lower child well-being, oftencoincides with less sensitive caregiving. It should be noted that theeffect size for quality of care as a predictor of child emotional well-being was quite small.

Another signicant predictor for child well-being was child age:within the age range of zero to four years relatively older childrenshowed higher levels of well-being, although it should be notedthat the effect size was small. Child gender, hours in care, care-giver's years of experience and working hours did not predict childwell-being. Because the inuence of noise, or more broadly, chaos,

C.D. Werner et al. / Journal of Enviro198may be conceptualized as the optimal t between the individual'sprocessing capacity and the environmental stimulus ow, the ef-fects of noise on well-being and development might be dependenton the dynamic Piagetian interplay of accommodation and assim-ilation, and on individual differences in the capacity to self-regulatethe inux and processing of noisy stimuli as suggested in theoptimal stimulation hypothesis (Hunt, 1961; McCall & McGhee,1977; Wachs, 1977).

The correlational nature of this study does not allow us to drawconclusions about cause and effect, for instance between noiselevels and child well-being or caregiving quality. Still, we tried toshed some light on the causes of high and low noise levels bylooking in detail to the sources of noise. The aim of this explorativeanalysis was to gain insight as to which kind of situations mayresult in very high or very low noise levels, and thus negativelyaffect child emotional well-being.

Sources of noise in this study could be distinguished along twodifferent dimensions. The rst dimension, representing child ac-tivities on the group versus background activities, was not clearlyrelated to noise levels: both types of noise sources seem to berelated to high and low noise levels. On the second dimensionhowever, outdoor activities were clearly related to high noiselevels, whereas indoor activities were related to low noise levels.The latter seem to consider relatively quiet indoor activities, such aschildren playing with soft toys like dolls and small cars, and care-givers moving about in the room.

With respect to child well-being, wemay speculate that outdooractivities with high noise levels are related to more rough play,which may in turn be related to lower child emotional well-beingbecause of accidents with gross motor equipment or peers. Onthe other hand, very quiet or even dull indoor situations with fewor no challenges may result in low well-being as well. We can alsocautiously relate the sources of noise to the caregiver sensitivitylevels, because sensitivity was rated from the same observations asthe ones used for the coding of noise sources. Bivariate correlationsin this study showed that higher average noise levels and lowernoise variability were related to higher caregiver sensitivity levels.

Possibly, higher average noise levels (with less noise uctuation)are observed in outdoor situations in which caregivers are espe-cially focused on supervising and responding to the children,anticipating children's more rough outdoor play or accidents thatmay occur when children are riding bicycles and climbing glides. Incontrast, in more neutral situations of quiet indoor play caregiversmay be occupied by other tasks (neutral sounds and handling objectsmay for instance include handing out sandwiches or clearing awaytoys) in which they are less focused on the children. It seems thatsome situations and activities may facilitate or evoke more sensi-tive caregiving than others. This context-specicity can be takeninto account when designing intervention programs aimed atimproving caregiver sensitivity and child well-being.

4.2. Implications for research and practice

This study adds to the scientic literature of environmentalchaos theory in center-based child care. We showed that in childcare centers noise levels stemming from both social and non-socialsources can exceed a threshold after which noise levels becomedetrimental to child emotional well-being. On the other hand, verylow levels of noise do not seem benecial to child well-being either,potentially because these noise levels occur when not enough ac-tivities or play material are provided. The effects of noise weresubstantial, so that we conclude that child well-being may beimproved substantially by taking into account noise levels as anindicator of child care quality.

It should be noted, however, that the only outcome we measured

ntal Psychology 42 (2015) 190e201was child well-being. Noise has been found to affect other outcomes,

care. Identical procedures, measurements and data analysis wereapplied which allowed for comparisons of noise levels and vari-ability across settings. It has become apparent that, compared tohome-based child care, center child care is a different ecologicalniche where higher average levels of noise are experienced on adaily basis and where other sources of noise play a role. Still, withthis study we further supported the theory that a threshold fornoise in relation to child emotional well-being is apparent, not onlyin home-based child care (Linting et al., 2013), but also in centerchild care. Noise levels form a major aspect of the quality of carethat should be integrated in other measures of child care quality.

An implication of our study is that noise levels in child carecenters should be regulated for optimal child well-being. To lowerdetrimental sound levels in noisy child care centers, interventionsto improve acoustics could be applied, such as the placing ofabsorbent plates (Kishimoto, 2012; Maxwell & Evans, 2000).However, adequate noise levels on the lower end should beanchored as well: adequate activities and caregiving should providea minimum level of stimulation. Thus, caregivers might be trainedto be sensitive to the lower and upper boundaries of noise level and

Appendix B. Procedures in nonlinear PCA

Before conducting the CATPCA, we took the steps explained byLinting and Van der Kooij (2012) to ensure the right method andprocedures in the analysis. For instance, stability of the CATPCA canbe at risk when categories have small marginal frequencies andstrongly affect the solution (Linting et al., 2007b). We dealt withthis issue of stability by requiring aminimumof ve observations ineach category: when a category had less than ve observations itwas merged with the closest category by recoding the center'sscores on this variable to the adjoining category score. This resultedin 14 dichotomous variables (Children's positive vocalizations andAdults' positive vocalizations had scores in category 2 and 3 only; 12variables had scores in category 0 and 1 only) and ve variableswith three categories (Children moving objects, Children movingaround, Adults' neutral sounds, Adults moving objects, and Toyseother).

One variable (Background noise: Other) retained the four originalcategories and as this variable's quantication plot showed a

14 dichotomous variables as numerical, too; for the remaining vevariables we applied an ordinal level (ordinal quantication, with

ple

utsiup oer,et grr slas (tans (pinlappningent)s (tans (pintes,ent).g. bil to

Musical instruments/muOther (please comment)

nmevariability that should be taken into account to optimize child well-being. Experiments are needed to evaluate the effectiveness ofthese kinds of interventions and to provide more insight in thecausal mechanisms between noise levels and child emotional well-being. Ultimately, effective interventions may help to protect pro-fessionals and children in child care from harmful levels of noise.

Author note

This study was supported by a grant from the Netherlands Or-for example in the area of cognitive or health development. Also,World Health Organization (WHO) target levels for acceptable noisefor adults (WHO, 1999), and guidelines of the American Academy ofAudiology (2011) for children imply lower thresholds thanwe foundin our study on noise in center care and child well-being. Beforedeciding about acceptable noise levels in child care other develop-mental outcome variables should be studied.

Our study is, in part, a replication of the investigation by Lintinget al. (2013) in a different setting: home-based care versus center

Category Denitions and exam

1 Outdoor Trafc/Other noise o2 Background Sounds in target gro

(music from CD playSounds outside targOther (e.g. front doo

3 Children Positive vocalizationNegative vocalizatioNeutral sounds (clapThrowing/pushing/sMoving around (runOther (please comm

4 Adults Positive vocalizationNegative vocalizatioNeutral sounds (clapMoving objects (plaOther (please comm

5 Toys Large motor toys: eMechanical/electricaBlocks, Lego, balls

C.D. Werner et al. / Journal of Enviroganization for Scientic Research (NWO Spinoza prize), awarded todiscretizing option ranking).After conducting nonlinear PCA, we checked for outliers in the

object plot. We explored two-dimensional and three-dimensionalsolutions. We considered variables to be tting adequately in asolution when their component loading on at least one of the di-mensions was higher than .30. Component loadings for sources ofnoise variables for the two-dimensional solution are presented instraight line, we treated this variable as numerical. We treated thesical toyse.g. plastic toys, play kitchenM.H. Van IJzendoorn.We are grateful to Joost van Ginkel (Centre forChild and Family Studies, Leiden University) for his assistance in themultilevel multiple imputation procedure. Correspondence con-cerning this article should be addressed to Harriet Vermeer, Centrefor Child and Family Studies, Leiden University, the Netherlands.Email: vermeer@fsw.leidenuniv.nl.

Appendix A. Coding system sources of noise

s

de the buildingn the background in target grouptelephone ringing, doorbell)oup (human conversation in hallway, corridor, other groups)mming)lking, laughing, singing)crying, screaming, ghting)g, coughing, eating, drinking)ing/moving objects (plates, cups, chairs), crawling, jumping)

lking, laughing, singing)scolding, shouting, irritated/angry vocalizations)g, coughing, eating, walking)cups, chairs, toys, kitchen tools)

cyclesys (e.g. cars)

ntal Psychology 42 (2015) 190e201 199Table B1.

2 Adults moving objects (kitchen tools, material, toys) .07 .68

5 Toys sounds: other (e.g. plastic kitchen tools) .33 .456 Toys: musical toys* .24 .24Cluster 27 Toys sounds: blocks (e.g. Lego) .48 .278 Children moving around (e.g. running, climbing, jumping) .70 .279 Children's positive vocalizations .71 .0110 Children's negative vocalizations .48 .11Cluster 311 Adult negative vocalizations* .21 .2512 Children's neutral sounds (clapping, coughing) .30 .4513 Toys sounds: motor development toys (bicycles etc.) .30 .5414 Background sounds from outdoor (e.g. trafc passing) .08 .30Cluster 415 Children's other sounds .43 .0116 Adults' other sounds .38 .0217 Background sounds: media (e.g. CD-player, doorbell) .38 .2518 Background sounds: human conversation outside group .30 .3919 Background sounds: other* .18 .18Other20 Adult positive vocalizations* .10 .12

*Component loadings of.30 on both dimensions: variable does not seem to t wellin the two-dimensional solution.

nme3 Adult neutral sounds (e.g. clapping, coughing, walking) .17 .534 Toys sounds: mechanical toys (e.g. small cars) .17 .38Table B1Component loadings of sources of noise variables for the two-dimensional solutionin CATPCA (values >.30 in bold).

Source of noise variable Componentloadings

Dimension

1 2

Cluster 11 Children moving, slapping, throwing objects .10 .41

C.D. Werner et al. / Journal of Enviro200Appendix C. Moderator analysis of cross-level interactions

References

Ainsworth, M. D. S., Bell, S. M., & Stayton, D. (1974). Infant-mother attachment andsocial development. In M. P. Richards (Ed.), The introduction of the child onto asocial world (pp. 99e135). London, United Kingdom: Cambridge UniversityPress.

American Academy of Audiology. (2011). Position statement: classroom acoustics.Retrieved fromhttp://www.audiology.org/publications-resources/document-library/classroom-acoustics.

Babisch, W., Schulz, C., Seiwert, M., & Conrad, A. (2012). Noise annoyance as re-ported by 8-to-14-year old children. Environment and Behavior, 44(1), 68e86.http://dx.doi.org/10.1177/0013916510387400.

Belojevic, G., Jakovljevic, B., Stojanov, V., Paunovic, K., & Ilic, J. (2008). Urban road-

Table C1Moderator analysis of cross-level interactions (N 245).

Cross-level interaction Moderator variable Beta SE t p

Variable level 1 Variable level 2

Well-being Average noiselevels

Child gendera .01 .02 .64 .53Child hours in careb .00 .00 .31 .76

Well-being Noise variabilityChild gender .02 .07 .30 .77Child hours in care .00 .00 .18 .86

Well-being Quality of careChild gender .08 .07 1.03 .30Child hours in care .00 .01 .49 .62

Well-being Caregiversensitivity

Child gender .08 .06 1.25 .21Child hours in care .01 .00 1.53 .13

a Boy 0, girl 1.b Child hours in care was entered as continuous variable.trafc noise and blood pressure and heart rate in preschool children. Environ-ment International, 34, 226e231. http://dx.doi.org/10.1016/j.envint.2007.08.003.

Corapci, F. (2010). Child-care chaos and child development. In G. W. Evans, &T. D. Wachs (Eds.), Chaos and its inuence on children's development: anecological perspective (pp. 67e82). Washington, DC: American PsychologicalAssociation.

De Kruif, R. E. L., Vermeer, H. J., Fukkink, R. G., Riksen-Walraven, J. M. A.,Tavecchio, L. W. C., & Van IJzendoorn, M. H. (2007). De nationale studie ped-agogische kwaliteit kinderopvang: Eindrapport project 0 en 1 [The national studyon child care quality: Final report project 0 and 1]. Amsterdam, theNetherlands: NCKO.

De Schipper, E. J., Marianne Riksen-Walraven, J., & Geurts, S. A. (2006). Effects ofchildecaregiver ratio on the interactions between caregivers and children inchild-care centers: an experimental study. Child Development, 77(4), 861e874.http://dx.doi.org/10.1111/j.1467-8624.2006.00907.x.

De Schipper, J. C., Tavecchio, L. W., Van IJzendoorn, M. H., & Van Zeijl, J. (2004).Goodness-of-t in center day care: relations of temperament, stability, andquality of care with the child's adjustment. Early Childhood Research Quarterly,19(2), 257e272. http://dx.doi.org/10.1016/j.ecresq.2004.04.004.

Enmarker, I., & Boman, E. (2004). Noise annoyance responses of middle schoolpupils and teachers. Journal of Environmental Psychology, 24, 527e536. http://dx.doi.org/10.1016/j.envp.2004.90.005.

Erickson, M. F., Sroufe, L. A., & Egeland, B. (1985). The relationship between qualityof attachment and behavior problems in preschool in a high-risk sample.Monographs of the Society for Research in Child Development, 147e166.

Evans, G. W. (2006). Child development and the physical environment. AnnualReview of Psychology, 57, 423e451. http://dx.doi.org/10.1146/annrev.psych.57.102904.190057.

Evans, G. W., & Wachs, T. D. (Eds.). (2010). Chaos and its inuence on children'sdevelopment: An ecological perspective. Washington, DC: American PsychologicalAssociation.

Eysel-Gosepath, K., Daut, T., Pinger, A., Lehmacher, W., & Erren, T. (2012). Soundlevels and their effects on children in German primary school. European Ar-chives of Oto-rhino-Laryngology, 269, 2475e2483. http://dx.doi.org/10.1007/s00405-011-1899-x.

Goldstein, H., & Woodhouse, G. (1996). Multilevel models with missing data. Paperpresented at the 11th International workshop on statistical modeling, Orvieto,Italy.

Groeneveld, M. G., Vermeer, H. J., Van IJzendoorn, M. H., & Linting, M. (2010).Children's wellbeing and cortisol levels in home-based and center-based childcare. Early Childhood Research Quarterly, 25, 502e514. http://dx.doi.org/10.1016/ecresq.2009.12.004.

Haines, M. M., Stansfeld, S. A., Brentnall, S., Head, J., Berry, B., Jiggings, M., et al.(2001). The West London schools study: The effects of chronic aircraft noiseexposure on child health. Psychological Medicine, 31, 1385e1396. http://dx.doi.org/10.1017/S003329170100469X.

Hambrick-Dixon, P. J. (1986). Effects of experimentally imposed noise on task per-formance of black children attending daycare centers near elevated subwaytrains. Developmental Psychology, 22, 259e264. http://dx.doi.org/10.1037/0012-1649.22.2.259.

Harms, T., Clifford, R. M., & Cryer, D. (1998). Early childhood environment rating scale-revised. New York, NY: Teachers College Press.

Helmerhorst, K. O. W., Riksen-Walraven, J. M., Vermeer, H. J., Fukkink, R. G., &Tavecchio, L. W. C. (2014). Measuring the interactive skills of caregivers in childcare centers: development and validation of the caregiver interaction prolescales. Early Education and Development, 1e12. http://dx.doi.org/10.1080/10409289.2014.840482.

Hox, J. (2010). Multilevel analysis: techniques and applications. New York, NY:Routledge.

Hunt, J. M. (1961). Intelligence and experience. New York: Ronald.Kishimoto, T. M. (2012). The integration of care and education: a case concerning

the problem of noise. European Early Childhood Education Research Journal,20(4), 493e503. http://dx.doi.org/10.1080/1350293x.2012.737240.

Kjellberg, A., Landstrom, U., Tesarz, M., Soderberg, L., & kerlund, E. (1996). Theeffects of nonphysical noise characteristics, ongoing task and noise sensitivityon annoyance and distraction due to noise at work. Journal of EnvironmentalPsychology, 16, 123e136. http://dx.doi.org/10.1006/jevp.1996.0010.

Klatte, M., Bergstrom, K., & Lachmann, T. (2013). Does noise affect learning? A shortreview of noise effects on cognitive performance in children. Frontiers inDevelopmental Psychology, 4. http://dx.doi.org/10.3389/fpsyg.2013.00578,1664e1078.

Lindstrom, F., Persson Waye, K., Sodersten, M., McAllister, A., & Ternstrom, S. (2011).Observations of the relationship between noise exposure and preschool teachervoice usage in day-care environments. Journal of Voice, 25(2), 166e172. http://dx.doi.org/10.1016/j.jvoice.2009.09.009.

Linting, M., Groeneveld, M. G., Vermeer, H. J., & Van IJzendoorn, M. H. (2013).Threshold for noise in child care: noise level and noise variability are associatedwith child wellbeing in home-based child care. Early Childhood ResearchQuarterly, 28, 960e971. http://dx.doi.org/10.1016/j.ecrq.2013.03.005.

Linting, M., Meulman, J. J., Groenen, P. F., & Van der Kooij, A. J. (2007a). Nonlinearprincipal components analysis: introduction and application. PsychologicalMethods, 12(3), 336e358. http://dx.doi.org/10.1037/1082-989X.12.3.336.

Linting, M., Meulman, J. J., Groenen, P. F., & Van der Kooij, A. J. (2007b). Stability of

ntal Psychology 42 (2015) 190e201nonlinear principal components analysis: an empirical study using thebalanced bootstrap. Psychological Methods, 12(3), 359e379. http://dx.doi.org/

10.1037/1082-989X.12.3.359.Linting, M., & Van der Kooij, A. (2012). Nonlinear principal component analysis with

CATPCA: a tutorial. Journal of Personality Assessment, 94(1), 12e25. http://dx.doi.org/10.1080/00223891.2011.627965.

Little, R. J. A. (1988). Missing-data adjustments in large surveys. Journal of Businessand Economic Statistics, 6, 287e296. http://dx.doi.org/10.1080/07350015.1988.10509663.

Manlove, E. E., Frank, T., & Vernon-Feagans, L. (2001). Why should we care aboutnoise in classrooms and child care settings? Child & Youth Care Forum, 30(10),55e64.

Matheny, A. P., Wachs, T. D., Ludwig, J. L., & Phillips, K. (1995). Bringing order out ofchaos: psychometric characteristics of the confusion, hubbub, and order scale.Journal of Applied Developmental Psychology, 16(3), 429e444. http://dx.doi.org/10.1016/0193-3973(95)90028-4.

Maxwell, L. (2010). Chaos outside the home: the school environment. InG. W. Evans, & T. D. Wachs (Eds.), Chaos and its inuence on children's devel-opment: An ecological perspective (pp. 83e95). Washington, DC: AmericanPsychological Association.

Maxwell, L. E., & Evans, G. W. (2000). The effects of noise on preschool children'spre-reading skills. Journal of Environmental Psychology, 20, 91e97. http://dx.doi.org/10.1006/jevp.19999.01444.

McAllister, A. M., Granqvist, S., Sjolander, P., & Sundberg, J. (2009). Child voice andnoise: A pilot study of noise in day cares and the effects on 10 children's voicequality according to perceptual evaluation. Journal of Voice, 23(5), 587e593.http://dx.doi.org/10.1016/j.jvoice.2007.10.017.

McCall, R. B., & McGhee, P. (1977). The discrepancy hypothesis of attention andaffect in infants. In I. Uzgiris, & R. Weizman (Eds.), The structuring of experience(pp. 179e210). New York, NY: Plenum.

Meulman, J. J., Heiser, W. J., & IBM SPSS. (2012). IBM SPSS categories 21.0. Chicago, IL:SPSS.

Nederlands Consortium Kinderopvang Onderzoek. (2006). Schalen leid-stervaardigheden. Handleiding (verkort). Amsterdam, the Netherlands: NCKO[Caregiver scales. Manual (shortened version)].

Perlman, M., Zellman, G. L., & Le, V. N. (2004). Examining the psychometric prop-erties of the early childhood environment rating scale-revised (ECERS-R). EarlyChildhood Research Quarterly, 19, 398e412. http://dx.doi.org/10.1016/j.ecresq.2004.07.006.

Rubin, D. B. (1986). Statistical matching using le concatenation with adjustedweights and multiple imputations. Journal of Business and Economic Statistics, 4,87e94.

Sala, E., Airo, E., Olkinuora, P., Simberg, S., Strom, U., Laine, A., et al. (2002). Vocalloading among day care center teachers. Logopedics Phoniatrics Vocology, 27,21e28. http://dx.doi.org/10.1080/140154302760146943.

Shield, B. M., & Dockrell, J. E. (2003). The effects of noise on children at school: areview. Building Acoustics, 10(2), 97e116. http://dx.doi.org/10.1260/135101003768965960.

Stansfeld, S. A., Clark, C., Cameron, R. M., Alfred, T., Head, J., Haines, M. M., et al.(2009). Aircraft and road trafc noise exposure and children's mental health.Journal of Environmental Psychology, 29, 203e207. http://dx.doi.org/10.1016/j.jenvp.2009.01.002.

Twisk, J. W. (2006). Applied multilevel analysis: A practical guide. Cambridge Uni-versity Press.

Van Buuren, S. (2011). Multiple imputation of multilevel data. In J. J. Hox, &J. K. Roberts (Eds.), The handbook of advanced multilevel analysis (pp. 173e196).New York, NY: Routledge.

Van der Kooij, A. J. (2007). Prediction accuracy and stability of regression with optimalscaling transformations (Doctoral dissertation). Retrieved from Open AccessLeiden University https://openaccess.leidenuniv.nl/handle/1887/12096.

Vermeer, H. J., Werner, C. D., Linting, M., VanIJzendoor, M. H, & Groeneveld, M. (July,2012). The ECERS-R brief: Psychometric properties (Poster presented at thebiennial meeting of the International Society for the Study of BehavioralDevelopment, Alberta, Canada).

Wachs, T. D. (1977). The optimal stimulation hypothesis and human development.In I. Uzgiris, & R. Weizman (Eds.), The structuring of experience (pp. 153e177).New York, NY: Plenum.

World Health Organization. (1999). Guidelines for community noise. Geneva: WHO.

C.D. Werner et al. / Journal of Environmental Psychology 42 (2015) 190e201 201

Noise in center-based child care: Associations with quality of care and child emotional wellbeing1. Introduction1.1. Aim of the study

2. Method2.1. Recruitment2.2. Participants2.3. Procedure2.4. Measures2.4.1. Child emotional well-being2.4.2. Noise levels and variability2.4.3. Sources of noise2.4.4. Caregiver sensitivity2.4.5. General child care quality2.4.6. Observed group size and caregiver-child ratio2.4.7. Demographic information

2.5. Data analysis2.5.1. Data inspection

2.5.2. Multilevel analyses2.5.3. Nonlinear data analysis2.5.4. Nonlinear principal component analysis

3. Results3.1. Descriptive statistics3.2. Multilevel analyses3.2.1. Moderators

3.3. Nonlinear analyses3.3.1. Average noise levels3.3.2. Noise variability3.3.3. Nonlinear model with covariates3.3.4. Nonlinear principal component analysis for sources of noise

4. Discussion4.1. Conclusions4.1.1. Child, caregiver, and care characteristics

4.2. Implications for research and practice

Author noteAppendix A. Coding system sources of noiseAppendix B. Procedures in nonlinear PCAAppendix C. Moderator analysis of cross-level interactionsReferences