Early Childhood Research Quarterly 42 (2018) 79–92

Contents lists available at ScienceDirect

Early Childhood Research Quarterly

journal homepage: www.elsevier.com/locate/ecresq

Linguistic environment of preschool classrooms: What dimensions support MARK

children’s language growth?
⁎
Laura M. Justicea, Hui Jianga, , Katherine Strasserb
a
The Ohio State University, United States
b
Pontificia Universidad Católica de Chile, Chile

A R T I C L E I N F O A B S T R A C T

Keywords: Individual differences in young children’s language acquisition reflect in part the variability in the language-
Preschool learning environment that they experience, both at home and in the classroom. Studies have examined various
Oral language dimensions of the preschool classroom language environment, including linguistic responsivity of early child-
Classroom quality hood educators, data-providing features of teachers’ talk, and characteristics of the systems-level general en-
Vocabulary growth
vironment, but no study has examined the unique contribution of each dimension to children’s language growth
over time. The goals of this study were to determine how best to represent the dimensionality of the preschool
classroom’s linguistic environment and to determine which dimensions are most strongly associated with chil-
dren’s language development. Participants were teachers in 49 preschool classrooms and a random sample of
children from each classroom (330 children between 40 and 60 months of age, M = 52 months, SD = 5.5).
Children’s grammar and vocabulary skills were measured at three time-points, and the classroom linguistic
environment was assessed with measures representing teachers’ linguistic responsivity, data-providing features
of teachers’ talk, and systems-level general quality. Using exploratory structural equation modeling (ESEM), we
determined that the classroom language environment is best characterized by a three-dimensional model. A
multilevel latent growth model subsequently showed that only one of the three dimensions, teachers’ commu-
nication-facilitating behaviors, predicted growth in children’s vocabulary from preschool to kindergarten.
Implications for teacher professional development are discussed.

1. Introduction adolescence (see Hayiou-Thomas et al., 2012). Some speculate that this
shift reflects the decreasing variability of children’s environments as
There are many universalities in young children’s language acqui- they become older and progress through the curricula of formal
sition, such as when children tend to speak their first word as well as schooling (Hayiou-Thomas et al., 2012). That is, prior to formal
when they start to use grammatical morphemes, yet there are also schooling, there is considerable variability in the language-learning
considerable individual differences, even within a relatively con- environment that children experience, and this variability appears to
strained cultural or linguistic context (Rowe, Raudenbush, & Goldin- have significant implications for children’s early linguistic trajectories.
Meadow, 2012). These individual differences reflect both the influence
of nature on children’s language development (i.e., one’s genetics or 1.1. Children’s learning–learning experiences in early-education settings
biology) as well as the influence of nurture, representing the child’s
environment (Hayiou-Thomas, Dale, & Plomin, 2012). Interestingly, Over the last several decades, numerous studies have sought to
longitudinal twin studies show that from birth to about age five, the document variability in young children’s language-learning environ-
dominant influence on language growth is the child’s environment, ments, including studies of both the home and preschool settings
accounting for about 60–70% of the variance in language skills, as (Dickinson & Smith, 1994; Girolametto & Weitzman, 2002; Justice,
compared to about 25% for genetic factors (the unaccounted-for re- McGinty, Zucker, Cabell, & Piasta, 2013; Landry, Smith, & Swank, 2006;
mainder reflects non-shared environmental factors and error) (Spinath, Yoder & Warren, 1999). This work has examined various dimensions of
Price, Dale, & Plomin, 2004). Over time, the dominant role of the en- children’s language-learning environments, often for the purpose of
vironment for influencing children’s language skills tends to diminish, understanding the relations between a specific dimension of the en-
with the contribution of genetics becoming increasingly prominent into vironment and children’s language development. Dimensions often

⁎
Corresponding author.
E-mail address: jiang.200@osu.edu (H. Jiang).

http://dx.doi.org/10.1016/j.ecresq.2017.09.003
Received 18 November 2016; Received in revised form 7 June 2017; Accepted 5 September 2017
0885-2006/ © 2017 Published by Elsevier Inc.
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

investigated within the early-education setting include (1) linguistic communication-facilitating behaviors, are specific behaviors that adults
responsivity of early childhood educators, (2) data-providing features use to create and sustain children’s participation in multi-turn con-
of teachers’ talk, and (3) systems-level general environment; each will versations and time spent in joint engagement. When engaging in ex-
be reviewed in turn. Although research to date suggests that each di- tended periods of conversation with young children, adults often have
mension is associated with children’s language growth, a key limitation to take an active role in maintaining the interaction, such as looking at
of this literature is that no study has examined the distinctiveness of the child expectantly to encourage him to contribute or to ask open-
these dimensions nor the unique contribution of each dimension to ended questions to cue a conversational turn (Adamson, Bakeman,
children’s language growth over time. The primary goal of the present & Deckner, 2004). In preschool settings, the frequency with which
study is to address these limitations by identifying what specific di- teachers are observed to use these strategies is associated positively
mension(s) of the preschool language-learning environment is instru- with the complexity of children’s talk during interactions with their
mental to advancing children’s language growth. In addressing this teachers (Girolametto & Weitzman, 2002) as well as children’s voca-
goal, we focus specifically on early childhood classrooms serving chil- bulary growth during an academic year (Cabell et al., 2015). The latter,
dren from low-income homes. Because children from low-income language-developing behaviors, represent responsivity behaviors that
homes often exhibit significant gaps in their language skills relative to serve to model for children advanced forms of language. Perhaps the
more advantaged peers, the preschool classroom environment can be most well-studied language-developing behaviors are recasts and ex-
especially important for supporting their language acquisition pansions, in which an adult responds to a child’s utterance with a more
(LoCasale-Crouch et al., 2007). Thus, the results of this work may be syntactically (recast) or semantically complex (expansion) form (Fey,
extremely relevant to settings that serve children from disadvantaged Cleave, Long, & Hughes, 1993).
backgrounds. Experts offer several reasons why linguistic responsivity serves to
There are significant practical implications for improving our un- stimulate language growth among young children. First, with respect to
derstanding of the language-learning environments of early-education communication-facilitating behaviors, experts suggest that these re-
settings, and whether there are specific malleable dimensions of the sponsive behaviors allow the child to maintain rather than shift her
environment that seem especially influential to children’s language current attentional focus, thus maximizing the allocation of cognitive
growth. For instance, some studies have suggested that teachers’ use of resources towards the child’s current attentional allocation (Landry
questions and comments when interacting with children are especially et al., 2006). Additionally, these responsive behaviors enhance the
important (Girolametto & Weitzman, 2002). Professional-development child’s understanding of the intentional nature of communicating (i.e.,
offerings, teacher-education coursework, and published curricula draw that talking to another recruits their interest and engagement), which
upon such findings to provide evidence-based guidance to educators serves to motivate the child to talk more often (Yoder & Warren, 1999).
who work in early-education settings. For instance, one professional Second, with respect to language-developing behaviors, especially re-
development program offered to early educators heavily emphasized casts and expansions in which adults extend children’s utterances with
the use of teacher questions as a way to improve children’s language in syntactically or semantically complex forms, experts argue that these
the classroom (Powell, Diamond, Burchinal, & Koehler, 2010). Al- provide children with a direct contrast between the child’s form and the
though trained teachers improved in their use of this language-facil- adult’s more complex form (Proctor-Williams & Fey, 2007). Because the
itating strategy, it appeared to have little benefit to children’s language adult’s extension maintained the child’s referential focus, thus limiting
growth over time (Powell et al., 2010). The present study is likely to working memory demands, the child’s attentional resources can focus
have direct bearing on the types of strategies we encourage teachers to on processing distinctions in the adult’s form. Such theories are im-
use in their classrooms, by pin-pointing those dimensions of the class- portant for interpreting the considerable empirical evidence indicating
room environment that positively affect children’s language growth. that adult use of communication-facilitating and language-developing
strategies are beneficial to young children’s language development
1.2. Frequently examined dimensions of early-education classrooms within various early caregiving settings, including both home and
preschool settings.
1.2.1. Linguistic responsivity
The first dimension of the early-education environment examined in 1.2.2. Data-providing features of teacher talk
this study was caregivers’ linguistic responsiveness. From a language- The second dimension of children’s language-learning environment
facilitation perspective, linguistic responsivity is observed when adults examined in this study is the “data-providing features” of adults’ talk
are sensitive to and reflective of the child’s interests and/or utterances when interacting with children (Hoff, 2003; Hoff & Naigles, 2002;
during conversations, often referred to as “following the child’s lead” Huttenlocher, Vasilyeva, Cymerman, & Levine, 2002). Data-providing
(Girolametto & Weitzman, 2002). For instance, when looking at a book features of input are relatively granular aspects of adult talk that pro-
together, an adult can be linguistically responsive by expanding the vide children with crucial information about linguistic forms and
child’s utterance (child: “bug”, mother: “it’s a big bug!”) rather than functions. From a very young age, children employ biologically en-
diverting the child’s focus to something else in the book (child: “bug”, dowed computational processes to the input to which they are exposed,
mother: “look at this mouse!”); such behaviors are referred to variously often referred to as statistical learning (Marcus, Vijayan, Rao, &
as expanding, extending, recasting, and contingent responding (Landry Vishton, 1999; Saffran, 2003; Saffran & Wilson, 2003). These processes
et al., 2006; Landry, Smith, Swank, Assel, & Vellet, 2001; Tamis- allow children to extract information from their environment to acquire
LeMonda, Bornstein, & Baumwell, 2001). Importantly, the frequency a seemingly infinite range of linguistic forms and functions that are
with which mothers and teachers use linguistically responsive beha- never directly taught to them, such as marking verbs for tense and
viors during interactions has been linked to children’s language growth marking nouns for plurality.
over time (Cabell, Justice, McGinty, DeCoster, & Forston, 2015; Landry Crucial for such processes is that the environment provide a suffi-
et al., 2006; Nicely, Tamis-LeMonda, & Bornstein, 1999; Yoder & cient corpus of data (i.e., input) for the child to analyze, including a
Warren, 1999). sufficient number of different word types and syntactic forms
Adults’ responsivity behaviors can be differentiated into those be- (Hoff & Naigles, 2002). Indeed, variability in these data-providing fea-
haviors that serve to promote children’s engagement in communication tures of adult talk within the home environment is associated with
routines, referred to as communication-facilitating behaviors, and those young children’s development of both vocabulary and grammar. For
that seek to provide advanced linguistic models, referred to as lan- instance, Hoff (2003) showed that the number of different words mo-
guage-developing behaviors (Girolametto, Pearce, & Weitzman, 1996; thers used when talking with their 2-year-olds was positively associated
Girolametto & Weitzman, 2002; Piasta et al., 2012). The former, with children’s vocabulary skills (r = 0.22), whereas Huttenlocher

80
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

et al. (2002) found that the percentage of syntactically complex sen- questions not only about the consistency of such effects but also the size
tences parents produced when talking with their 4-year-olds was posi- of the relations linking general estimates of quality to children’s lan-
tively associated with children’s grammar skills (r = 0.39). Within the guage outcomes. Such work shows that the relation between CLASS
early education classroom, studies find that data-providing features of scores and children’s language development are so small that they have
teachers’ talk are associated with children’s growth in complex syntax, little to no practical value (Keys et al., 2013). Current research there-
such as comprehension of syntactically complex sentences fore suggests that this third dimension of the language-learning en-
(Huttenlocher et al., 2002). A recent study sought to explore the me- vironment, namely a systems-level general index of classroom quality,
chanisms by which the data-providing features of teachers’ talk, espe- may insufficiently relate to children’s language growth over time.
cially teachers’ expression of syntactically complex forms, may serve to
accelerate children’s language development in the preschool classroom 1.3. Purpose of this study
(Justice et al., 2013). This work found significant, strong contingencies
between teacher and child talk during small-group interactions, such The present study was designed to substantially increase our un-
that when teachers produced a syntactically complex utterance, chil- derstanding of preschool classrooms as language-facilitating contexts
dren often followed with complex utterances of their own. These con- for young children, with an interest in (1) exploring how best to re-
tingencies may provide the mechanism through which children inter- present the dimensionality of the classroom linguistic environment that
nalize more syntactically advanced forms. preschool-aged children experience, and (2) determine those dimension
(s) that are most strongly associated with children’s language-devel-
1.2.3. System-level general quality of teacher–child interactions opment trajectories. For our purposes, the three theorized dimensions
The third dimension of children’s early language-learning environ- of the preschool classroom’s linguistic environment were teachers’
ment examined in this study is the systems-level, general quality of linguistic responsivity (communication-facilitating behaviors and lan-
teacher-child interactions (Hamre, Hatfield, Pianta, & Jamil, 2014). guage-developing behaviors), data-providing features of teachers’ talk,
This dimension has largely surfaced as the result of a measurement and systems-level general classroom quality. Two questions were ad-
model designed to examine the general quality of children’s early dressed. First, is the preschool classroom linguistic environment best
education settings (Hamre et al., 2014). The need for such a measure- represented as one dimension (uni-dimensional), two dimensions (bi-
ment model is based on the increasingly large number of preschool- dimensional) or three dimensions (tri-dimensional)? Second, of the
aged children enrolled in center-based care, often featuring an adult- identified dimension(s), which is most strongly associated with chil-
child ratio of about 1 adult to 10 children. As preschool participation dren’s language growth over time?
has grown, researchers and policy-makers have sought tools that dif-
ferentiate lower- from higher-quality programs, especially in reference 2. Method
to children’s outcomes, including language development.
Measurement models that seek to represent the systems-level, gen- 2.1. Participants
eral quality of the classroom environment appropriate the framework
that teacher-child linguistic interactions are a key proximal process This study involved teachers and children in 49 preschool class-
through which children’s language skills are developed in preschool rooms from a single mid-Atlantic state, all of which prioritized enroll-
settings (Hamre et al., 2014; NICHD Early Child Care Research ment for children from low-income settings; 37 of the classrooms were
Network, 2002). However, these measurement models do not serve to affiliated with Head Start, and 12 were sponsored by the state pre-
precisely catalog the fine-grained, granular proximal processes of tea- kindergarten initiative. The classrooms were enrolled in a larger ex-
cher-child interactions, but rather index the overall quality of the perimental study involving evaluation of the effects of professional
classroom environment. Derived initially from the work of the NICHD development (PD) on children’s language and literacy skills (Cabell
Early Child Care Research Network (2002), and then second-generation et al., 2011). Classrooms were randomly assigned to a professional
tools derived from this work, especially the Classroom Assessment development (PD) group or a placebo control group. In the former,
Scoring System (CLASS; Pianta, Karen, Paro, & Hamre, 2008), there is teachers received training on how to use specific strategies to facilitate
currently a heavy reliance on using systems-level indices to represent children’s language skills, whereas in the latter teachers received
the general quality of the language-learning environment available training focused on creative ways to use the block center in their
within early education settings (Hamre et al., 2014; Mashburn et al., classrooms. This assignment should not have affected the dimension-
2008). Application of the CLASS in early education settings serves to ality of the classroom environment nor the longitudinal relations be-
provide a general estimate of the linguistic environment based on the tween these dimensions and children’s language growth. Further, the
Language Modeling subscale of the tool or the composite of three PD had only modest effects on teachers’ practices (Piasta et al., 2012)
subscales which collectively capture the level of Instructional Support and little discernable effects on children’s language growth (Cabell
in a classroom. For both the subscale and the composite, classrooms et al., 2011). Therefore, for the purpose of the present study, we include
receive an overall score on a basic 1- to 7-point scale to delineate the all 49 classrooms irrespective of study condition.
overall level of the linguistic environment made available to children, The lead teachers in these classrooms were primarily female (96%),
and classrooms can be differentiated into those that provide low, and 67% were Caucasian. Of the remainder, 22% were Black, 2% were
average, and high levels of quality (Burchinal, Vandergrift, Native American, and 4% were multiracial (4% unreported). The ma-
Pianta, & Mashburn, 2010). jority of teachers (92%) had a two-year degree or higher (37%
In recent years, the use of a general index to document classroom Associate’s, 33% Bachelor’s, 22% Master’s), although 4% held only a
quality, including the quality of the language-learning environment, high school diploma (4% unreported). On average, teachers had about
has gained considerable appeal as a professional development tool for 12 years of teaching experience (M = 11.8 years, SD = 7.4). The ma-
teachers (Hamre et al., 2010), a component of quality rating systems in jority of teachers (n = 44) reported using The Creative Curriculum for
the child-care system (Sabol & Pianta, 2015), and even for determining Preschool (Dodge, Colker, & Heroman, 2002) as their primary classroom
whether early education programs should continue to receive federal curriculum.
funding (Mashburn, 2016). Early evidence suggested that general sys- For the purposes of the larger study, a random sample of children
tems-level scores of the preschool classroom were correlated with within each classroom was selected to participate in ongoing study
children’s language growth, suggesting the validity of this approach to assessments during the fall and spring of the preschool year and then
representing the quality of children’s language-learning environment one-year later, when most children were in kindergarten. To randomly
(Mashburn et al., 2008). However, recent meta-analytic work has raised select these children, caregiver consent was solicited for all children in

81
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 1
Variables used in main analyses with time-points.

Y1 fall Y1 winter Y1 spring Y2 spring

Children’ Language Skills

Grammar CELF-Sentence Structure (SS) x x x
CELF-Word Structure (WS) x x x
Vocabulary CELF-Expressive Vocabulary (EV) x x x
PPVT-3 (Receptive Vocabulary) x x x

Preschool Classroom Linguistic Environment

Teachers’ Linguistic Responsivity Communication strategies (CS) x x x
Language developing strategies (LDS) x x x
Data-Providing Features of Teachers’ Talk Mean length of utterance (MLU) x x x
Number of different words (NDW) x x x
Percentage of complex sentences (Comp) x x x
Noun phrase (NP) x x x
Systems-Level General Environment Concept development (CD) x x
Quality of feedback (QF) x x
Language modeling (LM) x x

each classroom who met an age eligibility criterion of 3 years, 4 months 2.3. Measures
by October of the study year. From among those children for whom
consent was received, five to eight children per classroom were ran- The primary measures utilized in the current study included as-
domly selected. The exact number of children selected per classroom sessment of children’s language skills and of the language environment
was based on the number of consents received; more children were of children’s preschool classrooms, representing three theorized di-
selected from classrooms with higher consent rates. mensions: teachers’ linguistic responsivity, data-providing features of
The final sample of 330 children was well-balanced with respect to teachers’ talk, and systems-level general quality. An overview of these
gender (174 males and 156 females) and relatively diverse (45% study measures, including a timeline for implementation, appears in
Caucasian, 33% African American, 5% Hispanic/Latino, with 16% Table 1. For each measure of language environment, average scores
multiracial, other, or unreported). The children were between 40 and across all available time points were used in the analyses.
60 months of age (M = 52 months, SD = 5.5) at the start of the study.
Noted previously, all children resided in low-income households based
2.3.1. Children’s language skills
on eligibility for participating in the preschool programs. Data provided
Two standardized measures were used to assess children’s language
by primary caregivers (79% reporting) indicated that the majority of
skills in the domains of morphosyntax and vocabulary: the Clinical
children (56%) lived in households with an annual income of less than
Evaluation of Language Fundamentals—Preschool (CELF; Wiig,
$25,000 (55.5%), and the majority of mothers had limited educational
Secord, & Semel, 2004) and the Peabody Picture Vocabulary Test
attainment (70% did not complete high school or high school was
(PPVT; D. Dunn & Dunn, 2007). Children were administered these
highest degree earned). About 16% of children had identified dis-
measures in one-on-one sessions with project staff in a quiet location in
abilities.
their program buildings. Prior to working in the field, project staff
completed a multi-pronged training program to ensure the quality of
2.2. Procedure their implementation of the measures.
The three core subtests of the CELF P-2 were administered: Sentence
Classrooms and teachers were recruited into the study in the spring Structure, Word Structure, and Expressive Vocabulary. For Sentence
or summer prior to the academic year. Teachers agreed to participate in Structure, children are provided with increasingly complex sentences
the study for a one-year period, with study activities including random and are asked to identify on a test plate which of four pictures best
assignment to a PD condition or control and ongoing observations in illustrates the given sentence. The subtest is designed to determine
their classrooms, including videotaping. Children were enrolled into the children’s knowledge of increasingly complex syntactic structures. For
study in the fall of the year, and participated in three assessment bat- Word Structure, the examiner uses a cloze procedure to elicit target
teries over an approximately 20-month period: fall of preschool, spring morphemes from children (e.g., the plural marker); the subtest is de-
of preschool, and spring of the following year. signed to assess children’s knowledge of inflectional and derivational
The experimental procedures of the larger study involved provision morphology. For Expressive Vocabulary, children are shown a set of
of PD to teachers in a treatment group (n = 25) or a placebo control of illustrations on test plates and are asked to provide the name of the item
PD on neutral topics for those in the control group (n = 24). In total, depicted; the subtest is designed to examine single-word, expressive
teachers received about 17 h of PD via a fall workshop (13 h) and a vocabulary breadth. For all three subtests of the CELF P-2, test devel-
winter refresher (4 h). Those in the treatment group received training opers report test-retest reliability coefficients of 0.78–0.90 and internal
on how to use specific strategies in their classrooms to facilitate chil- consistency (i.e., Cronbach’s alpha and split-half reliability) ranging
dren’s language skills; the PD is detailed extensively under separate from 0.78 to 0.87 (Wiig et al., 2004). As an additional measure of vo-
cover (see Piasta et al., 2012). Teachers in both groups were required to cabulary, the PPVT-III was also administered. In this assessment, chil-
videotape themselves interacting with small groups of children in their dren choose the illustration of four alternatives that best matches a
classrooms every two weeks and submit these via stamped, addressed target word spoken by the examiner. The measure is designed to assess
mailers to project staff. The videos were to be approximately 30 min in single-word receptive vocabulary breadth. Test developers report an
duration, and teachers were given instructions in how to capture these internal consistency of 0.95 (Cronbach’s alpha) and 0.94 (split-half
(e.g., how to set up the camera, target group size of children) and all reliability) as well as test-retest reliability for different age samples
materials needed for implementation (i.e., camera, tripod, recording ranging from 0.91 to 0.94 (Dunn & Dunn, 1997). Table 2 displays the
media, mailers). Teachers were permitted to keep the camera and descriptive statistics of each individual measure at the three time-
tripod following the study, thus providing an incentive for study par- points.
ticipation and collecting/submitting the videos to project staff. For the purpose of this study, a composite score of grammar was

82
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 2
Descriptive data for language measures at three time-points.

Measure Y1 Fall Y1 Spring Y2 Spring

M SD Range M SD Range M SD Range

CELF Sentence Structure 10.31 4.82 0 to 22 13.57 4.83 0 to 22 16.89 3.97 1 to 22

CELF Word Structure 9.62 5.14 0 to 21 12.87 5.55 0 to 23 16.61 4.50 3 to 24
CELF Expressive Vocabulary 14.03 7.93 0 to 34 19.00 8.60 0 to 37 25.84 7.17 6 to 40
PPVT Receptive Vocabulary 42.07 19.12 1 to 86 54.31 19.98 3 to 100 72.53 18.47 10 to 136
Grammar scaled composite 0.00 1.00 −2.20 to 2.36 0.73 1.07 −2.09 to 2.68 1.51 0.83 −1.54 to 2.89
Vocabulary scaled composite 0.00 1.00 −2.05 to 2.16 0.65 0.74 −2.14 to 2.81 1.65 0.91 −1.40 to 4.07

Note: CELF = Clinical Evaluation of Language Fundamentals (Wiig et al., 2004); PPVT = Peabody Picture Vocabulary Test-III (Dunn & Dunn, 1997)

created by scaling and averaging scores of CELF Sentence Structure and was conducted by research staff who were trained to 90% accuracy of
Word Structure (r = 0.629, 0.709, and 0.558 at each time point). coded intervals on master-coded video segments. A randomly selected
Similarly, a composite vocabulary score was computed based on CELF 10% of videos were double-coded, and exact agreement across the
Expressive Vocabulary and PPVT-III (r = 0.800, 0.734, and 0.722 at double-coded sessions was 88%. Scores per behavior were averaged
each time point). We used these composite scores to represent two key across intervals, to account for intervals that could not be coded due to
elements of language skills, namely grammar and vocabulary, in in- poor video or audio quality.
vestigating the relationship between language environment and chil- As part of the scale adaption, we re-examined the reliability and
dren’s language development. validity of the nine items as coded in the present study. In doing so, we
dropped two items that had low correlation with all the other items
(item 3, 6), and validated the factor structure for the remaining seven
2.3.2. Teachers’ linguistic responsivity
items using factor analyses. The seven behaviors represent two sets of
Teachers’ linguistic responsivity was assessed using an interval-
strategies, which we refer to as communication-facilitating and lan-
based coding scheme designed to document teachers’ use of nine re-
guage-developing strategies, consistent with prior descriptions of the
sponsivity behaviors derived from the Teacher Interaction and Language
TILRS items (see Piasta et al., 2012). Communication-facilitating strate-
Rating Scale (TILRS; M. Girolametto, Weitzman, & Greenberg, 2000).
gies are those that teachers use to encourage and maintain classroom
The TILRS was developed by its authors as a tool to assess teachers’ and
conversation (e.g., teacher uses open-ended questions to stimulate
parents’ conversational responsivity when interacting with toddlers and
conversation; items 1, 2, 4, 5), whereas language-developing strategies are
preschoolers (see Girolametto et al., 2000). A global rating scale in
those that teachers use to teach children new language forms or func-
which observers rate adults’ use of responsivity strategies on a scale of 1
tions, primarily by commenting and expanding (e.g., teacher expands
to 7 (e.g., “Wait and listen,” 1 = almost never, 7 = consistently), the
child utterance; items 5, 7, 8, 9). Note that item 5 (facilitating peer-to-
primary use of this tool was to help coaches work with teachers and
peer communication) loaded on both factors. Scale validation in-
parents to improve their use of these strategies when interacting with
formation is provided in Table 3. For our purposes, the extracted factor
their children. For the purposes of the larger study (see Piasta et al.,
scores of communication-facilitating and language-developing strate-
2012), the items in the TILRS (e.g., Wait and listen; Use a variety of
gies were used in analyses.
questions.) were more explicitly codified and adapted to an interval-
coding approach. This allowed coding of teachers’ responsive behaviors
to occur with greater measurement precision as well as a high level of 2.3.3. Data-providing features of teachers’ talk
inter-rater reliability. The same three videos used to examine teachers’ responsivity were
In total, teachers’ use of nine responsivity behaviors were coded for also coded to examine the data-providing features of teachers’ talk with
three videos submitted by teachers during week 2, week 14, and week children. As noted previously, the three videos coded corresponded to
24 of the project and thereby spanning the academic year (three tea- those taken at the beginning (week 2), middle (week 14), and end
chers failed to submit the week 24 video; therefore, we substituted their (week 22 or 24) of the year. Coding of the data-providing features of
previously submitted video from week 22). During these videos, tea- teachers’ talk involved first developing a transcript of the small-group
chers were interacting with a small group of up to six children in their interaction captured in each video. To develop this transcript, a 10-min
classroom using materials provided by the project (play-doh and related segment was selected from the mid-point of each video for transcrip-
manipulatives). Although each video lasted up to 30 min in duration, tion. The segment was slightly longer than that used to capture tea-
we coded a medial randomly selected segment per video segmented chers’ conversational strategies so to ensure an optimal corpus of
into 15 30-s intervals. Each interval was coded for teachers’ use of each utterances and words to analyze, given that some strategies involve
of nine responsive behaviors: (1) looking expectantly and being warm teachers talking less rather than more (e.g., looking expectantly to
and receptive to encourage interaction; (2) using a slow pace of con- encourage a child to talk). Other researchers have used 10-min video
versation to allow children to participate; (3) using comments to cue segments as a means to assess preschool teachers’ conversational stra-
children to take turns; (4) using open-ended questions to stimulate tegies (Girolametto & Weitzman, 2002), finding that these provide
conversation; (5) facilitating peer-to-peer communication; (6) stressing adequate information about teachers’ talk to examine its relations to
and/or repeating words to make them salient; (7) repeating and/or children’ language skills. Any videos that were less than 10 min in
expanding children’s utterances; (8) using comments and questions to length (44%) were coded in their entirety.
expand the discussion of certain objects or topics; and (9) using com- Transcription was conducted using the Systematic Analysis of
ments and questions to talk about feelings, to pretend, and/or to talk Language Transcripts software for Windows (SALT; Version 9). Parsing
about the past or future. Each of these nine behaviors represented a occurred at the level of the T-unit and adhered to SALT conventions for
distinct item on the coding scheme and was carefully defined in a marking morphemes, addressing fillers and mazes, and accounting for
coding catalog. unintelligible utterances and interruptions. Each T-unit, consisting of
To conduct coding, the selected interval was divided into 15 30-s no more than two independent clauses with dependent clauses or at-
intervals and each was scored for the presence (1) or absence (0) of the tached phrases, was entered as a separate line in the SALT software.
9 behaviors; an interval could be coded for multiple strategies. Coding Transcription was conducted by research assistants who had completed

83
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 3
Validation of the adapted Teacher Interaction and Language Rating (TILR) Scale.

Communication Facilitation Language Developing

Standardized Item Loadings 1. looking expectantly and being warm and receptive to encourage interaction 0.797***
2. using a slow pace of conversation to allow children to participate 0.649***
3. using comments to cue children to take turns (dropped)
4. using open-ended questions to stimulate conversation 0.431**
5. facilitating peer-to-peer communication 0.332* 0.564***
6. stressing and/or repeating words to make them salient (dropped)
7. repeating and/or expanding children’s utterances 0.685***
8. using comments and questions to expand the discussion of certain objects or topics 0.321*
9. using comments and questions to talk about feelings 0.378*
CFA Model Fit Chi-square test < βρ > χ2 = 10.185, df = 12, p = 0.600
RMSEA 0.000, 90% CI = (0.000, 0.127)
CFI 1.000
SRMR 0.064

*p < 0.05; ** p < 0.01; *** p < 0.001.

a multi-step tutorial involving completion of practice transcripts and a CLASS (Pianta et al., 2008). The CLASS is an observational tool de-
reliability test. The reliability test involved achieving 90% agreement signed to capture general qualitative features of the classroom en-
against four master-coded transcripts, with agreement calculated se- vironment, with a focus on the quality of the interactions among tea-
parately for accuracy of (1) T-unit segmentation, (2) word-for-word chers and children. The CLASS is often viewed as an index of general
transcription, and (3) SALT conventions and codes. Additionally, a classroom quality (Mashburn et al., 2008). The tool is organized to
subset of videos (10%) was independently transcribed to assess inter- capture three domains of classroom quality (Emotional Support, In-
rater reliability among the trained coders. We calculated the correla- structional Support, Classroom Organization), each of which is re-
tions between the doubly-transcribed sessions to examine relations for presented by three or four more specific dimensions; in total, 10 do-
all SALT standard measures, including total number of utterances, mean mains are examined (e.g., Positive Climate, Teacher Sensitivity). To
length of utterance in words and morphemes, type-token ratio, number administer the CLASS, observers rate each of the 10 dimensions on a
of different words, and number of bound morphemes; correlations holistic 7-point coding scheme for which 1 = low and 7 = high.
ranged from 0.89 to 0.99. For the current study, field assessors conducted two-hour video-
When transcribing, coders inserted codes into each teacher T-unit to taped classroom observations in the fall and spring of the preschool
capture the complexity of each T-unit in terms of its clausal complexity year. To select the optimal time for the observation, project staff
and noun phrase density, using a scheme adapted from Huttenlocher worked with teachers to identify an observational period that re-
et al. (2002). To code the clausal complexity of each T-unit, each was presented a customary day and during which a range of typical class-
assigned a mutually exclusive code to document whether the T-unit room routines could be observed. To conduct coding of the ten di-
contained no clauses (NO; e.g., “that boy”), one clause (SIMPLE; e.g., mensions, three 20-min cycles were randomly chosen from each
“the boy is coming”), or multiple clauses (COMPLEX; e.g., “the ball fell videotaped observation to be coded by trained and reliable CLASS co-
and I’ll pick it up”). The coding system used, with additional examples, ders. The codes for each cycle are then averaged for each of the ten
is described in prior reports (Justice et al., 2013). In addition, coders dimensions to arrive at a score for each classroom at each time-point.
also inserted codes into each T-unit to document each instance of a To ensure the reliability of the CLASS coding, 20% of the cycles were
noun phrase (NP). The coding of NP was based on Huttenlocher et al. randomly selected for double-coding by two independent observers,
(2002), which served to identify the number of NPs per T-unit in ma- and the intraclass correlations for double-coded video segments were
ternal speech. For instance, the T-unit “Juan, you need to put that down examined using a two-way mixed model across all project time-points
and help Xavier with his truck” would be coded to contain five noun and cohorts; all ICCs exceeded 0.7 for all domains.
phrases based on the scheme presented in Huttenlocher et al. [“Juan For the present study, we used the three dimensions of the CLASS
(NP), you (NP) need to put that (NP) down and help Xavier (NP) with that comprise the Instructional Support domain (Concept Development,
his truck (NP)”]. Quality of Feedback, Language Modeling), as these are designed to
The SALT software was used to automatically generate four mea- capture systems-level, general interactions supportive of children’s
sures of teachers’ data-providing features. The first two measures are language growth (Justice, Mashburn, Hamre, & Pianta, 2008), and prior
standard calculations computed by the software: mean length of work has shown that the Instructional Support domain is positively
utterances in morphemes (MLU) and total number of different words associated with children’s language growth in pre-kindergarten settings
(NDW). MLU represents the average number of morphemes per T-Unit (Mashburn et al., 2008). The Concept Development dimension is de-
for teachers. NDW represents the total number of different word roots signed to capture the extent of instructional discussions and activities
produced. In addition, using the COMPLEX and NP codes inserted into that promote children’s higher-order language and cognition skills,
each T-unit, we calculated a measure of T-unit complexity (COMP) by such as opportunities to compare/contrast, reason, and problem solve.
dividing the number of T-units with the COMPLEX code by the total (All descriptions of dimensions are derived from the CLASS coding
number of utterances per transcribed observation. The COMP index per manual; Pianta et al., 2008). The Quality of Feedback dimension cap-
each transcript indicates the percentage of teacher T-units that were tures the extent to which teachers provide quality feedback to children
complex (i.e., contained multiple clauses). In addition, we calculated a that serves to expand their learning, including the extent to which
measure of noun phrase (NP) by counting the total number of NPs teachers and children engage in back-and-forth exchanges and teachers
coded per observation. In total, four data-providing features of teachers’ provide responsive feedback to children. The Language Modeling di-
talk were used in analyses: MLU, NDW, COMP, and NP. mension captures the extent to which teachers provide children with a
language-stimulating environment, including the extent to which tea-
chers ask children open-ended questions, engage in multi-turn con-
2.3.4. Systems-level general environment
versations, and repeat and extend children’s conversational contribu-
The systems-level general environment was measured at two time-
tions.
points (fall, spring) via direct observation of each classroom using the

84
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 4
Descriptive data for the linguistic environment measures.

Construct Specific Index Y1 fall Y1 winter Y1 spring Y1 mean

M (SD) M (SD) M (SD) M (SD)

Data-Providing Features of Teachers’ Talk Mean length of utterance (MLU) 5.85 (0.98) 6.20 (0.91) 6.20 (1.11) 6.05 (0.83)
Number of different words (NDW) 186.02 (39.58) 188.84 (38.74) 187.71 (45.21) 186.34 (29.62)
Percent complex sentences (COMP) 0.28 (0.11) 0.31 (0.09) 0.29 (0.09) 0.30 (0.13)
Noun phrase (NP) 62.93 (33.05) 71.21 (27.50) 63.24 (26.66) 63.93 (19.64)
Systems-Level General Environment Concept development (CD) 2.91 (1.66) 2.40 (1.48) 2.66 (1.25)
Quality of feedback (QF) 3.00 (1.31) 3.54 (1.66) 3.27 (1.30)
Language modeling (LM) 3.29 (1.53) 3.54 (1.46) 3.43 (1.22)

For the present study, ratings were averaged across the two time all available information in each step of the analyses (Enders, 2010).
points for each dimension. Mean scores for each time-point are shown The final valid sample sizes were 49 classrooms and 320 children. All
in Table 4. analyses were carried out using Mplus 7.11 (Muthén & Muthén, 2005).

2.4. Analyses 3. Results

In order to examine the underlying dimension(s) of preschool 3.1. Exploring dimensions of the classroom linguistic environment
classrooms’ linguistic environment, we employed exploratory structural
equation modeling (ESEM) (Asparouhov & Muthén, 2009). Unlike con- Pearson correlation coefficients between the nine observed in-
ventional confirmatory factor analysis (CFA) that aims to test a parsi- dicators of the classroom linguistic environment are displayed in
monious measurement model based on strong a priori substantive Table 5. Using these indicators, we sequentially tested three ESEM
knowledge, ESEM allows for testing of a richer set of alternative, less models, representing uni-dimensional, bi-dimensional, and tri-dimen-
restrictive models. While being exploratory in nature, ESEM also gives sional structures respectively (Fig. 1; fit indices in Table 6). By a priori
access to all the CFA features (e.g., correlated residuals, model fit in- hypothesis, we co-varied MLU and percentage of complex sentences in
dices, parameter constraints), offering a great amount of flexibility in every model, as these two variables representing data-providing fea-
the situation of model uncertainty. Considering that the dimensionality tures of teachers’ talk both captured attributes of an average utterance
of the preschool classroom linguistic environment is yet unknown, we and were computed using the same denominator.
employed ESEM with oblique Geomin rotation (Asparouhov & Muthén, Results revealed that the preschool classroom language environ-
2009) as a preferable exploratory approach over confirmatory analyses ment is best represented multi-dimensionally, as the uni-dimensional
(Browne, 2001). model did not fit the data well (χ2 = 89.64, p < 0.001;
Following ESEM, we explored the predictive relationship between RMSEA = 0.211; CFI = 0.584; SRMR = 0.140). On the other hand,
the dimensions of the classroom linguistic environment and children’s both the bi-dimensional and tri-dimensional models provided reason-
language skills using multilevel latent growth model (Meredith & Tisak, able model fit (p-values of χ2 tests > 0.05; RMSEA < 0.08;
1990). Trajectories of children’s grammar and vocabulary development CFI > 0.950; SRMR < 0.08) and offer significant improvement in fit
from preschool to kindergarten were modeled using latent growth as compared to the uni-dimensional model. The tri-dimensional model
curves, the slope of which was predicted by classroom-level factor further boosted model fit compared to the bi-dimensional model
scores extracted from the final linguistic environment dimensionality (Δχ2 = 11.99, p = 0.101; ΔRMSEA = −0.072; ΔCFI = 0.033;
model. While the preschool classroom effects are modeled at level-2, we ΔSRMR = −0.024; lower ABIC value). Interestingly, the tri-dimen-
did not include the clustering of kindergarten classrooms due to a high sional model indicated a similar structure as what was revealed in the
percentage of missing data (30%) of kindergarten classroom affiliation. competing bi-dimensional model. Specifically, the first dimension is
This missing data is a result of the primary data collection occurring dominated by measures of teachers’ data-providing features (MLU,
during children’s year of preschool participation. NDW, Comp, and NP) along with the language-developing responsivity
Missing data ranged from 2% to 4% for teachers’ data-providing behaviors. The second dimension mostly represents the systems-level,
features, and 14% to 26% for children’s language measures. There was general language environment as measured by concept development,
no missing data for measures of the general language environment or quality of feedback, and language modeling. The third dimension re-
teachers’ responsivity. Instead of using listwise deletion, which tends to presents teachers’ communication-facilitating responsivity behaviors
greatly reduce sample size and produce biased results (Graham, 2012), exclusively, which do not load onto either factor in the bi-dimensional
we applied full information maximum likelihood (FIML) to incorporate model, and they alone dominate a third factor in the tri-dimensional

Table 5
Pearson’s correlation coefficients between measures of classroom linguistic environment (N = 49 classrooms).

1 2 3 4 5 6 7 8

1. Communication facilitation –
2. Language developing 0.245 –
3. Mean length of utterance −0.311 0.079 –
4. Number of different words −0.063 0.226 0.531 –
5.% complex sentences −0.043 −0.038 0.584 0.268 –
6. Noun phrases −0.203 0.177 0.583 0.689 0.396 –
7. Concept development 0.027 −0.024 0.113 0.244 0.118 0.041 –
8. Quality of feedback −0.096 0.115 0.114 0.400 0.054 0.195 0.522 –
9. Language modeling −0.020 −0.031 0.100 0.339 0.183 0.157 0.576 0.761

Note: For each measure, scores are averaged across all available time points in preschool year.

85
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Fig. 1. Graphical representation of the alternative

models tested in Exploratory SEM.
Note: CD = Concept development; QF = Quality of
Linguistic feedback; LM = Language modeling; LDS = Language
environment developing strategies; Comm = Communication strate-
gies; MLU = Mean length of utterance;
NDW = Number of different words; Comp =
Percentage of complex sentences; NP = noun phrases.

CD QF LM LDS CS MLU NDW Comp NP

Bi-dimensional model

Linguistic Linguistic
environment 1 environment 2

CD QF LM LDS CS MLU NDW Comp NP

Tri-dimensional model

Linguistic Linguistic Linguistic

environment 1 environment 2 environment 3

CD QF LM LDS CS MLU NDW Comp NP

model. Therefore, we concluded from the model fit indices that our data 3.2. Dimensions of the linguistic environment and children’s language
best support a tri-dimensional structure for the preschool-classroom growth
linguistic environment, consisting of data-providing and language-de-
veloping features of teachers’ talk (teacher talk; dimension 1), the sys- In order to explore how the multiple dimensions of the preschool lin-
tems-level general environment (system-level quality; dimension 2), and guistic environment are associated with children’s language-development
teachers’ communication facilitating behaviors (communication facilita- trajectories, we employed multilevel latent growth models as depicted in
tion; dimension 3). Fig. 2. Analyzing grammar and vocabulary development separately, we
The factor loading pattern and factor correlations from the tri-di- hypothesized that children’s language development from fall of preschool to
mensional ESEM model are displayed in Table 7. The factor loading the spring of kindergarten year (three time-points) can be summarized with
pattern reveals a clear three-dimension structure, in which each item a latent growth curve, and that the growth rate (indicated by the latent
loads highly onto one dimension while the loadings on the other two slope) is potentially predicted by the three dimensions of linguistic en-
dimensions are relatively low (< 0.30). Factor correlations are gen- vironment, the scores of which were estimated as factor scores in the tri-
erally small (−0.084 to 0.188), implying that the three dimensions are dimension ESEM. Results of the latent growth models are summarized in
distinct aspects of preschool classrooms’ linguistic environment. Table 8, and some of the coefficient estimates are marked in Fig. 2.

Table 6
Model fit of the ESEM models.

χ2
(df) p Δχ2 test RMSEA CFI SRMR AIC BIC ABIC

1 dimension 89.643 (26) < 0.001 0.211 0.584 0.140 1194.619 1247.590 1159.725
2 dimension 22.548 (18) 0.209 p < 0.001 0.072 0.967 0.059 1150.525 1218.630 1105.660
3 dimension 10.556 (11) 0.481 p = 0.101 0.000 1.000 0.035 1152.533 1233.881 1098.945

Note: RMSEA = Root mean square error of approximation; CFI = Comparative fit index; SRMR = Standardized root mean square residual; AIC = Akaike information criterion;
BIC = Bayesian information criterion; ABIC = Sample-size adjusted Bayesian information criterion.

86
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 7
Standardized loading pattern and factor correlation of the tri-dimensional ESEM.

Teacher Talk System-level Quality Communication Facilitation R2 (%)

λ S.E. λ S.E. λ S.E.

Mean length of utterance 0.631 0.102 49.0

Number of different words 0.790 0.098 75.8
Percent of complex sentences 0.339 0.155 13.5
Noun phrases 0.803 0.090 68.7
Language developing strategies 0.301 0.144 14.8
Communication strategies 0.999 0.001 > 99.9
Concept development 0.637 0.101 40.1
Quality of feedback 0.821 0.087 70.8
Language modeling 0.916 0.073 83.2
Teacher Talk –
System-level Quality 0.188 –
Communication Facilitation −0.084 −0.038 –

Note: The upper-part of the table contains the standardized loadings for each indicator, and the lower-part of the table contains the correlation coefficients between the three factors. For
clarity in the table, cross-loadings below 0.30 are not displayed.
S.E. = Standard error.

Fig. 2. Latent growth models predicting children’s

language development trajectories from preschool
classroom’s linguistic environment.

Intercept Slope

1 1 1 1 2.16
Grammar Y1 Grammar Y1 Grammar Y2
Fall Spring Spring

Grammar Grammar Grammar

Y1 Fall Y1 S prin g Y2 S prin g

1 1
1 2.16
1
Intercept Slope

-.02 .01 -.01

Communication
Teacher Talk System Quality
Facilitation

Intercept Slope

1 1 1 1 2.58
Vocab Y1 Vocab Y2
Vocab Y1 Fall
Spring Spring

Vocab Y1 Vocab Y1 Vocab Y2

Fall Sprin g Sprin g
1
1 1 1 2.58
Intercept Slope

.00 .00 .02*

Communication
Teacher Talk System Quality
Facilitation

87
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Table 8
Parameter estimates of two-level latent growth models.

Grammar Vocab

Coeff SE. p Coeff SE p

Parameter estimates
Within Baseline (intercept) Variance 0.88 0.08 < 0.001 0.87 0.07 < 0.001
Growth rate (slope) Variance 0.09 0.02 < 0.001 0.04 0.01 < 0.001
Between Baseline (intercept) Mean −0.01 0.07 0.903 −0.01 0.08 0.897
Baseline (intercept) Variance 0.04 0.05 0.489 0.14 0.05 0.006
Growth rate (slope) Mean 0.69 0.06 < 0.001 0.62 0.04 < 0.001
Growth rate (slope) Variance 0.00 0.01 0.838 0.01 0.01 0.034
Teacher talk → slope −0.02 0.02 0.37 0.00 0.02 0.892
System-level quality → slope 0.01 0.02 0.666 0.00 0.02 0.950
Communication facilitation → slope −0.01 0.02 0.455 0.02 0.01 0.047
Trajectory Y1 fall to Y1 spring 1.00 / / 1.00 / /
Y1 spring to Y2 spring 2.16 0.16 < 0.001 2.58 0.13 < 0.001

Intraclass correlation (ICC)

Y1 Fall 0.088 0.142
Y1 Spring 0.072 0.093
Y2 Spring 0.034 0.101

Note: Coeff. = Coefficient (unstandardized); S.E. = Standard error.

For grammar, the intraclass correlations were 0.088, 0.072, and responsivity when interacting with children (Girolametto & Weitzman,
0.034 for the three time-points respectively, indicating that 2002; Girolametto, Weitzman, & Greenberg, 2003), the data-providing
3.4% ∼ 8.8% of total variation in children’s grammar scores were at- features available within teachers’ talk to children (Huttenlocher et al.,
tributable to classroom differences. The latent growth model had ac- 2002; Justice et al., 2013), and the systems-level general quality of
ceptable fit (χ2 = 16.01, df = 9, p = 0.067; RMSEA = 0.049; preschool classrooms (Hamre et al., 2014). While each of these pro-
CFI = 0.984). The estimated growth in grammar was approximately posed dimensions of the linguistic environment has been linked to
0.69 standard deviations (SD) in the preschool year as compared to growth in children’s language skills, the unique contribution of each has
baseline, and an extra 0.80 SD in kindergarten. However, the class- not been assessed. Identifying which of these dimensions appears most
room-level growth rate was not predicted by any of the three dimen- strongly associated with children’s language growth over time would
sions of linguistic environment. In fact, with the majority of variation have significant influence on educational practices and policies, as
observed at the child level, and the very small variance of level-2 slope these might serve as the most important levers in the design of pro-
attributable to the classroom (coefficient = 0.00, p = 0.838), we could fessional development (PD) and curricular programs and practices. In
not expect to link any feature of the classroom linguistic environment to addition, such work could be highly informative to teacher-preparation
the growth of grammar. programs that prepare educators to work in early-education settings.
For vocabulary, the intraclass correlations were 0.142, 0.093, and The goal of the present study was to examine the dimensionality of
0.101 for the three time-points respectively, indicating that the linguistic environment of 49 preschool classrooms, and assess the
9.3%–14.2% of total variance in vocabulary scores were attributable to relations between the empirically determined dimensions and chil-
classroom differences. Again, fit was satisfactory for the latent growth dren’s longitudinal trajectories in vocabulary and grammar from pre-
model (χ2 = 10.84, df = 9, p = 0.287; RMSEA = 0.025; school to the end of kindergarten. Although there was insufficient be-
CFI = 0.997). On average, children’s vocabulary scores increased by tween-classroom variance in children’s grammatical development to
0.62 SD in the preschool year, and by 0.98 SD in the kindergarten year, link features of the classroom environment to growth in this domain of
demonstrating accelerating growth. Among the three dimensions of language, we found that one specific dimension of the preschool
preschool linguistic environment, teachers’ communication-facilitating classroom linguistic environment was associated with children’s long-
behaviors significantly predicted growth rate at the classroom level itudinal growth in vocabulary, namely teachers’ use of communication-
(coefficient = 0.02, p = 0.047). Specifically, with one unit increase in facilitating strategies. Here, we discuss this and additional findings of
teachers’ use of communication-facilitating behaviors, the growth rate note as well as implications for educational practice.
in vocabulary was expected to increase by 0.02 SD. On the other hand, The first finding of note in this study was establishing that the
teacher talk and systems-level general environment did not predict the preschool classroom linguistic environment is best represented tri-di-
classroom-level growth rate. mensionally. Dimension 1, teacher talk, reflected the data-providing
and language-developing features of teachers’ talk, including both the
lexical diversity and grammatical complexity of their utterances.
4. Discussion Dimension 2, system-level quality, reflected the general quality of the
classroom environment with respect to instructional support; and
Today, the majority of preschool-aged children in the United States Dimension 3, communication facilitation, reflected teachers’ use of
attend center-based preschool programs in the years prior to kinder- specific facilitative behaviors that serve to engender children’s con-
garten entrance. Many preschool programs, especially those supported versational participation. This finding helps us to recognize that the
by federal and state funds, include in their mission a goal to support linguistic environment of preschool classrooms is multi-dimensional,
children’s school readiness, including language development (Scott- and that the dimensions appear to reflect discrete aspects of the class-
Little, Kagan, & Frelow, 2006). Both theory and evidence support the room environment, given the relatively low inter-dimensional correla-
need for preschool classrooms to provide linguistically rich environ- tions. For instance, two variables represented in Dimension 1, namely
ments to children in order to support their language growth over time the mean length of utterances and percentage of syntactically complex
(Dickinson & Caswell, 2007; Dickinson & Smith, 1994). Such work has utterances of teachers’ talk, both reflecting important data-providing
generally emphasized three dimensions of the preschool classroom features of teachers’ talk, were very modestly correlated with the
environment as being especially important: teachers’ linguistic

88
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

systems-level general measure of quality represented by the Language challenging in the classroom than in a dyadic parent-child situation,
Modeling subscale of the CLASS (r = 0.093 and 0.179, respectively). and therefore teachers’ ability to use communication-facilitating stra-
Similarly, teachers’ use of communication-facilitating behaviors, re- tegies become crucial.
presenting in part teachers’ linguistic responsivity when talking with However, none of these hypotheses about the role of communica-
children, was also poorly correlated with the Language Modeling sub- tion-facilitating strategies can explain why the other aspect of re-
scale (r = 0.229). The importance of this finding is that it shows that sponsivity evaluated in this study, language-developing behaviors, did
the linguistic environment of preschool classrooms cannot be re- not make a contribution to children’s vocabulary growth. It is possible
presented monolithically in empirical or theoretical discussions. Rather, that communication-facilitating strategies moderate the impact of other
it seems necessary to disambiguate the nature of teacher talk, the sys- dimensions of classroom language. For example, language-developing
tems-level quality, and communication facilitating behaviors of the strategies such as recasts and expansions may only have a positive
teacher, and as well to recognize that the systems-level quality mea- impact when children participate and provide input for the adult to
sures often discussed in the literature (Keys et al., 2013)may not ade- recast or expand. It is possible that in the absence of sufficient com-
quately index the more proximal processes represented by Dimension 1, munication-facilitating behaviors, children do not develop the socio-
Teacher Talk, and Dimension 3, Communication Facilitation. cognitive skills necessary to benefit fully from linguistic input.
In examining the dimensionality of the linguistic environment of By demonstrating the potentially crucial contribution of teacher
preschool settings, it is worth noting that teachers’ linguistic re- behaviors that actively cultivate children to communicate in the
sponsivity, measured in terms of language-developing and commu- classroom, the results of this study help to elucidate why recent efforts
nication-facilitating behaviors, do not represent a single unitary di- to link systems-level general classroom quality measures to child lan-
mension. Rather, teachers’ language-developing behaviors, such as guage gains during preschool have not been successful. For instance,
their use of recasts and expansions, loaded with data-providing features Keys and colleagues (2013) conducted a meta-analysis examining the
of teachers’ talk, such as the syntactic complexity of teachers’ utter- relations between several systems-level preschool-classroom quality
ances. Theoretically, this is an intriguing finding as it suggests that indices and children’s language growth which aggregated data from
functional features of teachers’ talk that have long been believed to four large-scale studies. One of the four studies specifically employed
directly facilitate children’s language, such as expanding children’s the CLASS Instructional Support dimension, as we did in the present
utterances, share variance with the grammatical properties of teachers’ study, showing it to be only very modestly related to children’s lan-
utterances, which are also viewed as facilitative of children’s language. guage outcomes. The present results suggest that systems-level class-
It seems that features of teachers’ talk, whether functional forms (e.g., room quality measures may not adequately document the critical
questions) or grammatical properties, that directly serve to provide proximal processes that most contribute to children’s development in
children with advanced linguistic models represent a single dimension preschool settings.
of children’s linguistic environment. In this regard, the present study also raises questions regarding
The second finding of note is that the dimension most strongly as- some current directions for how to improve children’s language de-
sociated with children’s linguistic trajectories, specific to vocabulary, velopment in the preschool setting, specifically those which focus on
was teachers’ use of communication-facilitating behaviors. In fact, improving the systems-level general environment through intensive
teachers’ use of these behaviors was the only significant predictor of professional development of teachers (Early, Maxwell, Ponder, & Pan,
vocabulary growth over time. We propose that the mechanism behind 2017; Hamre et al., 2012; LoCasale-Crouch et al., 2016). The outcome
this finding is that the communication-facilitating behaviors measured of these studies is the CLASS observational tool, presumably because
in this study resulted in children talking more within the classroom, improvements in the systems-level general quality of the classroom
which directly facilitates children’s language development. That is, by environment will have downstream effects on children’s language skills,
using strategies to encourage children to talk more, teachers are able to among other cognitive outcomes (LoCasale-Crouch et al., 2016).
directly facilitate children’s language growth. At the same time, these However, recent evidence shows that even very intensive professional
strategies also likely served to encourage children’s active participation development efforts delivered to preschool teachers are ineffective in
in conversations, which has been shown to contribute to children’s increasing CLASS Instructional Support scores to the magnitude that
growth in vocabulary (Zimmerman et al., 2009). Conversations can be likely is needed to improve children’s language skills, with increases
especially important to children’s language development, as these typically in the range of about 0.1–0.3 of one point on the 7-point scale
provide a context in which adults can scaffold their interactions based (Early et al., 2017). Rather, professional development efforts provided
on the children’s needs (Konishi, Kanero, Freeman, Golinkoff, & Hirsh- to early educators should focus most intensively on helping them to
Pasek, 2014). The information provided by children’s talk may be even both elevate and execute the precise, proximal behaviors that serve to
more relevant for adults in group situations such as classrooms, given engage children in productive conversations.
the diversity in children’s language skills and the teacher’s need to While we were not particularly surprised to find that systems-level
adapt their language to different children at different times. general features of the classroom environment would not correlate with
Another way in which communication-facilitating strategies could children’s growth (see Keys et al., 2013), it is unexpected to find that
encourage vocabulary growth is by supporting the development of the data-providing features of teachers’ talk did not have unique re-
socio-cognitive skills that are crucial for language learning, such as joint levance for children’s language growth. Both theory and evidence point
attention (Tamis-LeMonda, Kuchirko, & Song, 2014). The episodes of to the importance of very granular features of adults’ talk, such as the
joint attention created by communication-facilitation behaviors, such syntactic complexity of their utterances, to children’s growth in both
as following the child’s lead or waiting expectantly for children to re- grammar and vocabulary (Hoff, 2003; Huttenlocher et al., 2002). There
spond, are key in helping children understand the intentional and are several possibilities regarding our non-significant findings. First,
communicative nature of language, and this understanding in turn can prior research linking data-providing features of teachers’ talk to chil-
support children’s subsequent vocabulary acquisition. Joint attention, dren’s language skills did not include other dimensions of the classroom
as well as other social cognition skills such as perspective taking and linguistic environment; thus, the observed relations may have reflected
theory of mind, are seen as central to language acquisition in social- the possible shared variance between data-providing features of tea-
pragmatic accounts of word learning, since they expand the available chers’ talk and other linguistically-supportive behaviors of teachers
information that children can use to identify the correct referent for a (Huttenlocher et al., 2002). Second, prior research linking data-pro-
new word (Butler & Tomasello, 2016; Grassmann, Stracke, & Tomasello, viding features of both teachers’ and parents’ talk to children’s language
2009; Tomasello & Akhtar, 1995). Creating joint attention episodes that growth has focused on children who were younger than those in the
foster the development of social cognition skills may be more present study; for instance, Huttenlocher et al. (2002) showed a

89
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

significant, positive correlation between teachers’ syntactic complexity linguistic environment variables had limited predictive power. Future
and children’s language growth for children who were three years of research evaluating individual differences via the examination of child-
age. Hoff’s (2003) research linking mothers’ syntactic complexity to specific language-learning environment features (e.g., how much in-
children’s vocabulary skills involved children who were two years of dividual child is exposed to teacher talk) is necessary to better under-
age. The children in the present study were older than those in the prior stand young children’s language acquisition.
works, averaging 4 years of age at the study start. It may be that data- To conclude, we highlight several key implications of this work for
providing features of linguistic input become less salient as a source of practice. First, the importance of teachers’ linguistic responsivity in the
language growth as children grow older, with opportunities to actively preschool classroom, especially their use of communication-facilitating
engage and participate in conversations with peers and adults becoming behaviors, should be emphasized. It appears that children’s engagement
a more prominent mechanism for growth. in extended conversations, a phenomenon explicitly facilitated by tea-
The present study has several important implications for practice chers’ behaviors, is extremely important to their language development,
and research, but before reviewing these we want to make clear the especially vocabulary (Cabell et al., 2015). To this end, we contend that
limitations of this study. First, there are dimensions of the linguistic pre- and post-professional development of teachers should explicitly
environment that were not assessed in this study that may be especially support teachers to learn to use conversation-facilitating behaviors.
significant, especially teachers’ use of questions (Zucker, Justice, However, with relatively little training, evidence points to teachers as
Piasta, & Kaderavek, 2010) and the conceptual focus of teachers’ talk being able to significantly increase their use of these behaviors (Piasta
(Dickinson & Smith, 1994). Thus, while our work shows that teachers’ et al., 2012). Second, we also suggest that less attention be directed
communication-facilitating behaviors are an especially important di- towards increasing the two dimensions of the classroom linguistic en-
mension of the classroom linguistic environment, they may not be the vironment that are not linked to children’s language growth (the tea-
only significant dimension. Future research needs to continue to in- cher talk and systems-level dimensions), when improving children’s
vestigate how best to represent the classroom linguistic environment language skills is a desired outcome. Specifically, it is extremely chal-
and ensure that all relevant features are represented. Second, and re- lenging to coach educators to use more syntactically complex or lexi-
latedly, the assessments used to represent the linguistic environment of cally diverse talk, as one’s linguistic skills are relatively intractable by
children’s classrooms did not transcend various contexts of discourse, adolescence (Farkas & Beron, 2004). Similarly, evidence shows that the
with two of the assessments focused specifically on teacher-child systems-level general quality of preschool classrooms is also very dif-
proximal interactions during free-play. It is well-established that book ficult to change (Early et al., 2017), with intensive efforts resulting in
reading is a particularly important context of the early education set- only very small, incremental improvements. We interpret the present
ting within which teachers can facilitate children’s language skills study as well as other prior works (Cabell et al., 2015;
(Dickinson & Smith, 1994). Thus, future investigations of the di- Dickinson & Smith, 1994; Girolametto & Weitzman, 2002; Girolametto
mensionality of the preschool classroom’s linguistic environment et al., 2003; Wasik & Bond, 2001; Zucker et al., 2010) to show that the
should explore how discourse contexts fit into such models. Third, the most prominent lever for improving children’s language skills in the
classrooms involved in this study served only children from dis- early education setting is to explicitly target the precise, proximal
advantaged backgrounds. These are a subset of children enrolled in processes that lead to high-quality, extended conversations among
center-based programming, and it is unknown whether the results in teachers and children.
this study would generalize outward to include all preschool classrooms
and an unselected sample of children. As preschool access continues to Author note
expand towards universal access, it is necessary to ensure that research
findings focused on targeted-enrollment programs can appropriately be The research reported herein was supported by Grant R305F05124
generalized. Fourth, with 49 classrooms in the sample, sample size was from the U.S. Department of Education, Institute of Education Sciences
limited in level-2 of the latent growth model. However based on the to the first author. The content of this publication does not necessarily
simulation study of Maas and Hox (2005), a sample size of 50 is suf- reflect the views or policies of the Institute, nor does mention of trade
ficient for producing unbiased parameter estimates and standard error names, commercial products, or organizations imply endorsement by
for fixed effects. Finally, especially for grammar growth, the majority of the U.S. Department of Education. We are grateful to the many tea-
individual variation lies at the child level, and the classroom-level chers, children, and research staff who made this study possible.

Appendix A

Equations of the two-level latent growth models are specified as below. Note that all equations are expressed in matrix forms in accordance with
conventions of the SEM literature.
For child i nested within classroom g, let ygi‘ = [ygi1, ygi2, ygi3] represent the child’s outcomes (grammar or vocabulary scores) over three time
points (year 1 fall, year 1 spring, and year 2 spring). At level 1, ygi is modeled as

ygi μ g = μ g + ΛW ηgi + εgi ,

where μg‘ = [μg1, μg2, μg3] represents the random intercepts of the longitudinal measurements; ΛW is the 3 × 2 within-level factor loading matrix
linking latent intercept and slope to the outcomes; ηgi‘ = [ηgi-Intercept, ηgi-Slope] is the vector of within-level latent growth parameters; and εgi‘ = [εgi1,
εgi2, εgi3] represent the within-level error. At level 2, the random intercept μg is modeled as
μ g = ΛB ηg + εg
,
= ΛB (α + ΓB ξ g) + εg

where ΛΒ is a 3 × 2 between-level factor loading matrix; ηg‘ = [ηg-Intercept, ηg-Slope] represents the between-level latent growth parameters; α is the
mean vector that contains the fixed intercept and slope; ΓB is a 2 × 3 matrix describing the relationship between the three classroom-level predictors
(teacher talk, system quality, and communication facilitation) and the latent growth parameters; ξg is the 3 × 1vector containing the predictors; and
εg‘ = [εg1, εg2, εg3] is the between-level error.
Given the unequal spacing between the three time points, and the complexities in early language development, we assume that the growth curve

90
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

may be non-linear. Therefore, the factor loading matrices are defined as

⎡1 0 ⎤
ΛB = ΛW = ⎢ 1 1 ⎥,
⎢ 1 β3 ⎥
⎣ ⎦
where β3 represents the growth rate from time point 2 (year 1 spring) to time point 3 (year 2 spring) relative to the growth rate from time point 1
(year 1 fall) to time point 2.
Further, assuming normal distribution of latent growth parameters and error terms, we define
ηgi ∼ N (0, ΨW ),

ηg ξ g ∼ N (α + ΓB ξ g , ΨB ),

εgi ∼ N (0, ΘW ), and

εg ∼ N (0, ΘB );

where Ψ is the 2 × 2 variance-covariance matrix of the latent growth parameters; and Θ is the 3 × 3 diagonal variance-covariance matrix of error
terms.

References as potential referents of novel words based on pragmatics. Cognition, 112, 488–493.
http://dx.doi.org/10.1016/j.cognition.2009.06.010.
Hamre, B., Justice, L. M., Pianta, R. C., Kilday, C., Sweeney, B., Downer, J. T., & Leach, A.
Adamson, L. B., Bakeman, R., & Deckner, D. F. (2004). The development of symbol-in- (2010). Implementation fidelity of MyTeachingPartner literacy and language activ-
fused joint engagement. Child Development, 75(4), 1171–1187. ities: association with preschoolers' language and literacy growth. Early Childhood
Asparouhov, T., & Muthén, B. (2009). Exploratory structural equation modeling. Research Quarterly, 25(3), 329–347.
Structural Equation Modeling: A Multidisciplinary Journal, 16(3), 397–438. Hamre, B., Pianta, R. C., Burchinal, M., Field, S., LoCasale-Crouch, J., Downer, J. T., &
Browne, M. W. (2001). An overview of analytic rotation in exploratory factor analysis. Scott-Little, C. (2012). A course on effective teacher-child interactions effects on
Multivariate Behavioral Research, 36(1), 111–150. teacher beliefs, knowledge, and observed practice. American Educational Research
Burchinal, M., Vandergrift, N., Pianta, R. C., & Mashburn, A. (2010). Threshold analysis of Journal, 49(1), 88–123.
association between child care quality and child outcomes for low-income children in Hamre, B., Hatfield, B., Pianta, R., & Jamil, F. (2014). Evidence for general and do-
pre-kindergarten programs. Early Childhood Research Quarterly, 25(2), 166–176. main‐specific elements of teacher–child interactions: Associations with preschool
Butler, L. P., & Tomasello, M. (2016). Two- and 3-year-olds integrate linguistic and children's development. Child Development, 85(3), 1257–1274.
pedagogical cues in guiding inductive generalization and exploration. Journal of Hayiou-Thomas, M. E., Dale, P. S., & Plomin, R. (2012). The etiology of variation in
Experimental Child Psychology, 145, 64–78. http://dx.doi.org/10.1016/j.jecp.2015. language skills changes with development: A longitudinal twin study of language
12.001. from 2 to 12 years. Developmental Science, 15(2), 233–249.
Cabell, S. Q., Justice, L. M., Piasta, S. B., Curenton, S. M., Wiggins, A., Turnbull, K. P., & Hoff, E., & Naigles, L. (2002). How children use input to acquire a lexicon. Child
Petscher, Y. (2011). The impact of teacher responsivity education on preschoolers' Development, 73(2), 418–433.
language and literacy skills. American Journal of Speech-Language Pathology, 20(4), Hoff, E. (2003). The specificity of environmental influence: Socioeconomic status affects
315–330. early vocabulary development via maternal speech. Child Development, 74(5),
Cabell, S. Q., Justice, L. M., McGinty, A. S., DeCoster, J., & Forston, L. D. (2015). 1368–1378.
Teacher–child conversations in preschool classrooms: contributions to children's Huttenlocher, J., Vasilyeva, M., Cymerman, E., & Levine, S. (2002). Language input and
vocabulary development. Early Childhood Research Quarterly, 30, 80–92. child synta. Cognitive Psychology, 45, 337–374.
Dickinson, D. K., & Caswell, L. (2007). Building support for language and early literacy in Justice, L. M., Mashburn, A. J., Hamre, B., & Pianta, R. C. (2008). Quality of language and
preschool classrooms through in-service professional development: effects of the literacy instruction in preschool classrooms serving at-risk pupils. Early Childhood
Literacy Environment Enrichment Program (LEEP). Early Childhood Research Research Quarterly, 23(1), 51–68.
Quarterly, 22(2), 243–260. Justice, L. M., McGinty, A. S., Zucker, T., Cabell, S. Q., & Piasta, S. B. (2013). Bi-direc-
Dickinson, D. K., & Smith, M. W. (1994). Long-term effects of preschool teachers' book tional dynamics underlie the complexity of talk in teacher–child play-based con-
readings on low-income children's vocabulary and story comprehension. Reading versations in classrooms serving at-risk pupils. Early Childhood Research Quarterly,
Research Quarterly, 105–122. 28(3), 496–508.
Dodge, D. T., Colker, L. J., & Heroman, C. (2002). The creative curriculum for preschool (4th Keys, T. D., Farkas, G., Burchinal, M. R., Duncan, G. J., Vandell, D. L., Li, W., & Howes, C.
ed.). Washington, D.C: Teaching Strategies Inc. (2013). Preschool center quality and school readiness: Quality effects and variation
Dunn, L. M., & Dunn, L. M. (1997). Peaboby picture vocabulary test (3rd ed.). Circle Pines, by demographic and child characteristics. Child Development, 84(4), 1171–1190.
MN: American Guidance Service. Konishi, H., Kanero, J., Freeman, M. R., Golinkoff, R. M., & Hirsh-Pasek, K. (2014). Six
Dunn, D. M., & Dunn, L. M. (2007). Peabody picture vocabulary test (PPVT). New York: Principles of language development: Implications for second language Learners.
Pearson. Developmental Neuropsychology, 39(5), 404–420. http://dx.doi.org/10.1080/
Early, D. M., Maxwell, K. L., Ponder, B. B., & Pan, Y. (2017). Improving teacher-child 87565641.201.931961.
interactions: A randomized control trial of Making the Most of Classroom Interactions Landry, S. H., Smith, K. E., Swank, P. R., Assel, M. A., & Vellet, S. (2001). Does early
and My Teaching Partner professional development models. Early Child Res Q, 38, responsive parenting have a special importance for children's development or is
57–70. consistency across early childhood necessary? Developmental Psychology, 37(3), 387.
Enders, C. K. (2010). Applied missing data analysis. Guilford Press. Landry, S. H., Smith, K. E., & Swank, P. R. (2006). Responsive parenting: establishing
Farkas, G., & Beron, K. (2004). The detailed age trajectory of oral vocabulary knowledge: early foundations for social, communication, and independent problem-solving skills.
Differences by class and race. Social Science Research, 33(3), 464–497. Developmental Psychology, 42(4), 627.
Fey, M. E., Cleave, P. L., Long, S. H., & Hughes, D. L. (1993). Two approaches to the LoCasale-Crouch, J., Konold, T., Pianta, R., Howes, C., Burchinal, M., Bryant, D., &
facilitation of grammar in children with language impairment: An experimental Barbarin, O. (2007). Observed classroom quality profiles in state-funded pre-kin-
evaluation. Journal of Speech, Language, and Hearing Research, 36(1), 141–157. dergarten programs and associations with teacher, program, and classroom char-
Girolametto, L., & Weitzman, E. (2002). Responsiveness of child care providers in in- acteristics. Early Childhood Research Quarterly, 22(1), 3–17.
teractions with toddlers and preschoolers. Language, Speech, and Hearing Services in LoCasale-Crouch, J., DeCoster, J., Cabell, S. Q., Pianta, R. C., Hamre, B., Downer, J. T., &
Schools, 33(4), 268. Howes, C. (2016). Unpacking intervention effects: Teacher responsiveness as a
Girolametto, L., Pearce, P. S., & Weitzman, E. (1996). Interactive focused stimulation for mediator of perceived intervention quality and change in teaching practice. Early
toddlers with expressive vocabulary delays. Journal of Speech, Language, and Hearing Childhood Research Quarterly, 36, 201–209.
Research, 39(6), 1274–1283. Maas, C. J. M., & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling.
Girolametto, L., Weitzman, E., & Greenberg, J. (2000). Teacher interaction and language Methodology, 1(3), 86–92. http://dx.doi.org/10.1027/1614-2241.1.3.86.
rating scale. Toronto, OR: The Hanen Centre. Marcus, G. F., Vijayan, S., Rao, S. B., & Vishton, P. M. (1999). Rule learning by seven-
Girolametto, L., Weitzman, E., & Greenberg, J. (2003). Training day care staff to facilitate month-old infants. Science, 283(5398), 77–80.
children's language. American Journal of Speech-Language Pathology, 12(3), 299–311. Mashburn, A. J., Pianta, R. C., Hamre, B., Downer, J. T., Barbarin, O. A., Bryant, D., &
Graham, J. W. (2012). Missing data: analysis and design. New York, NY: Springer Howes, C. (2008). Measures of classroom quality in prekindergarten and children’s
Science & Business Media. development of academic, language, and social skills. Child Development, 79(3),
Grassmann, S., Stracke, M., & Tomasello, M. (2009). Two-year-olds exclude novel objects 732–749.

91
L.M. Justice et al. Early Childhood Research Quarterly 42 (2018) 79–92

Mashburn, A. J. (2016). Evaluating the validity of classroom observations in the head Saffran, J. R. (2003). Statistical language learning mechanisms and constraints. Current
start designation renewal system. Educational Psychologist, 1–12. Directions in Psychological Science, 12(4), 110–114.
Meredith, W., & Tisak, J. (1990). Latent curve analysis. Psychometrika, 55(1), 107–122. Scott-Little, C., Kagan, S. L., & Frelow, V. S. (2006). Conceptualization of readiness and
Muthén, L. K., & Muthén, B. O. (2005). Mplus: Statistical analysis with latent variables: user's the content of early learning standards: The intersection of policy and research? Early
guide. Muthén & Muthén Los Angeles. Childhood Research Quarterly, 21(2), 153–173.
NICHD Early Child Care Research Network (2002). Child-care structure→ process→ Spinath, F. M., Price, T. S., Dale, P. S., & Plomin, R. (2004). The genetic and environ-
outcome: Direct and indirect effects of child-care quality on young children's devel- mental origins of language disability and ability. Child Development, 75(2), 445–454.
opment. Psychological Science, 13(3), 199–206. Tamis-LeMonda, C. S., Bornstein, M. H., & Baumwell, L. (2001). Maternal responsiveness
Nicely, P., Tamis-LeMonda, C. S., & Bornstein, M. H. (1999). Mothers’ attuned responses and children's achievement of language milestones. Child Development, 72(3),
to infant affect expressivity promote earlier achievement of language milestones. 748–767.
Infant Behavior and Development, 22(4), 557–568. Tamis-LeMonda, C. S., Kuchirko, Y., & Song, L. (2014). Why is infant language learning
Pianta, R. C., Karen, M., Paro, L., & Hamre, B. (2008). Classroom assessment scoring system. facilitated by parental responsiveness. Current Directions in Psychological Science,
Paul H. Brookes Publishing Company. 23(2), 121–126. http://dx.doi.org/10.1177/0963721414522813.
Piasta, S. B., Justice, L. M., Cabell, S. Q., Wiggins, A. K., Turnbull, K. P., & Curenton, S. M. Tomasello, M., & Akhtar, N. (1995). Two-year-olds use pragmatic cues to differentiate
(2012). Impact of professional development on preschool teachers’ conversational reference to objects and actions. Cognitive Development, 10, 201–224. http://dx.doi.
responsivity and children's linguistic productivity and complexity. Early Childhood org/10.1016/0885-2014(95)90009-8.
Research Quarterly, 27(3), 387–400. Wasik, B. A., & Bond, M. A. (2001). Beyond the pages of a book: Interactive book reading
Powell, D. R., Diamond, K. E., Burchinal, M. R., & Koehler, M. J. (2010). Effects of an and language development in preschool classrooms. Journal of Educational
early literacy professional development intervention on head start teachers and Psychology; Journal of Educational Psychology, 93(2), 243.
children. Journal of Educational Psychology, 102(2), 299. Wiig, E. H., Secord, W., & Semel, E. M. (2004). CELF Preschool 2: Clinical evaluation of
Proctor-Williams, K., & Fey, M. E. (2007). Recast density and acquisition of novel irre- language fundamentals preschool. Pearson/PsychCorp.
gular past tense verbs. Journal of Speech, Language, and Hearing Research, 50(4), Yoder, P. J., & Warren, S. F. (1999). Maternal responsivity mediates the relationship
1029–1047. between prelinguistic intentional communication and later language. Journal of Early
Rowe, M. L., Raudenbush, S. W., & Goldin-Meadow, S. (2012). The pace of vocabulary Intervention, 22(2), 126–136.
growth helps predict later vocabulary skill. Child Development, 83(2), 508–525. Zimmerman, F. J., Gilkerson, J., Richards, J. A., Christakis, D. A., Xu, D., Gray, S., &
Sabol, T. J., & Pianta, R. C. (2015). Validating Virginia's quality rating and improvement Yapanel, U. (2009). Teaching by listening: The importance of adult-child conversa-
system among state-funded pre-kindergarten programs. Early Childhood Research tions to language development. Pediatrics, 124(1), 342–349.
Quarterly, 30, 183–198. Zucker, T., Justice, L., Piasta, S., & Kaderavek, J. (2010). Preschool teachers’ literal and
Saffran, J. R., & Wilson, D. P. (2003). From syllables to syntax: Multilevel statistical inferential questions and children's responses during whole-class shared reading.
learning by 12-month-old infants. Infancy, 4(2), 273–284. Early Childhood Research Quarterly, 25(1), 65–83.

92