593923
EROXXX10.1177/2332858415593923Barbu et al.Assessing Approaches to Learning
research-article2015
AERA Open
July-September 2015, Vol. 1, No. 3, pp. 1–15
DOI: 10.1177/2332858415593923
© The Author(s) 2015. http://ero.sagepub.com
Assessing Approaches to Learning in School Readiness:
Comparing the Devereux Early Childhood Assessment to an
Early Learning Standards-Based Measure
Otilia C. Barbu
David B. Yaden Jr.
Deborah Levine-Donnerstein
Ronald W. Marx
University of Arizona
This study examines the psychometric properties of two assessments of children’s approaches to learning: the Devereux Early
Childhood Assessment (DECA) and a 13-item approaches to learning rating scale (AtL) derived from the Arizona Early
Learning Standards (AELS). First, we administered questionnaires to 1,145 randomly selected parents/guardians of first-time
kindergarteners. Second, we employed confirmatory factor analysis (CFA) with parceling for DECA to reduce errors due to
item specificity and prevent convergence difficulties when simultaneously estimating DECA and AtL models. Results indicated
an overlap of 55% to 72% variance between the domains of the two instruments and suggested that the new AtL instrument
is an easily administered alternative to the DECA for measuring children’s approaches to learning. This is one of the first
studies that investigated DECA’s approaches to learning dimension and explored the measurement properties of an instrument purposely derived from a state’s early learning guidelines.
Keywords: readiness, approaches to learning, DECA, EFA/CFA, parceling
A substantial body of recent research suggests that the
level of school readiness is an important and even a critical
indicator of whether an entering kindergarten child will succeed in school (Haskins & Barnett, 2010; Magnuson, Rohm,
& Waldfogel, 2007; K. Snow, 2006, 2011). Several longitudinal analyses of high-quality early childhood programs
over the past decade demonstrate that considerable benefits
accrue to both individuals and society when children have
positive experiences with well-trained caregivers prior to
school entry (Heckman, Moon, Pinto, Savelyev, & Yavitz,
2010; Rolnick & Grunewald, 2003). Conversely, when children’s high-quality early experiences are absent, lower
school performance in the early grades is likely, reducing the
possibility of greater success as children advance in school
(Pianta, Barnett, Burchinal, & Thornburg, 2009). Thus, educators, child development specialists, researchers, and policy makers could gain knowledge about children at risk at
the start of kindergarten (K. Snow, 2011) by examining
school readiness factors, and subsequently, they could design
policies and programs to ameliorate these risks (cf. also
Center on the Developing Child at Harvard University,
2011).
Approaches to learning is a key construct in many policybased definitions of school readiness, although insufficient
research exists on how to measure this construct efficiently
and accurately. We address this gap in the research by evaluating the psychometric properties of two assessments of
approaches to learning using data from parent and guardian
ratings. We compare findings from the Devereux Early
Childhood Assessment, a widely used instrument for assessing emotional intelligence in preschool settings (Chain,
Dopp, Smith, Woodland, & LeBuffe, 2010) with findings
from a researcher-developed, 13-item instrument designed
specifically to assess approaches to learning behaviors based
on the Arizona State Early Learning Standards (AELS). In
this comparison, we examined the psychometric validity of
this shorter instrument and its potential efficiency and usefulness for classroom assessment.
Development of State Standards for Early Childhood
Despite definitional differences regarding the nature of
school readiness (Graue, 1993, 2006; Kagan, 1990; ScottLittle, Kagan, & Frelow, 2006), an effort to develop early
learning standards for pre-kindergarten to third-grade children has arisen among regional and national policy makers
(Brito, 2012; Neuman & Roskos, 2005; Scott-Little et al.,
2006). In the United States, outcomes from the National
Creative Commons CC-BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 3.0
License (http://www.creativecommons.org/licenses/by-nc/3.0/) which permits non-commercial use, reproduction and distribution of the
work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (http://www.uk.sagepub.com/
aboutus/openaccess.htm).
Barbu et al.
Education Goals Panel (NEGP) in 1995 and the Goals 2000
legislation passed in 1994 (C. E. Snow & Van Hemel, 2008)
led most states to create standards for assisting preschool
professionals to guide the development of infants, toddlers,
and preschoolers (K. Snow, 2011). These standards included
the following domains: (a) physical well-being and motor
development, (b) social and emotional development, (c)
approaches to learning (our emphasis), (d) language development, and (e) cognition and general knowledge (see
Kagan, Moore, & Bredekamp, 1995, for the foundational
definitions). These dimensions also are included in the
United Nations Children’s Fund (UNICEF) conceptual
frame (Brito, 2012). Additional specificity in outlining the
primary domains of readiness has been provided in the
United States in the School Readiness Act of 2007, which
includes language, literacy, mathematics, science, social and
emotional functioning, creative arts, physical skills, and
approaches to learning.
In an analysis of 46 early learning standards documents
published after 1999, Scott-Little et al. (2006) noted the limited amount of attention given in many standards, in particular, to socioemotional development and approaches to
learning (see also Chen, Masur, & McNamee, 2010;
Niemeyer & Scott-Little, 2001; Thigpen, 2014, for similar
statements), two components identified as key to early success in school (NEGP, 1995; C. E. Snow & Van Hemel,
2008). In addition to the uneven discussion among the standards themselves, to our knowledge, no state has developed
assessments tailored to any of their standards. Rather, extant
measures are used whose norming samples may have different characteristics than their state’s population of children,
for whom the standards are targeted. Similar limitations
exist in the measurement of school readiness internationally
(Brito & Limlingan, 2012).
Moreover, concerns with instruments’ measurement accuracy with young children (e.g., Isquith, Gioia, & Espy, 2004;
Meisels & Fenichel, 1996; K. Snow, 2006, 2011), as well as
their ease of use and interpretation (Diamond, Justice, Siegler,
& Snyder, 2013; National Association for the Education of
Young Children, 2002), raise issues in school readiness
assessments. As these assessments were meant to support
instruction, identify children at risk and in need of special
services, and facilitate program evaluation and accountability, accuracy of measurement is paramount. Particularly, in
program evaluation and accountability, important and politically sensitive program evaluations often are conducted by
request of funders, such as federal agencies, state and local
governments, nongovernment organizations, and private philanthropy that require reliable and valid data for meaningful
investment decisions (Center on the Developing Child at
Harvard University, 2011). An additional concern is that the
high-stakes testing movement (Books, 2004; Kim &
Sunderman, 2005; Woodside-Jiron & Gehsmann, 2009) coupled with oversimplified and one-time-used assessments has
2
produced deleterious effects on vulnerable populations
(Gehsmann & Templeton, 2011–2012). This concern was
also reflected by UNICEF research (Brito & Lumlingan,
2012).
In light of these concerns (e.g., limited, reliable instrumentation based on specific learning criteria), we describe in
this article an instrument designed to measure the domain of
approaches to learning (AtL) based on one state’s early
learning standards that is both a valid indication of the construct and an easily administered instrument. In the sections
below, we provide additional information on the approaches
to learning domain, a related literature on children’s executive functioning (EF), and the context of the large-scale
study where the new AtL assessment was administered.
Assessing Approaches to Learning
School Readiness and Components of
Approaches to Learning
The approaches to learning construct, introduced by
Kagan et al. (1995) as a component of school readiness, has
been identified as an important domain related to children’s
positive early achievement outcomes in math, reading, and
socioemotional development (Fantuzzo, Bulotsky-Shearer,
Fusco, & McWayne, 2005; Ziv, 2013). According to Kagan
et al. (1995), this construct comprises a combination of traits
such as gender and temperament, predispositions and attitudes conditioned by culture, and learning styles. However,
in contrast to predispositions, Kagan et al. suggested that
learning styles are malleable and include variables that affect
how children attitudinally address the learning process: their
openness to and curiosity about new tasks and challenges;
their initiative, task persistence, and attentiveness; their
approach to reflection and interpretation; their capacity for
invention and imagination; and their cognitive approaches to
tasks (p. 23).
Given that certain styles of learning are favored over others in the U.S. educational system, Kagan et al. (1995) urged
that further research is needed to understand approaches to
learning so that all children, despite their diversity of learning styles, could have equal opportunities to learn with
appropriate pedagogical adjustments. Two decades after
Kagan et al.’s assertion that approaches to learning was an
underresearched area, their claim is echoed in more recent
surveys of the research literature (e.g., Scott-Little et al.,
2006; C. E. Snow & Van Hemel, 2008; K. Snow, 2011),
although efforts have been made to develop instruments to
measure, in particular, the malleable aspects of this domain.
Using varied definitions of approaches to learning,
researchers have studied characteristics such as initiative,
curiosity, persistence, engagement, and problem solving
(e.g., Bulotsky-Shearer, Fernandez, Dominguez, & Rouse,
2011; Li-Grining, Votruba-Drzal, Maldonado-Carreño, &
Hass, 2010). In addition, researchers also consistently find
Assessing Approaches to Learning
connections to children’s early cognitive and developmental
growth with characteristics defined as attentiveness, flexibility, and organization (Ziv, 2013), as well as resourcefulness, goal orientation, and planfulness (Chen et al., 2010;
Chen & McNamee, 2007). For the most part, studies consistently demonstrate that variables associated with the
approaches to learning domain have unique contributions to
children’s achievement beyond other important cognitive
and demographic variables such as intelligence, receptive
and expressive vocabulary, parental income, and education.
Related Definitions of Executive Functioning,
Self-Control, and “Everyday Behaviors”
The approaches to learning construct described above
shares a great deal in common with the research and literature on executive functions, the latter defined as involving
such metacognitive processes as found in working memory,
attention shifting, inhibitory control processes, planning,
error correction, resistance to interference, and memory
updating, to name a few (Blair, Zelazo, & Greenberg, 2005;
Carlson, 2005; Schmeichel & Tang, 2015, for additional terminology). As with the definitions of approaches to learning,
various terms are associated with executive functioning processes, which include such designations as “impulsivity,
conscientiousness, self-regulation, delay of gratification,
inattention-hyperactivity, executive function, willpower, and
intertemporal choice” (Moffit et al., 2011, p. 2693), descriptions that primarily reflect their disciplinary origin in the
medically oriented fields.
Attempting to bring some “ecological validity” or reallife applications to these constructs and behaviors, which
have been researched regularly in neuropsychological fields,
Galinsky (2010) has focused on the manifestation or applications of these executive functions in real life and in the
classroom as being critical life skills and has described them
as follows: focus and self-control, perspective taking, communicating, making connections, critical thinking, taking on
challenges, and self-directed and engaged learning.
Similarly, Isquith et al. (2004) have designated the practical
outcome of such executive function processes in working
memory or inhibitory self-control as “everyday behaviors”
with the claim that “the child’s everyday environments, both
at home and at school or day care, are important venues for
observing routine manifestations of the executive functions”
(p. 406). Other recent work to bring the esoteric, neurological, and clinical terminology into the public awareness is the
concept of “grit” described by Duckworth, Peterson,
Matthews, and Kelly (2007) as “perseverance and passion
for long-term goals” (p. 1087).
Thus, in comparing the executive function and approaches
to learning literatures (with few overlapping citations), the
latter body of research, although complementary to the executive function work, aims to focus on the malleability,
learning, and practical manifestations of behaviors related to
executive functioning rather than its actual nature, as related
to particular neurological and brain structures (see the special issue of Developmental Neuropsychology edited by
Blair et al., 2005, for a comprehensive review of this work).
It is clear from the rapid expansion of work on early learning
standards across the nation (Scott-Little et al., 2006; K.
Snow, 2011) that states have made dedicated efforts to outline the behaviors of young children, which will eventually
lead to successful achievement in kindergarten and beyond
(for recent examples, see Arizona Department of Education,
2005; Connecticut Early Learning and Development
Standards, 2014; North Carolina Division of Child Development,
2008).
However, common to both areas (EF and approaches to
learning) is the limited research on standardized, reliable,
easily administered assessments that practitioners can use to
gain information about very young children, in particular, in
this important area. Although there are a number of clinical
instruments or protocols to measure various aspects of EF
(Carlson, 2005), most of these are not designed for normal
classroom use, usually requiring one-on-one administration,
needing some props, and having limited psychometric stability. Thus, a consensus across all researchers examining
EF, approaches to learning, or other self-regulatory behavior
(e.g., McClelland, Acock, Piccinin, Rhea, & Stallings, 2013;
Poropat, 2014) is that there are few easily administered, psychometrically robust instruments available for teachers or
parents to use.
Approaches to Learning Instrumentation
Studies measuring the approaches to learning domain have
followed either one of two approaches by (a) using items initially developed for other instruments and purposes or (b)
developing dedicated instruments focused on approaches to
learning itself. An example of the first are secondary analyses
(e.g., Li-Grining et al., 2010) of the Early Childhood
Longitudinal Study–Kindergarten Cohort (1998–1999). This
study used a small subset of items taken originally from the
Social Skills Rating System (Gresham & Elliot, 1990) in
which parents and teachers rated a child’s behavior on aspects
of persistence in a task, curiosity, creativity, ability to concentrate, ability to work independently, and paying attention
(National Center for Educational Statistics, 2010).
The second approach includes instruments developed to
measure aspects of approaches to learning directly (e.g.,
Chen & McNamee, 2007; McDermott, Leigh, & Perry,
2002) and comprise rating scales of children’s behaviors in a
preschool classroom based on the frequency of various
activities (e.g., individual engagement with books or peer
interaction during play), as occurring very often, sometimes,
or never. For example, in their Bridging assessment, Chen
and McNamee (2007) measured approaches to learning
3
Barbu et al.
across different school tasks (e.g., moving to music, block
play, and counting) while rating children’s “Initial
Engagement,” “Focused Attention, Planfulness,” “Goal
Orientation,” and “Resourcefulness.” This assessment is
based on Vygotsky’s (1931/2012) sociocultural theory,
which infers that children’s behaviors may vary in the function of different task demands.
In contrast, McDermott et al.’s (2002) Preschool Learning
Behavior Scale (PLBS) is a 29-item instrument designed to
capture children’s overall behaviors across various activities
according to their “Competence/Motivation,” “Attention/
Persistence,” and “Attitude Toward Learning.” Developers
of both instruments designed them for early childhood and
kindergarten teachers to administer to children, but
McDermott et al.’s AtL instrument is the sole one that is
normed (see Fantuzzo, Perry, & McDermott, 2004).
Each of these instruments represents one of two major
philosophical views on how children’s learning may be
assessed. Bridging (Chen & McNamee, 2007) is a portfoliotype assessment, assessing a range of performances across
many activities over time and requiring an integrative, qualitative judgment related to the child’s development. The
Preschool Learning Behavior Scale (McDermott et al.,
2002), on the other hand, uses a quantitative rating scale to
index degrees of performance relative to previously identified latent factor structures. The merits of each view have
been debated extensively among early childhood researchers
and educators (Meisels & Fenichel, 1996; Pianta, Barnett,
Justice, & Sheridan, 2012). For example, Gilliam and Frede
(2012) stated that early childhood development is too “variable and rapid,” and children are “not consistent” in demonstrating their abilities for any test to capture a child’s ability
accurately in the brief assessment time. Nonetheless, the
selection of either a qualitative portfolio assessment or rating scale measures seems largely based on the user’s philosophical preferences, with those claiming a sociocultural
orientation leaning toward portfolio assessment and those
with more of a cognitive or psychometric perspective learning toward rating scale measures. Nonetheless, it is clear in
the research literature that the majority of studies examining
approaches to learning in young children favor the use of
rating scales that are examined via statistical analyses.
A standards-based measure of approaches to learning. The
new AtL assessment we describe below, as well as the
Devereux Early Childhood Assessment (DECA), fall into
the latter category and share the general advantages of teacher
and guardian rating scales in terms of amount of information
obtained, ease of administration, scoring, and prior reliability
and validity testing (see Campbell & James, 2007). Although
the DECA was designed primarily for early identification of
social and emotional problems in young children, the
Devereux Foundation (2003) has reported that the instrument
measures an approaches-to-learning construct. However, we
4
could not find published literature or research studies examining the psychometric properties of this aspect of the DECA
instrument or any descriptions from the Devereux Foundation of items dedicated to assessing approaches to learning
behaviors.
Related to the development of the shorter AtL instrument
described here, a recent review of research (Diamond et al.,
2013), funded by the Institute for Education Science (IES)
on early intervention and education, highlighted the need for
efficient, easily administered assessments with strong evidence of reliability and validity for use in research and educational applications. Diamond et al. (2013) suggested that
the research is inadequate in early assessment regarding
important characteristics, such as approaches to learning,
and additional research is needed to add to the knowledge
base in this arena.
The Current Study
In this article, we describe research regarding the psychometric properties of two assessments of approaches to learning, an important domain of school readiness. The context
for this study was an Arizona early childhood initiative, First
Things First (FTF), a comprehensive, statewide early childhood program intended to increase access to health and educational services for families and children, birth to age 5
years, and to improve preschoolers’ readiness for school.
The project was designed to include a researcher-developed
instrument created to assess approaches to learning as
defined by the AELS (Arizona Department of Education,
2005) and the widely used and standardized DECA (LeBuffe
& Naglieri, 1999a, 1999b).
The AtL scale was part of a demographic questionnaire in
a larger battery of readiness assessments that included direct
measures of language, literacy, math, and demographics
(e.g., height and weight). The researcher-designed items rating kindergarteners’ AtL were derived from the state’s early
learning standards, which specified that children should be
developing attitudes and behaviors characterizing initiative,
curiosity, engagement, persistence, reasoning, and problemsolving skills. The AtL instrument also was designed to
reduce participant response burden by decreasing time to
complete the assessment portfolio for each child and to
reduce overall cost in the next stage of the project.
The purpose of this study is to examine the psychometric
properties of the DECA and AtL assessments and to ascertain whether these instruments measure the same or different
approaches to learning domains based on parent/guardian
perceptions of their children, as defined by the AELS
(Arizona Department of Education, 2005) and the DECA
(LeBuffe & Naglieri, 1999a, 1999b). We accomplish these
goals by (a) providing evidence for validity and reliability of
the DECA and AtL instruments and presenting recent results
based on contemporary statistical techniques concerning
Assessing Approaches to Learning
DECA’s validity, (b) testing whether the parent/guardian
data from the DECA rating scale fit the hypothesized DECA
measurement model to determine the degree to which the
DECA’s model as a whole is consistent with the empirical
data, (c) developing confirmatory factor analysis (CFA)
models for DECA data by using the parceling procedure to
reduce the errors due to each item’s specificity and to prevent convergence difficulties when DECA and AtL models
were estimated simultaneously, and (d) evaluating the combined DECA-AtL model and examining the relationship
between DECA and AtL questionnaires.
the teacher sent home the child with a packet containing a
letter explaining the purpose of the study, a parent consent
form, and a child assent form. The research staff then called
parents to confirm they received the packet, further explained
the study, responded to any questions, and invited them to
participate.
Instrumentation
The parents/guardians of participating students completed two questionnaires: the newly developed AtL instrument and the DECA.
Method
Sample
We used a proportional, stratified random sampling
approach that included 1,145 kindergarten children attending
82 schools drawn from 48 districts across the state of Arizona.
We used this sampling design to ensure that the sample was
randomly selected and representative of the population. We
randomly selected the sample at the child and school level
that included three types of schools (public, private, and charter), from three different regions (northern, central, and
southern). This sample distribution, derived across the type
of schools and regions, was in close agreement with the state
proportions (see Barbu, Levine-Donnerstein, Marx, & Yaden,
2012; Barbu, Marx, Yaden, & Levine-Donnerstein, 2015;
Yaden et al., 2011).
The average age of the children in the sample was 5 years,
8 months with an age range from 5 to 6 years, 7 months.
Fifty-one percent of the children were male, with 49.1% of
all children identified as Hispanic. In the present sample,
5.5% of the children had an individualized educational plan,
which is required for children with identified special needs,
51.3% were eligible for free or reduced-price lunch, and
73.2% had attended an out-of-home child care, nursery
school, or pre-kindergarten program, including Head Start
(Barbu et al., 2012). For the analyses, the rating scales from
1,025 families with 1,145 children were examined (some
families had multiple children in the sample).
Procedures
For each participating school, a directory of first-time
kindergarten students was obtained. From these directories,
14 first-time kindergarten children were randomly selected
per school or from one class if only one class existed. From
schools with two kindergarten classrooms, seven children
were randomly selected from each classroom; from schools
with three classrooms, five children per classroom were randomly drawn; and from schools with more than three classrooms, first, three classes were randomly selected, and then
the targeted number of children (five per classroom) was
randomly drawn from each. After children were identified,
AtL rating scale. The newly developed AtL rating scale
included 13 items derived from the state’s early learning
guidelines. To illustrate its alignment with state standards
and the DECA, Table 1 provides a comparison of the AtL
items, the learning indicators from the AELS (Arizona
Department of Education, 2005), and the selected items
from the DECA with similarly worded items. The AtL instrument employs a 4-point scale: (a) proficient (coded as 3),
child demonstrates skill, knowledge, or behavior consistency (i.e., regularly); (b) in progress (coded as 2), child
demonstrates skill, knowledge, or behavior with some regularity; (c) not yet (coded as 1), child cannot perform the skill,
knowledge, or behavior; and (d) don’t know (coded as 0).
In a previous analysis, we investigated the psychometric
properties of the AtL instrument (Barbu et al., 2015) and
found a one-factor structure via exploratory factor analysis
(EFA) and CFA. Moreover, early childhood educators and
researchers from a tri-university research consortium
(University of Arizona, Arizona State University, and
Northern Arizona University) assessed the face validity and
content validity of this rating scale. These results, combined
with evidence of reliability (0.83 for latent variables and
ranging between 0.52 and 0.79 for manifest variables) and
structural validity of the instrument (Barbu et al., 2015), supported the educational utility of the AtL as a tool for measuring school readiness among kindergarteners with a
population similar to children in Arizona.
DECA. For comparison purposes with the researcher-developed AtL instrument and to index broader socioemotional
characteristics in the larger study, we administered the
DECA, a standardized, norm-referenced assessment of
within-child protective factors, designed originally to identify resilience in children ages 2 to 6 (LeBuffe & Naglieri,
1999a, 1999b). The DECA’s 37 items are measured on a
5-point Likert scale indexing behaviors as occurring never,
rarely, occasionally, frequently, or very frequently. They are
designed to measure four distinguishable latent variables:
initiative (IN—11 items), self-control (SC—8 items), attachment (AT—8 items), and behavioral concerns (BC—10
items). Used in practice, DECA provides an option for
5
Barbu et al.
TABLE 1
Comparisons Between the 13-Item Approaches to Learning (AtL) Scale, Statements From the Social Emotional Domain of the Arizona
Early Learning Standards,a and Selected Items of the Devereux Early Childhood Assessment.
AtL Scale Item
Arizona Early Learning Standards
Indicator/Strand/Concept
1. Sustains positive interactions with other Initiates and sustains positive interactions with adults and
children (e.g., When doing a puzzle, child friends.
asks if he can help. The children finish Strand 2: Social interactions with others
Concept 2: Cooperation
the puzzle together).
2. Sustains positive interactions with
Initiates and sustains positive interactions with adults and
familiar adults.
friends.
Strand 2: Social interactions with others
Concept 2: Cooperation
3. Has friends.
Child sees his friend crying and then gives her a hug.b
Strand 1: Knowledge of self
Concept 2: Recognition and expression of feelings
Child trades toys with a friend.
Strand 2: Social interactions with others
Concept 2: Cooperation
Child reminds friends that running is for outside.
Strand 3: Responsibility for self and others
Concept: Self-control
Child inquires why his friend is not at school.
Strand 4: Approaches to learning
Concept: Curiosity
Adjusts behavior for alternate activities and in different
4. Adjusts behavior to correspond to
different settings (e.g., child knows when settings of the learning environment.
Strand 3: Responsibility for self and others
to use a “quiet voice”).
Concept 1: Self-control
5. Follows rules.
Understands and follows rules in the learning environment.
Strand 3: Responsibility for self and others
Concept 1: Self-control
6. Manages transitions (e.g., When it is time Manages transitions, daily routines, and unexpected events.
Strand 3: Responsibility for self and others
for a story, child puts away the blocks
Concept 1: Self-control
and goes to hear the story).
7. Shows curiosity as a learner.
Shows interest in learning new things and trying new
experiences.
Strand 4: Approaches to learning
Concept 1: Curiosity
Makes decisions independently.
8. Makes independent decisions (e.g.,
instead of playing with friends, the child Strand 4: Approaches to learning
Concept 2: Initiative
decides to read a story).
9. Attends to tasks (e.g., child works on
building a Lego structure throughout the
course of the day).
10. Seeks help when encountering a
problem (e.g., child tells adult, “He took
my toy.”).
11. Copes with frustration (e.g., child says,
“We have to go inside, it’s raining. We
can come back out when it stops.”).
12. Takes risks during learning situations.
13. Shows respect for toys.
Devereux Early Childhood
Assessment Item: Subscale
33. Cooperate with others?
Subscale: Self-Control
9. Touch children/adults inappropriately?
Subscale: Behavioral Concern
10. Show affection for familiar adults?
Subscale: Attachment
9. Touch children/adults inappropriately?
Subscale: Behavioral Concern
20. Start or organize play with other children?
Subscale: Initiative
25. Share with other children?
Subscale: Self-Control
30. Accept another choice when his or her first
choice was unavailable.
Subscale: Self-Control
4. Listen to or respect others?
Subscale: Self-Control
30. Accept another choice when his or her first
choice was unavailable.
Subscale: Self-Control
19. Try or ask to try new things or activities?
Subscale: Initiative
36. Make decisions for himself or herself?
Subscale: Initiative
2. Does things for himself or herself?
Subscale: Initiative
24. Focus his or her attention or concentrate on a
task or activity?
Subscale: Initiative
31. Seeks help from adults/children when
necessary?
Subscale: Attachment
13. Handle frustration well?
Subscale: Self-Control
Continuously attends to a task.
Strand 4: Approaches to learning
Concept 3: Persistence
Seeks adult assistance when support is required.
Strand 4: Approaches to learning
Concept 5: Problem solving
Copes with frustration or disappointment.
Strand 4: Approaches to learning
Concept 3: Persistence
Is willing to take risks and consider a variety of alternatives. 12. Takes risks during learning situations.
Strand 4: Approaches to learning
Concept 6: Confidence
13. Shows respect for toys.
Shows respect for learning materials and toys.
Strand 3: Responsibility for self and others
Concept 2: Respect
a
During 2009 when the study was conducted, the second edition of the Arizona Early Learning Standards (Arizona Department of Education, 2005) was in
force. In this document, approaches to learning was one of four strands in the social and emotional standard. In the third edition (2013) of the standards, the
approaches to learning strand was recast as a standard itself. No new wording or descriptions were added in 2013.
b
There is no specific wording in the social-emotional domain about “having friends” per se, but multiple statements for all indicators occur across all strands
indicating that the child interacts with friends in various settings and situations.
6
Assessing Approaches to Learning
researchers to further group together the first three factors
(i.e., IN, SC, and AT) into a total protective factor (TPF) that
evaluates the frequency of positive behaviors. The TPF measures positive child outcomes and includes two domains:
social and emotional development and approaches to learning (Devereux Foundation, 2003). The instrument is used for
screening, progress monitoring (e.g., DECA can be administer two to three times per year), intervention planning, and
research (Henderson & Strain, 2009).
The DECA is considered a reliable instrument for assessing children’s protective factors (LeBuffe & Naglieri,
1999b), based on research findings from infants to preschoolers (Barbu et al., 2012; Brinkman, Wigent, Tomac,
Pham, & Carlson, 2007; Buhs, 2003; Bulotsky-Shearer,
Fernandez, & Rainelli, 2013; Chittooran, 2003; Jaberg,
Dixon, & Weis, 2009; Lien & Carlson, 2009; Meyer, 2008;
Ogg, Brinkman, Dedrick, & Carlson, 2010).
Although the DECA measures a range of important constructs, recent research using structural equation modeling
techniques has found issues related to DECA’s validity that
may limit its use; therefore, researchers have recommended
further investigation of DECA’s psychometric properties.
For example, Ogg et al. (2010) found 10 problematic item
pairs as a source of poor model fit of the DECA instrument
in their sample of 1,344 participants recruited from 25 Head
Start centers across a four-county region in the Midwest. In
addition, Barbu et al. (2012) found insufficient discriminant
validity of the DECA instrument based on samples of parents’ and teachers’ ratings of 1,145 entering kindergartners
in the Southwest. More recently, Bulotsky-Shearer et al.
(2013) found that DECA’s factor structure was not adequate
using a large sample of culturally and linguistically diverse
Head Start children (N = 5,197) in the Southeast.
A possible explanation for the departure of these recent
conclusions from the results reported in DECA’s technical
manual (LeBuffe & Naglieri, 1999a) and obtained with classical statistical analyses is that the traditional multivariate
procedures are incapable of assessing and correcting for
measurement errors (Byrne, 2009). Therefore, correlations
between two latent variables in structural equation modeling
(SEM) could be more than twice that for the individual
observed variables analyzed with classical statistics, and
hidden effects among latent variables (i.e., multicollinearity)
could remain undetected (Bollen, 1989).
Data Analysis
We approached the data analysis in three phases. First, we
developed CFA models for our sample DECA data based on
the theoretical structure explained above. The purpose of
assessing a model’s overall fit is to determine the degree to
which the model as a whole is consistent with the empirical
data. The null hypothesis states that the model fits the population data perfectly, and the aim is not to reject this null
hypothesis. Although a wide range of goodness-of-fit indices
have been developed to provide measures of a model’s overall fit, one is not superior to the others in all situations (Brown,
2006). In addressing the ordinal nature of the observed variables, Byrne (2009) suggested five goodness-of-fit indices to
test and respecify the hypothesized model: chi-square (χ2),
root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), comparative fit index (CFI), and
expected cross-validation index (ECVI). The following cutoff criteria (Hu & Bentler, 1999) were used as guidelines for
goodness-of-fit indices between the target model and the
observed data: (a) GFI values close to 0.95 or greater, (b)
RMSEA values close to 0.06 or below, and (c) CFI values
close to 0.95 or greater. In addition, the model with the smallest ECVI value suggests that the hypothesized model is well
fitted and represents a reasonable approximation to the population (Browne & Cudeck, 1993).
Second, we developed CFA models for our sample DECA
data by constructing parcels based on T. D. Little,
Cunningham, Shahar, and Widaman’s (2002) technique and
confirmed the correctness of each parcel by calculating their
internal consistency. A parcel is an aggregated indicator
composed of the sum or average of several items that measure the same construct. Our decision of how to select these
indicators for each parcel was based on the loadings obtained
for each factor, such that the coefficients’ center of mass
would be preserved. First, we identified the three items with
the highest loadings to anchor the three-parcel solution for
each factor (Nasser & Wisenbaker, 2003). Second, we added
three items with the next highest item-to-factor loadings to
the anchors in an inverted order such as the highest loaded
item from among the anchor items would be matched with
the lowest loaded item from among the second selections.
Finally, we placed lower loaded items with higher loaded
parcels. Thus, parcels resulting from this technique had a
different number of items, and they were selected to achieve
a reasonable balance, in agreement with the method presented above (T. D. Little et al., 2002).
In the present analysis, a parceling approach for the DECA
instrument was appropriate for two reasons. First, the errors
were reduced in the final variance-covariance matrix due to
each item’s specificity, in addition to the randomness of each
item’s errors. Therefore, the value of GFI was expected to
improve with a correct selection of factors forming each parcel (T. D. Little et al., 2002). Second, factor structures can be
difficult to determine when analyzing individual items from
a lengthy questionnaire. Practical convergence difficulties
could arise in LISREL 8.80 when analyzing a larger number
of manifest variables. In our case, this occurred when we
added 13 more items from AtL to our sample DECA data,
which already contained 37 measured variables.
The potential advantages of using parcels include the following: (a) Parcels better approximate normality than individual items (Brown, 2006), (b) they improve reliability and
relationships with other variables (Kishton & Widaman,
1994), and (c) models based on parcels may be less complex
7
Barbu et al.
than models based on individual items (i.e., fewer parameters,
smaller input matrix) (Brown, 2006). However, a variety of
disadvantages could exist in the following cases: (a) The
assumption of unidimensionality is not met (i.e., each indicator loads on a single factor and the error terms are independent), (b) the likelihood of improper or nonconvergent
solutions increases as the number of parcels decreases (Nasser
& Wisenbaker, 2003), and (c) the use of parcels may not be
feasible in situations in which too few items form a sufficient
number of parcels (Brown, 2006). In our case, the parceling
was a vital solution in solving the problem of convergence
when DECA and AtL models were estimated simultaneously.
Third, we examined the relationship between DECA and
AtL questionnaires to ascertain the extent to which these
instruments measured the same or different approach-tolearning domains. In this phase, the parameters for both
DECA and AtL models were estimated simultaneously, and
the hypothesized structure was analyzed as a function of the
overall goodness of fit for the combined model, followed by
an examination of standardized residuals. We investigated
the correlations among latent variables for the completely
standardized solution, because the two tests (i.e., DECA and
AtL) were measured using two different Likert scales.
across both missing data treatment methods; therefore, we
reported the MCMC imputation results in agreement with
data analyses in the larger project from which these data were
drawn (Yaden et al., 2011). In our case, Little’s MCAR test,
χ2(535, 984) = 951.11, p < .001, based on the evaluation of
the homogeneity of the available means for different patterns
of incomplete data (R. J. A. Little, 1988), was statistically
significant, indicating that the data for DECA and AtL instruments were not missing completely at random. However, the
Separate Variance Test indicated no relationship between
missingness and variables, and thus we proceeded under
MAR assumptions (Tabachnick & Fidell, 2013).
The original sample of 2,290 DECA and AtL questionnaires completed by the parents/guardians of 1,145 participating students was reduced to 2,075 (1,053 for DECA and
1,022 for AtL) by eliminating the participants without
records. Consequently, data from participants with at least
one response were imputed to complete the missing data,
using the MI procedure (Rubin, 1987) under the MAR
assumption, with a WLS estimator to correct the violation of
multivariate normality. Finally, we identified 984 parent/
guardian questionnaires for both DECA and AtL after eliminating the participants with only one questionnaire (i.e., one
DECA or one AtL questionnaire).
Results
Data Missingness
DECA Model Analysis
For the current study, less than 3% of the data for guardians were missing, and we concluded that these results raised
no concerns regarding data missingness. Data in this study
were ordinal based on DECA’s five rating categories and
AtL’s three rating categories for measurement. As a result,
we generated polychoric correlation matrices using PRELIS
2.0. Two possible approaches are implemented in PRELIS
2.0 when data are missing: (a) the expectation maximization
algorithm (EM; Dempster, Laird, & Rubin, 1977) and (b)
Markov chain Monte Carlo method (MCMC; Gilks,
Richardson, & Spiegelhalter, 1995). Collectively known as
multiple imputation (MI), these procedures replace the missing values across multiple variables under the assumptions
of either data missing at random (MAR) or data missing
completely at random (MCAR) and multivariate normality.
The assumption of multivariate normality of observed data
is usually violated when the outcomes are rated on Likerttype scales (Lubke & Muthén, 2004), but the weighted least
squares (WLS) estimator adjusts for this violation in large
samples with categorical outcomes (Brown, 2006).
We used MCMC imputation technique and full-information maximum likelihood (FIML) estimation to correct for
missing data and evaluate the consistency between the missing data treatment methods. The MCMC imputations were
based on generation of five imputed data sets. The parameter
estimates (i.e., factor loadings, available model fit statistics,
etc.) for MCMC imputation and FIML were nearly identical
In the first model (M1) for the DECA (see Table 2), we
followed the official prescription in assigning the indicators
for each of the four latent variables considered. The variance
of each latent variable was fixed to 1. The minimum fit function value was χ2(627, n = 1,053) = 2,764.16. The values of
RMSEA (.057), ECVI (2.78), and CFI (.97) indicated a reasonable fit of the model. Still, the value of the GFI (.74) was
low. Further attempts to improve the model by freeing the
covariances among the measured indicator errors, as suggested by the modification indices, did not provide a satisfactory improvement of the GFI measure. Consequently, we
used item parcels instead of individual items.
In parceling the DECA model, we averaged two, three, or
four original indicators together to generate a new indicator
(Bollen, 1989) and reduced the number of indicators for each
of the four factors to three. Table 3 shows a detailed diagram
of this item parceling selection process. Internal consistency
reliability of the parcels was acceptable, ranging from .64 for
the DECA 2 parcel to .87 for the DECA 12 parcel. In addition, the normality of the parcels indicated that the values for
skewness were less than an absolute value of sk = 1.92, and
the values for kurtosis were less than absolute value of kur =
6.13. Neither skewness nor kurtosis exceeded recommended
cutoffs, |2.00| and |7.00| (Curran, West, & Finch, 1996),
respectively, indicating adequate univariate normality.
We formed and investigated two parcel models. The first
model represented a parceled model consisting of four
8
Assessing Approaches to Learning
TABLE 2
Fit Statistics of the Devereux Early Childhood Assessment Instrument for Guardians.
Goodness of Fit
Model
M1
M2
M3
M4
M5
χ2
df
RMSEA
CFI
GFI
ECVI
∆χ2
2764.16
352.09
352.09
281.19
200.66
627
48
48
47
46
.057
.078
.078
.069
.057
.97
.97
.97
.97
.98
.74
.98
.98
.99
.99
2.78
.39
.39
.33
.25
70.90
80.53
Note: χ2 = chi-square fit statistic; RMSEA = root mean square error of approximation; CFI = comparative fit index; GFI = goodness-of-fit index; ECVI =
expected cross-validation index; M1 = original model; M2 = model with four exogenous latent variables; M3 = model with two exogenous and three endogenous latent factors; M4 = model with item parcel DECA_7 loaded on attachment (AT) and self-control (SC); M5 = model with item parcels DECA_7 and
DECA_8 loaded on AT and SC.
exogenous latent variables with three indicators per variable
(M2). The second model included a structural model with
two exogenous and three endogenous latent factors (M3).
The fitting parameters for both models were similar: RMSEA
(.078), ECVI (.39), CFI (.97), GFI (.98), and χ2(48, n =
1,053) = 352.09. Since including a structural relationship did
not lead to an improvement in the fitting of the original
model, this more complex model (M3) was abandoned, and
the next step consisted of improving the fitting of model M2.
Further examination of the modification indices suggested
that some pairs of parcels shared similar content. Therefore,
we inspected two nested models M4 and M5 (see Table 2).
The results indicated that the item parcels DECA_7 and
DECA_8 in M5 were dependent not only on the AT latent
variable as initially proposed but also on the SC variable.
Inspection of the goodness-of-fit indices RMSEA (.057),
ECVI (.25), CFI (.99), and GFI (.99) indicated a good model
fit (see Figure 1). In addition, inspection of the stem-leaf plot
of the standardized residuals revealed a symmetrically clustered distribution around zero, accounting for a reasonably
well-fitted model.
Therefore, on the basis of the fit statistics, in conjunction
with the following evidence—(a) a decrease of the ECVI
value, (b) the distribution of the standardized residual values, and (c) a significant reduction in χ2 between models M4
and M5, ∆χ2(1, n = 1,053) = 281.19 – 200.66 = 80.53—we
concluded that the parent/guardian data fit the hypothesized
DECA measurement model and considered model M5 to
represent the best model fit of the DECA instrument.
Combined DECA-AtL Model Analysis
Figure 2 presents the standardized factor intercorrelations, factor loadings, and residuals for the combined AtLDECA model for parent/guardian data. The overall
goodness-of-fit indices for this model, χ2(261, n = 984) =
925.08, RMSEA = .051, CFI = .97, and GFI =.98, indicated
a good model fit. This outcome was supported by standardized residuals below 2.00 and standardized factor loadings
below 1.00 (Brown, 2006). Further attempts of improving
the model fit by allowing the measurement errors to correlate did not result in a considerable change, and thus, the
nested models were discarded.
First, the BC measure correlated negatively with all the
other factors from both tests (i.e., IN, SC, AT, and AtL). This
result was consistent with the way in which the instruments
were constructed. The items measuring the BC factor were
designed to address developmental factors that are contraindicative of the factors in the other scales. Therefore, items
indicating large behavioral concerns were designed as indicators of a child’s low socioemotional development.
Second, correlations between the latent factor of the AtL
instrument and latent factors of the DECA instrument ranged
between .74 and .85 in absolute value, r (AtL, BC) = –.74, r
(AtL, IN) = .79, r (AtL, AT) = .83, and r (AtL, SC) = .85.
These results indicated that the AtL explained between 55%
and 72% of the variance in testing the DECA, and thus, the
four latent factors of the DECA instrument were strongly
related to the factor of the AtL instrument.
Discussion
This is one of the first studies that examined an assessment developed specifically to reflect children’s performance on a particular state standard related to approaches to
learning. Thus, aligned with the need for valid and reliable
instruments that are easy for parents/guardians to complete
(Diamond et al., 2013), we developed a 13-item instrument
based on the AELS (Arizona Department of Education,
2005) that examined approaches-to-learning behaviors
among kindergarteners. Additionally, this is one of the first
research efforts of which we know that investigated the
DECA’s approaches-to-learning dimension. Although
DECA is a widely used tool that measures the
9
Barbu et al.
TABLE 3
Item Parcel Selection for the DECA Data.
.38
Old Item
New Item
IN
DECA3
DECA12
DECA19
DECA2
DECA7
DECA16
DECA20
DECA24
DECA28
DECA32
DECA36
DECA4
DECA5
DECA34
DECA13
DECA25
DECA21
DECA30
DECA33
DECA1
DECA10
DECA6
DECA17
DECA29
DECA22
DECA31
DECA37
DECA8
DECA11
DECA18
DECA9
DECA15
DECA23
DECA14
DECA26
DECA27
DECA35
DECA_1
AT
BC
.26
DE CA2
.28
DE CA3
IN
.91
.89
.52
.68
DECA_2
.32
DE CA4
1*
.36
DE CA5
.27
DE CA6
SC
.97
.21
.94
DECA_3
.58
-.22
.91
.93
.30
DE CA7
1*
DECA_4
.35
AT
-.33
.71
DE CA8
.27
-.22
.25
DE CA9
.55
DE CA10
.36
DECA_5
DECA_6
1*
.99
.56
BC
DE CA11
.45
DECA_7
DECA_8
DECA_9
DECA_10
.90
.36
DE CA12
FIGURE 1. Standardized factor intercorrelations, factor
loadings, and residuals of the Devereux Early Childhood
Assessment (DECA) model after parceling (M5) for parent/
guardian samples. Overall fit of the M5 model: χ2(46, n = 1,053)
= 200.66, comparative fit index = .98, goodness-of-fit index =
.99, and root mean square error of approximation = .057. IN =
initiative; SC = self-control; AT = attachment; BC = behavioral
concerns.
DECA_11
DECA_12
Note: DECA = Devereux Early Childhood Assessment; IN = initiative; SC
= self-control; AT = attachment; BC = behavioral concerns.
socioemotional resilience of young children, we have found
no studies in the literature that explore the approaches-tolearning component of the instrument.
In previous analyses (Barbu et al., 2015), we found that
the 13-item AtL rating scale has a valid and reliable onefactor structure, and it can be reliably used to assess children’s approaches-to-learning behaviors. In this study, our
purpose was to ascertain the extent to which AtL and DECA
10
.62
1*
Factor Name
SC
DE CA1
instruments measured similar approaches-to-learning
domains (Arizona Department of Education, 2005; LeBuffe
& Naglieri, 1999a, 1999b) based on parent/guardian’s perceptions of their children. Through an analysis of a combined DECA-AtL model of the two instruments, we found
that although their Likert rating scales differed, they measured a similar approaches-to-learning domain, and thus,
used simultaneously, these instruments are redundant. Our
findings also supported the content validity of the AtL instrument, being based on a comparison of the AtL items with
those of the AELS and DECA (see Table 1) and an analysis
at the construct level between DECA and AtL instruments.
Unfortunately, an analysis at the item level between these
two instruments was not possible because DECA’s technical
manual (LeBuffe & Naglieri, 1999a, 1999b) does not identify
which of its items is included in the approaches-to-learning
domain. Moreover, our findings that the DECA subscales
(i.e., self-control, initiative, attachment, and behavioral concerns) strongly correlated with AtL calls into question their
Assessing Approaches to Learning
DECA’s parent/guardian empirical data and its hypothesized
measurement model, which indicated that DECA’s model as
a whole was consistent with these data; and (c) to demonstrate evidence that AtL and DECA questionnaires correlate
and both measure the approaches-to-learning domain.
Moreover, using CFA with the parceling technique when
simultaneously estimating the DECA and AtL models presents a robust method of convergence. This analytical
approach provides an approach for researchers to find new
dimensions of the approaches-to-learning domain in existent
questionnaires in which this construct is not apparent.
Limitations
FIGURE 2. Standardized factor intercorrelations, factor
loadings, and residuals of the Devereux Early Childhood
Assessment (DECA)–approaches to learning (AtL) combined
model for parent/guardian samples. Overall fit of the DECA and
AtL combined model: χ2(261, n = 984) = 925.08, comparative fit
index = .97, goodness-of-fit index =.98, and root mean square
error of approximation = .051. IN = initiative; SC = self-control;
AT = attachment; BC = behavioral concerns.
uniqueness in measuring broader aspects of social and emotional development, in addition to the more narrow
approaches-to-learning construct, which it states to measure
(Devereux Foundation, 2003). In a previous analysis (Barbu
et al., 2012), we found that the DECA subscales of self-control, initiative, and attachment lacked discriminant validity;
therefore, scores on these subscales should be viewed with
great caution regarding making decisions about children’s
performances or their resilience. Thus, we invite the DECA’s
authors to test and reevaluate the content and structural
validity of their instrument and encourage research to
explore the approaches-to-learning dimension of the DECA
instrument in behavior and psychoeducational studies.
The statistical procedures used in this study provided an
improved set of techniques: (a) to appropriately mitigate
missing parent/guardian data and increase sample size; (b) to
obtain the most accurate estimated model fit between
Three potential sources of limitations are highlighted in
this study. First, the use of the item parceling procedure
could affect the structural parameter estimates and might
lead to a biased estimate of model parameters in some situations, while allowing for a slight improvement of the model
fitting. However, researchers have demonstrated that parcels
do not outperform models based on individual items (Hau &
Marsh, 2004). Therefore, we concluded that this technique
was appropriate for our study.
A second limitation consists in data composed of ordinal
measurement scales. Although Likert scales are widely used
methods of capturing ratings from respondents in the social
and behavior sciences, they produce imprecise response
measures from a restrained number of categories (e.g.,
5-point rating scales); thus, information might be lost due to
the limited resolution of categories (i.e., less precision)
(Neibecker, 1984). In our case, considering that DECA had
five measurement categories and AtL included four, the
presence or absence of any additional categories in Likert
scales could lead to different results.
Finally, the Don’t know and Not yet categories of the AtL
rating scale could lead to misinterpretations of the results if
the difference between categories is not distinguished or the
categories are not coded appropriately. The Don’t know category indicated that the observer did not have enough information about the child or relevant settings to determine
whether the target behavior had occurred, whereas Not yet
indicated that the behavior was not observed in situations
where the rater had enough information about the child to
recognize it. In a preliminary analysis, we eliminated the
Don’t know category to investigate the coder effect, and we
found no significant difference between results. However,
this aspect should be examined empirically in future studies.
Conclusion
Researchers and program evaluators charged with assessing and evaluating early childhood programs have little
guidance in choosing what assessments might be appropriate for measuring early learning standards for particular
populations of children in different geographical areas. Also,
11
Barbu et al.
while standardized instruments aimed to measure various
abilities in the socioemotional domain exist (Denham, 2006;
Denham, Wyatt, Bassett, Echeverria, & Knox, 2009), the
main purpose of these instruments is primarily clinical. They
have not been designed for large-scale administration or to
address children’s readiness for kindergarten, as indicated
by their progression on specific early learning standards.
In a recent review of research, funded by the Institute of
Education Sciences from 2002 to 2006, Diamond et al.
(2013) strongly recommended that “research identifying
valid and reliable ways to measure children’s skills and capture their learning over time is greatly needed,” and furthermore, “there is a need to develop tools that can be readily
used within everyday educational settings by teachers and
other practitioners” (p. 29). Similarly, early childhood
researchers associated with the Center on the Developing
Child at Harvard University (2011) in their report, Building
the Brain’s “Air Traffic” Control System, stress the need for
policy makers, parents, caregivers, and researchers to have
access to accurate, reliable, and practical measurements of
these emergent self-regulatory behaviors such that the planning of intervention programs, normal curricula, and
research is based on valid, sensitive instrumentation that
captures the variability in these early manifestations of
approaches to learning. We heartily concur with such recommendations and suggest that researchers norm assessments
for young children with diverse backgrounds in specific geographical areas as past findings (Diamond et al., 2013) indicated inconsistent results for socioemotional and related
domains, such as approaches to learning.
Our goal was to create a statistically robust, easily administered assessment aimed to effectively measure the
approaches-to-learning domain based on Arizona’s early
learning standards, which are similar to the standards
adopted by many other states. The question of generalizability to populations in other states and regions, in addition to
assessments derived from other state standards, remains
open. The response to these questions requires further
research to ascertain AtL’s external validity beyond our sample, based on teacher and parent ratings. Moreover, our
intent was to examine and determine whether the DECA and
AtL instruments measured similar approaches-to-learning
domains based on parent/guardian perceptions of children.
In our next analysis, data from teachers’ responses to both
instruments will be examined to ascertain whether the relationship between DECA and AtL’s instruments is retained.
We suggest that the AtL questionnaire, with its 13 items,
serves as a more efficient and accurate tool for data collection of the approaches-to-learning domain than the DECA.
However, we need to test this instrument among populations
in other states with learning standards that similarly align
with those from Arizona.
In addition, future research should evaluate AtL’s efficacy against other instruments (i.e., Bridging assessment
12
and PLBS) that assess children’s learning behavior and to
explore its generalizability in measuring the approaches-tolearning domain as an important component of school readiness. At the same time, we invite researchers and practitioners
to test the AtL instrument with populations different from or
similar to children in Arizona and to evaluate the psychometric properties of this promising tool in behavior and psychoeducational studies.
References
Arizona Department of Education. (2005). Arizona early learning
standards. Phoenix, AZ: Author. Retrieved from http://www
.azed.gov/wpcontent/uploads/PDF/EarlyLearningStandards.pdf
Barbu, O.C., Levine-Donnerstein, D., Marx R.W., Yaden D.B.
(2012). Reliability and Validity of the Devereux Early
Childhood Assessment (DECA) as a Function of Parent and
Teacher Ratings. Journal of Psychoeducational Assessment,
31(5), 469–481.
Barbu, O.C., Marx R.W, Yaden D.B., & Levine-Donnerstein, D.
(in press). Measuring Approaches to Learning in Preschoolers:
Validating the Structure of an Instrument for Teachers and
Parents. Education, 3–13.
Blair, C., Zelazo, P. D., & Greenberg, M. T. (2005). The measurement of executive functioning in early childhood.
Developmental Neuropsychology, 28, 561–571.
Bollen, K. A. (1989). Structural equations with latent variables.
New York, NY: John Wiley.
Books, S. (2004). Poverty and schooling in the U.S.: Contexts and
consequences. Mahwah, NJ: Lawrence Erlbaum.
Brinkman, T. M., Wigent, C. A., Tomac, R. A., Pham, A. V., &
Carlson, J. S. (2007). Using the Devereux Early Childhood
Assessment to identify behavioral risk and protective factors
within a Head Start population. Canadian Journal of School
Psychology, 22(2), 136–151.
Brito, P. R. (2012). School readiness: A conceptual frame. New
York, NY: United Nations Children’s Fund.
Brito, P. R., & Lumlingan, M. C. (2012). School readiness and
transitions. New York, NY: United Nations Children’s Fund.
Brown, T. A. (2006). Confirmatory factor analysis for applied
research. New York, NY: Guilford.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 445–455). Newbury Park, CA: Sage.
Buhs, E. S. (2003). Devereux Early Childhood Assessment. In
B. S. Plake, J. C. Impara, & R. A. Spies (Eds.), The fifteenth
mental measurements yearbook. Lincoln, NE: Buros Institute
of Mental Measurements.
Bulotsky-Shearer, R. J., Fernandez, V., Dominguez, X., & Rouse,
H. L. (2011). Behavior problems in learning activities and
social interactions in Head Start classrooms and early reading,
mathematics, and approaches to learning. School Psychology
Review, 40(1), 39–56.
Bulotsky-Shearer, R. J., Fernandez, V., & Rainelli, S. (2013). The
validity of the Devereux Early Childhood Assessment for culturally and linguistically diverse Head Start children. Early
Childhood Research Quarterly, 28, 794–807.
Assessing Approaches to Learning
Byrne, B. M. (2009). Structural equation modeling with LISREL,
PRELIS, and SIMPLIS. Mahwah, NJ: Lawrence Erlbaum.
Campbell, J. M., & James, C. L. (2007). Assessment of social and
emotional development in preschool children. In B. A. Bracken
& R. J. Nagle (Eds.), The psychoeducational assessment of preschool children (pp. 111–135). Mahwah, NJ: Lawrence Erlbaum.
Carlson, S. M. (2005). Developmentally sensitive measures of
executive function in preschool children. Developmental
Neuropsychology, 28, 595–616.
Center on the Developing Child at Harvard University. (2011).
Building the brain’s “air Traffic control” system: How early
experiences shape the development of executive function
(Working Paper No. 11). Retrieved from www.developing
child.harvard.edu
Chain, J., Dopp, A., Smith, G., Woodland, S., & LeBuffe, P.
(2010). The Devereux Early Childhood Assessment literature review. Retrieved from http://www.devereux.org/site/
PageServer?pagename=deci_research
Chen, J. Q., Masur, A., & McNamee, G. (2010). Young children’s
approaches to learning: A socio-cultural approach. Early Child
Development and Care, 181, 1137–1152.
Chen, J. Q., & McNamee, G. D. (Eds.). (2007). Bridging:
Assessment for teaching and learning in early childhood classrooms, PreK–3. Thousand Oaks, CA: Corwin Press.
Chittooran, M. M. (2003). Devereux Early Childhood Assessment.
In B. S. Plake, J. C. Impara, & R. A. Spies (Eds.), The fifteenth
mental measurements yearbook. Lincoln, NE: Buros Institute of
Mental Measurements.
Connecticutt Early Learning and Development Standards: What
children, birth to five, should know and be able to do (2014).
Retrieved from www.ct.gov/oec/elds.
Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of
test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29.
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum
likelihood from incomplete data via the EM algorithm, Journal
of the Royal Statistical Society, 39, 1–38.
Denham, S. A. (2006). Social emotional competence as support for
school readiness: What is it and how do we assess it? Early
Education & Development, 17, 57–89.
Denham, S. A., Wyatt, T. M., Bassett, H. H., Echeverria, D., &
Knox, S. S. (2009). Assessing social-emotional development in children from a longitudinal perspective. Journal of
Epidemiology and Community Health, 63(11), 137–152.
Devereux Foundation, DECI. (2003). The Search Institute’s 40
developmental assets in conjunction with the Devereux Early
Childhood Assessment (DECA) Program. Retrieved from
http://www.devereux.org/site/DocServer/crosswalks-40Assets.
doc?docID=3361
Diamond, K. E., Justice, L. M., Siegler, R. S., & Snyder, P. A.
(2013). Synthesis of IES research on early intervention and
early childhood education (NCSER 2013–3001). Washington,
DC: National Center for Special Education Research, Institute
of Education Sciences, U.S. Department of Education.
Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R.
(2007). Grit: Perseverance and passion for long term goals.
Journal of Personality and Social Psychology, 92, 1087–1101.
Fantuzzo, J., Perry, M. A., & McDermott, P. (2004). Preschool
approaches to learning and their relationship to other relevant
classroom competencies for low-income children. School
Psychology Quarterly, 19(3), 212–230.
Fantuzzo, J. W., Bulotsky-Shearer, R., Fusco, R. A., & McWayne,
C. (2005). An investigation of preschool classroom behavioral
adjustment problems and social-emotional school readiness competencies. Early Childhood Research Quarterly, 20, 259–275.
Galinsky, E. (2010). Mind in the making: The seven essential life
skills that every child needs. New York, NY: William and
Morrow.
Gehsmann, K., & Templeton, S. (2011–2012). Stages and standards
in literacy: Teaching developmentally in the age of accountability. Journal of Education, 192, 5–16.
Gilks, W., Richardson, S., & Spiegelhalter, D. (1995). Markov chain
Monte Carlo in practice. New York, NY: Chapman & Hall.
Gilliam, W. S., & Frede, E. (2012). Accountability and program
evaluation in early education. In R. C. Pianta, W. S. Barnett,
L. M. Justice, & S. M. Sheridan (Eds.), Handbook of early
childhood education (pp. 73–91). New York, NY: Guilford.
Graue, M. E. (1993). Ready for what? Constructing meanings of
readiness for kindergarten. Albany: State University of New
York Press.
Graue, M. E. (2006). The answer is readiness—Now what is the
question? Early Education and Development, 17, 43–56.
Gresham, F. M., & Elliot, S. N. (1990). Social skills rating system.
Circle Pines, MN: Pearson.
Haskins, R., & Barnett, W. S. (Eds.). (2010). Investing in young children: New directions in federal preschool and early childhood
policy. Washington, DC: The Center on Children and Families
and the National Institute for Early Education Research.
Hau, K.-T., & Marsh, H. W. (2004). The use of item parcels in
structural equation modeling: Nonnormal data and small
sample sizes. British Journal of Mathematical and Statistical
Psychology, 57, 327–351.
Heckman, J. J., Moon, S. H., Pinto, R., Savelyev, P. A., & Yavitz,
A. (2010). The rate of return to the High Scope Perry Preschool
Program. Journal of Public Economics, 94, 114–128.
Henderson, J., & Strain, P. S. (2009). Screening for delays
and problem behavior (Roadmap to Effective Intervention
Practices). Tampa: University of South Florida.
Hu, L., & Bentler, P. (1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Isquith, P. K., Gioia, G. A., & Espy, K. A. (2004). Executive
function in preschool children: Examination through everyday
behavior. Developmental Neuropsychology, 26, 403–422.
Jaberg, P. E., Dixon, D. J., & Weis, G. M. (2009). Replication evidence in support of the psychometric properties of the Devereux
Early Childhood Assessment. Canadian Journal of School
Psychology, 24(2), 158–166.
Kagan, S. L. (1990). Readiness 2000: Rethinking rhetoric and
responsibility. Phi Delta Kappan, 72, 272–279.
Kagan, S. L., Moore, E., & Bredekamp, S. (1995). Reconsidering children’s early learning and development: Toward shared beliefs and
vocabulary. Washington, DC: National Education Goals Panel.
Kim, J. S., & Sunderman, G. L. (2005). Measuring academic proficiency under the No Child Left Behind Act: Implications for
educational equity. Educational Researcher, 34, 3–13.
Kishton, J. M., & Widaman, K. F. (1994). Unidimensional versus domain representative parceling of questionnaire items:
13
Barbu et al.
An empirical example. Educational and Psychological
Measurement, 54, 757–765.
LeBuffe, P. A., & Naglieri, J. A. (1999a). Technical manual for the
Devereux Early Childhood Assessment (DECA). Villanova, PA:
Devereux Foundation.
LeBuffe, P. A., & Naglieri, J. A. (1999b). Devereux Early
Childhood Assessment user’s guide. Lewisville, NC: Kaplan
Early Learning Company Publishing.
Li-Grining, C. P., Votruba-Drzal, E., Maldonado-Carreño, C.,
& Haas, K. (2010). Children’s early approaches to learning
and academic trajectories through fifth grade. Developmental
Psychology, 46(5), 1062.
Lien, M. T., & Carlson, J. S. (2009). Psychometric properties of the
Devereux Early Childhood Assessment in a Head Start sample.
Journal of Psychoeducational Assessment, 27(5), 386–396.
Little, R. J. A. (1988). A test of missing completely at random for
multivariate data with missing values. Journal of the American
Statistical Association, 83, 1198–1202.
Little, T. D., Cunningham, W. A., Shahar, G., & Widaman, K. F.
(2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling, 9(2), 151–173.
Lubke, G. H., & Muthén, B. O. (2004). Applying multi-group confirmatory factor models for continuous outcomes to Likert scale
data complicates meaningful group comparisons. Structural
Equation Modeling, 11, 514–534.
Magnuson, K. A., Rohm, C., & Waldfogel, J. (2007). Does prekindergarten improve school preparation and performance?
Economics of Education Review, 26(1), 33–51.
McClelland, M. M., Acock, A. C., Piccinin, A., Rhea, S. A., &
Stallings, M. C. (2013). Relations between preschool attention span-persistence and age 25 educational outcomes. Early
Childhood Research Quarterly, 28, 314–324.
McDermott, P. A., Leigh, N. M., & Perry, M. A. (2002).
Development and validation of the preschool learning behaviors scale. Psychology in the Schools, 39(4), 353–365.
Meisels, S. J., & Fenichel, J. (Eds.). (1996). New visions for the
developmental assessment of infants and young children.
Washington, DC: Zero to Three: National Center for Infants,
Toddlers, and Families.
Meyer, L. (2008). Fremont school readiness: Devereux Early
Childhood Assessment (DECA) outcomes 2007–08. Retrieved
July 8, 2011, from http://www.echofamilycenter.com/uploads/
child_outcomes/07-08/DECA07-08_School_Readiness-Final.pdf
Moffitt, M. E., Arseneault, L., Belsky, D., Dickson, N., Hancox,
R. J., Harrington, H., . . . Caspi, A. (2011). A gradient of childhood self-control predicts health, wealth and public safety.
Proceedings of the National Academy of Sciences, 108, 2693–
2698.
Nasser, F., & Wisenbaker, J. (2003). A Monte Carlo study investigating the impact of item parceling on measures of fit in
confirmatory factor analysis. Educational and Psychological
Measurement, 63, 729–757.
National Association for the Education of Young Children
(NAEYC). (2002). Early learning standards: Creating conditions for success. Washington, DC: Author.
National Center for Educational Statistics. (2010). Kindergarten
through first grade parent Social Rating Scale (SRS)
(Approaches to Learning scale items). Retrieved from nces.
ed.gov/pubs2010/data/2010070_atl_readme.pdf-2010-07-28.
14
National Education Goals Panel (NEGP). (1995). Reconsidering
children’s early development and learning: Toward common
views and vocabulary. Reston, VA: Author.
Neibecker, B. (1984). The validity of computer-controlled magnitude scaling to measure emotional impact of stimuli. Journal of
Marketing Research, 21(3), 325–331.
Neuman, S. B., & Roskos, K. (2005). The state of pre-kindergarten standards. Early Childhood Research Quarterly, 20,
125–145.
Niemeyer, J., & Scott-Little, C. (2001). Assessing kindergarten children: A compendium of assessment instruments.
Tallahassee, FL: SERVE. Retrieved from www.serve.org/
publications/rdakcc.pdf
North Carolina Division of Child Development. (2008). Approaches
to learning. Raleigh: North Carolina Department of Health and
Human Services, North Carolina Division of Child DevelopmentInfant-Toddler Foundations. Retrieved from http://ncchildcare
.nc.gov/pdf_forms/infant_toddler_foundations.pdf
Ogg, J. A., Brinkman, T. M., Dedrick, R. F., & Carlson, J. S.
(2010). Factor structure and invariance across gender of the
Devereux Early Childhood Assessment protective factor scale.
School Psychology Quarterly, 25(2), 107–118.
Pianta, R. C., Barnett, W. S., Burchinal, M., & Thornburg, K. R.
(2009). The effects of preschool education: What we know,
how public policy is or is not aligned with the evidence base,
and what we need to know. Psychological Science in the Public
Interest, 10, 49–88.
Pianta, R. C., Barnett, W. S., Justice, L. M., & Sheridan, S. M.
(2012). Handbook of early childhood education. New York,
NY: Guilford.
Poropat, A. E. (2014). Other-related personality and academic performance: Evidence and implications. Learning and Individual
Differences, 34, 24–32.
Rolnick, A., & Grunewald, R. (2003). Early childhood development: Economic development with a high public return.
Minneapolis: The Federal Reserve Bank of Minneapolis.
Retrieved from http://www.minneapolis.org/publications_
papers/pub_display.pdf
Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York, NY: John Wiley.
Schmeichel, B. J., & Tang, D. (2015). Individual differences in
executive functioning and their relationship to emotional processes and responses. Current Directions in Psychological
Science, 24, 93–98.
Scott-Little, C., Kagan, S. L., & Frelow, V. S. (2006).
Conceptualization of readiness and the content of early learning standards: The intersection of policy and research? Early
Childhood Research Quarterly, 21, 153–173.
Snow, C. E., & Van Hemel, S. B. (2008). Early childhood assessment:
Why, what & how? Washington, DC: National Academies Press.
Snow, K. (2006). Measuring school readiness: Conceptual and
practical considerations. Early Education and Development,
17, 7–41.
Snow, K. (2011). Developing kindergarten readiness and other
large-scale assessment systems: Necessary considerations in
the assessment of young children. Washington, DC: National
Association for the Education of Young Children.
Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics (6th ed.). Boston, MA: Allyn & Bacon.
Assessing Approaches to Learning
Thigpen, B. (2014). School readiness: A challenge and opportunity. Retrieved from http://orgs.wku.edu/ttas/docs/resources/
ta-paper-122008-school-readiness.pdf
Vygotsky, L. S. (2012). Thought and language. Cambridge, MA:
MIT Press. (Original work published 1931)
Woodside-Jiron, H., & Gehsmann, K. (2009). Peeling back the layers of policy and school reform: Revealing the structural and
social complexities within. International Journal of Disability,
Development, and Education, 56, 49–72.
Yaden, D. B., Cimetta, A., Marx, R. W., Taren, D., Perry, N., Enz,
B., Swadener, B., Horn, R., & Gallagher, L. (2011, January).
Arizona Kindergarten Readiness Study [Technical Report No.
1]. Tucson, AZ: University of Arizona, College of Education,
First Things First External Evaluation.
Ziv, Y. (2013). Social information processing patterns, social
skills, and school readiness in preschool children. Journal of
Experimental Child Psychology, 114, 306–320.
Authors
OTILIA C. BARBU, PhD, is scientist at the SPH Analytics and
research associate at the University of Arizona, College of Education.
Her research interests include mathematical and statistical modeling,
measurement, assessment, and evaluation in education and
healthcare.
DAVID B. YADEN, Jr., PhD, is professor of language, reading,
and culture in the Department of Teaching, Learning and
Sociocultural Studies at the University of Arizona, College of
Education. His research interests include early childhood development and assessment, the early acquisition reading and writing in
multilingual environments and the application of complex adaptive
systems theory to the emergence and development of literacy
knowledge and ability.
DEBORAH LEVINE-DONNERSTEIN, PhD, is senior lecturer
and senior researcher in the Department of Educational Psychology,
College of Education, at the University of Arizona. Her research
encompasses first generation undergraduates, witnessed race-ethnic prejudice models, effects of sleep among youth, and early childhood education.
RONALD W. MARX, PhD, is dean and professor of educational
psychology at the College of Education at the University of
Arizona, where he holds the Lindsey/Alexander Chair in Education.
His work in learning and instruction has focused on science education in urban school reform, and more recently, measurement issues
in early childhood education.
15