using sem to identify direct and indirect influences …using sem to identify direct and indirect...
TRANSCRIPT
Using SEM to Identify Direct and Indirect Influences
on Cognitive and Language Development of
Toddlers from Low-Income Families
Teresa M. Ober and Patricia J. Brooks
1. Introduction
Children growing up in poverty have elevated risk of cognitive and language
delays relative to their peers from more affluent families (Merz, Wiltshire, &
Noble, 2019). Poverty impacts neural development from infancy, placing infants
reared in economically disadvantaged and stressful environments at risk of
compromised neurocognitive development (Ursache & Noble, 2016). For such
children, factors such as poor prenatal care, low maternal educational attainment,
teenage pregnancy, food and housing insecurity, and parenting stress tend to co-
occur, with far-reaching impact on developmental trajectories for cognition and
language (Evans, Li, & Whipple, 2013). Cumulative risk indices may be more
predictive of children’s developmental outcomes than any single factor and
contribute to developmental cascades in abilities as they unfold over time (Cutuli
et al., 2017; Masten & Cicchetti, 2010).
The current study aimed to increase an understanding of structural relations
among factors influencing developmental trajectories of infants growing up in
poverty, with the goal of distinguishing direct and indirect influences of
contextual (home environment), maternal (educational attainment, maternal
distress), interactional (joint attention, negative interaction), and child (gestational
age, gender) factors on cognitive and language development. Building on prior
work (Poulakos, 2013), we developed a structural equation model (SEM) of
cumulative risk, using measures taken at age 14 months to predict individual
differences in cognitive and language outcomes at age 36 months. The models
examined developmental trajectories of 672 children (49.7% male) from the Early
Head Start Research and Evaluation Project (EHSRE; United States Department
of Health and Human Services Administration for Children and Families, 2011).
The EHSRE was a longitudinal randomized-control evaluation study of the
impact of enrollment in Early Head Start programs on social and cognitive
outcomes from infancy through childhood. The EHSRE Birth to Three Phase
involved administration of comprehensive assessments related to children’s
cognitive and communicative functioning, parenting stress, parent-child
* Teresa M. Ober is at the University of Notre Dame; Patricia J. Brooks is at the College of
Staten Island and The Graduate Center, CUNY. Please direct correspondence to Teresa
Ober at [email protected].
© 2020 Teresa M. Ober and Patricia J. Brooks. Proceedings of the 44th Boston University Conference on Language Development, ed. Megan M. Brown and Alexandra Kohut,
416-429. Somerville, MA: Cascadilla Press.
interactions, home environment, and family background collected when children
were 14, 24, and 36 months of age. For our study, we restricted the dataset to the
control group to avoid confounds associated with varying characteristics of Early
Head Start programs.
Our SEM included latent variables for maternal distress, joint attention, and
negative interactions involving the child. Maternal distress encompassed
variables related to maternal depressive symptoms, parenting stress, and
dysfunctional interactions with the child. Previous work indicates that children
exposed to maternal distress and depressive symptoms in infancy tend to have
lower receptive vocabulary knowledge relative to their peers (Ahun et al., 2017;
Letourneau, Tramonte, & Willms, 2013). Joint attention subsumed measures of
child sustained attention, child engagement of the parent, and parental
supportiveness in the context of semi-structured toy play protocol. Infants’ ability
to sustain attention to toys while engaged with a parent has been shown to predict
their subsequent vocabulary knowledge (Yu, Suanda, & Smith, 2019). Other
research emphasizes the importance of parental support and responsiveness to
children’s communicative bids in fostering cognitive and language development
(Tamis-LeMonda, Bornstein, & Baumwell, 2003). Negative interaction
subsumed measures of parental intrusiveness, parental negative regard, and child
negativity towards the parent. Negative parenting behaviors have been found to
be associated with diminished language development, particularly in infants
exhibiting negatively valanced affective responses (Laake & Bridgett, 2018).
In addition to the latent variables, our SEM included indices of the quality of
the home environment as the home is the primary context for development during
infancy. The extent to which the home contains age-appropriate toys and books
has been linked with language development outcomes (Melvin et al., 2017). We
included gestational age and gender in the models, as infants born prematurely are
at heightened risk of developmental delays (Bee et al., 1982). Gender was also
included as girls have been observed to outperform boys on language measures
(Bornstein, Hahn, & Haynes, 2004). Maternal education was also included as it
has been linked with children’s language development (Dollaghan et al., 1999),
and with associated factors including maternal responsiveness and the quality of
the home environment (Magnuson et al., 2009). After establishing the fit of the
SEM in predicting children’s cognitive and language outcomes at age 36 months,
we tested whether the model effectively accounted for variation in outcomes for
children of minority and non-minority parents. Determining the generalizability
of the SEM to different participant subgroups is important as parenting behaviors
associated with children’s cognitive and communicative development may vary
across racial/ethnic groups (Brooks-Gunn & Markman, 2005).
2. Method
2.1. Participants
The sample consisted of a subset of participants recruited from low-income
families for the EHSRE project (United States Department of Health and Human
417
Services Administration for Children and Families, 2011). The subset consisted
of children in the control group (i.e., not enrolled in Early Head Start programs)
who did not have missing data for the 36-month receptive vocabulary outcome (N
= 672; 327 boys and 345 girls). For the sample used here family income was on
average 60.2% (SD = 49.6%) of the poverty threshold.
Several key demographic variables were provided by maternal caregivers.
Mothers self-identified their race/ethnicity was as follows: 46.4% identified as
White/Caucasian, 37.1% as Black/African-American, 11.6% as Hispanic/Latino,
and 5.0% as Other. Race/ethnicity of the child’s caregiver was re-coded as a
binary variable (1=minority, 0=non-minority). Mothers indicated their highest
educational attainment by selecting one of three possible choices: 41.3% attended
some high school, 31.9% obtained a high school degree/GED, and 26.8% obtained
a degree higher than high school or GED diploma. Mothers also reported the
child’s approximate gestational age by selecting one of four choices. The infants’
gestational age was reported by the mother, with responses indicating that most
infants (85.2%) were born on time; 1.4% were born >2 months early, 11.4% were
born >3 weeks early, and 1.9% were born >3 weeks late. Many mothers (41.7%)
had given birth to the focal child before the age of 18. The majority (58.7%) were
neither married nor living with partner during the course of the study. A subset
(9.3%) of mothers reported speaking a language other than English at home.
2.2. Measures
2.2.1. Cognitive Ability and Vocabulary Knowledge
The child’s cognitive abilities were assessed at ages 14 and 36 months using
standardized scores from the Bayley Mental Development Index (MDI), a
subscale of the Bayley Scales of Infant Development, 2nd Edition (Bayley, 1993).
The MDI measures cognitive, communicative, and social-emotional development
of children under age 41 months. The child’s language ability was assessed at age
36 months using the Peabody Picture Vocabulary Test, 3rd Edition (PPVT; Dunn
& Dunn, 1997), an untimed assessment of receptive vocabulary. Four pictures are
presented at a time with the child asked to point to the one that matches a word
spoken aloud. Standardized scores for the Bayley MDI and the PPVT were
derived for each participant based on population norms, μ = 100, σ = 15.
2.2.2. Child and Maternal Characteristics
Demographic factors related to both the child and their mother were included
in the analyses; these variables were derived from a parent questionnaire
administered when the child was 14 months old. Child characteristics included
gender, coded as a binary variable (0=female, 1=male), and approximate
gestational age. Maternal characteristics included the mother’s level of
educational attainment and a set of measures used to create a latent variable for
maternal distress. The measures encompassed the 20-item Center for
Epidemiological Studies Depression (CES-D) Scale (Radloff, 1977), a self-report
418
assessment of depression symptoms experienced over the past week, and two
subscales of the short-form version of the Parental Stress Index (PSI-SF; Abidin,
1995). The 12-item Parent-Child Dysfunctional Interaction subscale indicates
whether the mother perceives her child to abuse or reject her or feels disappointed
in or alienated from her child. The 12-item Parental Distress Index indicates the
mother’s level of distress in the role of parent, stemming from personal factors,
e.g., a low sense of parenting competence, or stress due to perceived restrictions
associated with parenting, depression, or general lack of social support.
2.2.3. Home Environment
Examiners completed the Home Observation for the Measurement of the
Environment (HOME Inventory; Caldwell & Bradley, 1984), which uses a
combination of maternal self-report and observed measures to assess the quality
of cognitive stimulation and emotional support provided by the child's family at
home. For our analyses, 6p used the total score on the HOME Inventory.
2.2.4. Joint Attention and Negative Interaction
Two latent variables were derived from the 3-bag task, a semi-structured play
procedure where the parent and child are given three bags of interesting toys and
asked to play with the toys in a sequence (Early Head Start, 2005). The interaction
was videotaped with child and parent behaviors scored in accordance with the 3-
bag task coding scales (Ware et al., 1998). The coding scales focused three
positive and four negative aspects of parent-child interaction (possible scores for
each scale ranging from 1 to 7). The latent variable for joint attention was derived from scales assessing child sustained attention, child engagement of the parent,
and parent supportiveness. The latent variable for negative interaction was
derived from scales assessing parent intrusiveness, parent negative regard towards
the child, child negativity toward parent. Parent detachment was dropped from the
latent variable as its loading yielded a standardized estimate < .30.
3. Results
3.1. Descriptive Statistics
Table 1 provides descriptive statistics for measures taken at ages 14 and 36
months of age for the overall sample and subgroups; Ns vary due to missing data.
Note that six mothers did not report race/ethnicity. Standardized scores on the
Bayley MDI at 14 months approximated the normed population mean of 100.
However, at 36 months of age, standardized scores on both the Bayley MDI and
the PPVT were well below the normed population mean, suggesting delays in
cognitive and communicative development, with more than half of the children
scoring > 1 SD below the population mean of 100 on each of these tests. Scores
on the Bayley MDI and PPVT measured at 36 months of age were moderately
correlated, r(620) = .58, p < .001.
419
Table 1. Descriptive Statistics for Measures at 14 and 36 Months.
Overall Sample Minority Non-Minority
N M (SD) N M (SD) N M (SD)
14-month measures
Maternal Distress
Parent Distress Inventory 606 27.2 (9.4) 321 27.8 (9.6) 280 26.5 (9.0)
Parent-Child Dysfunction 604 17.3 (5.7) 319 17.6 (6.3) 280 16.9 (4.9)
CES Depression Scale 599 13.6 (9.8) 316 13.1 (9.4) 278 14.2 (10.4)
HOME Inventory 575 26.5 (3.3) 298 25.6 (3.5) 272 27.5 (2.8)
3–bag Task
Child Sustained Attention 540 5.0 (1.0) 280 4.8 (1.0) 256 5.2 (0.9)
Child Engagement 540 3.5 (1.1) 280 3.9 (1.8) 256 3.9 (1.0)
Parent Supportiveness 540 4.0 (1.0) 280 3.7 (1.0) 256 4.3 (1.0)
Parent Intrusiveness 540 2.4 (1.2) 280 2.7 (1.2) 256 2.1 (1.1)
Parent Negative Regard 540 1.4 (0.8) 280 1.6 (0.9) 256 1.3 (0.6)
Child Negativity 540 2.1 (1.1) 280 2.3 (1.2) 256 1.9 (1.0)
Bayley MDI 518 98.4 (10.8) 260 97.5 (9.9) 253 99.3 (11.6)
36-month measures
Bayley MDI 622 90.3 (12.7) 326 94.2 (12.3) 290 88.1 (12.6)
PPVT 672 82.3 (16.3) 357 78.4 (15.6) 309 86.9 (16.0)
3.2. Regression Analysis
We used regression analyses to examine the cumulative impact of the
variables as measured at 14 months in accounting for communicative and abilities
in children at 36 months of age, see Table 2. The regression analyses were
conducted with multiple imputation (10 iterations) in R (R Core Team, 2019)
using the MICE package (Van Buuren & Groothuis-Oudshoorn, 2011). Inspection
of a missing data plot suggested that data were missing at random for the selected
variables of interest. The regression model predicting Bayley MDI scores at 36
months was statistically significant, F(14, 657) = 22.68, p < .001, R2 = .33. The
regression model predicting PPVT at 36 months was also significant, F(14, 657)
= 16.65, p < .001, R2 = .26.
The results of the regression models for the two 36-month outcomes were
similar. Both models indicated significant effects of maternal education, home
environment (HOME Inventory), child sustained attention, parent supportiveness,
and infant cognitive ability (Bayley MDI at 14 months) in predicting individual
differences in cognitive and language outcomes at 36 months. The model
predicting scores on the Bayley MDI at 36 months also showed significant effects
of child gender (favoring girls) and gestational age, as well as a significant
negative effect of parent-child dysfunction. Note that despite the non-significance
of some of the variables, we retained the full set of 14-month variables as
predictors of 36-month outcomes, as preliminary analyses indicated patterns of
co-variance that would be accounted for in the SEM.
420
Table 2. Standardized regression coefficients for predictors of Bayley MDI and PPVT
scores at 36 months; all predictors were assessed at 14 months (N = 672).
Bayley MDI PPVT
Maternal Education .12 *** .13 ***
Child Gender (1 = Male) –.08 * –.06 †
Child Gestational Age .07 * .03
HOME Inventory .14 *** .16 ***
Maternal Distress Variables
Parent Distress Inventory –.03 –.08 †
Parent-Child Dysfunction –.12 ** .01
CES Depression Scale .02 .04
3–bag Task
Child Sustained Attention .09 * .20 ***
Child Engagement of Parent .03 .04
Parent Supportiveness .20 *** .10 *
Parent Intrusiveness –.01 .00
Parent Negative Regard –.04 .01
Child Negativity –.01 .02
Bayley MDI (14 months) .19 *** .18 ***
Note: ***p < .001, **p < .01, *p < .05, † p < .10
3.3. Structural Equation Modeling
As a first step in building the SEM, we used confirmatory factor analysis
(CFA) to evaluate the adequacy of the latent variables serving as indices of
maternal distress, joint attention, and negative interaction. CFA and SEM analyses
were conducted in the lavaan package (Rosseel, 2012) in R (R Core Team, 2019).
The latent variables had acceptable model fit indices. Standardized loadings for
manifest variables were > .50 for maternal distress and joint attention, and > .40
for negative interaction; see Figure 1.
Our SEM tested for direct and indirect influences of variables assessed at 14
months in predicting Bayley MDI and PPVT scores at 36 months. We used full
information maximum likelihood estimation to handle missing data (0%–24% for
variables at 14 months; 0%–7% for variables at 36 months) to reduce potential
bias resulting from listwise deletion (Acock, 2012). Model fit was assessed using
the chi-square test, comparative fit index (CFI), Tucker-Lewis index (TLI), root
mean square error of approximation (RMSEA), and standardized root mean
square residual (SRMR). After building a SEM for the full sample and trimming
non-significant paths, we conducted a measurement invariance analysis to test the
adequacy of the SEM for minority and non-minority subgroups. Measurement
invariance analysis tests assumptions of psychometric equivalence by comparing
the model fit of two groups against a theoretical model using a series of models
with increasingly constrained parameters (Putnick & Bornstein, 2017).
421
Fig. 1. Standardized factor loadings for the three latent variables: maternal distress,
joint attention, and negative interaction.
The SEM used to predict Bayley MDI scores at 36 months had acceptable
model fit indices using unconstrained factor loadings and intercepts; χ2(df = 80, n
= 672) = 178.56, CFI = .94, TLI = .92, RMSEA = .04 [.03, .05], SRMR = .05, R2
Bayley MDI (36m) = .345. Direct effects of joint attention, B = .37, z = 7.18, p <
.001, 14-month cognitive ability, B = .27, z = 6.46, p < .001, maternal distress, B
= –.16, z = –3.38, p < .01, and maternal education, B = .11, z = 2.94, p < .01, on
36-month Bayley MDI scores were observed. Additionally, there were significant
indirect effects of the home environment, B = .04, z = 3.34, p < .01, gender, B =
–.02, z = –2.06, p < .05, and gestational age, B = .04, z = 3.27, p < .01, via 14-month
cognitive ability, as well as of the home environment, B = .14, z = 5.28, p < .001,
and maternal education, B = .05, z = 3.16, p < .01, via joint attention.
The same SEM used to predict PPVT scores at 36 months also had acceptable
model fit indices using unconstrained factor loadings and intercepts; χ2(df = 80, n
= 672) = 178.56, CFI = .94, TLI = .92, RMSEA = .04 [.03, .05], SRMR = .05, R2
PPVT (36m) = .247. The model indicated significant direct effects of joint
attention, B = .33, z = 6.54, p < .001, 14-month cognitive ability, B = .21, z = 5.13, p < .001, maternal education, B = .11, z = 2.84, p < .01, and maternal distress, B
= –.10, z = –2.32, p < .05, in predicting 36-month PPVT scores. Again, there were
significant indirect effects of the home environment, B = .03, z = 2.94, p < .01,
and gestational age, B = .03, z = 2.86, p < .01, via 14-month cognitive ability, and
of the home environment, B = .10, z = 4.06, p < .001, and maternal education, B
= .04, z = 2.84, p < .01, via joint attention. The indirect effect of gender via 14-month cognitive ability approached significance, B = –.02, z = –1.94, p = .05.
422
Fig. 2a. SEM predicting Bayley MDI scores at 36 months in the minority subsample;
χ2(df = 194, n = 357) = 138.83, CFI = .92, TLI = .91, RMSEA = .04, SRMR = .07, R2 = .348.
Fig. 2b. SEM predicting Bayley MDI scores at 36 months in the non-minority
subsample; χ2(df = 194, n = 309) = 181.19, CFI = .92, TLI = .91, RMSEA = .04, SRMR =
.07, R2 = .321.
For each subgroup and outcome measure, we established partial invariance
after relaxing constraints on the loadings and residuals for several parameters of
the SEM. The model fit indices were acceptable for both minority and non-
minority subgroups; see Figures 2a, 2b for models predicting Bayley MDI and
Figures 3a, 3b for models predicting PPVT scores. Standardized estimates shown
423
in bold in the figures indicate significant differences between subgroups; all
coefficients > .09 were statistically significant (p < .05).
Fig. 3a. SEM predicting PPVT scores at 36 months in the minority subsample; χ2(df
= 194, n = 357) = 131.50, CFI = .92, TLI = .91, RMSEA = .05, SRMR = .06, R2 = .207.
Fig. 3b. SEM predicting PPVT scores at 36 months in the non-minority subsample;
χ2(df = 194, n = 309) = 186.37, CFI = .92, TLI = .91, RMSEA = .05, SRMR = .06, R2 =
.208.
The models indicate considerable stability in relations between 14-month and
36-month measures across subgroups, and the importance of joint attention,
cognitive ability (Bayley MDI), maternal education, and maternal distress as
424
direct influences on cognitive and language outcomes. Effects of other variables
were indirect.
4. Discussion
For children growing up in poverty, individual differences in developmental
outcomes have been linked to factors that originate in the first years of life
(Ursache & Noble, 2016). The current study used SEM to tease apart direct and
indirect influences of child and maternal characteristics, home environment, and
parent-infant interactional dynamics in infancy in predicting cognitive and
language outcomes of low-income minority and non-minority children at age 36
months. We conducted preliminary regression analyses to identify 14-month
variables of interest and used CFA to test the adequacy of three latent variables
serving as indices of maternal distress, joint attention, and negative interaction
between mother and infant. After testing the adequacy of the latent variables, we
established a best-fitting SEM for the full sample. We trimmed non-significant
paths from the full model, and then conducted a measurement invariance analysis
to determine whether the SEM showed acceptable fit indices for subgroups
comprising children from minority and non-minority backgrounds. Testing for
measurement invariance is important as children from minority groups experience
differential hardship and risk due to structural inequalities and societal prejudice,
which may affect psychometric equivalence (Brooks-Gunn & Markman, 2005;
Putnick & Bornstein, 2017).
The two outcome measures—Bayley MDI and PPVT scores—were
moderately correlated at 36-months. The Bayley MDI encompasses measures of
cognitive, social-emotional, and communicative skills (Bayley, 1993); hence one
would expect scores on this measure to be associated with receptive vocabulary
knowledge. Importantly, the SEMs reported in this paper yielded similar findings
across outcome measures (i.e., evidence of replication). Thus, the observed direct
and indirect relations between 14-month variables and cognitive and language
abilities at 36 months were not tied to results from a specific standardized test.
Starting with the child factors, we observed a direct effect of 14-month
cognitive ability (Bayley MDI) on 36-month cognitive and language outcomes
(Bayley MDI and PPVT), indicating considerable stability in intellectual abilities
over the first years of life (Bornstein et al., 2004). Both of the other child factors,
gestational age and child gender, had significant indirect effects on 36-month
outcomes via 14-month cognitive ability. Prior research indicates long-term
impact of prematurity on language development (Bee et al., 1982; Putnick et al.,
2017). Our findings suggest that early differences in general cognitive abilities
mediate the impact of prematurity on vocabulary development. Gender (favoring
girls over boys) has also been linked with individual differences in language
abilities at 2 to 5 year of age (Bornstein et al., 2004). Our findings again suggest
that early differences in general cognitive abilities mediate this relation.
With regard to maternal characteristics, both maternal educational attainment
had a significant direct effect in predicting children’s 36-month outcomes, while
425
the latent variable maternal distress did not appear to. The two maternal variables
exhibited significant negative covariance and contributed in opposite ways to the
quality of the home environment. Previous research has found that infants with
mothers exhibiting a great number of depressive symptoms tend to later possess
lower receptive vocabulary knowledge (Ahun et al., 2017; Letourneau, Tramonte,
& Willms, 2013). Our findings concur and provide some evidence that the
association is direct and relatively stable across subgroups, albeit somewhat weak
when other factors are taken into consideration. In line with previous research
(Christian et al., 1998; Dollaghan et al., 1999), we found maternal educational
attainment to be positively associated with children’s development outcomes at
36 months. Taken together, these results indicate the importance of supporting
maternal mental health and educational opportunities for mothers living in low
income communities as an effective means of fostering their children’s cognitive
and academic skill development.
Of all of the variables in the model, the latent variable joint attention had the
strongest direct effect on 36-month outcomes. Joint attention consisted of
measures of child’s sustained attention, child engagement of parent, and parent
supportiveness taken from the 3-bag task, an observational measure of parent-
infant interaction at 14 months. As indicated by the CFA, these three measures
were closely linked, as was expected (Adamson et al., 2019). In our preliminary
regression models, both child’s sustained attention and parent supportiveness
emerged as significant predictors of 36-month cognitive and language outcomes
(see also Brooks, Flynn, & Ober, 2018). In the context of joint attention,
supportive parents may aid children’s cognitive and linguistic development by
providing vocabulary that is relevant to the child’s focus of attention (Tamis-
LeMonda et al., 2001; Yu et al., 2019).
In addition to the longitudinal associations with cognitive and language
development, joint attention exhibited a significant concurrent association with
14-month cognitive ability. We interpreted this relation as bidirectional, given that
joint attention is thought to promote infants’ cognitive, social-emotional, and
communicative skills and vice versa (Adamson et al., 2019; Yu et al., 2019).
Notably, joint attention also had a significant negative covariance with the latent
variable negative interaction, consisting of measures of parent intrusiveness,
parent negative regard toward the child, and child negativity taken from the 3-bag
task. Previous research indicated that parental support and responsiveness to
infants were associated with advances in children’s language development
(Tamis-LeMonda et al., 2003) whereas parental negativity towards the child was
associated with decrements (Laake & Bridgett, 2018). However, in our models
there was no direct effect of negative interaction on 36-month outcomes. Thus,
intrusive parenting behaviors and negativity on the part of the parent towards the
child appeared to influence later outcomes only by diminishing opportunities for
joint attention.
Two other variables, maternal education and the quality of the home
environment, both had significant indirect effects on 36-month outcomes via joint
attention. That is, joint attention was enhanced for children of mothers with
426
greater educational attainment and for children living in more emotionally
supportive and cognitively stimulating homes. The quality of the home
environment has been found to have a significant and stable association with
language development, even after accounting for socio-economic status (SES;
Melvin et al., 2016). We found this to be true in regression analyses. However,
when entered into the SEM, the quality of the home environment unexpectedly
did not have a significant direct effect on 36-month outcomes. In addition to its
indirect effect via joint attention, the home environment also had a significant
indirect effect via 14-month cognitive ability.
The observed relations were not only stable across outcome measures, but
also for both minority and non-minority subgroups. The results of the
measurement invariance analysis suggested that the relations between 14-month
factors and 36-month outcomes were largely the same across the two subgroups,
even though a few of the parameters were left unconstrained (Steenkamp &
Baumgartner, 1998) and some of the standardized estimates varied in magnitude.
4.1. Conclusions
The SEMs served to highlight pathways through which individual differences
in developmental trajectories emerge and identify targets for early intervention.
We found that after accounting for the influence of various factors related to the
child, parent, and home environment, joint attention in infancy remained the
strongest predictor of subsequent cognitive and receptive vocabulary skills. The
observed direct and indirect influences on cognitive and language development
were largely stable across minority and non-minority subgroups. Although the
SEM models indicated a number of factors influencing developmental trajectories
of children growing up in poverty, the models did not aim to be comprehensive in
accounting for effects of SES as all of the children in the EHSRE study were
raised in low-income circumstances (i.e., there was restricted range on variables
associated with income). Within this low-income sample, risk factors appeared to
be heightened for minority children, which likely reflects systemic inequality in
our society (Ursache, & Noble, 2016). Further analyses of the EHSRE dataset
should examine how the constellation of factors considered here at 14 months
continue to influence children’s later cognitive and communicative abilities
beyond pre-school entry. Additional work is needed to understand more fully how
factors associated with poverty impact children’s developmental trajectories and
evaluate strategies for helping children overcome obstacles to their success.
References
Abidin, Richard R., & Brunner, Jack F. (1995). Development of a parenting alliance
inventory. Journal of Clinical Child Psychology, 24(1), 31-40.
Acock, Alan C. (2012). What to do about missing values. In Harris Cooper, Paul M. Camic,
Debra L. Long, A. T. Panter, David Rindskopf, Kenneth J. Sher (Eds). APA handbook of research methods in psychology, Vol. 3. Data analysis and research publication
(pp. 27-50). Washington, DC, US: American Psychological Association.
427
Adamson, Lauren B., Bakeman, Roger, Suma, Katherine, & Robins, Diana L. (2019). An
expanded view of joint attention: Skill, engagement, and language in typical
development and autism. Child Development, 90(1), e1-e18.
Ahun, Marilyn N., Geoffroy, Marie-Claude, Herba, Catherine M., Brendgen, Mara,
Séguin, Jean R., Sutter-Dallay, Anne-Laure, Boivin, Michel, Tremblay, Richard E.,
& Côté, Sylvana M. (2017). Timing and chronicity of maternal depression symptoms
and children's verbal abilities. The Journal of Pediatrics, 190, 251-257.
Bayley, Nancy. (1993). Bayley Scales of Infant Development (2nd Ed.). San Antonio, TX:
Psychological Corporation.
Bee, Helen L., Barnard, Kathryn E., Eyres, Sandra J., Gray, Carol A., Hammond, Mary A.,
Spietz, Anita L., Snyder, Charlene, & Clark, Barbara (1982). Prediction of IQ and
language skill from perinatal status, child performance, family characteristics, and
mother-infant interaction. Child Development, 53(5), 1134-1156. Bornstein, Marc H., Hahn, Chun-Shin, & Haynes, O. Maurice. (2004). Specific and general
language performance across early childhood: Stability and gender considerations.
First Language, 24(3), 267-304.
Brooks, Patricia J., Flynn, Rachel M., & Ober, Teresa M. (2018). Sustained attention in
infancy impacts vocabulary acquisition in low-income toddlers. In Proceedings of the 42nd Annual Boston University Conference on Language Development (pp. 86-99). Somerville, MA: Cascadilla Press.
Brooks-Gunn, Jeanne, & Markman, Lisa B. (2005). The contribution of parenting to ethnic
and racial gaps in school readiness. The Future of Children, 15(1), 139-168.
Caldwell, Bettye M., & Bradley, Robert H. (1984). Home Observation for Measurement of the Environment. Little Rock: University of Arkansas at Little Rock.
Christian, Kate, Morrison, Frederick J., & Bryant, Fred B. (1998). Predicting kindergarten
academic skills: Interactions among child care, maternal education, and family
literacy environments. Early Childhood Research Quarterly, 13(3), 501-521.
Cutuli, J. J., Montgomery, Ann Elizabeth, Evans‐Chase, Michelle, & Culhane, Dennis P.
(2017). Childhood adversity, adult homelessness and the intergenerational
transmission of risk: A population‐representative study of individuals in households
with children. Child & Family Social Work, 22(1), 116-125.
Dollaghan, Christine A., Campbell, Thomas F., Paradise, Jack L., Feldman, Heidi M.,
Janosky, Janine E., Pitcairn, Dayna N., & Kurs-Lasky, Marcia. (1999). Maternal
education and measures of early speech and language. Journal of Speech, Language, and Hearing Research, 42(6), 1432-1443.
Dunn, Lloyd M., & Dunn, Leota. M. (1997). PPVT-III: Peabody Picture Vocabulary Test. Circle Pines, MN: American Guidance Service.
Early Head Start. (2005). Videotaped Protocol Booklet–14 Months. Retrieved from
https://www.acf.hhs.gov/opre/resource/videotaped-protocol-booklet-14-months.
Evans, Gary W., Li, Dongping, & Whipple, Sara Sepanski. (2013). Cumulative risk and
child development. Psychological Bulletin, 139(6), 1342-1396.
Hu, Li-tze, & Bentler, Peter M. (1998). Fit indices in covariance structure modeling:
Sensitivity to underparameterized model misspecification. Psychological Methods,
3(4), 424-453.
Laake, Lauren M., &. Bridgett, David J. (2018). Early language development in context:
Interactions between infant temperament and parenting characteristics. Early Education and Development, 29(5), 730-746.
Letourneau, Nicole L., Tramonte, Lucia, & Willms, J. Douglas. (2013). Maternal
depression, family functioning and children's longitudinal development. Journal of Pediatric Nursing, 28(3), 223-234.
428
Magnuson, Katherine A., Sexton, Holly R., Davis-Kean, Pamela E., & Huston, Aletha C.
(2009). Increases in maternal education and young children's language skills. Merrill-Palmer Quarterly, 55(3), 319-350.
Masten, Ann S., & Cicchetti, Dante. (2010). Developmental cascades. Development and Psychopathology, 22(3), 491-495.
Melvin, Samantha A., Brito, Natalie H., Mack,. Luke J., Engelhardt, Laura E., Fifer,
William P., Elliott, Amy J., & Noble, Kimberly G. (2017). Home environment, but
not socioeconomic status, is linked to differences in early phonetic perception ability.
Infancy, 22(1), 42-55.
Merz, Emily C., Wiltshire, Cynthia A., & Noble, Kimberly G. (2019). Socioeconomic
inequality and the developing brain: Spotlight on language and executive function.
Child Development Perspectives, 13(1), 15-20.
Poulakos, Anthoula. (2013). Language development in context: Influences of perinatal risk and environmental characteristics on outcomes at 14 to 36 months (Doctoral
Dissertation). Order No. 3601689. Available from ProQuest Dissertations & Theses
Global (1467526239).
Putnick, Diane L., & Bornstein, Marc H. (2016). Measurement invariance conventions and
reporting: The state of the art and future directions for psychological research.
Developmental Review, 41, 71-90.
Putnick, Diane L., Bornstein, Marc H., Eryigit-Madzwamuse, Suna, & Wolke, Dieter.
(2017). Long-term stability of language performance in very preterm, moderate-late
preterm, and term children. The Journal of Pediatrics, 181, 74-79.
Radloff, Lenore Sawyer. (1977). The CES-D scale: A self-report depression scale for
research in the general population. Applied Psychological Measurement, 1, 385-401.
Rosseel, Yves. (2012). Lavaan: An R package for structural equation modeling and more.
Version 0.5–12 (BETA). Journal of Statistical Software, 48(2), 1-36.
Steenkamp, Jan-Benedict E. M. & Baumgartner, Hans. (1998). Assessing measurement
invariance in cross-national consumer research. Journal of Consumer Research,
25(1), 78-90.
Tamis‐LeMonda, Catherine S., Bornstein, Marc H., & Baumwell, Lisa (2001). Maternal
responsiveness and children's achievement of language milestones. Child Development, 72(3), 748-767.
United States Department of Health and Human Services Administration for Children and
Families (2011). Early Head Start Research and Evaluation (EHSRE) Study, 1996-2010: [United States]. Ann Arbor, MI: Inter-university Consortium for Political and
Social Research [distributor]. Retrieved from https://doi.org/10.3886/ICPSR03804.v5
Ursache, Alexandra, & Noble, Kimberly G. (2016). Neurocognitive development in
socioeconomic context: Multiple mechanisms and implications for measuring
socioeconomic status. Psychophysiology, 53(1), 71-82.
Van Buuren, Stefan & Groothuis-Oudshoorn, Karin. (2011). Mice: multivariate imputation
by chained equations in R. Journal of Statistical Software, 45(3), 1-67.
Ware, Anne, Brady, Christy, O’Brien, Claudia, Berlin, Lisa J. & Brooks-Gunn, Jeanne.
(1998). 14-month child-parent interaction rating scales for the three bag assessment. New York, NY: Center for Children and Families, Teachers College, Columbia
University.
Yu, Chen, Suanda, Sumarga H., & Smith, Linda B. (2019). Infant sustained attention but
not joint attention to objects at 9 months predicts vocabulary at 12 and 15 months.
Developmental Science, 22(1), e12735.
429
Proceedings of the 44th annualBoston University Conference on Language Development
edited by Megan M. Brown and Alexandra Kohut
Cascadilla Press Somerville, MA 2020
Copyright information
Proceedings of the 44th annual Boston University Conference on Language Development© 2020 Cascadilla Press. All rights reserved
Copyright notices are located at the bottom of the first page of each paper.Reprints for course packs can be authorized by Cascadilla Press.
ISSN 1080-692XISBN 978-1-57473-057-9 (2 volume set, paperback)
Ordering information
To order a copy of the proceedings or to place a standing order, contact:
Cascadilla Press, P.O. Box 440355, Somerville, MA 02144, USAphone: 1-617-776-2370, [email protected], www.cascadilla.com