Late Language Emergence at 24 Months: An Epidemiological Study of Prevalence, Predictors, and Covariates

Late Language Emergence at
24 Months: An Epidemiological
Study of Prevalence, Predictors,
and Covariates
Stephen R. Zubrick
Catherine L. Taylor
Purpose: The primary objectives of this study were to determine the prevalence of late
Curtin University of Technology,
language emergence (LLE) and to investigate the predictive status of maternal,
Telethon Institute for Child
family, and child variables.
Health Research
Method: This is a prospective cohort study of 1,766 epidemiologically ascertained
24-month-old singleton children. The framework was an ecological model of child
Mabel L. Rice development encompassing a wide range of maternal, family, and child variables.
David W. Slegers Data were obtained using a postal questionnaire. Item analyses of the 6-item
University of Kansas Communication scale of the Ages and Stages Questionnaire (ASQ; D. Bricker & J.
Squires, 1999; J. Squires & D. Bricker, 1993; J. Squires, D. Bricker, & L. Potter, 1997;
J. Squires, L. Potter, & D. Bricker, 1999) yielded a composite score encompassing
comprehension as well as production items. One SD below the mean yielded good
separation of affected from unaffected children. Analyses of bivariate relationships
with maternal, family, and child variables were carried out, followed by multivariate
logistic regression to predict LLE group membership.
Results: 13.4% of the sample showed LLE via the ASQ criterion, with 19.1% using the
single item of “combining words.” Risk for LLE at 24 months was not associated with
particular strata of parental educational levels, socioeconomic resources, parental
mental health, parenting practices, or family functioning. Significant predictors
included familial history of LLE, male gender, and early neurobiological growth.
Covariates included psychosocial indicators.
Conclusion: Results are congruent with models of language emergence and impair-
ment that posit a strong role for neurobiological and genetic mechanisms of onset
that operate across a wide variation in maternal and family characteristics.
KEY WORDS: late language emergence (LLE), late talkers
C
hildren’s language comprehension and production emerge be-
tween 12 and 24 months of age. Some otherwise healthy children
require more time to begin talking, a condition described here as
late language emergence ( LLE). The reasons for such variation at the
toddler stage of development are relatively unexplored. Variations in
family or maternal resources are thought to play a role, although actual
outcomes are mixed. More recently, genetic studies have focused on pos-
sible inherited risk for LLE (cf. Dale et al., 1998; Spinath, Price, Dale, &
Plomin, 2004). LLE is widely assumed to be the first diagnostic symptom
of children with language impairments. Tager-Flusberg and Cooper (1999)
called for studies of early identification of specific language impairment
(SLI) “ with particular emphasis on predicting which late talkers develop
SLI ” ( p. 1277).
1562 Journal of Speech, Language, and Hearing Research • Vol. 50 • 1562–1592 • December 2007 • D American Speech-Language-Hearing Association
1092-4388/07/5006-1562
A handful of studies have documented the phenom- societal) resources available to the child. The model rec-
enon (Fenson, Reznick, Bates, Thal, & Pethick, 1994; Paul, ognizes that the proximal and distal resources available
1996; Rescorla, 1989; Thal & Katich, 1996; Whitehurst to the child will vary over the life course due to changes
& Fischel, 1994) and provided valuable descriptive and in circumstances for better or worse. Although this frame-
interpretive information. At the same time, with few ex- work has not been used in previous studies of late talk-
ceptions, the studies were limited by small sample sizes ers, the independent variables linked to late talking can
and/or convenience sampling procedures and a small all be placed in this model and categorized as relating to
number of independent variables. In addition, much of the child (neurobiological and genetic mechanisms) or
the literature relies on relatively extensive parental ques- the maternal or family environment. Consistent with the
tionnaires to document children’s lexical development. ecological framework, information was collected on a
These instruments are often infeasible for large-scale wide range of variables to describe maternal and family
investigations of a wide range of possible predictors and attributes and socioeconomic factors, concurrent with
covariates of LLE. A few alternatives have been devel- extensive documentation of children’s perinatal status
oped, although they have not yet been used to evaluate and developmental and health outcomes. Findings are
predictors of LLE in single-born children in a study with summarized in a series of reports commissioned for pub-
a large number of participants and independent vari- lic policy application (Silburn et al., 1996; Zubrick et al.,
ables. The following factors remain unknown: (a) the prev- 1995, 1997).
alence of LLE in the general population of 24-month-old Candidate predictors of the emergence of language.
children and ( b) the extent to which a wide range of ma- Maternal and family variables —in particular, socio-
ternal, family, and child variables are predictive of late economic indictors —have been linked with the onset of
talking. These issues are addressed in this investigation language in young children. Mother ’s education level
of an epidemiologically ascertained sample of 1,766 and family socioeconomic status (SES) are thought to be
24-month-old singleton children who were participants proxy measures of environmental support for language
in a large-scale investigation of health outcomes. Par- learning. Mother’s education is reported to be associated
ticipants provided information on a wide range of tar- with the amount of talking to children (cf. Hart & Risley,
geted maternal, family, and child variables. 1995; Hoff-Ginsberg, 1994; Wells, 1985), which in turn
is predictive of vocabulary development in singletons
(Dollaghan et al., 1999; Huttenlocher, Haight, Bryk,
Ecological Model of Child Development Seltzer, & Lyons, 1991) and twins (Lytton, Conway, &
The participants in this study were recruited at Sauve, 1977) and is positively associated with a num-
birth (1995–1996) into an ongoing longitudinal study of ber of language indices in the first 3 years, including
children’s health and developmental outcomes known verb tenses (Hart & Risley, 1995) and utterance length
as Randomly Ascertained Sample of Children born in (Dollaghan et al., 1999). Furthermore, maternal and pa-
Australia’s Largest State ( RASCALS), based in Perth, ternal education is reported to be a predictor of lan-
Australia. According to current census data, Western Aus- guage impairment (Tomblin, 1996). LaBenz and LaBenz
tralia is demographically similar to some states in the (1980) document language outcomes of a national sam-
midwestern United States. For example, the population ple of 20,137 children, followed from birth to 8 years of
of Kansas is 2.7 million, and the population of Western age, and report that mother’s education predicted fail-
Australia is 1.8 million. Also, in each state, most mem- ure at age 8 years on language comprehension testing.
bers of the population live in urban areas. The states are The outcomes of these studies and the conclusions of
predominately Caucasian (86% for Kansas; 96% for Entwisle and Astone (1994)— that mother ’s education
Western Australia) and are native speakers of English, is the preferred index of “ human capital ” in the home
are well educated (86% high school completion in each when considering environmental contributions to young
state), and are family oriented (in Kansas, 55% of all children’s development—support consideration of levels
families are couple families with children and 9% are of mother’s education as a general risk index for chil-
sole-parent families; in Western Australia, 49% and 15%, dren’s language acquisition.
respectively). On a wide variety of behavioral and bio- Recent studies, however, yield mixed evidence with
logical assessments of children and adults, distributional regard to the chain of predictive effects sketched above.
outcomes conform to normative expectations for instru- In an investigation of 108 low-income toddlers, Pan, Rowe,
ments normed in the United States or the United Kingdom. Singer, and Snow (2005) found that maternal talkative-
This health outcomes study was guided by an eco- ness was not related to growth in children’s vocabulary
logical model of child development (Bronfenbrenner, 1979). production in the 24- to 36-month period observed. In-
This model views child development as a complex instead, maternal language and literacy (which was collin-
terplay between a child’s biogenetic endowment and ear with maternal education) was a significant predictor
the proximal (i.e., maternal and family) and distal (i.e., of growth; mothers with lower vocabularies/lower reading
Zubrick et al.: Late Language Emergence 1563

levels had children with lower levels of vocabulary pro- status, described as late talking, which is typically bench-
duction in spontaneous samples. Pan et al. (2005) noted marked to the 24-month age level. Two widely used crite-
that the outcomes are compatible with either a genetic ria are 50 words in reported vocabulary and the presence
view of shared linguistic aptitude between mothers and of 2–3 word combinations. Rescorla (1989) developed
children or an environmental input view that mothers the Language Development Survey ( LDS) as a parental
with higher language and literature skills interact with report measure, which comprises a 310-word vocabulary
their children differently than do mothers with lower checklist and a question about the presence and fre-
language and literacy skills. They conclude that the ma- quency of children’s early word combinations. Prevalence
ternal language and literacy effect, in their data, is not estimates from this instrument, using a criterion of fewer
entirely mediated by maternal input, suggesting a need than 50 words or no word combinations, yield estimates
to consider influences beyond input. Pan et al. (2005) of 10–20% of children as late talkers (Klee et al., 1998;
also investigated maternal depression as a predictor of Rescorla, 1989; Rescorla & Achenbach, 2002; Rescorla &
growth and found it to have direct effects that increased Alley, 2001; Rescorla, Hadick-Wiley, & Escarce, 1993).
with time. In their article, they note earlier reports that Paul (1996) also used this criterion to identify a group of
depressed mothers talk less to their children than do late talkers for longitudinal assessment. The MacArthur
healthy mothers (Breznitz & Sherman, 1987) and cau- Communicative Development Inventories: Words and
tion that in their study, the outcome measure was chil- Sentences (CDI/WS; Fenson et al., 1993) is a widely used
dren’s vocabulary use in interactions with their mothers. parent report measure that uses a 680-word checklist as
Further investigation of maternal depression as a pos- well as questions about early word combinations. Ac-
sible predictor is warranted. cording to the normative data provided by Fenson et al.
Family SES levels are also implicated as risk fac- (1993), at 24 months the bottom 5% of the distribution
tors. Although SES and maternal education are highly varies according to gender: For boys, it is an expressive
associated in the general population, there is strong rea- vocabulary of 70 words; for girls, it is 48 words. Es-
son to consider them as separate variables when eval- timates of word combinations are available for children
uating children’s development (cf. Entwisle & Astone, 16–30 months of age. Within this age range, for children
1994). For example, in the Western Australia Child Health who were reported to produce 50 words or less, 21% were
Survey (Zubrick et al., 1997), there was a stronger asso- reported to combine words “sometimes” and 1.3% were
ciation between the education and employment status reported to do so “often.” Children with an expressive
of caregivers and children’s academic competence than vocabulary of fewer than 50 words or not combining
income and family structure and academic outcomes. words by 24 months represented the bottom 10% of the
This finding would have been obscured if a composite CDI norming sample (Fenson et al., 1993). Thal, Tobias,
measure had been used. With regard to SES, in general, and Morrison (1991) used the CDI / WS criterion at or be-
as noted by Hoff-Ginsberg (1998), high SES mothers are low the 10th percentile to identify late talkers in a follow-up
reported to have higher levels of child-related adjust- study. Dale et al. (2003) used the 10th percentile; Feldman
ments that are thought to enhance children’s language et al. (2005) reported five levels to define delays (i.e., 5th,
acquisition. However, as she notes in her study of 63 high- 10th, 15th, 20th, and 25th percentiles).
and mid-level SES children ages 18–29 months (Hoff- Candidate predictors for LLE in children younger
Ginsberg, 1998), the effects of maternal differences and than 3 years have been identified from studies that have
child differences may be less detectable in the mid to differentiated late talkers and controls and compared
high levels of SES. Total vocabulary size in young chil- them on a select number of child, maternal, and family
dren is significantly related to SES, although at very low variables that have been linked theoretically and /or em-
levels of association and/or localized at the low end of the pirically to language development and language impair-
distribution (Fenson et al., 1994; Rescorla & Alley, 2001; ment. The child characteristics that have been examined
Rice, Spitz, & O’Brien, 1999; Wells, 1985). Pan et al. (2005) include gender, gestational age, perinatal and obstetric
reported no effect for family income in their study of low- risks, child behaviour, and motor development. The ma-
income toddlers. Overall, there is reason to regard mother’s ternal characteristics include mother’s education and
education and SES as risk factors for language emergence mother–child interaction. The family influences in-
in toddlers, although the associations may not be strong, clude SES, birth order, family size, and family history
and the available evidence yields mixed outcomes. of late talking (cf. Olswang, Rodriguez, & Timler, 1998;
Whitehurst & Fischel, 1994, for reviews).
LLE: Measurement, Prevalence, Candidate Among the candidate predictors, there is a strong,
replicated gender risk for late talking. In prevalence
Predictors, and Covariates studies, the proportion of boys who are late to talk is much
Parent report measures of children’s vocabulary higher than girls (Horwitz et al., 2003; Klee et al., 1998;
and early word combinations are used to determine LLE Rescorla, 1989; Rescorla & Achenbach, 2002; Rescorla &
1564 Journal of Speech, Language, and Hearing Research • Vol. 50 • 1562–1592 • December 2007
Alley, 2001; Rescorla et al., 1993), as well as in late talker their study also reported that living in a bilingual house-
cohort studies (Ellis Weismer, Murray Branch, & Miller, hold was a strong predictor, thereby confounding the inter-
1994; Paul, 2000; Rescorla, 2002; Whitehurst, Fischel, pretation of LLE risk with bilingualism.
Arnold, & Lonigan, 1992). This strong gender effect There is evidence that LLE influences the dynamics
seems to be a phenomenon of the lower tail of the dis- of parent–child interaction. Whitehurst and colleagues
tribution of children; it is not apparent across the full (1988) compared parental interactions for groups of late
distribution where the gender effects in favor of girls are talkers, age-matched controls, and language-matched
significant but small (Feldman et al., 2000; Fenson et al., controls. They reported differences for late talkers com-
1994; Huttenlocher et al., 1991; Rescorla & Achenbach, pared with age-matched controls and similarities be-
2002) or not evident at all (Bornstein, Tamis-LeMonda, & tween late talkers and language-matched controls. They
Maurice Haynes, 1999; Pan et al., 2005; Wells, 1985). concluded that the differences in parent interaction be-
Normal distributions that differ only modestly in their tween late talkers and age-matched normally developing
means can have very large relative differences at the children reflected parental adaptation to the language
extremes. abilities of the children. Paul and Elwood (1991) reported
Children’s birth history and perinatal status do not similar results.
appear to be viable risk indicators for LLE. Late talkers Feldman et al. (2005) call for investigation of the
do not have elevated rates of fetal and birth complica- role of a positive family history of language disorders or
tions compared with controls (Paul, 1991; Rescorla et al., delays as a potential predictor of outcomes. Hadley and
1993; Whitehurst et al., 1992). In the most recent epide- Holt (2006) investigated maternal education and posi-
miological study of SLI in kindergarten-aged children, tive family history as predictors of growth in tense mark-
adverse intrauterine and birth events were not risk ex- ing abilities in 2-year-old children with low levels of
posures for SLI (Tomblin, Smith, & Zhang, 1997). Sim- language development. Positive family history was re-
ilarly, in a recent twin study, prenatal, perinatal, and lated to differences in tense marking growth trajectories,
obstetric risks were not associated with lower levels of whereas maternal education was not a predictor. The
language performance in twins compared with single- finding contrasted with Hart and Risley’s (1995) finding
tons at 20 and 36 months (Rutter, Thorpe, Greenwood, that maternal education was associated with children’s
Northstone, & Golding, 2003). Marschik et al. (in press) production of verb tenses. Hadley and Holt studied chil-
reported that children ( N = 15) who scored below the dren in the low range of language abilities, whereas Hart
10th percentile on an Austrian adaptation of the CDI/WS and Risley studied children across the full range of the
at 18 months had lower Apgar scores than did controls distribution of language abilities. This suggests that the
and that 5 late talkers (and none of the controls) required influence of maternal education on performance is mod-
neonatal intensive care. Interestingly, 8 of the 15 late ulated by child characteristics. Hadley and Holt’s study
talkers scored within the normal range on the CDI / WS was the first to carry out growth curve analyses includ-
at 24 months. ing positive familial history as a predictive variable for
Delayed motor development has been reported in children’s late talking. This extends the findings from
several studies of late talkers. Rescorla and Alley (2001), previous studies that have reported higher levels of
Klee et al. (1998), and Carson et al. (1998) conducted di- familial risk in late talkers compared with controls (Ellis
rect assessment of motor abilities using standardized Weismer et al., 1994; Paul, 1991; Rescorla & Schwartz,
tests and reported that late talkers had lower levels of 1990).
motor development than did controls. None of the chil- Psychosocial development has been linked to late
dren in these studies had developmental disorders or talking. The temperament and behavior characteristics
syndromes associated with delayed motor development. of small numbers of late talkers have been investigated
Information about the influence of SES, parental in several studies. Caulfield, Fischel, DeBaryshe, and
education, and occupation on late talkers is very lim- Whitehurst (1989) studied 34 late talkers and 34 controls
ited, in part because of the predominately convenience (24–32 months); Carson et al. (1998) studied 11 late talkers
sampling methods that draw heavily from middle-class and 53 controls (24–26 months); Irwin, Carter, and
families (Rescorla, 2002). Using the CDI, Thal, Bates, Briggs-Gowan (2002) studied 14 late talkers and 14 con-
Goodman, & Jahn-Samilo (1997) reported at 16–25 months trols (21–32 months); Paul and James (1990) studied
a slight SES advantage for early talkers and a slight dis- 34 late talkers and 33 controls (24 months). Higher rates
advantage for late talkers, although this trend was not of problems were reported for late talkers compared with
confirmed by post hoc testing and was not present when controls in these studies. In contrast to these studies,
the children were 21–31 months. Using the MacArthur Rescorla and Achenbach (2002) did not find an associa-
Communicative Development Inventory ( MCDI)–Short tion between language delay and behavior problems in a
Form (Fenson et al., 1993), Horwitz et al. (2003) found general population sample of 278 children 18–35 months
SES and maternal education effects at 24 months, although of age.

Interpretation of the relation between psychosocial otitis media, and the family ’s SES levels. The proposed
development and LLE is not straightforward. There are risk factors were not subjected to formal analyses.
measurement confounds, in that many parent report What is needed is information on a relatively wide
measures of psychosocial ability include items such as range of maternal, family, and child variables as pos-
“talks with other children” that are confounded with lan- sible predictors of late talking in the same population
guage ability. Further, psychosocial differences could be of late talkers and controls, with formal multivariate
consequences of limited language ability (cf. Redmond analyses for risk assessment. The questions addressed
& Rice, 1998, 2002). Thus, psychosocial development in this study are: (a) What is the prevalence of LLE in an
does not carry the straightforward predictor status car- epidemiologically ascertained sample of 24-month-old
ried by other variables, such as gender and maternal children? and ( b) Which maternal, family, and child
education. variables are predictive of LLE?
The outcome of LLE is of considerable interest. Late
onset of language is a hallmark characteristic of chil-
dren with language impairments. In the case of children Method
with SLI, who do not demonstrate other developmental
limitations, late talking can be the first diagnostic symp-
Participants
tom. Available studies report that 17% (Rescorla, 2002) Data for this analysis came from the RASCALS
to 26% (Paul, 1996) of the late talkers have persistent study in Western Australia. RASCALS is an ongoing
SLI at 4–6 years, although the actual estimates are com- longitudinal postal study of a sample of children born in
plicated by the criteria used for diagnosis. These out- Western Australia in 1995 and 1996. The sample design
comes should be considered preliminary, given very small is an epidemiological prospective observational study
sample sizes. Rescorla (2002) followed 34 children; Paul of infants randomly ascertained from a total popula-
(1996) followed 31 children; Whitehurst and Fischel tion frame of birth notifications for the state of Western
(1994) followed 37 children; and Thal et al. (1991) followed Australia and followed annually from birth. These designs
10 children. Not surprisingly, outcomes of late talking are sometimes called cohort studies. In cohort studies,
include social and academic risk (i.e., reading ability), in the relationship between exposure and the incidence of
tandem with the likelihood of immature language com- an outcome is examined by following the entire cohort
petencies (Paul, 1996; Rescorla, 2002; Whitehurst & and measuring the rate of occurrence of new cases in dif-
Fischel, 1994). ferent exposure groups. The prospective follow-up allows
Limitations of current evidence. The samples of the investigator to identify participants with and with-
children studied have been small in size (with the ex- out the outcomes of interest. In a case-control study, the
ception of the twin study of Dale et al., 2003) and drawn individuals who develop the outcome condition (the cases)
from predominately middle- and upper-middle-class fam- are identified by some other mechanism than follow-up,
ilies. Furthermore, some of the samples have included and a group of participants (the controls) is used to rep-
an admixture of monolingual and bilingual children resent the participants who do not develop the outcome
or households in which multiple languages are spoken condition. As findings in this report focus on the 24-month
(Fenson et al., 1994; Klee et al., 1998; Rescorla, 1989; follow-up of the RASCALS cohort only, the data anal-
Rescorla et al., 1993), providing a possible confounding yses has been approached from the standpoint of a case-
effect of undetermined significance. Under such sam- control study nested in a cohort study (i.e., nested case
pling constraints, it is not possible to estimate the extent control study; Clayton & Hills, 1996).
to which the outcomes generalize to the general popu- The original sample was randomly drawn from a to-
lation of 24-month-old children or to interpret risk for tal population sample frame comprising statutory noti-
LLE independent of risk associated with English ac- fications of birth (Stanley, Read, Kurinczuk, Croft, &
quired simultaneously with one or more other languages. Bower, 1997). The RASCALS study was designed to sur-
The examination of potential predictors of late talking vey health-related behaviors (Kurinczuk, Parsons, Dawes,
has been limited to a few variables and has been fur- & Burton, 1999) as a way of identifying and evaluating
ther constrained by the small sample sizes and sampling health promotion opportunities from birth to 8 years and
confounds. as a way of investigating early causal pathways of men-
Multivariate models of risk have also been con- tal health problems in childhood. Four thousand and
strained by the available empirical evidence. Our review seven mothers responded to the initial questionnaire
of the literature yielded a single model (Olswang, et al. sent at 3 months postpartum. A comparison with data
1998). The model posits the child’s speech, language, available about all births in Western Australia (Stanley
and social development prior to age 3 years as predictors et al., 1997) showed that the 4,007 respondents were
of subsequent language growth, combined with the risk representative of all women with live births in that pe-
factors of positive family history, prolonged periods of riod, with the exception of a slight underrepresentation
of women with low birthweight babies (5.3% overall vs. who were either unmarried or not cohabitating, those
4.7% in the sample) and mothers aged younger than women who had an annual household income of $A16,000
20 years (6% overall vs. 3.6% in the sample). Because or less, and those women whose partner was absent from
metropolitan Aboriginal mothers were participating in a the household; as a result, 100 mothers were added to the
similar but more culturally appropriate study, they were sample. Thus, a total sample of 2,837 mothers and their
not recruited into the RASCALS study. singleton infants were selected for longitudinal follow-up,
Following the 3-month postpartum response, the of whom 2,224 (78%) agreed, when their infant was 1 year
study was converted to a longitudinal study and, for re- old, to participate. Of the 2,224 women who agreed to
source reasons, just less than a 70% random sample of participate, 1,880 (85%) returned a completed question-
mothers of singletons was drawn from the initial 4,007 naire when their child was 2 years old. These children
respondents. However, to ensure that “ hard-to-reach” are the focus of this report (see Figure 1).
groups remained in the RASCALS study in sufficiently Of potential concern is the representativeness of the
informative proportions, we also included all mothers study sample to the population from which it was drawn.
Figure 1. The study sample.

Our assessment of this suggests that the 2-year-old to the family, and to the child. The measures have been
follow-up sample is a reasonably representative sample used successfully in other population-based child health
of 2-year-old non-Aboriginal Western Australian single- surveys in Western Australia (Garton, Zubrick, &
ton children. The potential effects of sample attrition Silburn, 1995; Silburn et al., 1996; Zubrick et al., 1995;
were examined by comparing a range of early life char- Zubrick et al., 1997). The high completion rate (85%) and
acteristics present at 3 months of age for the respon- low level of missing data (2.2%) in this study provide fur-
dents at 2 years of age compared with the respondents at ther support for the suitability of the measures for di-
3 months. Small but significant variations were noted. verse population samples.
Participants at 2 years of age were more likely to be from Maternal characteristics. Each respondent was asked
families earning more than $A25,000 per year (74.5% vs. her age in years; current educational level; place of birth;
70.7%), c2(2) = 10.4, p < .01, and from married house- whether she was currently employed and, if so, the num-
holds (79.3% vs. 75.6%), c2(4, p < .03) = 11.7, with a higher bers of hours per week that she was in paid employment;
level of maternal education (27.8% vs. 23.3%), c2(4) = 23.6, whether she smoked during her pregnancy with the in-
p < .001. Aside from this, no significant differences were dex child; and whether she was a current smoker.
observed for mother’s place of birth, number of chil-
Each mother completed the 42-item Depression
dren or adults in the household, father absence, or receipt
Anxiety Stress Scale (DASS; Lovibond & Lovibond, 1995a,
of government benefits.
1995b). The DASS assesses symptoms of depression, anx-
iety, and stress in adults. Each item (i.e., “I feel sad and
Exclusions depressed”) is rated on a four-point Likert scale. Items are
As the principal focus of this study is on the phe- summed to generate a score for each of the three domains.
nomenology of late talking in children where English The scale has high reliability for the Depression (a = .91),
is their only language, prior to undertaking analysis, the Anxiety (a = .81), and Stress (a = .89) scales and has good
1,880 participants were assessed for eligibility and pos- discriminant and concurrent validity. Higher scores are
sible exclusion. Thirty-three participants were excluded associated with higher levels of distress.
because they did not speak English, another 66 children Mothers were also asked to complete the Parenting
who spoke both English and another language were ex- Scale ( PS; Arnold, O’Leary, Wolff, & Acher, 1993). This
cluded, and 7 children with known medical conditions or 30-item questionnaire measures three dysfunctional
syndromes were excluded. An additional 8 participants discipline styles: laxness ( permissive discipline); over-
lacked sufficient data from which to determine their pos- reactivity (authoritarian discipline; displays of anger,
sible eligibility and, hence, they were excluded. In total, meanness, and irritability); and verbosity (overly long
114 participants were excluded, leaving a final partici- reprimands or reliance on talking). The PS Total score
pant pool of 1,766 (see Figure 1). (range = 1–7) increases with increasingly dysfunctional
parenting, has good internal consistency (a = .84), good
Data Collection test–retest reliability (r = .84), and reliably discrimi-
nates between parents of clinical and nonclinical chil-
Data from the RASCALS participants were collected
dren where scores in excess of 3.1 denote “clinical” levels
by postal questionnaire. On or within 1 month of the
of dysfunctional parenting (Arnold et al., 1993). It has
study child’s first and subsequent birthdays up to age
been used extensively in research and been shown to be
8 years, the parents were mailed a questionnaire for self-
responsive to parenting interventions.
completion with a reply paid envelope. Mothers (who
represented the majority of the respondents) provided Family characteristics. Respondents were also
all details on themselves, their partners, and the study asked to provide details on the number of individuals
child. Nonrespondents received a reminder letter and who usually reside with the child and their biological
were subsequently contacted directly by the study re- and nonbiological relationship to one another. This al-
search assistant to confirm receipt of the questionnaire lowed a basic description of family structure (i.e., original,
and ascertain reasons for nonresponse. Only data col- step/blended, sole parent, or other). Although informa-
lected at ages 1 and 2 were used in the analysis reported tion on birth order was not gathered, the number of chil-
here. dren in the household under the age of 18 was recorded.
Total family income before taxes was gathered, as well as
receipt of government benefits. The residential address
Measures was linked to census track data permitting each child’s
In line with our ecological model, the variable space record to be populated with three small-area indices—the
is drawn from three broad domains of potential develop- Socioeconomic Indicators for Areas (SEIFA; Australian
mental influence: Characteristics related to the mother, Bureau of Statistics, 1998).
The SEIFA indicators used in this report mea- 10% of the control families, a statistically significant
sure disadvantage, resources, and occupation /education difference.
within the census collection district of the index house- Characteristics of the child: Birth status. The
hold. These indexes were developed by the Australian population database from which the RASCALS sample
Bureau of Statistics (1998). Each index summarizes a was drawn contains each child’s gender, birth date, race
different aspect of the socioeconomic conditions of the (Caucasian, Aboriginal, and Other), birthweight in grams,
Australian population using a combination of variables— low birthweight status (<2,500 grams), time to spontane-
in this case, from the 1996 Population and Housing ous respiration in minutes, and gestational age in weeks
Census. The Index of Relative Socioeconomic Disadvan- (Stanley et al., 1997). These data are collected by statute
tage (Australian Bureau of Statistics, 1998) is derived on all live births, stillbirths, and neonatal deaths in the
from variables that reflect or measure relative disadvan- state of Western Australia. An additional measure, the
tage. Variables used to calculate the index of relative Proportion of Optimal Birth Weight ( POBW ), is also de-
socioeconomic disadvantage include low income, low rived from these data.
educational attainment, high unemployment, and low-
POBW is a measure of the appropriateness of intra-
skill occupations. Lower scores are associated with greater
uterine growth and is routinely calculated from the birth
disadvantage. The Index of Economic Resources (Austra-
records of all children born in Western Australia. Be-
lian Bureau of Statistics, 1998) summarizes the income
cause birthweight is the end result of growth over the
and expenditure of families, such as income and rent
period of gestation, it is therefore determined both by the
living in the census district. Additionally, variables that
length of gestation and the rate of intrauterine growth.
reflect wealth, such as dwelling size, are also included.
The rate of intrauterine growth is determined by many
Lower scores reflect lower area economic resources. The
factors that are both pathological (maternal, fetal, or en-
Index of Education and Occupation (Australian Bureau
vironmental) and nonpathological ( genetic endowment
of Statistics, 1998) is designed to reflect the educational
[particularly fetal gender] and maternal environment).
and occupational structure of communities. The educa-
Thus, it is appropriate that fetal growth rate should vary
tion variables in this index show either the level of qual-
between individuals, as the nonpathological factors
ification achieved or whether further education is being
determining growth rate varies between individuals:
undertaken. The occupation variables classify the work-
Female newborns appropriately weigh less than male
force into the major groups of the Australian Standard
newborns of the same gestation, babies of small women
Classification of Occupations (ASCO) and the unemployed.
weigh less than babies of tall women, and a woman’s
This index does not include any income variables. Lower
first birth tends to weigh less than her subsequent
scores are associated with lower levels of education and
births. We define the optimal fetal growth rate for any
lower levels of job skill. Each index is standardized to
particular fetus as the median birthweight achieved by
have a mean of 1000 and a standard deviation of 100.
fetuses with the same values for the nonpathological
Mothers completed the 12-item General Factor determinants of fetal growth and duration of gestation1,
scale from the McMaster Family Assessment Device in the absence of any pathological determinants of fetal
( FAD; Miller, Epstein, Bishop, & Keitner, 1985). The growth. This median is expressed as the optimal birth-
12-item General Factor scale measures overall fam- weight once the values of the nonpathological determi-
ily functioning across six areas of family functioning: nants of growth have been specified.
problem-solving, communication, affective involvement,
The nonpathological determinants considered in
affective responsiveness, roles, and behavior control.
our statistical models were fetal gender, maternal age,
It has adequate test–retest reliability, has low correla-
height, and parity. Exclusion of pathological factors was
tions with social desirability, and shows evidence of both
achieved by limiting the sample from which optimal
concurrent and discriminative validity (Miller, Epstein,
birthweights were identified to singleton, live births with-
Bishop, & Keitner, 1985). Cronbach’s alphas on the gen-
out congenital abnormalities born to nonsmoking moth-
eral factor scale are on the order of .86 (Epstein et al.,
ers following pregnancies without any complications
1993; Miller et al., 1985; Zubrick et al., 1997). Higher
known to affect intrauterine growth (Blair, 1996). The me-
scores are associated with higher levels of dysfunction.
dian value of POBW is 100; values less than this signify
Finally, mothers were asked if there was a family infants that are undergrown, whereas values greater
history of late talking (i.e. “ Has anyone in your family than this represent growth in excess of optimal growth.
been slow in learning to talk? ”). Although this is a min-
imal estimate of family risk, there is evidence to support 1
Duration of gestation may be curtailed or prolonged, and this is usually the
validity. Rice, Haney, and Wexler (1998) investigated result of pathological factors; hence, abnormal duration of gestation may
19 families who were ascertained because of a child with be considered to reflect pathological factors. However, because delivery
must follow the period of intrauterine growth, duration of gestation is not a
SLI versus 41 control families. This question yielded determinant of growth and hence cannot be a pathological determinant of
39% of the SLI families with a positive history versus growth, although it is the primary determinant of birthweight.

Infants whose POBW is less than 85% are classified as the administration manual, and cutoffs were applied to
being growth restricted at birth. differentiate normal and abnormal performances (Squires
POBW is an important index of the child’s develop- & Bricker, 1993). Further details about the Communi-
mental status and is associated with increased risks cation scale are provided below.
for developmental and academic failure (Zubrick et al., Each mother completed a Child Behavior Checklist
2000). The advantage of this measure of appropriate- (CBCL) on the child (Achenbach & Edelbrock, 1991). The
ness of growth over birthweight is that it is both indi- CBCL is a checklist of 99 specific behavior problems
vidualized and takes into account the duration of gestation. (e.g., “nightmares,” “ too shy or timid,” “ wakes up often
The advantage over the commonly used percentile mea- at night,” “aches or pains (without medical cause), ” “gets
sures (sometimes termed small for gestational age) is in many fights,” “destroys his/ her own things”). Respon-
that it is more accurate and generalizable at the ex- dents are asked to identify if the item is not true (0), some-
tremes and, being a parametric ratio quantity, is more what true (1), or very true (2). The number of problems are
amenable to statistical manipulation. Where POBW can converted to normalized T scores (M = 50, SD = 10). The
be calculated, it is generally preferable to more tradi- higher the T score, the higher the problem behaviors.
tional measures such as gestational age and birthweight T scores can be computed for Total Problems, Internal-
(Blair, Liu, de Klerk, & Lawrence, 2005). izing and Externalizing Problems, and seven syndrome
Characteristics of the child: Developmental status. scales (Anxiety/Depression, Withdrawal, Sleep Problems,
Mothers were asked to complete the Infant /Child Mon- Somatic Problems, Aggressive Behavior, Destructive
itoring Questionnaires (now called the Ages and Stages Behavior, and Other Behavioral Problems). We report
Questionnaire [ASQ]; Bricker & Squires, 1999; Squires & T scores and Abnormal T scores for Total Problems and
Bricker, 1993; Squires, Bricker, & Potter, 1997; Squires Internalizing and Externalizing problems. The CBCL
et al., 1999) for children aged 24 months. The ASQ re- was chosen on the basis of its extensive use in Australia
quires the mother to observe her child and answer six and in numerous other settings and cultures (Hensley,
questions about her child’s development in each of five prin- 1988; Sawyer, Arney, et al., 2000; Sawyer et al., 2001;
cipal areas (which are also the names of the subscales)— Sawyer, Clark, & Baghurst, 1993; Sawyer, Kosky, et al.,
Communication, Gross Motor Skills, Fine Motor Skills, 2000; Sawyer, Sarris, Baghurst, Cornish, & Kalucy,
Adaptive Problem Solving, and Personal–Social Skills— 1990; Verhulst et al., 2003; Zubrick et al., 1995). Test–
for a total of 30 items. A final set of questions inquire retest reliability is reported to be on the order of 0.87
about the child’s overall development and are not used over 1 week and 0.75 over periods of 12 months (Achenbach
here. The authors of the ASQ have undertaken revision & Edelbrock, 1991). It has been demonstrated that the
of the instrument and, as a result, there is some slight CBCL is acceptable, relatively quick to administer, and
variation between the original items used in the RASCALS supplies adequate coverage of the phenomenology of
cohort, which are those in the original Infant/Child Mon- child behavior problems. Australian studies have found
itoring Questionnaires, and those now currently used in test–retest reliability of the parent CBCL to be 0.87 at
the ASQ (see Squires, Potter, & Bricker, 1999). 8 weeks and 0.75 at 6 months (Garton et al., 1995; Zubrick
The manual reports that at 24 months of age, the et al., 1997). The CBCL has also been used in previous
internal reliability for each of the ASQ scales is 0.75 for studies of late talkers (Carson et al., 1998; Rescorla &
Communication, 0.80 for Gross Motor Skills, 0.58 for Achenbach, 2002).
Fine Motor Skills, 0.57 for Adaptive Problem Solving, The Revised Dimensions of Temperament Survey
and 0.58 for Personal–Social Skills. Two week test–retest ( DOTS-R; Windle, 1992; Windle & Lerner, 1986) was
reliability, measured as a percentage of agreement be- completed by the mother of each study child. In pre-
tween parent-completed questionnaires, was 94%, and school children, this 54-item scale measures nine char-
interobserver reliability is similarly high. Extensive acteristics of temperament: (a) activity level–general
analyses of the validity of the ASQ, including receiver ( high scores signify high levels of energy and motor ac-
operator characteristics (ROC) analysis and assessments tivity), ( b) activity level–sleep ( high scores signify high
of concurrent validity using the Stanford–Binet Intel- motor activity during sleep; e.g., tossing and turning),
ligence Scale (Thorndike, Hagen, & Sattler, 1985), the (c) approach–withdrawal ( low scorers tend to withdraw
Bayley Scales of Infant Development (Bayley, 1969), and or move away from persons, objects, and situations),
the Revised Gesell and Armatruda Developmental and (d) flexibility–rigidity ( low scorers respond inflexibly to
Neurological Examination (Knobloch, Stevens, & Malone, changes in the environment), (e) mood ( low scores are
1980), show the ASQ to be valid for the identification characterized by negative affect), (f ) rhythmicity–sleep
of children at risk for developmental difficulties and in ( low scorers signify an irregular timing of daily sleep–
need of additional examination (Squires et al., 1999). Us- wake cycle), ( g) rhythmicity–eating ( low scores charac-
ing the RASCALS data, composite scores for each of the terize irregularity of eating habits pertinent to appetite
five areas were calculated using the method outlined in and quantity consumed), ( h) rhythmicity–daily habits
( low scores characterize irregularity of diurnal activities Following imputation, characteristics of the mother,
such as toileting, peak periods of vigor, and taking a the family, and the child were summarized (see Table 1).
rest), and (i) task orientation ( low scorers lack concen- Mothers were predominately between the ages of 24
tration and lack perceptual focus in the presence of ex- and 34 years at the time of the child’s birth. Australia
traneous stimuli and do not tend to stay with or continue mandates 10 years of compulsory education. Years (i.e.,
with an activity for relatively long periods of time). Grades) 11 and 12 are principally used for college entry
Alpha coefficients of internal consistency for each of the preparation. The majority of mothers completed 10 years
characteristics range from .62 to .89, and 6-week test– of education, and the distribution of maternal education
retest correlations range from .59 to .75 (Windle, 1992). was bimodal, with about one quarter of mothers having
We conducted a preliminary factor analysis to assess the fewer than 12 years of education, 19.3% having com-
suitability of Windle’s model for use with Australian pleted 12 years of schooling, 13.1% having completed a
children. Using Pearson product–moment correlations trade certificate, 13.7% having done some study toward
as input, a principal components analysis using the a postschool qualification, and another 29.2% having com-
RASCALS data revealed good factorability ( KMO = .88) pleted a postschool technical qualification or university
and communalities ranging from .306 to .839, with nine degree. Three quarters of the mothers were born in Aus-
factors accounting for 52% of the common factor var- tralia, and 40% of them were in paid employment, work-
iance. The number of nontrivial factors was determined ing an average of 22 hours per week. Mean maternal
by using Cattel’s scree plot in association with those DASS scores for depression, anxiety, and stress are com-
eigen values greater than or equal to 1.0. Following a parable to those of the normative sample (Lovibond
varimax rotation, the final factor structure revealed a & Lovibond, 1995b). The mean PS score reported by the
satisfactory correspondence with only four of the 54 var- RASCALS mothers was also commensurate with the
iables loading on factors different from those reported by means reported originally by Arnold et al. (1993) for
Windle (Windle, 1992; Windle & Lerner, 1986). In keep- their clinical and nonclinical groups and is comparable
ing with Windle’s recommendations, each of the dimen- to population means reported by Zubrick et al. (2005).
sion scores was coded into a dichotomous variable with a Families were predominately two-parent original
score of 1 indicative of dimension scores below the 30th families, with 13% of the remaining families being either
percentile for all but activity-level sleep and general ac- step/ blended or sole-parent families. The average num-
tivity level — these being coded 1 if above the 70th per- ber of children per family was two. Assessment of family
centile (Windle, 1992). income revealed a small proportion of families (5.5%)
Finally, the mother was asked the child’s day care earning $A16,000 or less per annum. Area SEIFA in-
status at the time of the interview and the hours per dicators for disadvantage, resources, and occupation /
week that the child attended or received daycare. education were well within population averages for these
measures. About 9% of families were classified as having
abnormal family function using the FAD. This compares
well to the population proportion of Western Australian
Results families reporting abnormal family function (10%; Silburn
All data were screened and distributions inspected et al., 1996). A family history of late talking was reported
for outliers and incorrect values. Missing data were in 13.5% of the families.
present, to some degree, in all modeled variables. The Children in the study were an average age of 2.1
average amount of missing data among the 1,766 sub- (SD = 0.13) years and were nearly all Caucasian (96.6%).
jects was 2.2% and ranged from zero (mother ’s place of With respect to neonatal characteristics, fewer low birth-
birth, child gender, and age) to 7.6% (dimensions of weight infants were in the study sample (3.7%) relative
temperament, rhythmicity–sleep). To address this prob- to the Western Australian population proportion (6.4%),
lem, we carried out data imputation via a multiple im- but otherwise, the neonatal characteristics of the study
putation procedure using SAS PROC MI (SAS Institute, sample were unremarkable, with mean birthweight, mean
2004). Five complete data sets were generated; each gestational age, and time to spontaneous respiration
subsequent analysis was performed on each of the data being comparable to Western Australian population av-
sets, and results were then combined. This imputation erages (Gee, 1996).
approach is preferable to single imputation, which sub- With respect to normative development, study
stitutes a single number for each missing value in that sample mean CBCL T scores were at the approximate
the multiple imputation approach accounts for the var- 50th percentile, and about 10% of the study children had
iability in plausible replacement values (Rubin, 1987). a CBCL Total T score in the clinical range. These are the
Using a Markov Chain Monte Carlo procedure, all data first Australian data to be gathered on children as young
were imputed at the item level before computing the as 2 years; however, the proportion of children scoring in
scale values. the clinical range is comparable to Western Australian

Table 1. Maternal, family, and child variables for total sample (N = 1,766a).
Total sample
Variable M (SD) %
Maternal characteristics
Age at child’s birth (years) 29.5 (4.9)
≤19 years 2.5
20–34 years 82.1
35+ years 15.4
Education
<12 years 24.6
12 years 19.3
Trade certificate 13.1
Postschool study 13.7
Completed postschool qualification 29.2
Mother’s place of birth
Australia 74.7
United Kingdom 14.8
New Zealand 3.9
Asia and India 2.4
Europe 1.8
North America 1.5
Africa 1.0
Employment
Currently in paid employment 40.1
Hours in paid employment per week 22.0 (12.7)
Cigarette use
Smoked before or during pregnancy 28.2
Current smoker 20.0
DASS scoresb
Depression 3.7 (5.3)
Anxiety 1.9 (3.3)
Stress 7.49 (6.8)
Clinical ranges
Depression
Normal 90.1
Mild 4.1
Moderate to severe 5.7
Anxiety
Normal 94.3
Mild 1.9
Stress
Normal 86.7
Mild 7.1
Parenting scorec 2.8 (0.57)
Parenting score in clinical range 28.8
Family characteristics
Family structure
Original 84.6
Step/blended 5.1
(Continued on the following page)
Table 1 Continued. Maternal, family, and child variables for total sample (N = 1,766a).
Total sample
Variable M (SD) %
Family structure
Sole parent 7.9
Other 2.4
Family size
No. children 2.06 (0.95)
1 29.8
2 or more 70.2
Income
$1–$154 wk / $1–$8,000 yr 0.7
$155–$308 wk / $8,001–$16,000 yr 5.5
$309–$481 wk / $16,001–$25,000 yr 11.5
$482–$769 wk / $25,001–$40,000 yr 25.9
$770 –$961 wk / $40,001–$50,000 yr 17.0
$962 or more wk / $50,001 or more yr 35.4
Not stated 3.8
Area SES indicators
Disadvantage 1007.9 (59.6)
Resources 1028.6 (63.3)
Occupation/education 987.3 (76.2)
Disadvantage (–1.0 SD) 3.4
Resources (–1.0 SD) 1.7
Occupation/education (–1.0 SD) 7.0
Family function
Abnormal family function 9.4
Child characteristics
Gender (male) 50.7
Age (years) 2.1 (0.13)
e
Race
Caucasian 96.6
Aboriginal 0.8
Other 2.4
Neonatal period
Gestational age < 37 weeks 6.2
<85% POBW 9.1
Low birth weight (<2500 gms) 3.7
Mean percent expected birth weight 100.4(12.3)
Mean birth weight - grams 339.7 (515.7)
Time to spontaneous respiration (mins) 1.26 (0.9)
Gestational age (weeks) 38.9 (1.7)
Family history of late talking
Yes 13.5
Ages and Stages Questionnaire (% abnormal)
Communication score 2.3
Gross motor score 2.0
Fine motor score 3.7
Adaptive score 8.3
Personal–social score 1.6

Table 1 Continued. Maternal, family, and child variables for total sample (N = 1,766a).
Total sample
Variable M (SD) %
Child Behavior Checklist
CBCL Total T score 48.1 (8.4)
CBCL Internalizing T 46.3 (8.4)
CBCL Externalizing T 49.5 (8.3)
Total T score Abnormal 10.4
Total Internalizing T score Abnormal 7.2
Total Externalizing T score Abnormal 16.3
Dimensions of temperament
Higher general activity level 37.1
Higher sleep activity level 38.9
Withdrawal orientation 23.8
Inflexible behavioral style 26.6
Negative mood quality 24.7
Irregularity in sleeping pattern 26.3
Irregularity in eating behavior 23.1
Irregular daily habits 25.5
Low persistence and high distractibility 26.3
Daycare status at interview
In day care 37.4
Hours per week in day care 15.5 (12.9)
a
Missing data were present in 44 of the 47 variables in this table.The average amount of missing data was
2.2% and ranged from 0% to 7.6%. Data imputation methods have been used (see Method section). bDepression
Anxiety Stress Scales (Lovibond & Lovibond, 1995a, 1995b). cParenting scale (Arnold et al., 1993). d*p < .05.
**p < .01. ***p < .001. eRacial status recorded on Midwive’s Notification Form.
population studies of 4- to 11-year-old children using the milestone method to develop a brief language screening
appropriate-for-age CBCL parent-reported measure instrument intended for public health assessments.
(Zubrick et al., 1995). The ASQ developmental mea- The ASQ Communication scale uses six items to
sures ranged from 1.6% (Personal–Social score) to 8.3% assess aspects of the child’s developing skills in speech
(Adaptive score) in the abnormal range, with 2.3% of the production and comprehension. Mothers were asked to
sample having ASQ Communication scores in the ab- report whether their child could (a) point to pictures on
normal range. A little over one third of the children were request, ( b) use two- or three-word phrases, (c) carry out
receiving daycare, with a mean number of 16 hours per simple directions on request, (d) name simple objects,
week. (e) point to body parts on request, and (f ) use personal
pronouns such as “me”, “ I,” and “ you.” The response cat-
egories for each item were (a) “Not Yet,” (b) “Sometimes,”
Determination of LLE and Prevalence and (c) “Yes.”
The scale of the study required an assessment of LLE In our sample, the Communication Scale had a
with minimal effort loading on the part of the parent Cronbach’s alpha of .71, essentially replicating the es-
respondents. The instrument used is the ASQ Commu- timate provided in the test manual. Because the manual
nication scale, which comprises a short list of language does not report validity estimates for the ASQ Com-
milestones drawn from the normative literature by Bricker munication subscale (only for the full instrument), we
and Squires (1999). The Communication scale is part of carried out analyses of criterion and concurrent validity
an instrument developed as a parent-report measure to for our ASQ outcome measure based on item response
screen for developmental impairments. Recently, Luinge, theory (ASQ IRT; described further below). Criterion
Post, Wit, and Goorhuis-Brouwer (2006) followed the same validity is hampered by the lack of an external “gold
standard” measure of LLE (McCardle, Cooper, & Freund, categorical responses. The GRM assumes (a) that the
2005; Tager-Flusberg & Cooper, 1999). We were, however, relationship between ability level and the probability
in a position to assess some aspects of concurrent validity of endorsing a particular item response category (or
against another measure of speech and language collected a higher category) is monotonic, ( b) that the items are
via parent report at the time of the survey. For approxi- unidimensional and have only one common factor, and
mately half of the cohort (N = 902), the LDS had been (c) that ability is distributed normally with a mean of
sought from the parent at the time the questionnaire was zero and a standard deviation of 1, even if the items do
completed. Of the children with LDS data, 888 also had not measure the entire range of the distribution. This
ASQ data. This permitted estimating concurrent valid- third assumption is not a necessary assumption of the
ity for our measure against the LDS. Both the ASQ IRT model but is merely an identification condition to set the
score and the LDS score are continuous variables and scale of ability and may be modified if desired. The major
were moderately correlated (.675; p < .001). Additionally, advantages of the IRT approach over other methods of
for those children for whom we had a parent-completed scaling include (a) use of all items rather than a reliance
LDS, we were able to calculate mean LDS scores for chil- on a single item; ( b) differential adjustment for item dif-
dren differentiated by LLE status on the ASQ. Children ficulty; (c) provisions for appropriate handling of miss-
defined with LLE on the ASQ measure had significantly ing data by determining estimates of ability that are
lower mean LDS scores than those children classified based on all of the items answered and that do not im-
in the normal range on the ASQ measure (MLLE = 62.5, pute the individual’s mean score for missing items;
SD = 52.5 vs. MNormal = 196.2, SD = 70.7; df = 198.1, and (d) use of a continuous estimate of (in this case)
t = 24.7, p < .001). communication /verbal ability, which is on a scale that is
We also assessed the correspondence between the not sample dependent.
LDS item, “ Does your child combine 2 or more words We commenced our assessment of the ASQ Com-
into phrasesI? ” and the ASQ item, “Does your child munication scale by testing the tenability of the dimen-
say 2 or 3 words togetherI? ”. Complete data were sionality assumption. To do this, we used a principal
available on both of these items for 896 of the chil- components analysis. In addition to this traditional
dren. Frequency distributions were obtained on both the analysis, we also used the DETECT algorithm (Stout,
LDS and ASQ items. Ninety percent of children were 1987) which is confirmatory in nature. The results of both
reported, on the LDS, to be combining two or more words of these analyses indicated that the six items represent
into phrases, and 89% of children were reported, on the only one dimension.
ASQ, to be saying two or three words together. Cross- Item characteristic curves ( ICCs; also known as
tabulation of these items indicated complete correspon- item response functions) for each of the six items were
dence of these items for 860 of these cases, c2(1) = then evaluated. For economy of space, an example of one
547.9, p < .001; k = .78. As initial reports of the validity of the items is shown in Figure 2. The lines show the
of the ASQ Communication scale, these findings sug- probability of endorsing a certain response at a given
gest an acceptable level of concurrent validity with an- level of ability. These figures show that with increasing
other measure frequently used to assess early language ability, the probability of a “not yet” response decreases,
emergence. whereas the probability of a “yes” response increases. At
the upper end of the ability scale, there is very little dif-
ference in the probability of a “yes” response. Thus, for
The Graded Response Model Item 2, measuring the use of two- or three-word phrases
To assess the suitability of the ASQ Communication (see Figure 2), a child with an ability of 1 SD above the
scale to identify children with LLE, we undertook an mean would have about the same probability of a “yes”
item response analysis using a type of polytomous item response as an individual with an ability of 3 SDs above
response theory (IRT) model known as the graded re- the mean. The graph in the right panel, the item infor-
sponse model (GRM; Samejima, 1969). The GRM models mation curve, represents how well the item can distin-
each of the three response categories simultaneously, guish or discriminate between different levels of ability.
creating a scaled value representing a person’s overall We can see that Item 2 is best at discriminating in-
ability on the test. In general, Likert-type scales with dividuals with ability near –1.5 SDs. This is where the
fewer than five response choices and a small number of item is most informative and where measurement error
items are difficult to summarize with a single “scale” score is the lowest. Our assessment of each of the ICCs showed
that has a quantifiable standard error of measurement. that the ASQ Communication scale measured the low
The GRM is well suited to the ASQ analyses because end of ability quite well.
it generates an ordering of persons on the ability scale Having determined the item parameters from the
where the responses for the scale are essentially ordered child’s response on each of the six items, these parameters

Figure 2. The item characteristic curve and information curve for Communication Item 2 (“Does your child say two or three words together
that are different ideas, such as ‘see dog,’ ‘mommy come home’ or ‘cat gone’?”).
were then used to create an estimate of each child’s of 19.1% of the sample who were not routinely combining
ability. This estimate gives the child’s most “ likely” abil- words in utterances.
ity level that explains the child’s responses. As shown in
the test information curve in Figure 3, we can see that
the six-item scale provides increasing discrimination
LLE—Bivariate Relationships With
and lower measurement error in the range from –1.0 to Maternal, Family, and Child Characteristics
–1.5 SDs below the mean. The IRT/GRM models do not Comparisons of maternal, family and child char-
generate an exact cutoff point for creating a dichotomous acteristics were made for children differentiated by LLE
variable for LLE, but the choice of the cutoff point is (see Table 2). Alpha levels were not adjusted for family-
guided, in part, by the range of scores within which the wise or study-wise error in order to detect any possible
scale is more precise in discriminating between different differences among the groups. When differences were
ability levels and also by the researcher ’s judgment evident, almost all of them were at conventional levels
based on previous research and clinical factors. of adjustment — that is, <.01 or .001. With respect to
For reasons of clinical benchmarking and to avoid maternal characteristics, no significant differences for
missing children with LLE, we chose –1.0 SD as the children with and without LLE were observed with re-
cutoff to demark those children with and without LLE gard to maternal age at the birth of the child, levels of
(cf. Feldman et al., 2005). Of the 1,766 children, 238 maternal education, mother ’s place of birth, maternal
(13.4%) were classified as having LLE (see Table 2). The uptake of paid employment, and cigarette use. There
13.4% estimate from the IRT composite can be compared were no significant differences among these groups in
to an alternative estimate. Following precedents in the their mean maternal DASS scores nor in the propor-
literature, the ability to combine words at 24 months tions of mothers reporting varying levels of clinical de-
was used as a criterion for grouping children. Of the pression, anxiety, and stress. The only statistically
sample, 10.7% of the children were reported to not com- significant difference observed with regard to maternal
bine words, 8.4% were reported as “sometimes,” and characteristics was in the Parenting score—the mothers
80.9% were reported as “yes,” yielding an overall estimate of children with LLE reported higher mean PS scores
Figure 3. Communication composite information curve.
(M = 2.9, SD = 0.6 vs. M = 2.8, SD = 0.6, ps < .01), with There was no significant difference between LLE
a correspondingly higher proportion falling within the groups on neonatal measures of birthweight, low birth-
clinical range (36.9% vs. 27.6%), c2(1) = 8.67, p < .01, weight status, and time to spontaneous respiration. How-
denoting a higher level of dysfunctional parenting. ever, children with LLE were significantly more likely to
Within families, LLE was associated with a family be born weighing less than 85% of their optimal birth-
history of late talking (22.2% vs. 12.1%), c2(1) = 18.2, weight (14.7% vs. 8.2%), c2(1) = 10.5, p < .01, and at less
p < .001, and with larger family size as measured by the than 37 weeks’ gestation. Because gestational age —
number of children in the family. When compared with specifically, prematurity—is frequently cited as a con-
children who did not have LLE, children with LLE were founding factor for LLE, separate investigation of this as
less likely to be the only child (20.1% vs. 31.4%), c2(3) = a possible threat to the validity of the findings is reported
16.6, p < .001. Otherwise, there were no significant dif- below.
ferences in the family characteristics of children with and With regard to development, significantly higher
without LLE in terms of family type (i.e., two-parent, proportions of children with LLE were in the abnormal
sole-parent), income, area level indicators of SES, and range on the ASQ Gross Motor, Fine Motor, Adaptive,
family function. and Personal–Social scores. Results on the ASQ Com-
With respect to characteristics of the child, there munication score, which is calculated from the six var-
were several significant differences between children iables used to define LLE status, revealed all children
with and without LLE. Children with LLE were signif- with LLE to fall in the abnormal range of the Communi-
icantly more likely to be male (70.8% vs. 47.6%), c2(1) = cation score. In terms of behavioral and emotional ad-
44.3, p < .001. Although comparisons of their mean ages justment, significantly higher proportions of children with
showed children with LLE to be significantly younger LLE were in the abnormal range on the parent-reported
(M = 2.08 years, SD = 0.104 vs. M = 2.11, SD = 0.135, CBCL Total Score (15.6% vs. 9.6%), c2(1) = 7.94, p < .001,
p < .001), this equates to a mean difference of 10 days in with corresponding and statistically significant elevations
age between these groups. In practical terms, 99.8% of in CBCL Internalizing problems (11.0% vs. 6.7%), c2(1) =
the children were between the ages of 23 and 24 months 5.64, p < .001, and Externalizing problems (23.8% vs.
of age. 15.1%), c2(1) = 11.4, p < .01.

Table 2. Maternal, family, and child variables for control and LLE children (N = 1,766a).
Language emergence
Normal (N = 1,528) Late (N = 238)
Variable M (SD) % M (SD) % c2 (t )b
Maternal characteristic
Age at child’s birth (years) 29.5 (5.0) 29.2 (4.7) (0.79)
≤19 years 2.5 2.6 .99
20–34 years 81.8 84.2
35+ years 15.7 13.2
Education
<12 years 23.9 29.6 6.82
12 years 19.3 18.5
Trade certificate 13.3 12.4
Postschool study 13.4 15.9
Completed postschool qualification 30.1 23.6
Mother’s place of birth
Australia 74.5 76.5 4.26
United Kingdom 15.0 13.2
New Zealand 4.0 3.4
Asia and India 2.2 3.8
Europe 1.7 1.3
North America 1.6 1.3
Africa 1.1 0.4
Currently in paid employment
Yes 40.7 36.7 1.37
Hours per week in paid employment 22.0 (12.7) 21.7 (12.9) (0.37)
Cigarette use
Smoked before/during pregnancy 28.0 30.0 0.41
Current smoker 19.6 23.0 1.48
Depression 3.6 (5.3) 4.0 (5.5) (–1.22)
Anxiety 1.9 (3.3) 2.1 (3.6) (–0.99)
Stress 7.4 (6.8) 7.6 (6.5) (–0.34)
Clinical ranges
Depression 0.47
Normal 90.3 89.1
Mild 4.1 4.2
Moderate to severe 5.6 6.7
Anxiety 0.51
Normal 94.5 93.3
Mild 1.9 2.1
Stress 2.98
Normal 87.1 84.2
Mild 6.7 9.8
Parenting score 2.8 (0.6) 2.9 (0.6) (–3.04)**
In clinical range 27.6 36.9 8.67**
Family structure
Original 84.9 82.8 0.73
Step/blended 5.0 5.6
Table 2 Continued. Maternal, family, and child variables for control and LLE children (N = 1,766a).
Language emergence
Normal (N = 1,528) Late (N = 238)
Family structure
Sole parent 7.7 9.0
Other 2.4 2.6
Family size
No. children 2.0 (0.9) 2.2 (1.0) (–2.85)**
1 31.4 20.1 16.6***
2 68.6 79.9
Income
$1–$154 wk / $1–$8,000 yr 0.6 0.9 9.41
$155–$308 wk / $8,001–$16,000 yr 5.5 5.6
$309–$481 wk / $1,6001–$25,000 yr 10.4 16.7
$482–$769 wk / $25,001–$40,000 yr 26.0 26.1
$770–$961 wk / $40,001–$50,000 yr 17.3 16.7
$962 or more wk / $50,001 or more yr 36.4 31.2
Not stated 3.7 3.0
Area SES indicators
Disadvantage 1008.6 (59.6) 1004.0 (59.3) (1.11)
Resources 1028.6 (63.3) 1028.4 (63.3) (0.05)
Occupation/education 988.3 (76.2) 981.0 (75.8) (1.36)
Disadvantage (–1.0 SD) 3.4 3.3 0.01
Resources (–1.0 SD) 1.7 1.4 0.11
Occupation/education (–1.0 SD) 6.8 8.5 0.91
Family functionc
Abnormal family function 9.1 11.8 1.75
Yes 12.1 22.2 18.2***
Gender (male) 47.6 70.8 44.3***
Age (years) 2.11 (0.135) 2.08 (0.104) (3.64)***

Race
Caucasian 96.6 96.4 0.34
Aboriginal 0.8 0.7
Other 2.3 2.9
Neonatal period
<85% POBW 8.2 14.7 10.5**
Low birthweight (<2,500 g) 3.4 5.9 3.56
<37 weeks’ gestation 5.5 10.7 9.6**
Mean POBW 100.6 (12.0) 99.5 (14.0) (1.28)
Mean birthweight (g) 3446.4 (498.0) 3396.9 (615.7) (1.35)
Time to spontaneous respiration (min) 1.3 (0.9) 1.3 1.0 (– 0.83)
Gestational age (weeks) 39.0 (1.6) 38.8 (2.1) (1.86)
Communication Score 0.0 16.9 264.3***
Gross Motor Score 1.4 5.9 21.2***
Fine Motor Score 2.7 8.1 18.1***

Table 2 Continued. Maternal, family, and child variables for control and LLE children (N = 1,766a).
Language emergence
Normal (N = 1,528) Late (N = 238)
Adaptive Score 6.3 21.1 59.3***
Personal–Social Score 0.6 8.1 73.1***
Total T score 47.8 (8.3) 49.8 (8.8) (–3.42)***
Internalizing T score 46.0 (8.3) 48.0 (8.8) (–3.46)***
Externalizing T score 49.2 (8.2) 51.1 (9.0) (–3.16)**
Total T score Abnormal 9.6 15.6 7.94**
Total Internalizing T score Abnormal 6.7 11.0 5.64*
Total Externalizing T Score Abnormal 15.1 23.8 11.43***
Dimension of temperament
Higher general activity level 36.9 38.2 0.15
Higher sleep activity level 38.0 44.1 3.22
Withdrawal orientation 23.6 25.0 0.22
Inflexible behavioral style 26.3 28.4 0.46
Negative mood quality 23.7 31.3 6.39*
Irregularity in sleeping pattern 25.6 31.3 3.44
Irregularity in eating behavior 23.2 22.6 0.04
Irregular daily habits 24.9 29.2 2.00
Low persistence and high distractibility 25.7 30.3 2.25
Daycare status at interview

In daycare 37.8 34.6 0.90
Mean hours in daycare 15.2 (17.3) 14.4 (–1.38)
Note. LLE = late language emergence; SES = socioeconomic status; POBW = Proportion of Optimal Birthweight.
a
A total of 49 participants had missing data on the outcome variable (“late talker ” status). b*p < .05. **p < .01. ***p < .001. c General Factor of the
McMaster Family Assessment Device (Byles, Byrne, Boyle, & Offord, 1988).
The only temperament difference between those variables (Hosmer & Lemeshow, 1989). The predictor
children with and without LLE was in negative mood variables may be continuous, dichotomous, discrete, or
quality. Relative to children without LLE, a significantly a mix of these types. Estimated effects of the predictor
greater proportion of children with LLE were reported variables are multivariately adjusted for the effects of
by their mothers to have negative mood quality (31.3% the other predictors. In this study, the associations be-
vs. 23.7%), c2(1) = 3.44. p < .05. Finally, there was no tween the outcome variable ( LLE) and the candidate
difference in the proportion of children with and without predictor variables were expressed as odds ratios. An
LLE who were enrolled in daycare nor in the amount of odds ratio is the ratio of the probability of an event’s oc-
daycare they received as measured by mean number of currence to the probability of the event’s nonoccurrence.
hours. In this study, the “event” is LLE, and because LLE is an
adverse outcome, the predictor variables are “risk ” var-
iables. Where predictors are categorical, these odds ra-
LLE—Multivariate Relationships With tios are calculated with reference to a specific base or
Maternal, Family, and Child Characteristics “reference” category.
The numerous relationships of maternal, family, The candidate predictor variables were selected from
and child characteristics with LLE (see Table 2) were Table 2. In fitting the logistic model, virtually all variables
further investigated using multivariate logistic regres- were used and, following Hosmer and Lemeshow (1989),
sion. Logistic regression allows the prediction of a discrete, most were coded to be categorical, rather than contin-
binary outcome (in this case, LLE) from a set of predictor uous. Two exceptions were made. First, the mother’s
country of birth was not entered in the model. The dis- score (OR = 2.39, 95% CI = 1.19–4.77), Adaptive score
tribution of this variable reflects differential bias in the (OR = 2.64, 95% CI = .66–4.21), and Personal–Social
exclusion of cases owing to English language require- Score (OR = 5.52, 95% CI = 2.05–14.86).
ments. Second, the child’s age in months at the time of
the interview was entered as a continuous variable. All
other variables were coded as categorical variables (see Potential Threats to Validity
Table 2).
These findings are based upon a well-defined and
To account for data imputation procedures (described well-described sample of children aged 2 years. Exclu-
in the first paragraph of the Results section), we under- sions from this sample included non-English back-
took logistic regression using SAS 9.1 (PROC LOGISTIC ground and medical conditions or syndromes known at
and PROC MIANALYZE; SAS Institute, 2004). Instead the time of the 2-year observation. To what extent might
of filling in a single value for each missing value, these “covert” disability—that is, conditions not known at the
procedures combine the results of the analyses of im- time of the 2-year assessment but associated with LLE—
putations and generate valid statistical inferences by impart bias to these findings? Although the focus of these
replacing each missing value with a set of plausible val- findings is on the phenomenology of LLE at 2 years, the
ues that represent the uncertainty about the correct study children were followed until 8 years of age.
value to impute (Rubin, 1976, 1987).
Subsequent examination revealed that 19 additional
All variables were entered into the model in a single children developed syndromal conditions that poten-
step with LLE as the response variable. For each of the tially were related to LLE. These children were assessed
predictor variables, parameter estimates ( betas), their on the ASQ Communication scale at age 2 years, and 37%
standard errors, 95% confidence intervals (CIs), degrees were in the normal range, whereas 63% were classified
of freedom, t values, and their probabilities, along with as having LLE. Of the 19 children, 10 were subsequently
the odds ratios and their 95% CIs, are shown in Table 3. found to have intellectual disabilities, 4 were diagnosed
There were no statistically significant associations with autism spectrum disorders, and the remaining 5
between the various maternal characteristics and LLE. were diagnosed with developmental syndromal condi-
No significant associations between LLE and maternal tions. The multivariate analysis (see Table 3) was re-
education, age, smoking, psychological state, or parent- peated without these children. Only one change occurred
ing style were observed. in the estimates: Prematurity was no longer a signifi-
In the variables characterizing the family, LLE was cant predictor of LLE status.
significantly associated with the number of children in Further inspection of the data revealed that an ad-
the family. Relative to singleton children, those children ditional 7 children had been born at less than 31 weeks’
with LLE were significantly more likely to have one or of gestation. Six of these children had ASQ Communi-
more siblings (odds ratio [OR] = 2.07, 95% CI = 1.39–3.09). cation scale scores. Fifty percent of these children were
Relative to families without a history of late talking, chil- measured at age 2 years to have LLE. All 7 of these chil-
dren with LLE were significantly more likely to be born dren were subsequently removed from the multivariate
to families in which a parent has a history of late talking analysis, along with the 19 children found later to have
(OR = 2.11, CI = 1.39–3.19). All other statistical associa- syndromal conditions. Aside from the nonsignificance of
tions between LLE and the set of family variables were gestational age, results revealed no substantive changes
nonsignificant. This included family type, income, local to those reported in Table 3.
area disadvantage, low economic resources, and low edu-
cation and occupational status, family function, and day-
care use.
Several characteristics of the child were associated
Discussion
with LLE status. Relative to female children, male chil- In this study of a large number (N = 1,766) of epi-
dren were significantly more likely to have LLE (OR = demiologically ascertained 24-month-old children, early
2.74, 95% CI = 1.96–3.83). LLE children were more language acquisition was assessed via a six-item parent
likely to be born at 32 weeks’ or less gestation (OR = 1.84, report scale that combined comprehension and produc-
95% CI = 1.04–3.25) and weigh 85% or less of their op- tion benchmarks. Item response analyses found the
timal birthweight (OR = 1.89, 95% CI = 1.18–3.01). All composite to measure the low end of ability quite well,
ASQ variables were significantly associated with LLE. providing acceptable levels of discrimination and mea-
Relative to children in each of the respective normal surement error. With a criterion of –1 SD from the mean,
categories, children with LLE were more likely to fall in 13.4% of the sample was identified as showing LLE.
the abnormal range of the ASQ on measures of Gross Using the criterion of “no or only occasional word com-
Motor score (OR = 3.12, 95% CI = 1.29–7.51), Fine Motor binations,” 19.1% of the sample was identified.

1582
Table 3. Multivariate logistic regression: Prediction of LLE status by maternal, family, and child variables (bolded entries are significant).
Journal of Speech, Language, and Hearing Research • Vol. 50 • 1562–1592 • December 2007
Parameter Estimate SE 95% CI min 95% CI max df t p Odds ratio 95% CI min 95% CI max
Intercept 3.7298 2.0435 –0.2755 7.7351 26164 1.83 0.068
Maternal factors
Educational level
<12 years –0.1064 0.2371 –0.5713 0.3586 2147.5 –0.45 0.654 0.899 0.565 1.431
12 years 0.0258 0.2539 –0.4722 0.5238 1876.4 0.1 0.919 1.026 0.624 1.688
Trade certificate –0.3990 0.2321 –0.8543 0.0563 1358.5 –1.72 0.086 0.671 0.426 1.058
Postschool study –0.1079 0.2769 –0.6510 0.4353 1274.9 –0.39 0.697 0.898 0.522 1.545
Completed postschool qualification ref 1.000
Age at birth of child
≤19 years –0.1813 0.5941 –1.3482 0.9855 580.43 –0.31 0.760 0.834 0.260 2.679
20 –34 years –0.4144 0.6277 –1.6463 0.8176 865.16 –0.66 0.509 0.661 0.193 2.265
35 + years ref 1.000
Employment status
No paid employment 0.1608 0.1780 –0.1882 0.5098 5108.8 0.9 0.366 1.174 0.828 1.665
In paid employment ref 1.000
Cigarette use
Nonsmoker during pregnancy ref 1.000
Smoked during pregnancy –0.2403 0.2707 –0.7725 0.2920 372.93 –0.89 0.375 0.786 0.462 1.339
Current nonsmoker ref 1.000
Current smoker –0.3242 0.2970 –0.9084 0.2600 343.89 –1.09 0.276 0.723 0.403 1.297
Depression Anxiety Stress Scale (DASS)
Depress – Normal ref 1.000
Depress – Mild –0.2274 0.4140 –1.0394 0.5846 1970.3 –0.55 0.583 0.797 0.354 1.794
Depress – Severe –0.1514 0.4810 –1.0946 0.7918 2569.6 –0.31 0.753 0.859 0.335 2.207
Anxious – Normal ref 1.000
Anxious – Mild 0.2318 0.6171 –0.9777 1.4414 9287.8 0.38 0.707 1.261 0.376 4.227
Anxious – Severe –0.2373 0.5379 –1.2918 0.8171 5106.2 –0.44 0.659 0.789 0.275 2.264
Stress – Normal ref 1.000
Stress – Mild 0.2232 0.3000 –0.3650 0.8115 2055.9 0.74 0.457 1.250 0.694 2.251
Stress – Severe –0.2487 0.4774 –1.1847 0.6874 3775 –0.52 0.603 0.780 0.306 1.988
Parenting Scale
Nonclinical ref 1.000
Clinical 0.3284 0.2010 –0.0758 0.7326 47.432 1.63 0.109 1.389 0.927 2.081

Table 3 Continued. Multivariate logistic regression: Prediction of LLE status by maternal, family, and child variables (bolded entries are significant).
Family factors
Family type
Original ref 1.000
Step/blended – 0.0459 0.3585 – 0.7487 0.6569 11048 –0.13 0.898 0.955 0.473 1.929
Sole parent 0.1704 0.3552 – 0.5260 0.8667 5459.5 0.48 0.632 1.186 0.591 2.379
Other – 0.3894 0.5275 – 1.4239 0.6451 2027.4 –0.74 0.461 0.677 0.241 1.906
Number of children in family
Number of children – 1 ref 1.000
Number of children – 2 or more 0.7315 0.2037 0.3322 1.1309 5836.8 3.59 0.000 2.078 1.394 3.099
Family income
<8K 0.1298 1.0090 – 1.8481 2.1077 10423 0.13 0.898 1.139 0.158 8.229
$8–$16K 0.5797 0.9896 – 1.3600 2.5193 23772 0.59 0.558 1.785 0.257 12.420
$16–$25K – 0.0232 0.9946 – 1.9727 1.9262 282892 –0.02 0.981 0.977 0.139 6.864
$25–$40K 0.1267 1.0046 – 1.8423 2.0958 94872 0.13 0.900 1.135 0.158 8.132
$40–$50K 0.0980 1.0015 – 1.8650 2.0609 72746 0.1 0.922 1.103 0.155 7.853
$50K or more ref 1.000
Socioeconomic indicators for area (SEIFA)

SEIFA disadvantage – No ref 1.000
SEIFA disadvantage – Yes – 1.2546 0.6877 – 2.6158 0.1066 123.52 –1.82 0.071 0.285 0.073 1.113
SEIFA low econ resource – No ref 1.000
SEIFA low econ resource – Yes 0.5416 0.7679 – 0.9694 2.0526 303.35 0.71 0.481 1.719 0.379 7.788
SEIFA low educ occ – No ref 1.000
SEIFA low educ occ – Yes 0.4475 0.3568 – 0.2541 1.1492 377.35 1.25 0.211 1.564 0.776 3.156
Family function
Zubrick et al.: Late Language Emergence
Normal ref 1.000

Poor 0.2721 0.2657 – 0.2491 0.7932 2162.1 1.02 0.306 1.313 0.780 2.210
No ref 1.000
Family history late talking – Yes 0.7480 0.2103 0.3348 1.1612 511.21 3.56 0.000 2.113 1.398 3.194
Currently in daycare
No ref 1.000
Yes – 0.1239 0.1790 – 0.4748 0.2270 19077 –0.69 0.489 0.883 0.622 1.255

1583
1584
Journal of Speech, Language, and Hearing Research • Vol. 50 • 1562–1592 • December 2007
Child factors
Gender
Female ref 1.000
Male 1.0087 0.1714 0.6726 1.3449 2754.1 5.88 <.0001 2.742 1.959 3.838
Proportion of optimal birth weight
POBW >85% ref 1.000
POBW <85% 0.6370 0.2383 0.1699 1.1040 58026 2.67 0.008 1.891 1.185 3.016
Ages and Stages Questionnaire results
Gross Motor – normal ref 1.000
Gross Motor – abnormal 1.1381 0.4484 0.2590 2.0171 5400.3 2.54 0.011 3.121 1.296 7.517
Fine Motor – normal ref 1.000
Fine Motor – abnormal 0.8717 0.3529 0.1794 1.5641 1081.3 2.47 0.014 2.391 1.196 4.778
Adaptive score – normal ref 1.000
Adaptive score – abnormal 0.9730 0.2372 0.5074 1.4386 769.19 4.1 <.0001 2.646 1.661 4.215
Personal–social – normal ref 1.000
Personal–social – abnormal 1.7099 0.5036 0.7208 2.6991 575.61 3.4 0.001 5.529 2.056 14.867
Total T Score – Normal ref 1.000
Total T Score – Abnormal 0.1111 0.3407 –0.5567 0.7790 34510 0.33 0.744 1.118 0.573 2.179
Internalizing – Normal ref 1.000
Internalizing – Abnormal 0.5158 0.3395 –0.1524 1.1839 288.85 1.52 0.130 1.675 0.859 3.267
Externalizing – Normal ref 1.000
Externalizing – Abnormal 0.2529 0.2730 –0.2847 0.7904 270.79 0.93 0.355 1.288 0.752 2.204
Dimension of Temperament scale
High general activity – no ref 1.000
High general activity – yes –0.0704 0.1819 –0.4275 0.2868 583.11 –0.39 0.699 0.932 0.652 1.332
High sleep activity – no ref 1.000
High sleep activity – yes 0.2410 0.1735 –0.1004 0.5825 283.57 1.39 0.166 1.273 0.904 1.790
Withdrawal orientation – no ref 1.000
Withdrawal orientation – yes –0.0695 0.2070 –0.4774 0.3384 224.95 –0.34 0.737 0.933 0.620 1.403
Inflexible style – no ref 1.000
Inflexible style – yes –0.0442 0.1960 –0.4298 0.3414 323.49 –0.23 0.822 0.957 0.651 1.407
Negative mood quality – no ref 1.000
Negative mood quality – yes 0.2970 0.1879 –0.0718 0.6658 940.91 1.58 0.114 1.346 0.931 1.946
Irregular sleep pattern – no ref 1.000

Child factors
Dimension of Temperament scale
Irregular sleep pattern – yes 0.1396 0.2214 –0.3058 0.5849 47.243 0.63 0.532 1.150 0.737 1.795
Irregular eat pattern – no ref 1.000
Irregular eat pattern – yes –0.2039 0.2114 –0.6193 0.2114 516.33 –0.96 0.335 0.816 0.538 1.235
Irregular daily habits – no ref 1.000 1.000 1.000
Irregular daily habits – yes 0.1341 0.1859 –0.2303 0.4984 52886 0.72 0.471 1.143 0.794 1.646
Low persist high distract – no ref 1.000
Low persist high distract – yes 0.2992 0.1856 –0.0650 0.6633 1006.1 1.61 0.107 1.349 0.937 1.941
Time to spontaneous respiration
<2 min ref 1.000
>2 min 0.0939 0.3028 –0.4996 0.6874 91758 0.31 0.756 1.098 0.607 1.989
Premature birth
No >36 weeks ref 1.000
Yes ≤ 36 weeks 0.6107 0.2898 0.0426 1.1788 4349.3 2.11 0.035 1.842 1.044 3.250
Zubrick et al.: Late Language Emergence
Age months –0.0094 0.0023 –0.0140 –0.0049 6988.8 –4.06 <.0001 0.991 0.986 0.995
Note. CI = confidence interval; econ = economic; educ = educational; occ = occupation; qual = quality; persist = persistence; distract = distraction
1585
Bivariate relationships with maternal, family, and child multivariate evaluation of the child’s biogenetic endow-
characteristics found the following maternal variables ment as well as the proximal (maternal and family) and
to be nonsignificant: age at child’s birth, education, birth- distal (societal) resources available to the child. The
place, paid employment, cigarette use, depression levels, large number of null findings is a noteworthy outcome.
anxiety, and stress. A parenting instrument found that Although the literature suggests positive predictor sta-
mothers of LLE children were more likely to use dys- tus for maternal education, maternal depression, family
functional parenting practices. Children with LLE were SES, and parental occupation, none of these variables
more likely to have a positive family history of late predicted LLE, either in bivariate comparisons or in
talking and were less likely to be only children. Non- multivariate analyses. Simply put, in this large and
predictors included family type (e.g., two-parent vs. sole- diverse sample of children and families, risk for LLE
parent), income, SES, family function, daycare enrolment, at 24 months was not associated with particular strata
or amount of time in daycare. At the child level, children of parental educational levels, socioeconomic resources,
with LLE were more likely to be male and younger ( by parental mental health, parenting practices, or family
a mean of 10 days’ difference). Neonatal nonpredictors functioning. Put another way, children with lower levels
included birthweight, low birthweight status, and time of proximal and distal resources are as likely as children
to spontaneous respiration; significant predictors were with higher levels of these resources to be beyond the
percentage of optimal birthweight and less than 37 weeks’ LLE category at 24 months.
gestation. Concurrent predictors at 24 months included The only environmental risk for LLE that we iden-
gross and fine motor development, adaptive scores, tified was the presence of siblings. There was a twofold
personal–social scores, psychosocial development, and increase in the risk for LLE for children with siblings,
temperament (i.e., negative mood quality). relative to only children. Although we did not examine
Multivariate analyses yielded the following signif- birth order effects directly, firstborn children are tem-
icant predictors, listed in order of odds ratio, from high- porary “only ” children, so our outcome is consistent with
est to lowest: Personal–social levels, gross motor skills, studies that report advantages in language development
gender, adaptive motor skills, fine motor skills, family for firstborn children that are attributed to the quantity
history, number of children, proportion of optimal birth- and quality of maternal speech (Fenson et al., 1994; Hoff-
weight, prematurity, and age. Ginsberg, 1998). According to the resource dilution model,
the addition of even one sibling would halve home re-
sources for language acquisition (Downey, 2001). In this
Prevalence study, the risk conferred by siblings was independent
Our estimate of 13.4% LLE in the general popula- of other home resources. It is possible that the number
tion falls in the same range as previous estimates that of children in the family may be a more sensitive proxy
have varied between 10% and 20% (Fenson et al., 1994; measure of home resources for language acquisition in
Horwitz et al., 2003; Klee et al., 1998; Rescorla, 1989; the low performance range than measures such as ma-
Rescorla & Achenbach, 2002; Rescorla et al., 1993). In ternal education and SES.
this sample, the prevalence estimate of 13.4% using a Alternatively, evidence of possible neurobiological
composite index of Receptive and Expressive Language and genetic contribution to LLE was more abundant.
was more conservative than 19.1% using the Expressive The first of these risk indicators was present at birth and
Language criterion, “combining words.” Our overall es- was related to male gender and suboptimal fetal growth.
timate of 19.1% of the sample who were not routinely A disproportionate number of male children had delayed
combining words at 24 months was comparable to the 19% language development, a finding that aligns strongly with
estimate for the CDI sample at 25 months (Bates, Dale, previous studies. Male children were at almost three
& Thal, 1996) and the ALSPAC sample at 25 months times the risk for LLE compared with female children.
(Roulstone, Loader, Northstone, Beveridge, & the ALSPAC In contrast to the strong disadvantage for male children
team, 2002). With specific regards to family history, 23% in the low performance range, there is only a modest ad-
of those children in families reporting a family history of vantage for female children across the full range of per-
late talking were found to have LLE as measured in this formance (Fenson et al., 1994; Huttenlocher et al., 1991;
study, versus 12.0% of those children in families who Wells, 1985). Children who were less than 85% of their
reported no such history. optimum birthweight or who were born earlier than
37 weeks’ gestation were at almost twice the risk for LLE.
As developed by Blair et al. (2005), the proportion of op-
Predictors timal birthweight is a population-based estimate of fetal
The comprehensive framework provided by the growth that is a more differentiated measure of fetal
Bronfenbrenner (1979) model established a set of po- growth than absolute birthweight. Our findings suggest
tential predictors unprecedented in the literature for that it is more sensitive to LLE than birthweight that is
not associated with risk for language delay (Paul, 1991; cohorts (Ellis Weismer et al., 1994; Paul, 1991; Rescorla
Rescorla et al., 1993; Whitehurst et al., 1992) or SLI & Schwartz, 1990; Whitehurst et al., 1991), and all but
(Tomblin et al., 1997). At the same time, there was no Whitehurst and colleagues (1991) reported an elevated
difference in the physical condition of the children at rate of affectedness for family members of late talkers.
birth, referenced to the time it took for the children to Negative mood quality, abnormal child behavior,
breathe independently. Prenatal, perinatal, and obstetric and dysfunctional parenting did not contribute to the
risks have not been implicated previously in the eti- risk for late language onset but were more frequent in
ology of LLE, although empirical data are scarce (Paul, children with LLE compared with control children.
1991; Rescorla et al., 1993; Whitehurst et al., 1992). The Differences in temperament, behavior, and parenting
results of this large epidemiological study that had ac- of children with and without early language delay at
cess to medical information collected at the time of de- 24 months have been reported previously (Carson et al.,
livery showed that some risks for LLE were present at the 1998; Carson, Perry, Diefenderfer, & Klee, 1999; Irwin
moment of birth. et al., 2002; Paul, Looney, & Dahm, 1991). Plomin and
Our finding that lower levels of motor, adaptive, and colleagues (2002) reported modest genetic associations
personal–social performance were predictors of LLE ex- between behavior problems and verbal and nonverbal
tends the findings of previous smaller-scale studies that abilities in 2-year-old twins. This finding provides pro-
compared late talkers and controls and reported lower visional support for the view that common biogenetic
levels of performance on concurrent measures of general mechanisms influence problematic temperament, ab-
development at 24 months (Carson et al., 1998; Klee normal behavior, and language delay in children. An
et al., 1998; Rescorla & Alley, 2001). The toddlers in alternative view is that language delay mediates child
previous studies—and the toddlers in our sample—did temperament and behavior. The conclusion we can draw
not have developmental conditions that might account from our data is that problematic child temperament,
for the group differences reported in previous studies or abnormal child behavior, and dysfunctional parenting
the significant prediction in our study. Our study cannot are more likely to be part of the psychosocial profile of
address, for example, the extent to which lower levels of late talkers than children with normal language devel-
motor performance or personal–social development are opment. The caveat is that the results here are inher-
etiological or phenotypic. ently ambiguous; the direction of influence is undetermined,
A further complication is that although we treated and the full interpretation is likely to be quite complex.
personal–social skills as a predictor in the analyses, the Overall, the results of this study are congruent with
domain of personal–social skills is difficult to interpret. models of language emergence and impairment that
Measurement confounds are an issue. The item that was posit a strong role for neurobiological and genetic mech-
most discriminating was a linguistic item—that is, “Does anisms of onset that operate across a wide variation in
your child call himself / herself ‘ I ’ or ‘me’ more often than maternal and family characteristics. This study points
his/ her own name?” Confounding of language emer- toward familial history of LLE, male gender, and early
gence and personal–social skills in this age range is a neurobiological growth as concomitant indicators of risk
difficult one to avoid in early assessments thus requires at 24 months.
caution in interpretation. Further, children with LLE
may find it more difficult to establish social interactions
because of their language limitations. It will be difficult Directions for Future Research
to sort out predictive status for this variable.
The import of LLE can be viewed in terms of recent
Yet, in general, our results indicate that children growth model studies of children with SLI (cf. Hadley
with LLE lag behind control children in multiple dimen- & Holt, 2006; Rice, Redmond, & Hoffman, 2006; Rice,
sions of development, and this maturational lag features Wexler, & Hershberger, 1998; Rice, Wexler, Marquis, &
in either the etiology or the phenotype of LLE. Although Hershberger, 2000; Rice, Tomblin, Hoffman, Richman,
temperament was not a risk factor for LLE, negative & Marquis, 2004). A consistent finding is that the af-
mood quality is biologically regulated and provides addi- fected group differs from unaffected children in the inter-
tional support for the role of maturational lag in the cept but not the trajectory of change over time, pointing
etiology of late language onset. to delayed onset of language as an important part of the
As suggested by Feldman et al. (2005), a positive phenotype of language impairment (cf. Rice, in press;
family history of late talking is predictive of LLE. Chil- Rice, 2007; Rice & Smolik, 2007; Rice, Warren, & Betz,
dren with positive histories had double the risk for LLE 2005, for more complete discussion). There is an impor-
compared with children in families with no family his- tant empirical gap, however, in fleshing out the connec-
tory, suggesting a genetic liability for LLE. Family ag- tion between LLE and later SLI. LLE status at 24 months
gregation data have been reported for four late talker is a limited predictor of later language impairment (cf.

Dale et al., 2003) because an indeterminate number of Bates, E., Dale, P. S., & Thal, D. (1996). Individual dif-
children “outgrow” their delayed onset of language. Two ferences and their implications for theories of language
development. In P. Fletcher & B. MacWhinney ( Eds.),
kinds of evidence are needed to help sort this out: The The handbook of child language ( pp. 96–151). Cambridge,
first is an accurate estimate of the proportion of chil- MA: Blackwell.
dren with LLE who at later ages are identified as having Bayley, N. (1969). Bayley Scales of Infant Development.
SLI; the second is identification of predictors of sub- San Antonio, TX: The Psychological Corporation.
sequent language impairment. Blair, E. (1996). Why do Aboriginal neonates weigh less? II.
It would be premature to use the predictors Determinants of birth weight for gestation. Journal of
evaluated here for screening purposes (i.e., identifying Paediatrics and Child Health, 32, 498–503.
2-year-olds at risk for subsequent SLI) because this use Blair, E. M., Liu, Y., de Klerk, N. H., & Lawrence, D. M.
requires measures and data collected at time points (2005). Optimal fetal growth for the Caucasian singleton
and assessment of appropriateness of fetal growth: Analysis
beyond the 24-month period of the current panel of data of a total population perinatal database. BMC Pediatrics,
on which this article is based. Longitudinal data are needed 5, 1471–2431.
to investigate whether predictors of LLE at 24 months Bornstein, M. H., Tamis-LeMonda, C. S., & Maurice
also predict language impairment later or whether the Haynes, O. (1999). First words in the second year: Con-
early predictors are modulated or supplanted later by tinuity, stability, and models of concurrent and predictive
other variables that did not predict LLE. That is, it correspondence in vocabulary and verbal responsiveness
is conceivable that predictive relationships change over across age and context. Infant Behavior & Development,
22, 65–85.
time. For example, maternal or family variables may
Breznitz, Z., & Sherman, T. (1987). Speech patterning of
come into play as a child’s vocabulary accelerates or as
natural discourse of well and depressed mothers and their
clausal structures emerge. It will be important to eval- young children. Child Development, 58, 395– 400.
uate the extent to which predictive relationships are
Bricker, D., & Squires, J. (1999). Ages and Stages Ques-
the same across all language dimensions. Further, long- tionnaires: A parent-completed, child-monitoring system
term outcome data can help sort out possible differences (2nd ed.). Baltimore: Paul H. Brookes.
in predictive relationships between children in the low Bronfenbrenner, U. (1979). The ecology of human develop-
performance range versus children in the normative range. ment. Cambridge, MA: Harvard University Press.
Evidence of this sort will provide a foundation for the de- Carson, D. K., Klee, T., Perry, C. K., Muskina, G., &
velopment of comprehensive etiological models of LLE Donaghy, T. (1998). Comparisons of children with delayed
and SLI. and normal language at 24 months of age on measures of
behavioral difficulties, social and cognitive development.
Infant Mental Health Journal, 19, 59–75.
Acknowledgments Carson, D. K., Perry, C. K., Diefenderfer, A., & Klee, T.
(1999). Differences in family characteristics and parenting
This work was funded by grants from the Health behavior in families with language-delayed and language-
Promotion Foundation of Western Australia and from Grants normal toddlers. Infant-Toddler Intervention: The Trans-
RO1DC05226, P30DC005803, and P30HD002528 from the disciplinary Journal, 9, 259–279.
National Institutes of Health. The RASCALS study was Caulfield, M. B., Fischel, J., DeBaryshe, B. D., &
originally created by Jennifer Kurinczuk, who provided details Whitehurst, G. J. (1989). Behavioral correlates of develop-
of the original sampling plan. Thanks to Meryl Biggs, Kaye mental expressive language disorder. Journal of Abnormal
Moore, and Deborah Parsons for help in maintaining the Child Psychology, 17, 187–201.
cohort and collection of data. We appreciate statistical Clayton, D., & Hills, M. (1996). Statistical models in epi-
consultation by Janet Marquis and Jonathan Templin. We demiology. London: Oxford University Press.
especially thank the children and families who participated Dale, P. S., Price, T. S., Bishop, D. V. M., & Plomin, R.
in the RASCALS study. (2003). Outcomes of early language delay: I. Predicting
persistent and transient delay at 3 and 4 years. Journal
of Speech, Language and Hearing Research, 46, 544 –560.
References Dale, P. S., Simonoff, E., Bishop, D. V. M., Eley, T. C.,
Oliver, B., Price, T. S., et al. (1998). Genetic influence on
Achenbach, T. M., & Edelbrock, C. S. (1991). Manual for the language delay in two-year-old children. Nature Neuro-
Child Behavior Checklist/2-3 and 1992 profile. Burlington, science, 1, 324–328.
VT: University of Vermont. Dollaghan, C. A., Campbell, T. F., Paradise, J. L., Feldman,
Arnold, D. S., O ’Leary, S. G., Wolff, L. S., & Acher, M. M. H. M., Janosky, J. E., Pitcairn, D. N., et al. (1999).
(1993). The Parenting Scale: A measure of dysfunctional Maternal education and measures of early speech and lan-
parenting in discipline situations. Psychological Assessment, guage. Journal of Speech, Language, and Hearing Research,
5, 137–144. 42, 1432–1443.
Australian Bureau of Statistics. (1998). 1996 census of Downey, D. B. (2001). Number of siblings and intellectual
population and housing. Socio-economic indexes for areas development: The resource dilution explanation. American
(Catalogue no. 2039.0). Canberra, Australia: Author. Psychologist, 56, 497–504.
Ellis Weismer, S., Murray Branch, J., & Miller, J. F. Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic
(1994). A prospective longitudinal study of language devel- regression. New York: Wiley Interscience.
opment in late talkers. Journal of Speech and Hearing Hresko, W. R., Kim Reid, D., & Hammill, D. D. (1999). Test
Research, 37, 852–867. of Early Language Development ( TELD-3). Austin, TX:
Entwisle, D. R., & Astone, N. M. (1994). Some practical Pro-Ed.
guidelines for measuring youths’ race/ethnicity and socio- Huttenlocher, J., Haight, W., Bryk, A., Seltzer, M., &
economic status. Child Development, 65, 1521–1540. Lyons, T. (1991). Early vocabulary growth: Relation to
Epstein, N. B., Bishop, D. S., Ryan, C., Miller, I. W., & language input and gender. Developmental Psychology, 27,
Keitner, G. (1993). The McMaster model view of healthy 236–248.
family functioning. In F. Walsh ( Ed.), Normal family Irwin, J. R., Carter, A. S., & Briggs-Gowan, M. J. (2002).
processes ( pp. 138–160). New York: Guilford. The social-emotional development of “late-talking” toddlers.
Feldman, H. M., Dale, P. S., Campbell, T. F., Colborn, Journal of the American Academy of Child & Adolescent
D. K., Kurs-Lasky, M., Rockette, H. E., et al. (2005). Psychiatry, 41, 1324–1332.
Concurrent and predictive validity of parent reports of child Klee, T., Carson, D., Gavin, W., Hall, L., Kent, A., & Reece,
language at ages 2 and 3 years. Child Development, 76, S. (1998). Concurrent and predictive validity of an early
856–868. language screening program. Journal of Speech, Language,
Feldman, H. M., Dollaghan, C. A., Campbell, T. F., Kurs- and Hearing Research, 41, 627–641.
Lasky, M., Janosky, J. E., & Paradise, J. L. (2000). Knobloch, H., Stevens, F., & Malone, A. F. (1980). Manual
Measurement properties of the MacArthur Communicative of developmental diagnosis: The administration and inter-
Development Inventories at ages one and two years. Child pretation of the Revised Gesell and Armatruda Develop-
Development, 71, 310–322. mental and Neurological Examination. New York: Harper
& Row.
Fenson, L., Dale, P. S., Reznick, S., Thal, D., Bates, E.,
Hartung, J., et al. (1993). MacArthur Communicative Kurinczuk, J. J., Parsons, D. E., Dawes, V., & Burton,
Development Inventories: User’s guide and technical P. R. (1999). The relationship between asthma and smoking
manual. San Diego, CA: Singular. during pregnancy. Women and Health, 29, 31– 47.
Fenson, L., Dale, P. S., Reznick, J. S., Bates, E., Thal, D., LaBenz, P. J., & LaBenz, E. S. (1980). Early correlates of
& Pethick, S. J. (1994). Variability in early communicative speech, language, and hearing. Littleton, MA: PSG Publish-
development. Monographs of the Society for Research in ing Company.
Child Development, 59, 1–173. Lovibond, P. F., & Lovibond, S. H. (1995a). The structure of
Garton, A. F., Zubrick, S. R., & Silburn, S. R. (1995). negative emotional states: Comparison of the Depression
The Western Australian child health survey: A pilot study. Anxiety Stress Scales (DASS) with the Beck Depression
Australian and New Zealand Journal of Psychiatry, 29, and Anxiety Inventories. Behavior Research and Therapy,
48–57. 33, 335–343.
Gee, V. (1996). Perinatal statistics in Western Australia: Lovibond, S. H., & Lovibond, P. F. (1995b). Manual for the
Thirteenth annual report of the Western Australian Mid- Depression Anxiety Stress Scales (2nd ed.). Sydney, Australia:
wives’ Notification System 1995. Perth, Australia: Health Psychology Foundation Monograph.
Department of Western Australia. Luinge, M. R., Post, W. J., Wit, H. P., & Goorhuis-Brouwer,
S. M. (2006). The ordering of milestones in language devel-
Hadley, P. A., & Holt, J. K. (2006). Individual differences
opment for children from 1 to 6 years of age. Journal of
in the onset of tense marking: A growth-curve analysis.
Speech, Language, and Hearing Research, 49, 923–940.
Journal of Speech, Language, and Hearing Research, 49,
984–1000. Lytton, H., Conway, D., & Sauve, R. (1977). The impact of
twinship on parent-child interaction. Journal of Personality
Hart, B., & Risley, T. (1995). Meaningful differences in the and Social Psychology, 35, 97–107.
everyday experience of young American children. Baltimore:
Paul H. Brookes. Marschik, P. B., Einspieler, C., Garzarolli, B., & Prechtl,
H. F. R. (in press). Events at early development: Are they
Hensley, V. R. (1988). Australian normative study of the associated with early word production and neurodevelopmen-
Achenbach Child Behaviour Checklist. Australian Psy- tal abilities at the preschool age? Early Human Development.
chologist, 23, 371–382.
McCardle, P., Cooper, J., & Freund, L. (2005). Language
Hoff-Ginsberg, E. (1994). Influences of mother and child on and genetics: Needs and opportunities. Applied Psycho-
maternal talkativeness. Discourse Processes, 18, 105–117. linguistics, 26, 129–135.
Hoff-Ginsberg, E. (1998). The relation of birth order and Miller, I. W., Epstein, N. B., Bishop, D. S., & Keitner, G. I.
socioeconomic status to children’s language experience (1985). The McMaster family assessment device: Reliability
and language development. Applied Psycholinguistics, 19, and validity. Journal of Marital and Family Therapy, 11,
603–629. 345–356.
Horwitz, S. M., Irwin, J. R., Briggs-Gowan, M. J., Olswang, L. B., Rodriguez, B., & Timler, G. (1998).
Heenan, J. M. B., Mendoza, J., & Carter, A. S. (2003). Recommending intervention for toddlers with specific lan-
Language delay in a community cohort of young children. guage learning difficulites: We may not have all the answers,
Journal of the American Academy of Child & Adolescent but we know a lot. American Journal of Speech-Language
Psychiatry, 42, 932–937. Pathology, 7, 23–32.

Pan, B. A., Rowe, M. L., Singer, J. D., & Snow, C. E. (2005). pathways to language: Methods, models, and research
Maternal correlates of growth in toddler vocabulary pro- directions. Mahwah, NJ: Erlbaum.
duction in low-income families. Child Development, 76, Rice, M. L. (2007). Children with specific language impair-
763–782. ment: Bridging the genetic and developmental perspectives.
Paul, R. (1991). Profiles of toddlers with slow expressive In E. Hoff & M. Shatz ( Eds.), Handbook of language
language development. Topics in Language Disorders, 11, development ( pp. 411–431). Cambridge, MA: Blackwell.
1–13. Rice, M. L., & Smolik, F. (2007). Genetics of language
Paul, R. (1996). Clinical implications of the natural history of disorders: Clinical conditions, phenotypes, and genes.
slow expressive language development. American Journal In G. Gaskell ( Ed.), Oxford handbook of psycholinguistics
of Speech-Language Pathology, 5, 5–21. ( pp. 685–700). Oxford, England: Oxford University Press.
Paul, R. (2000). Predicting outcomes of early expressive delay: Rice, M. L., Haney, K. R., & Wexler, K. (1998). Family
Ethical implications. In D. V. M. Bishop & L. B. Leonard histories of children with SLI who show extended optional
( Eds.), Speech and language impairments in children: Causes, infinitives. Journal of Speech, Language, and Hearing
characteristics, intervention and outcome ( pp. 195–209). Research, 41, 419–432.
Hove, East Sussex, England: Psychology Press. Rice, M. L., Redmond, S. M., & Hoffman, L. (2006). Mean
Paul, R., & Elwood, T. (1991). Maternal linguistic input length of utterance in children with specific language im-
to toddlers with slow expressive language development. pairment and in younger control children shows concurrent
Journal of Speech and Hearing Research, 34, 982–988. validity and stable and parallel growth trajectories. Journal
of Speech, Language, and Hearing Research, 49, 793–808.
Paul, R., & James, D. F. (1990). Language delay and parental
perceptions. Journal of the American Academy of Child & Rice, M. L., Tomblin, J. B., Hoffman, L., Richman, W. A.,
Adolescent Psychiatry, 29, 669–670. & Marquis, J. (2004). Grammatical tense deficits in
children with SLI and non-specific language impairment:
Paul, R., Looney, S. S., & Dahm, P. S. (1991). Communica- Relationships with nonverbal IQ over time. Journal of
tion and socialization skills at ages 2 and 3 in “late-talking” Speech, Language, and Hearing Research, 47, 816–834.
young children. Journal of Speech and Hearing Research
34, 858–865. Rice, M. L., Warren, S. F., & Betz, S. K. (2005). Language
symptoms of developmental language disorders: An over-
Plomin, R., Price, T. S., Eley, T. C., Dale, P. S., & view of autism, Down syndrome, fragile X, specific language
Stevenson, J. (2002). Associations between behaviour impairment, and Williams syndrome. Applied Psycho-
problems and verbal and nonverbal cognitive abilities and linguistics, 26, 7–28.
disabilities in early childhood. Journal of Child Psychology
and Psychiatry and Allied Disciplines, 43, 619–633. Rice, M. L., Spitz, R. V., & O’Brien, M. (1999). Semantic and
morphosyntactic language outcomes in biologically at-risk
Redmond, S. M., & Rice, M. L. (1998). The socioemotional children. Journal of Neurolinguistics, 12, 213–234.
behaviors of children with SLI: Social adaptation or social
deviance? Journal of Speech, Language, and Hearing Rice, M. L., Wexler, K., Marquis, J., & Hershberger, S.
Research, 41, 688–700. (2000). Acquisition of irregular past tense by children with
specific language impairment. Journal of Speech, Language,
Redmond, S. M., & Rice, M. L. (2002). Stability of behavioral and Hearing Research, 43, 1126–1145.
ratings of children with SLI. Journal of Speech, Language,
and Hearing Research, 45, 190–201. Rice, M. L., Wexler, K., & Hershberger, S. (1998). Tense
over time: The longitudinal course of tense acquisition in
Rescorla, L. (1989). The Language Development Survey: A children with specific language impairment. Journal of
screening tool for delayed language in toddlers. Journal Speech, Language, and Hearing Research, 41, 1412–1431.
of Speech and Hearing Disorders, 54, 587–599.
Roulstone, S., Loader, S., Northstone, K., Beveridge, M.,
Rescorla, L. (2002). Language and reading outcomes to age 9 & the ALSPAC team. (2002). The speech and language of
in late-talking toddlers. Journal of Speech, Language, and children aged 25 months: Descriptive data from the Avon
Hearing Research, 45, 360–371. Longitudinal Study of Parents and Children. Early
Rescorla, L., & Achenbach, T. M. (2002). Use of the Lan- Childhood Development and Care, 172, 259–268.
guage Development Survey (LDS) in a national probability Rubin, D. B. (1976). Inference and missing data. Biometrika,
sample of children 18 to 35 months old. Journal of Speech, 63, 581–592.
Language, and Hearing Research, 45, 733–743.
Rubin, D. B. (1987). Multiple imputation for nonresponse in
Rescorla, L., & Alley, A. (2001). Validation of the Language surveys. New York: Wiley.
Development Survey (LDS): A parent report tool for iden-
Rutter, M., Thorpe, K., Greenwood, R., Northstone, K., &
tifying language delay in toddlers. Journal of Speech,
Golding, J. (2003). Twins as a natural experiment to study
Language, and Hearing Research, 44, 434– 445.
the causes of mild language delay: I: Design, twin-singleton
Rescorla, L., Hadick-Wiley, M., & Escarce, E. (1993). differences in language, and obstetric risks. Journal of
Epidemiological investigation of expressive language delay Child Psychology and Psychiatry and Allied Disciplines,
at age two. First Language, 13, 5–22. 44, 326–341.
Rescorla, L., & Schwartz, E. (1990). Outcome of toddlers Samejima, F. (1969). Estimation of latent ability using a re-
with specific expressive language delay. Applied Psycho- sponse pattern of graded scores. Psychometrika Monograph
linguistics, 11, 393–407. Supplement, 34, 100–114.
Rice, M. L. (in press). How different is disordered language? SAS Institute. (2004). SAS/STAT 9.1 user’s guide. Cary, NC:
In J. Colombo, P. McCardle, & L. Freund ( Eds.), Infant Author.
Sawyer, M. G., Arney, F. M., Baghurst, P. A., Clark, J. J., study of late- and early-talking toddlers. Developmental
Graetz, B. W., Kosky, R. J., et al. (2000). Child and adoles- Neuropsychology, 13, 239–273.
cent component of the National Survey of Mental Health
Thal, D. J., Tobias, S., & Morrison, D. (1991). Language and
and Well Being: The mental health of young people in
gesture in late talkers: A one year follow-up. Journal of
Australia. Canberra, Australia: Mental Health and Special
Speech and Hearing Research, 34, 604–612.
Programs Branch, Commonwealth Department of Health
and Aged Care. Thorndike, R. L., Hagen, E. P., & Sattler, J. M. (1985).
Stanford-Binet Intelligence Scale (4th ed.). Riverside, CA:
Sawyer, M. G., Arney, F. M., Baghurst, P. A., Clarke, J. J.,
DLM Teaching Resources.
Graetz, B. W., Kosky, R. J., et al. (2001). The mental
health of young people in Australia: Key findings from the Tomblin, J. B. (1996). Genetic and environmental contribu-
child and adolescent component of the National Survey of tions to the risk for specific language impairment. In M. L.
Mental Health and Well-Being. Australian and New Rice ( Ed.), Towards a genetics of language ( pp. 191–210).
Zealand Journal of Psychiatry, 35, 806–814. Mahwah, NJ: Erlbaum.
Sawyer, M. G., Clark, J. J., & Baghurst, P. A. (1993). Child- Tomblin, J. B., Smith, E., & Zhang, X. (1997). Epidemi-
hood emotional and behavioural problems: A comparison of ology of specific language impairment: Prenatal and peri-
children’s reports with reports from parents and teachers. natal risk factors. Journal of Communication Disorders, 30,
Journal of Paediatrics and Child Health, 29, 119–125. 325–344.
Sawyer, M. G., Kosky, R. J., Graetz, B. W., Arney, F., Verhulst, F. C., Achenbach, T. M., van der Ende, J.,
Zubrick, S. R., & Baghurst, P. (2000). The National Erol, N., Lambert, M. C., Leung, P. W. L., et al. (2003).
Survey of Mental Health and Well Being: The child and Comparisons of problems reported by youths from seven
adolescent component. Australian and New Zealand countries. American Journal of Psychiatry, 160, 1479–1485.
Journal of Psychiatry, 34, 214 –220.
Wells, G. (1985). Language development in the pre-school
Sawyer, M. G., Sarris, A., Baghurst, P. A., Cornish, C. A., years. Cambridge, England: Cambridge University Press.
& Kalucy, R. S. (1990). The prevalence of emotional and
behavioural disorders and patterns of service utilisation in Whitehurst, G. J., Arnold, D. S., Smith, M., Fischel, J.,
children and adolescents. Australian and New Zealand Lonigan, C. J., & Valdez-Menchaca, M. C. (1991). Family
Journal of Psychiatry, 24, 323–330. history in developmental expressive language delay.
Journal of Speech and Hearing Research, 34, 1150–1157.
Silburn, S. R., Zubrick, S. R., Garton, A., Gurrin, L.,
Burton, P., Dalby, R., et al. (1996). Western Australian Whitehurst, G. J., & Fischel, J. (1994). Early developmental
Child Health Survey: Family and community health. Perth, language delay: What, if anything should the clinician do
Western Australia: Australian Bureau of Statistics and the about it? Journal of Child Psychology and Psychiatry and
TVW Telethon Institute for Child Health Research. Allied Disciplines, 35, 613–648.
Spinath, F., Price, T. S., Dale, P. S., & Plomin, R. (2004). Whitehurst, G. J., Fischel, J., Arnold, D. S., & Lonigan,
The genetic and environmental origins of language disabil- C. J. (1992). Evaluating outcomes with children with ex-
ity and ability. Child Development, 75, 445–454. pressive language delay. In S. F. Warren & J. Reichle ( Eds.),
Causes and effects in communication and language inter-
Squires, J., & Bricker, D. (1993). Infant/Child Monitoring vention ( Vol. 1, pp. 227–313). Baltimore: Paul H. Brookes.
Questionnaires procedures manual (Revised ed.). Eugene,
OR: University of Oregon. Whitehurst, G. J., Fischel, J. E., Lonigan, C. J., Valdez-
Menchaca, M. C., DeBaryshe, B. D., & Caulfield,
Squires, J., Bricker, D., & Potter, L. (1997). Revision of a
M. B. (1988). Verbal interaction in families of normal and
parent-completed developmental screening tool: Ages and
expressive-language-delayed children. Developmental Psy-
Stages Questionnaires. Journal of Pediatric Psychology,
chology, 24, 690–699.
22, 313–328.
Windle, M. (1992). Temperament and social support in
Squires, J., Potter, L., & Bricker, D. (1999). The ASQ user’s
adolescence: Interrelations with depressive symptoms and
guide (2nd ed.). Baltimore: Paul H. Brookes.
delinquent behavior. Journal of Youth and Adolescence,
Stanley, F. J., Read, A. W., Kurinczuk, J. J., Croft, M. L., & 21, 1–21.
Bower, C. (1997). A population maternal and child health
research database for research and policy evaluation in Windle, M., & Lerner, R. M. (1986). Reassessing the dimen-
Western Australia. Seminars in Neonatology, 2, 195–201. sions of temperamental individuality across the lifespan:
The Revised Dimensions of Temperament Survey ( DOTS-R).
Stout, W. (1987). A nonparametric approach for assessing Journal of Adolescent Research, 1, 213–230.
latent trait unidimensionality. Psychometrika, 52, 589–617.
Zubrick, S. R., Kurinczuk, J. J., McDermott, B. M.,
Tager-Flusberg, H., & Cooper, J. (1999). Present and future McKelvey, R. S., Silburn, S. R., & Davies, L. C. (2000).
possibilities for defining a phenotype for specific language Fetal growth and subsequent mental health problems in
impairment. Journal of Speech, Language, and Hearing children aged 4 to 13 years. Developmental Medicine and
Research, 42, 1275–1278. Child Neurology, 42, 14–20.
Thal, D. J., & Katich, J. (1996). Predicaments in early Zubrick, S. R., Silburn, S. R., Garton, A., Burton, P.,
identification of specific language impairment. In K. N. Cole, Dalby, R., Carlton, J., et al. (1995). Western Australian
P. S. Dale, & D. J. Thal ( Eds.), Assessment of communication Child Health Survey: Developing health and well-being
and language. Baltimore: Paul H. Brookes. in the nineties. Perth, Western Australia: Australian
Thal, D. J., Bates, E., Goodman, J., & Jahn-Samilo, J. Bureau of Statistics and the Institute for Child Health
(1997). Continuity of language abilities: An exploratory Research.

Zubrick, S. R., Silburn, S. R., Gurrin, L., Teoh, H. J.,
Received August 29, 2006
Shepard, C., Carlton, J., et al. (1997). Western Australian
Child Health Survey: Education, Health and Competence. Revision received January 16, 2007
Perth, Western Australia: Australian Bureau of Statistics Accepted March 2, 2007
and the TVW Telethon Institute for Child Health Research.
DOI: 10.1044/1092-4388(2007/106)
Zubrick, S. R., Ward, K. A., Silburn, S. R., Lawrence,
D. M., Williams, A. A., Blair, E., et al. (2005). Prevention of Contact author: Mabel Rice, Dole Center, 1000 Sunnyside
child behaviour problems via universal implementation of a Avenue, Room 3031, University of Kansas, Lawrence,
group behavioural family intervention. Prevention Science, KS 66045-7555. E-mail: mabel@ku.edu.
6, 287–304.

Late Language Emergence at 24 Months: An Epidemiological Study of Prevalence, Predictors, and Covariates

Uploaded by

Copyright:

Available Formats

Late Language Emergence at 24 Months: An Epidemiological Study of Prevalence, Predictors, and Covariates

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Late Language Emergence at 24 Months: An Epidemiological Study of Prevalence, Predictors, and Covariates

Uploaded by

Copyright:

Available Formats

Late Language Emergence at

Zubrick et al.: Late Language Emergence 1563

Zubrick et al.: Late Language Emergence 1565

Figure 1. The study sample.

Zubrick et al.: Late Language Emergence 1567

Zubrick et al.: Late Language Emergence 1569

Zubrick et al.: Late Language Emergence 1571

(Continued on the following page)

(Continued on the following page)

Zubrick et al.: Late Language Emergence 1573

Zubrick et al.: Late Language Emergence 1575

Zubrick et al.: Late Language Emergence 1577

Normal (N = 1,528) Late (N = 238)

Variable M (SD) % M (SD) % c2 (t )b

(Continued on the following page)

Normal (N = 1,528) Late (N = 238)

Variable M (SD) % M (SD) % c2 (t )b

Age (years) 2.11 (0.135) 2.08 (0.104) (3.64)***

(Continued on the following page)

Zubrick et al.: Late Language Emergence 1579

Normal (N = 1,528) Late (N = 238)

Variable M (SD) % M (SD) % c2 (t )b

Daycare status at interview

Zubrick et al.: Late Language Emergence 1581

Intercept 3.7298 2.0435 –0.2755 7.7351 26164 1.83 0.068

(Continued on the following page)

Socioeconomic indicators for area (SEIFA)

Normal ref 1.000

(Continued on the following page)

(Continued on the following page)

Zubrick et al.: Late Language Emergence 1587

Zubrick et al.: Late Language Emergence 1589

Zubrick et al.: Late Language Emergence 1591

You might also like