You are on page 1of 15

Behavior Predictors of Language

Development Over 2 Years


in Children With Autism
Spectrum Disorders

Karen D. Bopp
Pat Mirenda
Purpose: This exploratory study examined predictive relationships between 5 types of
Bruno D. Zumbo
behaviors and the trajectories of vocabulary and language development in young
The University of British Columbia,
children with autism over 2 years.
Vancouver, Canada
Method: Participants were 69 children with autism assessed using standardized
measures prior to the initiation of early intervention (T1) and 6 months (T2), 12 months
(T3), and 24 months (T4) later. Growth curve modeling examined the extent to which
behaviors at T1 and changes in behaviors between T1 and T2 predicted changes
in development from T1 to T4.
Results: Regardless of T1 nonverbal IQ and autism severity, high scores for inattentive
behaviors at T1 predicted lower rates of change in vocabulary production and language
comprehension over 2 years. High scores for social unresponsiveness at T1 predicted
lower rates of change in vocabulary comprehension and production and in language
comprehension over 2 years. Scores for insistence on sameness behaviors, repetitive
stereotypic motor behaviors, and acting-out behaviors at T1 did not predict the rate
of change of any child measure over 2 years beyond differences accounted for by
T1 autism severity and nonverbal IQ status.
Conclusions: The results are discussed with regard to their implications for early
intervention and understanding the complex factors that affect developmental outcomes.
KEY WORDS: autism, child development, problem behavior, predictors

R
esearch suggests that a number of variables are related to differ-
ences in the development of children with autism spectrum disorders
(ASD). For example, for children who receive early intervention, initial
IQ score (Gabriels, Hill, Pierce, & Rogers, 2001; Harris & Handleman, 2000;
Liss et al., 2001; Lovaas & Smith, 1998; Sallows & Graupner, 2005; T. Smith,
Eikeseth, Klevstrand, & Lovaas, 1997), chronological age at the onset of
intervention (Fenske, Zalenski, Krantz, & McClannahan, 1985; Harris &
Handleman, 2000; Lovaas, 1987), and autism severity score (DeMyer, 1973;
Eaves & Ho, 1996; Liss et al., 2001) have all been found to be related to
cognitive, language, and adaptive behavior outcomes over time. However,
even with early intervention, it is estimated that only approximately 15%
of individuals with autism are reasonably self-sufficient as adults and
another 15% to 20% function well with periodic support (Volkmar & Pauls,
2003). In addition, in a review of treatments for young children with ASD,
Schreibman (2000) noted a wide heterogeneity in the outcomes of inter-
vention and remarked that there is no “one size fits all” treatment for this
population. The lack of 100% effectiveness of early intervention leads one to

1106 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009 • D American Speech-Language-Hearing Association
1092-4388/09/5205-1106
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016
Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
conclude that there are other variables affecting outcomes at risk for isolation within or exclusion from typical home,
that have not yet been identified (Ingersoll, Schreibman, school, and community settings and thus may put him/
& Stahmer, 2001). Identifying these variables and un- her at risk for delayed or even disordered language devel-
derstanding the process of language development in opment. However, although these behaviors are often ob-
autism and the factors that influence differential out- served in clinical settings, almost no studies to date have
comes are critical for improving treatment efficiency. examined them as sources of variability in language
What are some of the factors that may influence dif- development. This study attempted to explore the rela-
ferential language outcomes in ASD? Performance-based tionship between these five behaviors and language
theory suggests that language disorders are secondary to development.
limitations in individual child factors such as cognitive
processing (e.g., motivation, attention, memory, and re-
tention) and social–affective skills (e.g., imitation, joint
Inattentiveness
attention, reciprocity, etc.; Evans, 2001). According to a Inattentiveness refers to behaviors that indicate high
social–interactionist perspective, language is developed distractibility (e.g., difficulty staying focused on relevant
through a motivation to interact socially with others; so- people or activities) and/or decreased awareness of ob-
cial experiences, in turn, motivate the acquisition of many jects, activities, or the environment. There is no direct
dimensions of language (Abbeduto & Boudreau, 2004). In empirical evidence of a relationship between inatten-
fact, a number of specific social–affective behaviors have tiveness and language development in autism; however,
been found to predict language development in this popu- there is some indirect evidence. Recent research has found
lation. These behaviors include, for example, motor imita- that the inability to both initiate and respond to joint at-
tion skills (Sallows & Graupner, 2005; Stone & Yoder, tention is closely related to language difficulties in this
2001; Toth, Munson, Meltzoff, & Dawson, 2006), verbal population (e.g., Charman et al., 2003; Rollins & Snow,
imitation skills (Sallows & Graupner, 2005; V. Smith, 1998; Sigman & McGovern, 2005; V. Smith et al., 2007;
Mirenda, & Zaidman-Zait, 2007), the ability to initiate Travis et al., 2001). Three basic components—attention,
and /or respond to joint attention (e.g., Charman et al., shared affect, and shared intentions—constitute the ca-
2003; Rollins & Snow, 1998; Sigman & McGovern, 2005; pacity for joint attention (Yoder & McDuffie, 2006).
V. Smith et al., 2007; Travis, Sigman, & Ruskin, 2001), Hypothetically, children who lack the first component,
and pretend play skills (e.g., Sigman & McGovern, 2005; attention, may be less likely to engage in joint attention
V. Smith et al., 2007; Toth et al., 2006; Yoder, 2006). interactions and this, in turn, may result in delayed lan-
Social–affective behaviors may not be the only fac- guage development over time (Mundy & Markus, 1997).
tors that influence language development in children
with autism. One area that has received little attention
to date is child “problem” behaviors and their potential Socially Unresponsive Behavior
impact. Of course, the extent to which any given behavior Socially unresponsive behavior refers to a decreased
can be deemed problematic depends both on one’s per- capacity to either initiate or respond to social and /or
spective and on the context in which the behavior occurs. emotional exchanges by other people and is one of the
The National Research Council (2001) described the com- hallmarks of ASD. Such behaviors include a failure to
plexity of this issue as follows: smile in response to others, avoidance of eye contact, and
failure to respond to verbal or physical overtures (e.g.,
From a child’s perspective, problem behaviors in-
one’s name being called or pointing by others). In their
clude the inability to understand demands of a class-
early description of three subgroups of children with au-
room or a parent and to communicate his or her needs
tism, Wing and Gould (1979) noted that the children who
and wants, severe difficulty in initiating and main-
were most socially impaired (i.e., those in their “socially
taining social interactions and relationships, confu-
aloof ” group) were also the most language impaired, es-
sion about the effects and consequences of any of his
pecially with regard to comprehension. More recently,
or her behaviors, and engagement in restrictive and
Mundy and his colleagues proposed a “social orienting
repetitive behaviors and interests that may limit the
model” to explain the relationship between social and lan-
child’s ability to learn and to fit in with peers. From a
guage development in autism (Mundy & Markus, 1997;
teacher’s or parent’s perspective, problem behaviors
Mundy & Neal, 2001). They hypothesized that early im-
include lack of compliance with or disruption of class-
pairments in social attention deprive the child with
room routines, tantrums, destruction of property,
autism of social information input during infancy and
and aggression against self or others. ( pp. 115–116)
preschool development and that this deprivation dis-
Problem behaviors such as acting-out, repetitive ste- rupts normal brain and behavioral development. This
reotypic motor behaviors, insistence on sameness, social cycle then acts like a negative feedback loop, affecting
unresponsiveness, and inattentiveness may place a child subsequent socio-communicative development. Dawson

Bopp et al.: Behavior Predictors 1107

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
et al. (2004) provided evidence to support this hypothesis Behavior Scales (VABS; Sparrow, Balla, & Cicchetti, 1984).
in a recent study in which structural equation modeling However, these same children also had more atypical,
was used to demonstrate that social orienting was di- stereotyped, repetitive, and idiosyncratic speech as mea-
rectly related to joint attention ability and indirectly re- sured on the ADI–R. With regard to RSMBs, associa-
lated to language ability in young children with autism. tions with a variety of developmental variables have been
Indeed, a recent study aimed at identifying a multidi- reported. Both lower adaptive behavior scores (Cuccaro
mensional model for the autism phenotype suggested et al., 2003; Szatmari et al., 2006; Venter, Lord, & Schopler,
that socio–communication skills—which are currently 1992) and lower IQ scores (Bishop, Richler, & Lord, 2006;
separated into separate social and communication do- Campbell et al., 1990; Szatmari et al., 2006) appear to be
mains in the DSM–IV–TR—should be combined to consti- related to more atypical RSMB scores. Similarly, lower-
tute a single domain (American Psychiatric Association, level language skills (Carcani-Rathwell, Rabe-Hasketh,
2000; Georgiades et al., 2007). Together, these findings & Santosh, 2006; Dadds, Schwartz, Adams, & Rose, 1988)
provide support for the hypothesized relationship be- and lower academic achievement outcomes, as measured
tween behaviors that reflect social unresponsiveness and by either school placement or standardized measures,
subsequent communication and language development. have been associated with more atypical RSMBs (Epstein,
Taubman, & Lovaas, 1985; Venter et al., 1992).

Restricted and Repetitive Behaviors


Restricted and repetitive behaviors (RRBs) include
Acting-Out Behavior
both repetitive stereotypic motor behaviors (RSMBs) such Acting-out behaviors are those that cause harm or
as hand flapping, rocking, and humming and insistence damage to the child him/herself, to another person, and/or
on sameness (IS) behaviors such as lining up objects or to the environment (e.g., tantrums, aggressive behaviors,
insisting that routines always occur in the same way and property destruction). They also include noncompliant
(American Psychiatric Association, 2000). By definition, behaviors (i.e., refusing to follow directions, “stubborn-
an ASD diagnosis requires the presence of RRBs; how- ness”) and other behaviors that indicate emotional dys-
ever, relatively few studies have examined the relation- regulation or distress (e.g., screaming, crying, whining).
ship between this broad category of behaviors and other Acting-out behaviors have often been linked to commu-
developmental variables. Bodfish, Symons, Parker, and nication, cognitive, and social skills. This may be because
Lewis (2000) examined the relationship between autism children’s inability to regulate their behavior during cog-
severity and rate of occurrence of stereotypy, self-injury, nitively demanding tasks or situations places them at a
compulsions, dyskinesia (i.e., repetitive, involuntary move- learning disadvantage over children who are able to do
ments), and akathisia (i.e., repetitive, restless movements so (Bronson, 2000). In addition, children with acting-out
such as pacing) in 32 adults with autism. They found that behaviors may be rejected by their peers and, as a result,
adults with more severe RRBs had more severe autism. may be deprived of opportunities to practice and develop
Similarly, Gabriels, Cuccaro, Hill, Ivers, and Goldson prosocial skills such as sharing and cooperating (Kaiser
(2005) examined the relationship between a number of & Rasminsky, 2003). Acting-out behaviors have also been
RRBs (i.e., repetitive motor phenomena, narrow or cir- viewed as communicative, as they can function to enable
cumscribed interests, and self-injurious behaviors) and an individual’s access to desired items, activities, or inter-
cognitive and adaptive communication skills in 14 chil- actions (“I want ____”) or escape from undesired tasks or
dren with autism (mean age: 10;7 [years;months]). They interactions (“I don’t want ___”; Bopp, Brown, & Mirenda,
found that children who had more RRBs had significantly 2004). Despite these potential links, no research has ex-
lower nonverbal cognitive and communicative abilities. amined the relationship between acting-out behavior and
These studies provide some evidence for a general rela- language development over time in children with ASD.
tionship between RRBs and other variables.
Recently, a number of researchers have examined
RSMBs and IS behaviors as distinct subclasses of be-
Summary
havior that appear to be related to different aspects of The literature examining child problem behaviors
development in individuals with ASD (e.g., Richler, Bishop, as predictors of language development in ASD is still in
Klienke, & Lord, 2007). With regard to IS behaviors, its infancy. Very few investigations have examined the
Szatmari et al. (2006) found that, in 339 children with relationships between these behaviors and development
pervasive developmental disorder (mean age: 8;5), those in children with autism under the age of 6. Finally, with
who had more atypical IS behaviors on the Autism Diag- some exceptions (e.g., Dissanayake, Sigman, & Kasari,
nostic Interview—Revised (ADI–R; Rutter, Le Couteur, 1996; Epstein et al., 1985), the majority of studies con-
& Lord, 2003) had better language abilities as measured ducted in this area have been descriptive or have only
by the communication subscale of the Vineland Adaptive examined relationships at one point in time, providing a

1108 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
restricted view of the relationships between child prob- Participants were 58 males and 11 females (mean
lem behaviors and other developmental domains. Thus, age at T1: 4;2; range = 1;9–6;0). Their mean CARS score at
the purpose of this exploratory study was to examine the T1 was 36.3 (range = 25.0–50.5). Scores below 30 (n = 12)
extent to which problem behaviors predicted changes were considered to be indicative of PDD-NOS, based on
in the developmental trajectories of vocabulary and lan- the results of a 2005 study by Perry, Condillac, Freeman,
guage skills in young children with ASD over 2 years. Dunn-Geier, and Belair, who examined the concordance
Specifically, we were interested in whether the presence rate between the CARS and DSM–IV clinical diagnoses
of problem behaviors prior to the onset of intervention in 274 preschool children (aged 2 to 6). The children’s eth-
and /or changes in problem behavior during the first nic backgrounds included European-Canadian (n = 37),
6 months of intervention were related to changes in child Asian (n = 21), Hispanic (n = 4), Caribbean/African (n = 2),
development over 2 years. Identifying these relation- Middle Eastern (n = 1), and multiple ethnicities (n = 4). At
ships is important for understanding the process of lan- baseline, mothers had completed some college or uni-
guage development in ASD and the factors that influence versity courses (on average) and were considered semi-
the differential outcomes reported in the literature. In the skilled workers (e.g., machine operator, grocery store
end, such information may also help to improve treat- clerk; Hollingshead, 1962). Fathers had completed some
ment effectiveness by allowing therapists to set priorities university training (on average) and were considered
based on empirical information rather than on simple skilled workers (e.g., department manager, administra-
heuristics. The specific questions addressed in this study tive assistant; Hollingshead, 1962). The families included
were: 55 two-parent families, three separated families, four di-
1. Do scores in one or more types of child problem be- vorced families, six other arrangements (e.g., common-
havior at the onset of intervention predict changes law), and one single-parent family.
in the developmental trajectories of vocabulary and All of the children received approximately 15 to
language skills in young children with ASD over 20 hours per week of year-round early intervention ser-
2 years? vices for 2 years. The interventions for all children were
2. Do changes in scores related to one or more types of eclectic and were administered from either one of three
child problem behavior over a 6-month period predict Early Intensive Behavior Intervention sites throughout
changes in the developmental trajectories of vocabu- the province of BC (n = 39) or from private behavior con-
lary and language skills in young children with ASD sultants and other professionals who were hired by the
over 2 years? family (n = 30). In general, intervention consisted of 1:1
instruction based on the principles of applied behavior
analysis as well as services from speech-language pathol-
ogists, occupational therapists, and other professionals.
Method The majority of children (87%) also attended preschool
Participants or school while they received intervention services. The
degree to which the five specific problem behaviors ex-
Data for this study were accessed from a database amined in this study were addressed during intervention
established for a project examining early intervention varied widely, depending on individual child and family
outcomes for 69 children with ASD and their parents in needs. Families reported that, in the first 6 months of in-
British Columbia (BC), Canada. Prior to entry into the tervention, only 2 children (3%) took prescription medica-
study, 55 children were diagnosed with autism and 14 tions that were related to problem behavior; 4 (6%) were
were diagnosed with pervasive developmental disorder, on a dairy-free diet; 4 (6%) were on a supplemental diet;
not otherwise specified (PDD-NOS). All diagnoses were and 7 (10%) were on another type of special diet. Because
made by experienced community-based clinicians who there were no differences between the two service deliv-
were not involved in the study. Specifically, 28 children ery groups with regard to language outcomes over 2 years
(41%) received a diagnosis from an autism team that in- (Mirenda, Bopp, Kavanagh, Smith, & Zaidman-Zait,
cluded, at a minimum, a psychologist or psychiatrist, a 2005), the two groups were combined for this analysis.
pediatrician, and a speech-language pathologist; 22 (32%)
received a diagnosis from a pediatrician; 12 (17%) re-
ceived a diagnosis from a psychiatrist; and 7 (10%) re-
Measurement
ceived a diagnosis from a registered psychologist or Data were collected at baseline (i.e., prior to the ini-
another qualified professional. In addition, the Child- tiation of early intervention, T1) and 6, 12, and 24 months
hood Autism Rating Scale (CARS; Schopler, Reichler, & later (T2–T4). Data collection occurred in each child’s
Renner, 1988) was administered at the outset of the home or early intervention center and was administered
study by psychologists who were naive to the original by an assessment team that included registered psychol-
diagnosis. ogists, certified speech-language pathologists, and trained

Bopp et al.: Behavior Predictors 1109

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
graduate students who acted as family interviewers. At Measures of Child Problem Behavior
each time point, the data collection location was consis-
tent for each child and data collection was typically con- The database did not include specific measures of
ducted by the same assessment team members. None of the five categories of problem behaviors that were uti-
the assessors were involved in service provision, either lized as independent (i.e., predictor) variables; thus, scores
to the children or to their families. for these variables were constructed in a seven-step
process.
Measures of Child Development Identify relevant predictor variables. The first step was
to conduct a thorough literature review to identify rele-
The dependent variables were each child’s raw scores vant variables that were potential predictors of language/
on measures of vocabulary and language at all four time communication abilities in young children with autism
points (T1–T4). All vocabulary and language measures and related disorders. The five problem behavior vari-
were administered by a certified speech-language pathol- ables that were selected from this review were all hypoth-
ogist. The independent variables, problem behaviors, esized to impact social–affective development and, as a
were not taken from standardized measures and were result, language development.
constructed in a process described below.
Identify relevant item indicators. The second step
Vocabulary skills. Single word receptive vocabulary was to identify individual item indicators related to the
skills were measured with the Peabody Picture Vocabu-
behavior variables, using existing assessment measures
lary Test—IIIA/B (PPVT–III; L. M. Dunn & Dunn, 1997),
in the data set. This process was completed by the first
which provides raw and standard scores for total words
and second authors’ independent examination of every
understood. The PPVT has been shown to be a reliable and
item in five measures that were collected at each assess-
valid measure of receptive vocabulary for persons from
ment time point but were not used as measures of child
2;6 (years;months) to over 90 years, with standard and
raw scores well correlated to chronological age and intel- language or vocabulary development. The five measures
lectual functioning. Internal reliability coefficients of raw were all parent-report measures and included the Au-
scores on the PPVT are reported to be a median alpha of tism Behavior Checklist (ABC; Krug, Arick, & Almond,
0.95 (range: 0.89–0.97; L. M. Dunn & Dunn, 1997). 1980/1983), the Vineland Adaptive Behavior Scales (VABS;
Sparrow et al., 1984), the Temperament and Atypical
Single word expressive vocabulary skills were as-
Behavior Scale (TABS; Bagnato, Neisworth, Salvia, &
sessed using the Expressive One-Word Picture Vocabu-
Hunt, 1999), the Sensory Profile (W. Dunn, 1999), and
lary Test (EOWPVT; Brownell, 2000), which provides
raw and standard scores of total words expressed. The the Parenting Stress Index—Short Form (PSI–SF; Abidin,
EOWPVT has been normed on individuals without dis- 1995). The interrater agreement for identification of po-
abilities, ages 2–18;11 and is a reliable and valid measure tential items related to the five behavior variables was
of expressive vocabulary. Standard and raw scores are 90%.
highly correlated with chronological age and measures of The ABC is a 57-item checklist of behaviors related
cognitive ability, language, academic achievement, and to autism and was completed by the parent or primary
receptive vocabulary. Internal reliability coefficients of caregiver. The VABS Interview Edition Survey Form is
scores on the EOWPVT are reported to be a median alpha a measure of social, daily living, communication, and mo-
of 0.96 (range: 0.93–0.98; Brownell, 2000). tor skills and was administered by a psychologist through
Language skills. Language skills were measured with a semistructured interview with the parent. The TABS
the Preschool Language Scale—3 (PLS–3; Zimmerman, consists of 55 questions regarding a child’s emotional dis-
Steiner, & Pond, 1992). This instrument measures both position and regulatory behavior and was completed by
receptive and expressive language abilities and provides the parent or primary caregiver. The Sensory Profile was
raw and standard scores for total language and for two also completed by the parent or primary caregiver and
subscales, Auditory Comprehension (PLS AC) and Ex- consists of 125 questions that report the frequency with
pressive Communication (PLS EC). The PLS–3 has been which a child responds to various sensory experiences. It
found to be a reliable and valid measure of auditory com- is used for children between the ages of 3 to 10 and pro-
prehension, language expression, and overall language. vides a standardized method of reporting children’s sen-
Standard and raw scores are highly correlated with chro- sory and processing abilities. Finally, the PSI-SF is a
nological age and other measures of language. For chil- parent/caregiver response form that consists of 36 key
dren ages birth to 6 years, 11 months, internal reliability items derived from the long form of the PSI and assesses
coefficients range from 0.47 to 0.88 for Auditory Compre- the level of parenting stress.
hension, 0.69 to 0.90 for Expressive Communication, and Choose item indicators. The third step was to choose
0.74 to 0.94 for overall Total Language Score (Zimmerman specific items for each problem behavior variable using
et al., 1992). an expert jury to address content validity (Sireci, 1998).

1110 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
The jury consisted of 15 masters- or doctoral-level the other occurred in the acting-out variable, where two
professionals whose academic focus was autism and who items, one from the Sensory Profile and another from the
had an average of 11 years (SD = 8.45) of experience ABC, both referred to temper tantrums and thus consti-
working directly with young children with ASD. Each tuted a subdomain. A decision was made to choose items
jury member was provided with a list of all of the poten- originally coded as dichotomous (i.e., those from the ABC
tial behavior items in random order. They were asked to and TABS) over items originally coded on a Likert-type
read definitions of the five categories of problem behav- scale (i.e., items from the PSI-SF, Sensory Profile, and
ior and to then sort the items from the list into the cate- VABS). On this basis, the temper tantrum item from the
gories. They were instructed that each problem behavior ABC was retained over the item from the Sensory Pro-
item could fall under only one behavior category, and file. The ABC item was also selected for eye contact to
they were provided with an “other” category for items that keep the origin of items as consistent as possible across
did not appear to fit any of them. Individual items that behavior variables.
were selected for a category by at least 80% of jury mem- Calculate coefficient alphas. The sixth step was to
bers were retained for use in the study. examine the psychometric properties of the items within
Convert items to dichotomous responses. The fourth each predictor variable to compute a composite score
step was to convert the individual item indicators into (Zumbo, Gelin, & Hubley, 2003). Coefficient alpha is used
dichotomous item responses (i.e., yes/no scores), as needed. to gauge the reliability of a measure and is based on the
In cases where ordinal or Likert-type scales were orig- average correlation among items (Cronbach, 2004). Typ-
inally used to score an item, the following steps were ically, a coefficient alpha score of ≥.70 indicates that
taken. For the VABS, “yes, usually” and “sometimes or items used to calculate a composite score are internally
partially” were both coded as “yes” (1), and “no” was consistent with little measurement error (Streiner &
coded as “no” (0); for the Sensory Profile, “always,” “fre- Norman, 1989). However, lower alpha values of .60 have
quently,” “occasionally,” and “seldom” were coded as “yes” been accepted as adequate (e.g., Clark & Watson, 1995;
(1), and “never” was coded as “no” (0); for the PSI-SF, Milfont & Gouveia, 2006), especially in exploratory
“strongly agree” and “agree” were coded as “yes” (1), and research.
“strongly disagree” or “disagree” responses were coded Calculate composite scores. In the final step in the
as “no” (0), whereas “not sure” was not coded. process, the sum of items was calculated (i.e., 1 = the be-
Eliminate item overrepresentation. The fifth step havior was present and 0 = the behavior was not pres-
was to eliminate overrepresentation of any single be- ent) to create a composite score for each of the predictor
havior within a predictor variable. The question for this variables at T1 (Zumbo et al., 2003). A summary of the
step was “Which item is most representative of a spe- items that were included in each behavior variable and
cific behavior of concern?” Answering this question in- their interrater reliability percentages based on the ex-
volved several phases. First, items related to the same pert jury judgments (Step 3) are available from the first
behavior were grouped into subdomains on the basis of author on request.
similar wording. For example, the three separate items
from the ABC, the Sensory Profile, and the TABS that
refer to avoidance or resistance of eye contact were Data Analysis
grouped into one subdomain within the variable “social
Following construction of the predictor variables,
unresponsiveness.”
individual growth curve modeling (IGCM) using SAS
Second, a series of decision rules was applied to elim- Proc Mixed computer software was used to answer each
inate redundant items within each subdomain. Multiple research question. Individual growth trajectories devel-
items from the same measure were retained even when oped through latent trajectory analysis are a rich and
they referred to similar behaviors. For example, Sensory flexible alternative to traditional methods for analyzing
Profile Items 27 and 28 both refer to rocking behaviors; longitudinal data (e.g., regression). IGCM was used in
however, both were retained because they were origi- order to take advantage of the multiwave data that were
nally considered to be separate items on this measure. available for investigation. IGCM enables the examination
This rule was applied to improve content validity such of change as a continuous process, such that the amount
that the constructed variables included sets of items that of change between time periods for a given participant is
sampled several similar, but slightly different, aspects of a result of that participant’s underlying growth trajec-
a specific behavior. tory (Francis, Fletcher, Stuebing, Davidson, & Thompson,
In only two instances did items with similar wording 1991). In addition, IGCM includes path analysis and re-
within a subdomain originate from different measures. gression to explore relationships among changes in in-
One of these occurred for the eye contact behavior in the dividual variables over time (Keith, 1993; Schumacker &
social unresponsiveness variable, as described previously; Lomax, 2004).

Bopp et al.: Behavior Predictors 1111

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Data analysis using SAS Proc Mixed was completed & Norman, 1989), and lower alpha values of .60 indicated
using a four-step process. In all four steps and in all adequate internal consistency (e.g., Clark & Watson, 1995;
models, an unstructured correlation structure was used Milfont & Gouveia, 2006).
(Singer, 1998); age was used as the measure of time to Composite scores were calculated for each category
take into account the wide age range at T1, and age was of problem behavior. Table 2 summarizes the mean, range,
centered (age minus the mean age at T1) to facilitate and standard deviation for each of the five problem be-
interpretation of the output (Singer and Willet, 2003). havior variables at T1 and T2 for all 69 children.
In Step 1, unconditional models were run separately for
each language measure (i.e., PPVT, EOWPVT, PLS AC,
Child Language Development (T1–T4)
and PLS EC). During this step, the shape of the growth
curve and the variance in the intercept and slope for the Table 3 summarizes the raw score, mean, range, and
group overall and between individuals was examined. standard deviation for each of the measures of vocabu-
In Step 2, separate conditional models were run for each lary and language development and autism severity from
language measure to determine the influence of T1 au- T1 to T4 for all 69 children.
tism severity and of T1 nonverbal IQ (NVIQ) scores on
each trajectory over 2 years. This was done because past Unconditional Model
research has found a relationship between autism se-
Before analyses were conducted to examine the pre-
verity scores and NVIQ (DeMyer, 1973; Eaves & Ho, 1996;
dictors of change in child development, unconditional
Liss et al., 2001) and between NVIQ and other develop-
models were examined to determine whether there was
mental outcomes (e.g., Luyster, Qui, Lopez, & Lord, 2007).
significant variance in the initial value and slope for each
Autism severity was measured by CARS total scores.
of the child measures over 2 years and to examine whether
NVIQ was missing for 9 of the children (13%) and was
the model was linear or quadratic. Table 4 presents these
estimated for the remaining 60 children (87%) by com-
findings and the Akaike Information Criterion (AIC), the
bining the t-scores from the Visual Reception and Fine
unconditional model goodness-of-fit statistic, for each mea-
Motor subscales of the Mullen Scales of Early Learning
sure. There is no statistical test available to assess the
(Mullen, 1995; see Luyster et al., 2007). In Step 3, con-
differences in AIC between models; however, models
ditional models were run for each measure that included,
with lower AIC values are preferred and are considered
when warranted, T1 CARS total scores and/or T1 NVIQ
a better fit (Tabachnick & Fidell, 2007). Overall, the re-
and each child behavior variable in isolation. In Step 4,
sults indicated that the children changed significantly
final conditional models were run for each measure and
over 2 years on all measures (i.e., the variance in the
included all variables from Steps 2 and 3 that were sig-
rate of change over 2 years was significant). With re-
nificant in predicting variance in the slope of a given
gard to the shape of the growth curves, the linear effect
measure.
was significant and the quadratic was not, in all cases.

Results Conditional Model


Predictor Variable Creation: Total Item Multilevel modeling was used to examine whether T1
autism severity (i.e., CARS total score), T1 NVIQ, and/or
Correlations and Composite Scores one or more categories of problem behavior, measured
Table 1 summarizes the total number of items chosen either prior to the start of intervention (T1) or with a
and the corresponding coefficient alpha found for each difference score between T1 and T2, predicted the indi-
predictor variable. Coefficient alpha ≥ .70 indicated that vidual difference variance in the rate of change of any
the items used to calculate the composite score were in- child measure over 2 years.
ternally consistent with little measurement error (Streiner T1 scores for autism severity and NVIQ as predictors
of vocabulary and language development. Table 5 pre-
Table 1. Coefficient alpha for each child behavior variable. sents the results of the relationship between T1 CARS
and NVIQ scores and changes in vocabulary and language
Total number Coefficient measures over 2 years. Overall, the results revealed
Child behavior variable of items alpha that high T1 CARS scores alone predicted significant
decreases in the slopes of PPVT, EOWPVT, and PLS EC.
Acting-out behavior 19 .84
High T1 NVIQ scores alone predicted significant in-
Repetitive sensory motor behavior 22 .75
creases in the slopes of PPVT, EOWPVT, and PLS EC.
Insistence on sameness behavior 6 .62
Socially unresponsive behavior 20 .78
However, when both T1 CARS and T1 NVIQ were com-
Inattentiveness 10 .56 bined, T1 CARS scores no longer significantly predicted
the individual difference variance in the slopes of either

1112 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Table 2. Scores for Time 1 (T1) behavior variables used to predict changes in child development over 2 years.

Child behavior variable T1 M (range) T1 SD T2 M (range) T2 SD

Acting-out behavior (maximum score = 19) 8.86 (1–18) 3.92 7.41 (0–17) 4.00
Repetitive sensory motor behavior (maximum score = 22) 10.25 (2–20) 3.68 9.07 (0–17) 3.83
Insistence on sameness behavior (maximum score = 6) 3.94 (1–6) 1.51 3.31 (0–6) 1.66
Socially unresponsive behavior (maximum score = 20) 5.78 (0–13) 2.96 4.04 (0–13) 3.15
Inattentiveness (maximum score = 10) 6.72 (3–10) 1.72 5.87 (1–9) 1.80

Note. T2 = Time 2.

PPVT or EOWPVT. In the final analysis, the inclusion such as those available in other statistical packages (e.g.,
of T1 NVIQ in the models for PPVT and EOWPVT and SPSS). However, the statistical effects can be illustrated
the inclusion of both T1 NVIQ and T1 CARS in the model by examining the average rate of change in language
for PLS EC resulted in models that were a better fit, as or vocabulary growth for a prototypical child with mean
reflected in the lower AIC values in Table 5 compared inattentiveness and/or social unresponsiveness scores at
to those in Table 4. Thus, both T1 CARS and T1 NVIQ T1 and with scores ±1 standard deviation above and be-
scores were included in all subsequent models for PLS low the mean (Brooks & Meltzoff, 2008; Singer & Willet,
EC, and T1 NVIQ scores alone were included in all sub- 2003). Figures 1–3 display these results.
sequent models for PPVT and EOWPVT. Figure 1 displays the results for the EOWPVT, sug-
T1 scores for child behavior as predictors of vocabu- gesting that a child with fewer inattentive and socially
lary and language development. Table 6 presents the unresponsive behaviors at T1 (–1 SD) would have a rap-
results of the child behaviors as predictors of vocabu- idly accelerating growth curve, whereas a child with more
lary and language development from T1 to T4 (i.e., over inattentive and socially unresponsive behaviors at T1
2 years). The results revealed that two of the five be- (+1 SD) would have a much slower acceleration in growth
havior variables measured at T1 predicted significant of vocabulary production over 2 years. In Figure 2 (PLS
changes in vocabulary or language development mea- AC), a child with fewer inattentive and socially unrespon-
sures. First, a high score on inattentive behavior at T1 sive behaviors at T1 (–1 SD) would have greater rate of
was predictive of less of an increase in the rate of change acceleration over the first 12-month period only, com-
for EOWPVT and PLS AC over 2 years, regardless of T1 pared to a child with more inattentive and socially un-
NVIQ. Second, a high score on social unresponsiveness responsive behaviors at T1 (+1 SD). Finally, in Figure 3
at T1 was predictive of less of an increase in the rate of (PPVT), a child with fewer socially unresponsive behav-
change for EOWPVT, PPVT, and PLS AC over 2 years, iors at T1 (–1 SD) would have a slightly more rapid growth
regardless of T1 NVIQ. Scores for IS behaviors, RSMBs, acceleration over the first 12 months only, compared to
and acting-out behaviors at T1 did not predict the rate of a child with more socially unresponsive behaviors at
change of any child measure over 2 years beyond differ- T1 (+1 SD).
ences accounted for by T1 CARS and NVIQ status. Difference scores for problem behavior from T1 to T 2
The IGCM software used in this analysis (SAS Proc as predictors of vocabulary and language development.
Mixed) does not yield conventional effect size estimates The results indicated that changes in any of the problem

Table 3. Raw scores for measures of child development from Time 1 to Time 4.

Time 1 Time 2 Time 3 Time 4

Child measure M (range) SD M (range) SD M (range) SD M (range) SD

PPVT 9.87 (0–85) 17.68 20.94 (0–97) 23.50 27.87 (0–108) 26.74 38.19 (0–125) 29.27
EOWPVT 10.51 (0–65) 14.79 19.07 (0–75) 18.05 24.45 (0–86) 22.48 33.90 (0–96) 26.29
PLS AC 12.96 (3–48) 9.15 19.62 (3–44) 11.05 23.56 (5–48) 12.78 28.86 (8–48) 13.25
PLS EC 13.96 (4–47) 7.36 18.01 (6–44) 8.38 21.21 (5–48) 9.90 25.67 (7–48) 12.19
CARS 36.34 (25–50.5) 5.72 35.54 (17.5–49) 6.75 34.83 (19–49.5) 7.19 34.53 (16.5–50.5) 7.83

Note. PPVT = Peabody Picture Vocabulary Test; EOWPVT = Expressive One Word Picture Vocabulary Test; PLS = Preschool Language Scale;
AC = Auditory Comprehension; EC = Expressive Communication; CARS = Childhood Autism Rating Scale.

Bopp et al.: Behavior Predictors 1113

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Table 4. Unconditional models of change in measures of child language development from
Time 1 to Time 4.

Value PPVT EOWPVT PLS AC PLS EC

Intercept estimate 12.78 12.35 14.77 14.77


t -value 5.88*** 6.54*** 11.55*** 15.39***
Slope estimate 1.11 0.93 0.63 0.48
t -value 11.59*** 10.59*** 13.12*** 12.01***
Goodness of fit (AIC) 2149.8 1988.5 1782.5 1571.2

Note. AIC = Akaike Information Criterion.


***p < .0001.

behaviors from T1 to T2 did not predict changes in any outcomes. More specifically, the results revealed that
of the vocabulary or language outcome measures over autism severity scores prior to the onset of intervention
2 years. were predictive of only one communicative measure over
time: language production. CARS scores at T1 when com-
bined in the model with T1 NVIQ did not predict dif-
Discussion ferential outcomes over 2 years in either vocabulary or
language comprehension skills. Autism severity scores
This study provides an initial exploration of the in- have often been linked to developmental outcomes for
terrelationships between five types of behaviors and vo- children with ASD (DeMyer, 1973; Eaves & Ho, 1996;
cabulary and language development over time. It is unique Liss et al., 2001); however, the results of this study sug-
in that individual growth curve modeling was used to ex- gest that they may have little bearing on children’s rate
plore the impact of problem behaviors at T1 and changes of development on a broad range of language skills. On
in problem behaviors 6 months later on changes in vo- the other hand, NVIQ prior to the onset of intervention
cabulary and language development over 2 years. was found to predict changes related to receptive and
Results of the conditional analyses related to autism expressive vocabulary development and expressive lan-
severity and NVIQ generally confirmed past research guage development over time. These results echo the find-
that has found a relationship between these two measures ings of previous research (Gabriels et al., 2001; Harris
themselves and between NVIQ and other developmental & Handleman, 2000; Liss et al., 2001; Lovaas & Smith,

Table 5. T1 CARS and nonverbal IQ (NVIQ) as predictors of vocabulary and language


development from Time 1 to Time 4.a

Value PPVT EOWPVT PLS AC PLS EC

T1 CARS alone
Slope estimate –0.04 –0.04 –0.01 –0.02
t -value –2.73* –3.10* –1.75 –3.77**
AIC 2137.5 1977.1 1769.5 1557.8
T1 NVIQ alone
Slope estimate 0.02 0.02 0.004 0.009
t -value 3.85** 3.38** 1.26 3.57**
AIC 1820.4 1676.1 1501.1 1295.6
T1 CARS plus T1 NVIQ
T1 CARS slope estimate –0.016 –0.03 –0.004 –0.02
t-value –0.88 –1.82 –0.45 –2.11*
T1 NVIQ slope estimate 0.02 0.015 0.003 0.007
t-value 3.09* 2.36* 0.87 2.45*
AIC 1815.5 1672.1 1492.5 1289.4

a
Centered age as calculated by age minus mean age at T1 was used as the measure of time.
*p < .05. **p < .001.

1114 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Table 6. Results from final models showing effects of child behavior predictors on language
outcomes from Time 1 to Time 4. Models also included T1 NVIQ and CARS scores.

Value PPVT EOWPVT PLS AC PLS EC

Inattentiveness slope estimate –0.11 –0.13 –0.07 –0.03


t-value –1.84 –2.19* –2.53* –1.23
Social unresponsiveness slope estimate –0.07 –0.06 –0.03 –0.02
t-value –2.68* –2.21* –2.10* –1.94
Acting-out slope estimate 0.02 0.04 0.004 0.006
t-value 0.98 1.59 0.30 0.58
RSMB slope estimate –0.04 –0.04 –0.02 –0.003
t-value –1.46 –1.49 –1.55 –0.31
Insistence on sameness slope estimate 0.02 0.09 0.02 0.05
t-value 0.35 1.54 0.43 1.79

Note. RSMB = Repetitive stereotypic motor behaviors.


*p < .05. **p < .001.

1998; Sallows & Graupner, 2005; T. Smith et al., 1997), change and that poor attending skills negatively impact
although they reveal little about specific child charac- children’s ability to learn from the environment (Edelson,
teristics that may affect differential outcomes. Schubert, & Edelson, 1998). The results also lend support
Five problem behaviors were examined as potential the common practice of providing focused instruction on
predictors of language development in children with au- attending skills early in the treatment process (e.g.,
tism. The results indicate a relationship between inat- Lovaas, 2003).
tentive behaviors and language development over time. The number of socially unresponsive behaviors at
Inattentive behaviors in this study included, for exam- T1 also predicted children’s vocabulary or language de-
ple, not paying attention to sights and sounds in the velopment over 2 years. Socially unresponsive behaviors
environment; being distracted by noise; not listening to in this investigation included, for example, rarely smil-
instructions or a story; and looking away from tasks to ing, not looking at faces, actively avoiding eye contact,
notice other actions in the room. The results suggest that and failing to respond to one’s own name. The results sug-
children with more inattentive behaviors at T1 made sig- gest that children with more socially unresponsive be-
nificantly less progress in expressive vocabulary and lan- havior at T1 made significantly less progress over 2 years
guage comprehension development over 2 years. Thus, in both vocabulary comprehension and production and
prior to the start of treatment, it appeared that inatten- in language comprehension. Although social deficits are
tiveness hindered language development. These results central to the diagnosis of autism, researchers have only
support the suggestion that inattention is a correlate of recently started to examine how these deficits impact

Figure 1. Average growth curve of Expressive One Word Picture Figure 2. Average growth curve of Preschool Language Scale Auditory
Vocabulary Test (EOWPVT) scores over 2 years for a prototypical Comprehension (PLS AC) scores over 2 years for a prototypical child
child with T1 inattentive (IA) and social unresponsiveness (SU) scores with T1 inattentive and social unresponsiveness scores that were at the
that were at the mean and ±1 standard deviation above and below mean and ±1 standard deviation above and below the mean.
the mean.

Bopp et al.: Behavior Predictors 1115

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Figure 3. Average growth curve of Peabody Picture Vocabulary Test (Bronson, 2000; Hauser-Cram, Warfield, Shonkoff, &
(PPVT) scores over 2 years for a prototypical child with T1 social Krauss, 2001; Kaiser & Rasminsky, 2003). In addition,
unresponsiveness scores that were at the mean and ±1 standard acting-out behaviors are often viewed as serving one or
deviation above and below the mean. more behavioral functions, at least some of which can
be viewed as communicative (Durand & Merges, 2001).
Again, because the acting-out variable in this investi-
gation was not taken from an existing valid and reliable
standardized measure, the items that comprised it may
not have represented the construct adequately. In addi-
tion, the relationship between acting-out behavior and
communication is complex, and additional factors (e.g.,
parenting skills related to acting-out behavior, variables
related to treatment, etc.) are likely to enter into the
equation. Additional research that clarifies the complex
dynamics involved in both the origins and treatment of
acting-out behavior are sorely needed (e.g., Lucyshyn,
2007).

other developmental domains such as language (Carter,


Davis, Klin, & Volkmar, 2005). The current findings sup- Limitations
port the notion that social–affective skills play a central This study attempted to examine behavioral predic-
role in language development in this population (Mundy tors of developmental trajectories in young children with
& Markus, 1997). Perhaps children with few social inter- ASD over time and, as is often case with exploratory stud-
action skills have fewer social experiences, which, in turn, ies, it has several limitations. First, a relatively small
results in fewer opportunities to hear, respond to, pro- sample size (N = 69) was used. Although small sample
duce, and learn language in social contexts. This study sizes are not uncommon in ASD research (e.g., Charman
adds to the growing body of research in this regard and et al., 2005; Sigman & McGovern, 2005; Siller & Sigman,
emphasizes the importance of early social behaviors on 2002; Stone & Yoder, 2001; Szatmari, Bryson, Boyle,
later language development. Streiner, & Duku, 2003), follow-up studies are needed to
Insistence on sameness and repetitive stereotypic examine the predictive relationships found here in a larger
motor behaviors were found to have a nonsignificant re- group of participants. Second, the timeline of 2 years for
lationship with language development after differences this investigation was relatively short for a longitudinal
in NVIQ and CARS scores were considered. This finding design. Follow-up at 5 years and beyond for this group of
was unexpected, as past correlational studies have iden- children would be optimal, in order to ascertain the sta-
tified relationships between one or both of these behav- bility of the findings.
ioral categories and language ability (Campbell et al., Third, the data were originally collected for a pur-
1990; Carcani-Rathwell et al., 2006; Dadds et al., 1988; pose other than that for which they were used in this in-
Szatmari et al., 2006; Venter et al., 1992). In addition, vestigation. This resulted in several limitations. First,
past intervention studies have identified an inverse re- information was not available about the extent to which
lationship between RSMBs and the acquisition of com- each child’s treatment specifically targeted the problem
munication skills (Koegel, Koegel, Hurley, & Frea, 1992; behaviors examined in this study. Second, items included
Lee & Odom, 1996). One explanation is that the IS and within each child behavior variable were limited to only
RSMB variables in this study were not sufficiently sen- those items available in the existing measures in the data
sitive to identify changes across time and /or included set. Thus, none of the predictor variables incorporated all
too wide a range of behaviors in these two areas. Past possible behavioral manifestations of the behavior con-
studies that have identified a relationship between com- structs. For example, the IS variable contained only six
munication and RSMBs have typically focused on only a items and did not include behaviors such as circumscribed
few ADI–R items to measure this variable (e.g., Bishop interests, which have been found to be related to child
et al., 2006; Cuccaro et al., 2003; Szatmari et al., 2006). development in other studies (e.g., Bishop et al., 2006).
Acting-out behaviors were also found to have no re- Third, all of the items used to construct the predictor
lationship with vocabulary or language development over variables originated from parent-report measures, and
2 years. This finding was also somewhat unexpected, as none of the behaviors were directly observed by an expe-
past researchers have suggested that acting-out behav- rienced professional. Finally, even though care was taken
iors may interfere with learning and/or restrict children’s in the construction of the predictor variables with re-
opportunities to practice and develop skills in other areas gard to psychometric and content validity, none of the

1116 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
variables were taken directly from existing valid and substantially to our current understanding of these com-
reliable standardized measures of child behavior. In plex relationships and for assisting service providers,
addition, the coefficient alpha for inattentive behavior families, and policy makers to make more informed deci-
was slightly lower (.56) than the .60 value that is typ- sions about how to maximize the effectiveness of their
ically used to reflect adequate internal consistency. In early intervention efforts.
preliminary analyses, we attempted to increase the alpha
by systematically removing one inattentive behavior item
at a time from the variable, but this did not improve Acknowledgments
the outcome. We decided to use the variable in the anal-
This research was completed by the first author in partial
ysis regardless, as the 10 behaviors that comprise it fulfillment of the requirements for a doctoral degree at the
were identified as relevant by 80% or more of the expert University of British Columbia. Portions of this research were
panel members. In the end, this appears to have been a presented at the International Meeting for Autism Research
reasonable decision, as this variable was found to be a in Seattle, May 2007; the American Speech-Language-Hearing
significant predictor of language development despite Association conference in Chicago, November 2008; and on a
the reduced power associated with lower alpha scores. public webcast at http://breeze.setbc.org/p65709571/. We are
grateful to the children and families who participated in this
study and to the many evaluators who were involved in data
Future Directions collection. Partial funding was provided by the Human Early
The results of this research provide a preliminary Learning Partnership, the BC Ministry for Children and Family
understanding of some of the child problem behaviors Development, and the National Alliance for Autism Research /
Autism Speaks postdoctoral fellowship program.
that may be related to language development in young
children with ASD. Identifying such predictive relation-
ships may provide guidance for the development of treat- References
ments that result in optimal outcomes (Koegel et al.,
1992). For example, many intervention programs tar- Abbeduto, L., & Boudreau, D. (2004). Theoretical influences
on research on language development and intervention in
get early socio-communicative behaviors such as imita-
individuals with mental retardation. Mental Retardation
tion (Ingersoll, Schreibman, & Quy, 2003; Schreibman & and Developmental Disabilities, 10, 184–192.
Ingersoll, 2005), an approach supported by recent re-
Abidin, R. R. (1995). Parenting Stress Index: Professional
search demonstrating predictive relationships between manual (3rd ed.). Odessa, FL: Psychological Assessment
imitation skills and development in other areas (e.g., Resources.
V. Smith et al., 2007; Toth et al., 2006). The results of American Psychiatric Association. (2000). Diagnostic and
the present study suggest that children who are more at- statistical manual of mental disorders (4th ed., text revision).
tentive and socially responsive are likely to make greater Washington, DC: Author.
gains in vocabulary and/or language development. These Bagnato, S. T., Neisworth, J. T., Salvia, J., & Hunt, F. M.
findings support the theory that children with autism (1999). Manual for the Temperament and Atypical Behavior
learn language much like typical children do, through forms Scale (TABS): Early childhood indicators of developmental
dysfunction. Baltimore: Paul H. Brookes.
of joint social engagement (e.g., Abbeduto & Boudreau,
2004; Luyster et al., 2007). If future research supports Bishop, S., Richler, J., & Lord, C. (2006). Association be-
tween restricted and repetitive behaviors and nonverbal IQ
this finding, specific approaches aimed at improving at-
in children with autism spectrum disorders. Child Neuro-
tentiveness and socially responsive behavior would ap- psychology, 12, 247–267.
pear to be critical as key components of early treatment
Bodfish, J. W., Symons, F. J., Parker, D. E., & Lewis, M. H.
packages. In addition, research is needed to examine the (2000). Varieties of repetitive behavior in autism: Compari-
extent to which the predictive relationships found in sons to mental retardation. Journal of Autism and Devel-
this investigation are common across young children opmental Disorders, 30, 237–243.
with ASD who receive various types of early interven- Bopp, K., Brown, K., & Mirenda, P. (2004). Speech-language
tion. This will add to our current, limited understand- pathologists’ roles in the delivery of positive behavior sup-
ing of the influence of problem behaviors on the widely port for individuals with autism spectrum disorders. Amer-
ican Journal of Speech-Language Pathology, 13, 5–19.
heterogeneous outcomes observed in young children
with ASD (Schreibman, 2000). Bronson, M. B. (2000). Self-regulation in early education.
New York: Guilford.
In conclusion, this exploratory investigation opens
Brooks, R., & Meltzoff, A. N. (2008). Infant gaze following
the door to future examinations of predictive relation- and pointing predict accelerated vocabulary growth through
ships between specific child behaviors and changes in two years of age: A longitudinal, growth curve modeling
child vocabulary and language development over time, study. Journal of Child Language, 35, 207–220.
using analytic techniques that take advantage of multi- Brownell, R. (2000). Expressive One Word Picture Vocabulary
wave data. This type of research has potential for adding Test. Novato, CA: Academic Therapy Publications.

Bopp et al.: Behavior Predictors 1117

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Campbell, M., Locascio, J. J., Chorocco, M. C., Spencer, Eaves, L. C., & Ho, H. H. (1996). Brief report: Stability and
E. K., Malone, R. P., et al. (1990). Stereotypies and tardive change in cognitive and behavioral characteristics of autism
dyskinesia: Abnormal movements in autistic children. Psy- through childhood. Journal of Autism and Developmental
chopharmacological Bulletin, 26, 260–266. Disabilities, 26, 557–569.
Carcani-Rathwell, I., Rabe-Hasketh, S., & Santosh, P. Edelson, M. G., Schubert, D. T., & Edelson, S. M. (1998).
(2006). Repetitive and stereotyped behaviors in pervasive Factors predicting intelligence scores on the TONI in in-
developmental disorders. Journal of Child Psychology and dividuals with autism. Focus on Autism and Other Develop-
Psychiatry, 47, 573–581. mental Disabilities, 13, 17–26.
Carter, A. S., Davis, N. O., Klin, A., & Volkmar, F. R. (2005). Epstein, L. J., Taubman, M. T., & Lovaas, O. I. (1985).
Social development in autism. In F. R. Volkmar, R. Paul, Changes in self-stimulatory behaviors with treatment. Jour-
A. Klin, & D. Cohen (Eds.), Handbook of autism and perva- nal of Abnormal Psychology, 13, 281–294.
sive developmental disorders, Vol. 1: Diagnosis, development, Evans, J. L. (2001). An emergent account of language impair-
neurobiology, and behavior (3rd ed., pp. 312–334). Hoboken, ments in children with SLI: Implications for assessment and
NJ: John Wiley & Sons. intervention. Journal of Communication Disorders, 34, 39–54.
Charman, T., Baron-Cohen, S., Swettenham, J., Baird, Fenske, E. C., Zalenski, S., Krantz, P. J., & McClannahan,
G., Drew, A., & Cox, A. (2003). Predicting language out- L. E. (1985). Analysis and intervention in developmental
come in infants with autism and pervasive developmental disabilities. Early Intervention, 5, 49–58.
disorder. International Journal of Language and Commu-
nication Disorders, 38, 265–285. Francis, D. F., Fletcher, J. M., Stuebing, K. K., Davidson,
K. C., & Thompson, N. M. (1991). Analysis of change: Mod-
Charman, T., Taylor, E., Drew, A., Cockerill, H., Brown, J., eling individual growth. Journal of Consulting and Clinical
& Baird, G. (2005). Outcome at 7 years of children diagnosed Psychology, 59, 27–37.
with autism at age 2: Predictive validity of assessments con-
ducted at 2 and 3 years of age and pattern of symptom change Gabriels, R. L., Cuccaro, M. L., Hill, D. E., Ivers, B. J., &
over time. Journal of Child Psychology and Psychiatry, 46, Goldson, E. (2005). Repetitive behaviors in autism: Rela-
500–513. tionships with associated clinical features. Research in
Developmental Disabilities, 26, 169–181.
Clark, L. A., & Watson, D. (1995). Constructing validity:
Basic issues in objective scale development. Psychological Gabriels, R. L., Hill, D. E., Pierce, R. A., & Rogers, S. J.
Assessment, 7, 309–319. (2001). Predictors of treatment outcome in young children
with autism: A retrospective study. Autism: The Interna-
Cronbach, L. J. (2004). My current thoughts on coefficient tional Journal of Research & Practice, 5, 407–429.
alpha and successor procedures. Educational and Psycho-
logical Measurement, 64, 391–418. Georgiades, S., Szatmari, P., Zwaigenbaum, L., Duku, E.,
Bryson, S., et al. (2007). Structure of the autism symptom
Cuccaro, M. L., Shao, Y., Grubber, J., Slifer, M., Wolpert, phenotype: A proposed multidimensional model. Journal
C. M., et al. (2003). Factor analysis of restricted and re- of the American Academy of Child and Adolescent Psychiatry,
petitive behaviors in autism using the Autism Diagnostic 46, 188–196.
Interview–Revised. Child Psychiatry and Human Develop-
ment, 1, 3–17. Harris, S. L., & Handleman, J. S. (2000). Age and IQ at
intake as predictors of placement for young children with
Dadds, M., Schwartz, S., Adams, T., & Rose, S. (1988). The autism: A four- to six-year follow-up. Journal of Autism and
effects of social context and verbal skill on stereotypic and Developmental Disorders, 30, 137–142.
task-involved behavior of autistic children. Journal of Child
Psychology and Psychiatry, 29, 669–676. Hauser-Cram, P., Warfield, M. E., Shonkoff, J. P., & Krauss,
M. W. (2001). Children with disabilities: A longitudinal study
Dawson, G., Toth, K., Abbott, R., Osterling, J., Munson, J., of child development and parent well-being. Monographs
et al. (2004). Early social attention impairments in autism: of the Society for Research in Child Development, 66(3, Serial
Social orienting, joint attention, and attention to distress. No. 266).
Developmental Psychology, 40, 271–283.
Hollingshead, A. B. (1962). Socioeconomic status and mental
DeMyer, M. K. (1973). Prognosis in autism: A follow-up study. illness. Sociology and Social Research, 46, 387–396.
Journal of Autism and Childhood Schizophrenia, 3, 199–246.
Ingersoll, B., Schreibman, L., & Stahmer, A. (2001). Brief
Dissanayake, C., Sigman, M., & Kasari, C. (1996). Long-term report: Differential treatment outcomes for children with
stability of individual differences in the emotional respon- autistic spectrum disorder based on level of peer social
siveness of children with autism. Journal of Child Psychol- avoidance. Journal of Autism and Developmental Disorders,
ogy and Psychiatry, 37, 461–467. 31, 343–349.
Dunn, L. M., & Dunn, L. M. (1997). Examiner’s manual for the Ingersoll, B., Schreibman, L., & Quy, T. (2003). Effect of
PPVT–III: Peabody Picture Vocabulary Test—Third Edition: sensory feedback on immediate object imitation in children
Form IIIA and Form IIIB. Circle Pines, MN: American with autism. Journal of Autism and Developmental Dis-
Guidance Service. orders, 33, 673–683.
Dunn, W. (1999). Sensory Profile: User’s manual. San Antonio, Kaiser, B., & Rasminsky, J. S. (2003). Challenging behavior
TX: The Psychological Corporation. in young children: Understanding, preventing, and respond-
Durand, V. M., & Merges, E. (2001). Functional communica- ing effectively. Boston: Allyn and Bacon.
tion training: A contemporary behavior analytic technique Keith, T. Z. (1993). Latent variable structural equation mod-
for problem behavior. Focus on Autism and Other Develop- els: LISREL in special education research. Remedial and
mental Disabilities, 16, 110–119. Special Education Research, 14, 36–46.

1118 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Koegel, R. L., Koegel, Hurley, C., & Frea, W. (1992). Im- Richler, J., Bishop, S., Klienke, J., & Lord, C. (2007). Re-
proving social skills and disruptive behavior in children with stricted and repetitive behaviors in young children with
autism through self-management. Journal of Applied Be- autism spectrum disorders. Journal of Autism and Devel-
havior Analysis, 25, 341–353. opmental Disorders, 37, 73–85.
Krug, D. A., Arick, J. R., & Almond, P. J. (1980/1983). Au- Rollins, P. R., & Snow, C. E. (1998). Shared attention and
tism Screening Instrument for Educational Planning (2nd ed.). grammatical development in typical children and children
Austin, TX: Pro-Ed. with autism. Journal of Child Language, 25, 653–673.
Lee, S., & Odom, S. L. (1996). The relationship between ste- Rutter, M., Le Couteur, A., & Lord, C. (2003). Autism Diag-
reotypic behavior and peer social interaction for children nostic Interview—Revised. Los Angeles: Western Psycholog-
with severe disabilities. Journal for the Association of Persons ical Services.
with Severe Handicaps, 21, 88–95.
Sallows, G. O., & Graupner, T. D. (2005). Intensive behav-
Liss, M., Harel, B., Fein, D., Allen, D., Dunn, M., et al. ioral treatment for children with autism: Four-year outcome
(2001). Predictors and correlates of adaptive functioning in and predictors. American Journal of Mental Retardation,
children with developmental disorders. Journal of Autism 110, 417–438.
and Developmental Disorders, 31, 219–230.
Schopler, E., Reichler, R. J., & Renner, B. R. (1988). The
Lovaas, O. I. (1987). Behavioral treatment and normal educa-
Childhood Autism Rating Scale (CARS). Los Angeles, CA:
tional and intellectual functioning in young autistic children.
Western Psychological Services.
Journal of Consulting and Clinical Psychology, 55, 3–9.
Schreibman, L. (2000). Intensive behavioral /psychoeduca-
Lovaas, O. I. (2003). Teaching individuals with developmental
tional treatments for autism: Research needs and future
delays. Austin, TX: Pro-Ed.
directions. Journal of Autism and Developmental Disorders,
Lovaas, O. I., & Smith, T. (1998). Intensive and early behav- 30, 373–378.
ioral intervention with autism: The UCLA young autism
project. Infants and Young Children, 10, 67–78. Schreibman, L., & Ingersoll, B. (2005). Behavioral interven-
tions to promote learning with individuals with autism. In
Lucyshyn, J. (2007, May). Transforming coercive processes in F. R. Volkmar, R. Paul, A. Klin, & D. Cohen (Eds.), Handbook
family routines: Experimental analyses of family centered of autism and pervasive developmental disorders, Vol. 2:
positive behavior support. Symposium presented at the 33rd Assessment, interventions, and policy (3rd ed., pp. 882–896).
annual convention of the Association for Behavior Analysis Hoboken, NJ: John Wiley & Sons.
International, San Diego, CA.
Schumacker, R. E., & Lomax, R. G. (2004). A beginner’s guide
Luyster, R., Qiu, S., Lopez, K., & Lord, C. (2007). Predicting
to structural equation modeling (2nd ed.). Mahwah, NJ:
outcomes of children referred for autism using the MacArthur-
Lawrence Erlbaum Associates.
Bates communicative developmental inventory. Journal of
Speech, Language, and Hearing Research, 50, 667–681. Sigman, M., & McGovern, C. (2005). Improvement in cogni-
Milfont, T. L., & Gouveia, V. V. (2006). Time perspective and tive and language skills from preschool to adolescence in
values: An exploratory study of their relations to environ- autism. Journal of Autism and Developmental Disorders, 35,
mental attitudes. Journal of Environmental Psychology, 26, 15–23.
72–82. Siller, M., & Sigman, M. (2002). The behaviors of parents of
Mirenda, P., Bopp, K., Kavanagh, P., Smith, V., & Zaidman- children with autism predict the subsequent development
Zait, A. (2005). Outcomes of autism early intervention in of their children’s communication. Journal of Autism and
British Columbia, Canada. Proceedings of the Autism Society Developmental Disorders, 32, 77–89.
of America annual conference (pp. 25–30). Nashville, TN: Singer, J. D. (1998). Using SAS PROC MIXED to fit multi-
ASA. level models, hierarchical models, and individual growth mod-
Mullen, E. M. (1995). Mullen Scales of Early Learning. Circle els. Journal of Educational and Behavioral Statistics, 24,
Pines, MN: American Guidance Service. 323–355.
Mundy, P., & Markus, J. (1997). On the nature of communi- Singer, J. D., & Willet, J. B. (2003). Applied longitudinal data
cation and language impairment in autism. Mental Retarda- analysis: Modeling change and event outcome. New York:
tion and Developmental Disabilities Research Reviews, 3, Oxford University Press.
343–349.
Sireci, S. G. (1998). Gathering and analyzing content validity
Mundy, P., & Neal, R. (2001). Neural plasticity, joint atten- data. Educational Assessment, 5, 299–321.
tion, and autistic developmental pathology. In L. M. Glidden
(Ed.), International review of research in mental retardation Smith, T., Eikeseth, S., Klevstrand, M., & Lovaas, O. I.
(Vol. 23, pp. 139–168). New York: Academic Press. (1997). Intensive behavioral treatment for preschoolers with
severe mental retardations and pervasive developmental
National Research Council. (2001). Educating children with disorder. American Journal of Mental Retardation, 102,
autism. Committee on Educational Interventions for Chil- 238–249.
dren with Autism, Division of Behavioral and Social Sci-
ences and Education. Washington, DC: National Academy Smith, V., Mirenda, P., & Zaidman-Zait, A. (2007). Pre-
Press. dictors of expressive vocabulary growth in children with
autism. Journal of Speech, Language, and Hearing Re-
Perry, A., Condillac, R. A., Freeman, N. L., Dunn-Geier,
search, 50, 149–160.
J., & Belair, J. (2005). Multi-site study of the Childhood
Autism Rating Scale (CARS) in five clinical groups of young Sparrow, S., Balla, D., & Cicchetti, D. (1984). Vineland
children. Journal of Autism and Developmental Disorders, Adaptive Behavior Scales: Expanded manual. Circle Pines,
35, 625–634. MN: American Guidance Service.

Bopp et al.: Behavior Predictors 1119

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Stone, W. L., & Yoder, P. (2001). Predicting spoken language Wing, L., & Gould, J. (1979). Severe impairments of social
level in children with autism spectrum disorders. Autism: interaction and associated abnormalities in children: Epi-
The International Journal of Research & Practice, 5, 341–362. demiology and classification. Journal of Autism and Devel-
Streiner, D. L., & Norman, G. R. (1989). Health measure- opmental Disorders, 9, 11–29.
ment scales: A practical guide to their development and use. Yoder, P. J. (2006). Predicting lexical density growth rate in
New York: Oxford University Press. young children with autism spectrum disorders. American
Szatmari, P., Bryson, S. E., Boyle, M. H., Streiner, D. L., & Journal of Speech Language Pathology, 15, 378–388.
Duku, E. (2003). Predictors of outcome among high func- Yoder, P. J., & McDuffie, A. S. (2006). Treatment of respond-
tioning children with autism and Asperger syndrome. Jour- ing to and initiating joint attention. In T. Charman & W. Stone
nal of Child Psychology and Psychiatry, 44, 520–528. (Eds.), Social and communication development in autism
Szatmari, P., Georgiades, S., Bryson, S., Zwaigenbaum, spectrum disorders: Early identification, diagnosis, and in-
L., Roberts, W., et al. (2006). Investigating the structure of tervention (pp. 117–142). New York: Guilford Press.
the restricted, repetitive behaviors and interests domain Zimmerman, I. L., Steiner, V. G., & Pond, R. E. (1992).
of autism. Journal of Child Psychology and Psychiatry, 47, PLS-3: Preschool Language Scale–3. San Antonio, TX: The
582–590. Psychological Corporation.
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate Zumbo, B. D., Gelin, M. N., & Hubley, A. M. (2003). The
statistics (5th ed.). Boston: Allyn and Bacon. construction and use of psychological tests and measures.
Toth, K., Munson, J., Meltzoff, A., & Dawson, G. (2006). In S. Carta (Ed.), The psychology theme of the Encyclopedia
Early predictors of communication development in young of Life Support Systems (EOLSS). Oxford, UK: EOLSS.
children with autism spectrum disorder: Joint attention, Retrieved June 21, 2006, from http://www.eolss.net.
imitation, and toy play. Journal of Autism and Develop-
mental Disorders, 36, 993–1005.
Received November 27, 2007
Travis, L., Sigman, M., & Ruskin, E. (2001). Links between
Revision received April 10, 2008
social understanding and social behavior in verbally able chil-
dren with autism. Journal of Autism and Developmental Accepted January 13, 2009
Disorders, 31, 119–130. DOI: 10.1044/1092-4388(2009/07-0262)
Venter, A., Lord, C., & Schopler, E. (1992). A follow-up study Contact author: Karen D. Bopp, The University of British
of high functioning autistic children. Journal of Child Psy- Columbia, 2125 Main Mall, Vancouver, BC, Canada,
chology and Psychiatry, 33, 489–507.
V6T 1Z4. E-mail: bopp@interchange.ubc.ca.
Volkmar, F. R., & Pauls, D. (2003). Autism. The Lancet, 362,
1133–1141.

1120 Journal of Speech, Language, and Hearing Research • Vol. 52 • 1106–1120 • October 2009

Downloaded From: http://jslhr.pubs.asha.org/ by a Univ Of Newcastle User on 11/02/2016


Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx

You might also like