Journal of School Psychology 42 (2004) 285 301

Elementary school students with AD/HD: predictors

of academic achievement
George J. DuPaul*, Robert J. Volpe, Asha K. Jitendra, J. Gary Lutz,
Kristi S. Lorah, Rosemary Gruber
Department of Education and Human Services, Lehigh University, Bethlehem PA, USA
Received 24 November 2003; accepted 26 May 2004

Abstract

Academic underachievement frequently is associated with attention-deficit/hyperactivity

disorder (AD/HD); however, the role of variables beyond AD/HD symptoms and cognitive
mediators is unknown. Further, whether prediction models vary (a) relative to non-AD/HD
students, (b) between math and reading, and (c) based on how achievement is defined has not
been examined. Multiple measures (e.g., teacher ratings and behavior observations) were
examined as predictors for concurrent achievement outcomes (standardized achievement test
scores and report card grades) in math and reading in two samples of 1st through 4th grade
children (136 with AD/HD, 53 without AD/HD). Teacher perceptions of academic skills were the
strongest predictors of achievement test scores for both groups, while academic skills and enablers
accounted for reading report card grades in children with AD/HD but not their normal
counterparts. Implications of these findings for school-based assessment and intervention for
students with AD/HD are discussed.
D 2004 Society for the Study of School Psychology. Published by Elsevier Ltd. All rights reserved.

Keywords: AD/HD; Academic achievement; Assessment; Prediction

* Corresponding author. School Psychology Program, Lehigh University, 111 Research Drive, Bethlehem PA
18015, USA. Tel.: +61 758 3252; fax: +61 758 6223.
E-mail address: gjd3@lehigh.edu (G.J. DuPaul).

0022-4405/$ - see front matter D 2004 Society for the Study of School Psychology. Published by Elsevier Ltd.
All rights reserved.
doi:10.1016/j.jsp.2004.05.001
286 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

Introduction

Numerous investigations indicate that the behavioral symptoms (i.e., inattention,

impulsivity, and overactivity) of attention-deficit/hyperactivity disorder (AD/HD) are
associated with concurrent and later academic underachievement (e.g., DeShazo Barry,
Lyman, & Klinger, 2002; Fergusson & Horwood, 1995; Rapport, Scanlan, & Denny, 1999;
Zentall, Smith, Lee, & Wieczorek, 1994). In fact, one of the most ubiquitous risks
associated with AD/HD is poor scholastic outcome including higher than average
frequencies of failing grades, grade retention, and school drop-out (e.g., Barkley, Fischer,
Edelbrock, & Smallish, 1990; Mannuzza, Gittelman-Klein, Bessler, Malloy, & LaPadula,
1993). The relationship with academic underachievement appears to be specific to AD/
HD-related behaviors and is not accounted for by comorbid conduct problems (Frick et al.,
1991; Hinshaw, 1992; Rapport et al., 1999).
Investigations, to date, have examined the association between AD/HD symptoms and
academic achievement at single or multiple points in time using correlational or structural
equation modeling analyses, primarily with large nonreferred samples. For example, a
series of longitudinal studies (Fergusson & Horwood, 1995; Fergusson, Horwood, &
Lynskey, 1993; Fergusson, Lynskey, & Horwood, 1997) conducted with a sample of over
700 children in New Zealand has demonstrated clear linkages between AD/HD behaviors
in elementary and middle school (based on maternal and teacher ratings) and later levels of
academic achievement (middle school through age 18). Specifically, structural equation
models indicated that early, high levels of AD/HD behaviors were associated with
concurrent and later, lower levels of academic achievement.
Rapport et al. (1999) replicated the findings of Fergusson et al. by assessing AD/HD
symptoms and scholastic achievement in an ethnically diverse sample of 325 Hawaiian
schoolchildren. The relationship between early AD/HD symptoms (based on teacher
ratings) and later academic achievement (based on a group-administered achievement test)
was confirmed by these investigators. Further analyses indicated that the influence of AD/
HD behaviors on scholastic status was mediated by both cognitive (e.g., memory) and
behavioral (e.g., task engagement) variables.
Studies conducted with clinical samples of children diagnosed with AD/HD and related
disruptive behavior disorders (i.e., Oppositional Defiant Disorder [ODD] and Conduct
Disorder [CD]) have confirmed the specific relationship between AD/HD symptoms and
academic achievement (Frick et al., 1991; Hinshaw, 1992). Recently, DeShazo Barry et al.
(2002) examined the degree to which measures of AD/HD symptoms and executive
functioning predicted academic underachievement in a sample of 66 children, including 33
children with AD/HD. Their findings demonstrated that symptoms of AD/HD
independently predicted underachievement over and above variance accounted for by
executive (i.e., cognitive) functioning. The total variance in achievement accounted for by
the combination of predictors ranged from 15% to 28% depending on the academic subject
area. These results imply that the greater the severity of AD/HD symptoms, the greater the
degree of underachievement. Alternatively, a considerable amount of variance remains
unaccounted for, thus necessitating identification of additional predictor variables.
The identification of factors that account for significant variance in achievement
outcomes is important for both assessment and treatment of this disorder. Specifically,
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 287

variables found to be significant predictors of achievement difficulties would need to be

assessed prior to treatment and could serve as possible targets for school-based
intervention. Given the results of prior investigations, several variables appear to have a
role in predicting academic achievement among children with AD/HD, including cognitive
factors (e.g., skill level), task engagement, and severity of AD/HD symptoms.
Furthermore, the educational literature is replete with investigations demonstrating
significant correlations between demographic variables (e.g., socioeconomic status
[SES], ethnicity) and academic outcomes in nonreferred samples (Diamond &
Onwuegbuzie, 2001; Jimerson, Egeland, & Teo, 1999). Rapport et al. (1999) has
theorized that the primary predictors of achievement in this population are cognitive
variables, chiefly memory, and rates of task engagement, which presumably are related to
the degree of attentional impairment. The primacy of these variables was supported by
data collected in a large, nonreferred sample (Rapport et al., 1999). To date, the relative
contributions of these and other potential predictor variables have not been examined
specifically for a referred sample of children with AD/HD.
Prior investigations of ac.ademic achievement in children with AD/HD have neglected
three other potentially important factors. First, studies typically have focused on one area
of academic achievement (i.e., either reading or math). Thus, it is unclear whether
predictors vary as a function of academic subject area. Given that instructional activities
and work assignments in reading and math may differ in terms of mode of delivery (e.g.,
lecture vs. group interaction), task expectations (silent reading vs. written answers), and
types of studentstudent interactions (e.g., independent seatwork vs. cooperative learning
exercises), this may be an important distinction. Second, norm-referenced, standardized
achievement tests have been the primary outcome measure of academic success. Although
such tests have many advantages (e.g., strong psychometric properties), these measures
typically are not used in determining student academic progress unless the child is referred
for an evaluation of special education eligibility. Alternatively, measures (e.g., report card
grades) that are used on a regular basis in school settings have not been included in prior
studies with this population. Finally, the degree to which variables vary in their utility for
predicting the academic achievement of children with AD/HD relative to their non-AD/
HD counterparts has not been examined.
The purpose of this investigation was to identify potential predictors of
contemporaneous academic achievement in a sample of children meeting research
criteria for AD/HD. Several research questions were addressed. First, to what degree
do AD/HD symptoms, conduct problems, classroom behavior, and academic skills
contribute to predicting academic achievement in children with AD/HD? Based on the
theory of Rapport et al. (1999) and the results of prior investigations, we conducted
hierarchical regression analyses to examine the relative contributions of putative
predictors over and above the variance in academic achievement accounted for by SES
and ethnicity. Although these analyses were exploratory, we hypothesized that all of
these variables would make independent, significant contributions to predicting
academic achievement. Second, do prediction models differ as a function of academic
subject (i.e., reading and math) and/or how achievement is defined (i.e., achievement
test score vs. report card grades)? Finally, do prediction models for children with AD/
HD differ from those found with non-AD/HD peers? Thus, separate regression analyses
288 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

were conducted for children with and without AD/HD in reading and math using two
different outcome (criterion) measures.

Method

Participants

Two groups of children participated in this study. The first group included 136 students
(106 boys, 30 girls) attending first through fourth grade (M, age=8.5 years; S.D.=1.2) in
public elementary schools in eastern Pennsylvania who were identified as having AD/HD,
as described below. The second group comprised 53 students (39 boys, 14 girls) from the
same schools and grades as the AD/HD group (M age=8.5 years; S.D.=1.1) who served as
normal controls. Both groups primarily comprised Caucasian children (AD/HD group,
n=84; Normal Control group, n=42), but also included children from Latino (AD/HD
group, n=32; Normal Control group, n=9) and AfricanAmerican (AD/HD group, n=14;
Normal Control group, n=2) backgrounds. Participants were recruited from schools in
urban, rural, and suburban settings. The percentage of students in these schools receiving
free and reduced lunch range from 0% to 78% (M=30.7%; S.D.=26.7%).
Students in the AD/HD group were referred to this study by their teachers due to
concerns regarding inattentive and/or hyperactiveimpulsive behavior as well as
difficulties with reading and/or math achievement. Furthermore, to be identified as
having AD/HD for the purposes of this study, children must have:

(a) received parent and teacher ratings on the AD/HD Rating Scale-IV (DuPaul, Power,
Anastopoulos, & Reid, 1998) at or above the 90th percentile on either the Inattention
or HyperactivityImpulsivity subscales using appropriate age and gender norms and
(b) met DSM-IV-TR (American Psychiatric Association, 2000) criteria for one of the
three subtypes of AD/HD based on a parent interview using the NIMH Diagnostic
Interview Schedule for Children-IV (NIMH-DISC-IV; Shaffer, Fisher, & Lucas,
1998).

Normal Control children were recruited from the same grades and schools as the AD/
HD participants but were in different classrooms. They were reported by their teachers to
be baverageQ in terms of classroom behavior and academic achievement. To be included in
the Normal Control group, children:

(a) received parent and teacher ratings on the AD/HD Rating Scale-IV below the 90th
percentile for their age and gender on both the Inattention and Hyperactivity
Impulsivity subscales and
(b) did not meet DSM-IV criteria for any of the disruptive behavior disorders (AD/HD,
ODD, or CD) according to a parent interview with the NIMH-DISC-IV.

For both groups, children with clinically significant symptoms of autism or

developmental disabilities, who had experienced brain damage, who had visual or hearing
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 289

impairments, or who were identified as cognitively impaired by their school were excluded
from participation.
Of the 136 children in the AD/HD group, 95 met study criteria for the Combined Type,
31 for the Predominantly Inattentive Type, and 10 for the Predominantly Hyperactive
Impulsive Type. In addition, 56 children met study criteria for comorbid Oppositional
Defiant Disorder and another 25 exhibited clinically significant Conduct Disorder
symptoms. All of the children in the AD/HD group were placed in general education
classrooms for at least a portion of the school day, with a small minority (n=5) receiving
part-time special education services. A total of 43 children (31.6%) in the AD/HD Group
were receiving psychotropic medication including psychostimulants (n=34), antidepres-
sants (n=1), and other medications (n=8). None of the Normal Control participants
received special education or psychotropic medication.
As expected, the two groups differed with respect to parent and teacher ratings on the
AD/HD Rating Scale-IV (see Table 1). Furthermore, the Normal Control group contained
a significantly higher percentage of Caucasian participants than did the AD/HD Group
( pb.05) and was of significantly higher ( pb.01) socioeconomic status as measured by the
Hollingshead Index (Hollingshead, 1975). Alternatively, the two groups were equivalent
in age and percentage of males.

Measures

Screening measures
The AD/HD Rating Scale-IV (DuPaul et al., 1998) is a behavior rating scale that
includes items directly related to the 18 symptoms of AD/HD based on the DSM-IV-TR
(American Psychiatric Association, 2000). Home and school versions are available for
completion by parents and teachers, respectively. Items are scored on a 0 (never or rarely)
to 3 (very often) basis. Normative data based on age and gender are available and the
psychometric properties of this instrument are well established (DuPaul et al., 1998).
The Computerized NIMH Diagnostic Interview Schedule for Children (Parent Version;
CDISC 4.0; Shaffer et al., 1998) is a structured diagnostic interview that is administered
using computer software. The Disruptive Behavior Disorders module was administered by

Table 1
Demographic characteristics of AD/HD and normal control groups
Measure Group t (185) or v 2 (1) p
AD/HD Normal
Age (in months) 102.9 (14.6) 103.8 (13.7) .36 .72
Male (%) 77.9 73.6 .40 .52
Caucasian (%) 61.8 79.2 5.2 .02
Hollingshead index (SES) 52.9 (24.5) 67.7 (20.9) 3.51 b.01
Teacher AD/HDRS-IV
Inattention 22.6 (4.2) 1.1 (1.5) 50.9 b.001
Hyperactive/impulsive 18.1 (7.3) 1.1 (1.9) 25.1 b.001
Parent AD/HDRS-IV
Inattention 19.7 (5.0) 1.7 (2.1) 34.7 b.001
Hyperactive/impulsive 17.6 (5.4) 1.7 (2.1) 29.2 b.001
290 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

a trained interviewer either in person or by telephone. Diagnostic decisions based on this

interview have been found to be highly reliable (Shaffer et al., 1998). All CDISC 4.0
interviews were audiotaped and a random subsample (21%) was reviewed by a second
trained interviewer to assess interdiagnostician agreement. Agreement was 100% across all
interviews with respect to overall diagnosis and subtype designation.

Criterion measures
Standard scores on the Broad Reading and Broad Math subtests of the Woodcock
Johnson III Tests of Achievement (WJ-III; Woodcock, McGrew, & Mather, 2001) served as
criterion measures of academic achievement. Broad Reading is a composite score based on
a students performance on three subtests: LetterWord Identification, Reading Fluency,
and Passage Comprehension. Broad Math is a composite score based on a students
performance on the Calculation, Math Fluency, and Applied Problems subtests. The WJ-
III is a widely used individually administered, norm-referenced achievement test that has
exemplary psychometric properties (Mather & Woodcock, 2001).
Report card grades in reading and math also served as criterion measures. Grades
provided by classroom teachers in these subject areas were converted to numerical scores
ranging from 1 (bFQ) to 5 (bAQ).

Predictor measures
Several measures were included as potential predictors of academic achievement
including teacher ratings of AD/HD symptoms, academic skills, achievement-related
behaviors, and social skills, as well as direct observations of classroom behavior.
Raw scores on the Inattention and HyperactivityImpulsivity subscales of the AD/HD
Rating Scale-IV (School Version; DuPaul et al., 1998) served as measures of AD/HD
symptomology, while the raw score on the Conduct Problems subscale of the Behavior
Assessment System for Children (BASC; Reynolds & Kamphaus, 1992) reflected levels of
disruptive behavior related to ODD and CD. The BASC is used frequently in research
studies with this population and has adequate psychometric properties (Reynolds &
Kamphaus, 1998). The Conduct Problems subscale includes 10 items that are scored on a 4-
point Likert scale ranging from bNeverQ to bAlmost Always.Q Raw scores on the
Mathematics, Reading/Language Arts, and Academic Enablers subscales of the Academic
Competency Evaluation Scale (ACES; DiPerna & Elliott, 2000) provided teacher
perceptions of childrens academic skills and achievement-related behaviors. The Mathe-
matics subscale contains 8 items, while the Reading/Language Arts subscale includes 11
items. The Academic Enablers subscale includes 40 items across four factors (Interpersonal
Skills, Engagement, Motivation, and Study Skills). For each of these three subscales, item
frequencies were scored on a 1 (bneverQ) to 5 (balmost alwaysQ) Likert scale. The ACES has
more than adequate levels of reliability and validity (DiPerna & Elliott, 2000). Finally,
teacher perceptions of childrens social skills were represented by standard scores on
the Social Skills subscale of the Social Skills Rating System (SSRS; Gresham & Elliott, 1990).
This subscale is comprised of 30 items rated on a 0 (bneverQ) to 2 (bvery oftenQ) basis.
The reliability and validity of the SSRS are well established (Gresham & Elliott, 1990).
An adapted version of the Behavior Observation System for Students (BOSS; Shapiro,
1996) was used to collect data on student behavior during math and reading class periods.
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 291

Using 15-s observation intervals, a momentary time sampling procedure was used to
categorize childrens engagement as either active (e.g., answering teacher questions or
writing) or passive (e.g., listening to teacher directions or reading silently). Childrens off-
task behavior was categorized on a partial interval basis as off-task-motor (e.g., fidgeting),
off-task-verbal (e.g., talking out without permission), or off-task-passive (e.g., looking
away from assigned task or activity). For the purposes of this study, the three off-task
categories were combined for a total-off-task percentage score. In addition, childrens
noncompliant behavior (e.g., refusing to follow a teacher directive prior to the initiation of
the next observation interval) was coded on a partial interval basis. Specific predictor
measures were the percentage of observation intervals for active engaged, passive
engaged, total off-task, and noncompliance.

Procedures

Local school districts were informed of a study examining the effects of academic
interventions for students with AD/HD and were asked to refer children with AD/HD
related behaviors who were also experiencing difficulties in reading and/or math. Parents
and teachers of referred children completed the AD/HD Rating Scale-IV and parents of
those children who met criteria on behavior ratings were then interviewed by phone or in
person using the CDISC 4.0. Parents of children who met inclusion criteria on the CDISC
4.0 then provided informed consent for participation in the study. A total of 250 students
were referred, of whom 144 were deemed eligible as participants with AD/HD. Consent
was provided for 136 children.
Teachers were then solicited to identify children as possible normal control participants
from the same grade levels as participants with AD/HD but in different classrooms to
avoid potential bias in teacher ratings due to a contrast effect. Students referred for the
Normal Control group were screened using the AD/HD Rating Scale-IV and CDISC 4.0.
Parents of those children who met criteria provided informed consent for participation. A
total of 60 students were nominated for this group, all of whom were deemed eligible.
Consent was provided for 53 normal control participants.
Once consent was obtained, predictor and criterion measures were collected
contemporaneously over approximately a 1-month period during the middle of the school
year (December to February). All measures were collected prior to academic interventions
being implemented in classrooms (i.e., during the baseline period of the larger study).
Trained graduate students in school psychology, special education, and counseling
psychology administered the WJ-III and conducted classroom observations using the
BOSS. Research assistants were blind to the purposes of the study and to group
membership of participating children. Behavior ratings were distributed directly to teachers
by data collectors and were returned to the investigators by mail. A stipend of $50 was
provided to teachers upon completion of the packet of rating scales. Schools or parents
provided photocopies of the most recent report card to provide grades in reading and math.
Behavioral observations were conducted for approximately 15 min each during one
math and reading class for each participant. For most students, these two observations
were conducted during the same school day. A second (reliability) observer was present
for 34% of these observations. Mean percentages of occurrence, nonoccurrence, and total
292 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

agreement across the behavioral categories and two subject areas ranged from 91.5% to
99.27% (M=96.56; S.D.=2.32). Mean Kappa coefficients ranged from .93 to .98 (M=.95;
S.D.=.02).

Data analyses

Prior to examining prediction models, group differences in academic achievement and

potential predictor variables were examined using Hotellings T2, followed by separate t-
tests for each variable. An alpha level of .01 was used for all analyses to control for
inflated experiment-wise Type I error. In order to establish whether boys and girls with
AD/HD could be combined for subsequent analyses, gender differences in criterion and
predictor variables were examined using Hotellings T2, followed by separate t-tests for
each variable.
Two different prediction models using hierarchical regression analyses were examined
for both math and reading achievements using WJ-III scores and report grades as criterion
measures in separate analyses. Each of these models was examined separately for the AD/
HD and Normal Control groups. In the initial stage of these analyses, ethnicity (coded as a
nominal level variable for white (1) vs. non-white (0)) and SES were entered as predictor
variables. This served as the bbaselineQ model wherein the contribution of additional
predictors was examined in sequential fashion using forward selection procedures. First,
the ACES math or reading/language arts raw score (representing teacher perception of
academic skills) was examined. Next, three variables related to engagement (ACES
Academic Enablers (AcEn) raw score, BOSS Active Engagement (AET), and BOSS
Passive Engagement (PET)) were entered. In the final stage, variables related to AD/HD
symptoms (AD/HD Rating Scale-IV Inattention and HyperactivityImpulsivity raw
scores, BOSS mean off-task (OT) percentage), social skills (SSRS Social Skills (SocSk)
standard score), and conduct problems (BASC Conduct Problems (Cond) standard score
and BOSS noncompliance (Nonc) percentage) were examined.

Results

Group differences in academic achievement and predictor variables

Table 2 displays the means and standard deviations for the criterion (academic
achievement) and predictor (teacher ratings and behavior observations) variables. A series
of Hotellings T2 analyses were conducted separately for three variable subsets including
academic achievement (WJ-III scores and report card grades), teacher behavior and
academic ratings (ACES Academic Enablers, ACES Academic Skills, SSRS Social Skills,
BASC Conduct Problems, AD/HD Rating Scale-IV Inattention and Hyperactivity
Impulsivity), and behavior observations (BOSS scores). Significant multivariate effects
were obtained for academic achievement (Wilks Lambda=.52, approximate
F (4,164)=38.2, p b.001), teacher ratings (Wilks Lambda=.10, approximate
F(6,111)=159.7, pb.001), and behavior observations (Wilks Lambda=.72, approximate
F(8,175)=8.56, pb.001).
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 293

Table 2
Group differences in achievement and predictor variables
Measure Group t (185) p ES
AD/HD Control
WJ-III broad reading 88.4 (14.9) 106.3 (12.1) 7.8 b.001 1.48
WJ-III broad math 94.5 (12.7) 113.4 (10.3) 9.7 b.001 1.83
Reading report card grade 3.2 (1.0) 4.6 (.5) 11.9 b.001 2.59
Math report card grade 3.3 (1.0) 4.7 (.7) 10.4 b.001 2.07
ACES academic skills 61.3 (17.24) 108.4 (18.3) 15.6 b.001 2.58
ACES academic enablers 100.7 (20.7) 174.5 (16.9) 21.7 b.001 4.36
SSRS social skills 82.8 (8.3) 112.2 (9.6) 19.6 b.001 3.05
BASC conduct problems 3.2 (2.9) 0.8 (2.9) 4.3 b.001 0.8
*AD/HDRS-IV inattention 21.6 (4.0) 0.8 (1.2) 28.7 b.001 17.3
*AD/HDRS-IV HypImp 14.5 (7.9) 1.5 (2.2) 10.3 b.001 5.9
BOSS AET reading 24.2 (15.7) 23.4 (15.3) 0.8 .40 .05
BOSS PET reading 49.4 (20.1) 59.3 (18.7) 3.3 .001 .53
BOSS OFT reading 10.4 (8.0) 5.2 (4.8) 5.7 b.001 1.08
BOSS NonC reading 0.7 (1.8) .06 (.35) 2.2 .03 1.83
BOSS AET math 26.6 (15.4) 28.6 (16.1) 1.5 .12 .12
BOSS PET math 41.3 (18.1) 52.2 (17.7) 3.6 b.001 .62
BOSS OFT math 12.6 (7.3) 6.1 (5.2) 6.5 b.001 1.25
BOSS NonC math 0.5 (1.3) 0 (0) 2.3 .02 .38
Note: AD/HDRS-IV completed by teacher; AET=Active Engagement; PET=Passive Engagement; OFT=Off-task;
NonC=Noncompliance.

Separate t-tests were conducted for each variable with most group differences
statistically significant at the pb.001 level (see Table 2). Specifically, children in the
Normal Control group obtained significantly higher WJ-III scores, report card grades,
teacher ratings of academic skills and enablers, as well as teacher ratings of social skills.
Effect sizes (calculated using Cohens d) were large and represented group differences of
more than one standard deviation. As expected, students in the AD/HD group were
reported by their teachers to display significantly greater frequencies of AD/HD symptoms
and conduct problems. Effect sizes for these group differences were also large.
Alternatively, for the behavior observations, the two groups were significantly different
(using an alpha level of .01) only with respect to passive engaged and off-task behavior.
Effect sizes for these differences were moderate (passive engaged) to large (off-task).
Although children in the AD/HD group were observed to be more noncompliant than
normal control children, these differences did not reach the threshold for statistical
significance.

Gender differences in achievement for children with AD/HD

A series of t-tests were conducted for each of the predictor and criterion variables to
examine whether there were differences between boys and girls with AD/HD on any of
these measures. None of these analyses revealed significant between-gender differences
(all p valuesN.05). Thus, subsequent multiple regression analyses were conducted using
both boys and girls in the AD/HD group.
294 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

Hierarchical regression analyses for math achievement

Table 3 presents correlations between math achievement predictor variables for the AD/
HD (above the diagonal) and Normal Control (below the diagonal) groups. Using the WJ-
III Broad Math standard score as the criterion variable, separate multiple regression
analyses were conducted for the Normal Control and AD/HD groups using the predictor
entry strategy described above. The baseline model (i.e., including SES and ethnicity as
predictor variables) for control participants was nonsignificant ( F(2,32)=2.1, p=.14) and
no further predictors were identified as statistically significant. Alternatively, a model
including SES, ethnicity, and ACES Math Skills score was not statistically significant for
predicting WJ-III Broad Math for participants with AD/HD ( F(3,57)=2.4, p=.07; R 2=.11;
see Table 4).
A similar set of analyses was conducted using math report card grade as the criterion
variable. As was the case for the WJ-III Broad Math score, the baseline prediction model
was nonsignificant for the normal control participants ( F(2,30)=2.4, p=.10; R 2=.27).
Nevertheless, the standardized beta coefficient for ethnicity was significant ( pb.05).
Alternatively, SES, ethnicity, and ACES Math Skills score were included in a statistically
significant prediction model for participants with AD/HD ( F(3,57)=3.38, p=.02; R 2=.15).
The standardized beta coefficient for ACES Math Skills was statistically significant
( pb.05), while coefficients for ethnicity and SES were not significant (see Table 4).

Table 3
Intercorrelations among predictor variables in math for AD/HD and normal control participants
Measure 1 2 3 4 5 6 7 8 9 10 11 12
1. Ethnicitya .19 .04 .06 .08 .18 .04 .06 .04 .05 .02 .17
2. SES .41* .12 .19 .01 .16 .07 .06 .12 .10 .05 .12
3. ACES Math .23 .06 .28* .03 .03 .05 .12 .06 .06 .06 .14
4. ACES AcEn .31 .27 .27 .06 .10 .27* .41** .17 .37** .07 .56**
5. BOSS AET .24 .17 .03 .24 .17 .09 .01 .32** .00 .11 .07
6. BOSS PET .30 .20 .13 .21 .75** .14 .09 .48** .03 .04 .09
7. ADHD HI .00 .00 .26 .28 .17 .24 .21 .20 .57** .02 .32*
8. ADHD IN .15 .16 .14 .54** .16 .28 .24 .09 .23 .02 .29*
9. BOSS OT .08 .12 .34 .14 .15 .37* .08 .12 .06 .01 .16
10. BASC Cond .07 .18 .02 .05 .02 .04 .08 .23 .04 .09 .40**
11. BOSS Nonc NAb NA NA NA NA NA NA NA NA NA .09
12. SSRS SocSk .27 .06 .12 .74** .11 .01 .23 .36 .23 .05 NA
Note: Intercorrelations for participants with AD/HD (n=136) are presented above the diagonal, and
intercorrelations for normal control participants (n=53) are presented below the diagonal.
ACES Math=ACES Math score; ACES AcEn=ACES Academic Enablers score; BOSS AET=BOSS Active
Engaged percentage; BOSS PET=BOSS Passive Engaged percentage; ADHD HI=ADHD Rating Scale-IV
HyperactivityImpulsivity score; ADHD IN=ADHD Rating Scale-IV Inattention score; BOSS OT=BOSS Off-
task percentage; BASC Cond=BASC Conduct Problems score; BOSS Nonc=BOSS Noncompliance percentage;
SSRS SocSk=SSRS Social Skills score.
a
Pointbiserial correlations are reported for ethnicity.
b
Correlations for BOSS Nonc are not reported for Control participants because the mean percentage for this
variable was equal to 0.
* pb.01.
** pb.001.
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 295

Table 4
Regression analysis summary for prediction of math achievement
Criterion measure Group Predictor(s) b R2
WJ-III broad math Control Ethnicity .33
SES .01 .12
AD/HD Ethnicity .10
SES .04 .01
ACES Math Skills* .33 .11
Math report card grade Control Ethnicity* .41
SES .13 .14
AD/HD Ethnicity .08
SES .22 .04
ACES Math Skills* .33 .15
Note: ACES=Academic Competency Evaluation Scales; WJ-III=WoodcockJohnson Tests of Achievement.
* pb.05.

Hierarchical regression analyses for reading achievement

Table 5 presents correlations between reading achievement predictor variables for the
AD/HD (above the diagonal) and Normal Control (below the diagonal) groups. Using the
WJ-III Broad Reading standard score as the criterion variable, separate multiple regression
analyses were conducted for the Normal Control and AD/HD groups using the predictor
entry strategy described above. For the control participants, a prediction model including

Table 5
Intercorrelations among predictor variables in reading for AD/HD and normal control participants
Measure 1 2 3 4 5 6 7 8 9 10 11 12
1. Ethnicitya .19 .07 .06 .08 .08 .04 .06 .10 .05 .10 .17
2. SES .41* .04 .19 .14 .21 .07 .06 .05 .10 .08 .12
3. ACES R/LA .46* .17 .35** .08 .12 .04 .28* .08 .06 .16 .09
4. ACES AcEn .31 .27 .37 .01 .08 .27* .41** .07 .37** .14 .56**
5. BOSS AET .06 .12 .00 .04 .53** .02 .02 .22 .05 .03 .03
6. BOSS PET .24 .04 .21 .24 .80** .16 .00 .04 .05 .57** .05
7. ADHD HI .00 .00 .06 .28 .08 .04 .21 .17 .57** .17 .32*
8. ADHD IN .15 .16 .29 .54** .06 .08 .24 .02 .23 .05 .29*
9. BOSS OT .20 .01 .15 .02 .02 .53** .12 .06 .06 .03 .06
10. BASC Cond .07 .18 .07 .05 .07 .02 .08 .23 .18 .03 .40**
11. BOSS Nonc .27 .02 .06 .31 .16 .04 .10 .00 .11 .01 .15
12. SSRS SocSk .27 .06 .20 .74** .18 .23 .23 .36 .13 .05 .35
Note. Intercorrelations for participants with AD/HD (n=136) are presented above the diagonal, and
intercorrelations for normal control participants (n=53) are presented below the diagonal.
ACES R/LA=ACES Reading/Language Arts score; ACES AcEn=ACES Academic Enablers score; BOSS
AET=BOSS Active Engaged percentage; BOSS PET=BOSS Passive Engaged percentage; ADHD HI=ADHD
Rating Scale-IV Hyperactivity-Impulsivity score; ADHD IN=ADHD Rating Scale-IV Inattention score; BOSS
OT=BOSS Off-task percentage; BASC Cond=BASC Conduct Problems score; BOSS Nonc=BOSS Non-
compliance percentage; SSRS SocSk=SSRS Social Skills score.
a
Pointbiserial correlations are reported for ethnicity.
* pb.01.
** pb.001.
296 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

Table 6
Regression analysis summary for prediction of reading achievement
Criterion measure Group Predictor(s)* b R2
WJ-III broad reading Control Ethnicity .18
SES .18 .17
ACES R/LA* .41 .28
ACES Acad. Enab.* -.52 .39
BOSS Off-task* .32 .47
AD/HDRS IA* .34 .55
AD/HD Ethnicity .10
SES .05 .02
ACES R/LA* .57 .35
Reading report card grade Control Ethnicity .09
SES .14 .07
ACES R/LA* .57 .32
AD/HD Ethnicity .03
SES .03 .01
ACES R/LA* .33 .19
ACES Acad. Enab.* .26 .24
Note: ACES R/LA=Academic Competency Evaluation Scale Reading/Language Arts score; ACES Acad.
Enab.=Academic Competency Evaluation Scale Academic Enabler score; BOSS=Behavior Observation System
for Schools; AD/HDRS IA=AD/HD Rating Scale-IV Inattention score.
* pb.05.

ethnicity, SES, ACES Reading/Language Arts, ACES Academic Enablers, BOSS Off-
task, and AD/HD RS-IV Inattention was significant ( F(6,28)=5.7, p=.001; R 2=.55).
Statistically significant standardized beta coefficients were found for ACES Reading/
Language Arts, Academic Enablers, BOSS Off-task, and AD/HD RS-IV Inattention (see
Table 6). For the participants with AD/HD, a significant prediction model including
ethnicity, SES, and ACES Reading/Language Arts was obtained ( F(3,57)=10.3, pb.001;
R 2=.35). The only standardized beta coefficient to achieve statistical significance was
ACES Reading/Language Arts (see Table 6).
A similar set of analyses was conducted using reading report card grade as the criterion
variable (see Table 6). For the normal control participants, a prediction model including
ethnicity, SES, and ACES Reading/Language Arts was significant ( F(3,29)=4.6, p=.01;
R 2=.32). The only predictor variable associated with a significant standardized beta
coefficient was ACES Reading/Language Arts (see Table 6). For the AD/HD group, a
model including four predictors (ethnicity, SES, ACES Reading/Language Arts, and
ACES Academic Enablers) was significant ( F(4,56)=4.45, p=.003; R 2=.24). Significant
standardized beta coefficients were obtained for ACES Reading/Language Arts and ACES
Academic Enablers (see Table 6).

Discussion

Although the association between AD/HD and academic underachievement is well

established, a relatively short list of target skills and behaviors has been identified that
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 297

would aid practitioners in the development of interventions for affected students. The
primary purpose of the current investigation was to address this issue by identifying
potential predictors of contemporaneous academic achievement in a sample of children
meeting research criteria for AD/HD.
Once the contributions of ethnicity and SES were considered, only a few of the
prediction variables selected for analysis played a role in predicting academic achieve-
ment, particularly teacher perceptions of academic skills, academic enablers, and
inattentive symptoms, as well as observations of off-task behavior. These findings were
not consistent across the models tested as they differed as a function not only of group
status but also that the academic content area and how achievement in that subject was
defined.
In general, there was better prediction of Reading than Math for both the AD/HD and
control group. Only one of the four regression models in math (for report card grades
among participants with AD/HD) was statistically significant, while all four reading
models were significant. In fact, the predictor set accounted for 3 to 5 times more variance
in WJ-III Broad Reading scores than for WJ-III Broad Math scores. About twice as much
variance was accounted for in reading report card grades than for math report card grades.
More variance may be accounted for in reading achievement because the predictor set may
be more reflective of skills and behaviors that impact reading performance. It also is
possible that reading reflects a narrower set of skills than math and skills in this area may
be more stable over the short-term. Finally, math may be the preferred subject area for
some children with AD/HD, perhaps due to differences in instruction content and type, and
this could lead to differences in classroom behavior across subject areas.
There were notable differences in the prediction of reading achievement between
children with AD/HD and their peers. First, report card grades for the AD/HD group were
predicted by teacher ratings of reading skills and academic enablers. In contrast, the only
significant predictor of reading grades for the control group was teacher rating of reading
skills. This pattern was reversed when the WJ-III Broad Reading score served as the
criterion variable. The ACES R/LA score was a significant predictor for both groups;
however, teacher ratings of academic enablers, observed off-task behavior, and teacher
ratings of inattentive behaviors added to the variance accounted for in the control group. In
fact, the regression model for normal controls accounted for 20% more of the variance of
WJ-III Broad Reading than did the model for the AD/HD group.
One could argue that the regression model for the AD/HD group is limited due to a
presumed restriction in range for predictor variables assessing classroom behavior (i.e.,
because the distribution of scores for participants with AD/HD will be in the extreme end
of the continuum). Alternatively, standard deviations for most of the predictor variables
were actually larger in the AD/HD group than for controls (see Table 2). Furthermore, if
relative restriction in range between groups was a factor, it is not clear why it only appears
to be an issue for prediction of WJ-III Broad Reading (and not for the other three criterion
variables). Another possible explanation for these results may be related to differences in
behavior across assessment settings. Specifically, the WJ-III was individually administered
in a setting separate from the classroom, while behavior observations were conducted in
the classroom. Students with AD/HD often behave very differently in one-to-one settings
(Barkley, 1998) and, therefore, observational data collected in a group setting may not
298 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

account for significant variance in achievement test outcomes. On the other hand, the
behavior of control students is presumably more consistent across settings; thus, classroom
behavior may be more likely to predict achievement test performance.

Advances over prior studies

The current study offers several unique contributions to the extant literature. First,
academic achievement was not defined as a unitary construct, instead, multiple sources of
information across academic content areas were examined. Similarly, the assessment of
child variables was extended beyond the core symptoms of AD/HD as assessed by single
raters. In addition to teacher ratings of the symptoms of AD/HD, teacher ratings of social
skills, academic skills and academic enablers, as well as direct observation of classroom
behaviors were obtained. The latter measures are particularly unique in studies examining
the prediction of academic achievement in students with AD/HD.
Moreover, unique to the current study was the examination of differential prediction
across groups (normals vs. AD/HD). Our data indicate that there may be important
differences in variables that predict academic outcomes in these two groups, particularly
when measuring achievement using report card grades. Finally, although academic
enablers have been implicated in the achievement of normal children (DiPerna, Volpe, &
Elliott, 2001), this is the first study to note the importance of academic enablers in the
prediction of achievement of children with AD/HD. Prior research indicates group
differences, but has not necessarily created a link between these variables and
achievement. At least with regards to reading achievement as assessed via report card
grades, the link between academic enablers and reading performance in youngsters with
AD/HD is clear, even after their symptoms of AD/HD have been considered.

Implications

The findings of the current study have several implications for those seeking to
understand and ameliorate the academic problems of students with AD/HD. The single
most important predictor variable for all four achievement outcomes examined in this
study was teacher perceptions of academic skills. Even when considered in the context of a
host of other putative predictors, math and reading skills emerged as primary predictors.
Thus, the reading and math skills of children with AD/HD must be targeted for
intervention rather than a sole reliance on reducing AD/HD symptoms and related
disruptive behaviors. Furthermore, reading report card grades being predicted by academic
enablers (in addition to reading skills) may indicate that teachers grading is affected or
biased by disruptive behavior or could indicate that enablers are more likely to disrupt
reading performance in the classroom but not necessarily affect skill attainment as
measured by individual achievement tests (which may be more direct measures of skill
level).
The relative roles of academic skills and enablers (as opposed to AD/HD symptoms) in
predicting academic outcomes suggest that perhaps it is time to move beyond the
traditional or medical approach of focusing almost exclusively on AD/HD symptoms
(MTA Cooperative Group, 1999) by also targeting academic skills and enablers for
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 299

intervention. Based on our findings and the results of previous studies (e.g., Fergusson &
Horwood, 1995; Rapport et al., 1999), one would not expect gains in academic
performance by targeting the symptoms of more severe forms of disruptive behavior
disorders (ODD, CD).

Limitations

The findings of the current study should be interpreted with some caution due to several
methodological limitations. First, the size of the control sample was relatively small, which
could limit the reliability of regression analyses and the generalization of our findings to
the population of nonreferred children. A further limitation with regard to sampling was
that there were differences in the ethnic distribution and SES across groups. For this
reason, we attempted to control for the effects of these variables by including them as
predictor variables in all of the regression models. A third issue concerning sampling was
that our AD/HD sample was preselected for academic difficulties. Although one could
argue that this may limit the generalizability of results, it should be noted that most
children with AD/HD (by definition) have academic impairments of one kind or another
(American Psychiatric Association, 2000). Finally, despite the establishment of structured
diagnostic criteria (i.e., DSM-IV; American Psychiatric Association, 2000), there is
considerable heterogeneity in clinical practice and research investigations with respect to
procedures for identifying children with AD/HD. This is particularly true when relatively
arbitrary cutpoints are used to screen participants who are not clinically referred, as was
done in this study. Thus, although we used identification procedures that are consistent
with current research practice, our sample may differ from other samples employed in
prior studies of AD/HD.
Although we conducted separate observations for Math and Reading instruction, the
observations were conducted on only one occasion in each subject area. The use of a cross-
sectional design limits conclusions regarding the relationship between predictor and
criterion variables over time. In addition, these 15-min observation occasions represent
relatively limited samples of behavior.
Finally, no information about the psychometric properties of report card grades is
available, thus questioning the validity of such measures. Despite such criticism, report
card grades are important criteria given that these outcomes are used in schools to make
decisions about childrens achievement on an everyday basis.

Conclusions

Both academic skills and enablers appear to be associated with academic achievement
in children with AD/HD and the nature of predictive relationships may differ from those
found for non-AD/HD peers. Predictors appear to vary as a function of the academic
content area and as how achievement is measured. In particular, childrens academic skills
are strong predictors of standardized achievement test performance, whereas academic
enablers also contribute to predicting report card grades, at least of children with AD/HD.
Despite the examination of multiple predictor variables, the majority of variance in
achievement outcomes was unexplained by these models. Thus, it remains important to
300 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301

evaluate the potential role of additional predictors such as more direct measures of
academic skills (e.g., curriculum-based measurement), academic self-concept (Hoza,
Pelham, Dobbs, Owens, & Pillow, 2002), and parent involvement. It is especially
important to examine the relationship between predictors and achievement in a
longitudinal fashion given the chronic academic difficulties experienced by children with
AD/HD. Further explication of factors that account for the academic difficulties of
students with AD/HD is critical as it is increasingly apparent that assessment and treatment
protocols must go beyond an exclusive focus on AD/HD symptoms as targets for
intervention.

Acknowledgements

The preparation of this manuscript was supported by NIMH Grant R01-MH62941. We

gratefully acknowledge the efforts of all teachers and students who participated in this
project as well as Emily Xin, Hillary Rogers, Karen Hailstones, Eve Puhalla, Cotie Strong,
Lauren Dullum and Erin Post who served as data collectors for this study.

References

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed.-text
revision). Washington, DC7 Author.
Barkley, R. A. (1998). Attention deficit hyperactivity disorder: A handbook for diagnosis and treatment (2nd ed.).
New York7 Guilford.
Barkley, R. A., Fischer, M., Edelbrock, C. S., & Smallish, L. (1990). The adolescent outcome of hyperactive
children diagnosed by research criteria: I. An 8-year prospective follow-up study. Journal of American
Academy of Child and Adolescent Psychiatry, 29, 546 557.
DeShazo Barry, T., Lyman, R. D., & Grofer Klinger, L. (2002). Academic underachievement and attention-
deficit/hyperactivity disorder: The negative impact of symptom severity on school performance. Journal of
School Psychology, 40, 259 283.
Diamond, P. J., & Onwuegbuzie, A. J. (2001). Factors associated with reading achievement and attitudes among
elementary school-aged students. Research in the Schools, 8, 1 11.
DiPerna, J. C., & Elliott, S. N. (2000). Academic competence evaluation scales. San Antonio, TX7 The
Psychological Corporation.
DiPerna, J. C., Volpe, R. J., & Elliott, S. N. (2001). A model of academic enablers and elementary reading/
language arts achievement. School Psychology Review, 31, 298 312.
DuPaul, G. J., Power, T. J., Anastopoulos, A. D., & Reid, R. (1998). ADHD Rating Scale, vol. IV. New York7
Guilford Press.
Fergusson, D. M., & Horwood, L. J. (1995). Early disruptive behavior, IQ, and later school achievement and
delinquent behavior. Journal of Abnormal Child Psychology, 23, 183 199.
Fergusson, D. M., Horwood, L. J., & Lynskey, M. T. (1993). The effects of conduct disorder and attention deficit
in middle childhood on offending and scholastic ability at age 13. Journal of Child Psychology and
Psychiatry, 34, 899 916.
Fergusson, D. M., Lynskey, M. T., & Horwood, L. J. (1997). Attentional difficulties in middle childhood and
psychosocial outcomes in young adulthood. Journal of Child Psychology and Psychiatry, 38, 633 644.
Frick, P. J., Kamphaus, R. W., Lahey, B. B., Loeber, R., Christ, M. A. G., Hart, E. L., & Tannenbaum, L. E.
(1991). Academic underachievement and the disruptive behavior disorders. Journal of Consulting and
Clinical Psychology, 59, 289 294.
 G.J. DuPaul et al. / Journal of School Psychology 42 (2004) 285301 301

Gresham, F. M., & Elliott, S. N. (1990). Social skills rating system. Circle Pines, MN7 American Guidance
Services.
Hinshaw, S. P. (1992). Externalizing behavior problems and academic underachievement in childhood and
adolescence: Causal relationships and underlying mechanisms. Psychological Bulletin, 111, 127 155.
Hollingshead, A. B. (1975). Four factor index of social status. New Haven, CT7 Yale University.
Hoza, B., Pelham, W. E., Dobbs, J., Owens, J. S., & Pillow, D. R. (2002). Do boys with attention-deficit/
hyperactivity disorder have positive illusory self-concepts? Journal of Abnormal Psychology, 111, 268 278.
Jimerson, S., Egeland, B., & Teo, A. (1999). A longitudinal study of achievement trajectories: Factors associated
with change. Journal of Educational Psychology, 91, 116 126.
Mannuzza, S., Gittelman-Klein, R., Bessler, A., Malloy, P., & LaPadula, M. (1993). Adult outcome of hyperactive
boys: Educational achievement, occupational rank, and psychiatric status. Archives of General Psychiatry, 50,
565 576.
Mather, N., & Woodcock, R. W. (2001). Examiners manual: WoodcockJohnson III tests of achievement. Itasca,
IL7 Riverside Publishing.
MTA Cooperative Group. (1999). A 14-month randomized clinical trial of treatment strategies for attention-
deficit/hyperactivity disorder. Archives of General Psychiatry, 56, 1073 1086.
Rapport, M. D., Scanlan, S. W., & Denney, C. B. (1999). Attention-deficit/hyperactivity disorder and scholastic
achievement: A model of dual developmental pathways. Journal of Child Psychology and Psychiatry, 40,
1169 1183.
Reynolds, C. R., & Kamphaus, R. W. (1992). Behavior assessment system for children. Circle Pines, MN7
American Guidance Service.
Reynolds, C. R., & Kamphaus, R. W. (1998). Behavior assessment system for children: Manual. Circle Pines,
MN7 American Guidance Service.
Shaffer, D., Fisher, P., & Lucas (1998). Computerized diagnostic interview schedule for children (CDISC 4.0).
New York7 New York State Psychiatric Institute.
Shapiro, E. S. (1996). Academic skills problems: Direct assessment and intervention (2nd ed.). New York7
Guilford Publications.
Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). WoodcockJohnson III tests of achievement. Itasca, IL7
Riverside Publishing.
Zentall, S. S., Smith, Y. N., Lee, Y. B., & Wieczorek, C. (1994). Mathematical outcomes of attention-deficit
hyperactivity disorder. Journal of Learning Disabilities, 27, 510 519.