This article was downloaded by: [UAA/APU Consortium Library]

On: 30 October 2014, At: 09:16

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Child Neuropsychology: A Journal on

Normal and Abnormal Development in
Childhood and Adolescence
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/ncny20

Confirmatory Factor Analysis of the

Behavior Rating Inventory of Executive
Function (BRIEF): Support for a
distinction between Emotional and
Behavioral Regulation
a a
Jens Egeland & Øyvind Fallmyr
a
Vestfold Mental Health Care Trust , Tønsberg, Norway
Published online: 05 Mar 2010.

To cite this article: Jens Egeland & Øyvind Fallmyr (2010) Confirmatory Factor Analysis of the
Behavior Rating Inventory of Executive Function (BRIEF): Support for a distinction between Emotional
and Behavioral Regulation, Child Neuropsychology: A Journal on Normal and Abnormal Development in
Childhood and Adolescence, 16:4, 326-337, DOI: 10.1080/09297041003601462

To link to this article: http://dx.doi.org/10.1080/09297041003601462

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014
 Child Neuropsychology, 16: 326–337, 2010
http://www.psypress.com/childneuropsych
ISSN: 0929-7049 print / 1744-4136 online
DOI: 10.1080/09297041003601462

CONFIRMATORY FACTOR ANALYSIS OF THE BEHAVIOR

RATING INVENTORY OF EXECUTIVE FUNCTION (BRIEF):
SUPPORT FOR A DISTINCTION BETWEEN EMOTIONAL
AND BEHAVIORAL REGULATION

Jens Egeland and Øyvind Fallmyr

Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

Vestfold Mental Health Care Trust, Tønsberg, Norway

Previous research has supported a three-factor division of the Behavior Rating Inventory of
Executive Function (BRIEF) when dividing the parent form in 9 instead of 8 subscales.
The present study investigated different factor models in the 8- and 9-scale division in both
the parent and teacher form of the Norwegian BRIEF version. Confirmatory Factor
Analyzes showed best fit for the three-factor model in a mixed healthy and clinical sample,
indicating a distinction between Emotional and Behavioral Regulation. This division is in
accordance with present knowledge of brain function and may increase the specificity of
executive dysfunction in clinical groups.

Keywords: BRIEF; Behavior Rating Inventory of Executive Function; Emotional Regulation;

Hot executive functions; CFA.

INTRODUCTION
Impaired executive functioning is central to several neuropsychiatric disorders of
childhood and is prevalent in for instance attention deficit/hyperactivity disorder (ADHD),
autism, conduct disorder, and Tourette Syndrome (Pennington & Ozonoff, 1996). The
concept of executive dysfunction consists of several quite different symptoms, such as
perseverations, impulsivity, lack of initiative, impersistance, intrusions of task-irrelevant
behavior, or inflexibility. It encompasses subfunctions of attention, such as working
memory and controlled attention, and problem solving and is considered independent of
intelligence. Overall, executive functions are supervisory functions modifying the output
of some other cognitive process. Traditionally, it is the neuropsychological domain that
has been most difficult to measure validly using laboratory tests, since such settings often
are too structured to uncover problems with self-organization (Salimpoor & Desrocher,
2006). Although impaired executive functioning in terms of impaired attention and
hyperactivity/impulsivity is a prerequisite of an ADHD diagnosis, several patients with
ADHD perform normal on neuropsychological tests (Egeland, in press). By the same
token, subjects with nonexecutive dysfunctions, such as impaired reading ability, may
score as if they were hyperactive-impulsive (McGee, Clark, & Symons, 2000).

Address correspondence to Jens Egeland, Department of Research, Vestfold Mental Health Care Trust,
Box 2267, N-3103 Tønsberg, Norway. E-mail: [email protected]

© 2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
 CFA OF BRIEF 327

One way of increasing the sensitivity and specificity of test and laboratory measures
has been to deconstruct global concepts such as attention and executive function into
subprocesses such as vigilance, sustained attention, and change in mental control (Egeland &
Kowalik-Gran, 2010). This is also concordant with new progress in research in cognition
and how the brain functions.
However, questionnaires gathering information from parents and teachers have typi-
cally been composed of items sampling problematic behavior with less emphasis on under-
lying cognitive processes. The Behavior Rating Inventory of Executive Function (BRIEF;
Gioya, Isquith, Guy, & Kenworthy, 2000) represents an attempt to obtain more detailed
information about the subprocesses of executive function. Exploratory factor analyses of
the original eight subscale division of the parent and teacher forms of BRIEF showed the
same two-factor solution in both normal controls and clinical subjects (Gioia et al., 2000).
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

However, a later reorganizing of the test into nine subscales has instead found evidence of
three underlying factors or dimensions in executive function (Gioia, Isquith, Retzlaff, &
Espy, 2002). In this latter study, the Monitor scale was divided into two separate subscales.
The later published adult version of the BRIEF (BRIEF-A; Roth, Isquith, & Gioia, 2005)
has incorporated this nine-scale division. Nevertheless, the parent and teacher versions,
applicable for children between 5 and 18 years of age, still applies the original eight-scale
division. It is still not settled whether a nine-scale division is superior to an eight-scale
division, and whether a two- or three-factor solution fits the data best. The Gioia et al.
(2002) study analyzed only the parent form and did not compare the eight- and nine-scale
division. In the present study, we analyze both forms, testing different factor models both
when dividing the Monitor scale into Task-Monitoring and Self-Monitoring (i.e. both an
eight-scale and a nine-scale division). Dividing the Monitor scale into two small scales,
consisting of only five and four items, could potentially compromise the reliability of the
scales. On the other hand, if such a division increases the test’s validity as a measure of
three distinct subprocesses of executive function, this may nevertheless be advantageous.
The original eight scales of BRIEF clustered together in a behavioral regulation
(BR) and a metacognition (MC) factor, giving credibility to a corresponding Behavior
Regulation Index (BRI) and a Metacognition Index (MCI) score. The MCI was considered
to measure the ability to use working memory to initiate, to plan, to organize, and to main-
tain future-oriented problem solving, while the BRI was thought to assess the ability to use
appropriate inhibitory control to shift cognitive set and to modulate emotions and behavior.
Recent research with subjects with autism spectrum disorders (Gilotty, Kenworthy, Sirian,
Black, & Wagner, 2002), myelomenigocele and hydrocepahalus (Mahone, Zabel, Levey,
Verda, & Kinsman, 2002), and ADHD (McCandless & O’Laughlin, 2007) has shown the
usefulness of the index scores successfully differentiating clinical subjects from healthy
controls and yielding diagnosis specific profiles differentiating the type of executive defi-
cit characteristic of different disorders. However, a closer look at the intercorrelations
between subtests has shown some problems regarding encompassing the Monitor subscale
in the MCI. In the Gilotty et al. study (2002) the Monitor subscale correlated higher with BRI
than did the other MCI subtests. In the ADHD study (McCandless & O’Laughlin) the
teacher Monitor scale in fact showed a higher correlation with BRI than with the MCI of
which it is a part. In the parent form, the Monitor scale correlated highly with both index
scores. This was in contrast to the other scales constituting the MCI, which correlated
higher with that index score than with the BRI. Another study using the BRIEF on ADHD
patients showed a higher correlation between teacher ratings on the Monitor scale and the
parent BRI, than with the parent MCI (Jarratt, Riccio, & Siekierski, 2005). These findings
 328 J. EGELAND AND Ø. FALLMYR

may indicate that even though the MCI may be a sensitive measure of a specific aspect of exec-
utive function, the Monitor scale does not fit as well into the two-factor division as the other
subtests. This is exactly what Slick, Lautzenhiser, Sherman, and Eyrl (2006) found in their fac-
tor analysis of the BRIEF in a group of young subjects with intractable epilepsy (i.e., that the
Monitor scale loaded equivalently on both the MCI and BRI). Thus, the suggestion by Gioia et
al. (2002) that the Monitor scale in fact consists of items related to two dimensions, namely
monitoring of task-related activities and monitoring of personal-behavior activities, could
underlie the psychometric problems related to this scale. Given the above mentioned problems
with encompassing the Monitor scale into the MCI factor, it is plausible that dividing the scale
could give a more robust factor structure. However, this is an empirical question, and the data
regarding the superiority of a nine-scale division and a three-factor solution is still scarce.
In the three-factor solution that Gioia et al. (2002) found in the parent form, the new
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

Self-Monitor scale and the three subtests loading on the original BRI factor split into
separate Emotional and Behavioral regulation factors. The third factor, Metacognition,
was thus deprived of the self-monitor items but retained the Task-Monitor scale, compared
to the original factor structure.
Testing the factor-structure of a translated test attempting to measure different
subprocesses of executive function is crucial to the validity of the index scores. The find-
ings will be relevant not only for potential users of the new Norwegian version but also, in
general, as the findings will reflect back on the still unsettled issue of the factor structure
of the original version.
In the present study three-factor models are compared: The One-Factor Model
presupposes the unity of executive functions (i.e., such high intercorrelations between the
BRIEF subscales that a division into stable subfunctions is not advisable). The Two-Factor
Model is the original division into MC and BR factors taken from the manual of the test.
In the three-factor model of the original eight subscales we place the BR Factor together
with the Inhibit subscale to comprise the Behavior Regulation factor, whereas the Shift
and the Emotional Control subscales constitute the Emotional Regulation factor.
In the two-factor model of the nine-scale partition of the scale, the Self-Monitor scale is
placed within the BR factor, while Task-Monitor is placed within the MC factor. The three-
factor model is the one tested by Gioia et al. (2002) for the parent form and is presented for the
teacher form in Figure 1. The metacognition factor in this model is comprised of the Initiate,
Working Memory, Plan/Organize, Organization of Materials, and Task-Monitor subscales.
The Emotional Regulation factor is comprised of the Shift and the Emotional Control sub-
scales, whereas the behavior regulation factor consists of the Self-Monitor and Inhibit scales.
Finding evidence of a division between emotional and behavioral regulation may
increase the utility of the BRIEF even further, as it may be linked to the differentiation between
“cold” executive and “hot” emotional processing. Brain-imaging studies have identified two
dissociable prefrontal brain systems for attention and emotion, the so-called dorsal and the ven-
tral systems (Dolcos & McCarthy, 2006). The theory of dual mechanisms of control asserts
that a proactive control system is responsible for selection of attentional focus in a top-down
fashion, whereas a reactive control system is more sensitive to emotionally salient bottom-up
input and may distract the subject from a preselected path (Fales et al., 2008). Whether a divi-
sion between emotional and behavioral regulation in the BRIEF could be linked to the hot/cold
dichotomy and the dual mechanisms of control must of course be investigated. However, this
is plausible if the factor structure of the test supports a three-factor model, which at face value
distinguishes between task-related and emotional-mental control. In her discussion of the rela-
tionship between behavior rating scales and laboratory assessments of executive function,
 CFA OF BRIEF 329
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

Figure 1 CFA BRIEF Teacher Form: Three-Factor Model based on nine subscales. Path diagram for the Three-
Factor Model, showing standardized regression weights and covariances between factors. WorkingM = Working
Memory; PlanOrga = Plan/Organize; Material = Organization of Materials; Task Moni = Task Monitor;
Emotiona = Emotional Control; SelfMoni = Self-Monitor; MetaCogn = Metacognition; EmotRegu = Emotional
Regulation; BehavReg = Behaviour Regulation.

Denckla (2002) emphasizes the fact that the three-factor structure in the Gioia et al. (2002)
study is closer to present knowledge of brain function, as it differentiates inhibitory behaviors
from more emotional behaviors. However, the empirical basis for this differentiation remains
scarce, as no replication of the factor study has been conducted prior to the present study. The
validity of the hot and cold executive function dichotomy will also depend on whether this dif-
ferentiation can be found across different developmental contexts. The parent and the teacher
forms of the BRIEF give us information about two such salient contexts.
 330 J. EGELAND AND Ø. FALLMYR

METHODS
Subjects
One hundred and fifty-eight children participated in the study. Forty-eight
fourth grade children (23 boys and 25 girls) were recruited from three primary schools
in the municipality of Tønsberg, Norway, and constituted a healthy control group. The
clinical sample consisted of 72 children referred to the school psychology service in
Tønsberg and 38 outpatients assessed in two Child Psychiatric Clinics of Vestfold
Mental Health Care Trust. The clinical group was mixed with regard to cause of
referral. Analysis of group differences on the BRIEF (Fallmyr & Egeland, in press)
indicated that the level of impaired executive functions was equivalent to that
reported in the mixed clinical sample used in the confirmatory factor analysis (CFA)
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

of Gioia et al. (2002). There were significantly more boys than girls in the clinical
sample (86 boys and 26 girls).
The healthy controls were not asked for their date of birth, so the mean age is estimated
to 10 years varying from 9½ and 10½ years according to their class level. The average age
of the clinical sample was 10.9 years (SD = 2.6).
Information letters describing the research project were distributed to all students by
the teacher. After the return of a signed written consent from the parents of the children,
parents and teachers received the BRIEF questionnaire. The return ratio from the healthy
controls was 49%, whereas almost all clinical subjects agreed to let their BRIEF protocol
be analyzed in the project. The Regional Committee for Medical Research Ethics and the
Norwegian Social Sciences Services had approved the study.

Instruments
The BRIEF is a questionnaire consisting of 86 items comprising eight clinical
scales. In the standard version, the Behavior Regulation Index score (BRI) is computed
from the Inhibit, Shift, and Emotional Control subtests. The Metacognition Index (MCI) is
based on the scores from the Initiate, Working Memory, Plan/Organize, Organization of
Materials, and Monitor subtests. The Global Executive Composite is a summary score that
incorporates all eight clinical scales.
The Norwegian version of the BRIEF was translated with permission from the
publisher Psychological Assessment Resources by psychologists Øyvind Fallmyr, Katrine
Ekerholt, and Jude Nicholas. It is used with the original American norms. Analyses of
within-scale consistency, correlations between the parent and the teacher form, based on
the present sample, were similar to the original version (Fallmyr & Egeland, in press). The
control group in the present study scored similarly to the original norms on the parent
form but somewhat below norms on the teacher form. There were no significant sex dif-
ferences in the present sample. Overall, the psychometric abilities of the Norwegian ver-
sion were considered satisfactory for clinical use in Norway.

Data Analyses
The goodness-of-fit measures for three different factor models are tested both with
the eight-scale and the-nine scale partition of both the parent and teacher forms of the
BRIEF. The factor models are referred in the introduction. We applied the LISREL 8.3
 CFA OF BRIEF 331

program (Jöreskog & Sörbom, 1993). The goodness-of-fit measures should be interpreted
as follows (Gregoire, 2004; Keith, 2005):
• Chi square (c2) – When applying chi-square in comparing the hypothesized and the
observed model, a low value indicates a good fit.
• c2 /df – If the chi-squares for two models are the same, the more constrained or more
parsimonious model is preferred. Parsimony is reflected in CFA models by df. Thus, the
smallest c2 in comparison to df represents the best fit of the data.
• Goodness-of-Fit Index (GFI) and Adjusted Goodness-of-Fit Index (AGFI). These measures
show how much better the model fits as compared to no model at all. The AGFI is
adjusted for degrees of freedom. Values can vary between zero and one. Values above
.90 indicate a good model fit.
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

• Comparative Fit Index (CFI) – Values of .95 or better suggest a good fit of the model to
the data, and values above .90 suggesting an adequate fit.
• Root mean square error of approximation (RMSEA) is a measure of approximate fit.
Smaller values suggest a better fit, with values of .06 or smaller suggesting a good fit
and those of approximately .08 suggesting an adequate fit.

RESULTS
Table 1 provides the scores of the healthy controls and the clinical sample on each
scale and factor index scores in both the eight- and nine-scale partition of the test for
both the teacher and parent forms. The clinical sample differed from the healthy

Table 1 BRIEF Scale Means and Standard Deviations, Healthy Controls (HC) and Clinical Sample (Clinical).

Parent Form Teacher Form

HC Clinical HC Clinical

M (SD) M (SD) M (SD) M (SD)

Inhibit 1.4 (.41) 1.7 (.52) 1.2 (.29) 1.9 (.63)

Shift 1.5 (.43) 1.7 (.52) 1.1 (.18) 1.8 (.52)
Emotional Control 1.6 (.51) 1.8 (.55) 1.1 (.28) 1.8 (.59)
BRI1 1.5 (.40) 1.8 (.45) 1.1 (.23) 1.8 (.51)
Initiate 1.5 (.31) 1.8 (.44) 1.3 (.38) 2.1 (.52)
Working Memory 1.5 (.42) 2.2 (.42) 1.1 (.33) 2.2 (.51)
Plan/Organize 1.6 (.39) 2.0 (.45) 1.2 (.31) 2.1 (.52)
Organization of Materials 1.9 (.47) 2.1 (.55)* 1.2 (.42) 1.8 (.62)
Monitor 1.6 (.40) 2.0 (.44) 1.3 (.35) 2.1 (.52)
Task-monitor 1.7 (.42) 2.1 (.52) 1.3 (.40) 2.1 (.59)
Self-monitor 1.6 (.58) 1.9 (.60) 1.2 (.39) 2.0 (.62)
MCI2 1.6 (.31) 2.0 (.36) 1.2 (.31) 2.0 (.43)
GEC3 1.6 (.32) 1.9 (.35) 1.2 (.27) 2.0 (.43)
MC94 1.6 (.28) 2.0 (.35) 1.2 (.31) 2.1 (.45)
ER94 1.8 (.49) 2.0 (.36) 1.1. (.23) 1.8 (.52)
BR94 1.5 (4.6) 1.8 (.51) 1.2 (.33) 2.0 (.60)

*The difference between HC and Clinical Sample significant at .05 level. All other comparisons: p < .001.
1
BRI = Behavior Regulation Index.
2
MCI = Metacognition Index.
3
GEC = Global Executive Composite.
4
Metacognition, Emotional Regulation, and Behavior Regulation index scores based on the nine-scale division.
 332 J. EGELAND AND Ø. FALLMYR

controls on all scales. Visually inspected, the differences between the Healthy Controls
(HC) and clinical subjects were greater for the teacher compared to the parent form.
Table 2 shows the correlation matrix between subtests. All correlations were significant.
The correlations between the five scales loading on the original MC factor and the MC
index in the eight-scale division ranged from .80–.93 in the teacher form and from .74–.87
in the parent form. For the nine-scale division the equivalent correlations with metacog-
niton (MC9) ranged from .87–.95 and from .67–.86, respectively. All three subscales
making up the BRI correlated .91 with that index score in the teacher form, whereas the
correlations ranged from .84–.91 in the parent form. In the teacher form, the subscales
Shift and Emotional Control correlated .95 with the Emotional Regulation (ER9), and
.93 and .95 in the nine-scale division of the parent form. The Self-Monitor and Inhibit
scales both correlated .97 with the Behavior Regulation index (BR9) in the teacher form
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

of the nine-scale division and .91 and .94 with BR9 in the parent form. The correlation
between the Self- and Task-Monitor subscales was low (.29) for the parent form and
moderate for the teacher form (.63). Nevertheless, this correlation was the lowest corre-
lation between Self-Monitor and any other subscale.
Since all correlations are significant, the question arises whether a division into
factors or index scores is at all relevant and, if so, which factor model fits the data best?
The results of the comparisons between the different factor models are presented in
Tables 3 to 6.
Table 3 shows the goodness-of-fit statistics for the different models for the parent
form. The one-factor model did not satisfy any of the criteria for an adequate or good fit
between the model and the observed results. The two- and three-factor models had an
adequate level of CFI, were close to an adequate level of GFI, but the AGFI and the
RMSEA were not satisfactory. The c2/df values showed the superiority of the original
two-factor model.
The results of the CFA for the teacher form are presented in Table 4. Again, the one-
factor model did not satisfy any of the fit criteria. The GFI was not satisfactory for the
two- and three-factor models, although approaching an adequate level for the three-factor
model. The CFI showed an adequate fit for the two-factor model and a good fit for the
three-factor model. The AGFI and RMSEA were below limits for adequate fit between
models and data. All fit indices were somewhat better for the three-factor model compared
to the two-factor model. The c2/df values showed that the three-factor model represented
the most parsimonious model.
Tables 5 and 6 present the fit indices for the nine-scale partition for parents and
teachers, respectively. None of the fit indices were satisfactory for the one-factor
model. All fit indices were better for the three-factor model than the two-factor model.
For the parent form all values were better than the comparable values for the best
model of the eight-scale partition. For the parents the GFI showed an adequate fit for
the three-factor model, while the CFI showed a satisfactory fit for the two-factor model
and a good fit for the three-factor model. For the teacher form, the GFI approached an
adequate level for the three-factor model. The AGFI and RMSEA were below satisfac-
tory levels for both models in both the parent and teacher forms. The c2/df clearly
showed that the three-factor model was superior to the two-factor model, by represent-
ing the most parsimonious solution in both the parent and teacher forms. The superior
three-factor model of the parent form is presented in Figure 1. The regression weight of
each subtest on their respective factors appears along the arrows linking the factors and
subtests.
 Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

Table 2 Correlation Matrix Between Subscales, Index Scores in the Standard Eight-Scale and the New Nine-Scale Division as well as the Global Executive Composite (Teacher
Form Above the Diagonal; Parent Form Below the Diagonal).

Inhibit Shift Emot.C. BRI Initiate WM Plan/org. Organ. of Mat. Monitor Task-mon. Self-mon. MCI GEC MC9 ER9 BR9

Inhibit .72 .76 .91 .55 .61 .64 .56 .85 .57 .89 .72 .85 .66 .77 .97
Shift .59 .82 .91 .62 .60 .68 .49 .70 .54 .69 .67 .82 .64 .95 .72
Emotional C. .64 .75 .91 .62 .60 .68 .49 .70 .54 .69 .67 .82 .64 .95 .72
BRI .84 .88 .91 .60 .61 .67 .47 .80 .57 .82 .71 .88 .66 .97 .89
Initiate .47 .63 .50 .61 .85 .84 .54 .72 .70 .63 .89 .84 .89 .58 .61
WM .50 .52 .42 .54 .62 .86 .65 .74 .71 .63 .93 .87 .92 .56 .64

333
Plan/organize .38 .48 .37 .46 .56 .75 .70 .80 .80 .66 .95 .91 .95 .64 .67
Org.of Mat. .36 .35 .33 .39 .52 .48 .50 .66 .65 .56 .80 .73 .80 .37 .57
Monitor .65 .53 .54 .65 .45 .62 .67 .41 .85 .94 .88 .92 .85 .71 .92
Task-monitor .28 .20 .21 .26 .28 .46 .57 .23 .77 .63 .83 .79 .87 .52 .62
Self-monitor .73 .62 .62 .75 .43 .52 .51 .41 .83 .29 .77 .85 .71 .71 .97
MCI .59 .62 .53 .65 .77 .87. .86 .74 .78 .57 .67 .96 .99 .64 .77
GEC .76 .80 .75 .87 .77 .80 .76 .65 .80 .49 .77 .94 .94 .82 .87
MC9 .52 .55 .47 .58 .75 .86 .87 .71 .76 .67 .56 .98 .90 .60 .70
ER9 .65 .93 .95 .96 .60 .49 .45 .36 .57 .22 .66 .61 .83 .71 .76
BR9 .91 .65 .68 .85 .48 .55 .48 .42 .81 .30 .94 .68 .83 .58 .58
 334 J. EGELAND AND Ø. FALLMYR

Table 3 Parent Ratings of the Eight-Scale Version and Goodness of Fit Indexes for
Three Hypothesized BRIEF Models (N = 152).

Model c2 (df) c2 /df GFI AGFI CFI RMSEA

One Factor 182.66 9.13 0.77 0.58 0.87 0.23

Two Factors 85.52 4.50 0.88 0.76 0.93 0.15
Three Factors 86.21 5.07 0.88 0.74 0.94 0.16

Table 4 Teacher Ratings of the Eight-Scale Version and Goodness of Fit Indexes for
Three Hypothesized BRIEF Models (N = 152).
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

Model c2 (df) c2 /df GFI AGFI CFI RMSEA

One Factor 329.54 16.48 0.63 0.34 0.84 0.33

Two Factors 145.82 7.67 0.79 0.61 0.91 0.22
Three Factors 96.05 5.65 0.85 0.69 0.95 0.18

Table 5 Parent Ratings of the Nine-Scale Version and Goodness of Fit Indexes for
Three Hypothesized BRIEF Models (N = 152).

Model c2 (df) c2 /df GFI AGFI CFI RMSEA

One Factor 242.13 8.97 0.74 0.56 0.86 0.23

Two Factors 106.16 4.08 0.86 0.77 0.94 0.12
Three Factors 78.18 3.26 0.90 0.81 0.96 0.14

Table 6 Teacher Ratings of the Nine-Scale Version and Goodness of Fit Indexes for
Three Hypothesized BRIEF Models (N = 152).

Model c2 (df) c2 /df GFI AGFI CFI RMSEA

One Factor 376.80 13.95 0.63 0.38 0.84 0.30

Two Factors 150.35 5.78 0.81 0.67 0.94 0.18
Three Factors 94.78 3.66 0.89 0.79 0.97 0.13

DISCUSSION
The study corroborates Gioia et al.’s (2002) finding of a three-factor solution for the
nine-scale partition of the BRIEF. The present study extends the Gioia et al. study by finding
that the three-factor model represents the best fit to the data also for the teacher form,
which was not examined in that study. By analyzing the different factor models in both the
eight- and nine-scale divisions, the present study gives the first empirical evidence of the
superiority of three-factor model based on the nine-scale division compared to the two-factor
solution based on eight scales, as applied in the current manual for the parent and teacher
forms. In fact, even in the eight-scale division, the three-factor model is superior to the
original two-factor model for the teacher form. However, even the best model only had a
moderately good fit to the data. Even though correlations between all subscales are signif-
icant, the CFAs of both the eight- and nine-scale partitions of both forms clearly warn
against considering executive functions as one homogeneous construct. The descriptive
 CFA OF BRIEF 335

statistics also show that the new Norwegian BRIEF version successfully differentiates
between normal controls and clinical subjects. Thus, also this translated version of the
BRIEF seems to measure subfunctions of executive function in ways that, at least on a
group level, are sensitive to presumed deficits.
The study indicates that the nine-scale partition of the test should be preferred to the
eight-scale partition, and that the results should be analyzed as measuring three distinct
aspects of executive function. Theoretically, the division between emotional and behav-
ioral regulation are interesting, but in need of further research in order to substantiate the
differential validity of such a division. Also, this division needs to be anchored to research
on brain function and neuropsychological assessment.
As mentioned above, the present study extends the findings of Gioia et al. (2002).
How do the results relate to other studies? The exploratory factor analysis of Slick et al.
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

(2006) only compared one- and two-factor solutions within the eight-scale partition, and
only in the parent form. They found that the two-factor model best fit the data but
commented that the Monitor scale loaded similarly on both factors. Including a three-fac-
tor model in our CFA in the eight-scale partition of the test and also in the teacher version,
as we do here, questions the validity of the division into BRI and MCI also in that original
partition into eight scales. The term “questions the validity” is an intentionally chosen
understatement, as the exploratory factor analyses presented in the test manual (Gioia
et al., 2000) gave support for the two-factor model in the parent form both in the norma-
tive sample and in a mixed clinical sample. Thus, the issue is not yet settled. Nevertheless,
the sum of the present evidence seems to favor a nine-scale division of the test: As
mentioned in the introduction, the correlation analyses of clinical groups typically show
that the Monitor scale correlates with both index scores. In addition to this, the fit indices
of the nine-scale partition is higher than in the eight-scale division, even when constraining
the model more by adding the ninth scale.
What are the possible implications of the nine-scale partition and the three-factor
model? Intuitively, monitoring instrumental task and social performance are quite differ-
ent. Much of the criticism of the low sensitivity of neuropsychological testing of executive
function points to the fact that neuropsychological testing often offers a strict structure
that prevents the planning and monitoring deficits evident in daily life to surface (Gioia &
Isquith, 2004). Additionally, even tests that with little extrinsic structure are nevertheless
“cold” in the sense of dealing with affect-neutral problem solving, such as the Wisconsin
Card Sorting Test. Monitoring one’s own behavior in a social context is more influenced
by emotions, which is not assessed in the neuropsychological examination. Possibly, the
monitoring of cold problem-solving behavior would correlate higher with neuropsycho-
logical tests than would warm emotional behavior, which instead must be tested in context
or rated by observers. Traditionally, executive functions have been locked into the “frontal
metaphor” (Pennington & Ozonoff, 1996). Central to the umbrella of executive functions is
the notion of modifying or supervising otherwise modular cognitive processes. Today it is
clear that this modulation is both cognitive and affective and includes both prefrontal as well
as limbic activity (Phelps, 2006). Thus, while being psychometrically sound as the CFA
shows, the separate measurement of emotional and behavioral regulation also fits well
with the present knowledge of the limitations of laboratory testing as well as knowledge of
brain function. Whether the division also increases the ecological validity and discrimina-
tory power between clinical groups remains to be tested in future research.
There are some limitations in the present study: Only about half of the children in
the fourth grade classes that were asked to participate actually took part in the study. There
 336 J. EGELAND AND Ø. FALLMYR

is a risk of a selection bias towards the best functioning children. Nonetheless, the scores
of the present healthy sample corresponded well with the original American norms for the
parent version, while the scores for the teacher version were somewhat below the original
norms (Fallmyr & Egeland, in press).
Another limitation concerns the limited experimental control due to lack of detailed
information about diagnosis and severity of symptoms of the clinical sample. However, as
there are no between-group comparisons in the present CFA part of the study, the lack of
descriptive data is not crucial.
Also the combination of clinical subjects and healthy controls within the same CFA
can be criticized. Most CFAs are performed either on normative samples, i.e., healthy
controls, or clinical samples. However, the factor structure of a normal group may not
correspond to impairment patterns in clinical groups. Small variability in a homogeneous
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

group may falsely lead to a unidimensional factor due to normal scores on all scales. By the
same token, small variability on a particular subscale among clinical groups may make that
scale difficult to place within a factor structure, even though the variability in a mixed
healthy and normal sample is large. Working Memory deficits are, for instance, highly sen-
sitive when differentiating healthy controls from clinical groups such as autism (Gilotty et
al., 2002 ), specific language impairment (Montgomery, 2003), state anxiety (Hadwin, Bro-
gan, & Stevenson, 2005), and ADHD (Slick et al., 2006). Since impaired working memory
is so prevalent in clinical groups, variability, which is the raw material of factor analyses,
would be low in clinical samples. Thus, this ecologically important dimension of executive
function could be overlooked. Performing a CFA with a mixed clinical and healthy sample,
thus, has the advantage of maximum variability. An alternative would be to perform CFAs
in the normal and clinical groups separately to see if the same factor structure emerges, but
the size of the present subsamples prohibited this alternative.
Taking the methodological limitations into account, this study lends support to the
multifactorial approach to executive function inherent in the BRIEF but suggests that the
partition into nine scales and three factors may improve the inventory.

Original manuscript received April 17, 2009

Revised manuscript accepted December 29, 2009
First published online March 5, 2010

REFERENCES
Denckla, M. B. (2002). The Behavior Rating Inventory of Executive Function: Commentary. Child
Neuropsychology, 8, 304–306.
Dolcos, F., & McCarthy, G. (2006). Brain systems mediating cognitive interference by emotional
distraction. The Journal of Neuroscience, 26, 2072–2079.
Egeland, J. (in press). Frequency of attention deficit in first episode schizophrenia compared to
ADHD. Applied Neuropsychology.
Egeland, J. & Kowalik-Gran, I. (2010) Measuring several aspects of attention in one test: The
factorstructure of Conners’ Continuous Performance Test. Journal of Attention Disorders, 13,
339–347.
Fales, C. L., Burgess, G. C., Schafer, A., Mennin, D. S., Gray, J. R., & Braver, T. S. (2008). Anxiety
and cognitive efficiency: Differential modulation of transient and sustained neural activity
during a working memory task. Cognitive, Affective, & Behvioral Neuroscience, 8, 239–253.
Fallmyr, Ø., & Egeland, J. (in press). Psykometriske egenskaper for den norske versjonen av spør-
reskjemaet BRIEF (Behavior Inventory of Executive Functioning) [Psychometric properties of the
Norwegian version of BRIEF]. Tidsskrift for Norsk Psykologforening.
 CFA OF BRIEF 337

Gilotty, L., Kenworthy, L., Sirian, L., Black, D. O., & Wagner, A. E. (2002). Adaptive skills and
executive function in autism spectrum disorders. Child Neuropsychology, 8, 241–248.
Gioya, G. A., & Isquith, P. K.(2004). Ecological assessment of executive function in traumatic brain
injury. Developmental Neuropsychology, 25, 135–158.
Gioya, G. A., Isquith, P. K., Guy, S. C., & Kenworthy, L. (2000). BRIEF: Behavior Rating Inven-
tory of Executive Function, professional manual. Lutz, Florida: Psychological Assessment
Resources.
Gioya, G. A., Isquith, P. K., Retzlaff, P. D., & Espy, K. A. (2002). Confirmatory factor analysis of
the Behavior Rating Inventory of Executive Function (BRIEF) in a clinical sample. Child
Neuropsychology, 8, 249–257.
Gregoire, J. (2004). Factor structure of the French version of the Wechsler Adult Intelligence Scale-III.
Educational and Psychological Measurement, 64, 463–474.
Hadwin, J. A., Brogan, J., & Stevenson, J. (2005). State anxiety and working memory in children:
Downloaded by [UAA/APU Consortium Library] at 09:16 30 October 2014

A test of processing efficiency theory. Educational Psychology, 25, 379–393.

Jarratt, K. P., Riccio, C. A., & Siekierski, B. M. (2005). Assessment of Attention Deficit Hyperactivity
Disorder (ADHD) using the BASC and BRIEF. Applied Neuropsychology, 12, 83–93.
Jöreskog, K., & Sörbom, D. (1993). LISREL 8: Structural equation modeling with the SIMPLIS
command language. Lincolnwood, IL: Scientific Software International.
Keith, T. Z. (2005). Using confirmatory factor analysis to aid in understanding the constructs
measured by intelligence tests. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary
intellectual assessment (pp. 581–614). New York: Guildford Press.
Mahone, E. M., Zabel, T. A., Levey, E., Verda, M., & Kinsman, S. (2002). Parent and self-report
ratings of executive function in adolescents with myelomeningocele and hydrocephalus. Child
Neuropsychology, 8, 258–270.
McCandless, S., & O’Laughlin, L. (2007). The clinical utility of the Behavior Rating Inventory of
Executive function (BRIEF) in the diagnosis of ADHD. Journal of Attention Disorders, 10,
381–389.
McGee, R. A., Clark, S. E., & Symons, D. K. (2000). Does the Conners’ CPT aid in ADHD Diagnosis?
Journal of Abnormal Child Psychology, 28, 415–424.
Montgomery, J. W. (2003). Working memory and comprehension in children with specific language
impairment: What we know so far. Journal of Communication Disorders, 36, 221–231.
Pennington, B. F. & Ozonoff, S. (1996). Executive functions and developmental psychopathology.
Journal of Child Psychology and Psychiatry, 37, 51–87.
Phelps, Elizabeth A. (2006). Emotion and cognition: Insights from studies of the human amygdala.
Annual Review of Psychology, 57, 27–53.
Roth, R. M., Isquith, P. K., & Gioia, G. A. (2005). Behavior Rating Inventory of Executive Function-
Adult Version (BRIEF-A). Lutz, FL: Psychological Assessment Resources, Inc.
Salimpoor, V. N., & Desrocher, M. (2006). Increasing the utility of EF assessment of executive
function in children. Developmental Disabilities Bulletin, 34, 15–42.
Slick, D., Lautzenhiser, A., Sherman, E. M. S., & Eyrl, K. (2006). Frequency of scale elevations and
factor structure of the Behavior Rating Inventory of Executive function (BRIEF) in children
and adolescents with intractable epilepsy. Child Neuropsychology, 12, 181–186.