ChIA

children’s
inventory of
Anger
Manual
W. Michael Nelson III, Ph.D., ABPP, and A. J. Finch, Jr., Ph.D., ABPP

Published by

W-344B
wps ®
Publishers Distributors
 1
Introduction

Self-report symptom inventories have a long history meaning of each item was kept intact. These features, as
in the assessment of various psychological disorders in both well as updated norms representing a much broader range of
adults and children. These tools are easy to administer, eco- children, further exploration of the statistical properties of
nomical, and readily analyzable. As is typically the case in the ChIA, and the development of a user-friendly Manual,
psychology, assessment instruments and clinical services to all make the ChIA an excellent choice for professionals who
children and adolescents follow the development of tests and are interested in assessing anger in youngsters.
interventions for adults. In contrast to the wide availability A more detailed description of the development of the
of self-report symptom scales in many areas for adults, there ChIA and its psychometric properties can be found in chap-
have been relatively few corresponding instruments for ters 4 and 5 of this Manual.
youngsters. The Children’s Inventory of Anger (ChIA) con-
tinues the tradition of the self-report measurement of anger Description of the ChIA
in children that was established over 20 years ago with the
initial development of the instrument (then referred to as the With only 39 items, the ChIA takes about 10–15 min-
CIA) in 1978. At that time, the test was developed by Nelson utes to administer. The items are very easy to read for
and Finch in response not only to the need for a child as- youngsters with at least a third-grade reading ability. The
sessment instrument to measure anger, but also to the hap- test administrator can read the items aloud to younger chil-
hazard fashion with which assessment and treatment of dren or those with reading difficulties. Item responses are
anger in children had been explored. given on a 4-point Likert scale. Each numerical choice on
Basically, the CIA was designed as a children’s coun- the scale is associated with a pictorial representation. The
terpart to the Novaco Anger Inventory (NAI; Novaco, 1975, response choices and their associated pictures are displayed
in press). Like Novaco in his development of the NAI, the in Figure 1.
authors considered that a thorough anger-related assessment In addition to a Total score, the ChIA yields an
should address all of the important parameters of anger re- Inconsistent Responding validity index (INC) and four sub-
actions, including the mode of expression and the effect on scale scores covering four different areas that might lead to
the personal relationship, as well as the frequency, intensity, an angry response in youngsters: Frustration (FRUST),
and duration of the anger response. As was true at that time, Physical Aggression (PHYS), Peer Relationships (PEER),
few instruments are available to assess any of these parame- and Authority Relations (AUTH). ChIA results are present-
ters in children. Thus the 1978 CIA was developed to con- ed as age-stratified standard scores based on a nationally
tribute to such a comprehensive assessment by providing a representative sample of 1,604 youngsters, ages 6–16.
self-report means to assess the intensity of anger experi- A complete description of administration and scoring
enced in response to a variety of potential anger-arousing procedures for the ChIA can be found in chapter 2 of this
situations. Manual. Guidelines for interpreting the test scores are pre-
In the course of this instrument being prepared for sented in chapter 3. A detailed description of the standard-
widespread use, the inventory was shortened from 71 items ization sample is provided in chapter 4, and the ChIA’s
to 39, and the acronym was changed to the ChIA (the test’s psychometric properties are described in chapter 5.
full name remains the same). The Total score obtained using
the current version of the ChIA is equivalent to that of the Purposes and Uses
earlier version, but an Inconsistent Responding validity
index and four subscale scores can now also be obtained. The ChIA is an efficient and cost-effective screening
The wording of some items was modified to make them procedure. Because administration, scoring, and interpreta-
more readily understood by youngsters, although the core tion do not require much time, the ChIA should prove useful

1
2 Children’s Inventory of Anger (ChIA)

in situations where a quick, routine assessment of a young-

ster’s anger problems is required. Principles of Use
The ChIA measures various aspects of a youngster’s
experience of anger. As such, it can be used as part of a The use of this test, as well as any other psychological
routine screening in a number of settings, including instrument, presupposes familiarity with this Manual and
schools, special education in regular classrooms, in-patient with basic principles of tests and measurements. This
and out-patient clinics, residential treatment centers, juve- knowledge involves an understanding of psychological in-
nile detention centers, and private practice offices. Because terpretation and the limitations of psychological testing. The
it is easily and quickly administered to children in groups, test itself can be administered and scored by technicians and
it can be used as a screening device to identify those who trained paraprofessionals. Nevertheless, only those with pro-
might benefit from further psychological evaluations in the fessional training can interpret the test results properly.
school system, in hospital settings, by large managed care Thus, the ultimate responsibility for interpretation must be
groups, or by other agencies. Potential test users include assumed by an individual who has an appropriate under-
clinical/counseling psychologists, social workers, school standing of the limitations of such screening and testing pro-
psychologists, physicians, counselors, psychiatric nurses, cedures. Users should also take precautions to safeguard the
pediatric nurses, teachers, remedial and special education confidentiality of test results, restricting their access to those
personnel, school officials, and/or pastoral counselors. with the ability to use the information appropriately.
It should be noted that the ChIA is not designed to re- In general, an individual whose only exposure to test-
place clinical diagnosis. It is useful both as a screening tool ing is gained from this Manual is not considered a qualified
to identify children for further evaluation and as a source of user of the ChIA. Qualified users of this test should be mem-
additional detailed information in the context of a thorough, bers of professional associations that endorse a set of stan-
multifaceted clinical assessment. In addition, because it is dards for ethical use of psychological or educational tests, or
designed to measure the subjective state of anger, the ChIA licensed professionals in the area of psychology, education,
can be a useful supplement to treatment planning and to medicine, social work, or an allied field.
evaluating changes in the experience of anger over time. Although the ChIA is designed for use across a wide
The ChIA can also be used to evaluate psychoeduca- age range and a variety of contexts, it is not intended to be
tional and clinical interventions designed to promote greater the only instrument in clinical assessment or a substitute for
self-management of anger. Self-reported changes in the ex- sound clinical judgement that is based upon various sources
perience of anger in specific general areas of experience of information about the youngster. The test cannot be used
(using ChIA subscale scores) or situations (using individual as a substitute for direct clinical observation of the child or
ChIA items) may be assessed. Integration of results from the ratings of important others (e.g., parents, teachers) regarding
ChIA with information from other sources about a young- overt aggressive behavior. It is also important for profes-
ster’s acting out (for example, parent and teacher ratings or sionals to consider any factors which may bias the child into
clinical observations) can result in a more comprehensive inaccurate reporting of his/her experience of anger (e.g., de-
view of the youngster. nial of responsibility, oppositionality, low verbal ability).

1 2 3 4

I don’t care. That bothers me, I’m really angry I can’t stand that!
That situation doesn’t but I’m not too angry or mad, but I I’m furious! I feel like
even bother me. I don’t or mad about it. I’ll think I can really hurting or killing
know why that would just forget it. control myself. that person, or
make anyone angry destroying that thing!
or mad.

Figure 1
The Four ChIA Response Choices and Their Associated Pictures
 4
Development and Standardization

In 1978, the original 71-item Children’s Inventory of children and adolescents. A self-report format was chosen,
Anger (then referred to as the CIA) was designed as an ob- as opposed to the observer ratings commonly used with chil-
jective self-report measure to identify a range of anger dren, because it seemed that anger, as an internal experience,
provocations, to identify individuals whom a child might was most directly measurable in this way. Pictorial represen-
view as a provocateur of his or her aggressive behavior, and tations of the Likert-type response choices were employed
to pinpoint events that evoke the anger response for a given based on previous research (Finch & Montgomery, 1973;
child. It was hoped that this information could be employed Stein, Finch, Hooke, Montgomery, & Nelson, 1975) show-
by researchers and clinicians to bridge the gap between the ing that impulsive/aggressive children are more likely to
assessment and treatment of anger problems in children. think in pictures than in words. Thus, the numbered response
Assessment was viewed as an initial step in the formulation scale coupled with the picture aids and explanatory state-
of a comprehensive cognitive-behavioral treatment regimen ments provide assistance to help children better recognize
that would be systematically applicable to anger manage- and anchor their ratings.
ment in children. Such multifaceted treatment programs The CIA was a self-report measure consisting of 71
have been shown to be an effective means to facilitate better statements describing provocation incidents. The child was
anger management in youth (e.g., Beck & Fernandez, 1998; asked to rate each of these items on a 4-point scale, indicat-
Nelson, 1998; Nelson & Finch, 1996, in press). ing the degree of anger he or she would experience if the in-
The present version of the ChIA is a result of more cident described by the item had actually happened. The
than 20 years of use in research and clinical settings. This choices on the scale represented increasing intensities of
chapter describes the development of the instrument and the anger from “I don’t care. That situation doesn’t even bother
standardization sample from which its normative data were me. I don’t know why that would make anyone mad
derived. (angry)” to “ I can’t stand that! I’m furious! I feel like really
hurting or killing that person, or destroying that thing!”
Development and Use of the CIA Each test item was designed to provide contextual informa-
(Early Version of the ChIA) tion by specifying the setting, people, or external events in
the provocation incident being described. The test items
In 1978, the authors saw that there was a significant were determined to be at the fourth-grade reading level. The
need for the development of an instrument to assess the sub- test was designed to be self-administered and had no time
jective feeling of anger in children. At that time, the con- limits, although it typically required 20 to 25 minutes to
cepts of anger and violence were being increasingly complete in either an individual or group format. The instru-
explored in the psychological literature (e.g., Bandura, ment yielded a Total score, but subscales for the test were
1973; Johnson, 1972). New developments in psychothera- not established.
peutic interventions in the area of self-control and stress- Through 20 years of use in research and clinical set-
inoculation (e.g., Meichenbaum, 1977) were being tings, the authors were able to amass a significant amount of
published. Novaco (1975) proposed a self-control program knowledge regarding the practical features of this inventory,
to deal with adult anger and aggression, and had developed and to address some of the problems that researchers and
the Novaco Anger Inventory (NAI) to assess the construct clinicians encountered in its use. One of the primary draw-
of anger in adults. It was at this time that the authors con- backs of the CIA was its length, which was often particular-
structed the original 71-item version of the Children’s ly problematic for youngsters with emotional or behavioral
Inventory of Anger (CIA; Finch, Saylor, & Nelson, 1987; difficulties. In addition, the wording of some of the items
Nelson & Finch, 1978). was awkward. These were some of the issues addressed in
The CIA was developed by generating a range of prob- the process of refining the CIA, which resulted in the present
lem situations that were thought to evoke angry responses in 39-item ChIA.

25
26 Children’s Inventory of Anger (ChIA)

Development of the ChIA The fact made apparent by both of these studies, in spite of
in its Present Form their conflicting results, is that the 1978 CIA was a multidi-
mensional instrument both for normal children and for those
In developing the current version of the ChIA, the who had been diagnosed with emotional disturbance.
original 71 items were evaluated for their readability, as well Furthermore, three of the six factors for the nonclinical
as their contribution to the distinct factors forming the sub- group were very similar to two of the derived factors in the
scales of the test. Items that did not clearly relate to a single clinical sample. Specifically, the first two factors for the
subscale were dropped. nonclinical group (“At the mercy of authority figures,” and
Some of the remaining items were slightly reworded “At the mercy of uncontrollable events”), are closely related
to eliminate passive constructions and shorten sentences. to the Mayhall et al. study’s Factor 2 (“Imposed events”) for
These changes were minor. For example, Item 1 was the psychiatric hospitalization group. The third factor for the
changed from “You are right in the middle of your favorite nonclinical group (“Injustice”) was closely related to Factor
TV show and your mother calls you to dinner” to “Your 3 (“Perceived Injustices”) in the Mayhall et al. study. Results
mother calls you to dinner in the middle of your favorite TV from these two separate factor analyses were encouraging
show.” Item 4 was changed from “You clean up your room because the statistically derived factors tended to make clin-
on Saturday and want to go out to play, but your mom says ical sense. Probably the most powerful reason to retain a fac-
you have to clean out your drawers and closet, too” to “You tor in a factor analytic study is its actual clinical utility.
clean up your room and want to go out to play. Your mom Subscales for the 39-item ChIA were derived based on
says you have to clean some more.” the results of additional factor analyses, rational inspection
In this way, the readability of the item set was reduced of item groupings, and inspection of corrected item-to-scale
from a fourth-grade level to a third-grade level. A reading correlations for individual items. Table 4 presents the inter-
ease index, based on a scale of 0–100 with higher numbers scale correlations of the resulting subscales, as observed for
representing easier reading, was raised from 78 to 91, plac- various subgroupings of the ChIA standardization sample
ing the test in the Very Easy range. Not surprisingly, the cor- (described later in this chapter). The relationships among the
relation between the Total scores for the 71-item CIA and scales are consistent across the subgroups represented in the
the present 39-item ChIA is .99. table. Moreover, although correlations among the subscales
are somewhat high, they are still substantially lower than the
The ChIA Subscales and correlations between subscale scores and the ChIA Total
Inconsistent Responding Index score, thus providing justification for separate scoring and
interpretation of the subscales.
Several factor analytic studies were conducted to Further support for subscale scoring and interpretation
examine the subscale structure of the ChIA. In the first fac- is apparent in the results of the final principal components
tor analytic study done with the 1978 version of the instru- factor analysis for the full standardization sample, repre-
ment, the 71-item CIA, the responses of 1,000 children in sented in Table 5. As can be seen in that table, the 39 items
the fourth through eighth grades were subjected to a princi- on the ChIA form four clear factors, accounting for 45% of
pal component factor analysis with a varimax rotation. Six the variance in item responses in the sample. These factors
factors with eigenvalues greater than one accounted for 75% correspond well to the four subscales of the test. The factor
of the total variance. These factors were reviewed indepen- structure did not vary for gender or age-based subgroupings
dently by a group of six mental health professionals who of subjects. Goodness-of-fit index values (Jöreskog &
worked with children. The reviewers described the six fac- Sörbom, 1984) indicated an exact fit (1.0) when the factor
tors in the following way: loadings represented in Table 5 were applied to ChIA item
1. At the mercy of authority figures responses for only males or only females from the sample.
2. At the mercy of uncontrollable events The same results were observed when the structure was ap-
3. Injustice plied to item responses for the youngest (6–8) and oldest
4. Embarrassment or threat to self-esteem (14–16) children in the sample. Further support for the use
5. Frustration of desires of the ChIA subscales, in the form of reliability and validity
6. Sibling conflict evidence, is presented in chapter 5.
In a second factor analysis (Mayhall, Nelson, Politano, & The Inconsistent Responding (INC) index for the
Wendall, 1986), the 1978 CIA was administered to 242 chil- ChIA was constructed by selecting pairs of items for which
dren (ages 6 to 18) hospitalized for emotional disturbance. responses were found to be highly correlated. Table 6 dis-
Four factors, which accounted for 41.6% of the variance, plays the correlations for these item pairs from the standard-
were described as follows: ization sample data. In the standardization sample, it was
1. Loss of possessions fairly unusual for more than five of these item pairs to re-
2. Imposed events with a loss of control or ceive responses that differed by more than one point. The
helplessness average number of item pairs differing by more than a point
3. Perceived injustices was 1.82 (SD = 1.69) in the standardization sample, as com-
4. Loss through accidental destruction of an object pared with an average of 4.33 such pairs (SD = 1.62) in a set
 4 Development and Standardization 27

Table 4 of 200 randomly generated ChIA protocols. Comparison of

Interscale Correlations Observed in actual ChIA protocols with randomly generated responses
the ChIA Standardization Sample revealed that when five or more INC item pairs differ by
more than a point, there is an 84% likelihood that the re-
FRUST PHYS PEER AUTH sponses were randomly generated. This likelihood increases
a to 93% for six or more INC item pairs, 97% for seven item
Full Sample
pairs, 98% for eight item pairs, and 99% for nine item pairs.
FRUST -
PHYS .58 - The ChIA Standardization Sample
PEER .73 .67 -
AUTH .69 .66 .72 -
The demographics of the ChIA standardization sam-
ple are presented in Table 7. The sample consists of 1,604
TOTAL .87 .84 .89 .88
children, ages 6 through 16 (M = 10.8 years, SD = 3.1 years),
Boysb spanning grades 1 through 11. The children were drawn
FRUST - from research sites across the country that were primarily
public school settings in both urban and rural areas, al-
PHYS .57 -
though a few private schools were also included. Thus, the
PEER .75 .67 - current ChIA sample provides a more stable comparison
AUTH .71 .67 .73 - base than was offered by the 1978 normative group for the
TOTAL .87 .83 .90 .90 CIA, which included only 679 normal public school chil-
Girls c dren in Virginia. That sample represented a cross section of
socioeconomic groups, but only included children in grades
FRUST - four through seven (ages 9–14).
PHYS .59 - The current standardization sample included equiva-
PEER .71 .68 - lent numbers of males and females. The ethnic backgrounds
AUTH .66 .64 .70 - that were reported included 63% White, 8% Black, 6%
TOTAL .86 .84 .89 .87
Hispanic, 3% Asian, and 3% “Other.” For 17% of the stan-
dardization sample, ethnic background was not indicated.
d
Ages 6–8 Socioeconomic status was also determined for most of the
FRUST - standardization sample by asking subjects to specify the edu-
PHYS .62 - cational level of the “head of the household.” As can be seen
in Table 7, the head of household educational levels reported
PEER .69 .72 -
by the sample were 26% with a high school education or less,
AUTH .74 .68 .74 - 22% college graduates, 21% with some postgraduate experi-
TOTAL .87 .85 .89 .90 ence, and 12% with at least some college experience. Head of
Ages 9–13e household educational level was not indicated for 18% of the
standardization sample. This sample is generally consistent
FRUST -
with US population demographics, although it includes a
PHYS .56 - somewhat higher proportion of individuals with college and
PEER .70 .66 - postgraduate educational experience, thereby representing a
AUTH .69 .65 .70 - slightly higher than average socioeconomic level.
TOTAL .86 .83 .88 .88
The Effects of Demographic
Ages 14–15f
Characteristics on ChIA Scores
FRUST -
PHYS .45 - The average ChIA raw scores obtained by the stan-
PEER .66 .54 - dardization sample are displayed in Table 8. Average scores
did not differ based on region or socioeconomic status. To
AUTH .67 .64 .73 -
examine the effects of other demographic characteristics,
TOTAL .81 .80 .86 .91 such as gender, age, and ethnic background on ChIA scores,
a
N = 1,604. all scores were first converted to T-scores based on these
b
N = 796. averages for the full sample. In this way, because all
c
N = 785. T-scores have a mean of 50 and a standard deviation of 10,
d
N = 459. it is easy to make comparisons between subgroups and to
e
N = 762. discern a common metric called effect size. Effect sizes are
f
N = 383. usually expressed in deviation units. An effect size of .5, for
example, means that a score for the groups being compared
 Table 5
Factor Structure of ChIA Responses for the Full Standardization Sample

Factor

I II III IV
Frustration (FRUST)
1. Your mother calls you to dinner in the middle of your favorite TV show. .48 .37
2. Your bike has a flat tire. .54
3. Your brother or sister or friend ignores you. .41
13. Somebody calls you a “chicken.” .57 .33
14. You want to show someone a new trick on your bike and you can’t do it. .53
15. You put all your change in the soda machine. It takes your money, but you don’t get a drink. .53 .32
25. Someone cuts in front of you in line. .32 .54
26. You brought your favorite candy bar in your lunch. When you go to eat it, it’s melted. .63
27. Your mom won’t buy your favorite cereal. .37 .56
28. On your bike, you come to a steep hill. You have to get off the bike and walk it all the way up the hill. .59
38. You have to do your homework and your brother or sister gets to watch TV. .37 .50
Physical Aggression (PHYS)
10. You are playing a game and someone on the other side tries to punch you. .61
36. Somebody punches you. .68
12. You get a piece of pie at lunch and someone knocks it out of your hand. .37 .44
35. Two bigger kids come and take your basketball and play “Keep Away” with it. .31 .57
23. Someone puts gum on your seat on the bus and you sit on it. .58
24. You bump into a stranger on the bus. He says he will beat you up if you get near him again. .32 .57
22. Your mom or dad slaps you. .57 .34
11. Someone spits at you. .68
37. Someone tries to trip you on purpose. .67
Peer Relationships (PEER)
7. The teacher’s pet gets to do all the fun jobs in class. .45 .37
8. In a game, someone on the other side tries to cheat. .53 .33
9. Someone bumps your desk on purpose and you mess up your work. .45 .34 .33
19. Your brother or sister says “No” when you ask for something. .59 .41
20. Your friends are playing a game and they won’t let you play. .55
21. Somebody says “I told you so” after something goes wrong. .62
32. You do something special for a friend and he or she won’t do anything for you. .55
33. Your friends pick you last to be on a team. .62
34. Your friend gets what he or she wants for Christmas, but you don’t. .55
Authority Relations (AUTH)
4. You clean up your room and want to go out to play. Your mom says you have to clean some more. .63
5. You have to do a job that your brother or sister was supposed to do. .30 .54
6. You want to go somewhere with a friend. Your dad says “No” for no reason. .55
16. A teacher gives you a lot of homework on the weekend. .56
17. Someone says that you are not old enough to do something. .33 .40
18. Someone turns the TV to another channel when you are watching a show. .34 .33 .42
29. Your mom says she does not want you to play with one of your friends. .44 .54
30. Your dad yells at you in front of other people. .33 .30 .50
31. You don’t have any homework, but your mom makes you study anyway. .41 .51
39. You have to go to bed at 9:30 and your friends get to stay up until 10:30 or 11:00. .43 .42

% Total Variance (45) 33 5 4 3

28
 4 Development and Standardization 29

Table 6
Correlations for the Inconsistent Responding (INC) Index Item Pairs

r
7. The teacher’s pet gets to do all the fun jobs in class.
38. You have to do your homework and your brother or sister gets to watch TV. .43

8. In a game, someone on the other side tries to cheat.

19. Your brother or sister says “No” when you ask for something. .45

9. Someone bumps your desk on purpose and you mess up your work.
12. You get a piece of pie at lunch and someone knocks it out of your hand. .43

10. You are playing a game and someone on the other side tries to punch you.
36. Somebody punches you. .44

11. Someone spits at you.

23. Someone puts gum on your seat on the bus and you sit on it. .44

17. Someone says that you are not old enough to do something.
18. Someone turns the TV to another channel when you are watching a show. .41

20. Your friends are playing a game and they won’t let you play.
33. Your friends pick you last to be on a team. .42

21. Somebody says “I told you so” after something goes wrong.
32. You do something special for a friend and he or she won’t do anything for you. .41

24. You bump into a stranger on the bus. He says he will beat you up if you get near him again.
37. Someone tries to trip you on purpose. .48

25. Someone cuts in front of you in line.

27. Your mom won’t buy your favorite cereal. .41

29. Your mom says she does not want you to play with one of your friends.
30. Your dad yells at you in front of other people. .45

31. You don’t have any homework, but your mom makes you study anyway.
39. You have to go to bed at 9:30 and your friends get to stay up until 10:30 or 11:00. .42

differs by one-half of a standard deviation. Therefore, an ef- used by the clinician who is trying to decide whether a
fect size of .5 is a difference of 5 T-score units between the T-score of, for example, 58T should be interpreted different-
scores for two groups. ly for a boy or for a girl. For practical purposes, the score
will be interpreted similarly, even if it is adjusted up or down
Meaningful Differences by 1⁄3 point to account for gender differences.
When scores obtained by large samples are compared Effect sizes are used to evaluate whether a statistically
using simple analysis of variance techniques, it very often significant difference is also likely to be a clinically mean-
happens that reliable differences identified between group ingful one (Cohen, 1992; Horst, Tallmadge & Wood, 1975).
scores are statistically significant simply because large sam- In general, effect sizes of .1 to .3 deviation units (or 1 to 3
ples were used, whereas the actual differences are quite T-score points) are considered small and not of much practi-
small—in some cases, a fraction of a scale score point. Such cal import, effect sizes between .3 and .5 deviation units (or
statistically significant differences are not very meaningful 3 to 5 T-score points) are considered moderate, and an effect
for clinical purposes. Knowing, for example, that boys reli- size above .5 deviation units (or 5 T-score points) is consid-
ably obtain test scores that are 1⁄3 point higher than those ob- ered large. Moderate to large effect sizes indicate clinically
tained by girls does not offer any information that can be meaningful differences, particularly when observed across
30 Children’s Inventory of Anger (ChIA)

Table 7 Table 8
Demographic Characteristics of Average Age and ChIA Raw Scores
the ChIA Standardization Sample in the Standardization Sample

N % M SD
Age Age 10.84 3.05
6 94 6
7 189 11
Frustration (FRUST) 22.22 6.50
8 176 11
Physical Aggression (PHYS) 26.62 6.00
9 180 11
Peer Relationships (PEER) 20.33 6.04
10 140 9
Authority Relations (AUTH) 25.15 6.41
11 141 9
Total Score 94.32 21.69
12 142 9
13 159 10
14 110 7
15 157 10
16 116 7 a variety of scores in a pattern that is consistent with other
knowledge about the comparisons being made.
Gender
Males 796 50 Gender
Average ChIA T-scores for males and females in the
Females 785 49
ChIA standardization sample are displayed in Table 9. All
Missing 23 1 but one of the scores (PEER) are significantly different from
Ethnicity a statistical standpoint, with girls consistently obtaining
lower scores than boys. The effect size in every case, how-
Asian 43 3 ever, is small and does not indicate that these differences
Black 133 8 ought to bear strong influence on the interpretation of ChIA
Hispanic 95 6 scores. These findings are consistent with findings for the
ChIA’s predecessor, where an effect size of .22 was ob-
Native American 12 1 served in a comparison of boys’ and girls’ CIA Total scores.
White 1,003 63
Other 47 3 Ethnic Background
Sample subgroupings of children with Native
Missing 271 17 American (n = 12), Asian (n = 43), and Other (n = 47) eth-
Head of Household Educational Level nic backgrounds were considered too small to provide use-
ful information with regard to this moderator variable. For
Did not graduate from high school 79 5
those groups with large enough representation to be useful
High school graduate 337 21 in this regard (Black, n = 133; Hispanic, n = 95; and White,
Some college 196 12 n = 1,003), average T-scores are displayed in Table 10.
When the effect of age is taken into account, none of the av-
College graduate 345 22
erage T-scores represent either clinically or statistically sig-
Postgraduate 332 21 nificant differences from the T-scores obtained by the full
Missing 315 18 standardization sample. These findings echo those for the
1978 CIA, where Total scores for children reporting White
Region and Black ethnic backgrounds were virtually identical.
Northeast 262 16
Age
Midwest 603 38
Of all the demographic characteristics examined in rela-
South 452 28 tion to ChIA scores, age had by far the strongest effect. Average
West 287 18 T-scores for each age level represented in the ChIA standard-
ization sample are displayed in Table 11. Average scores for
Note. N = 1,604. younger and older children vary considerably from those ob-
tained by children in the middle age range. For children ages 6
 4 Development and Standardization 31

through 8, 7/15 or 47% of the average T-scores shown in the purpose of obtaining and profiling ChIA standard scores.
table represent a moderate effect size, larger than .30, in com- Average T-scores for these three combined groups are included
parison with the full standardization sample. For youngsters in Table 11.
ages 14 through 16, 14/15 or 93% of the average T-scores indi- The scores displayed in Table 12 demonstrate the ef-
cate a moderate effect size. These results are consistent with the fectiveness of this age grouping strategy. These scores are
common sense notion that 6-year-olds’ reports about their average T-scores for each age level in the standardization
anger are likely to be different from reports obtained from 16- sample, based on the means and standard deviations for the
year-olds. Based on these observations, it was decided to strat- appropriate combined age group. The effect sizes indicated
ify the standardization sample into three combined age by differences in these average T-scores are negligible, and
groups—6 through 8, 9 through 13, and 14 through 16—for the none is in the moderate range.

Table 9 Table 10
Average T-Scores for Males and Females Average ChIA T-Scores for Various
in the ChIA Standardization Sample Ethnic Groups in the Standardization Sample

Males Females Effect Size Blacka Hispanicb Whitec

FRUST 50.9 49.1* .18 FRUST 50.8 54.0 49.7
PHYS 50.9 49.1* .18 PHYS 50.9 50.8 49.8
PEER 50.2 49.7 - PEER 50.4 53.1 49.9
AUTH 50.7 49.2* .15 AUTH 49.0 49.9 49.6
TOTAL 50.8 49.2* .16
TOTAL 50.3 52.3 49.7
*p < .05. a
n = 133.
b
n = 95.
c
n = 1,003.

Table 11
Average ChIA T-Scores for Individual and Combined Age Groups Based on the Full Standardization Sample

6 7 8 6–8 9 10 11 12 13 9–13 14 15 16 14–16

n 94 189 176 459 180 140 141 142 159 762 110 157 116 383

FRUST 56.3 55.4 54.0 55.0 52.1 51.1 49.4 48.2 47.5 49.7 45.1 44.2 45.3 44.8
PHYS 52.6 52.5 52.0 52.4 50.9 50.4 49.8 50.6 50.8 50.5 47.6 45.3 46.4 46.3
PEER 55.1 54.3 52.7 53.9 50.8 50.0 50.0 50.5 48.9 50.1 46.1 44.7 45.1 45.3
AUTH 51.0 50.4 50.8 50.7 50.3 51.6 50.8 51.1 51.4 51.0 48.4 46.3 46.7 47.0
TOTAL 54.5 53.6 52.7 53.5 51.2 50.9 50.0 50.1 49.5 50.4 46.4 44.3 45.2 45.2

Table 12
Average ChIA T-Scores for Each Age Level Based on the Appropriate Combined Age Group

6 7 8 9 10 11 12 13 14 15 16
n 94 189 176 180 140 141 142 159 110 157 116

FRUST 51.9 50.4 48.6 52.5 51.5 49.7 48.5 47.7 50.4 49.2 50.3
PHYS 50.3 50.3 49.5 50.5 49.8 49.2 50.0 50.2 51.4 49.0 50.0
PEER 51.3 50.6 48.9 50.8 50.0 50.0 50.6 48.8 51.2 49.5 49.7
AUTH 50.2 49.8 50.0 49.3 50.7 49.8 50.1 50.3 51.7 49.3 49.5
TOTAL 51.2 50.3 49.0 50.8 50.7 49.6 49.7 49.1 51.7 49.2 49.7

Note. The three combined age groups are 6–8, 9–13, and 14–16.
 5
Psychometric Properties

This chapter discusses the reliability and validity of internal consistency estimates in Table 13. Estimates are
the ChIA. Because the test items are a subset of the ChIA’s reported for a sample of 87 children, ages 6–11, who took
1978 predecessor, the CIA (Finch, Saylor, & Nelson, 1987; the ChIA once and then completed it again after a one-week
Nelson & Finch, 1978), much of the earlier research using interval. Estimates for the 24 children in this sample who
the CIA applies as well to the ChIA, and is included in this were 6 or 7 years old are reported separately.
overview. As shown in Table 13, the test-retest correlation for
the Total score is .75 for the entire sample and .66 for the
Reliability younger children. For the ChIA subscales, test-retest corre-
lations range from .65 to .75 (median = .71) for the full sam-
Reliability addresses the consistency or “dependability” ple and from .48 to .73 (median = .63) for the younger
of the test scores. There are several ways of investigating the children. With the exception of the PEER scale, these values
reliability of psychological tests. Two of the most common indicate adequate test-retest reliability for children of all
methods are to examine the consistency of responses to all ages. Although the sample of 24 subjects is small, these re-
items on a given scale (internal consistency) and to analyze sults suggest that, for younger children, the greatest inter-
the consistency of scores obtained by the same persons on pretive weight should be placed on the ChIA Total score,
different occasions (test-retest). The reliability of ChIA scores which appears to be a very reliable overall estimate of their
was examined for the entire standardization sample and also experience of anger. Also, for younger children, the least in-
separately for children 6 and 7 years of age, because young terpretive weight should be placed on the PEER score as it
children are notoriously unreliable reporters in some areas. seems that, although the younger children’s responses to
items on this scale are quite consistent during a single ad-
Reliability Studies Using the ChIA ministration of the test, they may be subject to unpredictable
Internal consistency. Internal consistency reliability changes within a very short period of time.
addresses whether all items on a given scale measure the
same dimension consistently or reliably. Internal consisten- Reliability Studies Using the 1978 CIA
cy is typically estimated using an overall summary coeffi- The reliability of the 1978 CIA was demonstrated in a
cient called coefficient alpha (Cronbach, 1951). Alpha series of investigations with both normal and clinically re-
coefficient values in the .60 to .70 range are considered ferred populations. For the initial study of internal consisten-
marginal, whereas those above .70 are considered adequate. cy (Finch, Saylor, & Nelson, 1983, 1987), 460 normal school
Internal consistency estimates for the ChIA are presented in children and 61 child psychiatry inpatients served as partici-
Table 13 for the full standardization sample and also for 6- pants (total N = 521). The 256 boys and 265 girls in this sam-
and 7-year-olds in the sample. The alpha coefficient for the ple ranged in age from 9 to 15 years (mean age = 10.9 years).
ChIA Total score is .95 for both groups, which is excellent. There were 302 White and 219 non-White children in the
Subscale alpha coefficients were also excellent and virtually sample.
identical for both groups. These ranged from .85 to .86 for Split-half reliability for the 71-item CIA ranged from
the full sample, and from .85 to .87 for the younger children .83 to .95 depending on the type of split (even/odd versus
in the sample. first half/second half) and statistical procedures employed.
Test-retest reliability. Test-retest reliability provides Using Pearson correlation procedures, the split-half reliabil-
an index of the temporal stability of ChIA responses. It is a ity was .91 for an even-odd split and .83 for a first-second
function of the actual reliability of the child’s behavior and half split. Using Spearman-Brown correlation procedures,
of any extraneous conditions that occur at the time of the reliability was .95 for the even-odd split and .91 for the first-
testing. Values above .60 are considered adequate. Test- second half split. Item-total correlations were also computed
retest estimates for the ChIA are displayed along with the for all 71 items, and ranged from .34 to .59. Sixty-four of

33
34 Children’s Inventory of Anger (ChIA)

Table 13
Cronbach’s Alpha and Test-Retest Reliability
in the ChIA Standardization Sample

Test-Rest
Alpha (1-week intertest interval)
Number
Scale of Items Full Samplea Ages 6–7b Ages 6–11c Ages 6–7d

Frustration (FRUST) 11 .85 .85 .73 .65

Physical Aggression (PHYS) 9 .86 .87 .69 .60
Peer Relationships (PEER) 9 .86 .87 .65 .48
Authority Relations (AUTH) 10 .85 .86 .75 .73
Total Score 39 .95 .95 .75 .66
a
N = 1,604.
b
N = 283.
c
N = 87.
d
N = 24.

the 71 items had correlations greater than .40 with the Total Brown split-half correlation coefficients were .93 for the
score. Finally, a Kuder-Richardson coefficient alpha of .96 first-second half reliability and .96 for odd-even reliability,
confirmed in conjunction with the other findings that the quite similar to the findings of Finch, Saylor, and Nelson
CIA had high internal consistency. (1983). However, Mayhall et al. (1986) found a test-retest
Test-retest reliability of the 1978 CIA was also exam- correlation of only .49 from day 3 of hospitalization to day
ined (Montgomery, Nelson, & Finch, 1979). In this study, 30 30, possibly due to the greater disturbance level of their
child psychiatric inpatients were individually administered sample. In addition, because the patients were in treatment,
the CIA at the beginning and end of a 3-month period. one might expect that they would respond differently, with
Participants were diagnosed primarily with behavior disor- some children improving while others did not. These results
ders based on DSM-II (American Psychiatric Association, indicate the need for further examination of the test-retest
1968). Ages of the children in the sample ranged from 9 years reliability of the instrument with specific clinical samples.
and 0 months to 15 years and 1 month (mean age = 11.4
years). The CIA yielded a 3-month test-retest reliability of .82 Validity
for this group.
Because of its potential use as a repeated measure to The validity of a test instrument is the extent to
examine treatment outcome, the 1978 CIA’s reliability over which a test can be shown to measure what it is supposed
repeated administrations was also studied. Participants were to measure. There is no absolute way of determining
55 boys and 53 girls, ranging in age from 7 to 12 years whether a scale actually measures a theoretical construct
(mean age = 9.6 years). All were enrolled in public schools because, by nature, the construct can never be measured di-
and completed the CIA four times at 2-week intervals in a rectly or perfectly in any fashion. Given that it cannot be
classroom setting, where the items were read aloud to the measured directly, validity must be inferred. The issue of
group. CIA mean raw scores decreased from 205.7 initially validity is particularly difficult in relation to the assess-
to 197.9 at 2 weeks, 190.3 at 4 weeks, and 191.2 at 6 weeks. ment of anger, which is a private experience not readily
Repeated measures analysis of variance confirmed a signif- available to external validation. For example, if a child is
icant decrease over trials, F(3,321) = 12.77, p < .001. Test- very angry, but does not behave in an aggressive fashion,
retest correlations among the trials ranged from .63 to .90 this does not make his or her self-report of anger invalid.
(all significant at p < .001). Overall, these evaluations of Aggressive behavior is only one of the behavioral alterna-
reliability suggested that CIA scores might decrease some- tives available to the individual when the affect of anger is
what over repeated administrations, but that moderate to experienced. Whether the anger will be expressed as ag-
high levels of agreement could be expected from one ad- gression depends on such factors as social expectations,
ministration to the next. environmental demands, self-control, past learning histo-
In another reliability study, Mayhall et al. (1986) ad- ry, and the stimulus source of the anger. With these diffi-
ministered the 1978 CIA individually to a group of 242 culties in mind, a series of studies were conducted to
emotionally disturbed children at 3 and 30 days following investigate the discriminant validity of both the ChIA and
admission to an inpatient psychiatric hospital. Spearman- its predecessor, the 1978 CIA.
 5 Psychometric Properties 35

The validity of a psychological test is typically assessed of the primary techniques used to examine construct validity.
by posing several different questions: As reported in chapter 4, the factor structure for the ChIA
• Does the test have interpretable item content corresponds closely to the subscale structure of the test, and
(content and “face” validity)? does not appear to vary across age or gender groupings.
• Do test scores measure a coherent, identifiable The concurrent validity of the ChIA was investigated by
theoretical construct or trait (construct validity)? examining its relationship to children’s performance on other
A closely related question is whether the test relevant measures. These results are displayed in Table 14. For
scores correlate with other instruments designed the 362 children who completed both the ChIA and the
to measure the same thing (concurrent validity). Attitudes Toward Guns and Violence Questionnaire (AGVQ;
• Do test scores discriminate among relevant Shapiro, 2000), correlations between the test scores were
groups in predictable ways (discriminant weak. This result underscores the relative independence be-
validity)? tween the experience of anger and its expression in an aggres-
It should be remembered that the question of test validity sive or violent manner. The highest correlations (r = .13) were
rests upon the accumulated weight of the evidence from the observed between the ChIA Frustration score and the AGVQ
validity studies conducted with a variety of populations and Aggressive Response to Shame score, and between the ChIA
employing various methodologies (Campbell & Fiske, Authority Relations score and the AGVQ Aggressive
1959). Response to Shame, Power/Safety, and Total scores.
For the 104 children who completed both the ChIA
Content Validity and the Aggression Questionnaire (AQ, Buss & Warren,
Content validity has to do with the development of the 2000), correlations between Total scores on the two instru-
individual test items. As described in chapter 4, the content ments were moderate (r = .44). As would be expected, the
validity and readability of the items were built into the de- strongest relationships for the ChIA subscales were with the
velopment of the original 1978 CIA. A large pool of items AQ Anger and Hostility scores (average r’s of .35 and .42,
was developed, each of which provided information with re- respectively). As with the AGVQ, these results again em-
gard to three different dimensions: (a) setting (home, school, phasize that experienced anger is only partially related to ex-
or both), (b) persons (parents, adults in authority, pressed aggression.
peers/friends, siblings, strangers), and (c) external events For the 45 children who completed both the ChIA and
(direct threat to self, environmental obstacles, direct threat the Novaco Anger Scale and Provocation Inventory (NAS
from significant others, perceived injustice). Thus, the CIA and PI; Novaco, 1975, in press), correlations between scores
items were developed to assess the range of situations in on the two instruments ranged from weak to moderate.
which a child’s anger is aroused, the type of situation most Notably, a relatively strong correlation was observed be-
likely to arouse that anger, the overall magnitude of a child’s tween the ChIA Frustration score and the PI Frustration score
proneness to provocation in each of these settings, and the (r = .38). The ChIA Frustration score also demonstrated a
individuals against whom this anger would most likely be relatively strong relationship with the NAS Anger Regulation
directed. This information was then further examined with score (–.37) and the PI Irritability score (r = .46). The ChIA
respect to the various settings within which children func- Authority Relations and Total scores were consistently and
tion and the individuals in those settings with whom the moderately related to NAS and PI scores (average r’s were
child interacts. .30 for each scale). The ChIA Peer Relationships and
At every stage of this developmental process, psycholo- Physical Aggression scores were more strongly related to PI
gists who specialized in working with children provided use- scores than to NAS scores. Social Desirability, as measured
ful feedback on the items as they were reworked. The items by the PI, was unrelated to ChIA scores.
were specifically written to be readable by children at the A group of 457 children also completed the ChIA and
youngest age for which the inventory would be appropriate. the Piers-Harris Children’s Self-Concept Scale (PHCSCS;
Over the years, the authors received feedback about the CIA Piers, 1996). As can be seen in Table 15, correlations be-
from professionals who had employed it. This information tween scores on these two instruments were negligible. This
provided the framework for refining the item set of the current is worth noting because it demonstrates that the ChIA ap-
ChIA. The original meaning of each test item was maintained pears to be selective in measuring the experience of anger in
to ensure that information gained over the years regarding the particular, rather than a generally poor psychological adjust-
1978 CIA would remain applicable to the current ChIA. The ment, unhappiness, or low self-regard.
correlation of .99 between the Total scores for the two forms The concurrent validity of the 1978 CIA has also been
demonstrates that this goal was achieved. explored. Finch and Eastman (1983) investigated multiple
measures of anger, drawn from various informants: the 1978
Construct Validity and Concurrent Validity self-reported CIA, peer nominations of anger, staff-noted
In the case of the ChIA, construct validity refers to the anger control problems, and teacher ratings of anger. The
extent to which the instrument measures the theoretical con- sample was a group of 38 boys who had significant emo-
struct of anger in particular, and the extent to which the sub- tional and/or behavioral difficulties and were in an inpatient
scales measure specific facets of anger. Factor analysis is one psychiatric unit. The boys ranged in age from 6 years and
36 Children’s Inventory of Anger (ChIA)

Table 14
Correlations Between ChIA Scores and
Scores on Related Measures of Anger and Aggression

Physical Peer Authority Total

Frustration Aggression Relationships Relations Score
Attitudes Toward Guns and Violence Questionnaire (AGVQ)a
Aggressive Response to Shame .13 .11 –.05 .13 .12
Comfort With Aggression .06 .01 .05 .06 .02
Excitement .05 .08 .04 –.02 .04
Power/Safety –.04 .10 .00 .13 .06
Total .08 .10 –.00 .13 .09
Aggression Questionnaire (AQ)b
Physical Aggression .27 .27 .36 .27 .33
Verbal Aggression .23 .13 .32 .22 .25
Anger .33 .34 .37 .35 .39
Hostility .46 .39 .44 .38 .47
Indirect Aggression .28 .20 .30 .25 .29
Total .40 .35 .45 .37 .44
Novaco Anger Scale (NAS) and Provocation Inventory (PI)c
Novaco Anger Scale
Angry Cognitions .20 .11 .25 .23 .23
Arousal .21 .05 .12 .24 .18
Angry Behavior .26 .10 .19 .39 .28
Anger Regulation –.37 –.18 –.18 –.21 –.28
Total Score .25 .09 .20 .32 .25
Provocation Inventory
Disrespectful Treatment .29 .18 .21 .27 .28
Unfairness .33 .39 .31 .36 .42
Frustration .38 .28 .20 .26 .33
Annoying Traits of Others .24 .32 .25 .43 .37
Irritability .46 .16 .30 .36 .37
Total Score .39 .30 .28 .37 .40
Social Desirability –.01 .05 –.09 –.11 –.05
a
N = 362.
b
N = 104.
c
N = 45.

10 months to 15 years and 3 months (mean age = 11 years the child in his unit who was best described by each item.
and 9 months). Scores for each child were determined by summing the pro-
The 1978 CIA was administered to each child individ- portion of the children in the unit who named him for each
ually by an examiner who did not know how the child had item and then transforming the scores. Example items from
scored on the other measures. The teacher ratings of anger the PNACP are “Gets angry easily,” “Student most likely to
were obtained by having each child’s special education start fighting over nothing,” and “Spends the least amount
teacher complete a rating scale constructed by a group of of time being angry.”
mental health professionals, using anger-oriented items In addition to the measures just described, a problem
from several rating scales. Other children in the inpatient list for each child was generated by the evaluating team (psy-
psychiatric unit completed the Peer Nomination of Anger chiatrist, psychologist, and social worker) and the child’s in-
Control Problems (PNACP; Finch, Saylor, & Nelson, 1987), dividual therapist. Two clinical child psychologists who did
a 40-item task developed for the study. The PNACP includes not have knowledge of the child’s scores on any of the other
items indicating anger management problems, as well as measures then reviewed his problem list and noted the pres-
anger management skills. On this scale, each child named ence or absence of anger control problems. They agreed on
 5 Psychometric Properties 37

Table 15
Correlations Between ChIA Scores and Scores on
the Piers-Harris Children’s Self-Concept Scale

Physical Peer Authority Total

Frustration Aggression Relationships Relations Score

Behavior –.01 .01 –.00 –.01 .00

Intellectual –.03 .06 .07 .02 .05
Physical .02 .01 .06 .03 .03
Anxiety –.05 –.05 –.05 –.03 –.05
Popularity –.09 –.05 –.06 –.01 –.06
Happiness .05 .06 .05 .04 .06
Total –.00 .01 .01 –.01 .00

Note. N = 457.

36 of the 38 cases (94.7%), and the senior psychologist’s rat- ing out, moodiness, and learning difficulties. On this latter
ings were accepted in the cases of disagreement. measure, teachers rated the frequency of each of the 11 de-
The 1978 CIA Total score did not correlate signifi- scribed behaviors on a 5-point scale, ranging from “seldom
cantly with the teacher ratings, but did correlate significant- or never” (1) to “all of the time” (5). In this study, the acting
ly with both portions of the PNACP (Problems, r = .36; out score was used as a gross measure of anger/aggression.
Skills, r = –.36; p < .05). In addition, those patients who The CIA Total score was significantly correlated with
were judged to have anger control problems on their prob- the CDI, but not with any of the other measures. It was not
lem lists were found to have scored significantly above the surprising that the CIA and CDI were associated, as anger is
mean on the CIA. Also, behavior rating scale items that per- theoretically related to depression, and both the CIA and the
tained to acting out were significantly correlated with the CDI are self-report measures. It was surprising, however,
CIA score. Again, the behavior rating items reflected ag- that none of the other measures were related to anger, be-
gression and not necessarily anger. None of the other behav- cause anger had been related to the Peer Nomination of
ior rating item groupings correlated with the ChIA. Anger Control Problems with the group of hospitalized chil-
The concurrent validity of the 1978 CIA was also dren in the Finch and Eastman (1983) study. One possible
explored in a multimethod assessment study of anger, explanation for this difference in results was the fact that the
involving the participation of 133 students in grades two hospitalized children were from small units (no more than
through eight at a private school in the southeast (Saylor, 10 children in a unit), which meant that each child was like-
Finch, Baskin, Furey, & Kelly, 1984). In addition to the CIA, ly to be very familiar with the other children in the same
students completed the Children’s Depression Inventory unit. Furthermore, all of the children in a given unit were in-
(CDI; Kovacs, 1992), the Peer Nomination Inventory of volved in group therapy with one another, and therefore
Depression (PNID; Lefkowitz & Tesiny, 1980), and an might be more accurate in the ratings of a private affect,
abbreviated form of the Peer Nomination of Anger Control such as anger. Again, the problem of differentiating the rela-
Problems (Finch & Eastman, 1983). tion between anger and aggression seemed apparent. The
The PNID is a 20-item scale composed of descriptive availability of subscale scores for the ChIA may be of help
questions (e.g., “Who often plays alone?” and “Who often in future research efforts to clarify these results.
sleeps in class?”), for which children are asked to nominate In a subsequent study (Wolfe et al., 1987), the 1978
the classmate who best fits each description. The child can CIA was administered to 95 hospitalized child psychiatric
list as many children as may be appropriate, or none if the inpatients as part of their comprehensive psychological eval-
item did not apply to any child. Each child receives a score uations. In addition, the Child Behavior Checklist-Teacher
on each item by counting the total number of nominations he Report Form (CBCL-TRF; Achenbach & Edelbrock, 1988)
or she received and dividing it by the number of children was completed for these children by special education teach-
participating. The item scores are then summed to obtain ers in the hospital. The CIA Total score was found to corre-
scores for depression, happiness, and popularity subscales. late significantly with the CBCL-TRF depression and
Teachers completed modified versions of the PNID anxiety subscales, which was consistent with findings with
and PNACP, as well as an 11-item measure for screening normal children (Saylor, Benson, & Einhaus, 1985; Saylor,
maladaptive behaviors, which yields subscale scores for act- Finch, Baskin, Furey, & Kelly, 1984). However, the CIA did
38 Children’s Inventory of Anger (ChIA)

not correlate with the aggression subscale, the internalizing ones most closely related to the level of violence involved in
scale, or the externalizing scale of the CBCL-TRF. Again, it their current offenses. Interestingly, there is a high negative
was found that overt aggression was not related to the CIA, correlation between the current self-report of the level of
but there does seem to be a relationship between anger as frustration and the level of violence involved in their past of-
measured by the CIA and depression as measured by both fenses. This appears to suggest that angry frustration is more
the CDI and teacher ratings on the CBCL-TRF. evident in youngsters with a less violent history.
Similarly, the 1978 CIA was administered to a group Average ChIA raw scores for a clinical group—
of 37 hospitalized child psychiatric inpatients, whose teach- youngsters in a residential treatment program for children
ers also completed the CBCL-TRF and the Conners Teacher from homes where there was domestic violence—are dis-
Rating Scale (CTRS; Goyette, Conners, & Ulrich, 1978). played in Table 17. Boys in this setting reported notably
The CTRS yields a total score as well as conduct problems, higher levels of anger on all scales than did girls (with large
hyperactivity, and inattentive/passive subscales. As in the effect sizes, of .71, .93, .73, .93 and .95 for Frustration,
previous study, the CIA Total score was significantly corre- Physical Aggression, Peer Relationships, Authority
lated with the CBCL-TRF depression subscale, but not with Relations, and Total score, respectively, for boys compared
any of its other scales. The CIA did not correlate signifi- with girls). Boys’ scores were also notably higher than aver-
cantly with any of the CTRS scores, although its correlation age scores for the standardization sample. Moderate to large
with the inattentive subscale approached significance. effect sizes ranging from .60 to 1.05 were observed for boys’
scores in comparison with average scores for the standard-
Discriminant Validity ization sample (displayed in Table 8). Girls’ scores were
Another means of determining the ChIA’s validity is about the same, or in the case of the Authority Relations
to evaluate the “performance” of the scale in identifying scale, even slightly lower than average scores for the stan-
children in correctional and clinical settings. ChIA raw dardization sample.
scores obtained by children in a residential juvenile delin- Results from two discriminant validity studies using
quent treatment center are presented in Table 16, alongside the 1978 CIA are also available. The first of these studies
raw scores obtained by a subgroup of the standardization examined whether the CIA Total score could distinguish be-
sample, matched for age. As would be expected, the young- tween 66 children hospitalized for emotional and/or behav-
sters in correctional settings had higher ChIA scores than ioral difficulties and 66 normal children matched on age,
did those in the standardization sample, especially for the gender, and ethnicity. Although the differences between
Frustration and Peer Relationships scales (effect sizes of .91 their CIA Total scores were not statistically significant, it is
and .66, respectively). Correlations between ChIA scores interesting that the inpatient children reported less anger
and severity ratings of current and past offenses are also pre- than did the normal controls, obtaining an average CIA Total
sented in Table 16. For youngsters in this correctional set- score of 189.7 as compared to an average of 202.0 for the
ting, the Frustration and Peer Relationships scores were the normal group (p < .052).

Table 16
Average ChIA Raw Scores for Youths in a Residential Juvenile Correctional Facility
and a Comparable Subgroup From the ChIA Standardization Sample

Correctional Standardization Correlations With Severity Ratings

Samplea Subsampleb Effect Current Offensec Past Offensed
Mean SD Mean SD Size r r

Frustration (FRUST) 22.2 4.9 18.4 4.2 .91 .27 –.47

Physical Aggression (PHYS) 26.0 5.7 24.3 5.6 .30 .06 .02
Peer Relationships (PEER) 20.1 4.8 17.2 4.4 .66 .21 .12
Authority Relations (AUTH) 23.2 5.6 23.4 5.3 .00 .16 .21
Total Score 95.6 17.1 87.4 16.8 .49 .23 –.08

Note. Mean age = 14.6 ± 0.7 (range: 13–16).

a
N = 24.
b
N = 218.
c
Current offense ratings: 0 = not related to a violent act; 1 = violent act against property, but not person(s); 2 = violent act against person(s).
d
Past offense ratings: 0 = no prior history of violence; 1 = prior history of violence against property, but not person(s); 2 = prior history of violence
against person(s).
 5 Psychometric Properties 39

Table 17
Average ChIA Raw Scores for Youths in a Residential
Treatment Program for Children From Homes With Domestic Violence

Total Samplea Boysb Girlsc

Effect Effect Effect
Mean SD Size Mean SD Size Mean SD Size

Frustration (FRUST) 24.9 6.9 .42 27.5 6.2 .82 22.9 6.8 .11
Physical Aggression (PHYS) 28.6 6.5 .33 31.7 5.4 .85 26.1 6.4 –.08
Peer Relationships (PEER) 21.8 6.6 .25 24.2 6.5 .65 19.8 6.1 –.08
Authority Relations (AUTH) 25.7 6.6 .08 29.0 5.9 .60 23.0 5.9 –.34
Total Score 105.7 23.9 .53 117.1 19.7 1.05 96.5 23.4 .10

Note. Effect size values shown here are based on comparison with average scores obtained by the standardization sample, displayed in Table 8.
a
N = 38; mean age = 10.7 ± 2.7 (range: 7–17).
b
n = 17; mean age = 11.7 ± 3.2.
c
n = 21; mean age = 12.5 ± 3.2

It was hypothesized that perhaps the hospitalized chil- ChIA Sensitivity to Change
dren in this study were somewhat more hesitant to acknowl- An additional measure of the validity of a scale is its
edge their anger because many of them were admitted to the ability to reflect change as a result of an intervention designed
residential facility for problems with aggression or anger. to change the construct that the scale is reportedly measuring.
Certainly, many hospitalized children were reluctant to share Saylor, Benson, and Einhaus (1985) found that the 1978 CIA
their feelings, and it may be that they were somewhat more Total scores of a group of emotionally disturbed children with
likely to give socially acceptable answers while hospitalized. anger management problems decreased significantly, in con-
On the other hand, the children with emotional disturbance trast to Total scores for children in a no-treatment control
may have been less aware of their own feelings of anger, and group, following stress inoculation for anger management.
consequently less able to report on them accurately. Another Similarly, employing a multiple-baseline single-subject
hypothesis is that individuals who are more aware of their design with several repetitions, Johnson (1980) found
anger may try to refrain from exhibiting behaviors that are changes in CIA Total scores following stress inoculation in a
likely to result in hospitalization. The availability of subscales group of children with emotional disturbance.
on the ChIA may help to clarify whether differential subscale
patterns will help differentiate between these two groups. It is Summary
hoped that further research in this area will help to define how
the experience of anger is related to clinical status. The Children’s Inventory of Anger (ChIA) was devel-
A second discriminant validity study compared the oped because there was a clear need for a standardized as-
1978 CIA Total scores for a group of 29 children with conduct sessment tool that could evaluate the subjective experience
disorders (CD), a group of 28 children with affective/anxiety of anger in children and adolescents. It is economical to use
disorders (A/A), and a group of 22 children dually diagnosed and easy to administer because it is a self-rating scale. Data
as having both of these disorders (CD + A/A). The a priori ex- show the ChIA to be a psychometrically sound, reliable, and
pectation was that the children with conduct disorders would reasonably valid measure of anger among children. Because
be more angry than a group of anxious/depressed children. it provides a unique subjective report that facilitates our un-
This assumption, however, fails to recognize the indepen- derstanding of the nature of anger in children, it is hoped
dence between anger and aggression that has been previously that the ChIA will help to improve the effectiveness of treat-
noted. The three groups did not differ significantly in their ment and intervention procedures.
CIA Total scores. Interestingly, however, the group that was In addition, the ChIA can facilitate researchers’ efforts
dually diagnosed had the highest scores, and the group with not only in experimental studies examining the correspon-
only affective/anxiety disorders had the lowest (CD group, dence between subjective and observable psychological
M = 189.52; A/A group, M = 186.82; CD + A/A group, constructs but also in evaluating psychotherapeutic and edu-
M = 199.55). Again, the availability of ChIA subscales may cational intervention strategies aimed at anger management.
help future researchers to clarify whether a consistent rela- It can also be helpful in the therapeutic process by providing
tionship exists between clinical diagnosis and reports of expe- data for developing the kind of anger hierarchies needed in
rienced anger in specific areas. behavioral and cognitive-behavioral interventions.
40 Children’s Inventory of Anger (ChIA)

The validation work reported here makes it clear that aggression, particularly in children, but this is not always the
many questions about the relationship between subjective case. Several characteristics of depression include increased
anger and overt behavior still remain. These questions are irritability, persistent anger, responding to events with angry
largely related to methodological problems associated with a outbursts, and/or an exaggerated sense of frustration
lack of adequate criteria against which to validate the inter- (DSM-IV; American Psychiatric Association, 1994). Anger
nal experience of anger, and also to a mistaken tendency to has also been viewed as an important characteristic of child
overestimate the connection between the internal experience and adolescent suicide. As previously indicated, the experi-
of anger and the externalized expression of aggression. ence of anger is private and subjective, which complicates
Experience using the instrument over the years with both its assessment. It should be noted, however, that this diffi-
hospitalized and normal children (Nelson & Finch, in press) culty is not unique to anger, as the assessment of childhood
has shown that the ChIA does provide clinical information depression has been plagued by the same dilemma (e.g.,
that is directly useful in case evaluation and treatment plan- Saylor, Finch, Baskin, Saylor, Darnell, & Furey, 1984).
ning. Informal reports from clinicians and researchers alike Clinically, however, a child’s subjective report of depression
have indicated that the ChIA provides a valuable source of must be taken as seriously as evidence for depression, even
information about a child. In the long run, however, demon- if “depressed behavior” has not been noted by outside ob-
stration of a test’s validity rests on data, not impressions. servers. Conversely, the child’s failure to admit to depres-
Consequently, it is hoped that researchers will continue, sive ideation does not rule out depression, especially when
with this improved instrument, to address the validity issues others observe depressive signs. The same is true for anger.
discussed in this chapter. In either case, assessment based on multiple sources and
Anger appears to be a multifaceted construct that must multiple methods is important with children. The ChIA can
be further defined. Anger may overlap with a number of provide a useful means of assessing this important dimen-
other constructs and behavioral problems. For example, irri- sion with children and adolescents.
tability and anger are frequently viewed as components of