Journal of Traumatic Stress, Vol. 13, No. 1, 2000
Trauma Exposure and Psychological Reactions
to Genocide Among Rwandan Children
Atle Dyregrov,1Leila Gupta,2Rolf Gjestad,1and Eugenie Mukanoheli2
A total of 3030 children age 8–19 years from Rwanda was interviewed about their
war experiences and reactions approximately 13 months after the genocide that
started in April 1994. Rwandan children had been exposed to extreme levels of
violence in the form of witnessing the death of close family members and others
in massacres, as well as other violent acts. A majority of these children (90%)
believed that they would die; most had to hide to survive, and 15% had to hide
under dead bodies to survive. A shortened form of the Impact of Event Scale
used in a group of 1830 of these children documented high levels of intrusion
and avoidance. While children living in shelters were exposed to more trauma,
they evidenced less posttraumatic reactions. Analyses showed that reactions were
associated with loss, violence exposure, and, most importantly, feeling their life
was in danger.
KEY WORDS: children; war; massacres; Impact of Event Scale.
Scientific reports about the psychosocial effects of war among civilian popu-
lations and the provision of mental health services have increased in recent years
& Zeidner, 1991; Klingman, 1992a–c; Kuterovac, Dyregrov, & Stuvland, 1994;
in warfare vary from conflict to conflict (e.g., distant or close, amount of violence
exposure, direct life threat, losses, civilian participation in the violence). Conse-
quently, one may also expect that the psychosocial effects they engender differ.
1Center for Crisis Psychology, Bergen, Norway.
0894-9867/00/0100-0003$18.00/1 C ?2000 International Society for Traumatic Stress Studies
4Dyregrov, Gupta, Gjestad, and Mukanoheli
some indication that threat to survival is especially important for the development
of posttraumatic symptoms in both adults (Fontana, Rosenheck, & Brett, 1992;
Hauff & Vaglum, 1993) and children (Carlson & Rosser-Hogan, 1994).
In the genocide that occurred in Rwanda in 1994, it is estimated that as
many as 1 million people of all ages were killed of a total population of about
7.5 million (mortality estimates made by the UN Human Rights Commission, the
sites throughout Rwanda). Most people were killed in their local communities by
perpetrators who were known to the victims. Survivors witnessed brutal killings
Psychosocial programs established for war-traumatized children emphasize
Tracing and documentation programs have been integral parts of these programs.
Following the genocide in Rwanda, a large number of children lost most, if not
all, of their family members through death, and additional centers for unaccom-
panied children were established for nearly 13,000 children around the country
[United Nations Children’s Fund (UNICEF) Situations Reports on Rwanda from
November 1994 and February 1995]. The term “unaccompanied center” is syn-
onymous with “orphanage,” but UNICEF advisors adopted the euphemism to en-
compass a broader category of children living in orphanages (i.e., those who may
be separated from their parents due to the war). For brevity, “center” is used for
these orphanages throughout this article.
UNICEF, in close collaboration with the Rwandan Ministry of Rehabilitation
and several nongovernmental organizations, developed and began implementing
a Trauma Recovery Program in October 1994. The Trauma Recovery Program
uses a community-based, training of trainers (TOT) approach which emphasizes
identification of trauma symptoms among children and use of simple methods of
expression, such as storytelling, drawing, writing, dancing, and drama based on
the Rwandan cultural context. The primary objective of the Trauma Program is to
build national capacity among Rwandan paraprofessionals and professionals who
and health providers) by strengthening their knowledge about child development,
trauma and grief theory, and listening skills. In addition to the TOT component,
the Trauma Recovery Program provided support to the Ministry of Education to
establish a National Trauma Center in 1995 in the capital Kigali for the following:
(1) outpatient treatment of severely traumatized children/families, (2) training of
trainers, (3) developing/sharing informational materials, (4) sensitizing the public
Rwandan Children in War5
about the effects of war-related violence on children through a radio message
program, and (5) conducting research. A key component of the Trauma Program
National Trauma Center.
The survey reported herein was undertaken to obtain a baseline assessment
of the nature and magnitude of exposure to traumatic events and the severity
of psychological reactions among a representative sample of Rwandese children
1 year postgenocide.
in their local language, Kinyarwanda. A quota sampling approach was used to
select the study sample. The quota sampling approach was based on four strata:
(1) age, 8–18 years; (2) gender, a balanced sample of approximately 50% for
each gender; (3) living situation (center vs. the community), where we attempted
to interview 50% from centers and 50% from communities; and (4) prefecture
at least 200 children from each of the 11 prefectures in Rwanda. Unfortunately,
this was not possible in 5 of the 11 prefectures due to the ongoing civil unrest and
interviewed in Kibungo, Ruhengeri, Byumba, and Gisenyi, and 195 in Cyangugu.
Overview of Sampling Methodology
Of a total of 77 centers and approximately 1700 primary and secondary
schools from 30 of 146 communes existing within all 11 prefectures into which
the survey. Due to lack of human resources and logistic support, it was impossible
for the Minister of Education to provide exact figures about the number of schools
destroyed during the war. At least 50 children living in each of the 31 centers
and 50 children from each of the 29 primary/secondary schools were interviewed.
An additional 30 secondary school students were interviewed from the capital
city, Kigali, to ensure access to adolescents for the second wave of data collection
in 1996. Individual children were selected by the Center Directors and School
Headmasters based on the desired age range and gender balance criteria above.
On the day of the interviews, the research assistants explained the purpose of the
study (which had been previously explained in writing when we obtained written
permission from each prefectorial official) to each Center Director/Headmaster
6 Dyregrov, Gupta, Gjestad, and Mukanoheli
so they could develop appropriate programs for them. Only two Center Directors
and one Headmaster refused to let their center/school participate in the study.
The Center Directors’ reasons were that they were “too short-staffed,” and the
Headmaster was “not interested.”
The survey questionnaire had three parts. Part I included basic demographic
information and specific questions concerning the nature of exposure to various
war events (Table 1). Part II included a 22-item revised version of the widely used
Table 1. Rwandese Children’s Exposure to War Scenes (N =3030)
N forProportion (%) Who
Answered AffirmativelyQuestion and ResponseEach Question
Have you experienced death in your family
due to the war?
If yes, were both parents killed?
Sister(s) and brothers?
Have you ever been threatened to be killed?
Did you believe that you would die?
Did you hide to protect yourself during the war?
Have you seen or witnessed any violence during
the recent war?
Have you witnessed with your own eyes someone
being injured or killed?
Did you hear someone being injured or killed?
Did you hear people screaming for help?
Have you been physically injured during the war?
What kind(s) of violence have you witnessed?
Someone being shot
Killings/injuries with pangas (machetes)
Rape or sexual assault
Destruction/looting of your house
Dead bodies/parts of bodies
Shelling or mortar fire
People being beaten with sticks
Many people killed at one time (massacres)
Children participating in killing(s) or injuring
Family members being killed
If yes, did you have to hide alone?
Did you ever hide under dead bodies?
Length of time you hid (4–8 weeks or longer)?
Rwandan Children in War7
adult version of the Impact of Event Scale (IES; Horowitz, Wilner, & Alvarez,
1979), which focuses on intrusive images, avoidance of reminders, arousal signs,
work with the original IES scale has shown problems with some of the items when
used with children (Dyregrov, Kuterovac, & Arpad, 1996; Yule, Bruggencate, &
Joseph, 1994). Based on research conducted with war-traumatized children in the
Middle East and former Yugoslavia, a child version was developed in collabora-
tion with the Center for Crisis Psychology, Bergen, Norway and the Institute of
Psychiatry, London, England (Dyregrov & Yule, 1995). By factor analysis, the
original 15 intrusion and avoidance items were reduced to eight items. New items
DSM-IV [American Psychiatric Association (APA), 1994] have been added. The
shorter scale was developed as a screening instrument that is easy to administer
and answer and has been shown to differentiate children with and without PTSD,
Grief Reaction Inventory used by Pynoos, Nader, Frederick, Gonda, and Stuber
(1987) (results not to be reported here).
Regarding the IES translation, the Primary Investigator and five Rwandan
professionals and paraprofessionals including a teacher/translator, a child psy-
chologist, a social worker, a medical doctor, and a researcher (who are all bilin-
gual), translated and back-translated all the questionnaire items from English to
the above group to ensure cultural appropriateness and linguistic accuracy. Each
item was discussed at length and consensus was reached for the simplest and most
appropriate translation. The final English–Kinyarwanda translated questionnaire
was then given to two independent professional translators to translate back to
English again. The overall correspondence for IES items was quite high, where
21 of the 22 (95%) Kinyarwanda–English translated items matched the English
items exactly. The instructions for answering the IES questions told the children
to “keep in mind the worst event that occurred to you during the fighting between
April and June 1994.”
cide began. Four female research assistants with similar educational backgrounds
(all four had high school diplomas; one had a teaching certificate beyond high
school) were trained and supervised by the Trauma Project Officer and Project
Assistant to administer the survey. The training for research assistants consisted
of three 1-hr mock interview practice sessions with each other and with chil-
dren of staff members. All training information was conveyed in French and
Kinyarwanda to ensure accurate understanding of all directions conveyed by the
8 Dyregrov, Gupta, Gjestad, and Mukanoheli
Interviewers were supervised in the field by the UNICEF Project Assistant
to ensure accuracy, and the first author conducted weekly debriefings with re-
search assistants to discuss/correct any coding errors after checking the completed
questionnaires. Before conducting interviews, written authorization was obtained
from all 11 prefect officials to administer the trauma survey in their respective
for unaccompanied children by the four Rwandan female research assistants, with
each interview lasting approximately 30–40 min, depending on the child’s age
and comprehension level. Younger children (under age 10) required more time
than older respondents. After completion of the survey, research assistants were
instructed to offer emotional support to any child who appeared excessively dis-
tressed during the interview. This information was recorded in the “Comments”
section for future follow-up by the UNICEF trauma team. Data were collected
between May and December 1995.
After all the interviews were completed, the data were coded, entered, and
checked by the Project Officer and the Project Assistant using EPI-Info 6.0, a sta-
tistical software package. The data were later entered on a SPSS file for further
statistical analysis. The analyses used were descriptive statistics (mean, standard
deviation, minimum score, maximum score, skewness, kurtosis), frequency anal-
yses, reliability analyses (Cronbach’s alpha), alpha if item deleted, average inter-
item correlation, Pearson product–moment correlations, and multiple regression.
The program used was Statistica for Windows, version 5.0.
were interviewed using an incorrect IES scale with only three categories instead
of four. For this reason, all computations using IES data are restricted to the
1830 cases with complete IES data. Forty-nine percent of the children were liv-
ing in centers, and the remainder were living in family settings with parents,
other relatives, or foster parents. The children ranged in age from 8 to 19 years,
with the majority of children aged 11–16 years. The gender balance was almost
evenly split at 51% males and 49% females. On average, the children had com-
pleted 5 years of formal primary/secondary school, with the range between 0 and
Rwandan Children in War9
The percentages of children exposed to traumatic events during the war are
of violence during the genocide. More than two-thirds of the sample actually saw
someone being injured or killed, and 78% experienced death in their immediate
family, of which more than one-third of these children witnessed the death of
their own family members. In addition, almost all the children saw dead bodies
or parts of bodies, and more than half of the children witnessed many people
being killed at one time (massacres), people being injured or killed with pangas
(machetes), and people being beaten with sticks. The majority of children saw
their homes destroyed/looted and/or heard people being injured/killed. Almost
all of the children interviewed believed that they themselves would die during the
war, and nearly two-thirds of them reported that their lives were threatened by the
militia. More than 80% of the children reported that they had to hide to protect
themselves during the war, and more than half of these children hid for 4–8 weeks
or longer, while one-fourth of them had to hide alone. The majority of these
children were exposed to strong sensory impressions such as people screaming
for help, visual images of graphic violence inflicted against family members, and
destruction of homes. Finally, 16% of the children reported that they had to hide
under dead bodies in order to survive the genocide.
An a priori categorization of exposure variables was made in which expo-
sure variables were grouped into three exposure indices: loss of family members,
threat of dying, and violence exposure of different types. One variable (length
of time in hiding) was dropped from these indexes, as this variable had different
categories than the other variables. The reason for using only three categories was
to reduce the number of further analyses. The drawback of this procedure, when
many variables containing different types of information are grouped, is loss of
information. Empirically, principal-component analyses (PCA) of all 30 items did
not reveal a clear factor structure. Eigenvalues showed seven factors and scree
plot revealed several possible numbers of factors. Exploration with different so-
lutions did not uncover any clear factor structure. These findings did not give any
clear foundation of empirically derived subscales, and we decided to use the most
simple factor structure, a three-factor solution. Internal consistency analyses were
done for each subscale (Cronbach’s alpha, alpha if item deleted, and item–total
correlation). Based on the results of these analyses, five items were left out of the
PCA of the exposure variables of the items in each index showed that a one-factor
solution (unrotated) explained 48.0% (loss), 34.0% (violence exposure), 60.2%
(threat), and 24.3% (exposure/total) of the variance in each index. Because of
the low percentage of explained variance in the scale exposure total and because
10 Dyregrov, Gupta, Gjestad, and Mukanoheli
Table 2. Responses (%) to the Revised Impact of Event Scale (N =1830)
Please Mark the Response that Characterizes How it Has
Been for You During the Last 2 Weeks
at AllRarely Sometimes Often
1. Do you think about the event when you do not want to?
2. Do you try to remove the event from your memory?
3. Do you worry you may not live to become an adult?
4. Do you have difficulty concentrating/paying attention?
5. Do you have waves of strong feelings about the event?
6. Do you startle more easily because of loud unexpected
sounds or feel more jumpy or nervous than before?
7. Do you stay away from situations or things that remind
you of the event?
8. Have you suddenly acted or felt like the event was
9. Do you try not to talk about the event?
10. Do pictures about event suddenly pop into your mind?
11. Have you heard or seen things that make you think
about what happened?
12. Do things that remind you about the event cause or
trigger bodily reactions (beating heart, trembling)?
13. Do you try not to think about the event?
14. Have you had trouble experiencing feelings such as
love, happiness or sadness?
15. Do you easily get irritable?
16. Do you feel that people around you do not understand
how you really feel?
17. Are you alert and watchful even when there is no
obvious need to be?
18. Do you have difficulty remembering what happened?
19. Have you been less interested in activities that you
used to enjoy (i.e., sports, hobbies, games)?
20. Do you feel that it is a waste of time to plan for the
future because you do not expect to live long?
21. Do you feel guilt about what happened?
22. Do you have problems/difficulty sleeping at night?
9.8 7.215.9 67.1
55.0 9.321.4 14.3
interitem correlations indicated a less homogeneous measure than for the other
scales, this index was dropped in further analyses.
The association between where the children lived (centers vs. community)
and the degree of exposure was as follows: exposure/loss–living, r(1505)=
.18, p<.001; exposure/violence–living, r(1505)=.12, p<.001; and exposure/
threat–living, r(1563)=−.14, p<.001. This correlation indicates that children
living in centers experienced more losses and violence exposure than children
living in the community. However, children living in the community experienced
more threats than children living in centers. These associations are small and
explain a low degree of variance in the dependent variables (3.3%, 1.4%, and
Rwandan Children in War11
In terms of children’s reactions to what they were exposed to during the war,
Table 2 shows that the majority of the sample said they thought about the event
reported that they often tried to stay away from situations or things that reminded
them of the event. Overall, the data indicate that many children continued to have
of the war, despite their attempts to remove the event from their memory and
to avoid these reminders. Many of the children also reported increased arousal
symptoms such as an inability to concentrate or pay attention.
arousal, and the sum of intrusion and avoidance) are presented in Table 3. The
results show that the skewness is within an acceptable range. Two subscales of the
IES were significantly correlated: intrusion and arousal [r(1806)=.53, p<.001]
and avoidance and arousal [r(1798)=−.09, p<.001]. Intrusion and avoidance
[r(1806)=.03, ns] did not correlate statistically significant.
that deleting any item gave no better reliability. If more items with the same
psychometric properties as the existing ones were added, a better Cronbach’s
alpha would have been obtained.
Table 3. Descriptive Statistics for Exposure and Symptom Measures
M SD Minimum Maximum Kurtosis Skewness Correlation
Loss 2.06 1.65.0 5.0
9.25 4.25.0 17.0
2.26 0.95 .0 3.0.01
15.9 6.06.0 28.0
12Dyregrov, Gupta, Gjestad, and Mukanoheli
Predicting Reactions (IES)
Forced blockwise regression analyses were used to analyze the relationship
between the independent variables, exposure/loss, exposure/violence, exposure/
threat, gender, age, and place of residence (centers or community) and the depen-
dent variables, intrusion, avoidance, and arousal. Two-way interaction variables
between the independent variables were also included in the analyses. Several
variables were grouped together in blocks and tested simultaneously. If a block
counted significantly, nonsignificant effects were removed. The rationale for our
controlling for the most important variables (exposure). This procedure involves
ber of reestimated analyses, the results of each step in the regression analyses (as
B-weights, beta-weights, differences in R2) and the correlation matrices are not
presented, only the final results. Given the large number of children in this study,
the chance of finding statistically significant results is high. Therefore, variables
explaining less than 1% of the variance in the dependent variables were removed,
even if they made a statistically significant contribution.
analysis is presented in Table 4. Gender, age, and the interaction variables did not
contribute significantly to this model. Children with the highest intrusion scores
were those who had lost someone, had experienced violence and threats, and were
living in the community. The beta-weights indicate that experience of threats have
the strongest influence, but the effect of place of residence is almost as strong as
the effect of threats.
Table 4. Multiple Regression with Intrusion as the Dependent Variable
.33 .14.42— 0.210.18.15
−.12.45 .39— 1.370.25.21
Note. N =1429. All regression parameters were statistically significant at p<.01.
Rwandan Children in War13
Table 5. Multiple Regression with Arousal as the Dependent Variable
.08 .17— .35
.24.13.42— .21.15 .13
−.12.46 .39— .98.16 .13
Note. N =1422. All regression parameters were statistically significant at p<.01.
Regression analysis with avoidance as the dependent variable resulted in just
one significant independent variable; exposure/threat. The results gave a B-weight
of.53andaninterceptof12.06. R =.09, R2=.01,andβ =.09(t =3.50, p<.01).
This relationship is small and shows that children exposed to threats scored higher
on avoidance than children lower on this scale.
The results for arousal as the dependent variable are quite similar to that of
intrusion. The same independent variables significantly predicted the dependent
variable. Results are presented in Table 5. Children with the highest arousal scores
living in the community.
The regression analyses show that the effects of age, gender, and the inter-
action variables do not explain significant portions of the variance in intrusion,
avoidance, and arousal. However, for intrusion and arousal all three exposure vari-
ables contribute significantly, whereas for avoidance, only the experience of threat
explains the variance.
Exposure and Distress Level
The results show that a majority of Rwandan children had experienced a
multitude of stressors, each of which alone would be regarded as extremely stress-
ful in a peaceful society. The amount of violence exposure, loss, and threat that
these children experienced is difficult to comprehend. Their daily lives include
14 Dyregrov, Gupta, Gjestad, and Mukanoheli
intrusions, avoidance reactions, and other posttraumatic problems, while they face
the formidable task of trying to make sense of incomprehensible events.
dren, participating in the killings have left strong impressions on these children’s
minds. More than 90% of the children believed that they would die during the war,
and about one-third worried that they may not live to become adults. In previous
research (Dyregrov & Yule, 1995), a cutoff score of 17, based on the intrusion and
avoidance score of the revised IES, was a good predictor of a PTSD diagnosis. In
the Rwanda group, 79% of the children interviewed had a score of 17 or higher
on the revised IES measure more than a year after the genocide. Although caution
should be used in interpreting this as evidence for a high rate of PTSD in this
group, it definitely indicates high levels of distress.
Dyregrov, Gjestad, and Raundalen (1999) found that time alone did little to
alleviate IES scores among Iraqi children and adolescents following the Gulf War.
fear of reprisal, it is not surprising that psychological distress persists among child
survivors in Rwanda. Similar persistence, although with more decline, has been
documented by Sack et al. (1993) for PTSD in children and adolescents following
exposure to massive war trauma.
Social and community supports are believed to be important factors in stim-
ulating and sustaining resilience in children (Holaday & McPhearson, 1997). The
extent of loss and trauma which affected all levels of society throughout Rwanda
may have rendered the traditional coping mechanisms and collective support less
viable and the whole adult community less receptive to children’s needs given the
adults’ traumas and grief. Elbedour, ten Bensel, and Bastien (1993) propose that
in times of war. In Rwanda these ties were brutally severed, and the entire social
fabric, including the church, was decimated.
The persistent consequences of armed conflict on children should make us
reluctant to think that children’s resiliency automatically will reduce childrens’
distress levels, as Eisenbruch (1988) has cautioned. The notion of resiliency in
children could easily become a new form of denial of trauma among children,
whereby political systems evade responsibility for helping war-traumatized chil-
dren (Dyregrov, Gupta, Gjestad, & Raundalen, 1996). Resiliency in children is
intimately tied to the availability of family and community resources (Holaday
& McPhearson, 1997; Walsh, 1996), resources that may be severely restricted in
called it an ultimate irony “that at the time when the human is most vulnerable to
the effects of trauma—during infancy and childhood—adults generally presume
the most resilience.” However, the documentation of high distress levels does not
Rwandan Children in War15
automatically mean that the children have become unable to function socially or
who emigrated to the United States have a high level of functioning. In contrast to
the adolescents in the Sack et al. study, the Rwandan children continue to live in a
threatening environment amidst a variety of traumatic reminders and with limited
family and community resources.
Relation Between Exposure and Reactions
The multivariate analyses clearly showed that exposure was related to the
degree of intrusive memories and thoughts, as well as arousal. Children with a
high intrusion and arousal score had lost someone; they experienced violence
exposure; and, most importantly, they felt threatened. Threat was the factor that
by place of residence, which had an almost-equal effect on predicting intrusion
levels. Living in the community was associated with higher intrusion scores. The
same pattern was evident for arousal as the dependent variable, regarding both
exposure and place of residence.
The results confirm the findings from other studies where the threat elements
of exposure have been found particularly predictive of distress (Nader, Pynoos,
Fairbanks, & Frederick, 1990). Our findings are also consistent with Carlson and
Rosser-Hogan (1994), who, in a study of survivors of the genocide in Cambodia,
related to survival. The group studied may be especially vulnerable because of the
high levels of disruptions in their primary relationships that have deprived them of
various forms of support from primary caretakers and their larger social milieu.
It should be remembered, however, that if other stressors not included in the
study [such as shortage of food (see Carlson & Rosser-Hogan, 1994)], as well as
more information about the children’s social environment, possible exposure to
new traumas, and mediating variables (i.e., coping style, personality), had been
assessed, a more complete understanding of what contributed to the children’s
reactions could have been achieved.
Living in the Community Versus at “Unaccompanied” Centers
That place of residence significantly predicted intrusion and arousal was not
altogether surprising given the unique circumstances in Rwandan society in the
aftermath of the genocide. Children living in the community evidenced higher
intrusion and arousal scores than those living at centers. This finding may be ex-
plained by the fact that the UNICEF Trauma Recovery Program initially targeted
the centers for training caregivers who worked with children on basic methods
16 Dyregrov, Gupta, Gjestad, and Mukanoheli
of trauma healing, whereas very few schools and family members received this
training in the first 6–10 months after the genocide. The higher distress level in the
community is even more compelling, as children living at centers initially experi-
enced more losses and greater violence exposures than children in the community,
although community children did report more threat to their life.
One of the primary goals of emergency programming for children involves
initial support for the speedy return of children to the community, preferably with
of caring for parentless children where no blood ties exist (Tolfree, 1996). This
study suggests that we should not adopt such strategies indiscriminately. Some
situations, such as the Rwanda genocide, may render communities less able to
care for children in the immediate aftermath of a widespread disaster such as this.
Wolff, Tesfai, Egasso, and Aradom (1995) showed that unaccompanied children
(4- to 7-year-old Eritreans) provided for in humane group care had their social and
cognitive development fostered, even when technical and material resources were
There may be certain characteristics associated with staying at centers that
of camaraderie among children who lost both parents and other family members
during the genocide, whereby they felt accepted and that their losses were not
unique. Moreover, children often received better basic care, such as food, shelter,
medical attention, recreational activities, and schooling, at these institutions than
in communities. The UNICEF/Trauma team prioritized trainings in centers. For
an opportunity to express their traumas to adults who were more knowledgeable
about trauma than the parents/caretakers in the communities.
Methodological and Cultural Limitations
variables have to be based on the assumption that the indicators are reflective and
not formative in nature. Several researchers have discussed the methodological
1991). There are reasons for regarding exposure variables as formative, if one
alpha would be wrong, as there would be no measurement error. In addition,
formative indicators do not necessarily have to be correlated. Cronbach’s alpha,
PCA, exploratory factor analysis (EFA), confirmatory factor analysis (CFA), and
are subjective in nature because the response is a subjective memory of what
happened earlier, the scale would contain measurement error. If the items were
Rwandan Children in War17
measuring different parts of a specific exposure, the items would correlate to a
certain degree. If this were the case, it would be proper to use test statistics and
derive scales based upon these analyses. There are logical reasons to suppose that
exposure variables have both reflective and formative aspects in the items and
that a separation of these aspects would improve the measurement of exposure to
violence. This topic should be examined in later studies.
Regarding the IES subscales, we found a small, statistically significant cor-
relation between avoidance and arousal, indicating that children high in avoidance
indicates that these two scales may be tapping related features. The fact that the
score. Psychometrically, the more items involved, the higher the reliability if the
this is not the case, and there is empirical support for dropping the IES-total from
the statistical analyses. It is also theoretically difficult to conceptualize what this
superior dimension really measures. Even if intrusion and avoidance were corre-
lated, it would be problematic to use the sum score of the constructs. Although
the Cronbach’s alpha would be high, this measure does not say anything about
the construct validity and theoretical dimensions of the measurement [cf. Messick
(1975) and Pedhazur & Schmelkin (1991) for a discussion about this subject].
actions and possible mediating variables. Working in war-affected countries often
presents security risks, and the lack of basic infrastructure and human resources
creates less than ideal conditions for undertaking rigorous research. Despite the
difficult circumstances, all 11 prefectures were represented in the study, although
fewer children than originally intended were interviewed in five prefectures. Chil-
dren in four of these five prefectures reported more exposure than children in the
Ideally, instruments developed for and cross-validated within a particular
instruments are lacking, the instruments used should at least have been translated
and back-translated into the original language (Pernice, 1994). This procedure
was thoroughly followed. However, there might be a bias toward acknowledging
to respond in a socially desirable way (Hurh & Kim, 1982). The distribution in
acknowledgment of various reactions (see Table 2) is, however, an indication that
and were capable of answering in a manner reflecting their true response. With
the kind of questions used, it would also be difficult for the children to know what
would be a desirable response.
Because violent acts continued to take place throughout the research period,
children may have been reluctant to disclose information about their exposure in