Journal of Abnormal Psychology Copyright 1995 by the American Psychological Association, Inc.

1995, Vol. 104, No. 1, 184-196 002I-843X/95/S3.00

Alternative Representations of War Zone Stressors: Relationships to

Posttraumatic Stress Disorder in Male and Female Vietnam Veterans
Daniel W. King, Lynda A. King, David M. Gudanowski, and Dawn L. Vreven
Central Michigan University

Four conceptualizations of war zone stressor experiences were denned: traditional combat, atroci-
ties-abusive violence, perceived threat, and malevolent environment. Items from the National Viet-
nam Veterans Readjustment Study (NVVRS) were reviewed for content, and stressor indexes were
created. Using retrospective self-report data from the NVVRS, intercorrelations among stressor
scores and between these scores and measures of posttraumatic stress disorder (PTSD) were com-
puted for all veterans and for men and women separately. Structural equation modeling procedures
followed. Results indicated that the four stressor indexes were internally consistent, reasonably dis-
tinct from one another, and influenced PTSD differentially. Men scored significantly higher than
women on all 4 indexes. Whereas the pattern of relationships among the variables was comparable
across genders, there was evidence that one path coefficient in the model differed for men and women.

An integral component of the diagnosis of posttraumatic as high-magnitude forms of more ordinary stressful life events.
stress disorder (PTSD) is exposure to a universally accepted Their conclusion was that the likelihood of a psychopathological
traumatic event, Criterion A in the currently operative revised response and the particular nature of that response lies with
third edition of the Diagnostic and Statistical Manual of Mental subjective meaning grounded in preexisting personality traits. A
Disorders (DSM-III-R; American Psychiatric Association, similar position was proposed by Lazarus, DeLongis, Folkman,
1987). Kilpatrick, Resnick, and Freedy (1992) and Kilpatrick and Gruen (1985), who argued that there is no pure stressor
and Resnick (1993) appropriately characterized Criterion A as independent variable that can be isolated from personal ap-
playing the "key gatekeeping function" (p. 1 and p. 244, respec- praisal, and that the entire process of experiencing and respond-
tively) for a PTSD diagnosis; that is, its presence is a prerequisite ing to a stressor involves cyclical, multifactorial person-envi-
to consideration of the remaining diagnostic criteria: symptoms ronment relationships.
of reexperiencing (Criterion B), avoidance and numbing (Cri- Green (1990), on the other hand, sought to segregate the ob-
terion C), and hyperarousal (Criterion D) that have persisted for jective stressor event from the person's perception of that event
at least a month (Criterion E). and suggested that advancements in our understanding of stress
However, there is ongoing debate as to how the stressor crite- reactions necessitate such a distinction. She then offered a ty-
rion in PTSD should be denned and operationalized. One as- pology of objective, generic stressor dimensions to aid the clini-
pect of this debate is whether the stressor can and should be cian in linking specific Stressors to particular outcomes, which,
considered a truly objective event or conceptualized in terms of in turn, could inform treatment efforts. Along the same lines,
the subjective meaning of that event. In one of the earliest anal- Green proposed that the identification of different stressor cat-
yses of this issue, Breslau and Davis (1987) challenged the dis- egories or dimensions may generate "more precise research" (p.
criminant validity of PTSD on the basis of the thesis that the 1635) on how environmental events interact with other factors
denning characteristic of Criterion A as an extraordinary stres- (including subjective perceptions) to produce stress reactions.
sor fails to distinguish persons with PTSD from persons with A second aspect of the Criterion A debate concerns the char-
other stress reactions. They observed that extreme Stressors do acterization of the stressor event with regard to magnitude: be-
not always produce long-term dysfunction. Moreover, they yond "usual human experience" and "markedly distressing"
contended that such Stressors are not qualitatively unique in (italics added), as DSM-III-R prescribes (American Psychiat-
producing a particular response (PTSD), but rather function ric Association, 1987, p. 250). There is some interest in whether
more usual, less intense events can produce PTSD symptoms.
J. R. T. Davidson and Foa (1991), for example, remarked that
certain fairly common negative life events—they cited loss of a
Daniel W. King, Lynda A. King, David M. Gudanowski, and Dawn job and marital separation—may cause a PTSD-like response.
L. Vreven, Department of Psychology, Central Michigan University. They also pointed out that what is usual or typical for one indi-
Dawn L. Vreven is now at Brown University.
The study reported here was supported by National Institute of Men-
vidual may be quite unusual, and presumably far more distress-
tal Health Grant MH49168 (Violence and Traumatic Stress Program). ing, for another individual. Sutker, Uddo-Crane, and Allain
We express our sincere appreciation to John Fairbank, David Foy, Ter- (1991) similarly noted that some evidence suggests that re-
ence Keane, and William Schlenger for their thoughtful suggestions and peated, less intense events may produce a cumulative effect
assistance. equal to the impact of a single high-intensity traumatic event,
Correspondence concerning this article should be addressed to Dan- and that what one might characterize as "daily hassles" may, in
iel W. King, Department of Psychology, Central Michigan University, some cases, produce significantly greater dysfunction than that
Mount Pleasant, Michigan 48859. resulting from more intense events.
184
 WAR ZONE STRESSORS 185

We note here that the draft criteria for the fourth edition of those experiences were individually perceived and internalized.
the Diagnostic and Statistical Manual of Mental Disorders This theme was echoed in Baum, O'Keefe, and Davidson's
(DSM-IV; American Psychiatric Association, 1993) introduce (1990) model of acute and chronic stress, where duration of
a subjective component into Criterion A: a response of "intense perceived threat was considered as important as the objective
fear, helplessness, or horror" (p. K-8). No official recognition is characteristics of the stressor situation; L. M. Davidson and
given, however, to the possible role of low-level stressors in the Baum (1993) made specific application of this model to Viet-
etiology of PTSD. In anticipation of the release of the new cri- nam veterans. Furthermore, Solomon, Mikulincer, and Hobfoll
teria, March (1993) provided a timely summary of the objectiv- (1987) provided quite compelling empirical support for the im-
ity-subjectivity and event magnitude issues and discussed the portance of subjective interpretations of war-related traumatic
advantages and disadvantages of a broader conceptualization of or stressful events.
Criterion A. He concluded there was a current lack of research A final type of war zone experience deals with discomforts
evidence to resolve these issues, implying that additional em- and deprivations in day-to-day life that most veterans, both
pirical work is required. Accordingly, the present study sought combatants and noncombatants, experience (Kaylor, King, &
to address aspects of objectivity-subjectivity and event magni- King, 1987). These relatively low-magnitude stressors—at least
tude for one class of stressor events, war zone experiences, using when compared to combat or atrocity situations—might in-
retrospective self-report data from a national sample of male clude lack of desirable food, poor living arrangements, annoy-
and female Vietnam veterans. ing climate, unpredictable and extended work schedules, and
the like. They are the typical "daily hassles," irritations, and
War Zone Stressors: Alternative Representations pressures within a harsh or malevolent war zone. Although
there is a dearth of empirical research on this type of war zone
In large part, the stressor construct in past military veteran experience, assessments of such personal discomforts and de-
research has been conceptualized in terms of reports of observ- privations were included in the hallmark National Vietnam Vet-
able behaviors related to traditional combat events: firing a erans Readjustment Study (Kulka et al, 1990a, 1990b), with
weapon, being fired on, seeing persons wounded or killed, and the resulting data merged with other stressor scores to form that
the like. Because the experiences and mental health outcomes of study's global stress index.
veterans of the Vietnam War essentially motivated the contem- A first purpose of the present study was to use these four rep-
porary study of traumatic stress, combat exposure has been a resentations of war zone stressors to examine the Criterion A
prototypical Criterion A stressor for PTSD. The Vietnam vet- objectivity-subjectivity and event magnitude issues in terms of
eran literature contains numerous studies that document a sig- relationships to PTSD. Reports of exposure to traditional com-
nificant relationship between this representation of Criterion A bat events and reports of exposure to atrocities or episodes of
and measures of PTSD and other adjustment variables (e.g., Car- extraordinarily abusive violence were intended to denote poten-
roll, Rueger, Foy, & Donahoe, 1985; Foy, Resnick, Sipprelle, & tially verifiable (the more "objective") aspects of the war zone
Carroll, 1987; Foy, Sipprelle, Rueger, & Carroll, 1984; Res- experience, with a concern for high magnitude versus very high
nick, Foy, Donahoe, & Miller, 1989). magnitude, respectively. Personal judgments of threat or harm
Exposure to atrocities or exceptionally abusive violence is a in the face of war zone incidents and circumstances were in-
second stressor conceptualization that has received attention in cluded to assess a more subjective conceptualization of the
the veteran literature. Using data from the Legacies of Vietnam stressor. Finally, exposure to the harsh or malevolent environ-
study (Egendorf, Kadushin, Laufer, Rothbart, & Sloan, 1981), ment of daily life exemplified stressors of lower magnitude, and,
Laufer, Gallops, and Frey-Wouters (1984) created an index of as will be explained in the Method section, involved largely sub-
abusive violence from responses to questions about such inci- jective appraisals.
dents as the torture of prisoners, severe mistreatment of civil-
ians, use of cruel weaponry or chemicals, and mutilation of bod- Women and War Zone Stress
ies. Scores on this index were only moderately related to scores
on a traditional combat experience measure and were signifi- Wolfe (1993) recently pointed to a general lack of empirical
cantly related to postwar psychological states of veterans. Much studies of female military veterans. Indeed, whereas serious re-
the same results were reported by Breslau and Davis (1987) with search on male veterans of the Vietnam War is about to enter
a sample of veteran psychiatric patients. Callers, Foy, Donahoe, its third decade, it was not until the mid-1980s that a few re-
and Goldfarb (1988) found a relationship between PTSD and searchers (e.g., Paul, 1985; Schnaier, 1986; Stretch, Vail, & Ma-
several traumatic violence variables, and Green and her associ- loney, 1985) began to examine the unique war zone stressors
ates (Green, Grace, Lindy, & Gleser, 1990; Green, Grace, that were encountered by women in Vietnam and the adjust-
Lindy, Gleser, & Leonard, 1990; Green, Lindy, Grace, & Gleser, ment problems and stress-related symptoms they experienced
1989) endorsed the importance of a variable formulated in on return. Women Vietnam veterans, mostly registered nurses,
terms of exposure to grotesque death and mutilation. Thus, generally did not engage in what would be called traditional
war-related experiences of even higher magnitude than "nor- combat activities, nor were they typically in a position to wit-
mal" combat appear salient to the etiology of PTSD. ness or participate in the commission of atrocious or extremely
With regard to subjectivity, Hendin and Haas (1984) used a violent acts. On the other hand, perceived threat (fear for one's
case study approach to investigate the adaptation of 10 nondis- safety and well-being in a guerrilla war in which there were no
tressed heavy-combat Vietnam veterans and introduced the clearly established lines of battle) and harsh day-to-day living
idea that postcombat stress disorders are not so much a function (like their male counterparts, many female noncombatants
of the objective experiences in the war zone, but rather how contended with personal discomforts and deprivations accom-
186 KING, KING, GUDANOWSKI, AND VREVEN

panying a Vietnam tour) are possibly quite salient in under- tesque or raise questions of morality. As with traditional combat, this
standing the etiology of PTSD and other conditions for women. type of war zone stressor was defined so as to reduce personal inter-
pretations or subjective judgments. A distinction between witnessing
Wolfe, Brown, Furey, and Levin (1993) assessed war-related
and participating in atrocities was not possible with this category, be-
stressors specific to the experiences of female veterans and civil-
cause the response options to the interview questions ranged from no
ians who served in Vietnam and suggested the potential for involvement through degrees of witnessing to degrees of active partici-
differential relationships between war zone experiences and pation. (Sample items: "To what extent were you involved in terroriz-
outcomes for men and women. ing, wounding, or killing civilians?" "To what extent were you involved
Therefore, a second purpose of the present study was to give in mutilation of bodies of the enemy or civilians?")
special attention to the effect of war zone stressors on women Subjective or perceived threat was defined in terms of personal judg-
who served in Vietnam. We analyzed data separately for male ments or individual assessments of events or circumstances as poten-
and female veterans and made gender-based comparisons. tially threatening or harmful. This category reflected idiosyncratic
emotional or cognitive appraisals of situations that may or may not ac-
curately represent objective or factual reality. Of particular importance
Method in the definition of this type of stressor was the emphasis on inter-
pretations of and feelings about war zone events or circumstances.
Sample
(Sample items: "How often [if ever] did you find yourself in a combat
The data for this study were taken from those generated in the Na- situation in which you thought you would never survive?" "In your
tional Vietnam Veterans Readjustment Study (NVVRS; Kulka et al., opinion, how often were you in danger of being killed or wounded in [or
1990a, 1990b). A national probability sample of 1,632 Vietnam veter- around] Vietnam?")
ans participated in the NVVRS, 1,200 (74%) men and 432 (26%) General milieu of a harsh or malevolent environment was defined in
women. These individuals had served in the theater of war operations, terms of events or circumstances representing repeated or day-to-day
Vietnam or its surrounding regions, sometime between August 5, 1964, irritations and pressures related to life in the Vietnam war zone. The
and May 7, 1975. African American and Hispanic men, women, and definition referred specifically to "daily hassles" or bothersome and ag-
veterans with service-connected disabilities were oversampled. The gravating living conditions that accumulate to the point of possibly
overall response rate was 83%, and all veterans were individually ques- causing personal distress and creating a sense of futility, helplessness, or
tioned in their homes for approximately 5 hr on an array of premilitary, emotional emptiness. Because the items that were eventually placed in
military, and postmilitary variables. this category required the respondent to judge degree of personal com-
The current study was the first in a series of inquiries into the etiology fort or discomfort, this type of war zone stress connoted a subjective
of PTSD using the NVVRS data and was intended to address one aspect appraisal. (Sample items: "How unpleasant did you find this situation
of the causal framework, the role of war zone stressors. The larger, on- in Vietnam? loss of sleep," "How unpleasant did you find this situation
going project (of which this study is a part) aims to specify and evaluate in Vietnam? the insects, disease, and filth.")
a full theoretical model for the etiology of PTSD. Therefore, only a por- We then recruited two clinical psychologists expert in the research
tion of the sample was used for the present study of the war zone stres- and treatment of PTSD and six psychology graduate students for a sort-
sors in the model, allowing for other portions of the sample to be avail- ing task. Each was presented with the randomly shuffled full set of 113
able for the examination of premilitary and postmilitary factors, and items and the four stressor definitions and asked to sort each item into
especially for cross-validation of a fully integrated model. Consequently, one of the four content categories (or into a "none" category, if neces-
a 25% subsample (n = 408) was randomly selected from the 1,632 Viet- sary). The results of this sorting task were compiled, and items for which
nam theater veterans. The sample was stratified by gender, yielding 300 there was substantial agreement in categorization were retained.
men and 108 women. Next, the items within each category were reviewed, and some were
deleted due to content redundancy. Because of differences in response
Measures formats, retained items were all scaled to have a mean of 0 and a stan-
dard deviation of 1. Item-total correlations and estimates of internal
Stressor indexes. Four stressor indexes were developed using a ra- consistency were then computed for each stressor category. On the basis
tional approach to instrument construction (Hulin, Drasgow, & Par- of these values, a few additional items that detracted from the homoge-
sons, 1983; Jackson, 1971; Nunnally, 1978). As a first step, the "Viet- neity of their set were eliminated; care was taken, however, to maintain
nam Experience" portion of the NVVRS interview protocol was re- coverage of the relevant content domains. Means, standard deviations,
viewed, and 113 items with content judged as representing some aspect item-total correlations, and estimates of coefficient alpha were com-
of trauma or stress in the war zone were extracted and typed on separate puted for the four resulting stressor indexes.1
cards. PTSD measures. Current PTSD was assessed using three measures.
At the same time, rather extensive, formal definitions of the pre- The first was Keane, Caddell, and Taylor's (1988) Mississippi Scale for
viously discussed four stressors were set forth. Traditional combat events Combat-Related PTSD. This 35-item instrument uses a 5-point Likert
were defined in terms of reports of events or circumstances that would response format and assesses the standard PTSD symptoms as well as
be considered observable stereotypical combat-related experiences. For the associated features of depression, substance abuse, and suicidal
this stressor category, there was an attempt to exclude any personal in- tendencies. The Mississippi Scale has been shown to demonstrate high
terpretations or subjective judgments about the events or circumstances. internal consistency, test-retest reliability (Hyer, Davis, Boudewyns, &
It also was defined so as to exclude representations of highly deviant war- Woods, 1991; Keane et al., 1988; Kulka et al., 1990a, 1990b; McFall,
related experiences. (Sample items ultimately selected for this category: Smith, Mackay, & Tarver 1990) and validity in the form of expected
"How often did you fire a weapon?" "How often [if ever] did you see relationships with combat exposure (Keane et al.) and other PTSD mea-
[the bodies of] dead Americans?") sures and assessment devices (Kulka et al., 1990a, 1990b; McFall,
Atrocities or episodes of extraordinarily abusive violence were defined Smith, Roszell, Tarver, & Malas, 1990).
in terms of reports of observable events or circumstances that would be A second measure of current PTSD was taken from Kulka et al.'s
considered extreme or highly deviant forms of war-related experiences.
This representation included events or circumstances that might be
viewed as "going beyond" what most would regard as the "normal" 1
A copy of all items in each of the four stressor indexes may be ob-
expectations of warfare. Events might be characterized as very gro- tained from Daniel W. King.
 WAR ZONE STRESSORS 187
(1990a, 1990b) adaptation of the Diagnostic Interview Schedule (DIS; particularly their potential differential influences, we used a structural
Robins, Helzer, Croughan, & Ratcliff, 1981). For this measure, a total equation modeling strategy. The model included six indicators of PTSD
of 16 of the 17 DSM-III-R symptoms were assessed, each being scored (four derived from the Mississippi Scale along with the DIS symptom
as either 0 (absent within the last six months) or 1 (present within the count and the predicted probability of PTSD), and the four stressor in-
last six months). The item reflecting a sense of a foreshortened future dexes, with two (traditional combat and atrocities-abusive violence)
was previously eliminated from the interview by the original NVVRS treated as "causes" of latent constructs, and the other two (perceived
researchers. A count of the total number of endorsed symptoms was threat and malevolent environment) treated as manifest indicators of la-
computed. tent constructs. As with the previous analyses, attention was directed at
Finally, current PTSD was documented using Kulka et al.'s (1990a, the role of gender. We used the LISREL 8 (Joreskog & Sorbom, 1993a)
1990b) predicted probability computation. This is a composite proba- multisample analysis procedure (pp. 255-271) to determine the extent
bility statement calculated for each individual on the basis of a logistic to which relationships between the war zone stressors and PTSD were
regression equation. The equation incorporated nine variables, includ- similar for men and women, thus evaluating gender as a moderator
ing scores on PTSD assessment devices, adjustment scores, and demo- (Hoyle, 1991). We used generalized least squares estimation; Hu, Bent-
graphic characteristics. It should be noted that the score on the Missis- ler, and Kano (1992) found that this estimation procedure seemed pref-
sippi Scale and the DIS symptom count were included as part of the erable to maximum likelihood estimation with samples of less than 500
predictor set in the regression equation. As a probability value, scores cases and when latent variables and residuals were independent. Further
on this predicted PTSD measure could range from 0 to 1. details on the logic for model specification and evaluation are provided
in the Results and Discussion section. The correlation matrices for men
Limitations and Cautions Regarding the Data and women and the means and standard deviations for both groups are
contained in the Appendix.
The stressor indexes for the current study were based on retrospective
self-reports of events and circumstances that occurred approximately
10 to 20 years prior to data collection. Without doubt, there are a num-
ber of inherent problems in data of this sort. First and foremost is the Results and Discussion
possibility of simply poor recall, especially for events or circumstances
that were encoded at a time of emotional arousal or stress (Burke,
Heuer, & Reisberg, 1992; Loftus & Burns, 1982; Christiansen, 1992).
Descriptive Statistics and Reliability of Measures
In addition, this problem can be compounded by a potential tendency
to reconstruct details surrounding the prior events, which, in turn, The uppermost section of Table 1 presents scale characteris-
weakens the distinction between objectivity and subjectivity. In this re- tics and internal consistency reliability estimates for the four
gard, Metts, Sprecher, and Cupach (1991) pointed out that all retro- stressor indexes computed on the combined sample of male and
spective self-reports are reconstructions that vary in accuracy and ob- female veterans. The items making up each of these measures
jectivity, and that one's current state serves as a context within which
were submitted to z-score transformations prior to calculating
interpretations of the past are framed. Nisbett and Wilson (1977) com-
a summative score for each individual. Hence, each item had a
mented further that individuals often draw on internal, culturally ap-
propriate implicit causal theories to assign meaning to prior stimuli to mean of 0. Because there were missing data on several items,
account for a current condition. however, the average of the item sums over persons does not
For PTSD, in particular, present-day symptomatology and accompa- exactly equal (but does approximate) 0 for three of the four in-
nying dysfunction may create an attributional bias whereby the veteran dexes. The coefficient alpha values for all four indexes are quite
appraises war zone stressor events as more intense than what was actu- acceptable, suggesting a reasonably high degree of homogeneity
ally experienced. In the language of research design (Cook & Campbell, across items within each set. Along the same lines, internal con-
1979), attributions of this type introduce ambiguity about the direction sistency was reflected in strong item-total correlations, which
of causal influence, an inevitable consequence of cross-sectional studies
averaged .56, .63, .55, and .58 for the measures of traditional
with retrospective data. Other factors that should be considered as po-
tentially threatening to the validity of the data are recognition-seeking combat, atrocities-abusive violence, perceived threat, and ma-
or compensation-seeking on the part of the veteran and the need to levolent environment, respectively.
present oneself in a socially desirable manner (D. W. King & King, Table 1 also provides means, standard deviations, and in-
1991). Moreover, this latter issue may be especially relevant to the as- ternal consistency reliability estimates for the measures of
sessment of the atrocities-abusive violence stressor category. Many vet- PTSD. The average scores are quite consistent with what would
erans may have been quite hesitant to offer candid acknowledgements be expected from the community sample of veterans that par-
of this type of experience to the NVVRS interviewers. ticipated in the NVVRS. The mean Mississippi Scale score, for
example, is well below recommended screening cut-points,
Analyses which have ranged from 89 (Kulka et al, 1990a, 1990b) to 107
In addition to the previously described analyses required for develop- (Keane et al., 1988). The mean number of DIS-assessed PTSD
ing the stressor indexes and documenting their scale characteristics and symptoms is less than 1. The mean predicted probability of
reliability, we also computed scale means, standard deviations, and PTSD suggests an estimated 18% prevalence in this sample; the
(where appropriate) estimates of coefficient alpha for the PTSD mea- sampling design-weighted estimate resulting from the original
sures. Then, we intercorrelated the stressor indexes for the sample as a study and incorporating all veterans was 15% (Kulka et al.,
whole and separately for the subsamples of male and female Vietnam 1990a, 1990b). Regarding reliability, the Mississippi Scale's .94
veterans. We conducted tests of the significance of the difference be- value for this subsample of veterans is virtually identical to the
tween the correlations for men and the correlations for women, followed
value computed across all 2,348 theater and era veterans who
by multivariate analysis of variance and univariate t tests to determine
the significance of the difference between means of men and women on participated in the NVVRS (Kulka et al., 1990a, 1990b). The
the four stressor indexes. reliability of the DIS symptom count is remarkably high, con-
Next, to determine the contributions of the four stressors to PTSD, sidering that it consisted of 16 dichotomously scored items.
188 KING, KING, GUDANOWSKI, AND VREVEN

Table 1
Descriptive Statistics and Reliability Estimates for All Measures in the Study
Index-measure No. of items M SD N

Stressor indexes
Traditional combat 36 0.02" 20.04 .94 405
Atrocities-abusive violence 9 0.00" 6.31 .88 402
Perceived threat 9 0.02" 5.68 .83 405
Malevolent environment 18 o.o r 11.62 .91 405
PTSD measures
Mississippi Scale 35 72.41 21.80 .94 406
DIS symptom count 16 0.83 2.32 .91 387
Predicted probability of PTSDb 0.18 0.30 396

Note. PTSD = posttraumatic stress disorder; DIS = Diagnostic Interview Schedule.

* Means of Stressor indexes were computed on standardized item scores. Thus, each item had a mean of 0 and a standard deviation of 1. Because of
missing data, when summative scores across items were computed for all respondents, the expected scale mean of exactly 0 did not obtain for the
traditional
b
combat, perceived threat, and malevolent environment indexes.
Predicted probability of PTSD measure did not consist of a collection of items; classical test theory estimate of internal consistency was not
appropriate.

Stressor Indexes: Intercorrelations and Male-Female significant Wilks's lambda value of .87, with an exact F(3,
Comparisons 396) = 15.13, p < .001. As shown in Table 3, the univariate t
statistics demonstrated that male veterans reported signifi-
The intercorrelations among the four Stressor indexes are pre- cantly higher levels of war zone stress than female veterans
sented in Table 2. For the total sample, the six correlation co- across all four indexes. Gender accounted for a significant 11 %
efficients ranged from .44 (atrocities-abusive violence with ma- of the variance in scores on the traditional combat index, 6% of
levolent environment) to .79 (traditional combat with perceived the variance in scores on atrocities-abusive violence, and 2% of
threat). Despite the potential for monomethod bias—all items the variance in scores on both perceived threat and malevolent
were administered within the same segment of the interview, environment.
with the large majority requiring an oral response—the rela-
tionships among the variables suggest a fair amount of discrim- Structural Equation Modeling
inant validity. For men separately, the sizes of the coefficients
(.43 to .82) and the pattern of correlations were quite compara- As noted previously, we designed the structural equation
ble to those for the whole group, not an unexpected observation modeling phase of this study to assess the simultaneous and
considering that men comprised 74% of the total sample. For differential contributions of the several stressor variables to
female veterans, coefficients were lower (.30 to .72), and their PTSD within an integrated model. Figure 1 displays the initial
pattern of relationships differed somewhat from that of their model that was proposed. It contained both measurement and
male veteran counterparts. In fact, of the six correlations among structural components. Regarding the measurement compo-
Stressor indexes, three (traditional combat with atrocities-abu- nent, the endogenous latent variable of PTSD had six manifest
sive violence, traditional combat with perceived threat, and indicators. Two of these were the DIS symptom count and the
atrocities-abusive violence with perceived threat) were signifi- predicted probability of PTSD. The other four were average
cantly different for men and women (z values of 2.98, 2.19, and item scores for "mini-scales" (Bernstein & Teng, 1989, p. 475)
2.22, respectively, allps < .05). or content-based "parcels" (MacCallum, Roznowski, & Neco-
The multivariate analysis of variance test of the difference witz, 1992, p. 494) formed by clustering the 35 items in the
for men and women on the four stressor indexes produced a Mississippi Scale. These PTSD parcels or item clusters repre-

Table 2
Relationships Among Stressor Indexes
Total sample" Men" Women0
Variable
1. Traditional combat — .68 .79 .61 — .67 .82 .62 — .44 .72 .56
2. Atrocities-abusive violence — .49 .44 — .51 .43 — .30 .41
3. Perceived threat — .70 .72 ,61
4. Malevolent environment — —
• M range from 401 to 405.
b
Ns range from 293 to 297.
c
N = 108 for all correlations.
 WAR ZONE STRESSORS 189

Table 3
Differences Between Male Participants' and Female Participants' Means on Stressor Indexes
Men Women
(N == 297) (N = 108)

Variable M SD M SD t P<
Traditional combat 0.13 21.35 -0.28 9.61 9.55" .001
Atrocities-abusive violence1" 0.10 6.83 -0.29 3.51 6.78* .001
Perceived threat 0.08 5.81 -0.14 5.08 3.19 .01
Malevolent environment 0.07 11.95 -0.16 10.06 3.21 .01

• Because of significant differences between within-group variances for men and women, these t values were
computed using separate group variance estimates rather than pooled estimates.
b
n = 294.

sented the dimensions of reexperiencing and situational avoid- was an exogenous variable considered to be manifest in re-
ance (11 items), withdrawal and numbing (11 items), arousal sponses to six indicators, average scores on randomly grouped
and lack of control (8 items), and self-persecution (guilt and triplets formed from the 18 items comprising the malevolent
suicidality; 5 items). They were previously identified and used environment index.
in a series of exploratory and confirmatory factor analyses of The creation of miniscales or parcels was advantageous for
the Mississippi Scale (L. A. King & King, in press). several related reasons. First, single items within a scale such as
Perceived threat was likewise considered an endogenous la- those in the present study are generally considered rather unsta-
tent variable in the measurement component and had three ble (Bagozzi & Heatherton, 1994; Bernstein & Teng, 1989). As
manifest indicators, simply the average scores on each of three Marsh (1994) noted, the formation of parcels (as averages or
item parcels, formed by randomly grouping the nine items into sums of subsets of items) yields indicators of latent constructs
triplets. The third latent variable in the model's measurement that are more stable than single items and can result in a higher
component was exposure to a malevolent environment. This cases-to-variables or cases-to-parameter-estimates ratio. In ad-

MALEVOLENT
ENVIRONMENT

ATROCITIES/
ABUSIVE
VIOLENCE

Figure 1. Initial most saturated model displaying relationships among war zone stressors and posttrau-
matic stress disorder (PTSD). Miss = Mississippi Scale; DIS = Diagnostic Interview Schedule; pred =
predicted probability of PTSD.
190 KING, KING, GUDANOWSKI, AND VREVEN

dition, the specification of latent variables with multiple indica- was .80; and Steiger's (1990) root-mean-square error of approx-
tors (as opposed to merely using the total or average item score imation was .045.
on a full scale) yields perfectly reliable latent variables within Given that the pattern and structure of relationships seemed
the structural model (Joreskog & Sorbom, 1979, 1993a). Be- to be relatively consistent across genders, the next task was to
cause a reasonable number of manifest indicators are necessary assess whether the values of the parameters representing factor
to achieve an acceptable measurement model (Marsh & Bailey, loadings were the same for men and women. To this end, first, a
1991), the formation of parcel scores as averages of item clusters measurement model with equivalent or invariant factors across
within denned scales seemed a suitable means of meeting this groups was fit to the data, x2(234, AT = 397) = 409.03, p < .001.
requirement for the data in this study. The difference in fit between this more constrained model and
The last two manifest variables, exposure to traditional com- the previous more saturated model was nonsignificant, x2(12,
bat and exposure to atrocities-abusive violence, were both spec- N = 397) = 12.77, p = .386. The interpretation, therefore, is
ified in the measurement component as causes rather than invariant measurement of variables across male and female vet-
effects of their respective latent constructs. According to Loeh- erans (Reise, Widaman, & Pugh, 1993).
lin (1992) and MacCallum and Browne (1993), there are certain At this stage, evaluation of the significance of the individual
instances in which the conventional treatment of latent vari- structural coefficients suggested a respecification and simplifi-
ables as causes of observed variables is contrary to the rationale cation of the model portrayed in Figure 1. In particular, results
underlying the model. In this case, it seemed more logical that indicated negligible contributions of the paths from traditional
exposure to combat or atrocities would cause the experience of combat to PTSD for both men (t = -0.26) and women (t =
trauma or stress rather than the reverse. MacCallum and \ .08) and from atrocities-abusive violence to perceived threat
Browne pointed out possible problems that can arise when in- for both men (t = —1.66) and women (t = —0.26). The revised
corporating such causal indicators in a model, including identi- model, with two deleted paths, is depicted in Figure 2, where
fication problems and redundancy of the constructs caused by traditional combat now only indirectly influences PTSD
the observed variables. In the present model, however, each sin- through its impact on perceived threat and atrocities-abusive
gle manifest indicator causes a separate exogenous latent con- violence only directly influences PTSD, with no indirect path
struct or "composite variable" (MacCallum & Browne, 1993, through perceived threat. This model was fit to the data, x2(238,
p. 534), a situation less likely to be of concern. N = 397) = 413.11, p < .001, and the difference in fit between
As shown in Figure 1, all three exogenous variables (tradi- this model and the measurement invariance model was nonsig-
tional combat, atrocities-abusive violence, and malevolent en- nificant, x2(4, N = 397) = 4.08, p = .395. The two models,
vironment) were allowed to covary, as designated by the three therefore, appear functionally equivalent; deleting the two paths
curved two-headed arrows. In addition, the initial structural did not significantly alter the fit. Consequently, for both male
component proposed direct causal paths from each of these and female veterans, the more parsimonious model shown in
three exogenous variables to each of the two endogenous vari- Figure 2 must be deemed a better representation of the relation-
ables (perceived threat and PTSD). Moreover, a path from per- ships between war zone stressors and PTSD.
ceived threat to PTSD was specified. Indirect paths from the A chi-square difference test likewise was used to examine
three exogenous variables to PTSD by way of influence on the whether the strength of relationships between war zone stressors
perceived threat endogenous variable were thus a part of the and PTSD differed as a function of gender. A model constrain-
model. This more saturated model presumed, therefore, that all ing the five structural coefficients among latent variables for
stressor representations had a direct impact on PTSD; further- men to be equivalent to the corresponding structural coeffi-
more, it presumed that exposure to combat, to atrocities-abu- cients for women was fit to the data, x2(243, N = 397) = 427.64,
sive violence, and to the daily discomforts and deprivations of p < .001. The chi-square difference test between this model and
the war zone influenced the occurrence of PTSD by virtue of the previous model yielded a significant result, x2(5, N =
creating a judgment or appraisal of impending threat or harm. 397) = 14.53, p < .001. Thus, one or more of the path coeffi-
No covariances among the residual elements of the model were cients representing relationships among the latent variables in
postulated. Figure 2 appeared to be nonequivalent across genders.
The multisample modeling procedure was based on system- To explore the source of gender differences in structural path
atic chi-square difference testing of hierarchically nested coefficients, the modification indexes for the rejected equiva-
models (Anderson & Gerbing, 1988), proceeding from more lent-paths model were examined. The modification index for
saturated to less saturated. The first step was to postulate a sat- the traditional combat-to-perceived threat path for women, irf
urated model corresponding to Figure 1 simultaneously for comparison to the sizes of all other modification indexes for
both men and women. The pattern and structure of relation- the structural component, was quite high. Because modification
ships were specified to be equivalent across men and women, indexes are estimates of the drop in a chi-square statistic should
but parameter estimates (loadings, path coefficients, and resid- a constrained parameter be freed, this observation suggested
uals) were free to take on values that best fit the data. This base that the particular equality constraint for this path might likely
model yielded a significant discrepancy index, x2(222, N = be the primary source of misfit: The effect of traditional combat
397) = 396.26, p < .001. Inspection of several other indexes, on perceived threat is not the same for men as for women. Ac-
however, suggested reasonably good support for the model. The cordingly, a less constrained model in which this path was al-
LISREL goodness-of-fit index was .81; Bentler and Bonett's lowed to be estimated separately for men and women was fit to
(1980) normed fit index was .99; Bentler's (1990) comparative the data, x2(242, N = 397) = 416.82, p < .001. Relaxing the
fit index was .99; Bollen's (1989) incremental fit index was .99; constraint, therefore, led to a significant decrease in the chi-
James, Mulaik, and Brett's (1982) parsimony normed fit index square statistic, x 2 (l, N = 397) = 10.82, p < .01, from that
 WAR ZONE STRESSORS 191

MALEVOLENT
ENVIRONMENT

ATROCITIES/
ABUSIVE
VIOLENCE

65
Figure 2. Revised model displaying relationships among war zone stressors and posttraumatic stress dis-
order (PTSD). Miss = Mississippi Scale; DIS = Diagnostic Interview Schedule; pred = predicted probability
ofPTSD.

produced by the model in which all structural coefficients were For the full model, the LISREL goodness-of-fit index was .80;
constrained to be equivalent. Furthermore, the model allowing the normed fit index (Bentler & Bonett, 1980) was .98; the com-
the combat-to-perceived threat path to vary for men and parative fit index (Bentler, 1990) was .99; the incremental fit
women did not differ significantly from the more saturated index (Bollen, 1989) was .99; the parsimony normed fit index
model in which all structural coefficients were allowed to vary (James et al., 1982) was .88; and the root-mean-square error
across genders, x2(4, N = 397) = 3.71, p = .45. Thus, the sim- of approximation (Steiger, 1990) was .043. Relative to the base
plest representation of gender differences in the model is one in model's overall fit indexes, the LISREL goodness-of-fit index
which the combat-to-perceived threat path differs for male and and the normed fit index dropped slightly, but the indexes rep-
female veterans, and all others are equivalent. resenting parsimony improved.
In the end, the model that was judged as most parsimonious Table 4 presents the total, direct, and indirect effects of the
generated the structural path coefficients displayed in Figure 3. latent stressor variables on the latent PTSD variable for both
These coefficients are scaled such that all variables are in a men and women. Overall, the Table 4 and Figure 3 findings give
pooled, common standard-score metric that takes into consid- strong evidence that the four stressors do impact PTSD and that
eration the differential dispersion of the variables within each their impact is manifest in different ways. A particularly striking
group. All paths are relatively substantial. In the male subsam- result is that the index with the largest total effect on PTSD was
ple, the structural component produced squared multiple cor- that reflecting exposure to the harsh or malevolent war zone
relations for the equations in which the latent endogenous vari- environment. Its effect was both direct and indirect, through the
ables were regressed on their respective causal factors of .85 for intervening variable of perceived threat. The results also suggest
perceived threat and .52 for PTSD. In the female subsample, that exposure to traditional combat experiences does not di-
the corresponding values were .63 and .43, respectively. Regard- rectly influence PTSD, but rather makes its impact indirectly
ing the observed variables, the squared multiple correlations in- through perceived threat.
volving the six manifest indicators of PTSD ranged from .39 to It is also quite interesting that the variable defined in terms
.85 for men and from .27 to .78 for women. Similar squared of exposure to atrocities-abusive violence had only a direct
multiple correlations for the three perceived threat indicators effect on PTSD, with no indirect impact through perceived
ranged from .64 to .78 for men and from .51 to .76 for women; threat. In retrospect, the negligible and nonsignificant effect of
for malevolent environment's six manifest indicators, these co- atrocities-abusive violence on perceived threat may partially be
efficients ranged from .63 to .74 for men and from .38 to .70 for attributed to the manner in which the former variable was op-
women. erationalized. Although the differentiation of witnessing atro-
192 KING, KING, GUDANOWSKI, AND VREVEN

MALEVOLENT
ENVIRONMENT

ATROCITIES/
ABUSIVE
VIOLENCE

Figure 3. Path coefficients and associated t statistics for the revised model. For the combat-to-perceived
threat path, separate values are given for male veterans (M) and female veterans (F). PTSD = posttraumatic
stress disorder.

cious acts as opposed to participating in them is somewhat tional combat events, exposure to atrocities or episodes of ex-
clouded by the response options to many of these items, the traordinarily abusive violence, subjective or perceived threat,
items still impart an implicit suggestion of active behavior on and general milieu of a harsh or malevolent environment. Using
the part of the respondent. To endorse these items and score retrospective self-report data from a national sample of male
high on the index may suggest admitting to being a perpetrator and female Vietnam veterans, indexes created to measure these
or the creator of threat. This is in contrast to the wording of representations were internally consistent and sufficiently dis-
most of the items in the traditional combat category, for which tinct from one another to argue for a fair amount of discrimi-
responding in the keyed direction suggests a sense of being the nant validity.
victim of war zone events, a circumstance that might reason- As one would expect, male veterans scored significantly
ably lead to perceptions of threat. higher on all four stressor indexes than their female veteran
counterparts. Differences in means were particularly pro-
Conclusions nounced for the traditional combat and atrocities-abusive vio-
In summary, the present study provided some support for lence indexes, two measures with content more likely descrip-
four representations of war zone stressors: exposure to tradi- tive of the male veteran experience. Whereas the measurement
component of the model was invariant across genders, evidence
for a fully invariant structural model was not convincing. The
Table4 pattern of relationships was generally similar for men and
Effects of War Zone Stressors on PTSD for the Revised Model women, but one path, from traditional combat to perceived
Total Direct Indirect threat, differed across genders.
Variable effect effect effect Rather noteworthy differential influences of the stressor in-
dexes on PTSD were evident. Considering the total effects of
Traditional combat each stressor variable on PTSD, malevolent environment ap-
Men .12 .00 .12
Women .29 .00 .29 peared to be the most potent factor for men and women. For
Atrocities-abusive violence .21 .21 .00 men, traditional combat had the least impact; for women,
Perceived threat .25 .25 atrocities-abusive violence had the least impact. The tradi-
Malevolent environment .48 .37 .12 tional combat variable was found not to have a direct effect on
Note. All effects are computed from the LISREL 8 common metric PTSD; rather, its influence was indirect through perceived
completely standardized solution. PTSD = posttraumatic stress disor- threat. Conversely, the atrocities-abusive violence stressor had
der. only a direct and not an indirect influence.
 WAR ZONE STRESSORS 193
What potential implications do these findings have for our Third, the results of this study point to the critical need to
understanding of the stressors that might contribute to PTSD? evaluate further what role is played by lower magnitude events,
First, the study affirms Green's (1990) observation that perhaps occurring over an extended period of time, in the etiol-
multiple representations of observable or verifiable stressor ex- ogy of PTSD. In this regard, the results are partially at variance
periences are worthy of consideration. In the present study, re- with the draft DSM-IV (American Psychiatric Association,
ports of exposure to traditional combat and atrocities-abusive 1993) Criterion A nosology. Whereas a DSM-based diagnosis
violence did not appear to make their impact on PTSD in the of PTSD requires the experiencing of an extraordinary event,
same way. Perhaps, to some degree, this reflects Green's passive the findings of this study suggest that it is nonetheless important
versus active categorization of stressors: in this case, a distinc- to assess perceptions of lower magnitude events, which may be
tion between the veteran as more of a victim of stressful events, occurring simultaneously with the more potent and tradition-
as suggested by the content of the traditional combat items, ver- ally recognized stressors. After all, like the global war zone ex-
sus the veteran as more of an "agent of his or her own stressor perience that was the focus of the present study, a number of
experience" (Green, 1990, p. 1638), as suggested by the content trauma-inducing environments may persist over time and may
and response options for the atrocities-abusive violence items. include exposure to multiple stressors of varying magnitudes. A
This differential mode of impact might motivate two different home situation in which there is abuse or domestic violence is a
approaches to clarify further the etiology of PTSD. Specifically, good example. It may have its own version of a persistent harsh
the indirect effect of traditional combat through perceived or malevolent environment, creating day-to-day pressures, dis-
threat (but no direct effect) seems to require study of other fac- comforts, or deprivations, in addition to likely incidents of se-
tors that covary with or cause one to process impinging events vere abuse or violence. It seems well worth knowing the extent
as being more or less personally threatening. On the other hand, to which lower magnitude stressors over some length of time
the direct (but not indirect through event appraisal) effect of create their own effects or exacerbate a person's reaction to an
atrocities-abusive violence primarily for male veterans may extremely traumatic incident. Or it may be the case that
suggest investigation into other factors that covary with or cause exposure to an extreme event or circumstance leads an individ-
one to be involved in atrocious acts. Furthermore, it is very ual to be overly sensitive or reactant to daily irritations and pres-
likely that guilt and shame, which were not a part of the present sures of lower magnitude.
study, mediate the relationship between atrocities-abusive vio- Fourth, the finding of a common pattern of stressor-PTSD
lence and PTSD. In other words, the data here are limited in relationships for male and female veterans may be viewed as
that they do not contain all appropriate subjective components support for the validity of PTSD. In the ideal, the explanatory
that are likely to qualify relationships between variables, in par- power of a construct is bound up in how well its relationships
ticular, between atrocities-abusive violence and PTSD. with other antecedents, correlates, and consequences hold up
Second, on the basis of the findings, it seems reasonable to across different groups, settings, and times (Cook & Campbell,
recommend that both objective and subjective aspects of stress- 1979). From this perspective, one would desire a common net-
ful events be assessed. This recommendation is consistent with work of relationships between relevant stressor variables and
the emerging Z>S'Af-/Kguidelines (American Psychiatric Asso- PTSD across men and women. Generally, this was the outcome
ciation, 1993). In this study, perceived threat, conceptualized in this study, with this network accounting for fairly substantial
and operationalized as a subjective personal appraisal, had a to- proportions of variance in PTSD (52% for men and 43% for
tal effect for male veterans that was greater than either tradi- women). On the other hand, the variation found for the one
tional combat or atrocities-abusive violence, both considered structural coefficient representing the impact of combat expe-
more observable, verifiable representations of war zone stress- riences is intriguing. Although female veterans reported sig-
ful events. For women, the total effect for perceived threat ex- nificantly lower levels of exposure to traditional combat events,
ceeded that for atrocities. As noted above, traditional combat the indirect effect of this variable on PTSD appeared markedly
only manifested an influence when its causal linkage to PTSD stronger for women (.29) than for men (.12). This outcome
was mediated by the subjective assessment of impending harm. should encourage additional gender-based investigations of war
We must return, however, to the obvious caution that accom- zone stressors.
panies any commentary on objectivity versus subjectivity in this Relatedly, the findings of the present study may reinforce the
study (and in most other studies where exposure to prior trau- need for continued work on stressor representations that are
matic or stressful events is of concern). All the data about war more exclusively the domain of women. The vast majority of
zone stressors were retrospective, self-reports of events, circum- the items used in the NVVRS were understandably directed at
stances, and perceptions thereof that occurred many years ear- "describing" the war zone experiences of male Vietnam veter-
lier. Among other problems (presented earlier), there is no ans, and, indeed, men tended to receive higher item and sum-
doubt that some amount of event reappraisal would necessarily mative index scores than women. Resnick, Kilpatrick, and Li-
cloud the veterans' accounts of what this study classified as the povsky (1991) have urged researchers to be more precise in their
more objective experiences. Here, as with the work of Solomon assessment of potential trauma among women (in their case,
et al. (1987), the items that operationalized the more objective victims of rape), so as to screen more accurately for Criterion A
traditional combat and atrocities-abusive violence indexes and subsequently diagnose PTSD. Likewise, the ongoing work
were intended to elicit accounts of observable events, whereas of Wolfe et al. (1993) to measure precisely various dimensions
the items that operationalized the more subjective perceived of women's war zone stress may be helpful in articulating causal
threat and malevolent environment indexes required reflection, factors for PTSD unique to the increasing number of women
deliberation, or evaluation. It is duly recognized, however, that who serve in the armed forces.
the issue of objectivity-subjectivity remains one of degree. Finally, in the present study, PTSD was treated as a single
194 KING, KING, GUDANOWSKI, AND VREVEN

latent variable with six manifest indicators, four of which were Foy, D. W., Sipprelle, R. C., Rueger, D. B., & Carroll, E. M. (1984).
derived from subsets of Mississippi Scale items grouped accord- Etiology of posttraumatic stress disorder in Vietnam veterans: Anal-
ing to a rational judgment of their content. An interesting ex- ysis of premilitary, military, and combat exposure influences. Journal
of Consulting and Clinical Psychology, 52, 79-87.
tension of this work might be to disaggregate PTSD into its sev-
Gallers, J., Foy, D. W, Donahoe, C. P., & Goldfarb, J. (1988). Post-
eral symptom clusters or subdimensions (e.g., reexperiencing, traumatic stress disorder in Vietnam combat veterans: Effects of trau-
active avoidance, passive avoidance, arousal) and model these matic violence exposure and military adjustment. Journal of Trau-
as latent endogenous variables in their own right. This strategy maticSlress, 1, 181-192.
could very likely produce a different pattern of relationships. Green, B. L. (1990). Defining trauma: Terminology and generic stressor
L. A. King and King's (in press) measurement model for Mis- dimensions. Journal of Applied Social Psychology, 20, 1632-1642.
sissippi Scale data, with a single second-order PTSD factor and Green, B. L., Grace, M. C., Lindy, J. D., & Gleser, G. C. (1990). War
four first-order symptom category factors, may provide a begin- stressors and symptom persistence in posttraumatic stress disorder.
ning point for such an inquiry. Journal of Anxiety Disorders, 4, 31-39.
Green, B. L., Grace, M. C., Lindy, J. D., Gleser, G. C., & Leonard, A.
References (1990). Risk factors for PTSD and other diagnoses in a general sam-
ple of Vietnam veterans. American Journal of Psychiatry, 147, 729-
American Psychiatric Association. (1987). Diagnostic and statistical 733.
manual of mental disorders (3rd ed., rev.). Washington, DC: Author. Green, B. L., Lindy, J. D., Grace, M. C., & Gleser, G. C. (1989).
American Psychiatric Association. (1993). DSM-IV Draft Criteria. Multiple diagnosis in posttraumatic stress disorder. The role of war
Washington, DC: Author. stressors. Journal of Nervous and Mental Disease, 177, 329-335.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling Hendin, H., & Haas, A. P. (1984). Combat adaptations of Vietnam vet-
in practice: A review and recommended two-step approach. Psycho- erans without posttraumatic stress disorders. American Journal of
logical Bulletin, 103, 411-423. Psychiatry, 141, 956-959.
Bagozzi, R. P., & Heatherton, T. (1994). A general approach to repre- Hoyle, R. H. (1991). Evaluating measurement models in clinical re-
senting multifaceted personality constructs: Application to state self- search: Covariance structure analysis of latent variable models of self-
esteem. Structural Equation Modeling: A Multidisciplinary Journal, conception. Journal of Consulting and Clinical Psychology, 59, 67-
1, 35-67. 76.
Baum, A., O'Keefe, M. K., & Davidson, L. M. (1990). Acute stressors Hu, L., Bentler, P. M., & Kano, Y. (1992). Can test statistics in covari-
and chronic response: The case of traumatic stress. Journal of Applied ance structure analysis be trusted? Psychological Bulletin, 112, 351-
Social Psychology, 20, 1643-1653. 362.
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psy- Hulin, C. L., Drasgow, F, & Parsons, C. K. (1983). Item response the-
chological Bulletin, 107, 238-246. ory: Application to psychological measurement. Homewood, IL: Dow
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness Jones-Irwin.
of fit in the analysis of covariance structures. Psychological Bulletin,
Hyer, L., Davis, H., Boudewyns, P., & Woods, M. G. (1991). A short
88, 588-606.
form of the Mississippi Scale for Combat-Related PTSD. Journal of
Bernstein, I. H., & Teng, G. (1989). Factoring items and factoring scales
Clinical Psychology, 47, 510-518.
are different: Spurious evidence for multidimensionality due to item
Jackson, D. N. (1971). The dynamics of structured personality tests:
categorization. Psychological Bulletin, 105, 467-477.
197'1. Psychological Review, 78, 229-248.
Bollen, K. A. (1989). A new incremental fit index for general structural
equation models. Sociological Methods and Research, 17, 303-316. James, L. R., Mulaik, S. A., & Brett, J. M. (1982). Causal analysis:
Breslau, N., & Davis, G. C. (1987). Posttraumatic stress disorder. The Assumptions, models, and data. Beverly Hills, CA: Sage.
stressor criterion. Journal of Nervous and Mental Disease, 175, 255- Joreskog, K. G., & Sorbom, D. (Eds.). (1979). Advances in factor analy-
264. sis and structural equation models. Cambridge, MA: Abt Associates.
Burke, A., Heuer, E, & Reisberg, D. (1992). Remembering emotional Joreskog, K. G., & Sorbom, D. (1993a). LISREL 8 user's reference
events. Memory and Cognition, 20, 277-290. guide. Chicago: Scientific Software.
Carroll, E. M., Rueger, D. B., Foy, D. W, & Donahoe, C. P. (1985). Joreskog, K. G., & Sorbom, D. (1993b). PRELIS 2 user's reference
Vietnam combat veterans with posttraumatic stress disorder: Analy- guide. Chicago: Scientific Software.
sis of marital and cohabitating adjustment. Journal of Abnormal Psy- Kaylor, J. A., King, D. W, & King, L. A. (1987). Psychological effects of
chology, 94, 329-337. military service in Vietnam: A meta-analysis. Psychological Bulletin,
Christiansen, S.-A. (1992). Emotional stress and eyewitness memory: A 102,251-211.
critical review. Psychological Bulletin, 112, 284-309. Keane, T. M., Caddell, J. M., & Taylor, K. L. (1988). Mississippi Scale
Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design for Combat-Related Posttraumatic Stress Disorder: Three studies in
and analysis issues for field settings. Chicago: Rand McNally. reliability and validity. Journal of Consulting and Clinical Psychology,
Davidson, J. R. T., & Foa, E. B. (1991). Diagnostic issues in posttrau- 56, 85-90.
matic stress disorder: Considerations for the DSM-IV. Journal of Ab- Kilpatrick, D. G., & Resnick, H. S. (1993). A description of the post-
normal Psychology, 100, 346-355. traumatic stress disorder field trial. In J. R. T. Davidson & E. B. Foa
Davidson, L. M., & Baum, A. (1993). Predictors of chronic stress (Eds.), Posttraumatic stress disorder: DSM-IV and beyond (pp. 243-
among Vietnam veterans: Stressor exposure and intrusive recall. 250). Washington, DC: American Psychiatric Press.
Journal of Traumatic Stress, 6, 195-212. Kilpatrick, D. G., Resnick, H. S., & Freedy, J. R. (1992, November).
Egendorf, A., Kadushin, C., Laufer, R. S., Rothbart, G., & Sloan, L. DSM-IV Post Traumatic Stress Disorder Field Trial: Criterion A and
(1981). Legacies of Vietnam: Comparative adjustment of veterans other stressful events histories associated with PTSD in clinical and
and their peers. Washington, DC: U.S. Government Printing Press. community samples. Paper presented at the annual meeting of the
Foy, D. W., Resnick, H. S., Sipprelle, R. C., & Carroll, E. M. (1987). Association for Advancement of Behavior Therapy, Boston, MA.
Premilitary, military, and postmilitary factors in the development of King, D. W., & King, L. A. (1991). Validity issues in research on Viet-
combat-related posttraumatic stress disorder. The Behavior Thera- nam veteran adjustment. Psychological Bulletin, 109, 107-124.
pist. 10. 3-9. King, L. A., & King, D. W. (in press). Latent structure of the Mississippi
 WAR ZONE STRESSORS 195

Scale for Combat-Related Posttraumatic Stress Disorder: Explor- Metis, S., Sprecher, S., & Cupach, W. R. (1991). Retrospective self-
atory and higher-order confirmatory factor analyses. Assessment. reports. In B. M. Montgomery & S. Duck (Eds.), Studying interper-
Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan, sonal interaction (pp. 162-178). New York: Guilford Press.
B. K., Marmar, C. R., & Weiss, D. S. (1990a). The National Vietnam Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know:
Veterans Readjustment Study: Tables of findings and technical appen- Verbal reports on mental processes. Psychological Review, 84, 231-
dices. New York: Brunner/Mazel. 259.
Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan, Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York:
B. K.., Marmar, C. R., & Weiss, D. S. (1990b). Trauma and the Viet- McGraw-Hill.
nam war generation: Report on the findings from the National Viet- Paul, E. A. (1985). Wounded healers: A summary of the Vietnam Nurse
nam Veterans Readjustment Study. New York: Brunner/Mazel. Veteran Project. Military Medicine, 150, 571-576.
Laufer, R. S., Gallops, M. S., & Frey-Wouters, E. (1984). War stress Reise, S. P., Widaman, K. F., & Pugh, R. H. (1993). Confirmatory factor
and trauma: The Vietnam veteran experience. Journal of Health and analysis and item response theory: Two approaches for exploring
Social Behavior, 25, 65-85. measurement invariance. Psychological Bulletin, 114, 552-566.
Lazarus, R. S., DeLongis, A., Folkman, S., & Gruen, R. (1985). Stress Resnick, H. S., Foy, D. W., Donahoe, C. P., & Miller, E. N. (1989).
and adaptational outcomes: The problem of confounded measures. Antisocial behavior and posttraumatic stress disorder in Vietnam
American Psychologist, 40, 770-779. veterans. Journal of Clinical Psychology, 45, 860-866.
Loehlin, J. C. (1992). Latent variable models: An introduction to factor, Resnick, H. S., Kilpatrick, D. G., & Lipovsky, J. A. (1991). Assessment
path, and structural analysis (2nd ed.). Hillsdale, NJ: Erlbaum. of rape-related posttraumatic stress disorder: Stressor and symptom
Loftus, E. F, & Burns, T. (1982). Mental shock can produce retrograde dimensions. Psychological Assessment, 3, 561-672.
amnesia. Memory and Cognition, 10, 318-323. Robins, L. E., Helzer, J. E., Croughan, J., & Ratcliff, K. S. (1981). Na-
tional Institute of Mental Health Diagnostic Interview Schedule. Ar-
MacCallum, R. C., & Browne, M. W. (1993). The use of causal indica-
chive of General Psychiatry, 38, 381-389.
tors in covariance structure models: Some practical issues. Psycho-
Schnaier, J. A. (1986). A study of women Vietnam veterans and their
logical Bulletin, 114. 533-541.
mental health adjustment. In C. R. Figley (Ed.), Trauma and its wake:
MacCallum, R. C., Roznowski, M., & Necowitz, L. B. (1992). Model
Vol. 2. Traumatic stress, theory, research, and intervention (pp. 97-
modifications in covariance structure analysis: The problem of capi-
119). New York: Brunner/Mazel.
talization on chance. Psychological Bulletin, 111, 490-504.
Solomon, Z., Mikulincer, M., & Hobfoll, S. E. (1987). Objective versus
March, J. S. (1993). What constitutes a stressor? The "Criterion A" is- subjective measurement of stress and social support: Combat-related
sue. In J. R. T. Davidson & E. B. Foa (Eds.), Posttraumatic stress reactions. Journal of Consulting and Clinical Psychology, 55, 577-
disorder: DSM-IV and beyond (pp. 37-54). Washington, DC: Amer- 583.
ican Psychiatric Press. Steiger, J. H. (1990). Structural model evaluation and modification: An
Marsh, H. W. (1994). Confirmatory factor analysis models of factorial interval estimation approach. Multivariate Behavioral Research, 25,
invariance: A multifaceted approach. Structural Equation Modeling: 173-180.
A Multidisciplinary Journal, 1, 5-34. Stretch, R. H., Vail, J. D., & Maloney, J. P. (1985). Posttraumatic stress
Marsh, H. W, & Bailey, M. (1991). Confirmatory factor analyses of disorder among Army Nurse Corps Vietnam veterans. Journal of
multitrait-multimethod data: A comparison of alternative models. Consulting and Clinical Psychology, 53, 704-708.
Applied Psychological Measurement, 15, 47-69. Sutker, P. B., Uddo-Crane, M., & Allain, A. N. (1991). Clinical and
McFall, M. E., Smith, D. E., Mackay, P. W, & Tarver, D. J. (1990). research assessment of posttraumatic stress disorder: A conceptual
Reliability and validity of the Mississippi Scale for Com bat-Related overview. Psychological Assessment, 3, 520-530.
Posttraumatic Stress Disorder. Psychological Assessment, 2, 114- Wolfe, J. (1993). Female military veterans and traumatic stress. PTSD
121. Research Quarterly, 4, 1-4.
McFall, M. E., Smith, D. E., Roszell, D. K., Tarver, D. J., & Malas, Wolfe, J., Brown, P. J., Furey, J., & Levin, K. B. (1993). Development
K. L. (1990). Convergent validity of measures of PTSD in Vietnam of a wartime stressor scale for women. Psychological Assessment, 5,
combat veterans. American Journal of Psychiatry, 147, 645-648. 330-335.

(Appendix follows on next page)

196 KING, KING, GUDANOWSKI, AND VREVEN

—• oo f^ t~~
gS^SigPigg^^Sg^SSg %
1 1 1 1-, +£

-^ .2 ON ON ON
— so <N <N •s'3 - - -
^ 1 ' 00 O 0 | -g (N № —
1 u
S) S h >>
|2 | § 3
oo oo
i1 IS?
X fi
CO » "53 'S
° u
^ t

r*- v> so ^G
m fN SO •* 00 so 00 — r- w> so C- — </1 Tf fN
—-• OO C*) V) c*°^
1 &"
8 "
^<
1 ' ' • oo oo fe 3 .S .So

111 1
K
.S3 is 13 '%

ro o ^ t^ G
O oo O t^ w^
is o c a
1 • • • • O O OO
o o '-C

SO OO so O
O r- —' r^ 1 •»"§
<3 1 O O OO
g |3
1
S § f>
>•
OO O
O t~-
C3 i3

do do
!
ft!
t - ~ —• • ' o - m —• —< o o o o —• oo so oo m os oo —•
SIC S!S
C• ^
g)r«1
8
fli
1 do do ^

< t^- Os 10 O
o
U*
^ O SO CN OO

|
i|
1 C O0 00
.2 I
^
_o _co

to •^•(N'^-r^imfNSO .-^•(N'^-mf^rOCNW^rs) T3 —' I— (N T>

14 o i •p
a
oo o o
1
S ss
S.'E 1 -__0 S
cx ^ 3 00 — OS 1^
< "2
^ C3
—-< OO f*} \D
CO
"2
CO
dd do
I
S^j
55 "> W
•o
M
C •Mr.aJ
CO
«OO-N vo«*«.noo»o,r,n* ej os i— oo — 1 8t
CO
§.§£
(/f ' 'I 2
- d -' d :f 1 1
1 — (N
^- r^
O 00
oo »/i
(NO —O
Eft*
6 . -- W3O
.0 -*-*

CO pas s a a s f i a s s s s s s s a so oo rf oo
t —
* 3 i>

!
(NO —O

cQ cd n>
a
G
(NO —' O .2 g -
T3 g P
a> *r o
SSsaa-^SSSSSU^S 1 OS OO VI X! y **""
• i
S
ro oo m r*i
(N r^i — (N
1 1
*§ ^ S

fN r*"i -^ fN rt V5 -Q
do do
.2 tf T3

«
C C C C C B 0

lit
t, 111111.2
^^^y-O C C C J C C C ' S ' ^
3 ^ ^ . _ j C / 5 — <NrO 2 S 2 £ 8 £J§ >

73

1 lt!in|l|lllllll! &" |
"* «
C
3 o

— ( N f i - ^ u - J s o r - ' o o a ^ d — r-irO'S-'uSsdr-' s > ^"1

5 "y H