Edward S. Kubany, Ph.D.
|Diplomat in Behavioral Psychology, American Board of
Professional Psychology (ABPP) Board Certified Expert in Traumatic Stress, American
Academy of Experts in Traumatic Stress (BCETS)
Trauma-Related Guilt Inventory (TRGI)
|I was born and raised in Chicago. I received a B.A.
from the University of Colorado, an M.A. from Long Beach State University, and a Ph.D.
from the University of Colorado. I am currently a research health scientist and clinical
psychologist at the Pacific Islands Division of the National Center for PTSD, Department
of Veterans Affairs (VA), in Honolulu, Hawaii. I also have a small private practice where
I work exclusively with trauma survivors.
It was a circuitous and fortuitous series of events that led me to develop the Trauma-Related Guilt Inventory (TRGI). Prior to its development, I had been practicing clinical psychology as a behavior therapist for several years, with no aspirations of becoming a test developer and with no special interest in psychological trauma. However, my interest in trauma and its effects soared in 1990 when I accepted a position with the Veterans Administration (VA) with responsibility for conducting interview assessments of Vietnam combat veterans to diagnose post-traumatic stress disorder (PTSD). PTSD is caused by exposure to traumatic events and is common among survivors of combat, physical and sexual abuse, and other kinds of traumatic life events.
During my first year at the VA, I interviewed more than 50 combat veterans and was intrigued to see themes of guilt emerge repeatedly as a source of anguish that seemed to exacerbate PTSD (for example, guilt about having been afraid, guilt about having caused accidental harm, guilt about not having been able to prevent harm or do more for the wounded or suffering). Equally important was the fact that veterans guilt typically had no rational basis. For example, many veterans concluded that they could have and "should have" prevented negative outcomes that were not foreseeable or preventable (Kubany, 1994). As a result, I started to conduct brief cognitive therapy to correct veterans' distorted beliefs about the importance of their roles in wartime trauma. [This initial interest in cognitive therapy blossomed into a major interest in therapeutic intervention research that paralleled my emerging career as a test developer (see Kubany & Manke, 1995)].
In reviewing the literature for an article on a cognitive model of guilt typology (Kubany, 1994), I discovered that guilt
|is a common sequel of many types of trauma (from
incest to technological disasters) and that trauma survivors, in general, tend to hold
distorted beliefs about their roles in the trauma. Curiously, many of my colleagues did
not share my view that trauma-related guilt is a critically important problem. One of the
reasons for this oversight, I concluded, was that problematic guilt was not being
identified during pre-therapy assessment because no validated measure of the construct
existed. I was struck by the need for a reliable and valid measure of trauma-related guilt
to raise clinician awareness of the need to focus on guilt management in trauma recovery
My interest in developing a test myself was aroused because I had a friend and colleague, Stephen N. Hayes, who was an expert in the field of assessment. Steve had written several books on assessment and had just completed a multi-study project to develop and validate a Marital Satisfaction Scale for Older Persons (Floyd, Haynes, et al., 1992). He was also an associate editor (and is now editor) of the journal Psychological Assessment. In 1992, I asked Steve to collaborate with me as my mentor in the development of a guilt scale. He replied, "Are you unprepared to spend three years to do a proper job?" I responded, "You can't be serious." Well, Steve was serious, and his prediction was accurate. We spent three and a half years developing, refining, and validating the TRGI and another two years cross-validating the inventory in a project funded by the VA.
We needed a definition of the guilt construct to guide scale development. In arriving at a definition, I was strongly influenced by appraisal theories of emotion, which view discrete emotions as undifferentiated emotional states with certain kinds of accompanying interpretations (Schacter, 1970) or labels (Staats, 1968). Most would agree that guilt evokes negative affect, and in discussions with clients about guilt issues over the years, I noted that they repeatedly referred to things they "should have" or "should not have" doneobligations they did not meet. Thus, consistent with appraisal theories of emotion, we defined guilt phenomenologically as "an unpleasant feeling with an accompanying belief (or beliefs) that one should have thought, felt, or acted differently."
|In developing the TRGI, we placed a major emphasis on
content validation. According to Haynes, Richard, & Kubany (1995), content validity is
the degree to which elements of an assessment instrument (for example, individual items,
response format, instructions to respondents) are relevant to and representative of the
targeted construct for a particular purpose. We wanted to be sure the TRGI would assess
all important facets or dimensions of guilt. In so doing, we hoped to generate items which
would comprehensively sample (or be representative of) the domain of trauma-related guilt
as it is experienced phenomenologically. Consistent with recommendations regarding content
validation (Haynes et al., 1995), dimensions of our construct were identified from
multiple sources of information: (1) our clinical work with trauma survivors, (2) a review
and analysis of the guilt literature, (3) examination of previously published guilt
scales, and, most importantly, (4) analyses of interviews that probed the phenomenology of
We conducted structured interviews with eighteen combat veterans with PTSD and one formerly battered woman. The interview format included nine open-ended questions (for example, "Explain what feeling guilty means to you," When you feel guilty about what happened, what thoughts go through your mind?). These interviews were transcribed and reviewed for characteristics of guilt by five experts in the field of psychological trauma. Characteristics noted by the reviewers were collated, and the summaries were reviewed for characteristics of guilt by five experts in the field of psychological trauma. Characteristics noted by the reviewers were collate, and the summaries were reviewed and evaluated jointly by three of the investigators. This one phase of instrument development alone took more than six months to complete and at the time seemed like a millennium.
We were also interested in another important aspect of content validationexternal validity. External validity is the extent to which the results of a piece of research can be generalized beyond an immediate study (for example, to other populations). As conceived, the TRGI was intended to be appropriate for use with and valid across trauma populations. However, if we developed and validated the TRGI with only one trauma population (such as combat veterans), we would not know how appropriate or valid the instrument would be for other trauma populations (such as battered
|women). We conducted three factor-analytic studies,
two of which included college students with histories of exposure to a wide variety of
traumatic events (accidents, sudden deaths of loved ones, sexual abuse, intimate partner
abuse, etc.). The third factor-analytic study was conducted with a sample of battered
women. Across the three samples, the factor structure of the TRGI was stable and robust.
We conducted test-retest studies with college students and with combat veterans and
obtained similar results. We conducted convergent and discriminant validity studies with
groups of battered women and combat veterans. In both samples, TRGI scales and subscales
were highly correlated with PTSD and depression severity and with several other measures
of psychological adjustmentreflecting strong convergent validity. In both samples,
TRGI scales and subscales were not significantly correlated with age or
educationreflecting discriminant validity. The combat veterans and battered women
differed on several dimensions, including gender, age, type of trauma, and trauma recency.
Yet the findings obtained with the two groups were very similar. Finally, no significant
differences in trauma-related guilt were observed between groups of caucasian, filipino,
and native hawaiian women, suggesting that the TRGI may also be valid across ethnic
groups. According to one of the anonymous reviewers of our article on the TRGI,
"Attention to issues of external validity occur too rarely in psychometric research,
and the authors deserve credit for attending to this key element of validation."
Developing the TRGI was only the beginning of my interest in developing psychometrically sound assessment instrumentation. With Steve Haynes and others, I have also developed and validated the Traumatic Life Events Questionaire (a comprehensive measure of prior trauma exposure) and the Distressing Event Questionnaire (a brief diagnostic measure of PTSD). Content validation was also an extremely important phase in the development of these two scales. I also hope to develop scales assessing trauma-related anger and trauma-related bereavement, problems which I believe account for unique variance in PTSD severity over and above variance accounted for by trauma-related guilt. Little did I know, six years ago, that a large part of my subsequent career would involve test development.
|Some Observations and Tips Which New Test
Developers May Want to Consider
1. We experimented with several different response formats for the TRGI before settling on a format almost identical to that used on the Mississippi Scale, a well-validated and highly respected measure of combat-related PTSD (Keane, Caddell & Taylor, 1988). We concluded that, if Keane et al.'s format worked well, a response format in the assessment of a construct related to guilt, a similar format might serve us well in assessing the construct of guilt. The lesson for new test developers may be, "Before you create your 'uniques' response format, check out the response formats on several other scales with strong psychometric properties, which assess constructs related to yours. You may not have to re-invent the wheel."
2. In the initial version of the TRGI, there were 11 reverse-scored items. Four of these items were left off the final version (for example, "What happened was not my fault). Interestingly, several of the reverse-scored items that loaded on a single factor did not load as highly as did other items which loaded on that factor. We think this may have occurred because some respondents were confused by some of the reverse-scored items and answered them in an unintended direction. Thus, including reverse-scored items on a scale may sometimes result in a trade-off between discouraging development of a response set and increasing unintentional errors.
3. Interestingly, the empirically determined factor structure of trauma-related guilt differed slightly from its hypothesized structure based on our multi-method efforts to identify different, independent dimensions of guilt. For example, hindsight bias items and responsibility items loaded on the same factor rather than on separate factors. This finding underscores a couple of observations. First, the factors or dimensions that a test developer thinks "should" comprise a psychological construct based on a rational or theoretical analysis may not always yield that factor structure in empirical analyses. Second, the empirically determined factor structure of a construct may sometimes have theoretical implications. In the case of the TRGI, the fact that responsibility and hindsight bias items
|loaded on the same factor suggests that
hindsight-biased thinking (the tendency for outcome knowledge to bias recollections of
pre-outcome knowledge which is akin to "Monday morning quarter-backing")
may underlie the repeated observation that trauma survivors tend to distort or exaggerate
the importance of their roles in traumatic events.
4. I strongly encourage new test developers to use multiple methods for generating items to maximize the likelihood that the test items will adequately represent the domain of the construct being measured. Such methods include but are not limited to (1) reference to theory; (2) analyses of structured interviews and/open-ended questionnaires on the phenomenology of the construct of interest; (3) systematic reviews and ratings by experts on various aspects of the test [for example, instructions to respondents; item wording; comprehensiveness of coverage of the construct's domain; (4) focus groups with potential respondents and/or experts to evaluate item wording, item relevance, and comprehensiveness of coverage [see Kubany et al., (1997) for examples of (2), (3) and (4)].
5. One last tip for potential test developers: Look for a mentor with established expertise in scale construction. Certainly I was blessed to work with someone with exceptional expertise in behavioral and psychological assessment, including test development. Realistically, however, I recognize that assessment role models are not available everywhere, although there are experts in statistical analysis in most communities who can serve as consultants for test developers with weak statistical backgrounds. One written source of "expertise" (besides this text) that I strongly recommend is the September 1995 issue of Psychological Assessment, which was a special issue on various methodological issues in psychological assessment research (18 articles).
Haynes, S. N., Floyd, F. J., Lemsky, C., Rogers, E., Winemillr, D., Heilman, N., Wele,
M., Murphy, T., Cardone, L. (1992). The Marital Satisfaction Questionairre for Older Persons. Psychological Assessment, 4, 473482.
Haynes, S. N., Richard, D. R., & Kubany, E. S. (1995). Content validity in psychological assessment: A functional approach to concepts and methods Psychological Assessment, 7, 238-247.
Keane, T. M., Caddell, J. M., & Taylor, K. L. (1988). The Mississippi Scale for Combat-Related PTSD: Three studies in reliability and validity. Journal of Consulting and Clinical Psychology, 56, 8590.
Kubany, E. S., Abueg, F. R., Kilauano, W., Manke, R. P., & A, Kaplan, A. (1997).
Development and validation of the Sources of Trauma-Related Guilt Survey--War-Zone Version. Journal of Traumatic Stress, 10, 235-258.
|Kubany, E. S., Haynes, S. N., Abueg, F. R., Manke, F.
P., Brennan, J. M., & Stahura, C. (1996). Development and validation of the
Trauma-Related Guilt Inventory (TRGI). Psychological Assessment, 8, 428444.
Kubany, E. S., & Manke, F. P. (1995). Cognitive therapy for trauma-related guilt: Conceptual bases and treatment outlines. Cognitive and Behavioral Practice 2, 2361.
Psychological Assessment (1995). Special issue: Methodological issues in psychological assessment research, 7, Washington, DC: American Psychological Association.
Schacter, S. (1964). The interaction of cognitive and physiological determinants of emotional state. In L. Berkowitz (Ed.), Advances in Experimental Psychology (Vol. 1, pp. 4980). New York: Academic Press.
Staats, A.W. (1975). Social behaviorism. Homewood, IL: Dorsey Press.
The McGraw-Hill Companies, Inc..