Abstract
Recent research has suggested that there may be an intergenerational effect associated with exposure to extensive or long-term trauma. With Holocaust survivors as its major focus, this research has reported symptoms associated with posttraumatic stress disorder (PTSD) in children of survivors, including depression, mistrust, aggression [1], intrusive symptoms and hyperalertness [2], emotional numbing and isolation, and symptoms specifically similar to those of their parents, such as nightmares with Holocaust imagery [3]. Other research has explored the transgenerational impact of war trauma on children of USA Vietnam veterans. Several clinical and empirical studies [4–10] have reported lower selfesteem, poorer family functioning and emotional and psychiatric disturbances in both wives and children of Vietnam veterans with PTSD. In Australia, a recent small clinical pilot study of Vietnam veterans with PTSD [11] found that although their children (n = 22) did not differ significantly from controls (n = 14) on measures of selfesteem or psychological disorder, there was a small number of children who had very low self-esteem, and some who had a high level of dysfunction. Further, there was a non-significant trend for them to report stress symptoms similar to those of their fathers. Both these children and their mothers reported more characteristics of family dysfunction (less cohesion and expressiveness and more conflict) than controls. Notably though, this small study did not use a non-PTSD Vietnam veterans control group, nor was the community control group matched in any way to the target group.
The most salient feature of the research findings on this issue to date is, however, the heterogeneity of adjustment found in war trauma survivor offspring. In some studies children are found to be functioning very well, while in others they are found to exhibit significant psychiatric disturbance. Perhaps the findings by some researchers of normal adjustment and resilience, and alternate findings by others of psychopathology and maladjustment in both survivors and their children, should be recognized as indicators of the heterogeneity of survivor responses to trauma [12, 13]. However, that so many reported symptoms in children resemble the posttraumatic symptoms of their traumatized parents (usually fathers), the link between parental Holocaust survival and PTSD in Israeli combat veterans [14], and the correlations between PTSD symptoms in Holocaust parents and children found by Yehuda and associates [13], all support the case for a transgenerational transmission of posttraumatic symptoms.
While the nature and aetiology of the process of such transgenerational transmission of the effects of warrelated trauma are unresolved, the current study aims to explore this phenomenon in the offspring of the largest surviving combat-trauma group in Australia: Vietnam veterans. The study by O'Toole et al. [15] found high lifetime prevalence rates of PTSD (18.7%) among 641 Vietnam veterans when the DSM-IV criteria were used. The Morbidity of Vietnam Veterans Survey by the Department of Veterans Affairs [16] found an even higher rate of 31% in its sample of more than 40 000 male and female Vietnam veterans. These prevalence rates are much higher than the mental health and wellbeing profile of adult Australians [17] would predict (2.3% of men and 4.2% of women, using ICD-10 criteria). The lifetime prevalence of PTSD in the general USA population is estimated to be about 7.8% [18].
Although children of Australian Vietnam veterans may have grown up with PTSD symptomatology in the family, little is known about the adjustment of this group. Westerink and Giarratano [11] investigated children's psychological distress, self-esteem and perceptions of family functioning, but they did not specifically investigate PTSD. Further, they did not include a non-PTSD Veteran control group in their study, and their community control group was poorly matched to the PTSD veteran group. Therefore, the present study aims to examine the self-esteem, posttraumatic stress symptomatology and ratings of family functioning within Vietnam veterans and their children, with particular attention paid to the role of their father's PTSD status in the children's adjustment. It will use a matched community sample as a control group, and the presence of PTSD symptomatology will be used to categorize the veteran group into PTSD and non-PTSD subgroups. Thus, there will be three samples at the adult level and three samples at the offspring level. Due to its size and purpose, this study is exploratory in nature. However, it is expected that veterans with PTSD symptomatology will show poor adjustment, and that their children will show similar patterns. On the other hand, veterans without PTSD symptomatology, and their children, are not expected to show (significant) differences on any of the measures when compared with the control groups. Based on this premise, this study has three hypotheses:
1. Children of veterans with PTSD will have lower self-esteem, increased stress symptomatology, and will report poorer family functioning than either children of civilians or children of Vietnam veterans without PTSD.
2. Children of Vietnam veterans without PTSD will not differ significantly from the civilian children on any of these variables.
3. On measures of self-esteem and posttraumatic stress symptomatology the scores of the children of veterans with PTSD will correlate with the scores of their fathers.
Method
Participants
One hundred and ninety-seven subjects, consisting of four initial groups participated in the study. Group 1 consisted of 59 male Vietnam veterans recruited via notices in national newsletters of three Vietnam veteran associations and a mail-out from one of these association inviting members to return a tear-off slip or to telephone the principal researcher to volunteer their participation. Using a snowballing technique which is usually recommended for sampling rare populations [19, 20], each member of group 1 was asked to recruit his oldest willing child to the study. Group 2 thus consisted of 55 children of the members of group 1. The members of group 1 also recruited Group 3 and 4 members using a snowballing technique. Group 3 consisted of civilian men (n = 44) of a similar age to the veterans, and group 4 consisted of eldest children of group 3 members (n = 39), of a similar age to the members of group 2.
A risk in the snowballing technique is that the participants choose other participants similar to themselves, thus biasing the sample and minimizing the group differences. However, should this be so, and subsequent analyses produce significant differences, these group differences would be more meaningful. Anecdotal comments from veterans in this study suggested that brothers, brothers-in-law, neighbours or work colleagues were chosen for the civilian participants. Veterans were aged from 46 to 60 (mean = 51.6, SD = 3.7), the civilian adults were aged 43–62 (mean = 50.0, SD = 4.8), and both the children of veterans and the civilian children were aged 14–34 (mean = 23.4, SD = 4.5 and mean = 22.3, SD = 4.6 respectively).
Materials
Each participant completed a set of questionnaires, which varied according to group, but contained a set of demographic questions and scales assessing family functioning, self-esteem and posttraumatic stress symptomatology. Veterans also completed a section on combat and service history. The scales are described below.
The Family Assessment Device (FAD), version 3 [21] consists of 60 positive and negative descriptions reflecting the six different subscales of the McMaster Model of Family Functioning: problem solving, communication, roles, behaviour control, affective responsiveness, affective involvement; and an additional composite subscale, global functioning. Respondents indicate whether they strongly agree (1), agree (2), disagree (3) or strongly disagree (4) with each statement.
The Rosenberg Self-esteem Scale [22] is a 10-item Guttman scale. Respondents indicate whether they strongly agree, agree, disagree or strongly disagree with each item. Scores range from 10 (very low selfesteem) to 40 (very high self-esteem).
The Mississippi Scale for PTSD (M-PTSD) [23] is a 35-item selfreport scale with a series of 5-point Likert responses, such as ‘not at all true’ to ‘almost always true’; or ‘never’ to ‘very frequently’. The veteran group completed the combat-related version and all other participants completed the civilian version [24].
The Combat Exposure Scale (CES) is a self-report scale devised to measure combatants’ subjective experience of war stressors [25]. It has seven items in a Likert-type structure, with items weighted differentially according to the severity of the combat experience. The researchers added a question (marked ‘optional’) to this section asking veterans whether they had ever received a diagnosis of a service-related disability and the nature of the diagnosis.
Acceptable levels of reliability and validity are well documented for all of these scales [21–28].
Procedure
Four similar, but not identical, questionnaires were devised and differentiated only by a code and colour to identify type of the questionnaire (veteran, veteran child, male civilian and civilian child) and to which of the four groups each participant, numbered 001–250, belonged.
Packages containing the four questionnaire sets were posted with a covering letter to the veteran. The letter explained the study and gave instructions to the veteran on how to recruit the other participants, which set of questionnaires to distribute to each of them, and which to complete himself. Each set was packaged in a separate sealed envelope and included an explanation of the study and a reply-paid envelope. In accordance with the procedures approved by the Deakin University Ethics Committee, informed consent was assumed for all adults by the voluntary return of the questionnaire and for children was assumed by their father's decision to give them a questionnaire. All questionnaires were completed anonymously, and each questionnaire set took between 20 and 45 min to complete.
Results
One hundred and ninety-seven questionnaires were returned, from 59 veterans, 55 children of veterans, 44 civilian adults and 39 civilian children: a return rate of 60.13%. If any participant was not the father or the child of another participant, their questionnaire was excluded, as was the one female civilian, the two veterans who were childless and the four veterans who left substantial parts of the questionnaires incomplete. Analyses were conducted on the data from the remaining 166 subjects, consisting of 50 veterans and 50 of their children, 33 civilians and 33 of their children. Missing data were analysed on adult and children groups separately, and the few cases of missing data found were spread evenly across the variables, the subjects and the cells. Therefore, all were recoded with the appropriate cell mean.
The veteran and civilian groups at each generation were similar on many demographic variables, with the most noticeable differences being related to current work status, education and occupation. A large number of veterans were retired and not in the labour force (35%); and both veterans and their children were more likely than their civilian counterparts to be unemployed (10% vs 0% for the adults; 12% vs 10% for the children). About two-thirds of civilian children were employed full-time, compared with about 50% of veteran children. For the older generation, civilians were generally more highly educated than veterans (27.3% of civilians had tertiary qualifications compared with 8.4% of veterans), while the trend was reversed although weaker for the younger generation (22.4% of veteran children compared with 15.2% of civilian children had tertiary qualifications).
All veterans completed the only optional question which asked whether the veteran had ever been diagnosed with a service-related disability, and if so, what was (were) the diagnosis (diagnoses). Extensive information was given, indicating a predominance of PTSD in the sample, and demonstrating that most of the current sample had at some time sought clinical assistance. Such information meant that veterans and their children could be categorized by the father's previous diagnosis of PTSD rather than the score on the M-PTSD. This decision was validated by the significantly higher scores for the diagnosed group on the M-PTSD later found by a one-way analysis of variance. One veteran who did not indicate a diagnosis of PTSD was placed in the PTSD category on the basis of his M-PTSD score exceeding the recommended cut-off of 107. In total, 30 veterans were allocated to the PTSD group, and 20 to the non-PTSD group.
The scores for the adult groups are presented in Table 1, which also shows the results of a MANOVA carried out on these data to confirm that there were differences in the adjustment of the groups. This analysis revealed a significant global difference among the three adult groups (Pillai's F = 5.56, df = 2,80, p < 0.01, η2 = 0.38, power = 1). Univariate F-tests revealed the groups differed significantly (p < 0.01) on measures for posttraumatic stress symptomatology, self-esteem and four measures of family functioning: affective involvement, affective responsiveness, communication and problem solving. There were no significant differences found between any adult groups on the two remaining measures of family functioning, behaviour control and roles. The Global Functioning scale, a composite scale of components of all six subscales of the FAD could not be included in the MANOVA due to the multicollinearity it presented.
Means (and standard deviations) for measures of PTSD symptomatology, self-esteem and family functioning variables by adult group
Separate one-way analyses of variance with concurrent Tukey's HSD post hoc comparisons were conducted to determine the source of the significant differences. The results were not surprising: both veterans without PTSD (n = 20) and civilians (n = 33) had significantly higher self-esteem (F = 14.92, df = 2,80, p < 0.0001) and significantly lower posttraumatic stress symptomatology than veterans with PTSD (n = 30; F = 86.52, df = 2,80, p < 0.0001). On measures of family functioning, veterans with PTSD perceived their families as significantly less healthy than did the other two groups on their respective families’ problem-solving effectiveness within and outside the family (F = 14.28, df = 2,80, p < 0.0001) and the ability of individual family members to respond with appropriate affect (F = 14.16, df = 2,80, p < 0.0001). Veterans with PTSD rated their family communication as significantly more indirect, vague and less healthy than did either veterans without PTSD or civilians (F = 10.46, df = 2,80, p < 0.0001). Veterans with PTSD also rated their family members as significantly less interested and involved in each other's activities than either veterans without PTSD or civilians (F = 5.38, df = 2,80, p < 0.01).
Indeed, as shown in Table 1, veterans with PTSD rated their family functioning at clinically dysfunctional levels on all measures except behaviour control, and although it might be expected that veterans do not find establishing and maintaining family rules problematic, the power and effect size for this finding are both poor so that no conclusions can be drawn. The findings of family dysfunction in other areas are strongly supported, except for roles, by both the power and the effect sizes. Some caution needs to be exercised with the interpretation of the subscales of family functioning as several are highly correlated. Therefore, the findings of significance for affective responsiveness and problem solving do not represent independent variance for each of these variables given the variance each shares with communication.
The findings for the offspring groups are presented in Table 2. A MANOVA carried out on this data to test hypotheses 1 and 2 revealed a significant global difference among the groups (Pillai's F = 1.75, df = 2,80, p < 0.05, η2 = 0.16, power = 0.92). Univariate F-tests revealed the three offspring groups differed significantly (p < 0.025) on two measures of family functioning, affective responsiveness and problem solving. There were no significant differences found between any offspring groups on the measures of self-esteem, PTSD symptomatology, or any of the remaining three measures of family functioning: behaviour control, communication or roles. However, scores were generally in the direction predicted, with the PTSD group (n = 30) scoring lower on self-esteem and higher on PTSD symptomatology and family functioning measures (except behaviour control), and the non-PTSD group (n = 20) and civilians (n = 33) having similar scores across all measures. All three groups rated communication in their families as unhealthy.
Means (and standard deviations) for measures of PTSD symptomatology, self-esteem and family functioning variables by offspring group
Separate one-way analyses of variance with concurrent Tukey's HSD post hoc comparisons were conducted on the two significant variables to determine the source of the differences. Children of veterans with PTSD were found to have significantly higher scores (mean = 2.53) on the measure of problem solving than civilian children (mean = 2.15, F = 4.59, df = 2,80, p < 0.025). With a cut-off score of 2.2 used to screen for health–pathology on this measure, children of veterans with PTSD rated their families at a clinical level of dysfunction, and civilian children rated their families as healthy in their respective abilities to resolve problems effectively both within and outside the family. Children of veterans were also found to have significantly higher scores (mean = 2.73) than civilian children (mean = 2.29) on the measure of affective responsiveness (F = 4.59, df = 2,80, p < 0.025. Although all three groups (mean = 2.31 for non-PTSD veteran children) rated their families at unhealthy levels (greater than 2.2), the rating of children of veterans with PTSD was at a clinical level [28].
A one-way analysis of variance with concurrent Tukey's HSD post hoc test was conducted also to compare the three offspring groups on the measure of global functioning, and a significant difference between all three groups was found (F = 6.65, df = 2,80, p < 0.01). Based on the cut-off score of 2.0, the children of veterans with PTSD rated their families as dysfunctional (mean = 2.38), followed by the children of veterans without PTSD who rated their families as borderline dysfunctional (mean = 2.03), while the civilian children rated their families as functional (mean = 1.96).
The third hypothesis predicted that scores for offspring of veterans with PTSD on measures for self-esteem and posttraumatic stress symptomatology would correlate with their fathers’ scores. However, no correlation was found between the self-esteem and M-PTSD scores of the veterans with PTSD and scores of their children. Subsequent correlations for the scores of the other two father–child groups found that neither of these correlated significantly, nor was a significant correlation found for the scores on these measures of the total sample of fathers and their offspring. The results are summarized in Table 3. A measure which registers scores for individual PTSD symptoms might have identified specific symptom correlations between father and child in the PTSD group, however, this study suggests that there is no association between the veterans’ overall levels of PTSD symptomatology or self-esteem and those of their children.
Father-child correlations of mean scores on measures of self-esteem, and PTSD symptomatology by group and PTSD status
Discussion
The results confirm that the PTSD veteran group was less well-adjusted than either of the control groups on measures of self-esteem, stress symptomatology and family functioning. As expected there were no differences between the civilian and non-PTSD veteran group in these domains.
The first hypothesis predicted that there would also be significant differences between the PTSD and other offspring groups on the measures used. This hypothesis was only partially supported: PTSD in the veteran father made no significant difference to his children's score when compared with the other two groups of children on the measure of posttraumatic stress symptomatology or the measure of self-esteem. However, the hypothesis was supported with regard to two aspects of family functioning, affective responsiveness and problem solving, and the overall measure of global functioning. For the latter, PTSD veteran offspring rated their families’ global functioning to be dysfunctional, and significantly worse, than did the non-veteran offspring, who viewed their families’ global functioning to be borderline. In turn, this was significantly worse than the civilian offspring who rated their families as functioning normally.
The second hypothesis, that the children of non-PTSD veterans would not differ significantly from civilian children on any measure, was met for all variables, except for global functioning, as mentioned above. Interestingly, all groups of offspring recorded equally elevated scores for communication. While unhealthy communication patterns would not be unexpected in a family coping with PTSD in one of its members, it is of concern that in this study all offspring groups reported dysfunctional (indirect and vague) communication in their families.
A feature of the results for offspring in the PTSD group on both the measure of self-esteem and the M-PTSD is the wider range of their scores compared with scores for either of the other offspring groups, thus reflecting the heterogeneity of the PTSD group of offspring. Indeed Westerink and Giarratano [11] reported such variability in the scores of their sample of children of Vietnam veterans, which although not significantly different from controls, included a number of very low self-esteem scores and scores indicating a high level of distress. This heterogeneity of the offspring groups may explain the lack of significant differences between M-PTSD scores of children of veterans with PTSD and scores for either of their peer groups. Yet such heterogeneity does not appear to have affected the adult samples. Danieli [3] has explained this kind of heterogeneity with the assertion that members of culturally similar groups who have lived through a similar traumatic experience cannot be expected to demonstrate the same responses to the trauma. Such heterogeneity must increase within the next generation for which the traumatized parent is simply one familial influence.
Several factors may have contributed to some children of veterans with PTSD having very low self-esteem and others having very high self-esteem. For example, age of the offspring studied may be relevant; however, the number of subjects in this study did not enable separate analyses on children subgrouped according to their age. Previous studies of Vietnam veteran children [6, 9, 10, 29] examined adolescent or younger children, ages at which children may have less resilience against family dysfunction and have not been through many potentially esteem-building stages of development and achievement, including higher education, work, relationships and independent living. Children of veterans in the present study were more likely to have been or be completing tertiary degrees and less likely to be employed than their civilian peers. Such variables may further explain the diversity in the self-esteem scores of the children of the veterans with PTSD. This does not however, explain why some USA studies [6, 9] have found lower self-esteem in the children of veterans with PTSD and this study did not. However, since the current findings are consistent with the only other reported Australian study, there may be differences between the USA and Australian Vietnam veterans, including fundamental character differences, the differing nature of their traumatic war experiences and postwar readjustment, including level of substance abuse or violence and marital status. Any of these may have influenced significantly their children's self-esteem.
The third hypothesis in this study was not supported. There was no evidence of a relationship between the self-esteem and posttraumatic stress scores of the PTSD veterans and those of their offspring. Thus, rather than support the notion of transgenerational transmission of posttraumaumatic stress symptomatology, the present study adds to the body of evidence that disrupted family functioning is the most consistent outcome for trauma survivors, particularly those with PTSD [6, 7, 11]. Hallmark characteristics of PTSD become evident in the nature of the disruption of family functioning. Emotional numbing in the parent with PTSD creates a barrier to effective communication [4]. That both fathers and children in the PTSD group in the present study reported poor ability for members of their families to experience appropriate affect supports Harkness’ [6] assertion that the strongest effect of PTSD is the emotional inaccessibility of the father. Similarly, over half the veterans with PTSD in the study by Jordan et al. [7] described their family functioning as poor on both adaptability and cohesion, and both children and partners of Vietnam veterans with PTSD in the Westerink and Giarratano [11] study reported high levels of family dysfunction.
Pilot studies often raise more questions than they answer, and the present study is no exception, providing several openings for future research into the issue of intergenerational transmission of the effects of war-related trauma. First, aspects of methodology could be improved. The transgenerational impact of posttraumatic stress may be a vulnerability to non-specific disorders, and more general measures of mental health could be used in further studies. In addition, the sample in this study has limitations in that the veterans’ children and the control groups were recruited using a snowballing technique. While this may have helped achieve matches in age and socioeconomic status across the groups, it may have also led to a matching in dysfunction and disorder, thereby reducing cross-group differences in the variables of interest. Further, bias may have resulted both from the inclusion of only one child per family, and that this child might have been the only one with whom the veteran had a satisfactory relationship.
A larger study could focus on all children from each family unit. Also, comparing those children currently living with their fathers with those who live independently might lend some insight to the persistence or dissipation of the dysfunction in children. Almost all of the previous studies [4–10] examined children living with their parents. Finally, the role of the wives, and their reactions to their husbands’ PTSD, are also unknown factors in the families studied, although Westerink and Giarratano [11] suggest that wives of PTSD veterans often act as a buffer between the veteran and the rest of the world. Interestingly, the majority of veterans in the current sample are still married to their first wife. Whether it is the stability of this family unit, or the roles these women played in providing a buffer zone between the children and the most difficult of their husbands’ behaviours, this study cannot answer. Certainly these questions are worth addressing in a more comprehensive study including sibling groups and mothers/wives. The dysfunction in families associated with a father's PTSD may have serious costs for the veteran community and their social network.