Posttraumatic Phenomena in a Longitudinal Study of Children Following aNatural Disaster

ALEXANDER COWELL McFARLANE, M.B., B.S., DIP. PSYCHOTHER., F.R.A.N.Z.C.P.

Abstract. This longitudinal studyexaminedthe prevalence of posttraumaticphenomenaand howthey relateto symptomatic and behavioral disorders in a population of schoolchildren exposed to an Australian bushfiredisaster. The prevalence of these phenomena did not changeover an 18-monthperiod, suggesting that they weremarkersof significant developmental trauma. The mothers' responses to the disasterwere better predictors of thepresence of posttraumatic phenomena in children than the children's direct exposure to the disaster. Both theexperience of intrusive memoriesby the mothers and a changed pattern of parenting seemed to account for thisrelationship. J. Amer. Acad. Child Adol. Psychiat.. 1987, 26. 5:764-769. Key Words: posttraumatic disorders,disasters, parenting, lifeeventsstress, imagery.

The effects of extreme adversity, such as disasters, onchildren's psychological development can be viewed fromseveral perspectives. First, children may develop psychiatricsymptoms or behavioral disturbances (Burke et al., 1982; Zivand Israeli, 1973). These may be documented using question­naires and structured interviews and are therefore the favoredfocus of epidemiologists. Alternatively, nightmares and recur­rent memories of the traumatic event as well as play centeredon the trauma may indicate attempts to master the experience(Newman, 1976; Terr. 1979). The clinical method has beenthe main avenue for investigating these posttraumatic phe­nomena. For example, Terr has extensively studied the vic­tims of the Chowchilla school bus kidnapping using thistechnique (Terr, 1979; 198Ia). Thus the first method hasfocussed on readily observable disturbances, whereas the sec­ond has investigated the psychodynamic process ofadaptationin the child, which is more difficult to document reliably,although it is an equally valid determinant of the child's long­term adjustment (Newman, 1976).

Terr (l98Ia) has emphasized that there are a number ofovert posttraumatic phenomena that are indicative of chil­dren's failure to adapt to a traumatic event. Clinicians havelong noted such phenomena in children observed to be clini­cally disturbed after a traumatic event (Bloch et al., 1956;Bodman, 1941; Carey- Trefzer, 1949) but have not systemati­cally studied their relationship to symptom formation. Ex­amination of this association is important for three reasons.

First, the clinical significance of symptoms experiencedafter extreme adversity may be quite different from thoseobserved in a normal clinical setting (Mcfarlane. 1985). Forexample, most reports of children exposed to threat of death(e.g., by bombing or shelling in war) indicate that there isapathy or decreased activity (Dunsdon, 1941) and no evidenceof increased anxiety or panic (Ziv and Israeli, 1973). Hencesuch children could score low on most symptom counts

Received July 7, 1986; revised Nov. 25, 1986; accepted Dec. 10,1986.

Dr. McFarlane is a Senior Lecturer in Psychiatry. Department ofPsychiatry, The Flinders University ofSouth Australia, Bedford Park.South Australia 5042.

The assistance of V. Blumbergs. S. Policansky, M. Higgins, H.Schinkel. Dr. C. M. McFarlane. and Dr. C. Irwin, and the cooperationof the headmasters, teachers, and parents of the schools studied, isgratefully acknowledged. This project was partially supported by theFlinders Medical Centre Research Foundation and the Channel 10Medical Research Foundation.

0890-8567/87/2605-O764$02.00/0© 1987 by the American Acad­emy of Child and Adolescent Psychiatry.

despite their difficulties. Therefore, the study of posttraumaticphenomena may be a valid and sensitive measure of theeffects of psychic trauma on children, in addition to obser­vations of symptomatic disorder.

Second, whereas Davidson et al. (1985) have suggested thata concurrent psychiatric disorder can intensify posttraumaticimagery, Horowitz (1973) and others (Brett and Ostroff, 1985)have emphasized the important role that posttraumatic phe­nomena play in the etiology and maintenance of symptomsafter exposure to extreme adversity. Therefore, the possiblereciprocal association of posttraumatic phenomena withsymptom formation in children after a natural disaster islikely to be of particular interest in assessing levels of morbid­ity caused by such an event and the etiological process in­volved.

Third, investigation of the etiology of posttraumatic phe­nomena in children exposed to extreme adversity could helpclarify the mechanisms of symptom formation in such set­tings. Previous research suggests several possibilities. The in­direct effect of the parents' reaction to the trauma has previ­ously been noted to be of primary importance (Ziv and Israeli,1973). Rosenbeck and Nathan (1985), for example, describeda child who developed the same posttraumatic imagery as hisVietnam veteran father when the child had obviously neverbeen exposed to the war. Similarly, one child who has beenexposed to a traumatic event can trigger posttraumatic behav­ior in unexposed children by involving them in posttraumaticplay that powerfully depicts the repeat anxiety of the trauma(Terr, 1981b).

Parents could convey this anxiety to their children bybecoming overprotective, a common parental response indisaster (Silber et al., 1958) and concentration camp victims(Phillips, 1978). Thus, parents' ability to contain the anxietygenerated by an extreme threat may be the major factorinfluencing their children's responses.

Against this background, this project aimed to investigatelongitudinally the psychological morbidity in a population ofprimary school children who were exposed to an Australianbushfire disaster. In this paper, the prevalence of posttrau­matic phenomena and their relationship to symptoms andbehavior problems are reported. The determinants of theseposttraumatic phenomena are also examined.

In particular, it was hypothesized that posttraumatic phe­nomena in children would be related more to posttraumaticmorbidity in their parents and overprotective parenting thanto the direct effects of the disaster, such as property loss,bereavement, or the physical threat of death or injury. Chil-

764

POSTTRAUMATIC PHENOMENA FOLLOWIN(; A DISASTER 765

dren who were separated from their parents in the first 3 daysafter the fire were hypothesized to be at risk of having a greaterdegree of preoccupation with the disaster in the long run.because parents playa major role in the containment ofchildren's distress immediately after extreme trauma.

Method

The Sample

This project aimed to study all the children from one well­demarcated area devastated by bushfire on February 16. 19X3.In this region. 120.000 hectares of agricultural and plantedforest lands were burnt. 200.000 stock were killed or de­stroyed. 14 people died. 40 homes were destroyed. manyothers were damaged. and 359 farming properties were fullyor partially devastated. The fire was extremely intense: flames800 feet high were measured and the cyclonic strength windssnapped the trunks of mature radiata pines.

The sample consisted of 808 children who attended the 6state primary schools serving the region. All but one schoolhad been physically threatened during the fire. The childrenwere studied on three separate occasions. 2. 8. and 26 monthsafter the fire. The complete sample was not studied on eachoccasion. with two schools not being examined in the thirdstage (N = 365). The sampling procedure is described in moredetail elsewhere (Mclarlanc et al., 1(87).

Quest il nina irc

The parent and teacher questionnaires. developed and val­idated by Rutter and associates (Rutter et al., 1970: Rutterand Graham. 1%7). were administered on each occasion.These instruments were chosen because they had been previ­ously validated in an Australian rural setting (Connell ct ul.,1(82). In the 8- and 26-month stages. a further 4 and 3questions. respectively. were added to the parent and teacherquestionnaires inquiring about posttraumatic phenomena.They were scored in the same way as the other items of thequestion nai res (doesn't apply. applies somewhat, cert0 inlvapplies). The last two replies were interpreted as meaning thatthe phenomena were present. The items were selected on thebasis of observations of previous reports (Bloch et al., 1956.Bodman. 1941. Carey-Trcfzcr. 1949: Newman. 1976: Terr.1979: Terr. 198 Ia) during the first stage of the project andreports of parents in another clinical setting (Mcf-arlane.1(86). In the parents questionnaire. the items were: I. l l asdreams Of nightmares about thetire. 2. .'11 limes plays gamesabout the tireOf paints pictures about it. 3. Is 11/1.\1'1 Of worriedbv reminders o( tlu: [ire (c.g.. sirens. strong winds. ctc.), 4.Spontaneouslv talks about thetire. In the teachers question­naire. items 2 and 3 were combined to read. ('01111111/1'.\ 10

plav. talk. Of paint about thelire. These questions had estab­lished face validity.

To investigate the reliability of these items. all the teacherswere interviewed when the S-month questionnaires were col­lected. and problems about these items were discussed. Insummary. affirmative responses appeared to be highly reliable.but there was potential for a substantial false negative re­sponse. because teachers may not have been aware of someof the children's experiences.

The 2-month questionnaire also included 31 questions that

recorded the extent of personal and property losses. exposureof the child and parents to the fire. the duration and natureof the separation between the children and parents duringand after the fire. as well as the nature and site of temporaryaccommodation. The 8- and 26-month questionnaires docu­mented the continuing impact of the disaster on the parentsand their livelihood. The extent of reconstruction. parentalattitudes to losses. intrusive thoughts and memories about thefire. family functioning (measuring irritability and withdraw­al). (Mcf-arlane, 1(87). and ovcrprotcctivcncss toward thechildren were recorded. The details. scoring. and raw data ofthese items are to be reported in more detail elsewhere. At 26months. the parents were asked to record whether any of 14life events likely to have an effect on the children's adjustmenthad occurred in the past 18 months. These items were takenfrom a children's life events inventory (Monaghan ct al.,1(79). For analytical purposes. the number of life events wasadded with no wcightings. a method shown to be acceptablein other life events research (Tausig, 1(82).

Statisticalstcthod

To investigate the effect of time on the prevalence ofposttraumatic phenomena. the mean score for the individualitems and total scores on the parent and teacher scales werecompared at 8 and 26 months using paired I tests. Therelationship between posttraumatic phenomena and symp­toms was then investigated using several methods of analysis.

First. the prevalence of posttraumatic phenomena in chil­dren defined as "cases" was compared with those who werenot cases at 8 months. as defined by the Rutter questionnaire(Table 2).

Next. the nature of continuous relationship between symp­toms and posttraumatic phenomena was examined. usingPearson product moment correlation coefficients (Tables 3and 4) and step-wise multiple regression analysis.

The longitudinal relationship between the emotional andbehavioral phenomena observed at 2 months and subsequentposttraumatic item scores was then examined using cross­lagged correlational analyses. This allowed an investigation ofthe degree to which posttraumatic phenomena observed inclose proximity to a disaster might predict subsequent emo­tional and behavioral problems, an important test of theetiological role these phenomena might play in the onset ofpsychiatric impairment.

The etiological association among the children's experienceof the disaster. its effects on their parents and families, andthe children's posttraumatic phenomena observed by the par­ents was next examined using Pearson's product momentcorrelation coefficients (Table 5), Multiple regression analysiswas then used to examine these etiological relationships inmore detail because of the interdependence of many of thedisaster related variables. All the variables listed in Table 5were included in the regression equations according to thetime period being investigated. Only those variables thatloaded significantly in the regression analysis arc reported (/1

<0,(15).The data were analyzed using the Statistical Package for

the Social Sciences (SPSS lnc., 19X31. and a probability levelof less than (l.05 was taken to be significant.

7(j(j M(TARLANI'

Results

......amplr Characteristics

The mean age of till: XOX children was X.2 years (S.D. =

2.2), 52.X"; were boys, and 94. I"; ofthe parents were married.The impact of the disaster on these children's families wassuhstantial: J2"; ( 150) had sustained property damage: 25";(IIX) of lathers. 1.1";, (60) of mothers, and X";, (J7) of thechildren had had an intense exposure to the fire or come closeto death; 27"; (12X) were bereaved: and 25";. (IlX) of thechildren had been separated from their parents for up to Jdays after the fire,

Return rates of 4J.2";" 57.6";, and 55.5"; were obtainedon the parents' questionnaires at 2, X, and 26 months, rcspcc­tivcly. No significant sampling bias was found to exist whencomparing the teacher scores for children whose parents rc­sponded with till: scores for nonrcspondcrs (Mclarlanc ct al.,I ()X7).

1'1'1'1'11/1'1/1'1' (1(1'( 1111ra11I1/1I1ic l'hctu 111/1'/11/

The prevalence of posttraumatic phenomena in these chil­drcn at Xand 26 months is shown in Table I. The prevalencehad not decreased in this period as measured hy either theparent total (paired I = 0.15, /' = O.X) or the teacher total(paired I = (U4, /1 = O.n

R1'11i1iOI/I/i ifl o( 1'0,111 ra /l1I/1I1 ic 1'11l'1I01l/1'1/1i 10 .\'.\'11//11011/1

The prevalence of posttraumatic phenomena in childrendefined as cases by the parent and teacher Rutter question­naircs is shown in Table 2. With one exception, posttraumaticphenomena were mort: common in children defined as cases,This was not so tin the prevalence of posttraumatic phcnorn­cna rated by the parents in children defined as cases by theteacher questionnaire (t = 1.90,/1 = 0.(7)

The correlational relationships between crnotional/bchav­ioral problems and posttraumatic phenomena art: reported inTable J for the parent items and in Table 4 for the teacheritems. The parent posttraumatic items correlated very signif­icantly with the parents' symptom scores both cross-section­ally and longitudinally (Table J). However, the S-rnonth totalsymptom score did not correlate with the 26-month posttrau­matic item score (I' = O.10, /1> 0.(5).

In contrast. the x-rnonth posttraumatic items as scored hythe teachers did not significantly predict the 26-month teachersymptom score (I' = (l.() I. /1 > (l.05, Table 4), whereas the X­month teacher symptom score was highly correlated with the26-month posttraumatic item score (I' = 0.60, /1 < 0.00 I). Ahigh degree of cross-sectional association existed (Table 4).

Using multiple regression analysis, X months after thedisaster, the parent symptom scores accounted for 19'J;, of thevariance of the posttraumatic phenomena, whereas this had

TAIlII I. /'U'\'II/CII('(' ott'osuraunuu« /'111'1101111'1111 IIl1d Ruttcrl'arcnt und Teacher Scoresa! 8 1I11c1 l(, ,\10111 hs .: l!icr Disaster

No. of subjects with parent question­naires

Total parental posuraum.uic

Dreams( iarncsReminders upsetTalks aboutRutter Parent scorv

No. of subjects with teachers question­naire

Teacher totalPlays, talksDreamsAfraid of sirensRutter Teacher score

Mean (S.D.)

1.2(, (1.1,7)

0.16 (0.44)0.16 (0.44)0.44 (0.6'1)

0.'1 1(0.6'1)6.9 (6.2)

0.4:1 (0.94)0.07 (0.22)0.04 (0.22)0.0.1 (0.20)

:1.2 (4 ..1)

X Months

S = 467

S = 719

lrcqucncvPercentage

:<~.X(·;

12.9"i,13.Y·i,

.14.9"i,4.1.1 Pi.

»i»:6.1 Pi,

.l6"i,3.~(';

2t> Months

Mean (S.D.)

S = 241

1.4 (1.72)

0.21 (0.47)0.11 (0 ..,4)

0.6'1 (O.7t»0.44 (O.M)7.6 (6.7)

.\' = .,09

0.41 (0.77)

0.02 (0. '-')0.0.1 (l.I X)0.04 (0.22).'-.1 (4.4)

FrequencyPercentage

)7.2";,

IX.4"i,l}.X(;;)

47.4%3'i.9';i,

26.Y'i1.6";,1.7"i,3.7"i,

IAIlII ~ Comparison 0( 'hildrcn n('!illcd III (·<I.\CS bv Rutter Qitcstiimnuircstor /i'IIi/PIIsllralllllllli" /lclIIs 11'1111 ,,\'Ollca,ICS at 8 .\lolIllIs"

/'1'

Cases

Mean (S.D.)

Noncases

Mean (S.D.)I'

Parcnl casesParent posttraumatic itcmsleacher posttraumalic items

Teacher ('asesParent posttraumatic itemsTeacher posttraumalic itl'ms

70

622.41 (2.17)O.IQ(I.Ot»

1.77(2.24)1.42 ( 1.'1'1)

_,47

\!'i

.174MI

1.06 (1.4'1',) 6 ..17 0.0000.4'1', (1.00) 2.4'1', 0.01

1.24 (1.60) 1.90 0.07OAO (0.'1',0) '1',.67 0.00

" ( 'ases defined using 12/1.1 and '1',19 cUlotl" f(,r the parent ano teacher ljuestionnaire, respectively ( 'on nell et al. 19'1',.,).

I'OSTTRAUMA1'1<' PHENOMENA FOLLOWIN( i A DISAS1TR 767

T"'II.I' ~. Pearson Correlation ('oefficients olParcnt PosttraumaticItems and Factors otParcnts Rill/a <!Ill'stionnaires"

PosttraumaticItems.

X Months(i\' = 4(7)

PosttraumaticItems.

26 Months(t\,' = 2J5)

TAIII.I 5, Pearson Product Moment Correlations oiPosttraumaticItem Total Wit]: Disaster and Fumily Related Variables at 2, 8, and

:!r, Month»

X-Month 26-MonthPosttraumatic Posttraumatic

Items Items

TAIII.I· 4. !'('(Ir.\I1// Correlation ('oefficiell/.\ o!' reacherl'osnraumatic Items and Factors o] reacher Rill/en Questionnaires

" Factor structure of Rutter questionnaires is to he reported else­where (Mclarlanc ct al.. 19X7).

*" ~ 0,05; **" ~ 0.01; *** II -s 0.001,

FI i%g.l' or 1'0.1'11ra umat it' Phenomena

As shown in Table 5. there was a strong association amongposttraumatic phenomena in the parents. overprotective par-

risen to 25'};, of the variance IX months later. Phobic symp­tomatology was consistently most strongly predictive of thesephenomena (X months. adjusted R' = 0.14. F = n.3h: 26months. adjusted R' = 0.23. F = 7I,5X).

Further examination of the longitudinal relationship he­tween the Rutter scores at 2 months and subsequent posttrau­matic item scores demonstrated a significant correlation he­tween the 2-month teacher symptom score and hoth the 20­month teacher (r = 0.49. II < 0.(0) and parent (r = (UX. JI <0.(00) posttraumatic items. In contrast, no significant corre­lation existed between the 2-month parent posttraumaticsymptom scores (r = 0.07. JI = 0.5) and the 26-month parentposttraumatic scores.

0,1 '1**0,24***0.16**(UX***(U5***0.2X***O.2~**

(U4***(U7***(UX***

o.n***O.2J**0.11O.~O*

0.1 ~*

0.20**0.20**

0.19**0.20**0.2.1***(U4***0..15***0.1 ~*

(UO***O.2h***

0.17***0.17***0.20***0.2X***(UX***0.1 ~**

0.4.1***(U2***

-0.14*0.15*0.15*0.25***0.1r-­0.19***0.26***

2-month variables (N = 4~0)

AgePersonal lossProperty lossSeparation alter disasterMaternal copingChild's exposureParental exposure

X-month variables (N = 446)Complete reconstructionRegaining incomeResolution of personal lossAnxiety of future tiresMaterial intrusive thoughtsMaternal copingFamily functioningOverprotection

26-month variables (N = 229)Completion of reconstructionPreoccupation with property lossResolution of bereavementMaternal intrusive thoughtsMaternal interfering imageryPaternal interfering imageryMaternal copingFamily functioningOverprotectionLife events

* n ~ O'()5; ** n -s 0.0 I; *** I' -s 0.00 I.

cnting, changed family functioning. parents' and children'sexposure to the tire. and posttraumatic phenomena in thechild.

The correlation between these variahlcs and the individualposttraumatic items. although not presented. was generally ofa magnitude similar to the total. !\ notable exception was thepresence of playing games ahout the lire. which was onlycorrelated with parental coping at 2 months (r = O.IX. I' =0.0 I) and parental ovcrprotcctivcncss at X months (r = 0.25.II = 0.0 I).

When the X-month posttraumatic phenomena were re­gressed against the 2-month fire related variables, the familybeing split up (adjusted R" = 0.11. F = 35.7). parentalexposure (adjusted R' = 0.15. F = 25.2). and age (adjustedR' = 0.16. F = 19.5) accounted for 10% of the variance.Separation of the child from the parents was highly related tothe other variables because this separation was most likely tohave occurred in those families that sustained property dam­age.

Regression analysis of the X-month variahlcs against the X­month posttraumatic items indicated that changed familyfunctioning (adjusted R' = 0.17. F= X3.X). maternal intrusivethoughts (adjusted R' = 0.20. F = 52.4). and continuingunrepaired property damage (reconstruction) (adjusted R' =

0.21. F = 3.(9) all contributed significantly to the variance,

0,100.17**0,12*0.070.21**0,1 '1**O.IX**0.51 ***0.1 X**O.2J***0,17**0.44***O.4X***(UO***

PosttraumaticItems.

26 Months(S = ~(9)

0.60***0.22***0,15**0.46***O.2J***0,60***(UX***0.40***0.56***0.46***

0.50***0.17***0.29***(UO***0.21 ***0.01O,()5

o.oo0.050.03

0.4 7***0.26***0,1 '1***0.26***O.~5***

(1.37***0.25**0 ..16***0.22***0.1 X***O,~ I***O.~6***

O.~7***

0,16**

PosttraumaticItems.

XMonthsC'" = 71'1)

":» -s 0,01; r:» -s 0.001.

Teacher Rutter score. XmonthsFactor I AntisocialFactor 2 NeuroticFactor ~ DistractablcFactor 4 Unpopular. antisocial

Teacher Rutter score. 26 monthsFactor I AntisocialFactor 2 NeuroticFactor ~ DistractablcFactor 4 Unpopular, antisocial

Parent Rutter score. XmonthsFactor I AntisocialFactor 2 RestlessFactor .1 AngryFactor 4 NeuroticFactor 5 PhohicFactor 6 Somatic

Parent Rutter score. 26 monthsFactor I AntisocialFactor 2 RestlessFactor .1 AngryFactor 4 NeuroticFactor 5 PhobicFactor 6 Somatic

MCFARI.ANE

When the 26-month posttraumatic items were regressedagainst the 2-month variables. the famil y being split up (ad­justed R: = 0 .13 . F = 22 .3) was the only variable predictinga significant percentage of variance . Regression analysis ofthe 20-month posttraumatic items against the X-month vari­ablcs indicated that both maternal fears of future fires (ad­justed R~ = 0 .12, F = 20.5) and continuing loss of incomebecause of the fire (adjusted R~ = 0.14, F = 14.2) accountedfor a significant percentage of the variance .

When the 26-month parental posttraumatic items wereregressed against the 26-month variables. intrusive thoughtsof the disaster in the mother (adjusted R ~ = O.ll . F = 29.4),life events in the preceding I X months (adjusted R: = 0 .1 X, F= 23 .20), and family functioning (adjusted R: = 0.21. F =19.6) all contributed significantly to the variance.

Discussion

Approximately one third of the children in this study werefound to have a continuing preoccupation with their exposureto a hushfire 20 months after the disaster. This suggested thatthe meaning and impact of this event were continuing to playa role in the psychological and personality development ofthese children. There appeared to have been little workingthrough of traumatic memories between X months and 26months after the lire-I W}(, of the children continued todream about the tire, indicating the involuntary nature oftheir preoccupation. These posttraurnatic phenomena possi­bly served an adaptive function. For example. playing gamesabout the fire seemed to be particularl y associated with thechild's mother not coping with the tire and being overprotec­tive. rather than with the physical impact of the disaster. Suchpla y may haw been important to the children's attempts towork through an event that had undermined their mother'srole as the mediator between the child and danger.

The data indicated that there was no one-to-one relation­ship between posttraumatic phenomena and psychologicaldisorder as measured by the Rutter questionnaires. Althoughsuch phenomena were significantly more common in childrendefined as cases, they were present in many children who didnot exhibit disorders. Similarly, they accounted for, at themost , 25';;, of the symptomatic disturbance in these children.However, this relatively small relationship may in part hecaused by the diluting effect of the background psychologicalmorbidity that would have existed in this community inde­pendent of the disaster. Thus because these phenomena arenot invariabl y associated with symptoms. they should beinvestigated in any study wishing to document the effects ofextreme adversity on children.

The longitudinal nature of this project allowed a prelimi­nary examination of the causal interaction between symptomsand posttraumatic phenomena. At both 2 and X months afterthe fire, the level of anxiety and behavioral disturbance dem­onstrated hy a child at school hut not at home was significantlyassociated with the intensity of posttraumatic phenomena 26months after the disaster. This suggests that the level ofemotional or behavioral distress experienced hy children soonafter disaster predicts the later presence of posttraumaticphenomena, particularly if the disturbances arc manifest whenthe children arc separated from the reassuring presence oftheir parents (c.g., at school).

In contrast. the intensit y of posttraumatic phenomena oh­served by the parents X months after the disaster was signifi­cantly correlated with symptoms I X months later. This raisesthe possibility that the posttraumatic phenomena observed byparents arc a predictor of subsequent symptomatic disturb­ance, supporting Horowitz's (197J) etiological hypothesis.These combined observations suggest a possible- reciprocalrelationship between symptoms and posttraumatic phenom­ena in children where a high level of anxiety soon after adisaster predicts imprinting of memories of the event, whichsubsequently serves to maintain and predict future sympto­matic disorder.

The findings that separation from parents in the daysimmediately after the tire, continuing maternal preoccupationwith the disaster, and changed family functioning were morepowerful determinants of posttraumatic phenomena in thechildren than were exposure to the disaster or the lossessustained provides some support for the previous observationthat children's responses to traumatic events arc determinedmore hy their parents' attitudes than by the intensity of thedanger experienced (Ziv and Israeli, 1973). This suggests thatchildren 's ability to maintain a harrier against psychic traumais critically dependent upon their parents' response to theevent.

Several mechanisms could explain this process. First. if amother were continuing to experience intrusive and unwantedmemories of the fire , she would lind it very difficult to containher child's anxieties and memories of the disaster. Second,children's preoccupations about the disaster may be deter­mined by their awareness of their parents' recurring imageryof the event. independent of their own experience of the tire(Rosenbeck and Nathan. 19X5).

Third. the mothers with the most posttraumatic imagerywere also the most overprotective. perhaps because of theircontinuing sense of foreboding and vulnerability. Such ma­ternal anxiety is likely to he easily perceived bv children ,making it more difficult for them to resolve their own anxie­ties and heightening their sensitivity to further trauma.Twenty-six months after the fires, adverse life events. unre­lated to the disaster and experienced in the previous IXmonths, were predictors of children's continuing preoccupa­tion with the disaster. Further distressing events may interferewith children's ability to contain the sense of threat causedby the original disaster. as well as im pa iring their adjustmentto subsequent traumas. a phenomenon noted by Newman(1976) after the Buffalo Creek disaster. Such events may alsomaintain the mothers' preoccupation with their vulnerability.indirectly eroding the children's attempts at mastery.

The role played hy the pattern of interactions in the famil ywas examined in more detail using the family function scale,which measured irritahility and withdrawal. two commoninterpersonal manifestations of posttraumatic stress disordersin adults. Although irritability in the parents was of particularimportance as a predictor of their children's posttraumaticsymptoms at X months, the mother's preoccupation with thefire and lear of future fires were of greater significance inpredicting the children's symptoms at 26 months. Th is sug­gests that the direct convcycncc of the parents' traumaticanxiety to the child may he as important as any increasedexpression of hostility and withdrawal in the family in the

POSTTRA UMATIC PH ENOM ENA FO I.1.0 W IN( i A DISASTER 769

et iology of the chi ld's distress. Th is co ntrasts with the findingof Rutter and Quinton ( 19X4) that the patterns of interact ionin a famil y. rather than psychiatr ic d isorder in a parent. aremore im po rta nt determinants of d isorder in children .

Fin all y, 21 I:;,. a t most. of the va riance of the predictors ofposttraumatic phenomena in child ren co uld he acco unted forat an y on e time. Other factors require future in vestigati on .Th e child's own temperament is one factor not examined inthi s stu dy . In particular. the associat ion between the respon sesof mother and child ma y. in part. reflect genetic factorsinfluencing hoth the child's pe rson alit y and his or her an xiet yproneness.

T hese result s need to he considered against the backgroundof methodological problems inh erent in the study. In partic­ular. the reliability of reporting o f posttraumatic items wasnot tested extensively, although suffic ient reliability existed tomake the observations of some value. Furt hermore. the re­spo nse rates were not particularly high . However, no biasappeared to have been introduced hy thi s problem when thetea cher qu esti onnaires were used to examine the characteri s­tics of the nonrcspondcrs. On the other hand, the valid ity ofthe find ings is suppo rted by thei r co nfirma tion of the rep ortso r other cl in ical and anecdotal studies (Bloch et a l.. 1956:Bodma n, IlJ41: Carcy-Trcfzcr, 1941J: New ma n. 1970: Silberct al ., IlJ5X: Terr. IlJ7lJ).

In conclusion , thi s epidem iological study suggests that post­traumat ic phenomena are a n important marker of de velop­mental trauma in latency-aged child ren exposed to extremead versit y. Such phenomena sho uld be systema tically docu­mented in an y future study or the impact of disasters onchild ren as on e measure of mo rb idity.

References

Bloch , D . A.. Silbe r. L & Perry, S. L (I <J5h) , So m e factors in th ee motiona l rea cti on of child re n to d isas ter. .'im rr. .I. l 'svchiat..I 13:416-422.

Rodman . F. ( I<J41). War condition s a nd thy' mental health of th echild . Brit , .\It'd ./.. Oct.: 4X6-4XX.

Bret t, E. A. & Ostroff R. (I<JX5). Im agery and post-traumatic ,tressdi sorder: an overview . . tmcr. .I. Psvchiat .. 142:417-424.

Bu rke . J . D .. Boris. J. F .. Burns, IJ. 1.. ct ul. (I <JX2), Changes inchild ren's behaviour after a natura l d isaste r . . inu 'r . .I. l'svchiut ..I .I<J: 10 I0-1 0 14.

Ca rc v-Trctz er . C. J . (I <J4 <J ). The result s o f a cl inic a l stud v of war­damaged child re n wh o a tte nded th e chi ld guida nce clinic. The

Hosp ital fo r Sick Children, Great O rmond Street, London . .I. ,\ft'II! .Sci .. lJ5:535-55lJ.

Con nel l, II. M .. Irvine, I.. & Rodney. J . ( 19X21. The pr evalen ce o fpsychi atric di sorder in rural school chi ld re n . ..1/1.1'1 . S . /.. .I. 1'.1.1'­chiut.. 16:43-46.

Da vid son, 1.. Swartz, M. & Stack. M . ( I <JX51. A diagn osti c a nd fam il ystudy of post-traumati c st ress d isorder. .:111 /('1'. .I. l 'svchiat .. 142 :<JO­<J3.

Dunsdon. M . I. ( 11)41I. A psych ol ogist 's co n t ribu t io n to ai r raidprobl ems. .\It'II! . II/III .. 2:37-4 1.

Ho ro witz, M . J . ( IlJ7] I. Ph ase orien ta ted treatment o f st ress res ponsesyn dromes. Amrr. .I. I'SI'CIl" llIef .. 27 :506-5 I5.

Mel -arlene. A. C. ( llJX5I, The effects o f st ressfu l lilc even ts a nddi sasters: research and th eoret ical issues . . /111I. S. /.. .I. Psvchiut ..IlJ:40 lJ-42I .

-- ( I<J X6) , Posttraumatic morbidity of a d isaster. A study of casespr esent ing for psychiatric trcatrn cn t,'. Ncrv. .vtcnt, I);, . 174:4-1 4.

-- (1 1)X71, Family functioning and o verpro tec t io n following a nat ­ural disaster: the longitudinal effec ts of posttraumatic morbid ity,Au st . N . Z. J. Psychiat., 21 : 2 10- 2 1X.

-- Policansky. S. K. & Irw in. C. I. (1 9 X7), A longitudinal study ofth e psychological morbidity in chi ld re n due to a natural di saster.l 's vchol . Mcd.

M onaghan , J . II .. Robinson. J . O. & Dodge, J . A. (1 1) 71) ), Thech ild re n's 1& events inv ento ry . J . Psychosorn . Re s.. 23 :63- 6 X.

Newman, C. J . ( llJ76) . C hild re n o f disaste r: cl ini ca l obs ervat ion s atBuff alo Creek. .·I lIl ef . .I. l'syrh iat .. 133:306- .\ 12.

Phillips, R. E. ( llJ7XI, Impact of Nazi hol ocaust o n ch ildren ofsurvivors. .·l ll/ ef . .I. l 'svrhiat .. 32 :.nO-37X.

Rosenbeck . R . & Na tha n, P. ( I<JX5), Scconda rv t ra u rna uzut io n inchi ld re n of Vietna m vete ra ns . 110.1('. ( 'o lllll1l1/lIi l' l 'svchiat.. 36 :53X­531).

Rutter. M . & G ra ha m P. ( 1'1(,7 ), A children 's behavi our questi onnai refor co m plet ion by tea chers: prel im inary findin gs. .I. Cliild Psvcho).l'svchiat .. X: I-I I.

-- & Q uin to n. D . (1 I)X4), Pa rental psyc hia tric di sorder : effect s o nchi ld re n. 1'.ITClio/. stcd., 14:X53-XXO.

-- T iza rd , 'J. & Whitmore. K. ( 1lJ701. Education. I/m /III andltcha viour. l .ondon : Lon gm an.

Silber, E.. Perry , S. E. & R1 och . D. A. ( IlJ5XI. Paucrns in pa rent-ch ildint e rac tio n in a di sa ster. l 's vc hiutr v, 2 1: I5<J-16 7.

SPSS Inc. ( ) lJX 3), SI'SS L'sef\ (i ll iel;'. New York: McGraw H ill.Ta usig . M . (1 9X21. Measu ri ng life even ts. .I. II/III . Soc. ltchuv.. 23:62­

64.Terr. I.. C. (llJ7lJl. Children ofChowch illa, Tlu: Psvchoanalvtic Su ulv

oltlu: Child, 34:547-627 . .. .-- (II)X Ia), Psychic trauma in ch ildren : o bse rva t io n following th e

Chowc hilla school-bus kidnapping. ..llI1e'f. J. Psvrhiat., I3X: 14-1 9 .-- (l lJX Ih) , "Forbidden games": posttraumatic child's pla y. This

Journal. 20:741-760.Z iv, D. & Israeli . R . (llJ73). Effec ts ofbom bardrncnt on the manifest

a nx iety level of children livin g in Kibbutzim . .I. ('0 11.\ 11/1. tlin.l 'sycho} : 40: 2X7- 2<J I.