10.1192/bjp.151.1.45Access the most recent version at doi: 1987, 151:45-51.BJP 

M Place, I Kolvin and S M MortonThe Newcastle Adolescent Behaviour Screening Questionnaire.

permissionsReprints/

permissions@rcpsych.ac.ukto To obtain reprints or permission to reproduce material from this paper, please write

to this article atYou can respond http://bjp.rcpsych.org/cgi/eletter-submit/151/1/45

from Downloaded

The Royal College of PsychiatristsPublished by on September 5, 2011http://bjp.rcpsych.org/

http://bjp.rcpsych.org/site/subscriptions/ go to: The British Journal of PsychiatryTo subscribe to

British Journal of Psychiatry (1987), 151,45—51

The Newcastle Adolescent Behaviour Screening Questionnaire

M. PLACE,I. KOLVINandS. M. MORTON

As part of a largerstudy, a brief rating-scale was developed which focusseson the mid-adolescent phase of development. Completed by teachers, the questionnaire has an inter-raterreliability of 0.78, with a test—retestcorrelation of 0.82. When the performances of variousscreening instruments were compared it became clear that no single questionnaire wasobviously more efficient than the others at detecting potential disturbance in an urbanadolescent population. Indeed, different questionnaires seemed to highlight particularfacets of functioning. The Newcastle Adolescent Questionnaire proved to be a reliable andvalidscreeningmeasure.

Several teacher questionnaires measure facets of thebehaviour of children and adolescents. Some of thesequestionnaires are standardised for specific stages ofdevelopment, e.g. the Devereux Scale (Spivack et al,1967), while others are used across the whole schoolage-range with only minor attempts to standardisefor age (Rutter, 1967;Conners, 1969).The Rutter B2Scale consists of 26 items and gives rise to two subscales, neurotic and antisocial behaviour, while theConners Scale consists of 39 items and generates foursub-scales: inattentive—passive;passivity; anxiety;and hyperactivity. The Behaviour Problem Checklist(Quay & Peterson, 1979) consists of 55 items and alsogenerates four factors. Although this instrument hasbeen well validated, especially in the younger agegroup, the factors show a bias toward detecting conduct and personality problems, so that the checklistmay miss a portion of the distressed adolescents inthe community (Place & Kolvin, in preparation).One of the few scales which is specific for adolescenceis the Adolescent Life Assessment Checklist (Gleseret a!, 1977), but this 40-item scale has only beenvalidated for selected samples (Boyle & Jones, 1985)and so its value as a community screening instrumentis unclear.

As part of a larger study, it was decided to developa questionnaire which would be very brief and wouldfocus on the mid-adolescent phase of development.Using clinical experience and published reports, weconsidered a large number of behavioural featuresfor inclusion. Seven were eventually chosen becausethey often give rise to concern in educational settings:they include a sense of confusion, excessive daydreaming, lability of mood and emotion, lack ofconfidence, and being easily upset in the face of educational failure (see Figure 1). In addition it wasdecided to include three more global ratings. The first

reflects more widespread neurotic behaviour, thesecond antisocial behaviour. As a final item wedecided to include educational difficulties which wereserious enough to warrant special attention. We suspected that a positive rating on this item would provea useful indicator not only of general educationaldifficulties but also of personal and interpersonalproblems.

The aim of the current paper is to explore the use ofthis brief questionnaire in an adolescent sampledrawn from the community. Our primary hypothesiswas that adolescents detected as being potentiallydisturbed using a ‘¿�multiple-criterionscreening' technique (described below in the ‘¿�Method'section)would prove to have significantly high scores on thisbrief questionnaire. This would in effect provide aform of validation of the questionnaire. A secondaim was to study its sensitivity and specificity compared with a psychiatric interview. Finally, it wasplanned to compare the scores obtained on thisinstrument with the scores obtained on longerscreening questionnaires by the same population.

MethodThe 4th and 5th-year pupils of four large comprehensiveschools agreed to participate in the study: 1446 adolescentsformed the population for this phase of the study, and theywere screened using a ‘¿�multiple-criterionscreen' method.This technique uses the number of extreme scores that anindividual obtains on a variety of instruments as a measureof potentialdisturbance.Thedetailedmethodologymaybefound elsewhere (Place el a!, 1985), but in brief the screenconsistedof:

(a) Rutter B2 Scale (Rutter, 1967, 1973)(b) a sociometricassessmenttechnique(MacMillaneta!,

1978)

45

46 PLACE ET AL

NewcastleAdolescentBehaviourQuestionnaire

Name of Child School

Date of Birth Form

Below are a series of questions describing behaviour often seen in adolescents. If thedescriptionapplies to this youth pleasemark the third column. If it does not applyplease mark in the first column. If it applies to a lesser degree please put a cross in thebox in column two.Pleaseconsideronlyrecentbehaviour,i.e.in the last 6 months.

ForOfficial

Yes Use Onlyo oo 0o oo oo o

o 0 0 0o 0 0 0

No Somewhat Yeso o 0

No Somewhato 0o 0o 0o 0o 0

I. Excessiveday dreaming2.Often lookspuzzledor confused3.Over-emotionalwhencheckedor frustrated4. Lacksconfidencein educationalsituations5. Tends to vary in mood6. Easily upset when cannot master

educational tasks7. Reserved and uncommunicativeIt would be most helpful if you could make anoverall assessment as to whether this youth:

is nervous, i.e. emotional, tense, worried andanxious, to the extent of impeding him in hisschool work or relationships

ForOfficial

UseOnly0

showsproblembehaviour,i.e.difficultand dis- 0ruptive behaviour, disobedient, etc.

has educational difficulties which merit special 0attention

0 0 0

0 0 0

(c) Junior Eysenck Personality Inventory (Eysenck,1965)

(d) General Health Questionnaire (Goldberg, 1972)(e) Rutter MalaiseInventory(Rutter eta!, 1970).

(a) Rutter B2 Scale This is a teacher-report instrumentwhich has been used in many settings, both in this country(Ryle & MacDonald, 1977; Rutter, 1973; Wolkind, 1974;MacMillan ci a!, 1980) and abroad (Mmdc, 1975;Zimmermann-Tansella ci a!, 1978). The instrument wasdesigned for the 6-13 age-group (Rutter, 1967), for whomit has an inter-rater reliability of 0.72 and test—retestreliabilityof 0.89. Little validation work has beencarriedout on the instrumentin the adolescentage-group.

(b) Sociometry This technique uses the responses ofpeers to gauge an individual's social standing. It allowsidentification of those youths who are likely to be avoidedby classmates (rejection) and those who would not bechosen as friends (isolation). It has been shown tobea rapidway of gathering useful information from younger children(MacMillan eta!, 1978)with a test—retestreliability of 0.87

for rejection and 0.72 for isolation (tests carried out 4—5weeks apart), but its effectiveness with adolescents seemsless clear (Place & Kolvin, in preparation).

(c) Self-Report The Junior Eysenck PersonalityInventory (Eysenck, 1965)was used as the main item ofself-report, although previous work with ll-year-olds hasshown that it tends to select a rather unique populationcompared with othermeasures (MacMillan eta!, 1980).Theneuroticism scale is reported to have a split-half reliabilityof 0.84and a test—retestreliabilityof 0.77(Eysenck,1965).

In addition, 605 youths completed the 30-item GeneralHealth Questionnaire (Goldberg, 1972) and 1237 completedthe Rutter MalaiseInventory,whichhasa test—retestreliabilityof 0.91(Rutteret a!, 1970).In an effortto reducethe workloadon teachersand also to minimisethe disruption of classesfor the pupils,all weretested,but not everyone with all the tests. This also ensured that the exerciseremained within the resources of the research team.

The next step was to compare the resultsobtained fromthe questionnaire with the data obtained from a psychiatricinterview. For these purposes a sub-sample of 82 youths

A. Educationaldifficulties1.0B.Disruptive anddisobedient0.161.0C.Tense, worried andanxious0.430.141.0D.Reserved anduncommunicative0.250.060.391.0E.Easilyupset0.290.370.420.21.0F.

Variablemood0.170.580.270.110.531.0G.Lacksconfidence0.440.180.440.430.470.321.0H.

Overemotional0.220.580.280.080.580.680.291.0I.Puzzled0.420.130.380.380.380.260.660.251.0J.

Daydreams0.360.190.390.430.290.280.520.250.611.0K.Total Score0.560.510.630.540.690.650.760.640.720.69 1.0

NEWCASTLE ADOLESCENT BEHAVIOUR SCREENING QUESTIONNAIRE 47

TABU@ICorrelation of items in the Newcastle Adolescent BehaviourQuestionnaire.All correlationsare signjficantat the 0.001 level,

exceptfor 5(P <0.05)

A B C D E F G H I I K

was selected on a random basis from the youths who hadan extreme score on at least one of the four screeninginstruments. The percentage levels used for cut-off werevaried to ensure the inclusion of at least 20 individuals oneach major criterion, but because of overlap the numberselectedon each was approximately30. In addition, 50adolescents were selected at random from the youths whohad no extreme scores on any of the measures. All of these132 youths were given a structured psychiatric interview,the interviewer being unaware of the questionnaire results.Psychiatric disturbance was rated as ‘¿�absent',‘¿�dubious',‘¿�moderate'or ‘¿�marked'and these ratings were then furthercombined to produce two groups—'no significant disturbance' (absent/dubious) and ‘¿�significantdisturbance'(moderate/marked).

For comparison, 30 adolescents who were resident in amedium-stay psychiatric unit were assessed with the instruments mentioned above, with the exception of the sociometric technique. Additional data were obtained fromthe teachers of this in-patient group to allow a study ofthe Newcastle Questionnaire's inter-rater and test—retestreliabilitiestobeundertaken: two of the teachers wereaskedto rate the youths' behaviour at the time of the survey andagain one month later.

Results

It is important to establish the reliability and validity of anynew behavioural measure, and this may be done in variousways

(a) Inter-rater reliability

It is necessaryto confirm that differentteacherswillgivesimilar ratings for similar behaviour. Greater agreementoccurs when the instructions on the use of the instrumentare carefully presented in a standard form, and it was forthis reason that we chose to follow the presentation used by

Rutter in his teacherquestionnaire(Rutter, 1967). To testreliabilityweusedthe 30adolescentpatients in a mediumstay psychiatric unit: community surveys yield smallnumbers of potentially disturbed youths and so the correlations found may be influencedby a lack of spread in thedata. The questionnaire'sinter-rater reliabilitywas foundto be 0.78 (n=30), whichis satisfactoryfor thiskindof briefinstrument.

(b) Test—Retestreliability

This was established using the same 30 in-patients as forinter-raterreliability:theywerereassessedafteronemonth.Such exercises, which are essential to test the reliabilityof an instrument,are often bedevilledby the variation ofbehaviour which occurs over time. Nevertheless, the correlation value obtained (0.82) is an acceptable one for thistype of screening instrument.

(c) Split-half reliability

After pairing items, one member of the pair was assignedrandomly to either the first or second half of the questionnaire. A Pearsoncorrelationof 0.84wasobtained betweenthese halves, and applying the Spearman Brown Prophecyformula the split-half reliability was calculated at 0.91. Thisis a highly satisfactory level of reliability for a scale of 10items, particularly as one of the items is not intended tomeasure behaviour directly.To confirmthisfinding,thesplit-halfreliabilityexercise

was repeated without the more global assessments beingincluded. This yielded a Pearson correlation coefficient of0.85,whichisalmostidenticalto the valueobtainedfor thefull scale.

In addition, the means and standard deviations of thesplit-halves were calculated, and proved to be almostidentical, indicating that they are equivalent measures ofbehaviour.

A B CControls Screen+ Psychiatric

(screen—)in-patientsn1144

30230Mean1.993.986.87s.d.2.733.943.35SignificanceA

vsB B vsC(i-test)t=10.2 t=3.88

P <OJXXII P<0.001A

vsCi=7.9l

P <0.0001

48 PLACE F@ AL

TABLEIIValidation: mean scores on the Newcastle Adolescent Behaviour Questionnairefor three different groups selected using

otherpreviously validatedscreeninginstruments

C. Psychiatric cases Adolescents whose difficulties weresosevereas to not permitthemto survivein thecommunityand so warranted admission to a psychiatric unit on anin-patient basis. These youths could be expected to be moredisturbed than the screen-positive adolescents who weresurviving in the community.

Table II demonstrates a highly significant step-by-stepincrease in mean score on the questionnaire as we movefrom group A (the least disturbed) to group C (the mostdisturbed).

Table III gives teachers' ratings on the individualquestions for each of these three groups, and confirms thatall the questions distinguished between the controls and thepotentially disturbed subjects. Such a finding adds furtherto the conclusion that the Newcastle Adolescent BehaviourQuestionnaire is a valid instrument foran urban adolescentpopulation.

(c) Sensitivity, specjficity and efficiencyCut-off scores of 3,4, 5 and 6 on the Newcastle Questionnairewereexploredin relationto thefindingson thepsychiatric assessment, and it was established that a score of 4 ormore proved to be optimal for identifying potential cases inthis particular population. Such an exploration tests thevalidity of the questionnaireateach particularcut-off score:it checks the proportion of ‘¿�correct'cases the instrumentdetects against the proportion it misses (sensitivity); theproportion of true negatives it correctly identifies(spec@f1city);and the proportion of correct selection in thesample as a whole (efficiency).

The method of selectingthe interviewsample did notpermit direct inferences about the total study population.The number of disturbed adolescents in the total samplewas estimated from the proportion of disturbed youthsfound at each scorein the interviewsample.This estimatewas then used to explorethe performanceof the questionnaire when the optimum cut-off was used (Table IV). Ourmethod ofcalculatingsensitivity, specificity and efficiency isdetailed elsewhere (Place ci a!, 1985).The results reveal thatthe instrument is capable of selectingpotential cases ofadolescent disturbance witha moderate degree of accuracy,but is far less efficient than the Behaviour ScreeningInstrument developed by Richman & Graham (1971) forpre-schcolers, particularly in terms of specificity, and lessefficient than scales developed for adult populations(Goldberg,1972).

Comparison with other measures

In order to establish the comparative efficiencyof thismeasure compared with other available scales which mightbeconsideredappropriate for screeningadolescents,all theinstruments were compared using their estimated performance in the study population when the optimum cut-off wasused: the sensitivity, specificity and misclassification ratewere established for each instrument. This is a rather crudeanalysis, but it showed that no instrument was clearlysuperior to the others. The sociometry scales showed a lowsensitivity (isolation 30%, rejection 19%); highest were theMalaise Inventory (61%) and General Health Question

Validity

(a) Content validity

Content validity is usually taken to mean that the itemsreflect the type of behaviour that the instrument is intendedto measure.We believethe itemschosen for this instrumenthavefacevalidity.However,theirrelationshiptoeach otherand to the total score was explored. As Table I shows,although some of the correlations are reasonably high andsome are low, all the correlations are positive. The averageinter-correlationof the items was 0.34, and this suggeststhat the itemsare measuringsimilaraspectsof behaviour.Cronbach has developed a formula for assessing theinternal consistency of a questionnaire (Bynner, 1969),which provides a coefficientbased on the average correlation betweenitemsand the numberof itemsused.For aninstrumentto have valid internalconsistency the Cronbachcoefficient should bemore than 0.5. The present scalegivesacoefficient of 0.69, which is satisfactory.

(b) Concurrentvalidity

This compares the scores obtained on the scale with anexternal criterion, either information obtained from aciical interview or scores on other instruments which areknown to be valid. We explored our instrument's ability todistinguish between three groups of youths:A.ControlsYouthswhodidnotemergeashighscorerson a number of previously validated instrumentsof childdisturbance, i.e. youths who were@'screen negative'.B. Screen positives Youths with a number of extremescores on previously validated instruments designed todetect disturbed behaviour.

A B CScreen Screen + Psychiatric

in-patients

(n=1144) (n=302) (n=30)SignificanceP<0.05P <0.01P<0.001Daydreams

2.9 11.3 8.9Puzzledandconfused 4.7 12.6 15.6Over-emotional 0.9 10.6 11.1Lacks confidence 4.9 20.2 17.8Variablemood 2.3 19.5 11.1Easily upset 1.2 9.6 13.3Reserved and uncommunicative 4.5 12.3 15.6Tense,worriedand anxious 0.9 9.3 8.9Disruptiveand disobedient 0.9 15.6 6.7Educational difficulties 0.09 7.0 6.7AvC

AvCAvCAvCAvB

AvBAvC AvB

AvBAvB

AvC AvBAvB

AvC AvBAvC AvBAvC AvB

nSensitivitySpecjudyMisclassjficationKTotal

population144651%82%26%0.32Boys74557%83%23%0.38Girls70146%80%30%0.26

49NEWCASTLEADOLESCENTBEHAVIOURSCREENINGQUESTIONNAIRE

TABLEIllValidation: percentage of high scores on each question of the Newcastle Adolescent Behaviourfor each of three different groups

(selected with previously validated screening instruments)

TABLEIVEfficiencyof the Newcastle Adolescent BehaviourQuestionnairein identifying cases in an adolescentpopulation, compared

with the estimatedprevalenceof disturbance.For thesecalculationsthe optimumcut-offscore was used (4 or above)

naire (57%). The Newcastle scale fell in-between,with51%. Although the Rutter B2 Scale was not very sensitive(32%), it had a lower misclassification rate (22%) than theNewcastle Scale (26%). The performanceof the Rutter B2Scale in this population shows that some of the items werenot as sensitive at this age-level as they had been in theyounger age-group.

TheNewcastleQuestionnairescoreswerecorrelatedwiththescoresobtainedon theothermeasures,and forcomparison theexercisewasrepeatedfor the Rutter B2Scale(TableV). The significant correlations are with the peer-reportmeasures and they are low; but they are of the same order asthoseobtained by the Rutter B2 Scale. Theinter-correlationof the NewcastleAdolescentQuestionnaireand the RutterB2 Scalegivesa reasonablyhigh coefficient(0.69)whichsuggests that the two scales have much in common. There isa moderate but statisticallysignificantcorrelation betweenthe Newcastle Questionnaire and the multiple-criterionscreen score. It is lower than the correlation between themultiple-criterion screen score and the B2 Scale, but the B2Scale contributes to the screening procedure and so a highcorrelation is to be expected.

Comparing the Newcastle and Rutter B2 Scales (both

teacher-report instruments) with questionnaires whichexploreother facetsof functioning wecan clearlyseethatcorrelation between some of the scales is poor. It has beenpointed out elsewhere (MacMillan ci a!, 1980) that theJunior Eysenck P1 neuroticism scale has little in commonwith other screening measures for 11-year-old subjects,and it is perhaps not surprisingto find a similar result inadolescence. However, all the self-report inventories correlatewellwitheachother,and theyshowequalvaliditywhencompared with a psychiatrist's assessment (Place & Kolvin,in preparation). If we assume that these inventories aretapping internalised distress, there seems little doubt thatthe potential cases that they selectare likelyto be distinctfromthosewhohaveshownbehaviourproblemsat school.This distinction between internalised distress and externalised problems of conduct is not new (Rutter ci a!, 1970),butcannot be tapped in adolescence unless specific questionnaires are used. If this is not done, any survey tends to reflectobjective behaviour problems and show minimal internalclimate problems. With such little overlap between scaleswhich measure such functions, the importance of usingboth sources of information is reaffirmed.

TheNewcastlescaleappearsto bemore sensitiveforboys

Multiple-criterionscreen0.33*0.55*Sociometry'Isolation0.19*0.21*Rejection0.19*0.20*Rutter

MalaiseInventory20.030.04GeneralHealthQuestionnaire20.020.03Junior

Eysenck PersonalityInventory2Neuroticism0.030.03Extroversion0.010.02Rutter

B2Scale30.69*

50 PLACE Fr AL

regarding the psychological adjustment of theirstudents―. In addition, this type ofresearch is popular with epidemiologists because the number of‘¿�missingcases' is lower than with other methods ofidentifying a population: this advantage may bereduced in an adolescent population, however,because absentee rates tend to be high in thisage-group (Fogelman et a!, 1980).

In screening populations it is usual to seek a briefmeasure which is reliable and valid and of low cost

(Boyle & Jones, 1985). Such a measure is alreadyavailable for pre-school children (BSQ: Richman& Graham, 1971). The Rutter B2 Scale is a wellestablished instrument for use with school childrenbut is not specific for adolescents. The NewcastleAdolescent Behaviour Questionnaire has beendeveloped in an attempt to fill this gap, but itsadvantages over the Rutter B2 Scale, other thanbrevity, have yet to be established. However, its verybrevity makes it suitable for investigating largepopulations.

One area of concern is the high misclassificationrate which is obtained when this instrument's scoresare compared with the findings at psychiatric interview. However, the other instruments tested show asimilar range of misclassification values. Althoughsome authors report minimal misclassification rates(Goldberg, 1972), others have obtained rates similarto those reported here when studying urban community populations (Tarnopolsky eta!, 1979).

The lack of correlation between the self-reportmeasures and the teacher-assessed scales couldreflect a lack of validity in the new scale or couldindicate that the self-rating scales are measuringdifferent components of behaviour. Since the lack ofcorrelation is similar for the well-validated RutterB2 Scale, and with evidence of the validity of theself-report measures being available, it seems likelythat the different types of instrument are detectingdifferent components of behaviour.

Goldberg(1972) has pointed out that when screening a population research workers assume that psychiatric disturbance is evenly distributed throughoutthe population in varying degrees of severity. Anindividual's score on a questionnaire may then beseen as a quantitative estimate of that individual'sdegree of potential disturbance. These assumptionsallow the researcher to decide upon a limit score, orcut-off, such that subjects with scores above it have ahigh probability of showing disturbance. However,like most screening instruments, such a questionnaire is not intended to make diagnostic distinctions.

Our results indicate that no single screening instrument is superior for detecting potential disturbancein adolescence. Indeed, different questionnaires

TABLE VCorrelations of Newcastle Adolescent Behaviour Questionnaire and Rutter B2 Scale with other measures of adolescent

behaviour

Newcastle Rutter B2

1. Peer report2. Self-report3. Teacher-assessed

P <0.0001 (n= 1446)

than for girls, a finding which is supported by the higherkappa score (Table IV). When consideredwith the abovefindings this suggests the hypothesis that behaviour scalesmay be more accurate at detecting disturbed boys whileself-report measures may be more accurate at detectingdisturbed girls. Such a theory is discussed in more detailelsewhere(Place& Kolvin,in preparation).

Discussion

New rating scales seem to appear frequently, anddisappear almost as quickly. Snaith (1981) has urgedthat existing measures be refined and improvedrather than perpetuating the development of moreand more new instruments. This is a most appropriate aspiration because the response to a rating scale

must be considered in the light of various factors.‘¿�OverallAgreement Set' (Cronbach, 1942) identi

fies respondents' desire to give the answer that isexpected or at least which appears less critical ofthemselves. This may be a particular problem inschool-based research which uses the teachers as asource of information, when the results may be usedto assess a school's ability to educate and control(Rutter et a!, 1979; Rutter 1980). Despite this limitation many surveys have depended upon schoolbased research (e.g. Rutter et a!, 1970; Offer, 1969;Coleman, 1974;Newman 1979)because, as Lewine eta! (1978) have pointed out, “¿�teachersare capable ofmaking sophisticated and valuable observations

NEWCASTLE ADOLESCENT BEHAVIOUR SCREENING QUESTIONNAIRE

seem to highlight particular facets of functioning,which are not necessarily detected by the others.Behaviouralmeasuresmust sample both internalandexternal qualities if they are to fully explore disturbance in an adolescent population. The Newcastle

References

Adolescent Questionnaire, although very brief, hasproved to be almost as reliableand as valid as most ofthe other screening instruments available for thisage-group.

PLACE, M. & K0LvIN, I. (In preparation) The relative validity ofscreening instruments in an adolescent population.

MACMILLAN, A. & NIcOL, A. R. (1985) Therelevance of a multiple criterion screen to an adolescent population. Psychological Medicine, 13,661-670.

QUAY, H. C. & PEmIIS0N, D. R. (1979) Manual For The BehaviourProblem Checklist. Rutgers State University, New Jersey.

RICHMAN, N. & GBAHAM, P. (1971) A behavioural screeningquestionnaire for use with 3-year-old children: preliminaryfindings. Journal of Child Psychology and Psychiatry. 12,5-33.

RUTTER,M. (1967) A children's behaviour questionnaire forcompletion by teachers; preliminary findings. Journal of ChildPsychology and Psychiatry, 8, 1—11.

(1973)Whyare Londonchildrenso disturbed?Proceedings of the Royal Society of Medicine. 66,1221-1225.

(1980)School influenceson children's behaviour anddevelopment. Paediatrics, 65,208-220.

TIzARD,J.&Wmmoas., K.(1970) Education. HealthandBehaviour. London: Longman.

,MAUGHAN, B., Moam.oiut, P., OUSTON,3. & Siam, A.(1979) Fifteen Thousand Hours: Secondary Schools and TheirEffects on Children. London: Open Books.

R@ A. & MACDONALD,3. (1977) Responses to reading as perceivedby boys with and without specificreading retardationand behaviouraldisorders.Journal of Child PsychologyandPsychiatry, 18,323-334.

SNAJm, R. P. (1981) Rating scales. British Journal of Psychiatry.138,512—514.

SPIVACK, G., HASHEs, P. E. & Sports, 3. (1967) Devereux Adolescent Behaviour Raring Scale Manual. Devereux Foundation:Pittsburgh, Pennsylvania.

TARNOPOLSKY,A., HMII), D. 3., MCLEAN,E. K., ROBERTS,H. &Wioan@is, R. D. (1979) Validity and uses of a screening questionnaire (G.H.Q.) in the community. British Journal of Psychiatry,134,508—515.

WOLKIND,S. N. (1974) Sex differences in the aetiology of antisocialdisorders in children in long-term residential care. British Journalof Psychiatry, 125,125-130.

ZIMMERMANN-TANSELLA,C., Mmiovucrn, S., T@tccoNi, A. &TANSEU.A, M. (1978) Children's behaviour questionnaire forcompletionby teachersin an Italian sample.Journalof ChildPsychology and Psychiatry, 19, 167—173.

BoYLE,M. H. &Jouns, S. C. (1985) Selecting measures of emotionaland behavioural disorders of childhood for use in generalpopulations. Journal of Child Psychology and Psychiatry, 26,137—160.

Bmusa, J. M.(1969) The YoungSmoker. London: HMSO.COI.EMAN, J. C. (1974) Relationships in Adolescence. London:

Routledge& Kegan Paul.Couunss, C. K. (1969)A teacher rating scale for use in drug studies

with children. American Journal of Psychiatry. 126,884-888.CRoNBACH, L. J. (1942) Studies of acquiescence as a factor in the

true-false test. Journal of Edocatlonal Psychology, 33,401-415.EYsENcK, S. B. G. (1965) Manualfor the Junior Eysenck Personality

Inventory. London: University of London Press.FOGELMAN,K., Tmaases@tai, A., LAMBERT,L. (1980) Absent from

school: Findings from the national children'sdevelopment study.In Out of School:ModernPerspectiveson Truancyand SchoolRefusal (eds. L. Hersov & I. Berg) Chichester John Wiley.

GLEsER,G., SELIGMAN,R., Wmiosr, C. & RAUH,J. L. (1977)Adolescents view their mental health Journal of Youth andAdolescence, 6,249-263.

GOLDBERG, D. P. (1972) The Detection of Psychiatric illness byQuestionnaire. London: Oxford University Press.

LEWINE, R. R. J., WATT, N. F., PRENTKY, R. A. & FRYER, J. H.(1978) Childhood behaviour in schizophrenia, personality disorder, depressionand neurosis. BritLchJournal of Psychiatry. 132,347—357.

MACMILLAN, A., W@um, L., GARSIDE, R. F., Kotvmi, I., LarcH,I. M. &NIcoL, A. R. (1978) The developmentand application ofsociometric techniques for the identification of isolated andrejected children. Journal of the Association of Workers forMaladjusted Children,6,58-74.

,K0LVIN, I., GARSIDE, R. F., NICOL, A. R. & LEITCH, I. M.(1980) A multiple criterion screen for identifying secondaryschool children with psychiatric disorder. PsychologicalMedicine, 10,265—276.

MINDF., K. (1975) Psychological problems in Ugandan schoolchildren. Journal of Child Psychology and Psychiatry. 16,49-59.

NEwMAN,B. H. (1979) Coping and adaptation in adolescence.Human Development, 22,255—262.

Oma, D. (1969)ThePsychologicalWorldof the Teenager.NewYork:BasicBooks.

5M. Place, MB,BS,MRCPsych,Consultant Child and Adolescent Psychiatrist, SunderlandDistrict Genera!Hospital; I. Kolvin, BA, MD, FRCPsych,Proftssor of Child and Adolescent Psychiatry, Fleming Hospita!,Newcas:le-upon-Tyne, and S. M. Morton, FRCP,DPM,DCH, Consultant Child and Adolescent Psychiatrist,Fleming Hospital

*Co@ponde@: Child, Adolescentand Family Psychiatry Unit. Swsderland District General Hospital. Kayil Road, Sunderland. Tyne& WearSR4 7TP