Journal of Consulting and Clinical Psychology1978, Vol. 46, No. 6, 1246-1257

An Observational Approach to the Assessment ofAnxiety in Young Children

Blair GlennonHarvard University

John R. WeiszUniversity of North Carolina at Chapel Hill

Clinical research on anxiety has long relied on assessment techniques that maybe inappropriate with young children (e.g., self-report inventories). The presentarticle describes an alternative to such techniques—a scale using observationalmethodology. To assess the reliability and validity of this instrument, the Pre-school Observational Scale of Anxiety (POSA), preschoolers were observed andscored on the scale during two test sessions. Session 1, with mothers absent,was expected to provoke relatively high anxiety; Session 2, with mothers pres-ent, was expected to provoke minimal anxiety. Total POSA scores assigned bytwo independent judges correlated .78 (p < .001), with highly significant inter-judge correlations for most of the scale items. Regarding the validity of theinstrument, it was found that (a) POSA scores were significantly correlatedwith teachers' and parents' inventory ratings of children's anxiety (all ps < .01),and (b) children's POSA scores were significantly higher in Session 1 thanSession 2 (p < .01). The findings suggest that the POSA may provide a meansof assessing situationally induced anxiety in children who are too young toaccurately report their internal states.

Anxiety has long been a topic of centralimportance in clinical research and practice.Achenbach (1974) noted that among person-ality traits emerging from trait theories,,"anxiety is perhaps the most frequently in-ferred and measured (p. 574)"; yet, demon-strably valid and reliable measurement ofanxiety has been difficult to achieve, particu-larly among children. Especially acute is theneed for accurate measures that can be usedto assess specific situational effects on anxietystates in children (see Spielberger et al.,1972). Although many investigators recognizethe need for such measures, particularly inresearch with children who are too young to

The study was supported in part by a researchgrant from the College of Human Ecology, CornellUniversity.

The authors are grateful to Barbara Cesario, JohnDoris, Laura Harrison, Robert Loheydc, Linda Mack,and the staff, parents, and children of the CornellNursery School for their assistance in the conduct ofthis study.

Requests for reprints should be sent to John R.Weisz, Department of Psychology, University ofNorth Carolina, Chapel Hill, North Carolina 27514.

accurately report their own internal states,most researchers also recognize the difficultiesthat inhere in applying previous anxiety as-sessment techniques to children.

Historically, the principal methods for mea-suring anxiety have been physiological mea-surements, projective techniques, self-reports,and behavior ratings by observers. Throughdirect gauging of autonomic activity, physio-logical measures bypass the problems of sub-jective judgments. However, agreement amongphysiological measures is frequently poor,since people have different styles of autonomicresponding (Phillips, Martin, & Meyers,1972), and the physiological instruments maytap emotions such as anger or joy rather thananxiety (Lazarus, 1966). Furthermore, theunusual instruments used for physiologicalmeasures may distract subjects and makenaturalistic observations difficult, particularlywhen the subjects are children.

Projective techniques have also been usedto obtain ratings of generalized anxiety (Mc-Reynolds, 1968), but most projective tech-niques require individual administration aswell as substantial time and expertise on the

Copyright 1978 by the American Psychological Association, Inc. 0022-006X/78/4606-1246$00.75

1246

ANXIETY IN CHILDREN 1247

part of the examiner, thus limiting their use-fulness. Furthermore, McReynolds (1968)has reported that projective techniques "aregenerally found to be unrelated to inventoryessays of anxiety" (p. 258); and Achenbach(1974), after reviewing the massive literatureon projective techniques, has concluded that"there is little evidence for the reliability orvalidity of most of the interpretations madefrom them" (p. 604).

Three main types of self-reports have beenused in assessing anxiety. First, direct self-ratings involve asking subjects specificallyhow anxious or nervous they feel. McReynolds(1968) has suggested that such self-ratingsare most useful for measuring current levelsof anxiety or changes in anxiety from onesituation to another. A second type of self-report is the Adjective Check List, on whichsubjects indicate which of several adjectives(e.g., "jittery") best characterize their mood.Adjective Check Lists have been used toassess either current anxiety level or charac-teristic anxiety (McReynolds, 1968). Thethird type of self-report is the inventorymethod, with questionnaires designed to de-termine how subjects feel in a variety of situ-ations. This method is usually aimed at as-sessing characteristic rather than current anxi-ety level. Inventories may focus on generalizedanxiety (i.e., across a wide variety of situa-tions) or anxiety related to particular types ofsituations (such as separations). Several in-vestigators have noted drawbacks of the vari-ous self-report methods (see Spielberger,1972); however, probably the most importantdrawback in the present context is that suchmethods assume both the ability and the will-ingness of subjects to correctly describe theirown feelings. The first assumption is clearlytenuous with young children. And, with regardto the second assumption, there is evidencethat many children give inaccurate reportsabout their anxiety due to defensiveness orsocial pressures (Sarason, Davidson, Light-hall, Waite, & Ruebush, 1960). Consequently,Sarason (1966) has stated that "the verbalresponse to our [children's self-report anxi-ety] scales may be telling us more about theself than the affect" (p. 79).

Behavior ratings of anxiety by observersmay involve global, subjective judgments

about the subjects' apparent nervousness orratings on many specific, concrete behaviorsthought to indicate anxiety (e.g., stuttering).Such approaches appear to offer distinct ad-vantages relative to those discussed above(see Spielberger et al., 1972). For instance,unlike physiological measures, observers' be-havior ratings can be made unobtrusively;unlike projective techniques, they do not nec-essarily require trained clinicians; and unlikeself-report measures, they do not rest ontenuous assumptions about the abilities andattitudes of the subjects. In addition, behav-ioral observations may be particularly usefulwith children, since they appear to disguiseexpression of their feelings less effectivelythan adults (Sarason, 1966).

The observers for the behavior rating scalesof children's anxiety are sometimes teachersor parents, with the scales including itemsrelevant to school or family situations (e.g.,cries at bedtime). However, these particularscales entail a risk of distortion due to theparent's or teacher's lack of care or objectiv-ity in observing, or to their bias or defensive-ness about reporting on a child whom theyknow personally (Sarason et al., 1960). Theobservers for the behavior rating scales, how-ever, may be trained observers (see Buss,Wiener, Durkee, & Baer, 1955), and the scalesmay comprise behaviors indicative of anxietyacross situations, thus being more widelyapplicable and less subject to distortion byobservers. The latter type of behavior ratingscale would, in principle, permit detection ofsubtle relationships between specific eventsand ensuing anxiety in a way that satisfiesmethodological requirements and is develop-mentally appropriate as well (see Weisz,1978).

It is true that different individuals mayreveal their anxiety through different be-haviors; also, different observers may havedifficulty achieving high agreement in theirbehavior observations. But these problems ofbehavior ratings should be surmountable.Clearly defining the behaviors to be observedand increasing the training time of observersshould reduce problems of low interrateragreement. And total frequency scores on rat-ing scales including the gamut of behaviorssuggestive of anxiety should give good indi-

1248 BLAIR GLENNON AND JOHN R. WEISZ

cations of the relative anxiety of individualswith widely varying anxiety manifestations.

The most accurate behavior rating scalewould require observers who are able to con-centrate fully on the occurrence of the targetbehaviors and to record the occurrence ofthese behaviors for later frequency analysis.Tt should be noted that in the typical clinicalinterview situation, the clinician would notbe able to meet these requirements. However,a second observer could independently recordthe indicators in clinical situations in whichprecise anxiety measurements are desired, orthe clinicians themselves could use the indi-cators of behavior rating scales as an aid inmaking less exact, more global ratings of theclient's anxiety. In the latter case, the clini-cians would not be using the scale in theprescribed way but would probably add ob-jectivity to their observations by reference toit.

In short, behavior observation measures ofanxiety seem to have clear advantages overother types of anxiety measures and rela-tively minor drawbacks. But despite the po-tential usefulness of such measures, we havefound little systematic research on behaviorindicators of anxiety in children: In fact,there appears to be no carefully validatedbehavior rating scale for anxiety in children.Perhaps the closest approach is that of Gross-man (1968), who used observer ratings ofanxiety with 6-year-olds. However, his scaleconsisted of a mixture of specific objectivelyobservable behaviors (e.g., nail-biting) andgeneral indicators requiring subjective in-ference (e.g., "reactions suggesting that thechild was frightened"); as described by theauthor, the scale appeared to contain only sixitems, thus showing limited sensitivity to therange of behaviors through which childrenmay display anxiety. Validity data consistedof only two nonsignificant correlation coeffi-cients relating the behavior scale to the Gen-eral Anxiety Scale for Children (Sarason,Davidson, Lighthall, & Waite, 1958), and toour knowledge, no independent studies havebeen done to support the use of the scale asan anxiety measure. Melamed and Siegel(1975) used their Observer Rating Scale ofAnxiety in a study on the reduction of anxietythrough filmed modeling in 4- to 12-year-old

children facing surgery. Again, only very lim-ited validity data on this scale are available.Also, only four items of the scale were pub-lished, and these represent an exceedinglymodest sampling of the possible behavioralmanifestations of anxiety.

In the present study, a much more de-tailed list of behavior items was assembledby means of a systematic search of the anxietyliterature. This list was used to form the Pre-school Observation Scale of Anxiety (POSA).The interobserver reliability of the POSA wasassessed by two independent judges. Thevalidity of the scale was determined (a)through assessment of its relation to threeindependent inventory measures of anxietyand (b) through an experimental manipula-tion of stressors. Significant positive relationsbetween the POSA scores and the inventoryscores were anticipated, but these relationswere expected to be modest, since (a) the in-ventory measures and the POSA were de-signed to assess somewhat different aspects ofanxiety, and (b) the inventory measures reliedon parents' and teachers' judgments, whereasthe POSA involved relatively objective obser-vations by trained observers. The experi-mental manipulation involved a contrast be-tween an initial mother-absent condition anda subsequent mother-present condition. Be-cause the format and experimental procedurewould be more familiar to the children in thesecond session, and because mothers would bepresent, it was predicted that children wouldshow lower levels of anxiety in Session 2 thanin Session 1, and thus score lower on thePOSA.

MethodSubjects

The 36 Cornell University Nursery School childrenwho formed the sample ranged in age from 32 to 59months (M = 47 months; SD = 7.7 months). MeanHollingshead (Note 1) sociocconomic status (SES)was 1.47 (1 = highest; 7 = lowest; SD = .94). Therewere 21 girls and 15 boys.

Procedure

The Preschool Observation Scale of Anxiety. Thefirst step in devising the scale was a systematic searchof the Education Index (1929-1976), PsychologicalAbstracts (1927-1976), Resources in Education/Re-

ANXIETY IN CHILDREN 1249

search in Education [1Q66 (year or origin) to 1976],and Current Index to Journals in Education [1969(year of origin) to 1976], for studies using or men-tioning behavioral indicators of anxiety in childrenor adults. Next, three child-clinical psychologistsexamined a list of items based on the literature searchto suggest additions, to make minor modifications,and to give their final approval to the scale. Somebehavior indicators suggested by the literature orthe clinicians (e.g., hand perspiration) were not in-cluded in the final scale because pilot testing indicatedthat they were too difficult to observe accurately andreliably. Table 1 includes a description of each be-havioral indicator of the final POSA, along with areference to the article (s) supporting the use of eachindicator in the scale.

Parent and teacher questionnaires: Independentanxiety measures. Several independent measures ofanxiety were obtained for the purpose of validatingthe POSA. One was a questionnaire, the ParentAnxiety Rating Scale (Doris, Mclntyre, Kelsey, &Lehman, 1971), completed by each child's parentsand comprised of six questions about the child'sseparation anxiety (PARSEP) and 19 questions aboutthe child's general anxiety (PARGEN). Two weeks be-fore school began, this questionnaire was sent to thechildren's parents, who completed it by the beginningof the nursery school year. A second independentmeasure of anxiety was the Teachers' SeparationAnxiety Scale (TSAS; Doris et al., 1971), comprisedof 11 items about the child's reaction to separationfrom his/her mother or father when left at thenursery school at the beginning of the school day.One of two teachers (one for each of two nurseryschool sessions) rated the child on each of the first10 consecutive days of the child's attendance atnursery school for the year. Ratings were made atthe end of each school day and pertained to theperiod from the child's arrival at the nursery schoolwith parent to the parent's departure. If the childwas not delivered to the nursery school by his/herparent, the child was not rated that day, and his/herTSAS score was prorated.

Experimental manipulation of anxiety. In addi-tion to obtaining questionnaire scores of the chil-dren's anxiety, we used a manipulation designed tocreate differing levels of anxiety in two experi-mental sessions, both involving cognitive tests. Thefirst session was expected to be more anxiety arous-ing, since it involved an unfamiliar adult who in-dividually tested each child in an unfamiliar settingduring one of the first few days of the nurseryschool year. In contrast, the second session occurredapproximately 2 weeks following the first session,after the children had had a chance to settle in totheir new surroundings; furthermore, the child'smother was present at the second session, along withthe same examiner who had tested the child at thefirst session. Finally, the children were given thesame three tests in the second session that theyhad already taken in the first session.

Tests were given because it was felt that anevaluative atmosphere would make the experimental

situation more anxiety arousing, especially in thefirst session when the materials would be unfamiliar.Also, performance on the cognitive tests by thehigh-anxious children compared to the low-anxiouschildren was of interest, since existing evidence is inconflict over this question. Some evidence sug-gests that high anxiety should interfere with testperformance, especially when the tests are fairlydifficult ones for the subjects (e.g., Feldhusen &Klausmeier, 1962; Tamaroff, 1976; Young & Brown,1973). However, other evidence (e.g., Denny, 1966;Hodges & Durham, 1972; Katahn, 1966; Spielberger,1966) suggests that anxiety may have facilitatingeffects on performance for children of higher socio-economic and intellectual levels. (See other con-flicting evidence in Fischer & Awrcy, 1973; Mazzei& Goulet, '1969.) Since, according to the evidence,anxiety might lead to performance enhancement insome children and performance decrements in otherchildren, no prediction was advanced regarding therelation between anxiety and task performance inthe present study. However, the relation betweentask performance and anxiety scores was reported inan effort to shed light on the controversy justdescribed.

Global anxiety self-ratings and ratings by theexaminer. In the Introduction to this article, weemphasized the disadvantage of both self-reports bychildren and global (and thus subjective) ratings byadults. To determine whether our negative assess-ment was correct, we included two such measuresin the present study, so that their effectiveness mightbe contrasted with that of the more specific andpresumably more objective POSA. At the end ofeach of the two testing sessions, the examiner ratedthe child's general anxiety level during the sessionon a scale of 1 to 6. Also, a teacher asked each childto choose one of six pictures depicting progressivelymore fearful facial expressions to show how anxiousthe child felt during the testing session. The teacherrather than the examiner asked this question, sinceit was felt that the children -would be more candidabout their feelings with their teacher. The word-ing of the question was the following:

These pictures show a picture of a child who ismore and more scared as you go from this toppicture to the bottom picture here [point]. Yousee, up here the child is not scared at all [point] ;here [second picture from top] the child is alittle more scared but still pretty happy; here[third from top], he's getting more scared; andhere [bottom], very, very scared. Which picturewould show how scared you felt when you wentdownstairs to the testing room with that lady?One of these up here where the child isn't scaredat all, or one of the bottom ones showing a childwho is more and more scared?

The observation periods. At the beginning ofthe first session, the examiner, a 26-year-old ex-perienced female teacher, approached each childindividually in the nursery school and told the child

1250 BLAIR GLENNON AND JOHN R. WEISZ

Table 1Items of the Preschool Observation Scale of Anxiety and Interrater AgreementDuring the First 10 Minutes of the First Session

ItemRange offrequency

Interraterreliability"

Concor-dance 1

Concor-dance 2

Physical complaint : Child says he or shehas a headache, stomachache, or hasto go to the bathroom (B,C,D,E,G,K,Q)

Desire to leave: Child says he or shewants to leave the testing room ormakes excuses about why he or shemust leave; desire or "need" to leavemust be explicit (D,E,P,Q).

Expression of fear or worry: Childcomplains about being afraid of orworried about something; must usethe word "afraid," "scared,""worried," or a synonym (F).

Cry : Tears should be visible

Scream (P).Whine or whimper (G,P).Trembling voice (F,G,I,M).Stutter <A,F,G,H,M,O).Whisper: Child speaks softly, without

vocal cords; should not be a playfulwhisper (E,G).

10. Silence to one question in the interval(E).

11. Silence to more than one question in theinterval (E).

12. Nail-biting: Child actually bites his orher nails in the testing room (F,G,I).

13. Lip-licking: Tongue should be visible (G).14. Fingers touching mouth area: not

counted if bites nails while touchingmouth.

15. Sucking or chewing object : notfingernails (G,P).

16. Lip contortions.17. Trembling lip (B).18. Gratuitous hand movement at ear area

(G,I,J,N,P).19. Gratuitous hand movement at top of

head (G,I,J,N,P).20. Gratuitous hand movement at an object

separate from body or at a part ofclothing separate from body(G,IJ,N,P).

21. Gratuitous hand movement at some partof body (not ear, hair, mouth, orgenitals) (G,IJ,N,P).

22. Gratuitous hand movement (N).23. Gratuitous leg movement (M, N).24. Gratuitous foot movement : below ankles,

distinguish from foot merely movingalong with leg (M,N).

25. Trunk contortions (e.g., arching back)(N).

0-0

0-5

0-0

0-17

0-18

0-120-90-20

0-20

0-20

.99*

.96*

.81*

.82*

.52*

.89*

.49*

.89*

.99

.91

.76

.75

.80

.81

.66

.81

.46

0-20-00-20-10-0

ncnc.69*ncnc

ncnc.97ncnc

ncnc.17ncnc

0-11

0-3

0-5

0-30-13

.67*

.74*

.82*

.83*

.94*

.87

.98

.95

.98

.89

.39

.50

.22

.46

.63

.79

0-10-130-0

0-3

0-3

.47*.67*nc

.56*

.62*

.99

.76nc

.96

.95

.33

.44nc

.39

.30

.61

.56

.35

.77

.56

.68

ANXIETY IN CHILDREN 1251

Table 1 (continued)

ItemRange offrequency

Interraterreliability"

Concor-dance 1

Concor-dance 2

26. Rigid posture: Part of body is heldunusually stiff or motionless for theentire 30-sec interval (B,G,N). 0-11 .77* .94 .16

27. Masturbation: touches genital area (K). 0-3 .81* .99 .5028. Fearful facial expression (E,I). 0-4 .92* .98 .2629. Distraction: Must be indicated by a

verbal reminder by the examiner tothe child to pay attention (B,I,L,P). 0-7 .86* .94 .38

30. Avoidance of eye contact: Examinershould be having clear trouble makingeye contact with child (G, M). 0-1 -.04° .98 .00

Total 26-107 .78* .92 .58

Note. Letters in parentheses following each item description refer to studies suggesting that item as ananxiety indicator. The following code was used: A: Boland (1953); B: Buss, Wiener, Durkee, and Baer(1955); C: Cowen, Zax, Klein, Izzo, and Trost (1965); D: Endler, Hunt, and Rosenstein (1962); E: Fink(1956); F: Grossman (1968); G: Insel and Spencer (1972); H: Kasl and Mahl (1965); I: McReynolds(1965); J: Melamed and Siegel (1975); K: Miller, Barrett, Hampe, and Noble (1971); L: Nottelman(1975); M: Paul (1966); N: Raskin (1962); 0: Santostefano (1960); P: Tamaroff (1976); Q; Wolff (1969).Items 14 and 16 were suggested by our clinician consultants. The range of frequency column gives the lowestscore obtained by any subject and the highest score obtained by any subject. The interrater reliabilitycolumn gives the correlation coefficients for the Pearson correlations between the scores of the two observers.The Concordance 2 column gives the interrater agreements on the occurrence of indicators in pairs ofadjoining intervals (the number of agreements divided by the sum of agreements plus disagreements).The Concordance 1 column gives the interrater agreement on both the occurrence and nonoccurrence ofindicators in pairs of adjoining intervals (again, the number of agreements by the sum of agreements plusdisagreements)." n = 33 for interrater reliability and concordance calculations.b nc = not calculated due to infrequent occurrence (i.e., no children or only one child had nonzero scoreson the indicator).0 Based on only two children with nonzero scores.*p < .001.

that she wanted him or her to go with her to an- preventing observers from habituating to the be-other room "to do some tests."x havioral indicators, this procedure was used to aid

The testing room was 5.7 m X 3.5 m, with one- data analysis: Behaviors -were given a score of 1way mirrors on top of its four walls and a micro- for each interval in which they occurred,phone about 30 cm above the child's head for During the testing session, the examiner gavetransmitting sounds from the testing .room to the each child the three tests (Digits, Blocks, and Sen-observation areas. A male and a female observer, tences) in standard form and in as neutral a man-both in their early 30s, sat behind the one-way ner as possible. The examiner was instructed to trymirrors at a distance of about 1.2 m from the child. to keep the children in the testing room for at leastThe observers sat around a corner from each other, 10 minutes. The Digits test was taken from theand were separated by about l.S m and a parti- Illinois Test of Psycholinguistic Abilities (Kirk, Mc-tion. Each observer spoke softly into a tape recorder Carthy, & Kirk, 1968) and involved repeating aeach time he or she observed the child emit one of series of orally presented digits from memory. Sen-the 30 behavioral indicators. The two observers tences were taken from Tamaroff's (1976) adaptationcould not hear each other speaking. Also, the ob- of the Sentences subtest of the Wechsler Preschoolservers were blind to the independent anxiaty ratings and Primary Scale of Intelligence (Wechsler, 1967).of the children whom they observed and to the This tes,t requires the child to repeat increasinglyfact that lower anxiety ratings were expected fox complex orally presented sentences from memory,the second session in comparison to the first session. Finally, Blocks was taken from Tamaroff's (1976)

A standard time-sampling procedure involving 30-sec intervals was used. A Davis Scientific Instru-ments General Purpose Time Interval Generator l Two children refused to do any tests in the first(Model BS01) emitted to each observer a beep and session, and one refused to do any in the seconda red-light flash at the 30-sec intervals. Aside from session.

1252 BLAIR GLENNON AND JOHN R. WEISZ

adaptation of the Block Construction Test of theYale Scale of Child Development. It requires thechild to copy visible block constructions with sepa-rate blocks, and it is timed.

In the second testing session, the same procedurewas followed, except for two variations. First, thechild was accompanied to the testing room by hisor her mother as well as by the examiner, andthe mother sat unobtrusively reading a magazinein the corner of the testing room during the session.Second, only the female observer was behind theone-way mirrors, since the second observer wasused just during the first session to obtain inter-rater reliabilities. Of the 36 children in Session 1,32 participated in Session 2. One child refused,and the parents of 3 children were unable to par-ticipate. Technical difficulties with recording equip-ment fur ther reduced the sample ( fo r whom com-plete data on both sessions were available) to 29.

Training for the observers. Both observers spent2-3 hours memorizing the behavioral indicators and6 hours in observation-training sessions. In training,the observers watched videotapes of three childrenbeing tested as other children would be tested inthe actual experiment. Training also involved pilotobservations of three children in the actual settingand conditions of the experiment. Finally, the ob-servers spent approximately 2 hours discussing dis-agreements in their observations and ways to achievebetter consensus. The detailed description of theitems of the POSA given in Table 1 includes all ofthe details that the observers devised in order tomaximize agreement.

Results

Scores for all subjects on the POSA werecalculated by using the number of 30-secintervals in which a given indicator occurredduring the first 10 min (20 intervals) ofthe sessions. The examiner had been in-structed to try to keep each subject in thetesting room and working for a full 10 min;thus, a 10-min interval was chosen as thetarget time, in part, because of concern thatthe experimenter's behavior might havechanged significantly after the 10-min periodhad expired.2

Interrater Reliabilities

To assess interrater reliability, Pearsoncorrelations between the ratings of the twoobservers were calculated using the Session 1scores for each indicator separately and forthe sum of all 30 indicators (see Table 1third column from the right). The most im-

portant correlation coefficient, that for the30 indicators together, was .78 (p < .001).The intraclass correlation was .77 (p <.001).

Even though the preceding analysis an-swered the central question about reliabilityof overall POSA and individual indicatorscores, we also sought to learn .the level ofspecificity at which observer agreement tookplace. Toward this end we used a demandingprocedure designed to gauge the degree ofconcordance between observers within ob-servation intervals. The procedure involvedgrouping every 2 adjoining intervals (result-ing in 19 interval groups) and counting howoften the observers agreed or disagreed asto whether an indicator occurred within eachinterval group. Adjoining rather than singleintervals were used for this agreement mea-sure, since the observers sometimes reportedthe same behavior at slightly different timesso that the interval cutoff occurred betweentheir reports. Using this method of measure-ment, a quotient of concordance was calcu-lated by dividing the number of agreementsbetween the observers by the number oftheir "agreements" plus "disagreements" foreach indicator separately and for the totalof all 30 indicators. Table 1 (last two col-umns) shows the results of these calculations(a) when agreements between observers thata given indicator did not occur were includedin .the "agreement" score and (b) when suchnegative agreements were excluded from theagreement score. Considering the rigorousnature of the procedure (particularly babove), the percentages shown in Table 1represent rather substantial agreement be-tween observers at the level of brief ob-servation intervals.

- The alternate scoring method of dividing thefrequency of intervals in which each indicator oc-curred during the total session by the number ofintervals in the total session would have posed anadditional problem, in that fatigue and familiaritycould have affected the scores of children havinglonger sessions. Unfortunately, using the 10-minutecutoff, all subjects were not engaged in the sameactivities during the target time, but this seemed tobe a less significant consideration than those notedabove.

ANXIETY IN CHILDREN 1253

Table 2Correlations Among Anxiety Measures and Test Scores

1.2.3.4.5.6.7.8.

POSAPARSEPPARGEN

TSAS

Self-ratingExaminer ratingBlocksDigits

2 3 4

.37* .30* .47**.41** .36*

.22

5

-.12.01

-.18-.05

6

-.02.06.07.19.14

7

-.04.05.03

-.07-.45**-.24*

8

.38*,28.05.07

-.35*-.10

.42**

Sentences

-.16.24.05

-.18-.24-.13

.62***

.72***

Note. POSA = Preschool Observation Scale of Anxiety; PARSEP = questions about the child's separationanxiety; PARGKN = questions about the child's general anxiety; TSAS = Teachers' Separation AnxietyScale.

* p < .05.**/> < .01.

*** p < .001.

Correlations With Independent Measuresoj Anxiety

As an initial step in assessing the validityof the POSA, the correlations of the POSAwith the PARSEP, PARGEN, and TSAS werecalculated using data from the first session(since correlations with data of the secondsession would have been less meaningful dueto the mothers' presence). As Table 2 indi-cates, all three correlations were significant.Other correlations noted in the table are alsoof interest. Note that the self-rating andexaminer ratings were not correlated with oneanother or with POSA, PARSEP, PARGEN, orTSAS scores. Thus, even though the POSAdid meet the first set of validity criteria thatwe established (i.e., significant correlationswith the three inventory measures), the self-ratings and examiner ratings did not. Thesignificant negative correlations between self-ratings and two of the cognitive tests sug-gests that self-ratings may have been influ-enced by the children's awareness of thequality of their test performance.

To assess the contributions of individualPOSA indicators to the correlations betweenthe scale and the three inventory measures,the correlations of each of the 30 items withthe three inventory scores were calculated.Most of these individual correlations werenonsignificant, suggesting that the predictivepower of the indicators lies mainly in theircombination with one another. Those corre-

lations that did attain statistical significancewere lip contortions with PARSEP (.39, p <.01) and TSAS (.43, / > < .01), gratuitoushand movement in the ear area with PARGEN(.32, p < .OS), gratuitous hand movementat the top of the head with PARSEP ( — .38,p < .01), gratuitous hand movement towardobject with TSAS (.53, p < .001), gratuitoushand movement at other body part ( — .31,p < .05), gratuitous leg movement with PAR-SEP (.34, p < .05) and TSAS (.35, p < .05),gratuitous foot movement with PARSEP (.35,p < .05) and TSAS (.46, p < .01), trunkcontortions with PARSEP (.29, p < .05) andTSAS (.42, p < .01), and masturbation withPARGEN (.39, p< .01) and TSAS (.58, p<.001).

Anxiety Scores in the First andSecond Sessions

As a second way of assessing ,the validityof the POSA, the mean POSA score for allchildren in the first session was comparedwith the mean score for all children in thesecond session (designed to be less anxietyproducing than the first). As predicted, thechildren obtained significantly higher POSAscores in the first than in the second session,t(2&)-2.53, p < .01 (one-tailed). Of the29 children, 22 had higher scores in the firstthan in the second session, x 2 ( l ) = 6.76, p< .01. Further comparisons between the firstand second sessions were m'ade for children

1254 BLAIR GLENNON AND JOHN R. WEISZ

Table 3Comparisons of POSA Scores Between Sessions

Session 1 Session 2

Group M SD M SD df

All childrenHighHighHigh

PARSEP scorersPARGEN scorersTSAS scorers

61.70.64,61.

40.3,4

21.319.19.22.

6,2.4

51. 349.749.047.6

14.15.14,14,

,0,5,3,0

2.4.3.2,

,53*,35**,17*,66*

28141517

Note. POSA = Preschool Observation Scale of Anxiety; PARSEP = questions about the child's separationanxiety; PARGEN = questions about the child's general anxiety; TSAS = Teachers' Separation AnxietyScale.

* p < .01, one-tailed.** p < .001, one-tailed.

who scored above the median on the PARSEP,PARGEN, and TSAS, since it was thought thatthese children might be especially sensitiveto the situational manipulations of stressors.Again, the high scorers on the three question-naire scales showed significantly more be-havioral indicators of anxiety in .the firstthan in the second session. The results forthese analyses are shown in Table 3. Notethat in each of the three groups, the magni-tude of the Session 1 - Session 2 difference isgreater than for the entire sample. In fact,of the four groups, children rated by theirparents as high in separation anxiety showedthe highest mean POSA scores during Session1 with mothers absent and the largest Ses-sion 1 - Session 2 difference—more than dou-bling the difference shown by the full sample.

Session 1 - Session 2 differences were alsocalculated for the children's self-ratings andthe examiner ratings. The change in the chil-dren's self-ratings from Session 1 to Session2 indicates a nonsignificant increase in anxi-ety, whereas the change in the examiners'ratings indicates a highly significant decrease.Anxiety ratings on the children's 6-point pic-ture scale averaged 2.22 in Session 1 and 2.59in Session 2 (p = .28). Ratings on the ex-aminer's 6-point scale averaged 3.07 in Ses-sion 1 and 2.00 in Session 2, t ( 2 9 ) - 3.71,p < .001. Thus, both the POSA and examinerratings of anxiety met the second validitycriterion, that is, significantly higher scoresin Session 1 than in Session 2.

Finally, Session 1 - Session 2 differencesfor each of the 30 POSA indicators were

calculated in an effort to gauge the contri-bution of the individual indicators to theoverall sessions difference in total POSAscores. As was true in the correlational analy-sis reported earlier, most individual itemeffects were nonsignificant, suggesting thatthe discriminative power of the indicatorslies principally in their combination with oneanother. However, there were five indicatorsthat had significantly different frequencies inSessions 1 and 2 (all in the predicted direc-tion): silence to one question (p < .05),touching mouth area (p < .01), gratuitousarm movement (p < .001), trunk contortions(p < .001), and rigid posture (p < .05).

Discussion

The results of the present study supportthe use of the POSA as a measure of anxietyin young children. Independent judgesachieved strong agreement on both totalPOSA scores and total scores for most ofthe 30 individual indicators. The large num-ber of indicators used seemed .to interferewith interobserver concordance at the microlevel of 1-min observation blocks (at leastby the most rigorous method of analysis);and this suggests that the two observers mayhave differed frequently in the particular be-havioral incidents that they observed. Yet,this is a relatively trivial limitation con-sidering the high interobserver correlationsobtained for total POSA and individual indi-cator scores. In sum, the difficulties intro-duced by requiring observers to watch for

ANXIETY IN CHILDREN 1255

30 indicators appear to be outweighed by thevalue of including a variety of potentialanxiety manifestations in order to captureindications of anxiety in individuals withdiffering expressive styles.

Two types of evidence support the viewthat the POSA yields a valid index of anxi-ety. As predicted, the scale was significantlycorrelated with all three inventory measuresof anxiety. Also as predicted, the POSAyielded significantly higher scores during apresumably high-anxiety test session thanduring a session designed to provoke lessanxiety. The data generally supported ouroriginal belief that broad-based measuressuch as the POSA would outperform simplerapproaches such as self-ratings by youngchildren 'and global judgments by examiners.Children's self-ratings and examiner ratingsdid not correlate significantly with any ofthe three inventory measures, with POSAscores, or with each other. Self-ratingsshowed very slight differences (and in thewrong direction) between Sessions 1 and 2;however, Session 2 ratings by the examinerwere significantly lower than Session 1 rat-ings. It is uncertain whether this latter find-ing derived from an expectation by the ex-aminer that children's anxiety would be lowerin the second session with mothers present.However, whatever the basis for the finding,it constitutes the only bit of evidence sup-porting the use of either self- or examinerratings. This pattern of findings seems toindicate the superiority of structured ob-servations of carefully delineated behaviorsover global, unstructured, and thus subjec-tive ratings, though this conclusion must bequalified by the fact that the structuredobservations were made by trained observers,whereas the global ratings were not.

As indicated in the Introduction, the rela-tion between anxiety and problem-solvingperformance appears to be quite complex.In the present study none of the inventorymeasures was significantly related to test per-formance, and the POSA was significantlyrelated only to Digits performance (r = .38).So, anxiety, as measured by the POSA andinventory scores, did not appear to have adebilitating effect on test performance for

children in the present study, and may havehad an enhancing effect on one test, perhapsdue to factors related to the high socioeco-nomic status of the present subjects.

Further research with the POSA shouldinclude investigations of the scale's capacityto reflect the effects of stressors other thanthose devised in the present study. Childrenvarying more widely in age and other demo-graphic characteristics than those of thepresent sample should be included. And, moreimportantly, there is a need to assess thescale's usefulness with clinical populations ofchildren who (unlike those of the presentsample) suffer from pronounced behaviorproblems. In addition, research using longerobservation periods than the 10-minute-plussessions of the present study could be useful;some behaviors that did not occur in therelatively brief sessions of the present in-vestigation might prove to be useful indicesof anxiety if children were given a lengthieropportunity to display them. Even the be-haviors .that occur infrequently might wellprove to be potent anxiety indicators whenthey do appear. Finally, the degree of anxietyis apt to be reflected not only by the fre-quency of anxiety behaviors but also by theirintensity. Although intensity may be difficultto quantify, its potential for increasing theprecision of anxiety measurement would seemto justify efforts in this direction.

For the present, however, the POSA rep-resents a potentially useful approach to theassessment of anxiety in children. Given thesensitivity of the POSA to situational varia-tions in stressors, the scale itself should beparticularly useful to reseachers concernedwith the interplay between specific situa-tional conditions and affective states (e.g.,Spielberger et al., 1972). In addition, thePOSA could contribute usefully to our under-standing of sex differences in anxiety mani-festations at various ages (see Maccoby &Jacklin, 1974, pp. 182-190), and to ourcapacity to evaluate therapeutic techniquesfor children (see Achenbach, 1974, pp. 606-6SO). Research on these topics, focusing onanxiety states in children, clearly must cir-cumvent the problems of projective tech-niques, self-reports, global and subjective

1256 BLAIR GLENNON AND JOHN R. WEISZ

observer reports, and physiological measures,all outlined in the Introduction. The evidencepresented in the present study suggests thatinstruments such as the POSA may providelogical, valid, 'and reliable alternatives tothese more traditional approaches.

Reference Note

1 Hollingshead, A. B. Two-factor index of socialposition. Unpublished manuscript, Yale Univer-sity, 19S7.

References

Achenbach, T. M. Developmental psychopathology.New York: Ronald Press, 1974.

Boland, J. L. A comparison of stutterers and non-stutterers on several measures of anxiety. SpeechMonographs, 1953, 20, 144.

Buss, A., Wiener, M., Durkee, A., & Baer, M. Themeasurement of anxiety in clinical situations.Journal of Consulting Psychology, 1955, 19, 125-129.

Cowen, E. L., Zax, M., Klein, R., Izzo, L. D., &Trost, M. A. The relation of anxiety in schoolchildren to school record, achievement, and be-havioral measures. Child Development, 1965, 36,685-695.

Denny, J. P. Effects of anxiety and intelligence onconcept formation. Journal of Experimental Psy-chology, 1966, 72, 596-602.

Doris, J., Mclntyre, A., Kelsey, C., & Lehman, E.Separation anxiety in nursery school children.Proceedings of the 79th Annual Convention ofthe American Psychological Asso-ciation, 1971,79, 145-146. (Summary)

Endler, N. S., Hunt, J. M., & Rosenstein, A. J. AnS-R Inventory of Anxiousness. Psychological Mon-ographs, 1962, 7(5(17, Whole No. 536).

Feldhusen, J. F., & Klausmeier, H. J. Anxiety, in-telligence, and achievement in children of low,average, and high intelligence. Child Development,1962, 33, 403-409.

Fink, R. M. How children tell us they arc afraid.Childhood Education, 1956, 32, 264-266.

Fischer, D. G., & Awrey, A. Manifest anxiety, testanxiety, and intelligence in concept formation.Journal of Social Psychology, 1973, 89, 153-154.

Grossman, B. D. Anxiety as a factor in the child'sresponsiveness to social reinforcement. Journal ofGenetic Psychology, 1968, 112, 165-174.

Hodges, W. F., & Durham, R. L. Anxiety, ability,and digit span performance. Journal of Person-ality and Social Psychology, 1972, 24, 401-406.

Insel, S. A., & Spencer, T. D. Emotional concomi-tants of reading readiness thinking in kindergar-ten. Washington, D. C.: National Center for Edu-cational Research and Development, December1972. (ERIC Document Reproduction Service No.ED 072 871)

Kasl, S. V., & Mahl, G. F. The relationship ofdisturbances and hesitations in spontaneous speechto anxiety. Journal of Personality and Social Psy-chology, 1965, 1, 425-433.

Katahn, M. Interaction of anxdety and ability incomplex learning situations. Journal of Person-ality and Social Psychology, 1966, 3, 475-479.

Kirk, S. A., McCarthy, J. J., & Kirk, W. D.Illinois Test of Psycholinguistic Abilities. Urbana:University of Illinois Press, 1968.

Lazarus, R. S. Psychological stress and the copingprocess. New York: McGraw-Hill, 11966.

Maccoby, E. E., & Jaddin, C. N. The psychologyof sex differences. Stanford, Calif.: Stanford Uni-versity Press, 1974.

Mazzei, J., & Goulet, L. R. Test anxiety and in-telligence in concept formation. Psychological Re-ports, 1969, 24, 842.

McReynolds, P. On the assessment of anxiety: I.by a behavior checklist. Psychological Reports,:1965, 16, 805-808.

McReynolds, P. The assessment of anxiety: A sur-vey of available techniques. In P. McReynolds(Ed.), Advances in psychological assessment (Vol.1). Palo Alto, Calif.: Science and Behavior Books,1968.

Mclamed, B. G., & Siegel, L. J. Reduction of anxi-ety in children facing hospitalization and surgeryby use of filmed modeling. Journal oj Consultingand Clinical Pychology, 1975, 43, 511-521.

Miller, L. C., Barrett, C. L., Hampe, E., & Noble,H. Revised anxiety scales for the Louisville Be-havior Check List. Psychological Reports, 1971,29, 503-511.

Nottelman, E. Test anxiety and off-task behaviorin evaluative situations. Washington, D.C.: Na-tional Institute of Education, April 197S. (ERICDocument Reproduction Service No. ED 113022)

Paul, G. L. Insight vs. desensitization in psycho-therapy. Stanford, Calif.: Stanford UniversityPress, 1966.

Phillips, B. N., Martin, R., & Meyers, J. Interven-tions in relation to anxiety in school. In C. D.Spielberger (Ed.), Anxiety: Current trends intheory and research '(Vol. 2). New York: Aca-demic Press, 1972.

Raskin, A. Observable signs of anxiety or distressduring psychotherapy. Journal of Consulting Psy-chology, 1962, 26, 389.

Santostefano, S. Anxiety and hostility in stuttering.Journal of Speech and Hearing Research, 1960,3, 337-347.

Sarason, S. B. The Measurement of anxiety inchildren: Some questions and problems. In C. D.Spielberger (Ed.), Anxiety and behavior. NewYork: Academic Press, 1966.

Sarason, S. B., Davidson, K. S., Lighthall, F. F., &Waite, R. R. Rorschach behavior and the per-formance of high and low anxious children. ChildDevelopment, 1958, 29, 277-285.

Sarason, S. B., Davidson, K. S., Lighthall, F. F.,Waite, R. R., & Ruebush, iB. K. Anxiety in ele-mentary school children. New York: Wiley, 1960.

ANXIETY IN CHILDREN 1257

Spielberger, C. D. The effects of anxiety on com-plex learning and academic achievement. In C.D. Spielberger (Ed.), Anxiety and behavior. NewYork: Academic Press, 1966.

Spielberger, C. D. Anxiety as an emotional state.In C. D. Spielberger (Ed.), Anxiety: Currenttrends in theory and research (Vol. 1). NewYork: Academic Press, 1972.

Spielberger, C. D., et al. Needed research on stressand anxiety. A special report of the OSOE-sponsored grant study: Critical appraisal of re-search in the personality-emotions-motivation do-main (IB;R Report No. 72-10). Washington,DjC.: Office of Education, July 1972. (ERICDocument Reproduction Service No. ED 113 649)

Tamaroff, M. H. Separation anxiety and cognitivefunctioning in nursery school children. (Doctoral

dissertation, Cornell University, 1975) Disserta-tion Abstracts International, 1976, 36, S820-B.(University Microfilms No. 76-9569)

Wechsler, D. Wechsler Preschool and Primary Scaleof Intelligence. New York: Psychological Corpo-ration, 1967.

Weisz, J. R. Transcontextual validity in develop-mental research. Child Development, 1978, 40,1-12.

Wolff, S. Children under stress. London: PenguinPress, 1969.

Young, F. A., & Brown, M. Effects of test anxietyand testing conditions on intelligence test scoresof elementary school boys and girls. PsychologicalReports, 1973, 32, 643-649.

Received August 17, 1977 •