A Multimethod Study of Problem Behavioramong Thai and American Children in School:Teacher Reports versus Direct Observations

John R. WeiszUniversity of California, Los Angeles

Wanchai Chaiyasit

Child Mental Health Center, Bangkok, Thailand

Bahr WeissVanderbilt University

Karen L. EastmanUniversity of California, Los Angeles

Elizabeth W. JaeksonUniversity of North Carolina, Chapel Hill

WEISZ, JOHN R.; CHAIYASIT, WANCHAI; WEISS, BAHR; EASTMAN, KAREN L. ; and JACKSON, ELIZABETHW. A Multimethod Study of Problem Behavior among Thai and American Children in School:Teacher Reports versus Direct Observations. CHILD DEVELOPMENT, 1995, 66,402—415. Previousliterature describes Thai children as unusually polite, deferent, and behaviorally restrained. Yet,in a recent study employing teacher reports, Thai children were reported to show many morebehavior problems than American children. Such a finding may reflect culture-linked differencesin the perspective of Thai versus American teachers. To explore this possibility, we used trainedobservers to conduct direct observations of Thai and American children's school behavior, andwe obtained teacher reports on the same children. Observational resuits were precisely theopposite of previous and present teacher-report findings: Observers reported twice as muchproblem behavior and off-task behavior in American children as in their Thai age-mates. Thispattern may reflect Thai-U.S. differences in teachers' style, societal values and practices, evenchild temperament. The findings support the value of direct behavior observation in cross-national research on child problems.

In efforts to transcend a "monocultural ribbean (Lambert, Weisz, & Knight, 1989),science" (Kennedy, Scheirer, & Rogers, Australia (Achenbach, Hensley, Phares, &1984), researchers have begun building a Grayson, 1990), Asia (Weisz, Suwanlert,base of data on child development and be- Chaiyasit, Weiss, Achenbach, &c Walter,havior in diverse cultures. The field of de- 1987), Europe (Achenbach, Verhulst, Baron,velopmental psychopathology has been a & Akkerhuis, 1987), and Africa (Weisz, Sig-part of this Zeitgeist, with researchers gener- man, Weiss, & Mosk, 1993).ating data on children's problem behavior incultures around the world, including North The data for most of this research haveAmerica (Achenbach & Edelbrock, 1983), consisted of reports on child problems fromSouth America (Montenegro, 1983), the Ca- individuals who are personally involved in

We dedicate this study to the memory of our friend, Frank Bostancic. The investigation wassupported by a grant from the National Institute of Mental Health (5 ROl MH38240). We thankKimberly Cane, Janet Clarke, Mongkoi Lugkam, Meredith Mullins, Valerie Proffitt, SomsongSuwanlert, and Ampai Thong ngen for their help in various aspects of the study. Correspondenceregarding the study should be addressed to John Weisz, Department of Psychology, Franz Hall,UCLA, 405 Hilgard Avenue, Los Angeles, CA 90024-1563.

[Child Development, 1995, 66,402-415. © 1995 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/95/6602-0012$01.00]

Weisz et al. 403

the lives of the youngsters being studied. Innearly all such studies, these individuals arethe children's parents (e.g., Achenbach etal,, 1990) or teachers (e.g., Weisz, Suwanlert,Chaiyasit, Weiss, Achenbach, et al , 1988).Certainly, parents and teachers have impor-tant perspectives on the behavior of the chil-dren they see so regularly, and informationfrom these sources is valuable. On the otherhand, it is probably unwise to rely exclu-sively on such reports, particularly in re-search that compares samples from differentcultures. As has been noted elsewhere (e.g.,Weisz, 1989; Weisz, Suwanlert, Chaiyasit,Weiss, et al., 1987), adults' reports on chil-dren's behavior and problems may be influ-enced by cultural milieu in a number ofways; societal values and expectations aboutchild development may color such adultjudgments as whether a particular child hasshown a particular behavior or problemmore than the average child, which behavioris unusual or abnormal, and whether certainbehavior has occurred more often than is ap-propriate for children of a particular age.Such influences may lead to an incompleteor distorted picture of cross-national differ-ences in child behavior.

A possible case in point can be found inresearch on child behavior in Thailand andtlie United States. To explain the problem,we need a brief primer on Thai culture andchild socialization. The population of Thai-land is 95% Buddhist; children are exposedto Buddhist training at home, school, andtemple. Consistent with Thai Buddhist pre-cepts, children are reared to be nonaggres-sive, obedient, and respectful of others—particularly authority flgures (e.g., teachers)and others who are older than they—and toavoid expressing anger or other strong emo-tions (Gardiner, 1968; Gardiner & Suttipan,1977; Moore, 1974; Sangsingkeo, 1969; Su-vannathat, 1979; Suwanlert, 1974; Weisz,1989). Children are taught to strive forkrengchai, an attitude of humility and self-effacement intended to avoid disturbing orinconveniencing others (Phillips, 1965; Su-vannathat, 1979). The attitude is embodiedin the tt>a«—a respectiful bow with handspressed together in a prayerful position—with which social interactions in Thailandbegin and end. As conveyors of both infor-mation and societal values, teachers are ac-corded special honor in Thai society; par-ents defer to teachers' judgments about theirchildren, children are taught to obey andhonor their teachers, both parents and chil-

dren use honoriflc terms (e.g., achaan) to re-fer to teachers, and the entire profession iscelebrated on National Teachers Day. Takentogether, the literature suggests that Thaichildren—compared to, say, American chil-dren—should show unusually high levels ofself-control and deference, and unusuallylow levels of problem behavior, particularlyin the presence of their teachers at school.However, research thus far has found justthe opposite.

Weisz, Suwanlert, Chaiyasit, Weiss,Achenbach, et al. (1988), relying on teacherreports of child behavior, found that Thaiprimary school children were rated as show-ing much more problem behavior—includ-ing more conduct problems—than theirAmerican counterparts. It is certainly possi-ble that this flnding reflects real Thai-U.S.behavioral differences. However, the Bnd-ings run counter to the literature notedabove and clash with our own impressionthat Thai children are more orderly, atten-tive, and well-behaved in their classes thanare American children. Such considerationsraise the possibility (noted in Weisz, Su-wanlert, Chaiyasit, Weiss, Achenbach, et al.,1988) that Thai teachers are stricter thanAmericans in their expectations and stan-dards for children—that is, that Thai teach-ers' views on what is average, appropriate,and/or "normal" child behavior lead themto set lower thresholds for problem identifi-cation than the thresholds set by Americanteachers. If this were the case. Thai teachersmight well report higher levels of problembehavior than would American teacherseven if both groups of teachers were observ-ing the same child behavior at school, andpossibly even if Thai children were actuallybetter-behaved or less problem-prone thanAmerican children.

One way to explore the plausibility ofsuch alternative explanations is to collectstructured observations of child behavior inThailand and the United States, using ob-servers who (a) have been trained to followwell-specified rules in their reports and(b) have not developed personal relation-ships with the children that might color theirreports of the behavior they observe. Thiswas the main objective of the present study.In developing the observational method forthis study, we faced a number of challengesraised by our emphasis on problem behaviorand by the cross-national nature of the proj-ect. One challenge was that some of the most

404 Child Development

clinically significant behaviors of childrenare low in base rate and thus unlikely toshow much variability during the observa-tional periods that are typically available toobservational researchers. A second chal-lenge, applicable to cross-national research,was the impracticality of training a team ofobservers and transporting them from coun-try to country to do the research; we triedvideotapes but found that they could notcapture individual child behavior or class-room activity as fully as live observers could.A third challenge was that observers wererequired who were fluent in the two lan-guages involved and who had some familiar-ity with customs, conventions, slang expres-sions, gestures, and the like, in the twocultures. A fourth challenge was that the listof behaviors to be observed needed to in-clude problems of clinical significance in thetwo countries.

The study was designed to address eachof these challenges. Because many clinicallysignificant problems are low in base rate(challenge 1), we trained the observers toidentify more than 100 specific problem be-haviors and then focused our primary analy-ses on problem summary scores. Because weneeded observers who were fluent in thelanguage and familiar with the culture of thechildren they observed, but we also neededat least one observer who was involved inthe observations in both countries (chal-lenges 2 and 3), we used an observer whowas well versed in the language and customsof both countries, and this observer servedas the reliability standard for both the Amer-ican and Thai observer teams. Finally, to in-sure that our sample of child behaviors to beobserved included problems that are clini-cally significant in the two cultures (chal-lenge 4), we drew from previous researchwith child clinical samples in both countries(Weisz, Suwanlert, Chaiyasit, & Walter,1987) and from expert clinician opinions toassemble the final list of behaviors.

The study included attention to the pos-sible impact of gender. In Thailand, boys ap-pear to receive more rigorous training thangirls in Buddhist principles and ideals, in-cluding the ideals of nonaggression, quies-cence, and self-control. For example, boysbut not girls are expected to live in a templefor a period of priesthood during their youth.Is such gender-differentiated training asso-ciated with a difEerent pattern of gender dif-ferences in Thailand than in the UnitedStates? Previous Thai-U.S. research (e.g.,Weisz, Suwanlert, Chaiyasit, Weiss, et al..

1987; Weisz, Suwanlert, et al., 1993) has notfound such differential differences, but theobservational methodology used here pro-vided a fresh opportunity to test the possi-bility.

So, the study took the following form.Teams of trained observers in Thailand andthe United States, with one shared teammember, coded multiple problems of ele-mentary school children in classroom set-tings. Individual children were the focus ofthe observations, and the child (not theclass) was the unit of analysis. We then ex-amined the observed prevalence of the prob-lems as a function of culture (Thai vs. UnitedStates) and gender. To fill out the picture ofThai and American child behavior in class-room settings, we also coded and analyzedthe children's on-task versus off-task behav-ior. Finally, to provide a comparison of ob-server reports and teacher reports for thesample of children, we obtained standard-ized problem reports from teachers on mostof the children for whom we had observa-tional data. This made it possible to assessproblem reports (on problem items that ap-peared on both the observer list and theteacher list) as a function of source (i.e., ob-server vs. teacher), culture, and the theoreti-cally important source x culture interaction.

MethodSubjects and Design of the Study

The sample included 144 children, 36in each cell of a 2 x 2 (culture x gender)design; thus, culture and gender were or-thogonal factors. The children were all pu-pils in elementary schools. To minimize thechance that results might reflect the idiosyn-crasies of any one school, we drew childrenfrom multiple schools in each country—sixin Thailand and five in the United States.All the Thai children were ethnic Thai; theAmerican sample, reflecting the diversity ofthe population, was 68% Caucasian and 31%African-American. Within the U.S. sample,we checked for Caucasian versus African-American differences in off-task ratings andtotal problem scores from the observationalmeasure (see below) and for interactions in-volving race in analyses of Over- and Under-controlled problems and observer versusteacher differences (see below). The onlysignificant race effect was found for off-taskratings; these are discussed in the relevantsection of the Results (below).

Children in the full sample ranged inage from 5 to 11, with an overall mean of

Weisz et al. 405

8.13 (SD = 1.93). We tested for group differ-ences, via a 2 X 2 (culture x gender) GLMANOVA on age; the analysis revealed no sig-nificant main effects or interactions (all ps >.50). Moreover, a nonsignificant chi-squarevalue for the (7 x 2) yearly age level x cul-ture table, x^(6, N = 144) = 1.14, p > .5,revealed no Thai-U.S. difference in age dis-tiibution.

We sought an SES classification systemtliat could be used to check for group differ-ences; none of the five SES systems pro-posed by various investigators for Thailandhad validity data or coding instructions de-tailed enough to permit us to apply the sys-tem to our data. So, we applied Hollings-head's (1975) nine-level SES scheme (9 =highest SES) for parent occupations to ourThai and U.S. samples. The Thai meanshould b(; interpreted with caution; it re-fiects use of an American system in a culturefor which it was not designed. A 2 x 2 (cul-ture X gender) ANOVA of SES scoresshowed only a main effect of culture, F(l,137) = 12.19, p < .001; the U.S. mean (5.70)was higher than the Thai mean (4.31), re-flecting a well-known difference between"developed" and "developing" countries.

Classes, Teachers, and Conditions duringData Collection

In the schools observed in both coun-tries, children stay with their primary(homeroom) teacher during most of eachschool day, and they change homerooms andprimary teachers each calendar year. Therewere about 35 students in most of the Thaiclasses; some U.S. classes contained 30 chil-dren, but most were smaller. Observationsand teacher reports were collected for thestudy after children had been with their pri-mary teachers for 2 months or more. Thaiteachers v/ere all ethnic Thai, and 87% werefemale; 76% had bachelor's degrees, 19%teacher training or vocational certificates(i.e., less than a bachelor's degree), and5% had master's degrees. American teach-ers were 80% Caucasian, 20% African-American, and 95% were female; 90% hadbachelor's degrees, and 10% had master'sdegrees. In both Thailand and the UnitedStates, classes involved a mixture of lectur-ing, questi^ons to and from the students, and

projects for groups of students to work ontogether. Most classrooms in both countrieswere laid out with combinations of group ta-bles, work stations, and individual, paired,or joined desks; school days in both coun-tries included some time at desks and sometime at tables or work stations, or sitting onthe fioor with groups. It is possible that Thaiteachers showed a more authoritative and di-recting style in interactions with their pupilsthan did American teachers, but we did notsystematically assess teacher style.

Observational Form and ObservationalProcedure

The observers used an observation formdeveloped for this study and patterned inpart after the Direct Observation Form(DOF) of the Child Behavior Checklist (seeAchenbach, 1986). The procedure involvedtime sampling, with one observational formused for each 10-min episode. The form in-cluded:

1. On/Off-task ratings.—The form in-cluded 10 boxes, with "on" and "off" writ-ten inside each. The observer's in-the-eartone sounded after each minute, at whichpoint the observer circled "on" if the childwas on-task and "off" if not; when the lastbox was filled, the 10-min observational epi-sode had ended.

2. Individual problem ratings.—Theform listed 111 individual problem behavioritems (e.g., makes odd noises, argues, teases,physically isolates self from others); besideeach item there was space for a rating of 0(not observed), 1 (slight or ambiguous occur-rence), 2 (definite occurrence with moderateintensity and less than 3 min duration), or 3(definite occurrence with strong intensity orgreater than 3 min duration).' All of the 96specific behaviors listed in the Achenbach(1986) Direct Observation Form were in-cluded in the present observation form, al-though some Achenbach items were modi-fied in an effort to clarify the observers' task.For example, to distinguish between Achen-bach items "9 Doesn't sit still, restless, orhyperactive" and "13 Fidgets," we changeditem 9 to "Doesn't sit still [body trunk move-ments]" and item 13 to "Fidgets [arm, hand,leg, or foot movements, including fiddling

' Using the same four-point rating scale, we also had observers make prosocial behaviorratings, using four categories: Helping or assisting, sharing or giving, praising or encouraging,and other efforts to benefit others. Classroom activities appeared to offer little opportunity forprosocial behavior. Ratings were 0 for 95.0% (for "helping or assisting") to 99.5% (for "otherefforts") of all observational sessions, and chance-corrected interobserver reliabilities were, ac-cordingly, not acceptable.

406 Child Development

with hair or objects in hands]," In additionto the Achenbach-based items, we addeditems based on problems listed by Thai par-ents in a survey of clinic referral problems(see Weisz, Suwanlert, Chaiyasit, & Walter,1987) and problems nominated as importantby Thai clinicians,^ For example, the Thaiemphasis on social harmony appeared to bereflected in such items as "69 Unwilling toshare things with others," and "109 Unfairto peers," "16 Impolite, rude, or verballyabusive to peers," and "32 Gets bullied byother children," The Thai emphasis on so-cial appropriateness was seemingly re-flected in "108 Clings to peers, or too depen-dent on peers," "111 Acts too old for his/her age," and "70 Risky or unsafe behavior,"The full list of 111 problem items was in-cluded in the reliability checks and mainanalyses reported below. Observers mayhave been aided in their task by the appar-ently low base rate of many problems; 47 ofthe items received a nonzero rating in fewerthan 1% of all observations in both countries(17 of these had no nonzero ratings).

Each child was observed for a minimumof four 10-min episodes; the range was 4-15,with a mean of 5,69 sessions per child. Ob-servational scores (e,g,, for Total problems)were computed as an average per 10-min ep-isode, so that all available observational datacould be used. Children were selected ran-domly from among those whose parents hadgiven permission, with the constraint thatwe maintain an age x gender balance. Allobservations occurred during regular aca-demic class activities; such nonacademic ac-tivities as lunch and recess were excluded.For each child, observation times were ran-domly selected from those class times schoolstaff made available, with the constraint thatno child be observed more than once on thesame day (this constraint was not applied toreliability assessment observations—see be-low). This randomization had the effect ofscattering observations throughout the day,morning and afternoon, in both Thailand(8:00 A,M,-3:30 P,M,) and the United States(8:30 A,M,-3:00 P,M,). The observation pe-riods were signaled by in-the-ear tones. Im-mediately after each 10-min observation epi-

sode, the observer fllled in one copy of theobservation form for the child.

Reliability Data on the ObservationalProcedure

The U,S, observer team consisted of fiveAmerican observers plus one Thai observerwho served as a reliability standard (RS) inboth countries; the RS was a clinical psy-chologist who was fluent in English andThai, had received his academic training inclinical psychology in the United States, andhad worked in clinical and school settingswith both American children in the UnitedStates and Thai children in Thailand, TheRS also served as a member of the observerteam for Thailand, which included two addi-tional Thai observers. We calculated inter-observer reliabilities separately for theAmerican observers versus the RS, and forthe Thai observers versus the RS, using In-traclass Correlation CoefBcients (ICCs) be-tween pairs of observers—that is, RS versusother; to aid in interpretation of the ICCs,we also report Pearson rs. For the U,S, ob-servers, the number of observational epi-sodes per pair ranged from 9 to 34, with amean of 16 and a median of 12; for the Thaiobservers, all observations were done intrios, and there were 47 observational epi-sodes. Overall, 12% of the total observa-tional time was observed for reliability pur-poses.

For On/Off-task ratings, ICCs for theU,S, observers ranged from 0.39 (low outlier,one observer pair with only nine shared ob-servations) to 0,97, with a median of 0.82 andweighted mean of 0.78, Corresponding Pear-son rs ranged from 0,44 to 0,97, with a me-dian of 0.82 and weighted mean of 0,79, On/Off reliability for the Thai observers was re-flected in an ICC of 0,94 for RS versus ob-server A and 0,95 for RS versus observer B(ICC for the two non-RS observers: 0,90),and Pearson rs of 0,95,

Total Problem score (i,e,, sum of all rat-ings across all problem items) ICCs for theU,S, observers ranged from 0,46 (same out-lier pair as above, based on nine observa-tions) to 0,76, with a median of 0,66 andweighted mean of 0,66; corresponding Pear-

^ In addition to the individual problem ratings, there was space at the end of the form foroverall ratings (0, 1, 2, or 3) for "overcontrolled or internalizing problem behavior," "undercon-trolled or externalizing problem behavior," and "overall level of problem behavior"; these wereincluded to provide a check on whether the individual problem ratings provided more reliableor valid information than that which could be obtained through simple global ratings. Theseglobal ratings did indeed prove unreliable, indicating that observation of the component individ-ual problem items was needed for empirically acceptable observational research of this sort.

Weisz et al. 407

son rs ranged from 0.43 to 0.78, with a me-dian of 0.77 and weighted mean of 0.71. TheTotal Problem score ICCs for Thai observerswere 0.64 for RS versus A and 0.76 for RSversus B (ICC for A vs. B was 0.62), with amean Pearson r of 0.75. Thus, the reliabili-ties all appeared to be acceptable, for boththe U.S. observer team and the Thai ob-server team. The ICCs for On/OfF-task weresomewhat higher, and the ICCs for TotalProblems were somewhat lower, than thecorresponding ICCs reported by Reed andEdelbrock (1983) for Achenbach's (1986)DOF, used with an exclusively Americansample.

Becaitse we assessed reliability by hav-ing multiple observers present for some ses-sions, it seemed useful to test whether thenumber of observers present during a ses-sion might have influenced children's prob-lem scores. We checked, separately for theThai and U.S. samples, by computing thecorrelation between number of observerspresent and total problem score. We did thecalculations for the U.S. sample once usingactual number of observers (1, 2, or 3, r =-0,16, df = 97) and once collapsing to 1versus >1 observers (r = 0.14, df = 97). Inthe Thai sample, only one analysis wasneeded because the number of observerswas either one or three; r for the Thai sam-ple -0.01 (df = 78). In all instances, thecorrelation between child problem scoreand number of observers was nonsignificant.

Teacher Report DataTeachers of the children in our Thai and

American samples were asked to provide re-ports of the children's problem behavior.American teachers filled in the problemportion of the Child Behavior Checklist—Teacher Heport Form (TRF; Achenbach &Edelbrock, 1986). This is a list of 118 prob-lem items (e.g., argues a lot, cries a lot, dis-turbs other pupils), each of which is to berated 0 ("not true of the pupil"), 1 ("some-what or sometimes true of the pupil"), or 2("very true or often true of the pupil"). Thaiteachers filled in the problem portion of theThai Youth Checklist (TYC; Weisz, Suwan-lert, Chaiyasit, Weiss, Achenbach, et al,,1988). This consists of the same 118 problemitems listed on the TRF, followed by 24 ad-ditional problem items based on clinic sur-veys in Thailand, and the input of Thai clini-cians. (Details of instrument developmentand the three-wave translation—back transla-tion process are provided in Weisz, Suwan-lert, Chiayasit, Weiss, Achenbach, et al.,1988.) For analyses involving teacher re-

ports in the present study, we focused onthe subset of 86 problem items which werepresent on both teacher checklists and onthe list of problems coded by observers.

ResultsWe analyzed the different observational

measures via 2 x 2 (culture x gender)ANOVAs. In addition to reporting F and pvalues, we interpreted all significant effectsusing Cohen's (1988) criteria for magnitude:ANOVA effects are considered small if theyaccount for l%-5.9% of the variance, me-dium if 5,9%-13.8%, and large if >13.8%,

On-Task/Off-Task Scores from theObservational Measure

We conducted a 2 x 2 (culture x gen-der) ANOVA of On/Off-Task scores; thesewere computed as the arcsine-transformedproportion of off-task ratings across the 1-min intervals in each observational period.The ANOVA yielded only one significant ef-fect, a main effect of culture, F(l, 140) =53.94, p < ,0001. This effect accounted for27.38% of the variance in On/Off-Taskscores; thus, it was a large effect, by Cohen's(1988) criteria. Means are shown in Table 1,Thai children were off-task in 10% of theobservations (SD = 0.11); American chil-dren were off-task in 23% of the observations(SD = 0.13), that is, more than twice as oftenas Thai children. Note that neither the gen-der nor the gender x culture effect was sig-nificant. For comparison purposes, note thatstudies using the Achenbach (1986) DOFhave reported percent off-task means rang-ing fi-om 11% (Reed & Edelbrock, 1983) to22% (McConaughy, Achenbach, & Cent,1988—22% for classroom observations only;off-task scores were lower for lunch and re-cess observations) for nonclinical samples ofAmerican children.

Within the U.S. sample, we found thatAfrican-American children were more oftenoff-task than Caucasian children (means:30% vs. 21%), F(l, 69) = 7.44, p < .01. Thus,we reran the ANOVA reported in the previ-ous paragraph, including only Caucasianchildren; the results resembled those de-scribed above, with a main effect of culture,F(l, 117) = 34.86, p < .0001, accounting for22.59% of the variance. Thai children wereoff-task in 10% of the observations, Ameri-can children in 21%.

Total Problem Scores from theObservational Measure

Total problem scores (i.e., sum of allproblem item ratings for each 10-min obser-

408 Child Development

TABLE 1

MEANS (and SDs) FOR OBSERVER-REPORT AND TEACHER-REPORT MEASURES

United States

Boys

Girls

Thai

Boys

Girls

% OfTTask

23(.13).26

(.14)21

(.12).10

(.11)11

(.11)09

(.10)

ObsTot

11.62(5.21)12.92(5.52)10.33(4.61)5.17

(2.21)5.54

(2.20)4.81

(2.19)

ObsOvr

.01(.02).01

(.01).01

(.02).01

(.02).01

(.01).01

(.02)

ObsUndr

.20(.10).22

(.11).17

(.09).08

(.05).08

(.05).07

(.04)

TRFTot

15.65(14.78)19.08

(17.29)12.77

(11.82)31.06

(21.82)34.17

(22.74)27.94

(20.70)

TRFOvr

.30(.36).26

(.36).33

(.37).69

(.49).70

(.46).68

(.53)

TRFUndr

.37(.43),51

(.51).26

(.31).58

(.50).68

(.57).48

(.39)

NOTE.—Total problem scores based on observer reports (ObsTot) and teacher reports (TRFTot) represent thesums of all ratings on individual problem items; because observers used a 0—3 scale and teachers a 0—2 scale, thetotal scores are not directly comparable without transformation. To permit comparison between observer-report meansfor OvercontroUed and UndercontroUed problems (ObsOvr and ObsUndr) and between teacher-report means forOvercontrolled and UndercontroUed problems (TRFOvr and TRFUndr), these means are all based not on sums ofratings but on the average rating per item.

vation) were subjected to a square root trans-formation because of skewness in the distri-bution. The 2 x 2 (culture x gender)ANOVA of these scores yielded a significantmain effect of culture, F(l, 140) = 111.80,p < .0001. The effect accounted for 43.24%of the variance in problem scores, thus easilyqualifying as a large effect in Cohen's (1988)system. The raw mean problem score was5.17 (SD = 2.21) for Thai youngsters and11.62 (SD = 5.21) for their American coun-terparts. Thus, as was true of the On/Off-Task scores, American children showedproblem scores that were more than twice assevere as those of Thai children.

For comparison purposes, we redid theANOVA, this time including only those 96observational items derived from Achen-bach's (1986) Direct Observation Form. Weagain found a significant main effect of cul-ture, F(l, 136) = 113.10, p < .0001, an effectaccounting for 42.65% of the variance. Totalproblem score means were 4.44 for Thaichildren and 10.05 for American children.Thus, the findings were quite similar tothose based on the full set of 111 problemitems. Means of 4.5 (Reed & Edelbrock,1983) and 3.5 (McConaughy et al., 1988)have been reported previously for nonclini-

cal American samples(1986) DOF.

with Achenbach's

UndercontroUed and OvercontrolledProblems on the Observational Measure

In an effort to shed some light on whichtype of problem was most involved in theobservational findings reported above, wefocused on the distinction between Over-controlled and UndercontroUed problems(see introduction, above). Because the ob-servations had taken place in school class-rooms, we grouped problems into Over- andUndercontroUed types according to whetherthey loaded on Internalizing versus Exter-nalizing factors, respectively, in principalcomponents analyses of the Teacher ReportForm of the Child Behavior Checklist (re-ported in Achenbach & Edelbrock, 1986) .̂^There were 16 TRF Overcontrolled itemsthat also appeared in our observational form(e.g., sad, worries, withdrawn); 27 TRF Un-dercontroUed items which also appeared inour observational form (e.g., argues, fidgets,swears).

We carried out a 2 x 2 x 2 (culture xgender x problem type) repeated-measuresANOVA of the two observational scores,with problem type (Over- vs. Undercon-

3 We used the 1986 TRF factors ratber tban the Achenbach (1991) integrative factors forparent, teacher, and youth self-reports, to (a) maintain a focus on factors specific to school-basedteacher ratings and (b) facilitate comparison witb our previous Tbai-U.S. teacher report compari-son (Weisz, Suwanlert, Ghaiyasit, Weiss, Acbenbacb, et al., 1988), wbicb relied on the 1986 TRFfactors.

Weisz et al. 409

trolled) as the within-groups factor; becausethe number of Over- and Undercontrolleditems differed, we used as dependent vari-able the mean score per item, averagingacross 16 and 27 items for Over- and Under-controlled, respectively; the range was thus0-3 for both scores. The analysis yielded avery highly signiflcant main effect of prob-lem type, F(l, 140) = 948,98, p < ,0001; Un-dercontrolled problems were observedmuch more often than Overcontrolled(means: 0,14 and 0,01 on the 0-3 scale). Thiseffect accounted for 76.61% of the varianceand was tlius large by Cohen's (1988) stan-dards. The only other significant effect wasa culture x problem type interaction, F(l,140) = 22,95, p < .0001; the effect wassmall, accounting for 1,85% of the variance.Breaking the interaction down, we foundthat the problem type effect was highly sig-niiBcant for both the Thai sample, F(l, 71) =319,39, p < ,0001, and the U,S. sample, F(l,71) = 658,54, p < ,0001, and, of course, thetwo effects ran in the same direction. View-ing the interaction from the other direction,we found that the effect of culture was non-significant for Overcontrolled problems(U,S, and Thai means were both 0,01 on the0-3 scale) but highly significant for Under-controlled problems (U,S, mean 0,20; Thaimean 0,08), F(l, 142) = 94,28, p < .0001(a large effect, accounting for 39,90% of thevariance). Thus, the large culture effect fortotal problem scores reflects, in part, the factthat American children showed substantiallyhigher rates of Undercontrolled problemsthan Thai children, but the two nationalgroups did not differ in their rates of Over-controlled problems. We did not recomputethis set of analyses for Achenbach itemsonly, because Over- and Undercontrolledscores were already based on Achenbachitems.

Source Effects: Observational Data versusTeacher Reports

Finally, we undertook an analysis in-tended to clarify the relation between thepresent observational findings and the find-ings obtained with teacher reports in an ear-lier study by Weisz, Suwanlert, Chaiyasit,Weiss, Achenbach, et al, (1988), In thatstudy. Thai youngsters were reported toshow much higher levels of problem behav-ior than v/ere their American age-mates,when Thai and American teachers fllled inThai and English teacher reports on theyoungsters (mean total problem scores: 30,9VS, 19,3; cf, U.S, means of 22,3 for 6-11-year-old boys and 15,1 for 6-11-year-old girls in

Achenbach & Edelbrock, 1986, app, C), Theobservational findings reported in the pres-ent study run in precisely the opposite direc-tion, J' is possible that the difference be-tween the findings of the two studiesresulted from differences in the groups ofchildren studied, in the schools selected forstudy, or in other factors unrelated to realdifferences in the behavior of Thai versusAmerican youngsters. To rule out these ex-traneous explanations, one needs to haveboth observational data and teacher reportsfrom the same samples of children. As notedabove, we arranged to obtain such data, ob-taining teacher reports via parallel Thai andAmerican instruments; we were able to ob-tain such teacher reports for all 72 of theThai children and for 57 (i,e,, 79%) of the 72American children.

This made it possible to conduct a 2 x2 x 2 (culture x gender x source [i,e,, ob-server report vs, teacher report]) ANOVA onTotal Problem Scores, summing across those86 problem items that were common to boththe teacher form and the observational form;the 86 items included Overcontrolled andUndercontrolled problems, as noted above,plus other problems (e,g,, clings to adults,clumsy, destroys own things, unliked) whichcould not be clearly classified into either cat-egory based on previous principal compo-nents analyses. To produce comparablescores for the TRF (0-2 scale) and the obser-vational measure (0—3 scale), we convertedthe observational measure to a 0-2 scale bymultiplying scores by % (we considered us-ing standard scores, but this would have setmeans for both measures at 0, thus eliminat-ing any source effects). Our analysis yieldeda main effect of source, F(l, 125) = 167,20,p < ,0001 (a large effect, accounting for22,75% of the variance); scores were higherfor teacher reports than for observer reports(means: 21,10 and 4,28), The main effectseems likely to have resulted partly from thefact that teachers had had much more expo-sure to the children than had the observers.

The most interesting and theoreticallysignificant finding was the culture x sourceinteraction, shown in Figure 1, F(l, 125) =48,29, p < ,0001 (a medium effect, account-ing for 6,57% of the variance). Teacher rat-ings showed a highly significant Thai-U,S,difference, with Thai children (27,59) show-ing much higher problem scores than Ameri-can children (mean 12,91), F(l, 127) =30,38, p < .0001 (a large effect, 19,30% ofthe variance), (For comparison purposes,prorating scores yielded Thai and U,S,

410 Child Development

30

Teacher Reports

Tha Idren

Observer Reports

American CriJIoren

FIG, 1,—Total problem scores of Thai and American children, as reported by the children's teach-ers and by trained observers.

means of 22.1 and 13.8 in Weisz, Suwanlert,Chaiyasit, Weiss, Achenbach, et al.'s [1988]teacher report study, and U.S. TRF meansof 16.0 for boys and 10.8 for girls in Achen-bach & Edelbrock's [1986] TRF manual[app, C].) Observer ratings also showed ahighly significant Thai-U.S, difference, butin precisely the opposite direction, withThai children (mean 2.73) showing muchlower problem scores than American chil-dren (mean 6.25), F(l, 127) = 92.94, p <,0001 (also a large effect, 42,26% of the vari-ance). To summarize this striking effect,when teachers did the rating. Thai childrenscored twice as high as American childrenin problem behavior; but when trained ob-servers rated the same children, in the sameclasses, on the same problem items, theAmerican youngsters scored twice as high astheir Thai counterparts.

We recomputed the Observer versusTeacher analyses using only the observa-tional items that were based on Achenbach'sDirect Observation Form; observation datawere transformed to a 0-1-2 scale, to matchthe TRF scale. Again, the 2 x 2 x 2 (cultureX gender x source) ANOVA yielded a sig-nificant main effect of source, F(l, 125) =147,51, p < ,0001; it aceounted for 32.04%of the variance. Mean problem scores were24.25 for teachers, 4.29 for observers. Theculture X source interaction was also sig-nificant once again, F(l, 125) = 32,70, p <.0001, accounting for 7,10% of the variance.

The pattern of means resembled that shownin Figure 1, Once again, observer ratingsyielded problem scores twice as high forU,S. children as for Thai youngsters (means6,25 vs, 2.73), whereas teacher ratingsyielded problem scores twice as high for theThai sample as for the American sample(means 31,05 vs. 15.64). The culture effectwas significant for both observer and teacherratings (both ps < .0001). And the source ef-fect was significant for both the Thai andAmerican samples (both ps < ,0001),

Teacher x Observer CorrelationsAs might be expected from the preced-

ing findings, teachers and observers did notshow very high levels of agreement in theirratings of the children. To assess teacher-observer correlations we again focused ononly those items present in the rating formsfor both informants. For Undercontrolledproblems, the teacher-observer correlationwas 0.33 in the United States (df = 56), 0.19in Thailand (df = 71; z for difference = 0.99,p = N,S,). For Overcontrolled problems, theteacher-observer correlation was 0.13 in theUnited States, 0.17 in Thailand (z = 0,82, p= N.S.), For total problems, the teacher-observer correlation was 0.32 in the UnitedStates, 0,15 in Thailand (z = 0.99, p = N.S,),Correlations between TRF and AchenbachDOF total problem scores (but not Over- andUndercontrolled), reported elsewhere formixed clinical-plus-nonclinical samples of

Weisz et al. 411

boys (Achenbach & Edelbrock, 1986; Reed& Edelbrock, 1983), have ranged from 0,44to 0,51 (possibly enhanced by the increasedbehavioral variability of the mixed clinicaland nonclinical samples),

Teacher-Report Findings for Over- andUndercontrolled Problems

Given the emphasis in previous Thai-U,S, research on Overcontrolled and Under-controlled problems, we tested for national-it)' and gender differences in teacher reportsfor the two problem types. We first carriedout a 2 X 2 X 2 (problem type x cultureX gender) repeated-measures ANOVA, withproblem type as the repeated-measure fac-tor. The analyses yielded two significant in-teractions: Problem type x culture, F(l,125) = 3,90, p = ,05, and problem type xgender, F(l, 125) = 7,39, p < ,01, The prob-lem type X culture interaction reflected thefact that, ailthough teachers rated Thai chil-dren signiflcantly higher than Americanchildren on both Over- and Undercontrolledproblems, the Thai-U,S, difference wasgreater for Overcontrolled problems (meanrating per item: 0,69 vs, 0,30; p < ,0001) thanfor Undercontrolled problems (means 0,58vs, 0,37; p < ,01), Scores for Over- and Un-dercontrolled problems did not differ sig-nificantly within either national sample. Theproblem t:/pe x gender interaction largelyreflected tlie fact that boys were rated higherthan girls on Undercontrolled problems(means 0,61 vs, 0,38; p < ,01) whereas boysand girls had virtually identical Overcon-trolled scores (means 0,51 vs, 0,52; p =N,S,), The Over- versus Undercontrolled dif-ference was significant among girls (p < ,05)but not among boys.

Checking the Impact of SES on FindingsGiven the mean SES difference be-

tween the Thai and American samples (seeabove), we tested whether our findingsmight have been influenced by SES, We re-computed all the ANOVAs reported above,but with SES covaried. Controlling for SESdid not change any of the flndings in a sub-stantial way; all effects that were previouslysignificant remained significant and withinthe same p-level category (e,g,, at p < ,0001),and all previously nonsignificant effects re-mained nonsignificant, with SES controlled.

Most Common Observer-reported andTeacher-reported Problems in Thai andAmerican Samples

Results reported thus far have comefrom purely quantitative analyses. To com-plement these findings with information of

a somewhat more qualitative sort, we pre-sent in Table 2 the 12 most frequently notedproblems in the Thai and U,S, samples,based on our observational data and teacherreports. The table reveals numerous similar-ities in the types of problems reported byobservers for Thai and American children,with nine of the top 12 observer-reportedproblems among Thai children also in thetop 12 for American children. The overallpattern suggests that observers in the UnitedStates were seeing problems similar to thoseseen in Thailand, but that U,S, observerswere seeing more frequent displays of thoseproblems. By contrast, teacher reports in thetwo countries seemed to reflect a greater dis-parity in the types of problems reported; thisis reflected in the fact that only four of thetop 12 teacher-reported problems amongThai children were in the top 12 for theAmerican sample.

The impression given by Table 2 wasreinforced by correlational analyses involv-ing the 86 problem items that were commonto both the teacher and observer forms. Themean correlation between Thai observa-tional scores and U,S, observational scores,across the 86 items, was 0,87 {df = 110,p < ,0001), whereas the mean correlation forteacher ratings was 0,57 {df = 118, p <,0001); the difference between the two coef-ficients was significant, z = 5,13, p < ,00001,

Discussion

As anticipated, the observational find-ings revealed a pattern of Thai-U,S, differ-ences quite different from the pattern shownin teacher reports, thus lending support tothe possibility that teacher reports are col-ored by culture-linked values and expecta-tions about child behavior. Teachers in pre-vious research (Weisz, Suwanlert, Chaiyasit,Weiss, Achenbach, et al,, 1988) and in thepresent study, reported levels of problembehavior that were twice as high for Thaichildren as for their American age-mates. Bycontrast, direct observations in the presentstudy showed the reverse pattern: Twice asmuch problem behavior among Americanchildren as among Thai children, a differ-ence so large that it accounted for 43% of thevariance in problem scores. Moreover, directobservations also showed that Americanyoungsters were off-task 23% of the time,compared to 10% for Thai children.

Although the present observationalfindings were derived from U,S, and Thaichildren, it is tempting to compare them to

412 Child Development

TABLE 2

TWELVE MOST COMMON PROBLEMS AMONG THAI AND AMERICAN CHILDREN AS REPORTED BY TRAINEDOBSERVERS AND BY TEACHERS

REPORTS BY TRAINED OBSERVERS

American Thai

1. Easily distracted (6) 73.82. Fidgets (10) 68.2

' llg

Doesn't sit still (8) 65.5Doesn't concentrate/pay attention(5) 46.7Talks when shouldn't (74) 36.7Sucks thumb, hand (25) 25.1Yawns (75) 23.5Picks nose, skin (42) 22.9Talks too much (94) 16.6

3.4.

5.6.7.8.9.

10. Shows off/clowns (50) 13.3

11. Sings/talks to self (73) 12.212. Teases (14) 11.6

1. Fidgets (10) 57.62. Easily distracted (6) 45.23. Doesn't sit still (8) 25.84. Picks nose, skin (42) 24.67

5. Eats/drinks nonfood (24) 22.56. Yawns (75) 19.07. Sucks thumb, hand (25) 16.88. Talks when shouldn't (74) 11.49. Apathetic, unmotivated (44) 6.3

10. Doesn't concentrate/pay attention(5) 6.3

11. Argues (2) 6.312. Shows off/clowns (50) 5.2

REPORTS BY TEACHERS

American Thai

1. Talks too much (93) 50.92. Talks out of turn (53) 45.63. Hurt when criticized (81) 42.1

4. Disturbs peers (24) 40.45. Easily distracted (78) 38.66. Messy work (72) 38.67. Demands attention (19) 36.8

8. Argues a lot (3) 35.19. Fails to finish things she or he starts

(4) 32.110. Self-conscious, easily embarrassed

(71) 31.611. Misdeeds with no guilt (26) 31.612. Disrupts class (67) 29.8

1. Can't concentrate/pay attention (8) .. 70.82. Messy work (72) 69.43. Fails to cany out assigned tasks

(100) 66.74. Dependent on adults (11) 66.75. Sulks a lot (88) 65.36. Easily distracted (78) 63.97. Fails to finish things she or he starts

(4) 63.98. Acts without thinking (41) 61.19. Hurt when criticized (81) 59.7

10. Apathetic, unmotivated (60) 59.7

11. Nervous, high-strung (45) 59.712. Can't sit still (10) 58.3

NOTE.—Numbers in parentheses are the item numbers from the observation form (top half of table) and TeacherReport Form (bottom half). Numbers in percent column for Trained Observers indicate the percent of observationsfor which nonzero ratings were recorded for the particular item; numbers in percent column for Reports by Teachersindicate the percent of children for whom nonzero ratings were recorded for the particular item.

findings reported by Stevenson and Stigler(1992) in their comparisons of youngsters inthe United States and two other Asian na-tions, Japan and Taiwan. Compared to chil-dren in the two Asian countries, Stevensonand Stigler (1992) found that "Americanchildren spent the least amount of time actu-ally engaged in academic activities in aca-demic classes" (p. 147). Moreover, "Ameri-can children spent more time out of theirseats, talked more to their peers at inappro-priate times, and engaged in other inappro-priate activities to a greater degree than didChinese and Japanese children. For exam-

ple, American fifth-graders were out of theirseats nearly 20 percent of the time; Chineseand Japanese fifth-graders, less than 5 per-cent" (p. 148). The findings summarized byStevenson and Stigler (1992) suggest that atleast some of our results may resemble pat-terns seen in other Asian versus Americancomparisons.

Our observational data do not providedirect evidence on reasons for the Thai-American differences, but several possiblereasons warrant attention. First, Thai in-structional procedures and classroom man-

Weisz et al. 413

agement practices appear to be somewhatmore strict, authoritative, and controllingthan the American norm. Thai practices, likethe Japanese and Taiwanese procedures de-scribed by Stevenson and Stigler (1992),may foster higher levels of classroom disci-pline than American schools, thus stimulat-ing more on-task activity and better behav-ioral control among Thai than Americanchildren. A second possibility is that broadcultural differences between Thailand andthe United States may be reflected in thefindings discussed here. For example, asnoted above (and see Gardiner, 1968; Gardi-ner & Suttipan, 1977; Moore, 1974; Suvan-nathat, 1979; Weisz, 1989), Thai children arereared to be deferent, compliant, and re-spectful of others, particularly those who areolder and those in positions of authority; andteachers are treated with special respectthroughout the society, addressed with hon-orific terminology, and supported by parentswhen difficulties arise with their children(cf. similar phenomena described by Steven-son & Stigler, 1992). A third possibility isthat Thai children differ from American chil-dren in temperament (see Weisz, Suwanlert,et al., 1993), manifesting a more behaviorallyand emotionally inhibited style that makesthem more attentive and less likely to showbehavior problems. Data from other Asiangroups at least raise this possibility (see e.g.,Freedman, 1979). Thus, the Thai-U.S. difFer-ences observed in the study may reflect suchdiverse influences as instructional styles andclassroom management practices, culturaltraditions (including respect for teachers)and child temperament. These causal possi-bilities warrant attention in future research.

Such causal speculations must be tem-pered by tlie fact that OvercontroUed prob-lems played essentially no role in our obser-vational findings, presumably because suchproblems are not very evident to external ob-servers during brief observation periods.Previous Thai-U.S. comparative findings us-ing parent reports (Weisz, Suwanlert, Chai-yasit, & Walter, 1987; Weisz, Suwanlert,Chaiyasit, Weiss, et al., 1987; Weisz, Su-wanlert, et al., 1993) indicate that Thaiyoungsters may be especially susceptibleto OvercontroUed problems. It is possible,then, that a measure more sensitive to chil-dren's internal processes might have shownhigher levels of at least this particular formof problem behavior in Thai than Americanschool children.

The most valuable general lesson of thisstudy may be that much can be learned

by directly observing children's behavior,rather than relying solely on the reports ofinformed but untrained informants, such asteachers. Our observational findings ran di-rectly counter to the findings of Weisz, Su-wanlert, Chaiyasit, Weiss, Achenbach, et al.(1988) based on Thai and American teachers'reports of their elementary school pupils'behavior. In the 1988 study. Thai teach-ers reported significantly higher levels ofOvercontrolled, Undercontrolled, and TotalProblems among Thai children than Ameri-can teachers reported for American children.In principle, it is possible that the differencebetween the earlier teacher report findingsof Weisz, Suwanlert, Chaiyasit, Weiss,Achenbach, et al. (1988) and the presentfindings resulted from between-study differ-ences in the schools or samples involved, orthe fact that the teacher reports involved asomewhat different collection of problemitems than did the observer reports. How-ever, one analysis reported above castsdoubt on these artifactual possibilities. Wecompared observers' reports and teachers'reports for the same children (i.e., childrenfrom the present study), from the sameschools, and using the same collection ofitems (i.e., only the 86 items that were in-cluded in both the teacher form and the ob-server form). The results showed quiteclearly that teacher reports for the presentsample generated higher problem ratings forThai than American children, whereas thereports of trained observers generated pre-cisely the opposite picture for the same chil-dren. The results suggest that Thai teachersmay evaluate their students' behavior usingmore exacting standards than Americanteachers do; perhaps Thai teachers are so ac-customed to high levels of respect, defer-ence, obedience, and behavioral control bytheir pupils that they rate as serious behav-ior problems what would be seen as normalor minor by American teachers. Alterna-tively, perhaps Thai teachers are more sensi-tive than American teachers to signs of be-havioral or emotional problems in children.Finally, Thai teachers may have been moreresponsive than American teachers to de-mand characteristics of the assessment pro-cedure—for example, they may have beenespecially conscientious about reading eachchecklist item carefully and trying to re-member all of each child's relevant past be-haviors.

Whose reports should we rely on—observers or teachers? As some of the pre-ceding discussion implies, we place some-

414 Child Development

what greater confidence in the observers'reports than in the teachers' ratings. Observ-ers were trained to use common standardsfor reporting on child behavior, even acrossthe two different cultures; and we have datasupporting the observers' interrater reliabil-ity. On the other hand, it is certainly possi-ble that teachers' day-to-day interactionswith their pupils give them a more compre-hensive picture of, and deeper insight into,the children's behavior than that whichcould be achieved by outside observers dur-ing a series of 10-min sessions. Our observ-ers focused on classroom behavior, whereasteachers are also privy to children's lunchtime and recess behavior, and at least onestudy of American children (McConaughy etal., 1988) points to lower levels of off-taskand problem behavior at lunch and recessthan in class; our exclusive focus on class-room behavior ruled out detection of possi-ble culture X situation interactions withinthe school setting.

To make the observer versus teacher de-bate even more complex, we must stress thatthere is no true "right answer" to the ques-tion of whether a particular child "has" aparticular problem. As we have discussedpreviously (Weisz, 1989; Weisz, Suwanlert,Chaiyasit, Weiss, Walter, et al., 1988), theconcept of a child problem, like the conceptof child psychopathology generally, involvestwo components that are very difficult to dis-entangle: Actual child behavior and the per-ception of that behavior through the lens ofsome observer—a teacher, a trained ob-server, a parent, or the child him- or herself."Actual behavior" and "perceived behav-ior" may be inextricably linked in reports byany informant; in a sense, it is precisely theblend of actual child behavior and culturallyconditioned perceptions that forms the es-sence of "child problems" in the first place.

Such reasoning, combined with thepresent findings, underscores the need toview the child's behavior through more thanone window. This need is especially pro-nounced in cross-national research on childbehavior and child problems. The perspec-tives of parent, teachers, trained observers,and the children themselves may all be im-portant sources of information. As suggestedabove, one might argue that trained observ-ers can provide the most unbiased and reli-able information on child behavior, becausesuch observers can be trained and held to

standards of reliability that may not apply toparents, teachers, or children. Yet, observerreports are almost certain to be based on aless complete sample of child behavior thanreports of the other three types of infor-mants. This is especially important when thetarget behavior of interest involves psycho-logical problems, many of which are quitelow in base rate. For example, potentiallysignificant problem behavior such as vandal-ism, suicidal talk or attempts, fire-setting,and stealing may never occur in the pres-ence of observers; information about suchproblems may be accessible only to infor-mants who share significant portions of thechild's life space—informants such as par-ents, teachers, or the children themselves.

Thus, for many reasons, some empiricaland some conceptual, no single source of in-formation on child behavior is likely to pro-vide a "magic bullet" or final "right an-swer." Instead, our closest approach to the"truth" about child behavior and child prob-lems is apt to involve a triangulation of re-ports from multiple sources, with each offer-ing a different base of information and adifferent perspective on the child.* What thepresent study suggests is that the reports oftrained observers will be an important partof that mix of information and perspectives.

ReferencesAchenbach, T. M. (1986). The Child Behavior

Checklist—Direct Observation Form. Burl-ington, VT: University of Vermont Depart-ment of Psychiatry.

Achenbach, T. M. (1991). Integrative guide forthe 1991 CBCL14-18, YSR, and TRF profiles.Burlington, VT: University of Vermont De-partment of Psychiatry.

Achenbach, T. M., & Edelbrock, C. S. (1983).Manual for the Child Behavior Checklist andRevised Child Behavior Profile. Burlington,VT: University of Vermont Department ofPsychiatry.

Achenbach, T. M., & Edelbrock, C. S. (1986).Manual for the Teacher's Report Form andteacher version of the child behavior profile.Burlington, VT: University of Vermont De-partment of Psychiatry.

Achenbach, T. M., Hensley, V. R., Phares, V., &Gray son, D. (1990). Problems and competen-cies reported by parents of Australian andAmerican children. yot*mai of Child Psychol-ogy and Psychiatry, 31, 265-286.

Achenbach, T. M., Verhulst, F. C , Baron, G. D.,

"* For a description of recent technology for integration of data on child problem behaviorfrom multiple sources, see Achenbach (1991).

Weisz et al. 415

& Akkerhuis, G. W. (1987). Epidemiologicalcomparisons of American and Dutch chil-dren; 1. Behavioral/emotional problems andcompetencies reported by parents for ages 4to 16. Journal of the American Academy ofChild and Adolescent Psychiatry, 26,317-325.

Cohen, J. (1988). Statistical power analysis forthe behavioral sciences (Rev. ed.). New York:Academic Press.

Freedman, D. G. (1979). Ethnic differences in ba-bies. Human Nature, 2, 36—43.

Gardiner, H. W. (1968). Expression of angeramong Thais: Some preliminary findings.Psychologia, 11, 221-228.

Gardiner, H. W., & Suttipan, G. S. (1977). Parentaltolerance of aggression: Perceptions of pie-adolescents in Thailand. Psychologia, 20,28-32.

Hollingshead, A. B. (1975). Four-factor index ofsocial status. Unpublished manuscript, YaleUniversity.

Kennedy, S.,, Scheirer, J., & Rogers, A. (1984). Theprice of success: Our monocultural science.American Psychologist, 39, 996-997.

Lambert, M. G., Weisz, J. R., & Knight, F. (1989).Over- and UndercontroUed clinic referralproblems of Jamaican and American childrenand adolescents: The culture-general and theculture-specific. Journal of Consulting andClinical Psychology, 57, 467-472.

McGonaughy, S. H., Achenbach, T. M., & Gent,G. L. (1988). Multiaxial empirically based as-sessment: Parent, teacher, observational, cog-nitive, and personality correlates of child be-havior profile types for 6- to 11-year-old boys.Journal of Abnormal Child Psychology, 16,485-509.

Montenegro, H. (1983). Salud mental del escolar:Estandarizacion del inventario de problemasconductuales y destrezas sociales de T.Achenbach en ninos de 6 a 11 anos. Santiago,Ghile: Centro de Estudios de Desarrollo y Es-timulacion Psisocial, Roman Diaz 26, Of 63.

Moore, F. J. (1974). Thailand: Its people, its soci-ety, its culture. New Haven, GT: HRAFPress.

Phillips, H. P. (1965). Thai peasant personality:The patterning of interpersonal behavior inthe village of Bang Chan. Berkeley: Univer-sity of California Press.

Reed, M. L., & Edelbrock, G. (1983). Reliabilityand validity of the Direct Observation Formof the Ghild Behavior Ghecklist. Journal ofAbnormal Child Psychology, 11, 521-530.

Sangsingkeo. P. (1969). Buddhism and some ef-fects on the rearing of children in Thailand.In W. Gaudill & T. Y. Lin (Eds.), Mental

health research in Asia and the Pacific (pp.286-295). Honolulu: East-West Genter Press.

Stevenson, H. W., & Stigler, J. W. (1992). Thelearning gap: Why our schools are failing andwhat we can learn from Japanese and Chi-nese education. New York: Summit Books.

Suvannathat, G. (1979). The inculcation of valuesin Thai children. International Social ScienceJournal, 31, 477-485.

Suwanlert, S. (1974). Some personality character-istics of Thai students. In W. P. Lebra (Ed.),Youth, socialization, and mental health inAsia and the Pacific (pp. 75-84). Honolulu:University Press of Hawaii.

Weisz, J. R. (1989). Gulture and the developmentof child psychopathology: Lessons from Thai-land. In D. Gicchetti (Ed.), The emergence ofa discipline: Rochester symposium on devel-opmental psychopathology (Vol. 1, pp. 89-117). Hillsdale, NJ: Erlbaum.

Weisz, J. R., Sigman, M., Weiss, B., & Mosk, J.(1993). Parent reports of behavioral and emo-tional problems among children in Kenya,Thailand, and the United States. Child Devel-opment, 64, 98-109.

Weisz, J. R., Suwanlert, S., Ghaiyasit, W., & Wal-ter, B. R. (1987). Over- and UndercontroUedreferral problems among Thai and Americanchildren and adolescents: The wat and waiof cultural differences, yournai of Consultingand Clinical Psychology, 55, 719-726.

Weisz, J. R., Suwanlert, S., Ghaiyasit, W., Weiss,B., Achenbach, T. M., & Eastman, K. L.(1993). Behavioral and emotional problemsamong Thai and American adolescents: Par-ent reports for ages 12-16.7ourna/ of Abnor-mal Psychology, 102, 395-403.

Weisz, J. R., Suwanlert, S., Ghaiyasit, W., Weiss,B., Achenbach, T. M., & Trevathan, D. (1988).Epidemiology of behavioral and emotionalproblems among Thai and American chil-dren: Teacher reports for ages 6-11. Journalof Child Psychology and Psychiatry, 30,471-484.

Weisz, J. R., Suwanlert, S., Ghaiyasit, W., Weiss,B., Achenbach, T. M., & Walter, B. A. (1987).Epidemiology of behavioral and emotionalproblems among Thai and American chil-dren: Parent reports for ages 6-11. Journal ofthe American Academy of Child and Adoles-cent Psychiatry, 26, 890-898.

Weisz, J. R., Suwanlert, S., Ghaiyasit, W., Weiss,B., Walter, B., & Anderson, W. W. (1988). Thaiand American perspectives on Over- and Un-dercontroUed child behavior problems: Ex-ploring the threshold model among parents,teachers, and psychologists. Journal of Con-sulting and Clinical Psychology, 56,601-609.