Journal of Emotional and Behavioral Disorders...(1979) found that females with ADHD were usually referred for learning prob-lems rather than behavior problems, whereas boys identified

http://ebx.sagepub.com

Journal of Emotional and Behavioral Disorders

DOI: 10.1177/106342660000800105 2000; 8; 38 Journal of Emotional and Behavioral Disorders

Diana Rogers-Adkinson and Mary-Beth Noll Robert Reid, Cynthia A. Riccio, Robert H. Kessler, George J. Dupaul, Thomas J. Power, Arthur D. Anastopoulos,

Gender and Ethnic Differences in ADHD as Assessed by Behavior Ratings

http://ebx.sagepub.com/cgi/content/abstract/8/1/38 The online version of this article can be found at:

Published by: Hammill Institute on Disabilities

and

http://www.sagepublications.com

can be found at:Journal of Emotional and Behavioral Disorders Additional services and information for

http://ebx.sagepub.com/cgi/alerts Email Alerts:

http://ebx.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

at Landspitali University Hospital on February 17, 2009 http://ebx.sagepub.comDownloaded from

http://www.hammill-institute.org

http://ebx.sagepub.com/cgi/alerts

http://ebx.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav


38

Gender and Ethnic Differences in ADHD asAssessed by Behavior Ratings

ROBERT REID, CYNTHIA A. RICCIO, ROBERT H. KESSLER,GEORGE J. DUPAUL, THOMAS J. POWER, ARTHUR D. ANASTOPOULOS,

DIANA ROGERS-ADKINSON, AND MARY-BETH NOLL

Attention-deficit/hyperactivity disorder (ADHD) is a common childhood disorder. Research suggests thatADHD is 4 to 9 times more frequent in males than females, and the possibility of underidentification infemales and overidentification in males has been suggested as an explanation for these statistics.As partof the diagnostic process, teachers are frequently asked to complete behavior rating scales. There is a

lack of empirical data concerning the extent to which gender differences are evident on such rating scales.This study investigated the use of the ADHD-IV Rating Scale-School Version, with male and female stu-dents from ages 5 to 18 years. Results suggest that the ADHD construct is consistent across gender;however, there are differences across gender and ethnicity. For Caucasian children, externalizing behav-iors are most salient in terms of discriminating between males and females. Implications for research and

practice are discussed.

TTENTION-DEFICIT/HYPERACTIV-ITY disorder (ADHD) is a com-

~ mon childhood disorder that is

frequently brought to the attention of

physicians and psychologists to conductevaluations due to behavioral concernsof both teachers and parents (Breen &

Altepeter, 1990). According to the Diag-nostic and Statistical Manual of MentalDisorders, 4th edition (DSM-IV,- Ameri-can Psychiatric Association, 1994), ADHDis characterized by a pattern of inatten-tion and/or hyperactivity-impulsivitythat is exhibited to an extreme level, suchthat it is developmentally inappropriaterelative to a person’s age. It occurs in anestimated 3% to 5% of school-age children.

The most evident difference betweenmales and females with ADHD is the

higher rate at which males are diag-nosed. Male-to-female ratios range from4:1 to 9:1, depending upon whether

community-based or clinical samples areused (APA, 1994). In fact, the behaviorsused to define the symptomatology ofADHD in the DSM-IV were identifiedfrom a sample pool composed predomi-nately of males (Frick et al., 1994; Laheyet al., 1994). Several hypotheses havebeen offered to explain the dispropor-tionate frequency of males with ADHD.These hypothesis tend to rely on biolog-

ical or child-rearing differences by gen-der (see, e.g., Barkley, 1989; Befera &

Barkley, 1984; Brown, Madan-Swain, &

Baldwin, 1991; Eme, 1992; Ernst et al.,1994; Gualtieri & Hicks, 1985; James &

Taylor, 1990; Kashani, Chapel, Ellis, &

Shekim, 1979), but the issue has not beenresolved.

Extensive research has been conductedon boys with ADHD, but comparablestudies of girls with the disorder are in-frequent, possibly due to the challengesof recruiting sufficient samples of girlswith the disorder (Arnold, 1996). Stud-ies on ADHD specific to gender dispar-ity have often been epidemiological innature (e.g., Barkley, 1989; Berry, Shay-witz, & Shaywitz, 1989; Faraone, Bie-derman, Keenan, & Tsuang, 1991; James& Taylor, 1990; McGee, Williams, &

Silva, 1987). Studies have found moresimilarities than differences in girls andboys with ADHD, and a number of stud-ies have found no gender differences inthe number, or severity, of ADHD symp-toms as a function of gender (Berry et al.,1985; Brown et al., 1991; Horn, Wagner,& Ialongo, 1989; James & Taylor, 1990;Kashani et al., 1979; Silverthom, Frick,Kuper, & Ott, 1996). However, some gen-der differences related to ADHD have

been identified. Seidman and colleagues(1997) found that, in contrast to researchfindings of impaired executive functionin boys with ADHD, girls with ADHDdid not demonstrate significant executivefunction deficits. Other research has

found that girls with ADHD show moresevere visual-spatial, cognitive, and lan-guage deficits than boys with ADHD



39

(Berry et al., 1985; Breen, 1989; Brownet al., 1991; Gordon & Mettelman, 1994;James & Taylor, 1990; Kashani et al.,1979; Taylor, 1986).

The available research on gender dif-ferences has revealed some differenceswith regard to the referral process forboys and girls. Kashani and colleagues(1979) found that females with ADHDwere usually referred for learning prob-lems rather than behavior problems,whereas boys identified as hyperactivewere more frequently referred for behav-ior problems than learning problems.This would be consistent with the re-

search findings that girls with ADHDtend to have more comorbid internal

manifestations of the disorder, whereasboys with ADHD have been noted to ex-press more aggressive overt types of be-havior (Breen & Altepeter, 1990; Brown,Abramowitz, Madan-Swain, Eckstrand,& Dulcan, 1989; deHaas, 1986; Gordon& Mettelman, 1994). Thus, gender-correlated behavioral patterns may bemore frequently identified as ADHD inboys than in girls due to the frequency ofdisruptive classroom behavior exhibitedby boys (Breen & Altepeter, 1990).

One important caution to keep in mindis that gender differences may vary as afunction of the sample used. For exam-ple, Gaub and Carlson (1997) found atrend for greater severity of inattentionamong females, but comparable levels ofhyperactivity in females and males, in aclinically referred sample. However, thispattern was not evident among nonre-ferred children.

It has been suggested that one causefor the gender disparity in referrals forADHD may rest in the scales used as partof the diagnostic process, and the use ofgeneral norms as opposed to gender-specific norms (Barkley, 1996). Genderdifferences have been noted on ratingscales used to assess children suspectedof having ADHD. Based on informationprovided in the manual for the BehaviorAssessment System for Children (BASC;Reynolds & Kamphaus, 1992), for thestandardization sample, males were

rated one third to one half of a standard

deviation higher than females on hyper-activity, attention problems, and othersubscales. For this reason, the BASC

manual includes separate norm tables formales and females, as well as a com-bined norm table, and recommends thatthe same-sex norms be used for clinical

diagnosis in order to identify those chil-dren whose ratings are significant forboth their age and gender. Similarly,DuPaul and colleagues (1997) reportedsignificant differences across gender oncomposite inattention and hyperactivityscores based on normative data from theADHD-IV Rating Scale-School Version(ARS; DuPaul, Power, Anastopoulous,& Reid, 1998). DuPaul et al. (1998) pro-vided separate norms for males andfemales.

In contrast, Silverthorn and col-

leagues (1996) concluded that separatenorms by gender were not warrantedbased on their finding that girls and boyswith ADHD did not differ on measuresof severity and that diagnostic cut-scoresidentified boys and girls with equivalentlevels of impairment. They further ar-gued that to use separate norms mightartificially reduce the difference in

prevalence rates for ADHD for girls andboys. Gordon (1996) argued that boysare more often identified with ADHD be-

cause they are more likely to demon-strate the severity of symptomatology towarrant this diagnosis; therefore, genderreferencing is not appropriate. Alterna-tively, the possibility has been suggestedthat ADHD manifests differently in girlsas compared to boys (Barkley, 1996;Gordon & Mettelman, 1994).

When considering the possibility ofdifferences in ADHD symptomatologyacross gender, ethnicity is a factor that

should also be considered. Serious con-cerns have been expressed regarding theassessment of ADHD with children from

ethnic minorities (Bauermeister, Berrios,Jimenez, Acevedos, & Gordon, 1990)and the use of behavior rating scales withthese children (Reid, 1995). However,ADHD among ethnic minorities remains

an understudied area. Studies that have

used behavior rating scales have reportedsignificant differences between Cau-

casian (see Note) and African Americanstudents (e.g., DuPaul et al., 1997; Ep-stein, March, Conners, & Jackson, 1998;Jarvinen & Sprague, 1995; Lambert,Sandoval, & Sassone, 1978; Reid et al.,

1998; Waechter, Anderson, Juarez, Langs-dorf, & Madrigal, 1979). However, it is

uncertain whether these differences were

due to real differences in behavior

among groups, rater bias due to ethnicityand/or socioeconomic factors, or a com-bination of the two. Only one study todate has addressed the issue of rater bias.

This study found evidence that ethnicstatus appeared to bias behavior ratings(Sonuga-Barke, Minocha, Tayor, & Sand-

berg, 1993). Because ethnic minoritiesconstitute a large portion of the popula-tion, and because research has demon-strated convincingly that there are in-terethnic differences on behavior ratingscales, we should attend to ethnicity as afactor in gender differences.

The purpose of this study was to ex-plore the extent and nature of gender dif-ferences in ADHD across different eth-

nic groups (i.e., Caucasian and AfricanAmerican). Although clinical diagnosisrequires multimethod, multisource in-

formation, we focused on teacher-

completed rating scale information, be-cause teacher ratings are often viewed asindicative of functioning in the schoolenvironment and are, thus, an importantcomponent in the assessment process(e.g. Barkley, 1989). Our exploration in-volved four interrelated analyses:

1. We analyzed descriptive statistics.2. We performed a multivariate analy-

sis to assess whether or not different

patterns of item means existedacross gender for two ethnic groups.

3. We used structural equation model-ing to investigate whether the sametwo-factor construct of ADHD was

appropriate for both males andfemales.

4. We conducted a discriminant func-tion analysis to determine the extentto which items representing ADHDsymptoms best separated childrenby gender.

Because of the possibility that genderdifferences may vary depending onsymptom severity, we conducted analy-ses on two samples: one representative ofunselected children and a second repre-senting children with severe or numeroussymptoms of ADHD.



40

. METHOD

Instrumentation

Teachers completed the ARS. The ARSwas selected because it reflects current

DSM-IV diagnostic criteria for ADHD.This scale consists of 18 items-9 that

address hyperactivity-impulsivity (HI)symptoms and 9 that address inattention

(IA) symptoms-directly adapted fromthe diagnostic criteria for ADHD spe-cified in the DSM-IV. The ARS has

demonstrated adequate reliability and

criterion-related validity (DuPaul, Power,McGooey, Ikeda, & Anastopoulos, 1998).

In order to minimize possible bias dueto response set, IA symptoms were des-

ignated as odd-numbered items, and HIsymptoms were designated as even-

numbered items. Teachers were in-

structed to select the single response foreach item that best described the fre-

quency of the specific behavior dis-

played by the target child over the past6 months or since the beginning of theschool year. The frequency for each itemwas delineated on a four-point Likertscale ranging from 0 (never or rarely) to3 (very often), with the higher scores in-dicative of greater ADHD-related behav-ior. Teachers were asked to rate the be-

havior of two students from their classroster (e.g., third male and fifth femaleon the class roster). Ratings were com-pleted between October and May in the1994-1995 or 1995-1996 school years.Estimated return rates ranged from 50%to 95 % (M = 85 %) across school districts.

Participants ’

Participants for this study included 3,322children and youth ages 5 to 18 (referredto as the Total group) taken from the nor-mative sample of the ARS (DuPaul et al.,1997). The Caucasian (CA) sample con-sisted of 1,338 males and 1,298 females.The African American (AA) sample con-sisted of 376 males and 310 females. Be-

cause the total behavior rating scores

decrease as age levels increase, the pos-sibility of an age by gender dependencywas tested for both CA and AA ethnic

groups. The results suggested that ages

were proportional across gender for bothgroups, AA group, x2( 12, N = 686) =7.14, p = .84; CA group, x2( 12, N =2636) = 12.39, p = .71.A subsample of participants who might

have been considered at risk for ADHD

(referred to as the At-Risk group)-whoscored above the 90th percentile on HI,IA, or both factors-was created. The90th percentile was selected because it isthe recommended cut-score (DuPaul,Power, Anastopoulos, & Reid, 1998).This resulted in an At-Risk group con-

sisting of 218 CA males, 203 CA fe-males, 64 AA males, and 53 AA females.

Participating teachers were largelywomen (82.0%) and Caucasian (93.4%).Other ethnic groups represented amongteachers included African American

(5.4%), Hispanic (0.7%), Native Ameri-can (0.1 %), Asian American (0.1 %), andother (0.4%). The majority of teachers(91.1 %) were general educators; the re-mainder (8.9%) were special educators.No dependency was found between spe-cial education or general education sta-tus and gender for either ethnic group,AA group, x2( 1, N = 686) = .43, p = .51;CA group, x2( 1, N = 2636) = .04, p = .85.

AnalysisDescriptive Data. Descriptive data

analysis followed the guidelines sug-

gested by Bracken and Barona (1991).Descriptive statistics for each groupwere computed along with reliability co-efficients (Cronbach’s alpha) for IA andHI factors. To assess the magnitude ofdifference across gender, effect sizes foreach item were computed for both theCA and AA groups. Because effect sizesare a standardized measure, they allowfor direct comparison of differences. Ef-fect sizes were calculated by subtractingthe female mean from the male mean,then dividing by the square root of thepooled variance.

Multivariate Analysis. To test for

differences in item means across genderand ethnicity, a 2 (gender) x 2 (ethnicity)multivariate analysis of variance was

performed. Separate analyses were con-ducted for the Total group and the At-

Risk group.

Structural Equation Modeling. This sanalysis was conducted only for the Totalgroup, because the number of AA malesand females in the At-Risk group was toosmall. To test whether the two-factor (HIand IA) model of ADHD based on theDSM-IV was consistent across gender,structural equation modeling (SEM) wasused to compare the factor structure, fac-tor correlations, item loadings, and itemuniqueness across gender, following theprocedures (LISREL 8) suggested byJoreskog and Sorbom (1993) and Ben-son (1987).

The two-factor model was shown to

adequately model observed data whenfactor analysis was conducted on theARS during the norming process (Du-Paul, Power, Anastopoulos, & Reid,1998). To determine if the factor struc-ture was invariant across genders, a four-step procedure was used. First, malesand females were compared with all

parameters (i.e., item factor loadings,factor correlations, and item uniqueness)constrained to be equal to those of themales. This step provides a baseline esti-mate of model fit. Second, separate es-timates of item-factor loadings were

computed for males and females. Third,separate estimates of item-factor load-

ings and factor correlations were com-puted for males and females. The laststep involved computing the separate es-timates of factor loadings, factor correla-tions, and item uniqueness (random mea-surement error and unique item variance)for males and females. The results from

Steps 2 through 4 were then compared tothe results of Step 1 to assess model fit

with each additional freed parameter.The chi-square difference test was usedto determine whether freeing a restraintimproved the overall fit of the model

(Bentler & Bonett, 1980). All parameterestimates were performed using covari-ance matrices and generalized least-

square estimation.

Discriminant Function Analysis. Todetermine the extent to which ARS items

were predictive of gender, we conducteddiscriminant function analysis using aprocedure suggested by Huberty (1984).A step-wise analysis was used to deter-mine the most parsimonious group of



41

predictor variables and to analyze eachvariable in terms of its effect on between-

gender discrimination. Separate analyseswere conducted for the Total group and

the At-Risk group.

RESULTS

Descriptive DataItem means for the Total group and theAt-Risk group are shown in Figures 1

and 2. The reliability of the scale for allgroups was consistently high. Cron-

bach’s alpha for both the HI and IAscales was greater than .90 for AA andCA participants for both the Total andAt-Risk groups. Table 1 shows between-

gender effect sizes by item, separatelyfor CA and AA ethnic groups within theTotal group and the At-Risk group. In

general, between-gender differences

were slightly greater for the CA groupthan for the AA group. However, forsome items such as Item 7 (fails to finishwork), this was not the case.

For most items, between-gender dif-ferences tended to be higher for the At-Risk group than for the Total group.However, in the case of the AA At-Risk

group, between-gender differences de-creased markedly for four of the nine IAitems: 1 (fails to attend/careless), 5 (doesn’tlisten), 7 (fails to finish work), and 9 (dif-ficulty with organization). As Table 1

shows, the average between-gender ef-fect sizes for the Total group were CA =

.40, AA = .36, and for the At-Risk group,CA = .52, AA = .34.

Multivariate AnalysisFor the Total group, significant effectswere found both for gender, F( 18, 3,310)= 12.66, p < .001; 112 = .065, and ethnic-ity, F(18, 3,310) = 9.75, p < .001; q2 =.053. All univariate tests were significant(p < .001 for all items) for both genderand ethnicity. A significant gender byethnicity interaction was also found,F(18, 3,301) = 2.57, p < .001; q2 = .014.However, none of the univariate tests at-tained significance. Only two items ap-

TABLE IBetween-Gender Effect Sizes for I 8 Items of the ARS

Note. ARS = ADHD-IV Rating Scale-School Version (DuPaul, Power, Anastopoulous, & Reid ( 1998).

proximated the .05 level (Item 4 [leavesseat], p = .064, and Item 7 [fails to finishwork], p = .085). For the At-Risk group,significant effects also were found bothfor gender, F( 18, 511 ) = 5.70, p < .001;q2 = .167, and ethnicity F( 18, 511) =5.82, p < .001; q2 = .170. With the ex-ception of Item 7 (fails to finish work),all univariate tests were significant(p < .05) for both gender and ethnicity.The gender by ethnicity interaction wasnot significant, F( 18, 511 ) =1.36, p = .15.

Because the MANOVA procedureuses a weighted composite to assess in-teractions, analysis is not straightfor-ward. The interaction for the total sam-

ple appears to be due to differences

among a small number of items, in com-bination with a wider separation of groupcentroids, between the CA groups thanthe AA groups. However, the practicalsignificance of the interaction is ques-tionable. Power was high, and the inter-action effect sizes (q2) were in the smallrange (Stevens, 1996).

SEM AnalysisTables 2 and 3 show the results of the

SEM analysis. The results at Step one,where all parameters are invariant, showthat all fit measures for both the CA

(Table 2) and AA (Table 3) groups are atlevels suggesting an acceptable fit acrossgender. The goodness of fit index is nearthe .90 level for both ethnic groups, sug-gesting an adequate fit. Also, the com-parative fit index and the non-normed fitindex are well above .90 for both groups.Additionally, the root mean square errorof approximation is at or below the .08level suggested as indicative of accept-able fit (Joreskog & Sorbom, 1993). Thisimplies that, from a measurement per-spective, there are no major qualitativedistinctions across gender for either theCA or the AA group. When separate es-timates of item uniquenesses were made,there was a statistically significant de-crease in chi-square for both groups.However, because the fit at Step one,where all parameters were constrainedacross gender, is acceptable, the mostconservative interpretation is that the

same two-factor ADHD model holds

across gender for both ethnic groups.



42

, . TABLE 2

. Test of Model Fit Across CA Males and Females

Note. GFI = goodness of fit index; RMSEA = root mean square error of approximation; CFI = comparative fit index; NNFI = non-normed fit index.*p < .01. &dquo; ’

’

TABLE 3 ’ ’ - <&dquo; ..

Test of Model Fit Across AA Males and Females

Note. GFI = goodness of fit index; RMSEA = root mean square error of approximation; CFI = comparative fit index; NNFI = non-normed fit index.*p < .0 I ...

Discriminant Function AnalysisCA Group. For the Total group, a sig-

nificant discriminant function for genderwith eight predictors was obtained,x2(8, N = 2636) = 302.36, p < .0001.Univariate statistics for each predictorare presented in Table 4. Group centroidsfor the male and female groups were.344 and -.354, respectively. For the At-Risk group, a significant discriminantfunction for gender with five predictorswas also obtained, x2(5, N = 421) =121.86, p < .0001. Univariate statis-tics for each predictor are presented inTable 4. Group centroids for the maleand female groups were .561 and -.612,respectively.

AA Group. For the Total group, a

significant discriminant function for

gender with four predictors was ob-

tained, x2(4, N = 686) = 48.09, p < .0001.Univariate statistics for each predictorare presented in Table 5. Group centroidsfor the male and female groups were.245 and -.297, respectively. For the At-

Risk group, a significant discriminantfunction with two predictors was alsoobtained, x2(2, N = 117) = 19.67,p < .0001. Univariate statistics for each

predictor are presented in Table 5. Groupcentroids for the male and female groupswere .396 and -.471, respectively.

DISCUSSIONThe results of this study suggest that gen-der has a significant effect on teacher rat-ings of ADHD symptomatology. Resultsalso suggest that there are no significantqualitative differences in symptomatol-ogy across gender and that the ADHDconstruct is consistent across gender.However, the results also suggest the

possibility of cross-gender differencesbased on ethnicity.

Uniqueness of SymptomatologyIf ADHD symptoms manifest differentlyacross gender, then we would expect tosee either a different pattern of item

means and effect sizes, or differences initem variance/covariance structure. Theresults of this study suggest just the op-posite. The observed pattern of item

means was strikingly similar across gen-der and ethnicity. As Figures 1 and 2

show, there is a consistent pattern: theAA males were seen as most severe, theAA females and CA males were indis-

tinguishable and distinctly separate fromthe AA males, and the CA females wereseen as least severe and were distinctlyseparate from the AA females and CAmales. The between-gender item effectsizes, which incorporate item variancefor each group, further indicate that dif-

ferences were consistent across items.

If females at risk for ADHD manifest

higher rates of inattentive or hyperactivebehavior, we would expect to see smaller

between-gender item effect sizes for theAt-Risk group than for the Total groupfor items on the IA or HI factors. How-

ever, this result was not generally ob-tained. Aside from smaller effect sizes

for a few items for the AA At-Risk group



43

.

TABLE 4. TABLE 4

Stepwise Discriminant Function Results for CA Group

Note. Item 2 = fidgets; Item I = talks excessively; Item 14 = blurts answers; Item 8 = difficulty playing quietly; Item 4 = leaves seat; Item I = fails to attend/careless;Item I I = avoids sustained mental effort; Item 7 = fails to finish work; Item 9 = difficulty with organization; Item 15 = easily distracted.

TABLE 5

Stepwise Discriminant Function Results for AA Group

Note. Item 13 = loses things; Item 17 = forgetful; Item 2 = fidgets; Item 7 = fails to finish work; Item 13 = loses things; Item 6 = runs/climbs excessively.

(see Table 1), there appear to be no con-sistent pattern of item effect size differ-ences between the Total group and the At-

Risk group across either HI or IA factors.

Finally, the results of the SEM analy-ses suggest that, in terms of factor struc-ture, factor loadings, and factor correla-tions, there is no difference between

gender across ethnic groups. Thus, theresults of both the descriptive analysesand the SEM analyses strongly suggestthat, within the CA and AA groups, theADHD construct was consistent across

gender for the total sample.We should caution that, because the

number of AA students in the male and

female At-Risk groups fell below the

limits of a minimum of 100 participantsper group (Loehlin, 1992), we did notuse SEM to compare the At-Risk groups’structures. Thus, the results of the SEM

analyses are limited to the Total groups.It is possible that structural differencesexist between males and females in the

more severely involved groups.

Effects of Gender onTeacher RatingsThe results of this study are consistentwith previous analyses of the effects ofgender on behavioral ratings (e.g., Rey-

nolds & Kamphaus, 1992; Trites, Blouin,& Laprade, 1982) and strongly supportthe notion of gender differences in theperceived severity of symptom expres-sion as assessed by teachers’ behaviorratings. The MANOVA results indicatesignificant cross-gender differences

when one considers both the Total groupand the At-Risk group. The observed

between-gender effect sizes for items inTable 1 indicate that these differences arein the moderate range and exist for most

scale items. This suggests that perceivedgender differences are broad in nature, asopposed to being limited to a subset ofitems or a single dimension (i.e., IA or HI).



44

FIGURE I . Item means for the Total group.

Furthermore, between-gender differ-ences increased slightly in the At-Riskgroup; thus, gender differences increasedas symptom severity increased. This isconsistent with the fact that males aremore likely to be identified as havingADHD than females. Interestingly, in theMANOVA analyses, when both genderand ethnicity are considered simultane-ously, gender accounted for somewhatmore variance than ethnicity, as evi-denced by the q2 statistics. Thus, the ef-fect of gender appears to be equal to orslightly greater than that of ethnicity onteacher ratings.

In the discriminant function analyses,the items that most distinguished malesfrom females in terms of a unique effect(i.e., when effects of correlation withother variables are parceled out) werenot consistent across ethnic groups. Forthe CA group, a distinct pattern emerged.Items from the HI factor that reflect ex-

ternalizing behaviors (i.e., fidgets, talksexcessively, blurts out answers, difficultyplaying quietly, and difficulty organizing

tasks) constituted the majority of itemsand had the greatest effect on male-

female separation for both the Total

group and the At-Risk group. Thus, forthe CA group, externalizing behaviorseems to play a major role in defininggender differences. This is likely due, atleast in part, to the fact that these behav-iors are disruptive to the classroom envi-ronment and are more likely to be salientto the classroom teacher. For the AA

group, no such pattern emerged. Feweritems attained significance, and onlyone item was significant for both

the Total group and the At-Risk group(loses things necessary for tasks and

activities).This difference between the CA and

AA groups appears to be the result of

two factors. First, there is less differencebetween genders for the AA group, asevidenced by the between-gender effectsizes. Second, the shared variance foritems appears to be greater for the AA

group; thus, colinearity (i.e., high itemintercorrelations) becomes a problem.

This suggests that teachers’ perceptionsof the ADHD-related behaviors of AA

students is consistent across gender, withfew distinguishing behaviors.Why have some studies that have used

behavior ratings (Breen, 1989; Breen &

Altpeter, 1990; Horn et al., 1989; James& Taylor, 1990; Silverthom et al., 1996)failed to find statistically significantbetween-gender differences? This is likelydue to two reasons. First, sample sizes inprevious studies were small, rangingfrom 39 (Breen, 1989) to 80 (Silverthomet al., 1996). Thus, some studies mayhave lacked the statistical power needed

to find gender differences of moderatesize, such as those identified in the pres-ent study. A related problem lies in thesmall numbers of females with ADHD in

the studies noted above, ranging from 13(Breen, 1989) to 18 (James & Taylor,1990). Consequently, the stability of re-sults may be an issue. A second reason

lies in the participant selection process.The present study used a randomized se-lection process from a school-based pop-



45

FIGURE 2. Item means for the At-Risk group.

ulation. In contrast, in studies that failedto find differences, the participants weredrawn from clinically referred groups.One study used students identified byICD-9 standards (James & Taylor, 1990),and another used students with pervasiveADHD (Horn et al., 1989). The relianceon clinically referred groups has beenquestioned on the grounds that childrenwho were clinically referred may not berepresentative, as they may constitute anextreme group (Epstein, Shaywitz, Shay-witz, & Woolston, 1991 ).

It is also possible that there is a higherbehavioral threshold for females (Eme,1992). If so, females diagnosed as hav-ing ADHD would need to demonstrateextreme levels of behavior to receive anADHD diagnosis. Therefore, clinicallyreferred females with ADHD in these

studies may not be representative of thegeneral population of females with

ADHD. However, we cannot rule out the

possibility that there may be a subgroupof females who do not differ signifi-cantly from males.

Same or Different Norms

Present results have implications for theissue of separate ADHD norms for malesand females and/or different ethnic groups.

Generally stated, should we make judg-ments of deviance or disorder on thebasis of an absolute standard or a relative

standard? An absolute standard impliesthat a single threshold should be used todefine behavior as disordered and thatthis threshold is valid across potentiallymitigating factors, such as gender or eth-nicity. A relative standard implies thatthere should be different thresholds fordifferent groups (e.g., males and femalesor AA and CA groups) and that devianceor disorder should be judged within thecontext of group membership.

Results of the present study argue forthe relative approach (separate standardsfor different groups) for using behaviorrating scales. The use of an absolutestandard implies the use of a single cut-score for all groups. This, in turn, wouldrequire pooling the scores of groups that

are demonstrably different in terms ofmeans and variances for all scale items.The result of this process is a &dquo;hybrid,&dquo;composed of two dissimilar groups, witha cut-score that would not accuratelyrepresent either group. Males would bemore likely to screen positive (exceedthe cut-score), due to the inclusion of fe-males ; females would be less likely toscreen positive, due to the inclusion ofmales. The fact that gender differences(i.e., between-gender effect sizes) weremore pronounced as severity increasedwould further exacerbate this situation.

The problem of accurate representa-tion is compounded when ethnicity is in-cluded. Our results demonstrated signif-icant differences in item means acrossthe AA and CA groups (males and fe-males combined). Similarly, mean dif-ferences across AA and CA groups have

been demonstrated in a number of previ-ous studies and across different behavior

rating scales (e.g., DuPaul et al., 1997;Epstein et al., 1998; Jarvinen & Sprague,1995; Lambert et al., 1978; Reid et al.,



46

1998; Waechter et al., 1979). In addition,for the Total group, we found ethnicityby gender interaction, which apparentlywas due to two considerations. First, asevidenced by the effect size differences,there was greater separation betweengenders for the CA group than for theAA group. Second, the discriminant

analysis showed that the CA and AAgroups differed in terms of those itemsthat contributed most to gender groupseparation (see Tables 4 and 5). We notetwo cautions with regard to ethnicity bygender interaction: The interaction ef-fects were small, and the absence of aninteraction for the At-Risk groups sug-gests that the clinical significance of theinteraction is minimal.

Another point in support of separatenorms relates to the nature of behavior

rating scales. As Barkley (1987) noted,behavior rating scales are &dquo;simply quan-tifications of adult opinions. As a resultthey are subject to the same sourcesof unreliability as those opinions ....&dquo;(p. 219). A number of potential sourcesof rater-based error have been identified

(Reid & Maag, 1994). One of them-halo effects-has direct relevance to theuse of separate norms. Abikoff, Court-

ney, Pelham, and Koplewicz (1993) re-ported that halo effects, which inflate be-havior rating scale scores, can occur

when teachers rate students with opposi-tional behaviors. Similar findings havebeen reported by Schachar, Sandberg,and Rutter (1986). Since males are morelikely to demonstrate these types of be-haviors, they would be more likely to besubject to halo effects and, thus, havespuriously higher ratings.

Halo effects may also differ acrossethnic groups. Epstein et al. (1998) re-ported that items on the Conners TeacherRating Scale reflecting oppositional be-havior (e.g., quarrelsome, defiant, unco-operative) loaded on different factors forAA and CA males. Other studies havealso suggested the possibility of halo ef-fects for AA students (Reid et al., 1998)and rater effects for students who are eth-

nically different (Sonuga-Barke et al.,1993). We see the combination of differ-ences across gender and ethnic groupsand the possibility of systematic rater in-fluence on scores as a strong rationale for

the use of separate norms for behavioral

ratings based on gender, at least, and,perhaps, ethnicity. Rater effects and haloeffects are topics requiring additional re-search.

In summary, the present results showthat, although the two-factor ADHD con-struct is consistent across genders, thereare broad-based significant differencesin item scores across gender and ethnic-ity. However, these differences do notappear to be consistent across ethnic

groups. For the CA group, a consistent

group of items best discriminated males

and females in both the total sample andthe clinical group; however, no such pat-tern was evident for the AA group. The

results of this study also suggest that sep-arate norms are appropriate for malesand females.

We wish to emphasize several cau-tionary limitations. Present results andarguments are based on teacher ratingson the ARS rating scale. Parent ratings,for example, may produce different re-sults. In addition, differences betweenethnic groups could be due to socioeco-nomic status. Future research shouldcontrol for this variable. Finally, the

value of cut-scores should be determined

by the relationships between cut-scoresand functional impairment. Differencescannot be used synonymously with de-viance. Further research will be requiredto determine the relationships amonggender, ethnicity, socioeconomic status,and functional impairment, as assessedby behavior rating scales.

About the Authors

ROBERT REID is an associate professor inthe department of Special Education and Com-munication Disorders at the University ofNebraska. His research focuses on children

with attention and self-regulation difficulties.CYNTHIA A. RICCIO is currently an as-sistant professor at Texas A & M Universityin College Station. Her primary research in-terests include ADHD, neuropsychologicalassessment, and linguistic and cultural differ-ences, as they impact assessment and differ-ential diagnosis. ROBERT H. KESSLER re-ceived his MA and EdS in school psychologyfrom the University of Alabama. GEORGEJ. DuPAUL, PhD, is a professor and coordi-nator of the school psychology program at

Lehigh University. Currently, he is investi-

gating the effects of early intervention andschool-based interventions for students withADHD. THOMAS J. POWER, PhD, is anassistant professor of school psychology inpediatrics at the University of Pennsylvaniaand is co-director of the Center for Attentionand Learning Problems at the Children’s

Hospital of Philadelphia. His research inter-ests include assessment and intervertion for

children with ADHD and health promotion inurban schools. ARTHUR D. ANASTO-

POULOS, PhD, is an associate professor inthe Department of Psychology at the Univer-sity of North Carolina at Greensboro, wherehe also directs a specialty clinic serving chil-dren, adolescents, and adults with ADHD.His research interests include the impact ofADHD on family functioning. DIANAROGERS-ADKINSON is an associate pro-fessor and coordinator of the Emotional/

Behavioral Disorders Program at the Univer-sity of Wisconsin-Whitewater. Her researchinterests include the interaction of culture and

behavior and language function in childrenwith EBD. MARY-BETH NOLL has taughtstudents with EBD for 16 years and has

taught in university teacher preparation pro-grams for the past 9 years. Currently, she isan associate professor and chairperson of theDepartment of Special Education at St. CloudState University, where she continues to beinvolved in planning a variety of state, re-gional, and national professional develop-ment opportunities for teachers of studentswith EBD. Address: Robert Reid, 202 L

Barkley Center, University of Nebraska-

Lincoln, Lincoln, NE 68583-0732.

Note

For purposes of this study, Caucasian is definedas persons of European-American ancestry.

References

Abikoff, H., Courtney, M., Pelham, W., &

Koplewicz, H. (1993). Teachers’ ratings ofdisruptive behaviors: The influence of haloeffects. Journal of Abnormal Child Psy-chology, 21, 519-533.

American Psychiatric Association. (1994).Diagnostic and statistical manual of men-tal disorders (4th ed.). Washington, DC:Author.

Arnold, L. E. (1996). Sex differences in ADHD :Conference summary. Journal of Abnor-mal Child Psychology, 24, 555-569.

Barkley, R. A. (1987). The assessment of at-tention deficit-hyperactivity disorder. Be-havioral Assessment, 9, 207-233.



47

Barkley, R. A. (1989). Hyperactive girls andboys: Stimulant drug effects on mother-child interactions. Journal of Child Psy-chology and Psychiatry, 30, 379-390.

Barkley, R. A. (1996). Gender is already im-plicit in the diagnosis of ADHD: Shouldn’tit be explicit? ADHD Report, 4(2), 3-6.

Bauermeister, J. J., Berrios, V., Jimenez, A. L.,Acevedos, L., & Gordon, M. (1990). Someissues and instruments for the assessmentof attention-deficit hyperactivity disorderin Puerto Rican children. Journal of Clini-cal Child Psychology, 19, 9-16.

Befera, M., & Barkley, R. A. (1984). Hyper-active and normal girls and boys: Mother-child interactions, parent psychiatric sta-

tus, and child psychopathology. Journalof Child Psychology and Psychiatry, 26,439-452.

Benson, J. (1987). Detecting item bias in af-fective scales. Educational and Psycholog-ical Measurement, 47, 55-67.

Bentler, P. M., & Bonett, D. G. (1980). Sig-nificance tests and goodness of fit in theanalysis of covariance structures. Psycho-logical Bulletin, 88, 588-606.

Berry, C. A., Shaywitz, S. E., & Shaywitz,B. A. (1985). Girls with attention deficitdisorder: A silent minority? A report on be-havioral and cognitive characteristics. Pe-diatrics, 76, 801-809.

Bracken, B., & Barona, A. (1991). State ofthe art procedures for translating, validat-ing, and using psychoeducational tests incross-cultural assessment. School Psychol-ogy International, 12, 119-132.

Breen, M. J. (1989). Cognitive and behav-ioral differences in ADHD boys and girls.Journal of Child Psychology and Psychia-try, 30, 711-716.

Breen, M. J., & Altepeter, T. S. (1990). Situ-ational variability in boys and girls identi-fied as ADHD. Journal of Clinical Psy-chology, 46(4), 486-490.

Brown, R. T., Abramowitz, A. J., Madan-

Swain, A., Eckstrand, D., & Dulcan, M.

(1989, October). ADHD gender differ-ences in a clinic-referred sample. Paperpresented at the meeting of the AmericanAcademy of Child and Adolescent Psychi-atry, New York.

Brown, R. T., Madan-Swain, A., & Baldwin,K. (1991). Gender differences in a clinic-referred sample of attention-deficit disor-dered children. Child Psychiatry and Hu-man Development, 22, 111-128.

Chou, C., & Bentler, P. (1995). Estimates andtests in structural equation modeling. InR. Hoyle (Ed.), Structural equation model-ing: Concepts, issues, and practice (pp.37-55). Thousand Oaks, CA: Sage.

deHaas, P. A. (1986). Attentional styles andpeer relationships of hyperactive and nor-mal boys and girls. Journal of AbnormalChild Psychology, 14, 457-467.

DuPaul, G. J., Power, T. J., Anastopoulos, A.,Reid, R. (1998). ADHD Rating Scale IVManual. New York: Guilford.

DuPaul, G. J., Power, T. J., Anastopoulos, A.,Reid, R., McGoey, K. E., & Ikeda, M. J.(1997). Teacher ratings of attention-

deficit/hyperactivity disorder symptoms:Factor structure, normative data, and psy-chometric properties. Psychological As-sessment, 9, 436-444.

DuPaul, G. J., Power, T. J., McGooey, K.,Ikeda, M., & Anastopoulos, A. D. (1998).Reliability and validity of parent andteacher ratings of attention-deficit hyper-activity disorder symptoms. Journal ofPsychoeducational Assessment, 16, 55-68.

Eme, R. F. (1992). Selective female afflictionin the developmental disorders of child-hood : A literature review. Journal of Clin-ical Child Psychology, 21, 354-364.

Epstein, J. N., March, J. S., Conners, C. K.,& Jackson, D. L. (1998). Racial differ-ences on the Conners Teacher RatingScale. Journal of Abnormal Child Psy-chology, 26, 109-118.

Epstein, M. A., Shaywitz, S. E., Shaywitz,B. A., & Woolston, J. L. (1991). The boun-daries of attention deficit disorder. Journal

of Learning Disabilities, 24, 78-86.Ernst, M., Liebenauer, L. L., King, A. C.,

Fitzgerald, G. A., Cohen, R. M., & Za-

metkin, A. J. (1994). Reduced brain me-tabolism in hyperactive girls. Journal ofthe American Academy of Child and Ado-lescent Psychiatry, 33, 858-868.

Faraone, S. V., Beiderman, J., Keenan, K., &

Tsuang, M. T. (1991). A family-geneticstudy of girls with DSM-III attention def-icit disorder. American Journal of Psychi-atry, 148, 112-117.

Frick, P. J., Lahey, B. B., Applegate, B.,Kerdyck, L., Ollendick, T., Hynd, G. W.,Garfinkel, B., Greenhill, L., Biederman, J.,Barkley, R. A., McBurnett, K., Newcorn,J., & Waldman, I. (1994). DSM-IV fieldtrials for the disruptive behavior disorders:Symptom utility estimates. Journal of theAmerican Academy of Child and Adoles-cent Psychiatry, 33, 529-539.

Gaub, M., & Carlson, C. L. (1997). Genderdifferences in ADHD: A meta-analysis andcritical review. Journal of the AmericanAcademy of Child and Adolescent Psychi-atry, 36, 1036-1045.

Gordon, M. (1996). Must ADHD be an equalopportunity disorder? ADHD Reports, 4(2),1-3.

Gordon, M., & Mettelman, B. B. (1994).Gender differences in ADHD referrals: IQ,laboratory measures, and behavior ratings.ADHD/Hyperactivity Newsletter, 20, 7-9.

Gualtieri, T., & Hicks, R. E. (1985). An im-munoreactive theory of selective male af-fliction. Behavioral and Brain Sciences, 8,427-441.

Horn, W. F., Wagner, A. E., & lalongo, N.(1989). Sex differences in school-agedchildren with pervasive attention deficit

hyperactivity disorder. Journal of Abnor-mal Child Psychology, 17, 109-125.

Huberty, C. J. (1984). Issues in the use andinterpretation of discriminant analysis.Psychological Bulletin, 95, 156-171.

James, A., & Taylor, E. (1990). Sex dif-

ferences in the hyperkinetic syndrome ofchildhood. Journal of Child Psychologyand Psychiatry, 31, 437-446.

Jarvinen, D. W., & Sprague, R. L. (1995).Using ACTeRS to screen minority childrenfor ADHD: An examination of item bias.

Journal of Psychoeducational Assessment,Special Issue on ADHD, 172-184.

Joreskog, K., & Sorbom, D. (1993). Manualfor the LISREL 8. Hillsdale, NJ: Erlbaum.

Kashani, J., Chapel, J. L., Ellis, J., & Shekim,W. O. (1979). Hyperactive girls. Journal ofOperational Psychiatry, 10, 145-148.

Lahey, B. B., Applegate, B., McBurnett, K.,Biederman, J., Greenhill, L., Hynd, G. W.,Barkley, R. A., Newcorn, J., Jensen, P.,Richters, J., Garfinkel, B., Kerdyck, L.,Frick, P. J., Ollendick, T., Perez, D., Hart,E. L., Waldman, I., & Shaffer, D. (1994).DSM-IV field trials for attention deficit

hyperactivity disorder in children and ado-lescents. American Journal of Psychiatry,151, 1673-1685.

Lambert, N. M., Sandoval, J., & Sassone, D.

(1978). Prevalence of hyperactivity in ele-mentary school children as a function ofsocial system definers. American Journal

of Orthopsychiatry, 48, 446-463.Loehlin, J. C. (1992). Latent variable models.

Hillsdale, NJ: Erlbaum.

McGee, R., Williams, S., & Silva, P. A.

(1987). A comparison of girls and boyswith teacher-identified problems of atten-tion. Journal of the American Academy ofChild and Adolescent Psychiatry, 26,711-717.

Reid, R. (1995). Assessment of ADHD withculturally different groups: The use of be-havior rating scales. School PsychologyReview, 24, 537-560.

Reid, R., DuPaul, G. J., Power, T. J., Anasto-poulos, A. D., Rogers-Adkinson, D., Noll,M. B., & Riccio, C. A. (1998). Assessingculturally different students for attention



48

deficit hyperactivity disorder using behav-ior rating scales. Journal of AbnormalChild Psychology, 26, 187-198.

Reid, R., & Maag, J. W. (1994) How many"fidgets" in a "pretty much": A critique ofbehavior rating scales for identifying stu-dents with ADHD. Journal of School Psy-chology, 32, 339-354.

Reynolds, C. R., & Kamphaus, R. W. (1992).Behavior assessment system for children.Circle Pines, MN: American GuidanceService.

Schachar, R., Sandberg, S., & Rutter, M.

(1986). Agreement between teachers’ rat-ings and observations of hyperactivity,inattention, and defiance. Journal of Ab-normal Child Psychology, 14, 331-345.

Seidman, L. J., Biederman, J., Faraone, S. V.,Weber, W., Mennin, D., & Jones, J. (1997).

A pilot study of neuropsychological func-tion in girls with ADHD. Journal of theAmerican Academy of Child and Adoles-cent Psychiatry, 36, 366-373.

Silverthorn, P., Frick, P. J., Kuper, K., & Ott,J. (1996). Attention deficit hyperactivitydisorder and sex: A test of two etiologicalmodels to explain the male predominance.Journal of Clinical Child Psychology, 25,52-59.

Sonuga-Barke, E., Minocha, K., Tayor, E., &

Sandberg, S. (1993). Inter-ethnic bias inteachers’ ratings of childhood hyperactiv-ity. British Journal of Developmental Psy-chology, 11, 187-200.

Stevens, J. (1996). Applied multivariate sta-tistics for the social sciences (3rd ed.).Hillsdale, NJ: Erlbaum.

Tabachnick, B. G., & Fidell, L. S. (1983).Using multivariate statistics. New York:

Harper & Row.

Taylor, E. (1986). The overactive child.

Philadelphia: Lippincott.Trites, R. L., Blouin, A. G., & Laprade, K.

(1982). Factor analysis of the Conners

Teacher Rating Scale based on a large nor-mative sample. Journal of Consulting andClinical Psychology, 50, 615-623.

Waechter, D., Anderson, R., Juarez, L. J.,Langsdorf, R., & Madrigal, J. F. (1979).Ethnic group, hyperkinesis, and modes ofbehavior. Psychology in the Schools, 16,435-439.

Wechsler, D. (1974). Wechsler intelligencescale for children-Revised. New York: Psy-chological Corp.

(continued from p. 37)

Faggot, B. I. (1984). The consequences ofproblem behavior in toddler children.

Journal of Abnormal Child Psychology,12, 385-396.

Gewirtz, J. L., & Boyd, E. F. (1977). Experi-ments on mother-infant interactions un-

derlying mutual attachment acquisition:The infant conditions the mother. In T. Al-

loway, P. Pliner, & L. Kranes (Eds.), At-tachment behaviors: Advances in the studyof communication and affect (Vol. 3,pp. 109-143). New York: Plenum.

Glass, G. V., & Hopkins, K. D. (1984). Sta-tistical methods in education and psychol-ogy. Englewood Cliffs, NJ: Prentice Hall.

Kauffman, J. M. (1993). Characteristics ofemotional and behavioral disorders ofchildren and youth. Englewood Cliffs, NJ:Merrill.

Kazdin, A. (1987). Treatment of antisocialbehaviors in children: Current status and

future directions. Psychological Bulletin,102, 187-203.

Murray, A. D. (1979). Infant crying as an elic-itor of parental behaviors: An examinationof two models. Psychological Bulletin, 86,191-215.

National Association of State Boards of Edu-cation. (1997). Schools without fear (Re-port of the NASBE study group on vio-

lence and its impact on schools and learn-ing). Alexandria, VA: Author.

Neel, R., Meadow, N., Levine, P., & Edgar, E.(1988). What happens after special educa-tion : A statewide follow-up study of sec-ondary students who have EBD. Behav-ioral Disorders, 13(3), 209-216.

Nelson, J.R. (1996). Designing schools to

meet the needs of students who exhibit dis-

ruptive behaviors. Journal of Emotionaland Behavioral Disorders, 4, 147-161.

Nelson, J.R., Martella, R., & Galand, B.

(1998). The effects of teaching school ex-pectations and establishing consistent con-sequences on formal office disciplinaryactions. Journal of Emotional and Behav-ioral Disorders, 6, 153-161.

Patterson, G. (1982). Performance models forantisocial boys. American Psychologist,41, 432-444.

Patterson, G. R., Reid, J. B., & Dishion, T. J.

(1992). Antisocial boys. Eugene, OR:Castalia Press.

Shores, R. E., Jack, S. L., Gunter, P. L., Ellis,D. N., DeBriere, T. J., & Wehby, J. H.

(1993). Classroom interactions of childrenwith behavioral disorders. Journal of Emo-tional and Behavioral Disorders, 1, 27-39.

Shores, R. E., Gunter, P. L., Denny, S. L., &

Jack, S. L. (1993). Classroom management

strategies: Are they ecological contexts forcoercion? Behavioral Disorders, 17,178-190.

Stevens-Long, J. (1973). The effect of behav-ioral context on some aspects of adult dis-

ciplinary practice and affect. Child Devel-opment, 44, 476-484.

Thomas, A., Chess, S., & Birch, H. G. (1968).Temperament and behavior disorders in

children. New York: New York UniversityPress.

Todd, A. W., Horner, R. H., Sugai, G., &

Sprague, J. R. (1999). Effective behavior

support: Strengthening school-wide sys-tems through a team-based approach. Ef-fective School Practices, 17(4), 23-37.

Wagner, M., & Shaver, D. (1989). Educa-tional programs and achievements of sec-ondary special education students: Find-ings from the National LongitudinalTransition Study. Menlo Park, CA: SRIInternational.

Walker, H. M., Block-Pedego, A., Todis, B.,& Severson, H. (1991). School archivalrecords search (SARS). Longmont, CA:

Sopris West.Wickens, T. D. (1989). Multiway contingency

tables analysis for the social sciences.

Hillsdale, NJ: Erlbaum.



Documents

Journal of Emotional and Behavioral Disorders...(1979) found that females with ADHD were usually referred for learning prob-lems rather than behavior problems, whereas boys identified