This action might not be possible to undo. Are you sure you want to continue?
UNITED STATES AIR FORCE RESEARCH LABORATORY
NEAR IDENTITY OF COGNITIVE STRUCTURE IN SEX AND ETHNIC GROUPS
Thomas R. Carretta Training Effectiveness Branch Warfighter Training Research Division
Malcolm J. Ree Force Development Branch Cognition and Performance Division
HUMAN EFFECTIVENESS DIRECTORATE 7909 Lindbergh Drive Brooks AFB TX 78235-5352
Approved for public release; distribution unlimited.
AIR FORCE MATERIEL COMMAND AIR FORCE RESEARCH LABORATORY HUMAN EFFECTIVENESS DIRECTORATE WARFIGHTER TRAINING RESEARCH DIVISION 6001 South Power Road, Building 558 Mesa AZ 85206-0904
CARROLL. specifications. or if you no longer want to receive our technical reports. You may write or call the STINFO Office at DSN 240-3877. The fact that the Government formulated or supplied the drawings. or other data.NOTICES This report is published in the interest of scientific and technical information exchange and does not constitute approval or disapproval of its ideas or findings. The Office of Public Affairs has reviewed this paper. if your address changes. USAF Chief. CARRETTA Project Scientist DEE H. Brooks AFB TX 78235-5118. does not license the holder or any other person or corporation. specifications. where it will be available to the general public. use.walker@platinum. . Building 125. Using Government drawings. THOMAS R. E-mail address is shirlev. or Commercial (210) 536-3877. or convey any rights or permission to manufacture. Rm 158. Warfighter Training Research Division Please notify AFRL/HEOP.mil. and it is releasable to the National Technical Information Service.brooks. or other data included in this document for any purpose other than Government-related procurement does not in any way obligate the US Government. Colonel. ANDREWS Technical Director LYNN A. This paper has been reviewed and is approved for publication.af. 2509 Kennedy Circle. or sell any patented invention that may relate to them. including foreign nationals.
Regressions between pairs of groups indicated that there was no mean difference in loadings between males and females and among the ethnic groups. (510) 536-3622 12a. PRICE CODE 20. FUNDING NUMBERS Near Identity of Cognitive Structure in Sex and Ethnic Groups 6. Send comments regarding this burden estimate or any other aspect of mis collection o' information. Perceptual speed. Ethnic groups. Asian American. NUMBER OF PAGES 19 16. Ree 7. SPONSORING/MONITORING Air Force Research Laboratory Human Effectiveness Directorate Warfighter Training Research Division 6001 South Power Road. Verbal ability 17. 1215 Jeffereon Davis Hignway. Although many of the tests of the differences in the loadings were statistically significant. TITLE AND SUBTITLE 2. distribution unlimited. and to the Office of Management and Budget. Spatial ability.July 1996 to September 1996 5. and Perceptual Speed. SECURITY CLASSIFICATION OF THIS PAGE 19. approaching r = 1. searching existing data sources.05). ASVAB. ABSTRACT (Maximum 200 words) Comparisons of aptitude factor structures were made in large samples of young Americans who took a multiple aptitude test battery. Air Force Officer Qualifying Test: Armed Services Vocational Aptitude Battery. Training Effectiveness Branch 7909 Lindbergh Drive Brooks AFB TX 78235-5352 9. SECURITY CLASSIFICATION OF ABSTRACT 15. SECURITY CLASSIFICATION OF REPORT 18. The factor model that was used had been statistically confirmed. along with previous research. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. Arlington. Suite 1204. SUBJECT TERMS Academic aptitude. VA 22202-4302. Black. Hispanic. Paperwork Reduction Project (0704-0188).62205F .REPORT DOCUMENTATION PAGE Form Approved OMB No. REPORT TYPE AND DATES COVERED August 1998 Interim . Math ability. The most notable differences occurred for a test of aviation knowledge. Correlations between factor loadings for the sex groups and for all pairs of ethnic groups were very high. Math. It had a lower loading for men than for women on both hierarchical g and on the lower-order Technical Knowledge factor. including suggestions for reducing this burden. Aviation knowledge. DC 20503 1. Spatial. Washington. It included hierarchical g and five lower-order factors representing Verbal. The model showed good fit for both the sexes (male and female) and ethnic group White. 13. DISTRIBUTION CODE Approved for public release. DISTRIBUTION/AVAILABILITY STATEMENT 12b. 14.0. Intellect. Direetorate for Information Operations and Reports. Confirmatory factor techniques that imposed statistical constraints tested if the loadings of the tests were the same for both sexes and for Whites versus each of the other ethnic groups. AFOQT. Technical knowledge. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) PE PR TA WU . and Native American) comparisons. Carretta Malcolm J. gathenng and maintaining the data needed. Technical Knowledge. AGENCY USE ONLY (Leave blank) 4. Bldg 558 MesaAZ 85206-0904 11. LIMITATION OF ABSTRACT Unclassified NSN 7540-01-280-5500 Unclassified Unclassified UL Standard Form 298 (Rev 2-89) Prescribed by ANSI Std Z-39-18 298-102 COMPUTER GENERATED . to Washington Headquarters Services. present a consistent picture of near identity of structure of intellect for sex and ethnic groups. Carretta. PERFORMING ORGANIZATION Air Force Research Laboratory Human Effectiveness Directorate Warfighter Training Research Division. Sexes. It also had a lower loading for Whites than for Blacks and Asian Americans on the lower-order Technical Knowledge factor. The proportions of total and common variance accounted for by g and the five lower-order factors were similar for men and women and for all five ethnic groups. REPORT DATE 3. SUPPLEMENTARY NOTES AL/HR-TP-1997-0006 Air Force Research Laboratory Technical Monitor: Dr Thomas R. most differences were small in magnitude (less than . 0704-0188 Public reporting burden for this collection of information is estimated to average 1 hour per response.1123 -Bl -01 8. and completing and reviewing the collection of information. including the time for reviewing instructions. AUTHOR(S) Thomas R. These findings.
..-:::.' r. 6 8 n FIGURES Figure No. 1 Structural Model of the Air Force Officer Qualifying Test TABLES Table No.:.CONTENTS Page SUMMARY INTRODUCTION Studies of Mean Differences Studies of Factorial Similarity METHOD Participants Measures Procedures 1 1 1 2 4 4 4 5 RESULTS DISCUSSION REFERENCES . 1 2 3 Percentage of Total Variance Accounted for by Factors Within the Groups Factor Loadings by Sex and Ethnicity for the Hierarchical g Factor Loadings by Sex and Ethnicity for the Residualized Lower-Order Factors Page 7 8 o in ...
We would like to thank Paul J. and Dr Joseph L. Air Force aircrew personnel. Brooks AFB. (1995). M. Weeks for his comments. S. Pilot Selection and Classification Support. 149-155. which is dedicated to research into the selection and classification of U. TX 78235-5352. R. Send electronic mail to Thomas. & Ree. Send written correspondence to AFRL/HEAB. IV . 7909 Lindbergh Drive. This paper has been published previously as Carretta.afmil.PREFACE This effort was conducted under Work Unit 1123-B1-01. 7P.Carretta(S)williams. Personality and Individual Differences. J. Rioux for his technical assistance in preparing the data files. Near identity of cognitive structure in sex and ethnic groups. T.
While informative.45s for verbal. and Perceptual Speed. The model showed good fit for both the sexes (male and female) and ethnic group (White. these studies did not address differences in factor structure among groups. The factor model that was used had been statistically confirmed. and Spuehler (1975) provided a summary of mean difference studies among groups. most differences were small in magnitude (less than . and NativeAmerican) comparisons. along with previous research. Spatial. and .05). Math. Although many of the tests of the differences in the loadings were statistically significant. Hyde (1981) in a meta-analysis of male-female differences in cognitive abilities. The current effort remedies that deficit by providing an investigation of the structure of cognitive abilities for sex and racial/ethnic groups.0. Black. Fifer. The most notable differences occurred for a test of aviation knowledge. present a consistent picture of near identity of structure of intellect for sex and ethnic groups. For example. Lindzey. Correlations between factor loadings for the sex groups and for all pairs of ethnic groups were very high. Technical Knowledge. It included hierarchical g and five lower-order factors representing Verbal. Regressions between pairs of groups indicated that there was no mean difference in loadings between males and females and among the ethnic groups. These findings. and . Lesser. It also had a lower loading for Whites than for Blacks and Asian-Americans on the lower-order Technical Knowledge factor. Studies of Mean Differences There have been several studies of aptitude differences between sex and ethnic groups.NEAR IDENTITY OF COGNITIVE STRUCTURE IN SEX AND ETHNIC GROUPS SUMMARY Comparisons of aptitude factor structures were made in large samples of young Americans who took a multiple aptitude test battery. and Clark (1965) investigated differences among ethnic groups on abilities.45s for visual-spatial ability. Similarly. Willerman (1979) reviewed several studies comparing average differences in factors for American and foreign racial and ethnic groups. reported median differences of . They compared mean profiles of scores for Americans identified as Black. approaching r = 1. Hispanic.43s for quantitative. Jewish. Asian-American. It had a lower loading for men than for women on both hierarchical g and on the lower-order Technical Knowledge factor. Loehlin. Chinese. The proportions of total and common variance accounted for by g and the five lower-order factors were similar for men and women and for all five ethnic groups. Confirmatory factor techniques that imposed statistical constraints tested if the loadings of the tests were the same for both sexes and for Whites versus each of the other ethnic groups. INTRODUCTION There have been numerous studies of group differences in mean aptitude scores. .
Indians. 1980. across sex and ethnic groups using a hierarchical model of ability (see Vernon. (1974) compared the structure of ability for two ethnic groups as measured by 15 cognitive tests. . Space perception. More informative analyses were conducted by Ree and Carretta (1995) who investigated ' the comparative structure of another multiple aptitude battery. More recently. Both studies concluded that it was unlikely that underlying dimensions of ability varied. Lindzey. Ree & Earles. 1969) and conducted statistical tests of the similarity of the factor loadings. the Black and Jewish groups had their highest average scores on the verbal factor. Lynn and Owen (1994) studied three ethnic groups in South Africa (Whites. For example. Loehlin. S. In each case the White mean was higher. 1991). similar patterns of intercorrelations among aptitude measures have been found for Blacks and Whites in studies by Nichols (1970) and Scarr-Salapatek (1971).1 cr and the mean standardized White-Indian difference was 1. in contrast. Each of these common factors was mostly general cognitive ability (Jensen.Puerto Rican. Studies ofFactorial Similarity Several studies of factorial similarity have been conducted. Humphreys and Taber (1973) compared factor structures for high and low socio-economic status boys from Project Talent and found no discernible differences. These previous studies all investigated the relationships among groups for common factors. Michael (1949) compared the factor structure of World War IIU.0 c. the ASVAB. Their participants were Hawaiians of either Japanese or European ancestry. was relatively lower for the Black and Jewish groups and highest for the Chinese and Puerto Rican groups. They found not only mean differences. In American samples. They concluded that the same four factors were present in both groups and that the factor loadings were nearly identical for the two groups. Army Air Force pilot candidate selection tests for Blacks and Whites and found virtually no differences. They found only small differences on the Verbal/Math and Speed factors. while the Chinese and Puerto Rican groups had their lowest average scores on this factor. there were notable differences on the Technical Knowledge factor. and Blacks). The mean standardized White-Black difference was about 2. but also differences in the pattern of mean profiles among the groups. DeFries et al. and Spuehler (1975) noted that although the ethnicity of the participants in Project Talent was not specifically identified. it could be expected that the ethnic composition of the two groups would differ significantly. However. It should be noted that lack of English as a first language for the Black group may have contributed to the observed White-Black difference on the tests. These differences in dimensionality would give rise to different factor structures that could be statistically tested. Most of the differences in this factor were due to a single specific knowledge test of electronics information.
Carretta and Ree (1996) used confirmatory factor methods (Bentler. and Navigator-Technical) and the same with a hierarchical g added. This resulted in a model with hierarchical g and residualized lower-order Verbal/Math (combined from Skinner & Ree. two models based on the operational AFOQT composites (not including the Academic Aptitude composite which would have been redundant with the Verbal and Quantitative composites and would have caused a linear dependency in the data). As with Skinner and Ree (1987). Math. They examined seven structural models including: g only. Skinner and Ree (1987) conducted exploratory factor analyses of the AFOQT using a sample of 3.43. The correlations among the 16 tests ranged from .The ASVAB is relatively factor poor. Quantitative. 1987). The Skinner and Ree (1987) model with hierarchical g and five 5 lower-order factors was selected as it showed both good fit and interpretability. Technical Knowledge. The two Skinner and Ree (1987) models were 5-factors (Verbal.77. Earles and Ree (1991) in a reanalysis of the Skinner and Ree (1987) data. They found an hierarchical estimate of general cognitive ability. 1987). Although the sample contained both men and women and several ethnic groups. Technical Knowledge (Aircrew Interest/Aptitude). Based on a principal factors analysis with communalities in the principal diagonal and Kaiser-Harris Type II oblique rotation. Pilot. no group-specific analyses were conducted. scores based on the factors cannot have the same meaning in research or in practical application. Model 7 combined the Verbal and Math factors to produce a Vernon-like model (Vernon. More recently.. and two models that simplified the Skinner and Ree (1987) solution by combining some of the lower-order factors. The current study investigated whether the factor structure of abilities was the same for men and women and for five ethnic groups. with an average correlation of . demonstrated the hierarchical structure of the AFOQT. 1969). Differences in factor structure or loadings could occur as a result of differences in personal choice (e. These factors showed an average correlation of . and Perceptual Speed. g. Math. The Carretta and Ree (1996) analyses did not address possible group differences as sex and ethnic group comparisons were not made.000 US Air Force officer applicants. 1989) to test the structure of the AFOQT in the Skinner and Ree (1987) sample. . Spatial. If the factors or loadings of tests on the factors differ among or between groups. and Technical Knowledge factors.17 to . based on correlated lower-order factors. but kept all the others. leisure interests) and opportunity. course enrollment.22 to .50.36 with a range of . Spatial. two models based on Skinner and Ree (1987). The Air Force Officer Qualifying Test (AFOQT) has a greater number of tests and a greater diversity of factors. they described a 5-factor solution: Verbal. Model 6 was a simplification of Skinner and Ree (1987) with hierarchical g that combined the Spatial and Perceptual Speed factors. they did not test the fit of the model or perform group specific analyses. and Perceptual Speed) and 5-factors with a hierarchical g added. The two models based on operational composites were a 4-factor solution (Verbal. Spatial/Perceptual Speed (combined from Skinner & Ree.g.
Mechanical Comprehension (MC).7% White. Scale Reading (SR). AFOQT scores are used along with other measures of aptitude and educational achievement (e.5% Asian-American. Finally. Math Knowledge (MK). Their ages at time of testing ranged between 18 and 27 years. The Carretta and Ree (1996) factor structure was used as the basis for the current study as it had been statistically confirmed and was meaningfully interpretable. and Hidden Figures (HF). and speed. Reading Comprehension (RC). Sorrentino. Skinner & Ree. Instrument Comprehension (IC). Arithmetic Reasoning (AR).g. Table Reading (TR).9% Native-American.2% Black. A more detailed description of the tests was provided elsewhere (Carretta & Ree. college grade point average. These qualities made it superior to the other models. 4. Skinner & Ree 1987). 1990. Measures The AFOQT is a multiple-aptitude test battery used for the commissioning of US Air Force officers through Officer Training School (OTS) and the Reserve Officer Training Corps (ROTC).7% Hispanic.4% male and 78. mathematical. Its reliability has been examined (Earles & Ree. quantitative. Aviation Information (AI). Block Counting (BC). differences in factor loadings cannot necessarily be interpreted to indicate test bias. 12. 3. and spatial factors. sex.. 1987) and it has ' been validated for pilot and navigator training (Arth. They were 81. . Rotated Blocks (RB). and 0. Electrical Maze (EM). and ethnicity of the participants were collected by self-report. 1991. Word Knowledge (WK). & Burke. Steuck. All of these tests stem from the theoretical model offered by Kelley (1928) and contain several of the factors he posited such as spatial. type of degree. The age. Data Interpretation (DI). It differs from the others in its measurement of specialized aviation knowledge. verbal.968 applicants for US Air Force officer commissions tested between 1981 and 1993. METHOD Participants The participants were 269. previous flying experience) to qualify applicants who pass medical and physical requirements for pilot and navigator training. and the Armed Services Vocational Aptitude Battery because it' contains measures of verbal.The consequences of group differences in aptitude measurement on operational scores are a function of the magnitude of the differences in factor loadings and the number of factors that contribute to operational scores. 1995. These consequences extend to using the test for research purposes. The AFOQT is similar to many other multiple aptitude tests such as the Differential Aptitude Tests General Aptitude Test Battery. The goal of this study was to determine if the factor structure was the same for men and women and for five ethnic groups. General Science (GS). The AFOQT consists of 16 tests that include Verbal Analogies (VA).
Procedures The factor structure tested across groups in this study was confirmed by Carretta and Ree (1996) and represented a 6-factor solution (See Figure 1). MK. and Perceptual Speed (DI. This procedure lets each group-specific (e. EM. AI. BC.g. All 16 AFOQT tests were used in this study. GS). TR).Carretta.g. BC.. 1994. HF). 1982. GS). it is necessary to estimate the structural model parameters with the loadings for both groups constrained to the same value and then to estimate the model parameters with the constraint removed. Technical Knowledge (MC. The hierarchical factor is g (all 16 tests) and the five lower-order factors are Verbal (VA. Spatial (MC. . WK. RC. 1957) with the effects of the hierarchical factor removed from the five lower-order factors. 1992. SR. To determine if the factor loadings for a test were the same for two groups (e.1995. DI. IC. Maximum likelihood estimation procedures were used. Carretta & Ree. Koonce. 1994).. Math (AR. SR). Structural Model of the Air Force Officer Qualifying Test A chi-square test was conducted to determine if the loadings for a variable on a factor for two groups were the same (Bentler. Whites and Blacks). 1989). The model is specified in residualized form (Schmid & Leiman. Olea & Ree. VA RC WK GS \R Dl VIK SR MC EM BC RB HF IC AI TR Figure 1. RB.
The proportions of total variance accounted for by g were 33% for Whites. The tests of equality of the loadings were done by separately comparing the majority group (Whites) with each of the four minority groups: Whites versus Blacks.052 (x2 = 143. Several goodness-of-fit statistics were considered. Balla. The models were evaluated with all constraints in place. The percent of total variance accounted for by each factor is shown for each group in Table L^ proportlon of total variance accounted for by g was similar for males (38%) and females (40^).927 with a RMSEA of . 35% for Blacks. 33% for Asian-Americans. the Tucker-Lewis Incremental Fit Index (TLI) (Tucker & Lewis. Additionally. In this study. Pearson correlations of factor loadings for pairs of groups (males and females. 59%. . the CFI was chosen to evaluate the goodness-of-fit of the models. The proportions of total and common variance accounted for by the five lower-order factors also were similar for both sexes and across all ethnic groups. 59%. 1990). A Type I error rate of p < .01 was used for all statistical tests. 34% for Hispanics. Regressions were computed and the regression constant (y intercept) was evaluated to determine whether a constant bias existed. The proportion of common variance accounted for by g also was similar for males and females (66% versus 67%). and the Comparative Fit Index (CFI) (Bentler. the RMSEA that gives an indication of amount of error per degree of freedom. The BBNNT has been shown to be susceptible to sample size effects (Marsh. df = 203). Whites versus Asian-Americans. 1989). and 59%. The CFI value for fitting the 5 group ethnicity comparison was . and Whites versus Native-Americans.522. which has been shown to be less susceptible to the effects of sample size with a smaller sampling variance than the TLI. The CFI value for fitting the male versus female comparative model was .089.936 with a RMSEA of . & McDonald. These included the Bentler-Bonett NonNormed Fit Index (BBNNI) (Bentler. If the chi-square test is significant. is presented. the root mean square error of approximation (RMSEA) (Browne & Cudeck. 1988). the loadings of males versus females for each test on each of the six ' factors were tested. separate parameter estimates for each group provide a better fit and are statistically justifiable. 59%. Whites versus Hispanics.Whites or Blacks) factor loading estimate have a different value. The proportions of common variance for which g accounted for these groups were similar: 60%. RESULTS No special problems occurred in estimating the parameters for the models. 1973). df= 563). Bentler (1989) extended the TLI to develop the CFI. As a result. 1993).031 (X2= 146. The most important ethnic comparisons are those between the majority and each minority group. and 33% for NativeAmericans. all pairs of ethnic groups) were examined to further explore the relationship of factor loadings between groups.
Native-Americans.62 3. . The correlation of the loadings for males and females for the hierarchical g factor was . . Pearson correlations between factor loadings for the sex groups and among the ethnic groups were very high.87 4.684 versus . approaching r = 1.99 for pairs of the five ethnic groups.717.90 3. . .57 5.93 ■ 33. Asian-Americans.94 7.795.69 2. IC. These were the loadings for EM. .92 2. Blacks.15 39. Although many of the univariate tests for differences in factor loadings between males and females were statistically significant (30 of 37).90 5.78 3.05.88 " 3.772.00 6.58 2.73 2.Table 1. These included EM and IC on g for Whites versus Blacks.41 5.73 3.800 on this lower-order factor. AI showed the largest male/female difference. . Results for the residualized lower-order factors showed the same range.11 6. and AI on g and for AI on the lower-order Technical Knowledge factor.94 7.97 to .92 3.79 2. and AI on the lower-order Technical Knowledge factor for Whites versus Blacks and Whites versus Asian-Americans. Hispanics. Tables 2 (g) and 3 (five lower-order factors) also show the loadings for the five ethnic groups. Although many of the univariate tests for differences in loadings between Whites and the other ethnic groups were statistically significant.87 2.42 5. only four pairs of loadings differed by more than .79 7.05 in magnitude.75 2.05 35.97 and ranged from .75 2.97 33. EM on the lower-order Spatial factor for Whites versus Blacks.81 7.31 2. The largest differences in loadings occurred for the AI test on the lower-order Technical Knowledge factor (Whites.77 The loadings of the tests on the factors for males and females are shown in Tables 2 (g) and 3 (five lower-order factors).0.42 33.58 2.59 3. .88 3.695. Percent of Total Variance Accounted for bv Factors Within the Groups Percent of Total Variance Group g Verbal Math Spatial Technical Knowledge Perceptual Speed Males Females Whites Blacks Hispanics Asian-Americans Native-Americans 38. few were larger than .88 2.81 5.78 2.709).20 2.52 32.30 5.16 7.
556 . indicating that there was no mean difference in loadings.712 . 1996).667 . Further.628 . .759 .516 .372 .446 .439 .611 .706 .584 .456 .572 .618 .627 .713 .660 .589 . The AFOQT is somewhat less g-saturated than the military enlistment test battery.606 .513 . g accounted for more of the total variance than did the sum of Ree rcSn^SS")™8 " COnSiStem Wlth reSUltS f°r °ther 3PtitUde tCStS (JCnSen' '980.568 .610 .585 .725 .564 .477 .738 .520 .524 .737 .673 .588 .463 .705 .457 .692 .521 .710 .534 .612 .484 .574 .619 .521 .326 .564 .609 .724 .332 .362 .652 .485 .Table 2.623 .705 .531 .432 .588 .715 .619 .635 .682 .583 .441 . Native-Am.615 .744 .710 .488 .432 .705 .646 .577 .539 .637 .661 .518 .484 Regressions were computed between the sex groups and between all pairs of the ethnic groups.448 .634 . VA AR .705 .570 .738 .707 . the Armed Services Vocational Aptitude Battery (Carretta & Ree.375 .596 .446 .435 .442 . """""" s Sex Ethnicity Test Male Female White Black Hispanic Asian-Am.730 .578 . All the regressions yielded intercepts of approximately zero. Hierarchical g accounted for the greatest amount of variance for both sexes and all ethnic groups and differed little by sex or ethnicity.527 .633 .RC DI WK MK MC EM SR IC BC TR AI RB GS HF .723 .512 .601 .446 . DISCUSSION Fit statistics indicated that the data gave a good fit to the model for both sexes and for the five ethnic groups.446 .693 .457 .615 .336 .667 .753 .494 .585 .604 .607 .724 .730 .580 .730 .512 .
244 .213 .322 .390 .106 .629 .788 .093 .310 .312 .684 .147 .338 .298 .464 .481 .236 .278 .333 .342 .253 .772 .148 .455 .281 .292 .458 .736 .272 .675 .214 .334 .232 .695 .328 .795 .106 .432 .630 .464 .317 .321 .404 .274 .768 .440 .573 .717 .577 .248 .347 .249 .628 .317 IC .728 .253 .422 .149 .205 .323 .764 .335 .376 .661 .646 .277 .451 .146 .289 .299 .376 .284 .415 .109 .094 .265 .379 .311 AI .464 .298 .391 .107 .207 .449 .381 .279 .800 GS .675 .143 .310 .317 .241 .234 Perceptual DI SR BC TR Speed .197 .371 .248 Technical Knowledge MC .288 .452 .239 .214 .222 .244 .278 .345 .335 .463 .Table 3.326 .413 .405 .406 .651 .298 .237 .234 .781 .295 .206 .110 .250 .247 .415 . Factor Loadings by Sex and Ethnicity for the Residualized Lower-Order Factors Sex Factor/Test Male Female White Ethnicity Black Hispanic Asian-Am Native-Am Verbal VA --RC WK GS Math AR DI MK SR Spatial MC EM BC RB HF .621 .709 .270 .778 .326 .322 .247 .301 .246 .667 .662 .345 .337 .312 .145 .151 .207 .324 .313 .624 .296 .348 .294 .310 .206 .458 .
Among the residualized lower-order factors. For the ethnic group comparisons. Ree and Carretta. and NativeAmericans (5. where the reverse was true. It is likely that these differences can be attributed to opportunity or individual decisions to learn aviation principles and flying. All three of these tests had lower loadings for males than for females on g and AI had a lower loading for males on the lower-order Technical Knowledge factor.4% and 9.9% and 10.4%>than for females (5. Also. 1974. Hispanics (5. The correlational and regression analyses were consistent with the findings of the comparisons of the structural model. For example. 10 . EM also had a lower loading on the lower-order Spatial factor for Whites than for Blacks. Instrument Comprehension (IC). the proportions of variance accounted for by the lower-order factors were similar across groups. and Aviation Information (AI) tests.4% and 10.6% and 9.3% and 7. The two largest sources of test variance for all groups are Verbal and Technical Knowledge. the following stand out. AI had a lower loading for Whites than for Blacks and Asian-Americans. These differences in loadings are most likely the consequences of personal choice and opportunity. 1949) that the structure of cognitive ability is nearly identical across sex and ethnic groups.5%). Asian-Americans (5.7%). These results further expand our understanding by including additional ethnic groups and by using a hierarchical model of human abilities. the most notable finding was the great similarity in percent-oftotal-variance-accounted-for by each of the factors for all groups..1% and 9. The results of this study add to the findings (DeFries et al.3%) and less for Whites (5. The most notable differences in factor loadings occurred for the Electrical Maze (EM). On the lower-order Technical Knowledge factor.6%). 1994) found differences in loadings for males and females on technical knowledge tests on a similar battery and attributed them to differences in high school course enrollment patterns.3% and 9. EM and IC had lower loadings on g for Whites than for Blacks. The Verbal factor accounted for more variance than did the Technical Knowledge factor for all groups except females.6%). The similarity in the proportions of total and common variance for g for the sex and ethnic groups suggested that the combination of unreliability and unique variance was similar across groups. Michael. However.2%) than for Blacks (6. The largest sex and ethnic group differences occurred for the Technical Knowledge factor which accounted for less total and common-variance for males (4. The reasons for these differences are not apparent.
& Burke. G. 338-339. R. Testing structural equation models. Bollen and J. Vandenberg. T.. Ability factors as a function of advantaged and disadvantaged groups. P. L. Carretta. Understanding the relations between selection factors and pilot training performance: Does the criterion make a difference? The Internationaljournal of Aviation Psychology. (1973). Carretta. J. Bentler. W. C. & Taber. BMDP Statistical Software. Air Force Officer Qualifying Test (AFOQT): Predictors of undergraduate pilot training and undergraduate navigator training success (AFHRL-TP-89-52). 4. (1992). Carretta. P. R. Lang (Eds. (1991). T. Armstrong Laboratory. The Internationaljournal of Aviation Psychology. A. (1990). Pilot-candidate selection method: Sources of validity. In K. 10. M. T. T. G. S. Browne.. Steuck. Military Psychology. R. Science 183. R. 103-117.). Journal of Business and Psychology. (1989). 892-901. C. S. Comparative fit indexes in structural models. J. 8. Brooks AFB. American Psychologist. A. & Cudeck. (1996). TX: Manpower and Personnel Research Division. M. Los Angeles. How large are cognitive gender differences? A meta-analysis using a2 and d. R. Psychological Bulletin 1-07. (1994). J. W. 29-42. E. 11 . J. (1993). DeFries. R. Air Force Human Resources Laboratory. Earles. 379-388. R. M. Kuse. T. Near identity of cognitive structure in two ethnic groups.. G.. Newbury Park. Ashton. F. Hyde. T.. A. Air Force Officer Qualifying Test validity for predicting pilot training performance... M. S. C.. (1981). 107-115. Bentler. J. Humphreys. (1990). J. Factor structure of the Air Force Officer Qualifying Test: Analysis and comparison. K. 95-105. Sorrentino. (1995). 36. Carretta. 2.. 0. T. R.. Air Force Officer Qualifying Test (AFOQT): Estimating the general ability component (AL-TP-1991-0039). CA: Sage.238-246. Brooks AFB. TX: Manpower and Personnel Division. EQS structural equations program manual.. & Johnson. Alternative ways of assessing model fit. & Ree. CA.REFERENCES Arth. M. Human Resources Directorate. Wilson. J. (1974). & Ree. C. 9. McClearn. G.. & Ree.. Journal of Educational Measurement. E. & Ree. M. M. J.
J & Earles. 174. Koonce. & Owen. & Spuehler. W. Scarr-Salapatek. Factor analyses of tests and criteria: A comparative study of two AAF pilot populations.. J. 79.. 121. social class. Mental abilities of children from different socialclass and cultural groups. R. M. Educational and Psychological Measurement. 12 . (1991). T.. Loehlin. 1285-1295. 53. (1971). (1994). Psychological Bulletin. G. P. Goodness-of-fit indexes in confirmatory factor analysis: The effect of sample size. N. W. R. R. and Environmental Medicine. H. C. 268-277. 63. 55. (1995). Fifer.27-36. San Francisco: Freeman. J. Psychological Monographs. A. (1988). 103. G.. Unpublished doctoral dissertation. M J & Carretta. R. (1970). Predicting pilot and navigator criteria: Not much more than g Journal of Applied Psychology. (1994).874.391-410. University of Minnesota. and Blacks in South Africa. T. Monographs of the Societyfor Research in Child Development. (1975). Spearman's hypothesis and test score differences between Whites Indians. The Journal of General Psychology. J. (1980). T.. & McDonald. J. M. M. J. D. M. S. Ree. Race differences in intelligence. and IQ. (1965). Educational and Psychological Measurement. (1994). Lindzey. Aviation. Science. J. Marsh. Lesser. Factor analysis of the ASVAB: Confirming a Vernonhke structure. Crossroads in the mind ofman: A study of differentiate mental abilities. 459-463. The effects of heredity and environment on intelligence test performance in 4 and 7 year white and Negro sibling pairs. L. MI: University Microfilms 71-18. 845-851. Lynn.321-332. Palo Alto. & Carretta. Nichols. & Clark. Space. Personnel Psychology. Ann Arbor.1166-1169. Race. 55-84. Kelley. (1928).. 44. R. (1982). A. P.. Ree. Olea. (1949). K. M.. J. Ree. B.. S. Balla. Validation of a proposed pilot-trainee selection system. Michael. Predicting training success: Not much more than?. Bias in mental testing. R. G. CA: Stanford University Press. & Ree. Group differences in aptitude factor structure. 54. New York: The Free Press. L. H.Jensen.
22. L. J. Psychometrika. L. (1957). Tucker. R. (1987). P.53-61. E. The psychology of individual and group differences. 38. Brooks AFB. Air Force Officer Qualifying Test (AFOQT): Item andfactor analysis of form O (AFHRL-TR-86-68). C.H. San Francisco CA: W. The development of hierarchical factor solutions. Air Force Human Resources Laboratory. Intelligence and cultural environment. M. & Lewis. Skinner. Willerman. London: Methuen. A reliability coefficient for maximum likelihood factor analysis. & Leiman.Schmid. J.. Freeman 13 . J. Psychometrika. (1973). Vemon. M. TX: Manpower and Personnel Division. 1-10. (1969). J. (1979). & Ree...
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.