ACT and general cognitive ability

1 ACT and general cognitive abilityKatherine A. Koenig , Meredith C. Frey, Douglas K. DettermanDepartment of Psychology, Case Western Reserve University, United StatesReceived 1 July 2006; received in revised form 16 March 2007; accepted 27 March 2007 Available online 2 May 2007 AbstractResearch on the SAT has shown a substantial correlation with measures ofgsuch as the Armed Services Vocational AptitudeBattery (ASVAB). Another widely administered test for college admission is the American College Test (ACT). Using the NationalLongitudinal Survey of Youth 1979, measures ofgwere derived from the ASVAB and correlated with ACT scores for 1075participants. The resulting correlation was .77. The ACT also shows significant correlations with the SAT and several standard IQtests. A more recent sample (N=149) consisting of ACT scores and the Raven's APM shows a correlation of .61 between Raven's-derived IQ scores and Composite ACT scores.

2 It appears that ACT scores can be used to accurately predict IQ in the generalpopulation. 2007 Elsevier Inc. All rights :ACT; general cognitive ability ; SAT; Advanced progressive matrices; ASVABA primary concern of college-bound adolescents isperformance on a college admissions test. One of themost widely used tests is the American College Test(ACT). The ACT is accepted by colleges throughout theUnited States and is administered to over 1 millionstudents annually. Designed in 1959 as an alternative tothe SAT, the ACT purports to closely parallel highschool curriculum and to measure the preparedness ofthe test-taker for more advanced education. Accordingto the ACT web site: The ACT is ACT is not an aptitude or an IQ test (Facts aboutthe ACT).Frey and Detterman (2004)showed that theSAT was correlated with measures of general intelli-gence .82 (.87 when corrected for nonlinearity).

3 Inaddition, a correlation of .92 was found between SAT IVerbal+Math and ACT composite scores in a sample of103,525 students, and ACT Math correlated .89 withSAT I Math (Dorans, Lyu, Pommerich, & Houston,1997). Given the similarity between the SAT and theACT it is not unreasonable to expect that the ACTwouldshow similar correlations with general intelligence,despite claims to the contrary. However, to the best ofour knowledge, the relationship between the ACT andgeneral intelligence has never been investigated in alarge ACT is composed of four sections measuringMathematics, English, Reading, and Science, with acomposite score that is the average of the four subtestscores. The score range for each subtest is 1 36 with a2003 average of Composite and subtest scores havevaried little in the past decade, though changes to theACT were implemented in 2005 in the form of anoptional writing test (Facts about the ACT).

4 Available online at 36 (2008) 153 160 Corresponding author. Department of Psychology, Case WesternReserve University, 10900 Euclid Avenue, Cleveland, OH 44106,United Koenig).0160-2896/$ - see front matter 2007 Elsevier Inc. All rights research has focused on the usefulness of theACT for predicting success in andStanley (2002)found that ACT scores show a .70correlation with college graduation rates. In addition,ACT scores have been shown to correlate with collegeGPA from .54 to .63, and the ACT math subtestcorrelates with math GPA from .48 to .64 (Koretz &Berends, 2001; Pettijohn, 1995; Sibert & Ayers, 1989;Snowman, Leitner, Snyder, & Lockhart, 1980). Com-posite ACT scores are generally better at predictingcollege GPA than is high school GPA, especially at highlevels of ability (Noble & Sawyer, 2002). In datagathered at St. Norbert College, ACT composite scorescorrelated with final college GPA about.

5 50, and thecorrelation between ACT composite scores and highschool GPA was found to be about .55. (St. NorbertCollege, 2002).In general , tests of academic achievement correlatewith IQ scores about .50 (Brody, 1997; Petrill &Wilkerson, 2000). Several studies have explored therelationship between IQ and ACT scores specifically(Lewis & Johnson, 1985; Steinberg, Segel, & Levine,1967). These studies used relatively small samples andfound moderate to high correlations between verbal,performance, and full scale IQ and English, Mathe-matics, and Composite ACTscores. In addition, the ACTcomposite scores show gender effects, with malesscoring significantly higher than females (Mau &Lynn, 2001). This does not mean that the ACT is abiased (1987)used Item Response Theoryto analyze a sample of over 8000 individual scores on theACT Mathematics and English subtests and found nogender or race psychometric similarities between measures ofacademic achievement and measures of IQ are (2006)correlated scores on the SAT and ACTwith performance on three highlyg-loaded cognitivemeasures (college GPA, the Wonderlic Personnel Testand a word recall task).

6 Theg, or general , factor is acommon element among all tests of mental ability , thefirst shared factor that is extracted through factoranalysis. Coyle performed a factor analysis that showedhighg-loading for raw ACT and SAT scores, and theraw scores were significantly predictive of scores onmeasures of cognitive ability . Coyle also calculatedchange scores on the SAT and ACT (all subjects hadtaken the exams twice). Change scores did not correlatewithg, indicating that a change in score on a test ofacademic achievement does not represent a change , change scores may represent change in a groupfactor, such as memory or spatial ability . This isconsistent with research that shows that for tests ofcognitive ability test retest change scores are not relatedtog(Jensen, 1998, pp. 314 316).There is also considerable research on the relation-ship between IQ, academic achievement, and herit- ability .

7 It is well-established that the genetic influenceon IQ is significant. As an individual ages, there isevidence that the heritability of IQ increases, so thatenvironment accounts for less individual variance(Plomin, 1986). Twin studies show that levels ofheritability for academic achievement are only slightlylower than levels of heritability for IQ. For example, in asample of 91 adult male twin pairs,Lichtenstein andPederson (1997)found that the heritability of educa-tional attainment was .42. In a sample of 132 dizygotictwin pairs and 146 monozygotic twin pairs, aged 6 12,Thompson, Detterman, and Plomin (1991)found thegenetic contribution to academic achievement was , while the shared family environment effect was . achievement appears to follow the samepattern of heritability, with heritability increasing withage (see review inPetrill & Wilkerson, 2000).In a study of the Queensland Core Skills Test (QCST),Wainwright, Wright, Geffen, Luciano, and Martin(2005)investigated genetic and environmental contribu-tions to performance.

8 The QCST is a test of academicachievement given to students in the 12th year ofschooling. It includes writing, multiple choice, and shortresponse, and is designed to test reasoning and the abilityto integrate information. 326 dizygotic twin pairs and256 monozygotic twin pairs ranging from 15 to 22 yearswere administered the QCST and the MultidimensionalAptitude Battery (MAD), a measure of IQ. The adjustedheritability on the QCST was found to be .64. Acorrelation of .81 was found between MAD Verbal IQand QCST scores and .57 between MAD Math IQ andQCST scores. The authors also found that the geneticinfluences responsible for the heritability of IQ over-lapped almost completely with those responsible for theheritability of academic achievement. This is similar tothe findings ofThompson et al. (1991), who also foundthat genetic influences can best explain the covariancebetween cognitive ability and achievement.

9 AccordingtoWainwright et al. (2005), this finding makes intuitivesense. Tests of academic achievement in many respectsmeasure what a student has been exposed to andassimilated during his or her education. Many widely-used IQ tests include subtests (such as Vocabulary) thatdepend on knowledge an individual has been exposed tothrough culture. Indeed, in many cases Verbal IQcorrelates more highly with measures of academicachievement than Performance IQ (Thompson et al.,1991; Wainwright et al., 2005). Koenig et al. / Intelligence 36 (2008) 153 160As discussed inFrey and Detterman (2004), theability to predict IQ from widely used tests such as theSAT and ACT can increase the accuracy of estimates ofpre-morbid functioning in clinical populations. Clini-cians currently use a number of demographic variablesand current performance on psychological measures topredict pre-morbid functioning in individuals whosustain, for example, an injury causing brain damage(Baade & Schoenberg, 2004).

10 While demographicvariables alone can be useful in prediction, the additionof tests of current functioning increases predictionsubstantially (Axelrod, Vanderploeg, & Schinka, 1999).However, premorbid ACT scores may provide a moreefficient and more accurate means of a large number of students have taken theACT, the potential impact of a more accurate estimateof IQ is a review of existing research,Baade andSchoenberg (2004)looked at 15 studies of academicachievement and IQ. Their review finds a highcorrelation between a variety of achievement tests(including the ACT) and scores on the WAIS orWISC. The authors suggest the use of the predicted-difference algorithm to calculate IQ from test scores, butcaution that at the time of the review no large scaleresearch had looked at the relationship between many ofthe measures of academic achievement and IQ. Thevalidity of the predicted-difference method described inthe article depends on a high correlation between IQscores and measures of academic achievement, andconfirmation of the relationship is growing field of cognitive epidemiology wouldalso benefit from a widely-used test of cognitiveabilities.