This article was downloaded by: [Dalhousie University]On: 16 April 2013, At: 08:14Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Clinical and ExperimentalNeuropsychologyPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/ncen20

The Flynn effect and its relevance toneuropsychologyMerrill Hiscock aa University of Houston, Houston, TX, USAVersion of record first published: 12 Jun 2007.

To cite this article: Merrill Hiscock (2007): The Flynn effect and its relevance to neuropsychology, Journal of Clinicaland Experimental Neuropsychology, 29:5, 514-529

To link to this article: http://dx.doi.org/10.1080/13803390600813841

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial orsystematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distributionin any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that thecontents will be complete or accurate or up to date. The accuracy of any instructions, formulae, anddrug doses should be independently verified with primary sources. The publisher shall not be liable forany loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever causedarising directly or indirectly in connection with or arising out of the use of this material.

© 2007 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business

http://www.psypress.com/jcen DOI: 10.1080/13803390600813841

JOURNAL OF CLINICAL AND EXPERIMENTAL NEUROPSYCHOLOGY2007, 29 (5), 514–529

NCEN The Flynn effect and its relevance to neuropsychology

Flynn Effect in Neuropsychology Merrill Hiscock

University of Houston, Houston, TX, USA

Evidence from several nations indicates that performance on mental ability tests is rising from one generation tothe next, and that this “Flynn effect” has been operative for more than a century. No satisfactory explanation hasbeen found. Nevertheless, the phenomenon has important implications for clinical utilization of IQ tests. This art-icle summarizes the empirical basis of the Flynn effect, arguments about the nature of the skill that is increasing,and proposed explanations for the cause of the increase. Ramifications for clinical neuropsychology are discussed,and some of the broader implications for psychology and society are noted.

A sustained upward drift in mean cognitive abil-ity now has been documented in several countriesand is thought to have been occurring since theadvent of the industrial revolution (Neisser,1998b). This “rising curve” is often referred to asthe Flynn effect, after the New Zealand politicalscientist James R. Flynn who discovered and char-acterized the phenomenon (e.g., Flynn, 1984, 1987,1998a, 1999, 2006a). Implications of this phenome-non are numerous and profound, ranging fromfundamental questions about the respective role ofgenes and environments in determining a person’sintellectual capability to practical questionsabout the definition of mental retardation. TheFlynn effect has the potential to change manyaspects of neuropsychological theory and prac-tice. This paper discusses implications of theFlynn effect for understanding and measuringcognitive ability before addressing some specificimplications of the Flynn effect for clinical neu-ropsychology.

IQ tests in neuropsychology

IQ tests play a prominent role in clinical neuropsy-chology (Kaplan, Fein, Morris, & Delis, 1991).Diaz-Asper, Schretlen, and Pearlson (2004), forexample, regard intelligence testing as “a corner-stone of neuropsychological assessment” (p. 90).

Neuropsychologists’ reliance on IQ tests is para-doxical, as the tests were not developed for the pur-pose of identifying deficits produced by braindamage, nor are they especially well suited to thatobjective (Boake, 2002; Lezak, 1988; Lezak,Howieson, & Loring, 2004). The substantial corre-lation between IQ and neuropsychological test per-formance in normal populations complicates theinterpretation of low scores on neuropsychologicaltests in clinical settings (Diaz-Asper et al., 2004).Nonetheless, the results of IQ tests complementinformation about developmental, social, educa-tional, and occupational history in providing acomprehensive portrayal of the patient. At the veryleast, the IQ test helps the clinician to estimate pre-morbid cognitive functioning, to formulate expec-tations of performance on other tests, and todetermine the level of discourse at which to engagethe patient. Multifaceted IQ tests such as theWechsler and Stanford–Binet tests also provide acognitive profile, which informs the clinician aboutrelative strengths and weaknesses and may consti-tute a starting point from which to search for morespecific deficits.

Normative data

Irrespective of how the IQ test is employed by theclinician, its usefulness depends on the adequacy of

The author thanks James R. Flynn for his helpful comments and updated information.Address correspondence to Merrill Hiscock, Department of Psychology, University of Houston, Houston, TX 77204–5022, USA

(E-mail: mhiscock@uh.edu).

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 515

normative data. Even profile analysis, whichentails within-individual (ipsative) comparisons,will be corrupted if the norms for one or moresubtests are inadequate or inappropriate for theparticular patient population being served. Age-stratified and up-to-date norms are an importantasset that enhances the usefulness of commerciallyavailable IQ tests. If norms are inaccurate or inap-propriate, results will be difficult to interpret if notmisleading.

With respect to IQ tests as well as other psycho-logical tests, normative problems arise underthree circumstances. One circumstance, relativelycommon in neuropsychology, involves normativedata that may be inaccurate even for the popula-tion from which the data were obtained. This mayoccur when the sample size is inadequate, whenthe test materials are not standardized, or whenthe test is administered in a manner that deviatesfrom the procedures used in clinical evaluation.Lezak et al. (2004) and Spreen and Strauss (1998)provide several examples of these sources of vari-ability in normative neuropsychological test data.If the norms do not represent accurately thepopulation on which the norms are based, theyare even less likely to be appropriate for the clini-cal population to which the norms are beingapplied. A second variety of normative problemsis observed when the normative data are repre-sentative of the normative population but thenormative population is discrepant in some signi-ficant way from the population to which thenorms are to be applied. The respective popula-tions may differ in age, education, socioeconomiclevel, or general health status. In the UnitedStates, for example, norms based on patients at aVeterans Administration Medical Center may notreflect characteristics of the general population(e.g., Burke, 1985). The third category of norma-tive problems pertains to norms that have becomeold. This is the problem to be considered in thepresent paper.

Currency of norms is generally regarded asa favorable attribute, but seldom does one findexplicit justification for this belief. If a test’s con-tent is appropriate for use with a contemporarypopulation, then why should it matter whethertest norms were compiled 5 years ago or 50 yearsago? It seems that there are two reasons for mis-trusting older norms, one of which is rather obvi-ous. The obvious reason is that tests sometimesare revised with respect to content, administra-tion, or scoring rules. Periodic revisions of theWechsler and Stanford–Binet IQ tests, forexample, are accompanied by updated norms.The less obvious reason for preferring recent

norms is that the distribution of the ability meas-ured by the test may change over time.

THE FLYNN EFFECT

IQ gains, when considered across various testsgiven to children and adults in different countries,are roughly twice as large for “culture-reduced”tests as for tests with learned content (i.e., verbaltests). In studies of data from 20 different nations,Flynn (1987, 1994) found that scores on Raven’sProgressive Matrices Test (RPM) increased about0.6 IQ points per year between 1930 and 1990. Thisconclusion was supported by more intensive analy-ses of data from the United Kingdom and fromfour countries—the Netherlands, Israel, Norway,and Belgium—in which the Raven’s test had beenadministered to huge military samples (Flynn,1987, 1998b, 1999).

Although the rate at which Wechsler andStanford–Binet IQ has increased is not as dramaticas the rate at which scores on Raven’s test haveincreased, the Flynn effect does apply to theWechsler and Stanford–Binet tests, as well as testssuch as the Otis Intermediate Test of Mental Abil-ity and the California Test of Mental Maturity(Flynn, 1987). In fact, Flynn’s discovery of IQgains over time began with his observation, fromtables in the Wechsler Intelligence Scale forChildren–Revised (WISC-R) manual, that childrenwho completed both the WISC-R and the olderWISC scored an average of 8 points higher on theWISC. Thus, the average IQ for the respectivestandardization samples must have increased dur-ing the interval between the norming of the WISCin 1947–1948 and the norming of the WISC-R in1971–1973. In Flynn’s (1999) words, “Clearly fromone sample to the other, over a period of 25 years,Americans had gained 8 IQ points” (p. 6). Figure 1shows typical mean increases over time for scoresfrom the Wechsler and Stanford–Binet tests andfor scores from the RPM.

The gain of 8 IQ points per 25 years, or about0.3 points per year, was confirmed by examinationof Wechsler and Stanford–Binet results forapproximately 7,500 individuals between the agesof 2 and 48 years who had completed the originaland revised versions of the same test (Flynn, 1984,1987). Subsequently, after examining differencesbetween WISC-R and Wechsler Intelligence Scalefor Children–Third Edition (WISC-III) scores, andbetween Wechsler Adult Intelligence Scale–Revised (WAIS–R) and Wechsler Adult Intelli-gence Scale–Third Edition (WAIS-III) scores,Flynn (1998c) noted a discrepancy between the

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

516 HISCOCK

annual IQ gain as inferred from the children’s tests(0.312 points) and that from the adult tests (0.171IQ points). The source of the discrepancy hasbeen identified by more recent analyses, in whichWAIS-III normative data were compared withnormative data from older as well as newer IQ tests(Flynn, 2006b). Those analyses indicate that IQsderived from the WAIS-III are inflated by 2.34points. Consequently, the Flynn effect is underesti-mated when the WAIS-III is compared with oldertests and overestimated when the WAIS-III is com-pared with more recent tests. Flynn (2006b) con-cludes that Wechsler and Stanford–Binet IQcontinues to rise at the rate of 0.3 points per year.He recommends subtracting 2.34 points from IQsbased on the WAIS-III to correct for its tendencyto yield “inflated IQs even in the year in which itwas normed” (p. 179).

The gains manifested on Wechsler IQ tests arenot equally distributed across subtests. In keepingwith the principle that increases for culture-reduced tests are greater than increases for tests oflearned content, Performance IQ has risen moredramatically than Verbal IQ. Flynn (1998c) hasused scores from about 1,000 children who tookthe WISC-R and WISC-III to calculate a 0.358 IQpoints-per-year increase in Performance IQ and a0.218 points-per-year increase in Verbal IQ.According to Flynn (1999), gains on the more cul-turally influenced Wechsler subtests such as Arith-metic, Information, and Vocabulary are “small ornil” in English-speaking countries. Likewise, gainson achievement tests are insignificant or nonexistent.

The Similarities subtest from the Wechsler IQtests stands as an exception to the rule that cultur-ally influenced tests are relatively invulnerable tothe Flynn effect. Gains on the Similarities subtest

are as large as those observed for Raven’s RPM(Flynn, 1999). As pointed out by Flynn, the rapidincrease in Similarities scores over time is informa-tive because the test has a verbal format and doesnot appear to demand spatial reasoning. The abil-ity that is increasing most dramatically thusappears to be broader than spatial ability or someother specific manifestation of nonverbal intelligence.

Characteristics of Raven’s test

As noted previously, the most striking evidence forrising IQ comes not from the Wechsler orStanford–Binet tests, but from Raven’s ProgressiveMatrices Test (RPM), an untimed measure ofintellectual ability that was first published by JohnC. Raven in 1938. The test requires no reading orspeaking and only simple written responses. It iseasily administered, it has satisfactory reliability,and the results correlate moderately well withscores from other tests that are purported to meas-ure general intellectual ability (Burke, 1985;Llabre, 1984; Raven, 1960; Spreen & Strauss,1998). The RPM is considered to be a culture-freetest (Jensen, 1980), or at least a culture-reducedtest, for which there are vast amounts of data fromseveral Western nations (Flynn, 1999). Althoughthe most widely used version of the RPM is theStandard Progressive Matrices, the easier ColoredProgressive Matrices and the more difficultAdvanced Progressive Matrices are available for usewith low- and high-functioning groups, respectively.

Raven’s Matrices test does not appear to be usedfrequently by American neuropsychologists. None-theless, it is an especially attractive test for theo-rists who posit the existence of a generalintellectual factor such as Spearman’s g (e.g.,Jensen, 1998; Spearman, 1904). In the words ofNeisser (1998a), “Whatever g may be, we at leastknow how to measure it. The accepted best meas-ure, which has played a central role in analyses ofthe world-wide rise in test scores, is the Raven Pro-gressive Matrices” (p. 9). Among those theoristswho endorse Cattell’s (1957) decomposition of ginto fluid intelligence (ability to solve novel prob-lems) and crystallized intelligence (domain-specificknowledge acquired over time), Raven’s Matricestest is also regarded as the quintessential measure offluid intelligence. Snow, Kyllonen, and Marshalek(1984), who conducted a multidimensional scalinganalysis of scores from various cognitive abilitytests, found that Raven’s test occupied the territoryat the very center of their model. In other words,the RPM was the best measure of the domain-inde-pendent abilities that are required for various

Figure 1. Smoothed curves depicting the differential increase inmean scores on Wechsler, Stanford–Binet, and Raven’s Matri-ces tests during the middle and latter parts of the twentieth cen-tury. Adapted from Neisser (1998a). Copyright © 1998 by theAmerican Psychological Association. Adapted with permission.

Score Increases on Raven's Matrices Compared with Those on Wechsler and Stanford-Binet Tests

100

105

110

115

120

125

130

1932 1942 1952 1962 1972 1982 1992 2002

Year Data Were Collected

ecnerefeR ot e vita le

R ser oc S Q I

Wechsler and Stanford-Binet

Raven's Matrices

o

o

o oo

x

x

x

x

o x

o

x

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 517

kinds of figural, verbal, and numerical problemsolving.

What is rising?

Three aspects of the Flynn effect provide poten-tially important clues as to the identity of the cog-nitive characteristic that is increasing over time.One clue is the constancy of the slope across time.Since 1932 and probably prior to that, test scoreshave been increasing at a rate of 3 to 6 IQ pointsper decade, depending on the IQ test used. Thepreponderance of evidence indicates that scores arecontinuing to rise at a constant rate (Flynn,2006b). Any characterization of the cognitive abil-ity that is increasing must account for the apparentinexorability of the increase. There is at least oneexception, however. The Scandinavian countriescurrently are showing little or no rise in their testscores (Flynn, 2006a). As large IQ increases wereseen in Norway prior to 1968, Flynn suggests thatScandinavia might have experienced earlyincreases that have since abated. This raises thepossibility that IQ increases in other industrializednations will also end.

A second clue consists of evidence that IQ gainshave been comparable for young children, olderchildren, and adults (Flynn, 1984) and for variousnations, ethnic and linguistic groups, and geo-graphic regions (Flynn, 1999, 2006a). The IQincrease can be seen in children who have not yetbeen exposed to formal education and who, beingyoung children, have had limited exposure to otheraspects of the culture into which they were born.The increase has been documented in developedand developing nations. It has been reported inEurope, North American, South America, Asia,and Africa.

The third clue, which has been discussed above,consists of findings that the scores on culture-reduced tests, or tests of fluid intelligence, show anincrease twice as large as that observed for tests oflearned information, or tests of crystallized intelli-gence. The increase represents largely an enhance-ment of people’s ability to solve certain kinds ofproblems rather than their acquisition of moreinformation from the culture in which they live.

Attempts to describe the cognitive skill that isincreasing over time fall into three categories.Some authors accept the rising scores at face valueand assume that they reflect actual increases ingeneral intelligence or in the ability to adapt to thenew cognitive demands of a changing culture(Greenfield, 1998; Martorell, 1998; Sigman &Whaley, 1998; Williams, 1998). For example,

Greenfield (1998) suggests that the increasing testscores indicate a rise in “culturally phenotypicintelligence,” by which she means an adaptation tothe increasing importance of visual electronicmedia and the decreasing emphasis on traditionalprint media. This characterization accords wellwith the greater rise of nonverbal IQ than verbalIQ, although it fails to explain why verbal IQshould rise at all. Greenfield’s argument does notaccount for the ubiquity of the increase across indi-viduals from different cultures, socioeconomic lev-els, and educational backgrounds unless oneassumes that environmental shifts are uniformacross diverse groups. It is also difficult to recon-cile Greenfield’s characterization of the cognitiveskill with the absence of age differences in the IQincrease,and even more difficult to reconcile it withthe long period over which scores have been rising.As Flynn (1999) has observed, “Ravens gains werelarge before the television era, much less before thecomputer-game era” (p. 9).

Lynn (1998) has proposed a modified version ofthe position that IQ gains reflect actual increases inintelligence. According to Lynn, only the 3-point-per-decade increase in IQ, as observed in theWechsler data, reflects a genuine rise in intelli-gence. The larger increases on other tests, such asthose shown on Raven’s Matrices by northernEuropean and Israeli military conscripts, areattributed to mathematics education. Lynninvokes the experimental study of Carpenter, Just,and Shell (1990) to argue that Raven’s test requiresthe application of “the mathematical principles ofaddition, subtraction, progression, and the distri-bution of values” (Lynn, 1998, p. 212). If increas-ing proportions of adolescents from northernEurope and Israel remain in school and becomemore proficient with mathematics than were earliercohorts, then, according to Lynn, this additionallearning may account for the size of the increase inRaven’s performance.

Insofar as Lynn’s explanation concerns degreesof IQ increases, rather than the presence orabsence of increases, the quality of the extantevidence is inadequate for evaluating the explana-tion. The magnitude of the Flynn effect varies toowidely across studies. For example, Flynn (1987)found that the rise in Raven’s scores in differentstudies of children between 8 and 16 years rangedfrom 0.19 to 1.25 IQ points per year. Given themarked variability across samples, one cannot con-clude that the gain for children is smaller than thegain for older individuals, who are more likely tohave benefited from improvements in mathematicseducation. Even if a reliable difference betweenchildren and adults were to be established, the

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

518 HISCOCK

Colored Progressive Matrices test used withyounger children may not be comparable in itscognitive demands to the Standard ProgressiveMatrices test used with adolescents and youngadults. The Colored Matrices test is even less likelyto be comparable to the Advanced Matrices testthat was examined by Carpenter et al. (1990). Thedegree to which mathematical reasoning is requiredby the different forms of Raven’s test is difficult tospecify, but it does not appear to be a salientdemand of either the Colored or Standard test(Carlson & Jensen, 1980; Costa, 1976; Villardita,1985; Waltz et al., 1999). Also, Lynn’s account failsto explain why the magnitude of the Flynn effect isso large for the Wechsler Similarities subtest.

Flynn, in the 1980s and 1990s, articulated a thirdpoint of view regarding the nature of the abilitythat is being indexed by the rising test scores. Spe-cifically, he asserted that the IQ gains have no rela-tion to any form of real-world intelligence. In his1987 article in Psychological Bulletin, Flynn used theterm “abstract problem-solving ability (APSA)” todescribe the characteristic that is increasing overtime. According to Flynn, APSA is the ability tosolve problems on an IQ test, and this ability canbe highly divergent from “the real-world problem-solving ability called intelligence” (Flynn, 1987, p.188). Flynn (1999) distinguished among (a) IQ testscores, (b) psychometric g, and (c) real-world intel-ligence. Real-world intelligence can be inferredfrom psychometric g, which in turn can be inferredfrom scores on IQ tests. The inferential chain canbe represented as follows:

Flynn accepted the validity of inferring psychomet-ric g from IQ tests (which, he acknowledged,measure g and not much else) but he disputed thesecond inference. According to Flynn, neither aca-demic achievement-test results nor “real-worldachievements” support the conclusion that humansare becoming more intelligent.

If the cognitive ability that increases over time isas narrow as Flynn (1998a, 1999) initially con-strued it to be, then the relationship between testscores and their meaning is isomorphic. Whendivorced from real-world manifestations, intelli-gence becomes only “what the tests test” (Boring,1923). The unknown ability that is rising is psycho-metric intelligence, and psychometric intelligence isdefined as that which IQ tests measure. Given thistautology, any change in test scores reflects noth-

ing more than a change in the ability to performthe tasks required by the test. The problem withthis interpretation is its incompatibility witha wealth of evidence that connects psychometric gwith real-world criteria. Intelligence theorists whoagree on little else agree that, under the right cir-cumstances, IQ test scores do predict educationaland occupational outcomes (cf. Herrnstein & Murray,1994; Sternberg, Grigorenko, & Bundy, 2001).

Flynn’s new model of IQ rise

Flynn’s original interpretation of the rising-curvephenomenon leaves us with two strikingly diver-gent characterizations of IQ disparities. On the onehand, population-wide increases in IQ are devoidof real-world significance and are environmentallydetermined. (There is no reason to believe that dif-ferential reproduction rates could account for therapid rise of scores from one generation to thenext; Flynn, 1994.) On the other hand, individualdifferences in IQ within a cohort are known to beassociated with people’s real-world attainmentsand seem to be influenced substantially by geneticinheritance (e.g., Bouchard, Lykken, McGue,Segal, & Tellegen, 1990; Herrnstein & Murray,1994; Jensen, 1973; Plomin, 1990). How can IQassume both guises? This paradox requires yetanother interpretation of the Flynn effect, an inter-pretation that reconciles the intergenerational risein IQ with the individual differences in IQ that arefound within cohorts.

The new interpretation of the Flynn effect(Flynn, 2003, 2006a) rests on a quantitative modelof the genesis of IQ differences that has been for-mulated by Dickens and Flynn (2001a, 2001b).The Dickens–Flynn argument begins with a reas-sessment of evidence that has been used to estab-lish the heritability of IQ. While acknowledgingthat genes contribute to IQ patterns obtained fromtwin studies and adoption studies, Dickens andFlynn (2001b) assert that genetic differencesamong individuals are amplified by correlationsbetween genes and environment. For instance,a young child who demonstrates an aptitude fora particular activity may be encouraged by parentsand teachers to pursue that activity. Additionalresources will be made available to the child, andthe child’s proficiency will increase. The increasedproficiency will lead to successes that, in turn, willincrease the child’s interest in the activity and themotivation to improve his or her knowledge andskill. Dickens and Flynn refer to this kind ofmatching of genetic characteristic and environ-ment, and the positive feedback loop that ensues,

IQ test scoresreal - world intelligence

→→

psychometric g

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 519

as a “multiplier effect.” Similar arguments havebeen made by others (e.g., Bell, 1968; Jensen, 1975;Plomin, 1986; Scarr, 1992). However, Dickens andFlynn use the principle of gene–environment corre-lation for a specific purpose—that is, to show that,in behavior genetics studies, heredity is credited forindividual differences produced by this interactiveprocess when, in fact, only a portion of the differ-ences is the direct consequence of genetic endow-ment. Thus, the true heritability of IQ is not asgreat as that reflected in heritability indices, andthe conflict between Flynn effect and findings frombehavior genetics is reduced accordingly.

The Dickens and Flynn (2001b) model alsoincorporates a new means of determining the trueimpact of environmental variables that are not cor-related with individuals’ intrinsic characteristics.These “exogenous environmental effects” can bedivided into transient and persistent components.The authors argue that, at any specific time, thevariance explained by a persistent exogenous effectmay be weak relative to the variance attributableto transient factors. However, the same persistentfactor will appear to be much stronger when it iscompared with the average of all the exogenousvariables that have been present at different times.In the words of Dickens and Flynn, “this collectiveaveraging further diminishes the importance of ran-dom individual environment influences, whereasconsistent factors acquire an impact beyond whatwe would expect viewing the individual in isola-tion” (p. 351).

The next task for the Dickens–Flynn model is toaccount for the large magnitude of IQ increasesacross generations, which appears to be muchgreater than the magnitude of the individual differ-ences produced by gene–environment correlations.To explain the size of transgenerational IQincreases, Dickens and Flynn (2001b) and Flynn(2003, 2006a) emphasize the distinction betweensocial multipliers (exogenous environmentaleffects) and individual multipliers (gene–environmentcorrelations). Whereas individual multipliers influ-ence an individual’s position in a societal peckingorder, those environmental factors have little directeffect but instead exert their influence throughinteraction with the individual’s inherited abilities.Social multipliers, in contrast, have a pervasiveeffect on an entire cohort. To borrow one ofDickens and Flynn’s most effective examples, rein-forcement for basketball accomplishments (anindividual multiplier) accrues almost entirely tothose individuals who have high levels of aptitudefor the game. Unusually tall or talented individualswill capitalize on their own individual multipliersand rise to the top of the basketball hierarchy.

However, the markedly increased popularity ofbasketball, presumably attributable to the wide-spread televising of professional games, constitutesa social multiplier. As an entire culture becomesprogressively more interested in basketball, manymore children play the game, and the general levelof skill, coaching, and competition increases dra-matically. The distribution of basketball perform-ance is shifted upward, and, over a sufficient timespan, the magnitude of that shift will be muchgreater than the effect that individual multipliershave on elite players. Gould (1996) has reacheda similar conclusion from his analysis of major-league baseball statistics. As the overall skill levelrises, the variability at the right tail of the distribu-tion is reduced, and the gap between the mostextraordinary batters and “ordinary” batters isreduced accordingly. To use Dickens and Flynn’sterminology, the baseball statistics seem to showthat the influence of individual multipliers on themost elite players is constrained by biological lim-its that are less likely to constrain the influence ofsocial multipliers on other players.

Athletics may provide the clearest examples ofsocial multipliers, but Flynn (2003, 2006a) appliesthe same reasoning to intellectual, academic, occu-pational, and avocational realms. According toFlynn’s new theory, the IQ increase does reflect anactual increase in intellectual prowess, and thecausal factors are environmental. Nevertheless,this increase in IQ can be reconciled with evidencethat individual differences in IQ predict real-worldoutcomes as well as evidence that the individualdifferences are determined in part by genetic inher-itance. Flynn’s specific account of the environmen-tal factors that cause IQ to rise are summarized atthe end of the following section.

Why are the scores rising?

Several writers have offered opinions and claimsregarding variables that might be causing a sus-tained international rise in IQ. Often the putativecause depends on the writer’s perception of thatwhich is changing. Thus Greenfield (1998), whointerprets the Flynn effect as in increment in visu-ospatial ability, argues that the effect stems froman emerging cultural shift away from traditionalverbal communication media and toward new vis-ual and interactive media. Writers who equate therise in test scores with an increase in general intelli-gence are more likely to favor causal explanationssuch as improved nutrition (Lynn, 1990, 1998;Sigman & Whaley, 1998) or greater exposure toformal education (Williams, 1998). As already

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

520 HISCOCK

noted, Lynn (1998) regards part of the increase inRaven’s scores as a consequence of more extensiveeducation in mathematics. He attributes theremainder of the Flynn effect—the “genuine” IQincrease—to nutritional improvements. However,as Flynn (2003) has pointed out, improved nutri-tion should benefit primarily those individuals whoare most deprived. Consequently, if nutritionalimprovement were the primary cause of IQincreases, there should be disproportionate gainsat low IQ levels, and this does not seem to be thecase, at least in the United States.

Williams (1998) has mentioned a large numberof other environmental changes that might havehad a favorable effect on the intelligence of chil-dren in various countries. Her list of possiblecauses includes: (a) a greater emphasis on proce-dural knowledge than on declarative knowledge;(b) teaching to the test; (c) improved health; (d)smaller families; (e) better educated parents; (f)changes in parenting style; (g) a trend towardurbanization; and (h) changing patterns of stresson pregnant women and their babies. When weadd these eight variables to the previously men-tioned factors—better nutrition, more formal edu-cation, and a cultural shift from verbal tononverbal media—we have a large choice of expla-nations for the Flynn effect. Some explanations aremore plausible than others. Some are better able toaccount for the long history of the rising curve,whereas others are better able to account for itspervasiveness across nations in recent decades. Theexplanations are not mutually exclusive; two ormore may be partially correct. Yet it is possiblethat none of the explanations is correct. This is theposition originally taken by Flynn: “I believe it isfair to say that up to now, efforts to identify theenvironmental factors that have caused IQ gainshave not come to much” (Flynn, 1998a, p. 49).

Building on the work of Dickens and Flynn(2001a, 2001b), Flynn recently has constructeda framework for explaining the rising test scores.Flynn (2006a) attributes the rise in IQ (especially inthe United States during the first half of the twentiethcentury) primarily to increases in the number ofyears of formal education. He notes that the aver-age number of years of public education in theUnited States rose from 8 to 10 between WorldWar I and World War II (Tuddenham, 1948). For-mal education exemplifies the power of a socialmultiplier. As students spend more years in school,“each student is surrounded by fellow studentswho are more competent, better students makebetter teachers for the next generation of students,parents become more serious about schooling andhomework, the lengths of the school day and

school year tend to increase” (p. 15). The transgen-erational impact of increased education presuma-bly explains why the Flynn effect is apparent inyoung children.

IQ gains since World War II, according to Flynn(2006a), can be attributed to a shift of emphasisfrom reading, writing, arithmetic, and other “disci-plined” learning to “on-the-spot problem-solvingskills.” This educational shift seems to be associ-ated with several demographic trends, such asgreater urbanization and affluence, decreasingfamily size, changes in the kinds of work thatpeople do, and the increasing importance of leisureactivities. Perhaps many contemporary workplacesrequire more on-the-spot problem solving (fluidintelligence) than did workplaces of the past, andperhaps some contemporary leisure activities (e.g.,playing video games, playing poker via the Inter-net, and using computer software) require morefluid intelligence than did leisure activities of thepast. As occupational and avocational activitiesevolve, and as educational systems adjust to thosesocietal changes, successive generations of childrenwill continue to perform better on tests of abstractproblem solving while demonstrating little or noimprovement in traditional academic skills. Alter-natively, societal emphasis may turn in a differentdirection. In either case, Flynn would invoke theconcept of the social multiplier to explain a rapidincrease in the standard of performance withinselected intellectual realms.

IMPLICATIONS FOR CLINICAL NEUROPSYCHOLOGY

Certain practical implications of the Flynn effectare especially relevant to clinical neuropsychology,and those are the implications that are emphasizedhere. It should be noted, however, that Flynn’sconcept of social multipliers resembles the venera-ble neuropsychological principle of dissociation(e.g., Luria, 1973). Flynn (2006) has described ananalogy between cultural variables, which maycause certain cognitive skills to be developed pref-erentially, and brain damage, which may selec-tively impair some cognitive functions withoutaffecting other functions. Both kinds of dissocia-tion—societal and pathological—are means offractionating the global pattern of general intelli-gence. Thus, in principle, findings from clinicalneuropsychology and cognitive neuroscience canbe used to guide hypotheses about how normalbrains might be expected to change as social multi-pliers reinforce particular cognitive skills whileneglecting other skills.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 521

Old tests and old norms

The most obvious implication of rising test normsfor the clinical neuropsychologist is the risk ofbeing misled by old tests and old norms. Yet, in thepast, when a new test became available, clinicianstypically were more worried about being misled bythe new test and the new norms. When the WAIS–Rwas introduced in 1981, clinical neuropsycholo-gists recognized that this test, with its new norms,yielded lower scores than did the 1955 WAIS(Bornstein, 1987; Chelune, Eversole, Kane, &Talbott, 1987; Reitan & Wolfson, 1990). Some ofthese authors acknowledged that the discrepancywas partly attributable to a population-wide rise inIQ, but they were more concerned about differ-ences between the WAIS and WAIS–R that mightalter subtest profiles. As neuropsychologists, theirprimary focus was on the neuropsychologicalvalidity of pattern analysis. Readers were cau-tioned that decision rules based on the older testmight not be applicable to the newer test.

Transitional problems notwithstanding, theFlynn effect constitutes a compelling reason toadopt new tests and new norms. Nearly 60 yearshave passed since norms for the original WISCwere obtained in 1947 and 1948. The WAIS waspublished more than 50 years ago, and theWechsler–Bellevue tests were published more than60 years ago. If Full Scale IQ has been increasingby 3 points per decade, then the results from thoseold tests and norms would be inflated by 15 to 18points. If the earlier tests continue to be used rou-tinely for testing, and if the norms are not adjustedfor the Flynn effect, the average IQ for the generalpopulation will be 115 to 118 instead of 100. Apatient with a tested IQ of 100, obtained currentlyfrom any of the early tests, would be expected tohave an IQ of 82–85 if tested on a current WechslerIQ test. This would be the case despite the appro-priate age adjustments having been made in thecalculation of IQ.

This is an extreme example, but the same prob-lem will be experienced to some degree whenevera test with old norms is used to measure IQ or toestimate premorbid IQ. Hiscock, Inch, andGleason (2002) describe an instance in whichawareness of the Flynn effect led to a dramaticchange in the way Raven’s Matrices scores wereinterpreted. In the 1980s, before the Flynn effectwas widely known, RPM raw scores were con-verted to IQs or z-scores on the basis of Raven’snorms from 1960 or from norms published by Peckin 1970 and reprinted in Lezak (1983). Irrespectiveof which norms were used, Hiscock et al. foundthat patients with traumatic brain injuries tended

to score above the population average on theRPM. It appeared at the time that Raven’s Matricestest was singularly insensitive to the effects of headtrauma and that RPM scores might even be usefulfor estimating premorbid intellectual ability. Sub-sequently, the availability of more recent norms(J. Raven, Raven, & Court, 1995) allowed theinvestigators to make retrograde as well as antero-grade adjustments for the Flynn effect. Using anadjustment of 0.6 IQ points per year, Hiscock et al.found almost perfect congruity between the effectsof retrograde and anterograde corrections. Afteradjustment for the Flynn effect, patients’ RPMscores were commensurate with their WAIS–R IQ.It was concluded that the RPM is neither more norless sensitive than the WAIS–R to the effects oftraumatic brain injury.

Researchers who track their participants psy-chometrically in long-term longitudinal studiesoften face a difficult dilemma. Does one choose atest and stay with that test for the duration of thestudy even though the test and its norms have beensuperseded? Or does one switch to the newer testand thereby jeopardize the comparability of earlyand late assessments? Similar obsolescence prob-lems arise in clinical practice when an IQ test withsuperseded norms is retained in a battery of tests inorder to preserve the validity of the interpretiverules (Reitan & Wolfson, 1985) or when an old ver-sion of an IQ test is administered to a patient whohas taken the same test in the past. In bothinstances, the IQ derived from the old test willoverestimate the patient’s actual IQ. On the otherhand, switching to a newer version of the test maynecessitate comparing scores based on differentsets of items.

It may be counterintuitive to avoid retesting apatient with an IQ test that had been administeredseveral years ago. The clinician may reason that, ifthe test had been appropriate for the patient 20years ago, it should still be appropriate so long asthe necessary age adjustments are made. The Flynneffect does not imply that the test has become inap-propriate. It implies that the norms have becomeinappropriate. The WAIS norms for 60-year-olds,for example, apply to individuals who wereapproximately 60 years old when the test wasnormed and not to individuals who since havebecome 60 years old. Members of that earliercohort of 60-year-olds, if alive today, would bemore than 110 years old. Unless norms for the oldIQ test can be collected from contemporary 60-year-olds, the IQ obtained from the WAIS will notbe comparable to the IQ obtained from the WAISin the years immediately following its publicationin 1955. This argument, of course, is based on

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

522 HISCOCK

psychometric considerations and does not precludereadministering the old test for the purpose of mak-ing item-by-item comparisons. It may be instructive,for instance, to note differences between present andpast performance on Block Design or Vocabularysubtests as indicated by the respective raw scoresand by the specific items that were passed andfailed.

Admittedly, a revised IQ test may have beenaltered in ways that compromise the clinical inter-pretation of patterns of scores across subtests. Thisis precisely the problem that had caused consterna-tion after release of the WAIS–R (Bornstein, 1987;Chelune et al., 1987; Reitan & Wolfson, 1990). Onthe other hand, analysis of patterns from a previ-ous version of the test will also be compromised ifthe Flynn effect has produced differential increasesin the various subtest scores. For instance, the so-called ACID profile from the children’s WechslerIQ test (relatively low scaled scores on Arithmetic,Coding, Information, and Digit Span) is com-monly interpreted as an indicator of poor atten-tion, impaired auditory short-term memory, andlearning difficulties (e.g., Rourke, Bakker, Fisk, &Strang, 1983). Although that interpretation mighthave been valid with respect to the 1949 WISC dur-ing the 1950s, the same pattern would require a dif-ferent interpretation today because some of thesubtest scaled scores would have been affectedmore than others by nearly 60 years of exposure tothe Flynn effect (Flynn & Weiss, 2006). Continuingto use an old test and its old norms is not an effect-ive means of preserving the validity of clinicaldecision rules.

Age-related change

Another important implication of the Flynn effect,which may be less obvious than the first, concernsage norms. Clinical neuropsychologists who workwith adult clients are accustomed to norms thatindicate age-related change in performance, andclinicians know that some tests are especially sensi-tive to the effects of aging. The Wechsler DigitSymbol subtest is one example. On the WAIS–R,the average raw score for Digit Symbol decreasessubstantially between the ages of 25 and 65 years.A raw score of 58 is equivalent to a scaled score of10 (average) if the examinee is 25 years old. How-ever, if the examinee is 65 years old, the same rawscore is equivalent to a scaled score of 15, which is1.67 standard deviations above average.

The norms for different age categories areobtained using a cross-sectional (rather than longi-tudinal) methodology. Different cohorts are

selected to represent 25-year-olds and 65-year-olds.Consequently, the performance differences betweenage groups on Digit Symbol may be confounded bycohort effects, or differences between the groupsother than age per se (Cozby, 2003). The Flynneffect is based on birth date, not age, and thus is apotential confounding variable. Although thereremains some ambiguity about the strength of theFlynn effect in different cohorts, the preponder-ance of evidence indicates that scores are rising bysimilar amounts in all age groups (Flynn, 1984,1998a; Hanson, Smith, & Hume, 1985; Lynn,1998). Therefore, for the purpose of discussing agenorms, the Flynn effect is assumed to be independ-ent of age. In other words, the following examplesare based on the assumption that age effects andthe Flynn effect are additive.

Given that age and cohort effects are additive,one effect can be disentangled from the other usinglogic similar to that applied by Schaie (1994) tocross-sectional data collected as part of the SeattleLongitudinal Study. Having found divergentcohort gradients for different indices of mentalabilities, Schaie recognized that his cross-sectionalpatterns either underestimated or overestimatedthe actual age-related changes in those abilities.Similarly, if one knows the magnitude of the Flynneffect with respect to a specific Wechsler subtest,then one can subtract the Flynn-effect componentfrom the age-related decrement observed in thecross-sectional norms. The best estimate of trueage-related deterioration of performance is the tab-ulated age-related change minus the change attrib-uted to the Flynn effect. These calculations, inother words, provide an estimate of the degree towhich normative age-related changes in test per-formance are contaminated by cohort effects.

WAIS–R Digit Symbol norms can be used toillustrate the procedure. As mentioned previously,comparison of norms for 25-year-olds and 65-year-olds indicates that performance falling at the aver-age (z=0) for the younger cohort would fall wellabove average (z=+1.67) for the older cohort. Inother words, mean performance of the two agegroups differs by 1.67 standard deviations, whichreflects a decline of about 0.4 standard deviationsper decade. But a difference of 0.20 standard devia-tions (3 IQ points) per decade would be expectedon the basis of the Flynn effect alone. That is,about half of the apparent age-related deteriora-tion in normative Digit Symbol performance canbe attributed to the Flynn effect. If Digit Symbolperformance is more susceptible to the Flynn effectthan are most other WAIS–R subtests—whichseems plausible, given that the Flynn effect isstrongest for culture-reduced tests—then more

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 523

than half of the apparent age-related decline couldbe attributed to the Flynn effect. Dickinson andHiscock (2005) have undertaken additional analysesin which the magnitude of the Flynn effect for eachsubtest is calculated individually from data publishedin the WAIS-III administration manual (Wechsler,1997). These calculations indicate that 60% of thenormative age-related decline in Digit Symbol per-formance is attributable to the Flynn effect.

The previous example relies on an indirectmethod of decomposing normative differencesbetween age groups into age-related and birth-date-related sources. A more direct method can beapplied to a set of Raven’s Matrices data in whichraw scores for 10 cohorts are available (Flynn,1998a). A graph of the data is shown in Figure 2.The curve depicts the median raw scores for 10 Brit-ish cohorts with birth dates ranging from 1877 to1967. The five points on the left represent five agegroups (25, 35, 45, 55, and 65 years) that were testedin 1942, and the five points on the right represent fivegroups with the same mean ages that were tested in1992. These data allow two kinds of analysis. Com-parisons among different age groups tested duringthe same year reflect the additive effects of age and

birth date. The data also allow comparisons betweengroups of the same age that were tested 50 yearsapart. The latter comparisons isolate the effects ofbirth date in the absence of age differences. The dataand calculations are shown in Table 1.

In Table 1, the 10 data points from Figure 2(from left to right) are denoted as Groups A to J.Groups A to E were tested in 1942, and Groups Fto J were tested in 1992. The average annual differ-ence between age groups tested in the same yearcan be used to estimate the combined effects ofbirth date and age. For the 1942 data, the averagedifference between groups (Average 1 in the table)was 0.51 points/year. Average 3, shown at the bot-tom of Table 1, represents the mean differencebetween each age group in 1942 and the corre-sponding age group in 1992. The average increaseis 0.36 points/year. This constitutes an estimate ofthe magnitude of the Flynn effect alone for the50-year interval from 1942 to 1992. Consequently,the proportion of the age-related decline that canbe attributed to the Flynn effect is 0.36 divided by0.51, or 71%. Consequently, if the 1942 mediansare used as age-group norms, no more than 29% ofthe differences between any two groups could be

Figure 2. Median raw scores for 10 samples of British adults with birth dates ranging from 1877 to 1967. The five data points on theleft represent groups tested in 1942 at ages 65 to 25 years (from left to right). The five data points on the right represent groups tested in1992 at ages 65 to 25 years (from left to right). The dashed line indicates that a raw score falling at the 90th percentile in 1942 fell at the5th percentile in 1992. Data from Raven et al. (1995). Figure adapted from Flynn (1998a). Copyright © 1998 by the American Psycho-logical Association. Adapted with permission.

Raven's Progressive Matrices Scores in Great Britain

0

5

10

15

20

25

30

35

40

45

50

55

60

1877 1887 1897 1907 1917 1927 1937 1947 1957 1967

SER

OCS

WA

R 50th percentile

5th percentile

90th percentile

Year of birth

Year tested

Age when tested

1942 1942 1942 1942 1942 1992 1992 1992 1992 1992

65 2555 3545 35 25 65 55 45

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

524 HISCOCK

attributed to an actual decline in performanceassociated with aging per se. The majority of thedecline in performance from one age group to thenext could be accounted for by the Flynn effect.

The same method yields surprising results whenapplied to the 1992 data. It is evident from Figure2 that the slope of the curve is much smaller for the1992 data than for the 1942 data. As indicated byAverage 2 in Table 1, the combined effect of birthdate and age for groups tested in 1992 is only 0.16points/year, which is less than the estimated magni-tude of the Flynn effect alone. The flattening of thecurve for the 1992 data seems to reflect a ceilingeffect, as the median score for 25- and 35-year-oldsindicates that half of the individuals in thosecohorts had raw scores above the 86% level. Otherdata indicate that Raven’s Matrices scores are con-tinuing to rise as rapidly as in the past (Neisser,1998b). In fact, the unremitting upward shift in thedistribution has prompted the development ofa more difficult version of the Standard Progres-sive Matrices (J. Raven, Raven, & Court, 1998).

Even though the refined cohort norms that areavailable for Raven’s Matrices and the WechslerIQ tests are not available for most neuropsycho-logical instruments, it sometimes is possible toestimate the degree to which the Flynn effect hasinfluenced performance on specific neuropsycho-logical tests. For instance, Dickinson and Hiscock(2006) found six comparable sets of norms for theTrail Making Test that were published between1968 and 2004. Analysis of those norms showedthat overall performance on Part B of the test hasimproved significantly over the interval of 36years. Mean completion time has decreased from

85 s to 60 s. The magnitude of this cohort effect isonly slightly less than the difference betweenmedian times for 30- and 65-year-olds in normativedata (e.g., Spreen & Strauss, 1998). This impliesthat a substantial portion of the age-related slow-ing of performance is attributable to the Flynneffect (at least in the middle and upper ranges ofthe distribution). The lesser degree of improvementon Part A of the Trail Making Test between 1968and 2004 was not statistically significant. Resultsfor other tests were either negative, or indetermi-nate because of a paucity of data. This kind of ana-lysis obviously is limited by the availability,quality, and comparability of archival data.

An alternative strategy for estimating the magni-tude of the Flynn effect for a neuropsychologicaltest entails comparing age-related declines as esti-mated from both cross-sectional and longitudinaldata. If the age-related change observed in cross-sectional data is commensurate with the changeobserved in longitudinal data, then the cross-sectional results can be presumed to index a trueage-related deterioration. If, however, cross-sectional data show greater decline than do longitu-dinal data, the difference probably reflects a cohorteffect. When this strategy was applied to scoresfrom the Boston Naming Test by Connor, Spiro,Obler, and Albert (2004), the cross-sectional datashowed a linear decline that was more than 50%greater than the decline seen in the longitudinaldata, and the quadratic component of the declinewas twice as great in the cross-sectional data as inthe longitudinal data. The authors ruled out prac-tice effects and attrition as factors that might havecaused the cross-sectional results to overestimate

TABLE 1 Age and year-of-testing effects on Raven's Progressive Matrices raw scores for British adults tested in 1942 or 1992

Group Year of birth Year tested Agea Raw score

A 1877 1942 65 24B 1887 1942 55 30C 1897 1942 45 35D 1907 1942 35 41E 1917 1942 25 44F 1927 1992 65 49G 1937 1992 55 51H 1947 1992 45 53I 1957 1992 35 55J 1967 1992 25 55

Average 1 (average of annual birth date and age differences combined for 1942 data):Average [(E – A)+(E – B)+(E – C)+(E – D)+(D – A)+(D – B)+(D – C)+(C – A)+(C – B)+(B – A)]=0.51Average 2 (average of annual birth date and age differences combined for 1992 data):Average [(J – F)+(J – G)+(J – H)+(J – I)+(I – F)+(I – G)+(I – H)+(H – F)+(H – G)+(G – F)]=0.16Average 3 (average of annual difference between comparable age groups in 1942 and 1992):Average [(F – A)+(G – B)+(H – C)+(I – D)+(J – E)]=0.36aIn years.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 525

the actual amount of age-related decline. Connoret al.’s results seem to indicate that one third to onehalf of the age-related change observed in the cross-sectional data is a cohort (Flynn) effect.

These attempts to disentangle age-related changesfrom birth-date effects are only as accurate as thedata on which they are based and only as valid asthe explicit and implicit assumptions on which theanalyses depend. The resulting estimates of age-related and birth-date components cannot be con-sidered to be definitive, even for the WAIS–R DigitSymbol subtest and Raven’s Matrices test. None-theless, it seems clear that year of birth accounts fora substantial proportion of what appears to be anage-related decline in normative performance. Quitepossibly, the differential strength of the Flynn effectacross tests accounts for the apparent differences insensitivity to the effects of aging. What appears tobe differential sensitivity to aging may be, in fact,differential sensitivity to the Flynn effect (Dickinson& Hiscock, 2005). Alternatively, the characteristicsof a test that make it sensitive to aging may be thevery characteristics that make a test sensitive to theFlynn effect as well. A statistical technique knownas age-period-cohort (APC) modeling may proveuseful in future attempts to identify the relative con-tribution of age and cohort effects to changes inmental ability and neuropsychological functioning(Holford, 1991).

A better understanding of how the Flynn effectinfluences age-group norms will not only benefitclinical neuropsychology and the psychology ofaging, but will also help to elucidate the Flynneffect. If researchers can specify which neuropsy-chological tests are “contaminated” by the Flynneffect and the degree to which they are contami-nated, then it may be possible to identify the char-acteristics of a test that render it susceptible to theFlynn effect. For Digit Symbol and the otherPerformance subtests from the Wechsler IQ tests, asubstantial proportion of the apparent age-relatedperformance decline can be attributed to the Flynneffect; 70-year-olds perform worse on Digit Sym-bol partly because their skills have actuallydecreased over the previous 50 years of their lives,but the drop in performance is due largely to thefact that they were born 50 years earlier than the20-year-olds in the same normative sample. Thismay also apply to some neuropsychological tests.

BROADER IMPLICATIONS OF THE FLYNN EFFECT

The problems associated with rapidly changingnormative test performance are hardly trivial

matters for the clinical neuropsychologist. Never-theless, as implied by the title of Flynn’s (1999)article, “Searching for Justice: The Discovery ofIQ Gains over Time,” the rising-curve phenome-non has broader ramifications for psychology andsociety. The impetus behind much of Flynn’seffort to document and explain the rising curve ishis interest in understanding the average IQ dif-ference between racial or ethnic groups (e.g.,Herrnstein & Murray, 1994; Jencks & Phillips,1998; Jensen, 1969). Flynn points out that themean IQ difference between generations born 30years apart, as measured by Raven’s Matrices, islarger than the mean IQ difference betweenAmerican whites and blacks. Moreover, becausechildren and their grandparents share the samegenes, we can be sure that the intergenerationalIQ difference is not genetic irrespective of theextent to which within-group variability isgenetic. Thus, the Flynn effect constitutes a com-pelling example of large between-group IQ differ-ences that are completely environmental. Flynnargues that IQ differences between ethnic orracial groups are analogous to intergenerationaldifferences. The differences between ethnicgroups are also likely to be environmental in ori-gin notwithstanding data showing within-groupdifferences to be genetically influenced. Thereader is referred to pages 12–16 of Flynn (1999)for a detailed discussion of this explanation forthe IQ gap between white and black Americans.

Flynn (2000) also has noted the dilemma thatthe rise in IQ has created for psychologists whouse IQ tests to classify people as mentallyretarded. After questioning the justification forusing an IQ of 70 as a fixed value that separatesthe retarded from the nonretarded, Flynn uses IQnorms to show that the percentage of Americanswho have met this criterion has ranged at differ-ent times during the past 50 years from 0.5 to4.3%. As scores rise throughout the population,fewer and fewer individuals meet the IQ criterionfor mental retardation. Then, when a new IQ testis published, along with new norms, there is a dra-matic increase in the proportion of the populationwith scores falling below 70. Flynn concludes thatawareness of the rising scores, and of the adjust-ments provided by newer tests and newer norms,gives psychologists a choice. Either they can selectthe version of the IQ test that is more likely toyield the desired classification, or they can disre-gard IQ testing and classify individuals solely onthe basis of adaptive functioning. In a morerecent publication, Flynn (2006b) offers a thirdoption—that is, calculating the interval betweenthe time at which a test was normed and the time

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

526 HISCOCK

at which it was administered and then adjustingthe IQ downward to counteract the Flynn effect.Flynn’s article provides all the information neces-sary to make the adjustments, including a list ofthe years during which the various Wechsler andStanford–Binet IQ tests were normed.

Whereas Flynn’s analysis of the mental-retardation problem is based on IQ test norms,Kanaya, Scullin, and Ceci (2003) have used datafrom several American school districts to showhow changes in IQ test norms actually affect thediagnosis of mental retardation (MR). From theiranalysis of longitudinal data, the authors foundthat the magnitude of the Flynn effect for chil-dren in the borderline and mild MR ranges of IQwas similar to that for children in the middle ofthe IQ distribution. Moreover, children in theborderline range who were tested initially with theWISC–R and then retested with the WISC-IIIwere more than twice as likely to be classified asMR as children who were evaluated twice withthe same test. As expected, the Flynn effect wasmost likely to affect the classification of childrenwho scored close to 70, the cutoff for MR. Eventhough a WISC-III IQ that falls a few pointsbelow 70 is equivalent to a WISC–R score slightlyabove 70, children were nearly three times aslikely to be placed into the MR category if theyscored 66–70 on the WISC-III than if theyobtained a comparable score of 71–75 on theWISC–R.

Kanaya et al.’s (2003) analyses confirm Flynn’s(2000) portrayal of the Flynn effect as a forcethat, over time, progressively inflates IQ scoresuntil a new test, with new norms, causes thescores to “plummet back toward baseline” (p.787). The authors caution that “there is reason tobelieve that many students are diagnosed as MRbased on the year in which they are tested and testnorms used rather than on their cognitive ability”(pp. 786–787). In addition to educational conse-quences of IQ classifications (e.g., eligibility forspecial-education services), Kanaya et al. alsomention several financial, legal, and occupationalconsequences. The most dramatic consequence inthe United States is eligibility for the death pen-alty. The U.S. Supreme Court ruled in 2002 that amentally retarded individual cannot be executedafter conviction for a capital crime. Although theintellectual classification of adult defendants isnot determined solely by IQ tests, IQ is often animportant factor. Thus, a defendant’s intellectualclassification literally may be a matter of life ordeath. Accordingly, the decision to execute or notexecute a person may hinge on whether theperson’s IQ has been properly corrected for the

Flynn effect. With increasing frequency, courts inthe United States are recognizing the need forsuch corrections (see Flynn, 2006b).

CONCLUSIONS

A pervasive increase over time in performance onIQ tests is well established. The magnitude of theincrease is especially marked when culture-reducedtests of general intelligence, such as Raven’s Matri-ces, are used. The Flynn effect also raises scoresfrom Wechsler and Stanford–Binet IQ tests, andthe increases are sufficiently large as to present aninterpretive problem for practitioners who admin-ister IQ tests to their clients. Not only does theFlynn effect cause published norms for Full ScaleIQ to become progressively less appropriate overtime, but it also causes different subtest norms tochange at different rates. Clinical neuropsycholo-gists who use old versions of IQ tests not only willoverestimate IQ but also will risk misinterpretingipsative indicators such as Verbal–Performancedisparities and subtest profiles.

Rapidly changing norms are especially problem-atic in situations that require reevaluation of a cli-ent after a new version of a test has beenintroduced. If the psychologist chooses the previ-ous version of a test—the version used in the initialassessment—changes in performance can be meas-ured directly but only at the cost of sacrificing thecurrency of the norms. To the degree that the sub-tests have been influenced unequally by the Flynneffect, the decision rules that are commonlyapplied to patterns of scores obtained from the testwill be undermined. If the psychologist opts for therevised version of the test, the benefit of up-to-datenorms is offset by interpretative ambiguity associ-ated with altered content. The psychologist at firstwill not know whether clinical decision rulesremain valid for the new test and will have to awaitthe availability of empirical evidence to support thecontinued use of those rules. The clinician’sdilemma becomes particularly acute if the testresults will influence classification (e.g., mentalretardation) or intervention (e.g., special education).

Norms, after being compiled and published, arestatic but the Flynn effect continues. Conse-quently, potential Flynn-effect problems do notend with the acceptance of a new test and currentnorms. The norms will gradually grow old and failto reflect the recent rise in ability. As time passes,retesting an individual with the same test requiresthat the Flynn effect be taken into account even ifthe test has not been revised, and new norms havenot been disseminated.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 527

Neuropsychologists who evaluate elderlypatients face an additional Flynn-effect problem—namely, the confounding of cross-sectional age-group norms with year of birth. This is the classicalcohort effect, to which cross-sectional data arealways susceptible. Normal decrements in subtestscores with increasing age reflect an unknown mix-ture of actual age-related deterioration of perform-ance and birth-date-related decline (i.e., the Flynneffect). Preliminary analyses suggest that, for twoof the tests most sensitive to age-related change—Raven’s Matrices and Wechsler Digit Symbol—theFlynn effect may account for more than 50% of thedecline in raw scores between the ages of 20 and 70years. If this finding is confirmed in futureresearch, it will necessitate modifying the interpreta-tion of what previously has been regarded as biolog-ically based decline in mental ability. Disentanglingage-related decline from the Flynn effect is mademore difficult by the likelihood that the tests mostsensitive to age-related biological deterioration arethe very tests that are most vulnerable to the Flynneffect (Dickinson & Hiscock, 2005).

IQ may be tip of the Flynn-effect iceberg. Theworst case scenario for clinical neuropsychologyinvolves widespread Flynn-effect contamination ofneuropsychological test norms. This is a plausibleconcern because neuropsychological instrumentstend to be relatively culture free, and many of thetest norms are old. The problem is compounded tothe extent that the various neuropsychological testsand norms have been published at different times.Not only are the tests likely to be differentially vul-nerable to the Flynn effect, but their respectivenorms are out of date in varying degrees.

The uneven quality of normative data for neu-ropsychological tests is a problem that has beenrecognized and, to some extent, redressed inrecent years (e.g., Heaton, Grant, & Matthews,1991). However, if normative neuropsychologicaltest scores tend to rise substantially over time,either norms will have to be updated frequentlyor else old norms will have to be adjusted mathe-matically for the passage of time between collec-tion of the normative data and collection of theclinical data. It should also be noted that newnorms may underestimate the actual magnitudeof the Flynn effect if those norms reflect thepooling of current and previous normative data,or if they have been accumulated over a pro-tracted period.

The Flynn effect may be regarded by the clini-cal neuropsychologist as an unwelcome sourceof complexity that can make the interpretationof test scores—already a challenging undertak-ing—even more difficult. On the other hand,

neuropsychological tests may hold clues that willlead to a better understanding of the Flynneffect. The search for rising scores on neuropsy-chological tests will reveal, at the very least, thescope of the Flynn effect. Is it restricted to intel-lectual tests, or does it affect various tests of per-ception, attention, processing speed, learning,memory, working memory, language, motorskill, and executive functions? If the Flynn effectdoes extend beyond the realm of IQ tests, asindicated by the recent findings of Connor et al.(2004) and Dickinson and Hiscock (2006), thenext question would concern its differentialimpact on different kinds of tests. Does it influ-ence some categories of tests (e.g., workingmemory) more than others (e.g., language)?Does it affect speeded performance more thanperformance on untimed tests? Does it affectbroad-band tests more than tests of specificfunctions? Does it affect tests according to thestrength of the association between test perform-ance and IQ? Whereas a detailed analysis of theFlynn effect in neuropsychological testing mightnot lead to an immediate and complete explana-tion of the phenomenon, it will answer somequestions of practical and theoreticalimportance.

Original manuscript received 11 May 2005Revised manuscript accepted 15 May 2006

First published online 20 December 2006

REFERENCES

Bell, R. Q. (1968). A reinterpretation of the direction ofeffects in studies of socialization. PsychologicalReview, 75, 81–95.

Boake, C. (2002). From the Binet–Simon to theWechsler–Bellevue: Tracing the history of intelligencetesting. Journal of Clinical and Experimental Neu-ropsychology, 24, 363–405.

Boring, E. G. (1923, June 6). Intelligence as the tests testit. The New Republic, 35–37.

Bornstein, R. A. (1987). The WAIS–R in neuropsycho-logical practice: Boon or bust? The Clinical Neuropsy-chologist, 1, 185–190.

Bouchard, T. J., Jr., Lykken, D. T., McGue, M., Segal,N. L., & Tellegen, A. (1990). Sources of human psy-chological differences: The Minnesota study of twinsreared apart. Science, 250, 223–250.

Burke, H. R. (1985). Raven’s Progressive Matrices:More on norms reliability, and validity. Journal ofClinical Psychology, 41, 231–235.

Carlson, J. S., & Jensen, C. M. (1980). The factorialstructure of the Raven Coloured Progressive Matri-ces Test: A reanalysis. Educational and PsychologicalMeasurement, 40, 1111–1116.

Carpenter, P. A., Just, M. A., & Shell, P. (1990). Whatone intelligence test measures: A theoretical accountof the processing in the Raven Progressive MatricesTest. Psychological Review, 97, 404–431.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

528 HISCOCK

Cattell, R. B. (1957). Personality and motivation structureand measurement. New York: Harcourt, Brace &World.

Chelune, G. J., Eversole, C., Kane, M., & Talbott, R.(1987). WAIS versus WAIS–R subtest patterns: Aproblem of generalization. The Clinical Neuropsy-chologist, 1, 235–242.

Cole, M. (1996). Cultural psychology: A once and futurediscipline. Cambridge, MA: Belknap Press.

Costa, L. D. (1976). Interset variability on the RavenColored Progressive Matrices as an indicator of spe-cific ability deficit in brain-lesioned patients. Cortex,12, 3–40.

Cozby, P. C. (2003). Methods in behavioral research (8thed.). Boston: McGraw-Hill.

Diaz-Asper, C. M., Schretlen, D. J., & Pearlson, G. D.(2004). How well does IQ predict neuropsychologicaltest performance in normal adults? Journal of theInternational Neuropsychological Society, 10, 82–90.

Dickens, W. T., & Flynn, J. R. (2001a). Great leap for-ward: A new theory of intelligence. New Scientist, 21,44–47.

Dickens, W. T., & Flynn, J. R. (2001b). Heritability esti-mates versus large environmental effects: The IQ para-dox resolved. Psychological Review, 108, 346–369.

Dickinson, M. D., & Hiscock, M. (2005, February). Theinfluence of the Flynn effect on age norms for theWechsler IQ test [Abstract]. Poster session presentedat the annual meeting of the International Neuropsy-chological Society, St. Louis, MO.

Dickinson, M. D., & Hiscock, M. (2006, February). TheFlynn effect in neuropsychological assessment[Abstract]. Poster session presented at the annualmeeting of the International NeuropsychologicalSociety, Boston.

Flynn, J. R. (1984). The mean IQ of Americans: Massivegains 1932 to 1978. Psychological Bulletin, 95, 29–51.

Flynn, J. R. (1987). Massive IQ gains in 14 nations:What IQ tests really measure. Psychological Bulletin,101, 171–191.

Flynn, J. R. (1994). IQ gains over time. In R. J. Stern-berg (Ed.), The encyclopedia of human intelligence(pp. 617–623). New York: Macmillan.

Flynn, J. R. (1998a). IQ gains over time: Toward findingthe causes. In U. Neisser (Ed.), The rising curve:Long-term gains in IQ and related measures (pp. 25–66). Washington, DC: American PsychologicalAssociation.

Flynn, J. R. (1998b). Israeli military IQ tests: Genderdifferences small; IQ gains large. Journal of BiosocialScience, 30, 541–553.

Flynn, J. R. (1998c). WAIS-III and WISC-III IQ gainsin the United States from 1972 to 1995: How to com-pensate for obsolete norms. Perceptual and MotorSkills, 86, 1231–1239.

Flynn, J. R. (1999). Searching for justice: The discov-ery of IQ gains over time. American Psychologist,54, 5–20.

Flynn, J. R. (2000). The hidden history of IQ and specialeducation: Can the problem be solved? Psychology,Public Policy, and Law, 6, 191–198.

Flynn, J. R. (2003). Movies about intelligence: The limi-tations of g. Current Directions in PsychologicalScience, 12, 95–99.

Flynn, J. R. (2006a). Efeito Flynn: Repensando ainteligência e seus efeitos [The Flynn effect: Rethinkingintelligence and what affects it]. In C. Flores-Mendoza

& R. Colom (Eds.), Introdução à psicologia das difer-enças individuais [Introduction to the psychology ofindividual differences] (pp. 387–411). Pôrto Alegre,Brasil: ArtMed. (English trans. available fromjim.flynn@stonebow.otago.ac.nz).

Flynn, J. R. (2006b). Tethering the elephant: Capitalcases, IQ, and the Flynn effect. Psychology, PublicPolicy, and Law, 12, 170–189.

Flynn, J. R. (2006c). Towards a theory of intelligencebeyond g. Behavioral and Brain Sciences, 29, 132–134.

Flynn, J. R., & Weiss, L. G. (2006). American IQ gainsfrom 1932 to 2002: The significance of the WISC sub-tests. Manuscript submitted for publication.

Gould, S. J. (1996). Full house: The spread of excellencefrom Plato to Darwin. New York: Harmony Books.

Greenfield, P. M. (1998). The cultural evolution of IQ.In U. Neisser (Ed.), The rising curve: Long-term gainsin IQ and related measures (pp. 81–123). Washington,DC: American Psychological Association.

Hanson, R., Smith, J. A., & Hume, W. (1985). Achieve-ments of infants on items of the Griffiths scales: 1980compared to 1950. Child: Care, Health and Develop-ment, 11, 91–104.

Heaton, R. K., Grant, I., & Matthews, C. G. (1991).Comprehensive norms for an expanded Halstead–Reitan battery: Demographic corrections, researchfindings, and clinical applications. Odessa, FL: Psy-chological Assessment Resources.

Herrnstein, R. J., & Murray, C. (1994). The bell curve:Intelligence and class structure in American life. NewYork: Free Press.

Hiscock, M., Inch, R., & Gleason, A. (2002). Raven’sProgressive Matrices performance in adults withtraumatic brain injury. Applied Neuropsychology, 9,129–138.

Holford, T. R. (1991). Understanding the effects of age,period, and cohort on incidence and mortality rates.Annual Review of Public Health, 12, 425–457.

Jencks, C., & Phillips, M. (1998). The Black–White testscore gap. Washington, DC: Brookings InstitutionPress.

Jensen, A. R. (1969). How much can we boost IQ andscholastic achievement? Harvard Educational Review,39, 1–123.

Jensen, A. R. (1973). Educability and group differences.New York: Harper & Row.

Jensen, A. R. (1975). The meaning of heritability in thebehavioral sciences. Educational Psychologist, 11,171–183.

Jensen, A. R. (1980). Bias in mental testing. London:Methuen.

Jensen, A. R. (1998). The g factor: The science of mentalability. Westport, CT: Praeger.

Kanaya, T., Scullin, M. H., & Ceci, S. (2003). The Flynneffect and U.S. policies: The impact of rising IQscores on American society via mental retardationdiagnoses. American Psychologist, 58, 778–790.

Kaplan, E., Fein, D., Morris, R., & Delis, D. C. (1991).WAIS–R as a neuropsychological instrument. SanAntonio, TX: The Psychological Corporation.

Lezak, M. D. (1983). Neuropsychological assessment(2nd ed.). New York: Oxford University Press.

Lezak, M. D. (1988). IQ: R.I.P. Journal of Clinical andExperimental Neuropsychology, 10, 351–361.

Lezak, M. D., Howieson, D. B., & Loring, D. W. (2004).Neuropsychological assessment (4th ed.). New York:Oxford University Press.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013

FLYNN EFFECT IN NEUROPSYCHOLOGY 529

Llabre, M. M. (1984). Standard Progressive Matrices. InD. J. Keyser & R. C. Sweetland (Eds.), Test critiques(Vol. 1). Kansas City, MO: Test Corporation ofAmerica.

Luria, A. R. (1973). The working brain: An introductionto neuropsychology (B. Haigh, Trans.) New York:Basic Books.

Lynn, R. (1990). the role of nutrition in the secularincreases of intelligence. Personality and IndividualDifferences, 11, 273–286.

Lynn, R. (1998). In support of the nutrition theory. InU. Neisser (Ed.), The rising curve: Long-term gains inIQ and related measures (pp. 207–215). Washington,DC: American Psychological Association.

Martorell, R. (1998). Nutrition and the worldwide rise inIQ scores. In U. Neisser (Ed.), The rising curve:Long-term gains in IQ and related measures(pp. 183–206). Washington, DC: American Psycho-logical Association.

Neisser, U. (1998a). Introduction: Rising test scores andwhat they mean. In U. Neisser (Ed.), The rising curve:Long-term gains in IQ and related measures (pp. 3–22).Washington, DC: American Psychological Association.

Neisser, U. (Ed.). (1998b). The rising curve: Long-termgains in IQ and related measures. Washington, DC:American Psychological Association.

Plomin, R. (1986). Development, genetics, and psychology.Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Plomin, R. (1990). The role of inheritance in behavior.Science, 248, 183–188.

Raven, J. C. (1938). Progressive Matrices: A perceptual testof intelligence: Individual form. London: H.K. Lewis.

Raven, J. C. (1960). Guide to the Standard ProgressiveMatrices. London: H.K. Lewis.

Raven, J., Raven, J. C., & Court, J. H. (1995). Manualfor Raven's Progressive Matrices and VocabularyScales: Section J. General overview. Oxford, UK:Oxford Psychologists Press.

Raven, J., Raven, J. C., & Court, J. H. (1998). Manualfor Raven’s Progressive Matrices and VocabularyScales: Section 1. General overview. Introducing paral-lel versions of the CPM and SPM together with a morepowerful version of the SPM (SPM-Plus). Oxford:Oxford Psychologists Press.

Reitan, R. M., & Wolfson, D. (1985). The Halstead–Reitan Neuropsychological Test Battery: Theory andclinical interpretation. Tucson, AZ: NeuropsychologyPress.

Reitan, R. M., & Wolfson, D. (1990). A consideration ofthe comparability of the WAIS and WAIS–R. TheClinical Neuropsychologist, 4, 80–85.

Rourke, B. P., Bakker, D. J., Fisk, J. L., & Strang, J. D.(1983). Child neuropsychology: An introduction to the-ory, research, and clinical practice. New York: Guilford.

Scarr, S. (1992). Developmental theories for the 1990s:Development and individual differences. Child Devel-opment, 63, 1–19.

Schaie, K. W. (1994). The course of adult intellectualdevelopment. American Psychologist, 49, 304–313.

Sigman, M., & Whaley, S.E. (1998). The role of nutritionin the development of intelligence. In U. Neisser(Ed.), The rising curve: Long-term gains in IQ andrelated measures (pp. 155–182). Washington, DC:American Psychological Association.

Snow, R. E., Kyllonen, C. P., & Marshalek, B. (1984).The topography of ability and learning correlations.In R. J. Sternberg (Ed.), Advances in the psychology ofhuman intelligence (pp. 47–103). Hillsdale, NJ:Lawrence Erlbaum Associates, Inc.

Spearman, C. S. (1904). “General intelligence,” objec-tively determined and measured. American Journal ofPsychology, 15, 201–209.

Spreen, O., & Strauss, E. (1998). A compendium of neu-ropsychological tests: Administration, norms, and com-mentary (2nd ed.). New York: Oxford UniversityPress.

Sternberg, R. J., Grigorenko, E. L., & Bundy, D. A.(2001). The predictive value of IQ. Merrill-PalmerQuarterly, 47, 1–41.

Tuddenham, R. D. (1948). Soldier intelligence in WorldWars I and II. American Psychologist, 3, 54–56.

Villardita, C. (1985). Raven’s Colored ProgressiveMatrices and intellectual impairment in patients withfocal brain damage. Cortex, 21, 627–634.

Waltz, J. A., Knowlton, B. J., Holyoak, K. J., Boone, K.B., Mishkin, F. S., Santos, M. de M., et al. (1999). Asystem for relational reasoning in human prefrontalcortex. Psychological Science, 10, 119–123.

Wechsler, D. (1997). The Wechsler Adult IntelligenceScale–Third Edition. San Antonio, TX: PsychologicalCorporation.

Williams, W. M. (1998). Are we raising smarter childrentoday? School- and home- related influences on IQ.In U. Neisser (Ed.), The rising curve: Long-term gainsin IQ and related measures (pp. 125–154). Washington,DC: American Psychological Association.

Dow

nloa

ded

by [

Dal

hous

ie U

nive

rsity

] at

08:

14 1

6 A

pril

2013