The IQ Halo effect

On desirable correlates of intelligence
psychology, survey, IQ
2013-04-012017-02-02 notes certainty: highly likely importance: 9

Increased IQ cor­re­lates with con­se­quen­tially morally desir­able actions, atti­tudes, and beliefs.

“All good things tend to go togeth­er, as do all bad ones.” –Ed­ward Lee Thorndike

mil­i­tary: AFQT IQ test (sub­test of the ASVAB) excel­lent pre­dic­tor of train­ing costs, suc­cess

the con­sen­sus reply to Bell Curve:

  1. IQ is strongly relat­ed, prob­a­bly more so than any other sin­gle mea­sur­able human trait, to many impor­tant edu­ca­tion­al, occu­pa­tion­al, eco­nom­ic, and social out­comes. Its rela­tion to the wel­fare and per­for­mance of indi­vid­u­als is very strong in some are­nas in life (ed­u­ca­tion, mil­i­tary train­ing), mod­er­ate but robust in oth­ers (so­cial com­pe­tence), and mod­est but con­sis­tent in oth­ers (law-abid­ing­ness). What­ever IQ tests mea­sure, it is of great prac­ti­cal and social impor­tance.

  2. A high IQ is an advan­tage in life because vir­tu­ally all activ­i­ties require some rea­son­ing and deci­sion-mak­ing. Con­verse­ly, a low IQ is often a dis­ad­van­tage, espe­cially in dis­or­ga­nized envi­ron­ments. Of course, a high IQ no more guar­an­tees suc­cess than a low IQ guar­an­tees fail­ure in life. There are many excep­tions, but the odds for suc­cess in our soci­ety greatly favor indi­vid­u­als with higher IQs.

  3. That IQ may be highly her­i­ta­ble does not mean that it is not affected by the envi­ron­ment. Indi­vid­u­als are not born with fixed, unchange­able lev­els of intel­li­gence (no one claims they are). IQs do grad­u­ally sta­bi­lize dur­ing child­hood, how­ev­er, and gen­er­ally change lit­tle there­after.

  4. Although the envi­ron­ment is impor­tant in cre­at­ing IQ differ­ences, we do not know yet how to manip­u­late it to raise low IQs per­ma­nent­ly. Whether recent attempts show promise is still a mat­ter of con­sid­er­able sci­en­tific debate.

The con­tro­versy over The Bell Curve (Her­rn­stein & Mur­ray, 1994) was at its height in the fall of 1994. Many crit­ics attacked the book for sup­pos­edly rely­ing on out­dat­ed, pseu­do­sci­en­tific notions of intel­li­gence. In crit­i­ciz­ing the book, many crit­ics pro­moted false and highly mis­lead­ing views about the sci­en­tific study of intel­li­gence. Pub­lic mise­d­u­ca­tion on the topic is hardly new (Sny­der­man & Roth­man, 1987, 1988), but never before had it been so angry and extreme…It is obvi­ously not the case that there is no dis­agree­ment about these impor­tant issues or that sci­en­tific truth is a mat­ter of major­ity rule. A sig­nifi­cant minor­ity of the experts who were con­tacted dis­agreed in part or in whole with the state­ment, and many of the sign­ers would have writ­ten the state­ment some­what differ­ent­ly. Rather, the les­son here is that what have often been car­i­ca­tured in the pub­lic press as dis­cred­it­ed, fringe ideas actu­ally rep­re­sent the solid sci­en­tific cen­ter in the seri­ous study of intel­li­gence. As Sny­der­man and Roth­man’s (1988) sur­vey of IQ experts and jour­nal­ists revealed, the media, among oth­ers, have been turn­ing the truth on its head.

“Gen­eral Men­tal Abil­ity in the World of Work: Occu­pa­tional Attain­ment and Job Per­for­mance”, Schmidt & Hunter 2004

The accu­mu­lated evi­dence has become very strong that GMA is cor­re­lated with a wide vari­ety of life out­comes, rang­ing from risky health-re­lated behav­iors, to crim­i­nal offens­es, to the abil­ity to use a bus or sub­way sys­tem (Got­tfred­son, 1997; Lubin­ski & Humphreys, 1997). In addi­tion, the more highly a given GMA mea­sure loads on the gen­eral fac­tor in men­tal abil­ity (the g fac­tor), the larger are these cor­re­la­tions. The rel­a­tive stand­ing of indi­vid­u­als on GMA has been found to be sta­ble over peri­ods of more than 65 years (Deary, Whal­ley, Lem­mon, Craw­ford, & Starr, 2000). Find­ings in behav­ior genet­ics, includ­ing stud­ies of iden­ti­cal twins reared apart and together (e.g., Bouchard, Lykken, McGue, Segal, & Tel­le­gen, 1990), have shown beyond doubt that GMA has a strong genetic basis (e.g., Bouchard, 1998; McGue & Bouchard, 1998).

  • Lubin­ski, D., & Humphreys, L. G. (1997). Incor­po­rat­ing gen­eral intel­li­gence into epi­demi­ol­ogy and the social sci­ences. Intel­li­gence, 24, 159 -201
  • Deary, I. J., Whal­ley, L. J., Lem­mon, H., Craw­ford, J. R., & Starr, J. M. (2000). The sta­bil­ity of indi­vid­ual differ­ences in men­tal abil­ity from child­hood to old age: Fol­low-up of the 1932 Scot­tish men­tal sur­vey. Intel­li­gence, 28, 49 -55

Peo­ple’s rank­ings or rat­ings of the occu­pa­tional level or pres­tige of differ­ent occu­pa­tions are very reli­able; cor­re­la­tions between mean rat­ings across stud­ies are in the .95 to .98 range, regard­less of the social class, occu­pa­tion, age, or coun­try of the raters (Daw­is, 1994; Jensen, 1980, pp. 339 -347). These occu­pa­tional level rat­ings cor­re­late between .90 and .95 with aver­age GMA scores of peo­ple in the occu­pa­tions (Jensen, 1998, p. 293). Indi­vid­ual level cor­re­la­tions are of course not this large. In the U.S. Employ­ment Ser­vice’s large data­base on the Gen­eral Apti­tude Test Bat­tery (GATB; Hunter, 1980), the indi­vid­ual level cor­re­la­tion between the GMA mea­sure derived from that bat­tery and occu­pa­tional level is .65 (.72 cor­rected for mea­sure­ment error; Jensen, 1998). Much mil­i­tary data exist from both world wars (when sam­ples of draftees were very rep­re­sen­ta­tive of the U.S. male pop­u­la­tion) show­ing an increase in mean GMA scores as occu­pa­tional level (as deter­mined by rat­ings of the sort dis­cussed here) increases (Har­rell & Har­rell, 1945; Stew­art, 1947; Yerkes, 1921). Table 1, show­ing find­ings for 18,782 White enlisted men in the Army Air Force Com­mand (Har­rell & Har­rell, 1945), presents typ­i­cal find­ings. The GMA mea­sure used was the Army Gen­eral Clas­si­fi­ca­tion Test (Schmidt, Hunter, & Pearl­man, 1981). Mean GMA scores clearly increase with occu­pa­tional lev­el.

Wilk, Des­marais, and Sack­ett (1995), using the 3,887 young adults in the National Lon­gi­tu­di­nal Sur­vey-Y­outh Cohort (NLSY; Cen­ter for Human Resource Research, 1989) for whom the required data were avail­able, showed that over the 5-year period from 1982 to 1987, GMA mea­sured in 1980 pre­dicted move­ment in the job hier­ar­chy. Those with higher GMA scores in 1980 moved up the hier­ar­chy, whereas those with lower GMA scores moved down in the hier­ar­chy. In a larger fol­low-up study that was based on some­what differ­ent method­ol­o­gy, Wilk and Sack­ett (1996) exam­ined two large gov­ern­ment data­bas­es: the National Lon­gi­tu­di­nal Study of the Class of 1972 (NLS-72) and the National Lon­gi­tu­di­nal Sur­vey of Labor Mar­ket Expe­ri­ence-Y­outh Cohort (NLSY). In both data­bas­es, Wilk and Sack­ett found that job mobil­ity was pre­dicted by the con­gru­ence between indi­vid­u­als’ GMA scores (mea­sured sev­eral years ear­lier) and the objec­tively mea­sured com­plex­ity of their jobs. If their GMA exceeded the com­plex­ity level of their job, they were likely to move into a higher com­plex­ity job. And if the com­plex­ity level of their job exceeded their GMA lev­el, they were likely to move down into a less com­plex job.

In another study drawn from this same large data­base, Mur­ray (1998) found that GMA pre­dicted later income even with unusu­ally thor­ough con­trol for socioe­co­nomic sta­tus (SES) and other back­ground vari­ables. This con­trol took advan­tage of the large vari­abil­ity of GMA within fam­i­lies and was achieved by use of a sam­ple of male full bio­log­i­cal sib­lings, hence con­trol­ling for home back­ground and many other vari­ables (e.g., schools, neigh­bor­hood­s). Mur­ray found that the sib­lings with higher GMA scores received more edu­ca­tion, entered more pres­ti­gious occu­pa­tions, had higher income, and were employed more reg­u­lar­ly. When the sib­lings were in their late 20s (in 1993), a per­son with aver­age GMA was earn­ing on aver­age almost $18,000 less per year than his brighter sib­ling who had an IQ of 120 or higher and was earn­ing more than $9,000 more than his duller sib­ling who had an IQ of less than 80. This pat­tern of find­ings held up even in a sub­-sam­ple of per­sons who were all from “advan­taged” homes (his “utopian” sam­ple).

Judge, Hig­gins, Thore­sen, and Bar­rick (1999) related GMA mea­sures taken at around 12 years of age to occu­pa­tional out­comes in the age range of 41 to 50 years. They found that child­hood GMA scores pre­dicted adult occu­pa­tional level with a cor­re­la­tion of .51 and pre­dicted adult income with a cor­re­la­tion of .53. Ball (1938) found that GMA mea­sured in child­hood cor­re­lated .47 with occu­pa­tional level 14 years later and .71 with occu­pa­tional level 19 years lat­er. Other such stud­ies include Brown and Reynolds (1975), Dreher and Bretz (1991), Got­tfred­son and Crouse (1986), Howard and Bray (1990), Siegel and Ghis­elli (1971), and Thorndike and Hagen (1959).

Results for GMA are typ­i­fied by the find­ings of the large study con­ducted by Hunter (1980; Hunter & Hunter, 1984) for the U.S. Employ­ment Ser­vice using the data­base on the Gen­eral Apti­tude Test Bat­tery (GATB). On the basis of 425 valid­ity stud­ies (N ϭ 32,124) con­ducted on civil­ian jobs span­ning the occu­pa­tional spec­trum, Hunter and Hunter (1984) and Hunter (1980) reported the results shown in Table 2. Hunter assigned each job to one of five job fam­i­lies based on com­plex­ity (i.e., the infor­ma­tion pro­cess­ing require­ments of the job, mea­sured using U.S. Depart­ment of Labor job analy­sis data for each job). This is the largest data­base avail­able using a mea­sure of per­for­mance on the job (mea­sured using super­vi­sory rat­ings of job per­for­mance). As can be seen, valid­ity for pre­dict­ing per­for­mance on the job ranges from .58 for the high­est com­plex­ity jobs (pro­fes­sion­al, sci­en­tific, and upper man­age­ment jobs) to .23 at the low­est com­plex­ity level (feeding/ off-bear­ing job­s). Job Fam­ily 2 (2.5% of all jobs in the econ­o­my) con­sists of com­plex tech­ni­cal jobs such as com­put­er-sys­tems trou­ble shoot­ing or com­plex man­u­fac­tur­ing set-up jobs. Job Fam­ily 3, with almost 63% of all jobs in the econ­o­my, includes skilled work­ers, tech­ni­cians, mid-level admin­is­tra­tors, para­pro­fes­sion­als, and sim­i­lar jobs. Job Fam­ily 4 is semi­skilled work. Clear­ly, GMA pre­dicts per­for­mance on higher level jobs bet­ter that it does for lower level jobs. How­ev­er, there is sub­stan­tial valid­ity for all job lev­els. In par­tic­u­lar, GMA pre­dicts per­for­mance even for the sim­plest 2.4% of jobs (Job Fam­ily 5). Other find­ings are reported in Table 3. On the basis of 194 stud­ies (N ϭ 17,539) of per­for­mance in cler­i­cal jobs, Pearl­man, Schmidt, and Hunter (1980) reported a mean GMA valid­ity for job per­for­mance of .52. For law enforce­ment jobs, Hir­sh, Northrup, and Schmidt (1986) reported a mean valid­ity for job per­for­mance of .38. In a large scale, mul­ti­-year mil­i­tary study on enlisted Army per­son­nel (called “Project A”), McHen­ry, Hough, Toquam, Han­son, and Ash­worth (1990) reported that GMA pre­dicted “Core Tech­ni­cal Pro­fi­ciency” with a valid­ity of .63 and “Gen­eral Sol­dier­ing Per­for­mance” with a valid­ity of .65. Both job per­for­mance mea­sures were based on hand­s-on work-sam­ple mea­sures. (Va­lidi­ties were not as high for rat­ings of “Effort and Lead­er­ship” [.31], “Per­sonal Dis­ci­pline” [.16], and “Phys­i­cal Fit­ness and Mil­i­tary Bear­ing” [.20], which are sec­ondary cri­te­rion mea­sures with fewer cog­ni­tive demand­s.) Using sim­i­lar job sam­ple mea­sures of job per­for­mance, Ree, Ear­les, and Tea­chout (1994) reported a mean value of .45 across seven Air Force jobs. Validi­ties for the pre­dic­tion of per­for­mance in train­ing pro­grams are even larg­er. As can be seen in Table 2, in the GATB train­ing data­base (90 stud­ies; N ϭ 6,496) used by Hunter and Hunter (1984), GMA pre­dicted per­for­mance in job train­ing pro­grams for all job fam­i­lies for which data existed with a cor­re­la­tion above .50. The data­base for train­ing per­for­mance is larger for mil­i­tary jobs. Hunter (1986) meta-an­a­lyzed mil­i­tary data­bases total­ing over 82,000 trainees and reported an aver­age valid­ity of .63 for GMA. On the basis of 77,958 Air Force trainees, Ree and Ear­les (1991) reported a very sim­i­lar value of .60. On the basis of 65 stud­ies with an N of 32,157, Pearl­man et al. (1980) reported a mean valid­ity of .71 for GMA in pre­dict­ing train­ing per­for­mance of cler­i­cal work­ers, whereas Hirsh et al. (1986) found a mean value of .76 for pre­dict­ing per­for­mance in police and other train­ing acad­e­mies for law enforce­ment trainees. These find­ings and oth­ers are shown in Table 3. Addi­tional data of this sort are pre­sented in Schmidt (2002).

Differ­en­tial or spe­cific apti­tude the­ory hypoth­e­sizes that per­for­mance on differ­ent jobs requires differ­ent cog­ni­tive apti­tudes and, there­fore, regres­sion equa­tions com­puted for each job and incor­po­rat­ing mea­sures of sev­eral spe­cific apti­tudes will opti­mize the pre­dic­tion of per­for­mance on the job and in train­ing. In the last 10 years, research has strongly dis­con­firmed this the­o­ry. Differ­en­tially weight­ing spe­cific apti­tude tests pro­duces lit­tle or no increase in valid­ity over the use of a mea­sure of GMA. An expla­na­tion for this find­ing has been dis­cov­ered. It has been found that spe­cific apti­tude tests mea­sure GMA; in addi­tion to GMA, each mea­sures some­thing spe­cific to that apti­tude (e.g., specifi­cally numer­i­cal apti­tude, over and above GMA). The GMA com­po­nent appears to be respon­si­ble for the pre­dic­tion of job and train­ing per­for­mance, whereas the fac­tors spe­cific to the apti­tudes appear to con­tribute lit­tle or noth­ing to pre­dic­tion. The research show­ing this is pre­sented and reviewed in Hunter (1986); Jensen (1986); Thorndike (1986); Olea and Ree (1994); Ree and Ear­les (1992); Ree et al. (1994); Schmidt, Ones, and Hunter (1992); and Sack­ett and Wilk (1994), among other sources. A par­tic­u­larly dra­matic refu­ta­tion of spe­cific apti­tude the­ory comes from the large sam­ple mil­i­tary research con­ducted by Hunter (1983b) for the Depart­ment of Defense on the per­for­mance of mil­i­tary per­son­nel in job train­ing pro­grams. Four large sam­ples were ana­lyzed sep­a­rate­ly: 21,032 Air Force per­son­nel, 20,256 Mari­nes, and two Army sam­ples of 16,618 and 79,926, respec­tive­ly. In all sam­ples, test data were obtained some months prior to mea­sure­ment of per­for­mance in job train­ing pro­grams. In all sam­ples, causal analy­sis mod­el­ing (with cor­rec­tions for mea­sure­ment error and range restric­tion) was used to pit spe­cific apti­tude the­ory against GMA in the pre­dic­tion of per­for­mance. In the case of all four sam­ples, mod­els with causal arrows from spe­cific apti­tudes to train­ing per­for­mance failed to fit the data. How­ev­er, in all the sam­ples, a hier­ar­chi­cal model show­ing a sin­gle causal path from GMA to per­for­mance - and no paths from spe­cific apti­tudes to per­for­mance - fit the data quite well. …Train­ing per­for­mance is deter­mined only by GMA, with the stan­dard­ized path coeffi­cient from GMA to per­for­mance being very large (.62). The find­ings for the other three sam­ples were essen­tially iden­ti­cal (Hunter, 1983b). It is well known that analy­sis of causal mod­els with cor­re­la­tional data can­not prove a the­o­ry. How­ev­er, such analy­ses - espe­cially when sam­ples are very large, as here - can dis­con­firm the­o­ries. The­o­ries that do not fit the data are dis­con­firmed. In these stud­ies, spe­cific apti­tude the­ory is strongly dis­con­firmed.

McDaniel (1985) ana­lyzed United States Employ­ment Ser­vices (USES) data for groups whose level of job expe­ri­ence extended beyond 5 years. Con­trol­ling for differ­ences in vari­abil­ity of GMA across groups, McDaniel cor­re­lated GMA with per­for­mance rat­ings for each level of expe­ri­ence to 12 years and up. The results are sum­ma­rized in Table 4. As the level of expe­ri­ence increas­es, the pre­dic­tive valid­ity does not decrease. Valid­ity goes from .36 for 0 - 6 years, up to .44 for 6 -12 years, up to .59 for more than 12 years (although the last value is based on a very small sam­ple). If any­thing, McDaniel’s data sug­gest an increase in the valid­ity of GMA for pre­dict­ing per­for­mance rat­ings as level of worker expe­ri­ence increas­es.

Many peo­ple may also believe that per­son­al­ity is more impor­tant than GMA in deter­min­ing ulti­mate occu­pa­tional lev­el. How­ev­er, research sup­ports the con­clu­sion that per­son­al­ity is less impor­tant than GMA in both areas. In recent years, most per­son­al­ity research has been orga­nized around the Big Five model of per­son­al­ity (Gold­berg, 1990) …As indi­cated ear­lier, Judge et al. (1999) found that three of the Big Five per­son­al­ity traits mea­sured in child­hood pre­dicted adult occu­pa­tional level and income. For Con­sci­en­tious­ness, these lon­gi­tu­di­nal cor­re­la­tions were .49 and .41, respec­tive­ly; these val­ues are only slightly smaller than the cor­re­spond­ing cor­re­la­tions in this study for GMA (dis­cussed in the Lon­gi­tu­di­nal Stud­ies sec­tion, above) of .51 and .53, respec­tive­ly. For Open­ness to Expe­ri­ence (which cor­re­lates pos­i­tively with GMA), the cor­re­la­tions were .32 and .26. Final­ly, Neu­roti­cism pro­duced lon­gi­tu­di­nal cor­re­la­tions of -.26 and -.34, for occu­pa­tional level and income, respec­tive­ly. Because of the unique nature of Judge et al.’s (1999) study, we con­ducted addi­tional analy­ses of the data from this study. Because occu­pa­tional level and income were highly cor­re­lated (r ϭ .83) and loaded on the same fac­tor, we com­bined them into one equally weighted mea­sure of career suc­cess. After cor­rect­ing for the bias­ing effects of mea­sure­ment error, we found that the three Big Five per­son­al­ity traits pre­dicted this index of career suc­cess with a (shrunk­en) mul­ti­ple cor­re­la­tion of .56. It is inter­est­ing to exam­ine the stan­dard­ized regres­sion weights (be­tas). For Neu­roti­cism, ␤ ϭ -.05 (SE ϭ .096); for Open­ness, ␤ ϭ .16 (SE ϭ .10); and for Con­sci­en­tious­ness, ␤ ϭ .44 (SE ϭ .123). Hence, in the regres­sion equa­tion, Con­sci­en­tious­ness is by far the most impor­tant per­son­al­ity vari­able, and Neu­roti­cism appears to have lit­tle impact after con­trol­ling for the other two per­son­al­ity traits. How­ev­er, it is also impor­tant to con­trol for the effects of GMA. When GMA is added to the regres­sion equa­tion, the (shrunk­en) mul­ti­ple cor­re­la­tion rises to .63. Again, it is instruc­tive to exam­ine the beta weights: Neu­roti­cism, ␤ ϭ -.05 (SE ϭ .096); Open­ness, ␤ ϭ -.03 (SE ϭ .113); Con­sci­en­tious­ness, ␤ ϭ .27 (SE ϭ .128); and GMA, ␤ ϭ .43 (SE ϭ .117). From these fig­ures, it appears that the bur­den of pre­dic­tion is borne almost entirely by GMA and Con­sci­en­tious­ness, with GMA being 59% more impor­tant than Con­sci­en­tious­ness (i.e., .43/.27 ϭ 1.59). In fact, when only GMA and Con­sci­en­tious­ness are included in the regres­sion equa­tion, the (shrunk­en) mul­ti­ple cor­re­la­tion remains the same, at .63. The stan­dard­ized regres­sion weights are then .29 for Con­sci­en­tious­ness (SE ϭ .102) and .41 for GMA (SE ϭ .096). These analy­ses sug­gest that Con­sci­en­tious­ness may be the only per­son­al­ity trait that con­tributes to career suc­cess. …The best meta-an­a­lytic esti­mate for the valid­ity of Con­sci­en­tious­ness, mea­sured with a reli­able scale, for pre­dict­ing job per­for­mance is .31 (Mount & Bar­rick, 1995). Hence, the valid­ity of GMA is 60% to 80% larger (de­pend­ing on the GMA valid­ity esti­mate used) than that of Con­sci­en­tious­ness. How­ev­er, Con­sci­en­tious­ness mea­sures con­tribute to valid­ity over and above the valid­ity of GMA, because the two are uncor­re­lated (Schmidt & Hunter, 1998). As noted above, Hunter and Hunter (1984) esti­mated the valid­ity of GMA for medium com­plex­ity jobs (63% of all jobs) to be .51. The mul­ti­ple cor­re­la­tion pro­duced by use of mea­sures of both GMA and Con­sci­en­tious­ness in a regres­sion equa­tion for such jobs is .60, an 18% increase in valid­ity over that of GMA alone (Schmidt & Hunter, 1998). The best meta-an­a­lytic esti­mate of the valid­ity of Con­sci­en­tious­ness for pre­dict­ing per­for­mance in job train­ing is .30 (Mount & Bar­rick, 1995). The mul­ti­ple cor­re­la­tion pro­duced by simul­ta­ne­ous use of GMA and Con­sci­en­tious­ness mea­sures is .65 (vs. .56 for GMA alone; Schmidt & Hunter, 1998).

, Kell et al 2013

Youth iden­ti­fied before age 13 (N = 320) as hav­ing pro­found math­e­mat­i­cal or ver­bal rea­son­ing abil­i­ties (top 1 in 10,000) were tracked for nearly three decades. Their awards and cre­ative accom­plish­ments by age 38, in com­bi­na­tion with spe­cific details about their occu­pa­tional respon­si­bil­i­ties, illu­mi­nate the mag­ni­tude of their con­tri­bu­tion and pro­fes­sional stature. Many have been entrusted with oblig­a­tions and resources for mak­ing crit­i­cal deci­sions about indi­vid­ual and orga­ni­za­tional well-be­ing. Their lead­er­ship posi­tions in busi­ness, health care, law, the pro­fes­so­ri­ate, and STEM (science, tech­nol­o­gy, engi­neer­ing, and math­e­mat­ics) sug­gest that many are out­stand­ing cre­ators of mod­ern cul­ture, con­sti­tut­ing a pre­cious human-cap­i­tal resource. Iden­ti­fy­ing truly pro­found human poten­tial, and fore­cast­ing differ­en­tial devel­op­ment within such pop­u­la­tions, requires assess­ing mul­ti­ple cog­ni­tive abil­i­ties and using atyp­i­cal mea­sure­ment pro­ce­dures. This study illus­trates how ulti­mate cri­te­ria may be aggre­gated and lon­gi­tu­di­nally sequenced to val­i­date such mea­sures.

Table 1 reveals the rich­ness and scope of par­tic­i­pants’ activ­i­ties. One indi­ca­tion of the cal­iber of their con­tri­bu­tions is the pres­tige of the orga­ni­za­tions that have awarded them grants. The data on cre­ative accom­plish­ments speak for them­selves, but a few sum­mary remarks are in order. In the arts and human­i­ties, 24 indi­vid­u­als had pro­duced 128 cre­ative writ­ten works (e.g., poems, nov­els, ref­er­eed pub­li­ca­tion­s), an aver­age of 5.3 accom­plish­ments per indi­vid­ual. In the same domain, 52 peo­ple had pro­duced 1,069 achieve­ments in the fine arts (e.g., music, sculp­ture), an aver­age of 20.6 accom­plish­ments per per­son. STEM achieve­ments are also note­wor­thy. Fifty-nine indi­vid­u­als had pro­duced ref­er­eed STEM pub­li­ca­tions, in areas rang­ing from bio­chem­istry to engi­neer­ing; the total num­ber of STEM pub­li­ca­tions pro­duced was 392 (6.6 per per­son). In the case of soft­ware devel­op­ment and patents, 117 peo­ple had made 820 con­tri­bu­tions, an aver­age of 7 per indi­vid­ual. Thir­ty-one indi­vid­u­als had received more than $25 mil­lion in grants, an aver­age of $825,635 per per­son. The tally of awards and sig­nifi­cant accom­plish­ments for these 320 indi­vid­u­als was 2,749, or an aver­age of 8.6 per per­son.

…enough infor­ma­tion is pro­vided to make clear that a num­ber of par­tic­i­pants are work­ing for world-class orga­ni­za­tions and hold impor­tant posi­tions of impact and respon­si­bil­ity in For­tune 500 com­pa­nies, tech­nol­o­gy, law, and med­i­cine. For the pro­fes­so­ri­ate in our sam­ple, Table 3 lists uni­ver­si­ties that either awarded them tenure or attracted them with tenure, plus some of their ref­er­eed pub­li­ca­tion out­lets. In total, 11.3% of par­tic­i­pants had earned tenure at accred­ited insti­tu­tions; 7.5% had tenure at research-in­ten­sive insti­tu­tions (Carnegie Foun­da­tion, 2010). This lat­ter per­cent­age is many, many times the base-rate expec­ta­tion, given the 2% base rate for doc­tor­ates in the United States and the fact that only a tiny frac­tion of the indi­vid­u­als with doc­tor­ates have tenure at research0in­ten­sive insti­tu­tions.

Although it would be diffi­cult to quan­tify par­tic­i­pants’ col­lec­tive accom­plish­ments in a sin­gle num­ber, by any stan­dard, it appears that many indi­vid­u­als iden­ti­fi­able by age 13 as hav­ing pro­found math­e­mat­i­cal and ver­bal rea­son­ing abil­ity develop into truly out­stand­ing con­trib­u­tors in their respec­tive fields. Not only did par­tic­i­pants choose pres­ti­gious occu­pa­tions by age 38 (Fig. 2 and Table 2), but the orga­ni­za­tions employ­ing them were impres­sive as well (Ta­bles 2 and 3). Although a num­ber of our data counts do not reflect the qual­ity of par­tic­i­pants’ con­tri­bu­tions, the orga­ni­za­tions employ­ing par­tic­i­pants (e.g., For­tune 500 com­pa­nies, major law firms, large med­ical facil­i­ties, and research uni­ver­si­ties) and bestow­ing awards on them (e.g., the U.S. Depart­ments of State and Jus­tice, the National Sci­ence Foun­da­tion, Intel Cor­po­ra­tion, NASA, and The Wall Street Jour­nal) afford rea­son­able qual­ity appraisals of their cre­ative prod­ucts as well as the respon­si­bil­i­ties, resources, and trust that they have earned. More than 7% of par­tic­i­pants held tenure at research-in­ten­sive uni­ver­si­ties (in­clud­ing many con­sid­ered the best in the world) by the time they were age 38. The 14 attor­neys were pre­dom­i­nantly work­ing in posi­tions of sig­nifi­cant respon­si­bil­ity for major firms or orga­ni­za­tions. The 19 physi­cians were also highly accom­plished: Seven were assis­tant pro­fes­sors, 2 were direc­tors of major pri­vate prac­tices, and 1 codi­rected a hos­pi­tal organ-trans­plant cen­ter serv­ing more than 3 mil­lion peo­ple. Rather than work­ing for estab­lished orga­ni­za­tions, 14 indi­vid­u­als founded com­pa­nies of their own. Two indi­vid­u­als were vice pres­i­dents at For­tune 500 com­pa­nies; 2 oth­ers were For­tune 500 senior hard­ware or soft­ware engi­neers. Sev­eral par­tic­i­pants were active in gov­ern­ment agen­cies at local and fed­eral lev­el­s-one advised the pres­i­dent of the United States on national pol­icy issues. Although par­tic­i­pants’ accom­plish­ments are impres­sive in vari­ety and scope, it is impor­tant to note the mag­ni­tude of indi­vid­ual differ­ences in out­put, even in this excep­tion­ally tal­ented sam­ple. Within sev­eral accom­plish­ment group­ings, some indi­vid­u­als far out­stripped their intel­lec­tual peers. For exam­ple, in the arts and human­i­ties, one indi­vid­ual pro­duced 500 musi­cal pro­duc­tions, account­ing for more than 57% of the musi­cal pro­duc­tions reported here; three indi­vid­u­als pro­duced 100 soft­ware con­tri­bu­tions each, or nearly 44% of the total report­ed. Seven par­tic­i­pants received more than $1 mil­lion in grant fund­ing each; col­lec­tive­ly, their fund­ing amounted to nearly $20 mil­lion, more than 77% of the total sam­ple’s grant fund­ing; one indi­vid­ual alone received $9 mil­lion in grant fund­ing. Final­ly, one per­son founded three com­pa­nies, and another was respon­si­ble for rais­ing more than $65 mil­lion in pri­vate equity invest­ment to fund his com­pa­ny. These find­ings mir­ror those in Gal­ton’s (1869/2006) inves­ti­ga­tion of the Cam­bridge Uni­ver­sity “wran­glers,” the 40 top-s­cor­ing stu­dents out of the approx­i­mately 100 hon­ors math­e­mat­ics grad­u­ates each year (400-450 stu­dents grad­u­ated from Cam­bridge annu­al­ly). Wran­glers were rank-ordered accord­ing to their scores on their final math­e­mat­ics exam (a 44-hr test spread over 8 days). Although being even a low-ranked wran­gler was enough for a grad­u­ate to obtain a fel­low­ship at a small col­lege, Gal­ton found that the high­est-ranked wran­gler tended to do more than twice as well on the final exam as the sec­ond-ranked wran­gler and approx­i­mately 4 times bet­ter than the low­est­-ranked wran­glers. Exam­in­ers empha­sized that the units of mea­sure­ment they employed were designed to index equal inter­vals, such that twice the score range trans­lated into approx­i­mately twice the knowl­edge. Such out­ly­ing indi­vid­ual differ­ences in accom­plish­ments, even among the most tal­ent­ed, are read­ily observed through­out his­tory (Mur­ray, 2003). This is one rea­son why O’Boyle and Agui­nis (2012) argued that, given the out­put of truly out­stand­ing per­form­ers, per­for­mance in gen­eral is bet­ter mod­eled through Paret­ian (power law) dis­tri­b­u­tions as opposed to Gauss­ian (nor­mal-curve) dis­tri­b­u­tions (Si­mon­ton, 1999a, 1999b).

Sipe and Curlette (1996) have found in their meta-syn­the­sis of edu­ca­tional research that on the indi­vid­ual level the effect of intel­li­gence on edu­ca­tional attain­ment was .6 (r = .5). The effects of other vari­ables (mo­ti­va­tion, SES, teacher edu­ca­tion, etc.) were small­er.

The inbreed­ing depres­sion had been cal­cu­lated by Schull and Neel (1965) from 1854 cousin mar­riages in Japan on the WISC and showed an over­all 7.5 point decre­ment (0.50 SD) in the off­spring, with each sub­test show­ing a greater or lesser amount. There is no non-ge­netic expla­na­tion for why Black­-White differ­ences in the US should be more pro­nounced on those sub­tests show­ing the most inbreed­ing depres­sion among the Japan­ese in Japan (Jensen also demon­strated inbreed­ing depres­sion effects on the Raven Matri­ces in India; Agrawal, Sin­ha, & Jensen, 1984).

Miller 2012, Sin­gu­lar­ity Ris­ing:

g - the let­ter used as short­hand for gen­eral men­tal abil­ity - is, in the words of Linda Got­tfred­son, a pro­lific scholar of human intel­li­gence, “prob­a­bly the best mea­sured and most stud­ied human trait in all of psy­chol­o­gy.” [Got­tfred­son, Linda S. 2002. “Where and Why g Mat­ters: Not a Mys­tery.” Human Per­for­mance 15 (1/2): 25-46. ]

econ­o­mist Garett Jones notes, “Across thou­sands of stud­ies on the cor­re­la­tion across men­tal abil­i­ties across pop­u­la­tions, no one has yet found a reli­able neg­a­tive cor­re­la­tion [be­tween per­for­mances on two differ­ent com­plex men­tal tasks]”; [117. Jones, Garett. 2011b. “National IQ and National Pro­duc­tiv­i­ty: The Hive Mind Across Asia.” Asian Devel­op­ment Review 28 (1): 51-71 ]

Some might chal­lenge IQ’s impor­tance by claim­ing that all chil­dren have about the same aca­d­e­mic poten­tial, but for social rea­sons we feel the need to grade and rank chil­dren even though the rank­ings arise from differ­ences that are very small. But differ­ences in IQ cor­re­late with starkly dis­sim­i­lar lev­els of real aca­d­e­mic per­for­mance. (One strik­ing piece of evi­dence is that “the nineti­eth per­centile of nine-year-olds . . . per­forms in read­ing, math, and sci­ence at the level of the twen­ty-fifth per­centile of sev­en­teen-year-olds.”[120. Got­tfred­son, Linda S. 2005. “Sup­press­ing Intel­li­gence Research: Hurt­ing Those We Intend to Help,” In Rogers H. Wright and Nicholas A. Cum­mings (ed­s.), Destruc­tive Trends in Men­tal Health: The Well-In­ten­tioned Path to Harm. New York: Tay­lor and Fran­cis, 155-86. . Foot­note omit­ted.] Because schools sort by age rather than abil­i­ty, we don’t find smart nine-year-olds in higher grades than not-so-s­mart sev­en­teen-year-olds, even when the for­mer are more capa­ble than the lat­ter.)

IQ tests taken by chil­dren have been found to go a long way toward pre­dict­ing life span.[126. Got­tfred­son, Lin­da. S., and Ian J. Deary. 2004. “Intel­li­gence Pre­dicts Health and Longevi­ty, But Why?” Cur­rent Direc­tions in Psy­cho­log­i­cal Sci­ence 13 (1): 1-4. ] High­er-IQ indi­vid­u­als also have bet­ter den­tal health, even when con­trol­ling for income and eth­nic­i­ty.[S­ab­bah, Wael, and Aubrey Shei­ham. 2010. “The Rela­tion­ships Between Cog­ni­tive Abil­ity and Den­tal Sta­tus in a National Sam­ple of USA Adults”, Intel­li­gence 38 (6): 605-10] We don’t under­stand the causes of the health-IQ rela­tion­ship, but plau­si­ble expla­na­tions include child­hood ill­nesses that may both reduce a child’s IQ and shorten his life span and the pos­si­bil­ity that high­er-IQ indi­vid­u­als make bet­ter health deci­sions, get into fewer acci­dents, choose to live in a health­ier envi­ron­ment, fol­low doc­tors’ advice bet­ter, and more often fol­low direc­tions when tak­ing med­ica­tion. Genet­ics would also explain part of the cor­re­la­tion if the same genes that con­ferred high intel­li­gence also boosted longevi­ty.[128. Got­tfred­son and Deary (2004).] The pos­i­tive rela­tion­ship between a man’s semen qual­ity and his IQ sup­ports the the­ory that genes play a role in the cor­re­la­tion.[Ar­den, Ros­alind, Linda S. Got­tfred­son, Geoffrey Miller, and Arand Pierce. 2009. “Intel­li­gence and Semen Qual­ity are Pos­i­tively Cor­re­lat­ed.” Intel­li­gence 37 (3): 277-82] Com­pound­ing IQs impact on inequal­i­ty, high­er-IQ peo­ple tend to be more phys­i­cally attrac­tive.[Kanaza­wa, Satoshi 2011. “Intel­li­gence and Phys­i­cal Attrac­tive­ness”. Intel­li­gence 39 (1): 7-14 ] Fur­ther­more, it’s pos­si­ble to make a decent guess at peo­ple’s intel­li­gence just by look­ing at them. From an arti­cle in the online mag­a­zine Slate : “In 1918, a researcher in Ohio showed a dozen pho­to­graphic por­traits of well-dressed chil­dren to a group of physi­cians and teach­ers, and asked the adults to rank the kids from smartest to dumb­est. A cou­ple of years lat­er, a Pitts­burgh psy­chol­o­gist ran a sim­i­lar exper­i­ment using head­shots of 69 employ­ees from a depart­ment store. In both stud­ies, seem­ingly naive guesses were com­pared to actual test scores and turned out to be accu­rate more often than not.” Stare at a com­puter screen until a big green ball appears, and then hit the space bar as quickly as you can. You have just taken a par­tially reli­able IQ test, since a per­son’s IQ has a pos­i­tive cor­re­la­tion with reac­tion time.[132. Got­tfred­son (2002).]

…mag­netic res­o­nance imag­ing show­ing a strong pos­i­tive cor­re­la­tion between brain size and IQ[136. Jones (2011b).]

A per­son’s IQ is large­ly, but not com­plete­ly, deter­mined by age eight.[Heck­man, James, Jora Stixrud, and Ser­gio Urzua. 2006. “The Effects of Cog­ni­tive and Noncog­ni­tive Abil­i­ties on Labor Mar­ket Out­comes and Social Behav­ior”. Jour­nal of Labor Eco­nom­ics 24 (3): 41 1-82 ] Tests given to infants mea­sur­ing how much atten­tion the infant pays to novel pic­tures have a pos­i­tive cor­re­la­tion with the IQ the infant will have at age twen­ty-one.[Hunt, Earl. 2011 . Human Intel­li­gence. Cam­bridge: Cam­bridge Uni­ver­sity Press ] The Scot­tish Men­tal Sur­vey of 1932 has helped show the remark­able sta­bil­ity of a per­son’s IQ across his adult life.[Deary, Ian, Martha C. White­man, John M. Starr, Lawrence J. Whal­ley, and Helen C. Fox 2004. “The Impact of Child­hood Intel­li­gence on Later Life: Fol­low­ing Up the Scot­tish Men­tal Sur­veys of 1932 and 1947”. Jour­nal of Per­son­al­ity and Social Psy­chol­ogy 86 (1): 130-47 ] On June 1, 1932, almost every child in Scot­land born in 1921 took the same men­tal test. Over sixty years lat­er, researchers tracked down some of the test tak­ers who lived in one par­tic­u­lar part of Scot­land and gave them the test they took in 1932. The researchers found a strong cor­re­la­tion between most peo­ple’s 1932 and recent test results.

IQ is the sin­gle best pre­dic­tor of job per­for­mance. [Got­tfred­son, Linda S. 1997. “Why g Mat­ters: The Com­plex­ity of Every­day Life.” Intel­li­gence 24 (1): 79–132]

The effect of IQ on wages might be mit­i­gated by the pos­si­bil­ity that some high­-IQ peo­ple are drawn to rel­a­tively low-pay­ing pro­fes­sions. Con­sid­er, for exam­ple, Ter­ence Tao, a rea­son­able can­di­date for the smartest per­son alive today. Ter­ence works as a math pro­fes­sor, and accord­ing to Wikipedia, his great­est accom­plish­ment to date is coau­thor­ing a the­o­rem on prime num­bers. Prime num­ber research does­n’t pay well, but you can’t do it, and would­n’t find it inter­est­ing, unless you had a super-ge­nius level IQ. Sim­i­lar­ly, a poet with an extremely high IQ might have become a lawyer had her IQ been a bit lower because then she would­n’t have under­stood the sub­tleties of poetry that drew her into a poorly remu­ner­ated pro­fes­sion. I sus­pect that many math pro­fes­sors and poets would have higher incomes if some brain injury low­ered their IQs just enough to force them out of their pro­fes­sions. If you have an IQ of 135, then you’re already smarter than 99% of human­i­ty. [I’m assum­ing an IQ stan­dard devi­a­tion of 15 for this and all other IQ cal­cu­la­tions used in this book.] Would you do bet­ter in life if your IQ went well above 135? Research by Heck­man says yes. He found that among men with IQs in the top 1% of the pop­u­la­tion, hav­ing a higher IQ boosts wages through­out one’s entire work­ing life, and this effect exists even after tak­ing into account an indi­vid­u­al’s level of edu­ca­tion.[­Gen­sowski, Miri­am, James J. Heck­man, and Peter Save­lyev. 2011. “The Effects of Edu­ca­tion, Per­son­al­i­ty, and IQ on Earn­ings of High­-A­bil­ity Men” Work­ing paper [See also for exam­ple SMPY results like Kell et al 2013 ]]

A researcher at the Lon­don School of Eco­nom­ics has even shown that one-fourth of the differ­ences in wealth between differ­ent US states can be explained by differ­ences in the aver­age IQ of their pop­u­la­tion.[Kanaza­wa, Satoshi. 2006. “IQ and the Wealth of States.” Intel­li­gence 34 (6): 593-600. ]

a British study showed that IQ tests given to a group of ten and eleven-year-olds strongly cor­re­lated with the level of trust these sub­jects expressed when they became adult­s.[S­tur­gis, Patrick, Sanna Read, and Nick Allum. 2010. “Does Intel­li­gence Fos­ter Gen­er­al­ized Trust? An Empir­i­cal Test Using the UK Birth Cohort Stud­ies”. Intel­li­gence 38 (1): 45-54 ]

Hav­ing a low IQ makes you, on aver­age, more dis­posed to crime.[Beaver, Kevin M., and John Paul Wright. 2011. “The Asso­ci­a­tion between Coun­ty-Level IQ and Coun­ty-Level Crime Rates”. Intel­li­gence 39 (1): 22-26 ] Since crime reduces pro­duc­tive eco­nomic activ­i­ty, this is another means by which high IQ con­tributes to eco­nomic growth. The higher a per­son’s IQ, the more likely he is to sup­port eco­nomic poli­cies that most econ­o­mists con­sider to be healthy for a nation’s econ­o­my.[­Ca­plan, Bryan, and Stephen C. Miller. 2010. “Intel­li­gence Makes Peo­ple Think Like Econ­o­mists: Evi­dence from the Gen­eral Social Sur­vey”. Intel­li­gence 38 (6): 636-47 ] Con­se­quent­ly, if you live in a democ­racy and trust econ­o­mists’ judg­ments, you should want your fel­low vot­ers to be smart. Hav­ing a high IQ makes you more long-term ori­ent­ed.[Warn­er, John T, and Saul Fleeter. 2001. ] Future-ori­ented peo­ple are more likely to make the kind of cal­cu­lated long-term invest­ments crit­i­cal to eco­nomic growth.

“Intel­li­gence makes peo­ple think like econ­o­mists: Evi­dence from the Gen­eral Social Sur­vey”, Caplan & Miller 2010:

Using data from the Gen­eral Social Sur­vey (GSS), we show that the esti­mated effect of edu­ca­tion sharply falls after con­trol­ling for intel­li­gence. In fact, edu­ca­tion is dri­ven down to sec­ond place, and intel­li­gence replaces it at the top of the list of vari­ables that make peo­ple “think like econ­o­mists.” Thus, to a fair degree edu­ca­tion is proxy for intel­li­gence, though there are some areas-in­ter­na­tional eco­nom­ics in par­tic­u­lar-where edu­ca­tion still dom­i­nates. An impor­tant impli­ca­tion is that the polit­i­cal exter­nal­i­ties of edu­ca­tion may not be as large as they ini­tially appear.

, Carl & Bil­lari 2014:

An exten­sive empir­i­cal lit­er­a­ture has estab­lished that gen­er­al­ized trust is an impor­tant aspect of civic cul­ture. It has been linked to a vari­ety of pos­i­tive out­comes at the indi­vid­ual lev­el, such as entre­pre­neur­ship, vol­un­teer­ing, self­-rated health, and hap­pi­ness. How­ev­er, two recent stud­ies have found that it is highly cor­re­lated with intel­li­gence, which raises the pos­si­bil­ity that the other rela­tion­ships in which it has been impli­cated may be spu­ri­ous. Here we repli­cate the asso­ci­a­tion between intel­li­gence and gen­er­al­ized trust in a large, nation­ally rep­re­sen­ta­tive sam­ple of U.S. adults. We also show that, after adjust­ing for intel­li­gence, gen­er­al­ized trust con­tin­ues to be strongly asso­ci­ated with both self­-rated health and hap­pi­ness. In the con­text of sub­stan­tial vari­a­tion across coun­tries, these results bol­ster the view that gen­er­al­ized trust is a valu­able social resource, not only for the indi­vid­ual but for the wider soci­ety as well.

…Two recent stud­ies have doc­u­mented a strong cor­re­la­tion between gen­er­al­ized trust and intel­li­gence [16], [17]. Stur­gis et al. [“Does intel­li­gence fos­ter gen­er­al­ized trust? An empir­i­cal test using the UK birth cohort stud­ies”] analyse data from the U.K., and show that intel­li­gence at age 10-11 pre­dicts gen­er­al­ized trust at age 34, even after con­di­tion­ing on a large num­ber of socio-e­co­nomic vari­ables, includ­ing self­-rated health and hap­pi­ness. Sim­i­lar­ly, Hooghe et al 2012 [“The cog­ni­tive basis of trust. The rela­tion between edu­ca­tion, cog­ni­tive abil­i­ty, and gen­er­al­ized and polit­i­cal trust”] exam­ine Dutch data, and find that a large part of the asso­ci­a­tion between gen­er­al­ized trust and edu­ca­tion is accounted for by cog­ni­tive abil­i­ty.

Hooghe et al 2012:

Pre­vi­ous stud­ies - mainly based on UK data - indeed show a pos­i­tive rela­tion between intel­li­gence and gen­er­al­ized and polit­i­cal trust (Ya­m­ag­ishi, 2001; Yam­ag­ishi, Kikuchi & Kosugi, 1999; Stur­gis, Read & Allum, 2010, p. 52; Schoon & Cheng, 2011; Schoon et al., 2010). In this line of rea­son­ing, Deary, Batty and Gale (2008, p. 1) stat­ed: “bright chil­dren become enlight­ened adults”.

  • Yam­ag­ishi, T. (2001). “Trust as a form of social intel­li­gence”, pp. 121-147 in K. Cook (ed.), Trust in soci­ety. New York: Rus­sell Sage Foun­da­tion
  • Yam­ag­ishi, T., Kikuchi, M. & Kosugi, M. (1999). “Trust, gulli­bil­ity and social intel­li­gence”. Asian Jour­nal of Social Psy­chol­ogy, 2(1), 145-161
  • Stur­gis, P., Read, S. & N. Allum (2010). “Does intel­li­gence fos­ter gen­er­al­ized trust? An empir­i­cal test using the UK birth cohort stud­ies”, Intel­li­gence, 38(1), 45-54
  • Schoon, I. & H. Cheng (2011). “Deter­mi­nants of polit­i­cal trust. A life-long learn­ing model”. Devel­op­men­tal Psy­chol­ogy, 47(3), 619-631
  • Schoon, I., Cheng, H., Gale, C., Bat­ty, D. & Deary, I. (2010). “Social sta­tus, cog­ni­tive abil­i­ty, and edu­ca­tional attain­ment as pre­dic­tors of lib­eral social atti­tudes and polit­i­cal trust”. Intel­li­gence, 38(1), 144-150

“Sup­press­ing Intel­li­gence Research: Hurt­ing Those We Intend to Help”, Got­tfred­son 2005

The results of a 1984 sur­vey (Sny­der­man & Roth­man, 1988) of experts on intel­li­gence and men­tal test­ing there­fore sur­prised even Jensen. The experts’ modal response on every ques­tion that involved the “hereti­cal” con­clu­sions from Jensen’s 1969 arti­cle was the same as his (Jensen, 1998, p. 198). (The experts’ mean response over­es­ti­mated test bias, how­ev­er, because there is none against blacks or lower social class indi­vid­u­als; Jensen, 1980; Neisser et al., 1996; Sny­der­man & Roth­man, 1988, p. 134; Wig­dor & Gar­ner, 1982). Here in abbre­vi­ated form are the sur­vey’s major ques­tions and the 600 experts’ respons­es.

  • Q: What are the impor­tant ele­ments of intel­li­gence?

  • A: “Near una­nim­ity” (96-99%) for abstract think­ing or rea­son­ing, prob­lem solv­ing abil­i­ty, and capac­ity to acquire knowl­edge (p. 56).

  • Q: Is intel­li­gence best described as a sin­gle gen­eral fac­tor with sub­sidiaries or as sep­a­rate fac­ul­ties?

  • A: A gen­eral fac­tor (58%, or 67% of those respond­ing; p. 71).

  • Q: What her­i­tabil­ity would you esti­mate for IQ differ­ences within the white pop­u­la­tion?

  • A: Aver­age esti­mate of 57% (p. 95).

  • Q: What her­i­tabil­ity would you esti­mate for IQ differ­ences within the black pop­u­la­tion?

  • A: Aver­age esti­mate of 57% (p. 95).

  • Q: Are intel­li­gence tests biased against blacks?

  • A: On a scale of 1 (not at all or insignifi­cant­ly) to 4 (ex­treme­ly), mean response of 2 (some­what, p. 117).

  • Q: Are intel­li­gence tests biased against lower social class indi­vid­u­als?

  • A: On a scale of 1 (not at all or insignifi­cant­ly) to 4 (ex­treme­ly), mean response of 2 (some­what, p. 118).

  • Q: What is the source of aver­age social class differ­ences in IQ?

  • A: Both genetic and envi­ron­men­tal (55%, or 65% of those respond­ing; p. 126).

  • Q: What is the source of the aver­age black­-white differ­ence in IQ?

  • A: Both genetic and envi­ron­men­tal (45%, or 52% of those respond­ing; p. 128).

The sup­pos­edly fringe sci­en­tist, Jensen, was actu­ally in the main­stream because the main­stream had silently come to him, where it remains today (Got­tfred­son, 1997a). Mean­while, pub­lic opin­ion was still being pushed in the oppo­site direc­tion, cre­at­ing an ever greater gulf between received opin­ion and sci­en­tifi­cally informed thought.

…And why keep silent when the media pro­mul­gate clear false­hoods as sci­en­tific truth­s-e­spe­cially when, as Sny­der­man and Roth­man (1988) demon­strat­ed, the media por­tray expert opin­ion on intel­li­gence as the oppo­site of what it really is?

Early in my career I reported that bright boys who had attended a school for dyslex­ics did not enter the usual high­-level jobs (med­i­cine, law, sci­ence, and col­lege teach­ing) but had nev­er­the­less suc­ceeded at a high level by enter­ing pres­ti­gious or remu­ner­a­tive occu­pa­tions that required above-av­er­age intel­li­gence but rel­a­tively lit­tle read­ing or writ­ing, specifi­cal­ly, top man­age­ment and sales posi­tions. A col­league accused me in that sem­i­nar of say­ing that “blacks can’t make it because they are dumb.”

The best informed, who are often called upon for expert com­ment, can­not endorse clear false­hoods with­out jeop­ar­diz­ing their own stand­ing within the dis­ci­pline, but they some­times dis­pute minor issues in a man­ner that the unin­formed mis­take for whole­sale repu­di­a­tion (Got­tfred­son, 1994a; Page, 1972).

The impli­ca­tion of ABC’s Novem­ber 22, 1994, national news­cast was surely not lost on view­ers when, while expos­ing the sup­pos­edly unsa­vory his­tory of intel­li­gence research behind The Bell Curve, news anchor Peter Jen­nings fol­lowed pho­tographs of Jensen and other sup­posed race sci­en­tists with footage of Nazi sol­diers and what appeared to be death camp doc­tors and pris­on­ers.

Crit­ics have asso­ci­ated a belief in the hered­i­tary basis of intel­li­gence with evil intent so fre­quently and for so long that merely men­tion­ing “IQ” is enough to trig­ger in many minds the words “pseu­do­science,” “racism,” and “geno­cide.” Even cur­rent APA pres­i­dent Robert Stern­berg keeps the mali­cious asso­ci­a­tion alive by reg­u­larly ridi­cul­ing and belit­tling empir­i­cal­ly-minded intel­li­gence researchers (e.g., com­par­ing Jensen, in a book meant to honor him, to a child who would not grow up; Stern­berg, 2003), refer­ring to their work as “qua­si­-science” (“Sci­ence and pseu­do­science,” 1999, p. 27) that has “recre­ated a kind of night of the liv­ing dead” (Stern­berg, 1997, p. 55), and sprin­kling his descrip­tions of it with men­tions of racism, slav­ery, and even Soviet tyranny (e.g., Stern­berg, 2003; see also Stern­berg, 2000, Stern­berg & Wag­n­er, 1993). But why should we assume that a belief in the her­i­tabil­ity of many human differ­ences is dan­ger­ous and a belief in man’s infi­nite mal­leabil­ity is not? Crit­ics have yet to explain. Why is the for­mer belief always yoked to Hitler, but the lat­ter never to Stal­in, who out­lawed both intel­li­gence tests and genetic think­ing? Stalin killed at least as many as did Hitler in his effort to reshape the Soviet cit­i­zenry (Cour­tois, 1999).

Behav­ior geneti­cists dis­tin­guish between two types of envi­ron­men­tal influ­ence: shared and non-shared (also called between-fam­ily and with­in-fam­ily effect­s). Shared influ­ences are those that make sib­lings more alike. Pos­si­ble such influ­ences would include parental income, edu­ca­tion, child-rear­ing style, and the like, because they would impinge on all sib­lings in a house­hold. Non-shared influ­ences are those that affect indi­vid­u­als one per­son at a time and there­fore make sib­lings less alike. Lit­tle is yet known about them, but they might include ill­ness, acci­dents, non-ge­netic influ­ences on fetal devel­op­ment, and the con­cate­na­tion of unique expe­ri­ences. To the great sur­prise even of behav­ior geneti­cists, shared envi­ron­men­tal effects on intel­li­gence (within the broad range of typ­i­cal envi­ron­ments) wash away by late ado­les­cence. IQ differ­ences can be traced to both genes (40%) and shared envi­ron­ments (25%) in early child­hood, but genetic effects increase in impor­tance with age (to 80% in adult­hood) while shared effects dis­si­pate (Plom­in, DeFries, McClearn, & McGuffin, 2001). For exam­ple, adop­tive sib­lings end up no more alike in IQ or per­son­al­ity by ado­les­cence than are ran­dom strangers, and instead become sim­i­lar to the bio­log­i­cal rel­a­tives they have never met.

Cur­rently one of the biggest puz­zles for fam­ily effects the­ory is that aca­d­e­mic achieve­ment gaps do not nar­row even in set­tings where all the sup­pos­edly impor­tant envi­ron­men­tal resources are present (Banchero & Lit­tle, 2002). For exam­ple, its adher­ents are now argu­ing among them­selves (Lee, 2002) about the proper cul­tural expla­na­tion for the large black­-white achieve­ment gaps that per­sist in the most socioe­co­nom­i­cally advan­taged, inte­grat­ed, lib­er­al, sub­ur­ban school dis­tricts in the United States, such as Shaker Heights, Ohio (Og­bu, 2003) and Berke­ley, Cal­i­for­nia (Noguera, 2001). More­over, black­-white test score gaps (IQ, SAT, etc.) tend to be larger at higher socioe­co­nomic lev­els. This find­ing con­tra­dicts the pre­dic­tions of fam­ily effects the­o­ry. It is con­sis­tent with g-based the­o­ry, how­ev­er, because the lat­ter pre­dicts that black and white chil­dren of high­-IQ par­ents will regress part way from their par­ents’ mean toward differ­ent pop­u­la­tion means, IQ 100 for whites and IQ 85 for blacks.

The fic­tions about intel­li­gence essen­tially deny that it exists, which vir­tu­ally no one really believes. Many peo­ple just want a more “demo­c­ra­tic” view of it. Not sur­pris­ing­ly, psy­chol­o­gy’s sup­ply has risen to meet pub­lic demand, and the new egal­i­tar­ian per­spec­tives on human intel­li­gence were instantly blessed by opin­ion mak­ers. Chief among them are the “mul­ti­ple intel­li­gence” the­o­ries by psy­chol­o­gists Howard Gard­ner (1983, 1998) and Robert Stern­berg (1997). The eager accep­tance of their the­o­ries by edu­ca­tors, psy­chol­o­gists, and oth­ers has occurred despite nei­ther of them pro­vid­ing cred­i­ble evi­dence that their pro­posed intel­li­gences actu­ally exist, that is, as inde­pen­dent abil­i­ties of com­pa­ra­ble gen­er­al­ity and prac­ti­cal impor­tance to g. Gard­ner has rejected even mea­sur­ing his eight intel­li­gences, let alone demon­strat­ing that they pre­dict any­thing (Hunt, 2001; Lubin­ski & Ben­bow, 1995). Study-by-s­tudy dis­sec­tions of Stern­berg’s mul­ti­ple-in­tel­li­gence research pro­gram reveal no such evi­dence (Brody, 2003a, b; Got­tfred­son, 2003a, c). If any­thing, they con­firm that all three of his pro­posed intel­li­gences are just differ­ent fla­vors of g itself, as prob­a­bly are most of Gard­ner’s too (Car­roll, 1993, p. 641).

In 1991, the U.S. Con­gress voted over­whelm­ingly to out­law race-norm­ing in employ­ment after it learned that the Labor Depart­ment had already been race-norm­ing its employ­ment tests for a decade and that the U. S. Equal Employ­ment Oppor­tu­nity Com­mis­sion (EEOC) had started threat­en­ing pri­vate employ­ers if they did not adopt the “sci­en­tifi­cal­ly-jus­ti­fied” prac­tice. The racial pref­er­ences that race-norm­ing entails are hardly triv­ial. What the NRC report did not say was that blacks scor­ing at the 15th per­centile in skill level on DOL’s test would have been judged equal to whites and Asians scor­ing at the 50th per­centile, and blacks at the 50th per­centile would be rated com­pa­ra­bly skilled as whites and Asians at the 84th (Blits & Got­tfred­son, 1990a). Sel­dom being apprised of such facts, most peo­ple greatly under­es­ti­mate how dis­crepant the pools of qual­i­fied appli­cants are from which racial bal­ance is sup­posed to emerge. Another illus­tra­tion, per­ti­nent to the next exam­ple, is that about 75% of whites vs. only 28% of blacks exceed the min­i­mum IQ level (~IQ 91)-a ratio of 3 to 1-usu­ally required for min­i­mally sat­is­fac­tory per­for­mance in the skilled trades, fire and police work, and mid-level cler­i­cal jobs such as bank teller (Got­tfred­son, 1986, pp. 400-401). The poten­tial pools become increas­ingly racially lop­sided for more cog­ni­tively demand­ing jobs. Work­ers in pro­fes­sional jobs such as engi­neer, lawyer, and physi­cian typ­i­cally need an IQ of at least 114 to per­form sat­is­fac­to­ri­ly. About 23% of whites but only 1% of blacks exceed this min­i­mum….De­vel­op­ing tests that mea­sure cog­ni­tive skills more effec­tively tends only to worsen the pro­scribed dis­parate impact. Adding rel­e­vant non-cog­ni­tive pre­dic­tors to the mix does lit­tle to reduce the racial imbal­ance (Schmitt, Rogers, Chan, Shep­pard, & Jen­nings, 1997).

The police selec­tion test devel­oped in 1994 for Nas­sau Coun­ty, NY, rep­re­sents one such “tech­ni­cal advance.” The 10 mem­bers of a joint Nas­sau Coun­ty-U.S. Depart­ment of Jus­tice (DOJ) team had set out to develop a police selec­tion test with less dis­parate impact (more racially bal­anced result­s). The county had not been able to sat­isfy the DOJ’s employ­ment dis­crim­i­na­tion unit in sev­eral tries under its var­i­ous con­sent decrees since 1977. (Re­call the 3 to 1 ratio given above for the pro­por­tion of whites vs. blacks exceed­ing the abil­ity level below which per­for­mance in police work tends to be unsat­is­fac­to­ry.) Seven of the team’s eight psy­chol­o­gists con­sti­tuted a Who’s Who of APA’s large Divi­sion 14 (In­dus­trial and Orga­ni­za­tional Psy­chol­o­gy), four of them hav­ing pre­vi­ously served as its pres­i­dent. Sev­eral years and mil­lions of dol­lars lat­er, this high­-pow­ered team claimed to have suc­ceeded in devel­op­ing a test that vir­tu­ally elim­i­nated dis­parate impact while simul­ta­ne­ously improv­ing selec­tion valid­i­ty. Water could run up-hill, after all. Once again, lead­ing psy­chol­o­gists found a seem­ingly sci­en­tific solu­tion to an intractable polit­i­cal-le­gal dilem­ma. DOJ imme­di­ately began press­ing other police juris­dic­tions nation­wide to replace their more “dis­crim­i­na­tory” tests with the new selec­tion bat­tery. A close look at the sev­er­al-vol­ume tech­ni­cal report for the Nas­sau test bat­tery revealed that the team had suc­ceeded in reduc­ing dis­parate impact by, in effect, ger­ry­man­der­ing the test to assess only traits on which the races differed lit­tle or not at all (Got­tfred­son, 1996a, b). The joint Nassau-DOJ team had admin­is­tered its nearly day-long, 25-part exper­i­men­tal bat­tery to all 25,000 appli­cants, but set­tled on the bat­tery’s final com­po­si­tion only after exam­in­ing the scores it yielded for differ­ent races. The exper­i­men­tal bat­tery was then appar­ently stripped of vir­tu­ally all parts demand­ing cog­ni­tive abil­i­ty. The only parts actu­ally used to rank appli­cants were eight non-cog­ni­tive per­son­al­ity scales (all com­mer­cial prod­ucts owned by mem­bers of the team) and being able to read above the 1st per­centile of cur­rently employed police offi­cers (near illit­er­a­cy). Selec­tion for cog­ni­tive com­pe­tence had been reduced to lit­tle more than the toss of a coin, despite the team’s own care­ful job analy­sis hav­ing shown that “rea­son­ing, judg­ment, and infer­en­tial think­ing” were the most crit­i­cal skills for good police work. The new police test was made to appear more valid than the coun­ty’s pre­vi­ous ones by, among other things, omit­ting key results required by legal and pro­fes­sional guide­li­nes, trans­form­ing the data in ways that arti­fi­cially reduced the appar­ent valid­ity of the cog­ni­tive sub­tests rel­a­tive to the non-cog­ni­tive ones, and mak­ing a series of sta­tis­ti­cal errors that more than dou­bled the final bat­tery’s appar­ent pre­dic­tive valid­ity (from .14 to .35). When exposed, the test cre­ated a scan­dal in Divi­sion 14 (“The Great Debate of 1997” in Hakel, 1997, p. 116), partly because other lead­ing selec­tion psy­chol­o­gists expected its use would pro­duce less effec­tive polic­ing and degrade pub­lic safety (Schmidt, 1996).

Even the most objec­tive, most care­fully vet­ted pro­ce­dures for iden­ti­fy­ing tal­ent are instantly pro­nounced guilty of bias or “exclu­sion” when they yield dis­parate impact in hir­ing, col­lege admis­sions, place­ment in gifted edu­ca­tion, and the like. Indeed, the very notions of objec­tiv­ity and merit are now under attack by influ­en­tial intel­lec­tual elites (Far­ber & Sher­ry, 1997). When faith­ful and fair appli­ca­tion of the law yields dis­parate impact in arrest or incar­cer­a­tion rates, Amer­i­can jurispru­dence must be con­sid­ered inher­ently racist (see argu­ments in Cren­shaw, Gotan­da, Peller, & Thomas, 1995). When earnest, socially lib­eral teach­ers fail to nar­row the stub­born achieve­ment gaps between races and class­es, they must be uncon­sciously dis­crim­i­na­tory and require diver­sity train­ing. Because Amer­i­can insti­tu­tions still rou­tinely and almost every­where fail to yield the desired racial bal­ance, the Amer­i­cans who cre­ated and sup­pos­edly con­trol those insti­tu­tion­s-ma­jor­ity Amer­i­can­s-must be judged deeply, uncon­scious­ly, invet­er­ately racist and to have cre­ated a soci­ety where appear­ances to the con­trary are just a smoke­screen to hide their built-in priv­i­leges. Under the equipo­ten­tial­ity fic­tion, there can be no other legit­i­mate expla­na­tion, and any attempt at one serves only to evade respon­si­bil­i­ty.

…Fewer but still many social sci­en­tists hold to a fourth false cre­do-that intel­li­gence has lit­tle or no func­tional util­i­ty, at least out­side schools. More­over, they often add that the advan­tages and dis­ad­van­tages of high or low IQ are mostly “socially con­structed” to serve the inter­ests of the priv­i­leged. This view was artic­u­lated in an influ­en­tial arti­cle pub­lished soon after Jensen’s 1969 arti­cle by econ­o­mists Samuel Bowles and Her­bert Gin­tis (1972/1973). They argued that higher IQ does not have any func­tional util­i­ty, even within schools, and that IQ tests are sim­ply a tool cre­ated by the upper classes to main­tain and jus­tify their priv­i­leges. They dis­missed talk of “objec­tiv­ity” and “merit” as just smoke blown to obscure this fact. Psy­chol­o­gist Robert Stern­berg implies much the same when he sug­gests that the g fac­tor dimen­sion of intel­lec­tual differ­ences is an arti­fact of West­ern school­ing (Stern­berg et al., 2000, p. 9) and that using cog­ni­tive tests such as the SAT to sort peo­ple is akin to the way slav­ery and reli­gious prej­u­dice were once used to keep dis­fa­vored groups down (Stern­berg, 2003).

How­ev­er, when crit­ics argue that IQ differ­ences have lit­tle or no func­tional mean­ing beyond that which cul­tures or their elites arbi­trar­ily attach to them for selfish pur­pos­es, they simul­ta­ne­ously turn atten­tion away from the very real prob­lems that lower intel­li­gence cre­ates for less able per­sons. As Her­rn­stein and Mur­ray (1994) note, the crit­ics gen­er­ally have lit­tle con­tact with the down­trod­den they would pro­tect. These bright opin­ion mak­ers may be liv­ing com­fort­ably with their fic­tions and benev­o­lent lies, but low­er-IQ indi­vid­u­als must live daily with the con­se­quences of their weaker learn­ing and rea­son­ing skills. Their dis­tant pro­tec­tors would seem to be the lim­ou­sine lib­er­als of intel­li­gence.

I focus below on every­day tasks that high­er-IQ indi­vid­u­als con­sider so sim­ple that they do not real­ize how such tasks might cre­ate obsta­cles to the well-be­ing of oth­ers less cog­ni­tively blessed.

Func­tional lit­er­acy and daily self­-main­te­nance. Cit­i­zens of lit­er­ate soci­eties take for granted that they are rou­tinely called upon to read instruc­tions, fill out forms, deter­mine best buys, deci­pher bus sched­ules, and oth­er­wise read and write to cope with the myr­iad details of every­day life. But such tasks are diffi­cult for many peo­ple. The prob­lem is sel­dom that they can­not read or write the words, but usu­ally that they are unable to carry out the men­tal oper­a­tions the task calls for-to com­pare two items, grasp an abstract con­cept, pro­vide com­pre­hen­si­ble and accu­rate infor­ma­tion about them­selves, fol­low a set of instruc­tions, and so on. This is what it means to have poor “func­tional lit­er­a­cy.” Func­tional lit­er­acy has been a major pub­lic pol­icy con­cern, as illus­trated by the U.S. Depart­ment of Edu­ca­tion’s var­i­ous efforts to gauge its level in differ­ent seg­ments of the Amer­i­can pop­u­la­tion. Tests of func­tional lit­er­acy essen­tially mimic indi­vid­u­al­ly-ad­min­is­tered intel­li­gence tests, except that all their items come from every­day life, such as cal­cu­lat­ing a tip (see extended dis­cus­sion in Got­tfred­son, 1997b). As on intel­li­gence tests, differ­ences in item diffi­culty rest on the items’ cog­ni­tive com­plex­ity (their abstract­ness, amount of dis­tract­ing irrel­e­vant infor­ma­tion, and degree of infer­ence required), not on their read­abil­ity per se or the level of edu­ca­tion test tak­ers have com­plet­ed. Lit­er­acy researchers have con­clud­ed, with some sur­prise, that func­tional lit­er­acy rep­re­sents a gen­eral capac­ity to learn, rea­son, and solve prob­lem­s-a ver­i­ta­ble descrip­tion of g.

The National Adult Lit­er­acy Sur­vey (NALS; Kirsch, Junge­blut, Jenk­ins, & Kol­stad, 1993) groups lit­er­acy scores into five lev­els. Indi­vid­u­als scor­ing in Level 1 have an 80% chance of suc­cess­fully per­form­ing tasks sim­i­lar in diffi­culty to locat­ing an expi­ra­tion date on a dri­ver’s license and total­ing a bank deposit slip. They are not rou­tinely able to per­form Level 2 tasks, such as deter­min­ing the price differ­ence between two show tick­ets or fill­ing in back­ground infor­ma­tion on an appli­ca­tion for a social secu­rity card. Level 3 diffi­culty includes writ­ing a brief let­ter explain­ing an error in a credit card bill and using a flight sched­ule to plan trav­el. Level 4 tasks include restat­ing an argu­ment made in a lengthy news arti­cle and cal­cu­lat­ing the money needed to raise a child based on infor­ma­tion in a news arti­cle. Only at Level 5 are indi­vid­u­als rou­tinely able to per­form men­tal tasks as com­plex as sum­ma­riz­ing two ways that lawyers chal­lenge prospec­tive jurors (based on a pas­sage dis­cussing such prac­tices) and, with a cal­cu­la­tor, deter­min­ing the total cost of car­pet to cover a room.

Although these tasks might seem to rep­re­sent only the incon­se­quen­tial minu­tiae of every­day life, they sam­ple the large uni­verse of mostly untu­tored tasks that mod­ern life demands of adults. Con­sis­tently fail­ing them is not just a daily incon­ve­nience, but a com­pound­ing prob­lem. Liken­ing func­tional lit­er­acy to mon­ey-it always helps to have more-, lit­er­acy researchers point out that rates of socioe­co­nomic dis­tress and pathol­ogy (unem­ploy­ment, adult pover­ty, etc.) rise steadily at suc­ces­sively lower lev­els of func­tional lit­er­acy (as is the pat­tern for IQ too; Got­tfred­son, 2002a)…­Such dis­ad­van­tage is com­mon, too, because 40% of the adult white pop­u­la­tion and 80% of the adult black pop­u­la­tion can­not rou­tinely per­form above Level 2. Fully 14% and 40%, respec­tive­ly, can­not rou­tinely per­form even above Level 1 (Kirsch et al., 1993, pp. 119121). To claim that low­er-a­bil­ity cit­i­zens will only be vic­tim­ized by the pub­lic know­ing that differ­ences in intel­li­gence are real, stub­born, and impor­tant is to ignore the prac­ti­cal hur­dles they face.

Health lit­er­a­cy, IQ, and health self­-care. The chal­lenges in self­-care for low­er-IQ indi­vid­u­als are espe­cially strik­ing in health mat­ters, where the con­se­quences of poor per­for­mance are tal­lied in excess mor­bid­ity and mor­tal­i­ty. Health psy­chol­o­gists have ignored the role of com­pe­tence in health behav­ior, focus­ing instead on voli­tion. Patient “non-com­pli­ance” is indeed a huge prob­lem in med­i­cine, but health lit­er­acy researchers, unlike health psy­chol­o­gists, have con­cluded that it is more a mat­ter of patients not under­stand­ing what is required of them than being unwill­ing to imple­ment it (re­views in Got­tfred­son, 2002a, in press).

…For exam­ple, 26% of out­pa­tients in sev­eral large urban hos­pi­tals could not deter­mine from an appoint­ment slip when the next visit was sched­uled and 42% could not under­stand instruc­tions for tak­ing med­i­cine on an empty stom­ach. Among those with “inad­e­quate” lit­er­a­cy, the fail­ure rates on these two tasks were 40% and 65%, respec­tive­ly. Sub­stan­tial per­cent­ages of this low-lit­er­acy group were unable to report, when given pre­scrip­tion labels con­tain­ing the nec­es­sary infor­ma­tion, how to take the med­ica­tion four times a day (24%), how many times the pre­scrip­tion could be refilled (42%), or how many pills of the pre­scrip­tion should be taken (70%). Tak­ing med­ica­tions improp­erly can be as harm­ful as not tak­ing them at all, and the phar­macy pro­fes­sion has esti­mated that about half of all pre­scrip­tions are taken incor­rect­ly. As in other per­for­mance domains, train­ing and moti­va­tion do not erase the dis­ad­van­tages of lower com­pre­hen­sion abil­i­ties. For instance, many patients who are under treat­ment for insulin-de­pen­dent dia­betes do not under­stand the most ele­men­tal facts for main­tain­ing daily con­trol of their dis­ease. In one study, about half of those with “inad­e­quate” lit­er­acy did not know the signs of very low or very high blood sug­ar, both of which require expe­di­tious cor­rec­tion, and 60% did not know the cor­rec­tive actions to take. Like hyper­ten­sion and many other chronic ill­ness­es, dia­betes requires con­tin­ual self­-mon­i­tor­ing and fre­quent judg­ments by patients to keep their phys­i­o­log­i­cal processes within safe lim­its dur­ing the day. Per­sis­tently high blood sugar lev­els can lead to blind­ness, heart dis­ease, limb ampu­ta­tion, and much more. For per­sons in gen­er­al, low func­tional lit­er­acy has been linked to num­ber and sever­ity of ill­ness­es, worse self­-rated health, far higher med­ical costs, and (prospec­tive­ly) more fre­quent hos­pi­tal­iza­tion. These rela­tions are not elim­i­nated by con­trol­ling for edu­ca­tion, socioe­co­nomic resources, access to health care, demo­graphic char­ac­ter­is­tics, and other such vari­ables.

Because health lit­er­acy is a rough sur­ro­gate for g, it pro­duces results con­sis­tent with research on IQ and health. To take sev­eral exam­ples, intel­li­gence at time of diag­no­sis cor­re­lates .36 with dia­betes knowl­edge mea­sured one year later (Tay­lor, Frier, Gold, & Deary, in press). IQ mea­sured at age 11 pre­dicts longevi­ty, inci­dence of can­cer, and func­tional inde­pen­dence in old age, and these rela­tions remain robust after con­trol­ling for deprived liv­ing con­di­tions (Deary, White­man, Starr, & Whal­ley, in press). Another prospec­tive epi­demi­o­log­i­cal study found that the motor vehi­cle death rate for men of IQ 80-85 was triple and for men of IQ 85-100 it was dou­ble the rate for men of IQ 100-115 (O’­Toole, 1990). Youth­ful IQ was the best pre­dic­tor of all-cause mor­tal­ity by age 40 in this large national sam­ple of Aus­tralian Army vet­er­ans, and IQ’s pre­dic­tive value remained sig­nifi­cant after con­trol­ling for all 56 demo­graph­ic, health, and other attrib­utes mea­sured (O’­Toole & Stankov, 1992). As in edu­ca­tion, equal resources do not pro­duce equal out­comes in health. Like edu­ca­tional inequal­i­ties, health inequal­i­ties increase when health resources become equally avail­able to all, such as hap­pened to the British gov­ern­men­t’s dis­may after it insti­tuted free national health care. Health improves over­all, but least for less edu­cated and lower income per­sons. They seek more but not nec­es­sar­ily appro­pri­ate care when cost is no bar­ri­er; adhere less often to treat­ment reg­i­mens; learn and under­stand less about how to pro­tect their health; seek less pre­ven­tive care, even when free; and less often prac­tice the healthy behav­iors so impor­tant for pre­vent­ing or slow­ing the pro­gres­sion of chronic dis­eases, the major killers and dis­ablers in devel­oped nations.

…In­fus­ing more knowl­edge into the pub­lic sphere about health risks (smok­ing) and new diag­nos­tic options (Pap smears) results in already-in­formed per­sons learn­ing the most and more often act­ing on the new infor­ma­tion. This may explain why an SES-mortality gra­di­ent favor­ing edu­cated women devel­oped for cer­vi­cal can­cer after Pap smears became avail­able.

…After it became clear that health inequal­i­ties could not be explained by inequal­i­ties in mate­r­ial resources and access to health care, it became fash­ion­able in health epi­demi­ol­ogy to blame class and race differ­ences in health on the psy­chic dam­age done by social inequal­i­ty. We are now to believe that social inequal­ity per se is lit­er­ally a killer (Wilkin­son, 1996). Physi­cians, like teach­ers, are increas­ingly being accused of racism and given sen­si­tiv­ity train­ing when they fail to pro­duce racial par­ity in out­comes (Satel, 2000). Mind­ful of ide­o­log­i­cally cor­rect thought, health lit­er­acy researchers who men­tion intel­li­gence do so only to reject out of hand the notion that lit­er­acy might reflect intel­li­gence, because any such notion would be racist and demean­ing.

In the mean­time, inad­e­quate learn­ing and rea­son­ing abil­i­ties put many peo­ple at risk of tak­ing med­ica­tions in health-dam­ag­ing ways, not grasp­ing the mer­its of pre­ven­tive pre­cau­tions against chronic dis­ease and acci­dents, and fail­ing to prop­erly imple­ment poten­tially more effec­tive but com­plex new treat­ment reg­i­mens for heart dis­ease, hyper­ten­sion, and other killers.

…To inten­tion­ally ignore differ­ences in men­tal com­pe­tence is uncon­scionable. It is social sci­ence mal­prac­tice against the very peo­ple whom the “untruth” is sup­pos­edly meant to pro­tect.

Got­tfred­son, Linda S. 2002. “Where and Why g Mat­ters: Not a Mys­tery.” Human Per­for­mance 15 (1/2): 25-46.

g is a highly gen­eral capa­bil­ity for pro­cess­ing com­plex infor­ma­tion of any type. This explains its great value in pre­dict­ing job per­for­mance. Com­plex­ity is the major dis­tinc­tion among jobs, which explains why g is more impor­tant fur­ther up the occu­pa­tional hier­ar­chy. The pre­dic­tive validi­ties of g are mod­er­ated by the cri­te­ria and other pre­dic­tors con­sid­ered in selec­tion research, but the result­ing gra­di­ents of g’s effects are sys­tem­at­ic. The pat­tern pro­vides per­son­nel psy­chol­o­gists a road map for how to design bet­ter selec­tion bat­ter­ies.

…One of the sim­plest facts about men­tal abil­i­ties pro­vides one of the most impor­tant clues to the nature of g. Peo­ple who do well on one kind of men­tal test tend to do well on all oth­ers. When the scores on a large, diverse bat­tery of men­tal abil­ity tests are fac­tor ana­lyzed, they yield a large com­mon fac­tor, labeled g. Pick any test of men­tal apti­tude or achieve­men­t-say, ver­bal apti­tude, spa­tial visu­al­iza­tion, the SAT, a stan­dard­ized test of aca­d­e­mic achieve­ment in 8th grade, or the Block De- sign or Mem­ory for Sen­tences sub­tests of the Stan­ford-Bi­net intel­li­gence test- and you will find that it mea­sures mostly g. All efforts to build mean­ing­ful men­tal tests that do not mea­sure g have failed…In con­trast, no gen­eral fac­tor emerges from per­son­al­ity inven­to­ries, which shows that gen­eral fac­tors are not a nec­es­sary out­come of fac­tor analy­sis. (See Jensen, 1998, and Got­tfred­son, 1997, 2000a, 2002, for fuller dis­cus­sion and doc­u­men­ta­tion of these and fol­low­ing points on g.)

The impor­tant point is that the pre­dic­tive validi­ties of g behave law­ful­ly. They vary, but they vary sys­tem­at­i­cally and for rea­sons that are begin­ning to be well under­stood. Over 2 decades of meta-analy­ses have shown that they are not sen­si­tive to small vari­a­tions in job duties and cir­cum­stance, after con­trol­ling for sam­pling error and other sta­tis­ti­cal arti­facts. Com­plex jobs will always put a pre­mium on higher g. Their per­for­mance will always be notably enhanced by higher g, all else equal. Higher g will also enhance per­for­mance in sim­ple jobs, but to a much smaller degree.


Gen­eral intel­li­gence is an impor­tant human quan­ti­ta­tive trait that accounts for much of the vari­a­tion in diverse cog­ni­tive abil­i­ties. Indi­vid­ual differ­ences in intel­li­gence are strongly asso­ci­ated with many impor­tant life out­comes, includ­ing edu­ca­tional and occu­pa­tional attain­ments, income, health and lifes­pan1,2. Data from twin and fam­ily stud­ies are con­sis­tent with a high her­i­tabil­ity of intel­li­gence3, but this infer­ence has been con­tro­ver­sial. We con­ducted a genome-wide analy­sis of 3511 unre­lated adults with data on 549 692 SNPs and detailed phe­no­types on cog­ni­tive traits. We esti­mate that 40% of the vari­a­tion in crys­tal­lized-type intel­li­gence and 51% of the vari­a­tion in flu­id-type intel­li­gence between indi­vid­u­als is accounted for by link­age dis­e­qui­lib­rium between geno­typed com­mon SNP mark­ers and unknown causal vari­ants. These esti­mates pro­vide lower bounds for the nar­row-sense her­i­tabil­ity of the traits. We par­ti­tioned genetic vari­a­tion on indi­vid­ual chro­mo­somes and found that, on aver­age, longer chro­mo­somes explain more vari­a­tion. Final­ly, using just SNP data we pre­dicted approx­i­mately 1% of the vari­ance of crys­tal­lized and fluid cog­ni­tive phe­no­types in an inde­pen­dent sam­ple (P = 0.009 and 0.028, respec­tive­ly). Our results unequiv­o­cally con­firm that a sub­stan­tial pro­por­tion of indi­vid­ual differ­ences in human intel­li­gence is due to genetic vari­a­tion, and are con­sis­tent with many genes of small effects under­ly­ing the addi­tive genetic influ­ences on intel­li­gence.

“Com­mon DNA Mark­ers Can Account for More Than Half of the Genetic Influ­ence on Cog­ni­tive Abil­i­ties”:

For nearly a cen­tu­ry, twin and adop­tion stud­ies have yielded sub­stan­tial esti­mates of her­i­tabil­ity for cog­ni­tive abil­i­ties, although it has proved diffi­cult for genome-wide-as­so­ci­a­tion stud­ies to iden­tify the genetic vari­ants that account for this her­i­tabil­ity (i.e., the miss­ing-her­i­tabil­ity prob­lem). How­ev­er, a new approach, genome-wide com­plex-trait analy­sis (GCTA), for­goes the iden­ti­fi­ca­tion of indi­vid­ual vari­ants to esti­mate the total her­i­tabil­ity cap­tured by com­mon DNA mark­ers on geno­typ­ing arrays. In the same sam­ple of 3,154 pairs of 12-year-old twins, we directly com­pared twin-s­tudy her­i­tabil­ity esti­mates for cog­ni­tive abil­i­ties (lan­guage, ver­bal, non­ver­bal, and gen­er­al) with GCTA esti­mates cap­tured by 1.7 mil­lion DNA mark­ers. We found that DNA mark­ers tagged by the array accounted for .66 of the esti­mated her­i­tabil­i­ty, reaffirm­ing that cog­ni­tive abil­i­ties are her­i­ta­ble. Larger sam­ple sizes alone will be suffi­cient to iden­tify many of the genetic vari­ants that influ­ence cog­ni­tive abil­i­ties.

…Cog­ni­tive abil­i­ties pre­dict edu­ca­tional attain­ment, income, health, and longevi­ty, and thus con­tribute impor­tantly to the intel­lec­tual cap­i­tal of knowl­edge-based soci­eties (Deary, 2012). Since the 1920s, twin and adop­tion stud­ies have inves­ti­gated the genetic and envi­ron­men­tal ori­gins of indi­vid­ual differ­ences in cog­ni­tive abil­i­ties; scores of such stud­ies have con­sis­tently yielded esti­mates of sub­stan­tial her­i­tabil­ity (i.e., the extent to which genetic vari­ance can account for observed, or phe­no­typ­ic, vari­ance; Deary, John­son, & Houli­han, 2009). Meta-analy­ses of these stud­ies have yielded her­i­tabil­ity esti­mates of about .50 for gen­eral cog­ni­tive abil­i­ty, the most well-s­tud­ied cog­ni­tive trait (Plom­in, DeFries, Knopik, & Neu­houser, 2013).

…One of the most far-reach­ing results of GWA stud­ies is to show that there are no genes of large effect size in the pop­u­la­tion, which means that the her­i­tabil­ity of com­plex traits is prob­a­bly due to many genes of small effect size, and this means that asso­ci­a­tions will be diffi­cult to detect and repli­cate (Plom­in, 2012). For exam­ple, the first GWA stud­ies of gen­eral cog­ni­tive abil­ity (Davies et al., 2011; Davis et al., 2010) were pow­ered to detect asso­ci­a­tions that account for as lit­tle as .01 of the vari­ance, but they came up emp­ty-handed because the asso­ci­a­tions with the largest effect accounted for less than .005 of the vari­ance. One of many pos­si­ble rea­sons for the miss­ing-her­i­tabil­ity prob­lem is that the com­mon SNPs (i.e., SNPs for which the fre­quency of the less fre­quent allele is greater than .01) incor­po­rated in com­mer­cially avail­able DNA arrays miss the con­tri­bu­tion of rare DNA vari­ants (Cir­ulli & Gold­stein, 2010). Another pos­si­bil­ity is that her­i­tabil­ity has been over­es­ti­mated by twin and adop­tion stud­ies.

…The study reported here addressed both of these pos­si­bil­i­ties by com­par­ing twin-based esti­mates of her­i­tabil­ity for cog­ni­tive abil­i­ties with esti­mates from a new method that is pop­u­la­tion based rather than fam­ily based. The method, called genome-wide com­plex-trait analy­sis (GCTA), can be used to esti­mate genetic vari­ance accounted for by all the SNPs that have been geno­typed in any sam­ple, not just sam­ples con­sist­ing of spe­cial fam­ily mem­bers such as twins or adoptees (Lee, Wray, God­dard, & Viss­cher, 2011; Yang, Lee, God­dard, & Viss­cher, 2011; Yang, Mano­lio, et al., 2011)…GCTA does not iden­tify spe­cific genes asso­ci­ated with traits. Instead, it uses chance sim­i­lar­ity across hun­dreds of thou­sands of SNPs to pre­dict phe­no­typic sim­i­lar­ity pair by pair in a large sam­ple of unre­lated indi­vid­u­als. The essence of GCTA is to esti­mate genetic influ­ence on a trait by pre­dict­ing phe­no­typic sim­i­lar­ity for each pair of indi­vid­u­als in the sam­ple from their total SNP sim­i­lar­i­ty. In con­trast to the twin method, which esti­mates her­i­tabil­ity by com­par­ing phe­no­typic sim­i­lar­ity of iden­ti­cal and fra­ter­nal twin pairs, whose genetic sim­i­lar­ity is roughly 1.00 and .50, respec­tive­ly, GCTA relies on com­par­isons of pairs of indi­vid­u­als whose genetic sim­i­lar­ity varies from .00 to .02. GCTA extracts this tiny genetic sig­nal from the noise of hun­dreds of thou­sands of SNPs using the mas­sive infor­ma­tion avail­able from a matrix of thou­sands of indi­vid­u­als, each com­pared pair by pair with every other indi­vid­ual in the sam­ple; for exam­ple, the 3,000-plus indi­vid­u­als in the present sam­ple pro­vided nearly 5 mil­lion pair­wise com­par­isons

GCTA has been used to esti­mate her­i­tabil­ity as cap­tured by geno­typ­ing arrays for height (Yang et al., 2010), weight (Yang, Mano­lio, et al., 2011), psy­chi­atric and other med­ical dis­or­ders (Lee et al., 2012; Lee et al., 2011; Lubke et al., 2012), and per­son­al­ity (Vinkhuyzen, Ped­er­sen, et al., 2012). GCTA was first applied to cog­ni­tive abil­ity in a study of 3,500 unre­lated adults, which yielded her­i­tabil­ity esti­mates of .40 and .51 for crys­tal­lized and fluid intel­li­gence, respec­tively (Davies et al., 2011). The GCTA esti­mate for gen­eral cog­ni­tive abil­ity was .47 in a meta-analy­sis across three stud­ies involv­ing nearly 10,000 adults (Chabris et al., 2012) and .48 in a study of nearly 2 thou­sand 11-year-old chil­dren (Deary et al., 2012)…GCTA has been used to esti­mate her­i­tabil­ity as cap­tured by geno­typ­ing arrays for height (Yang et al., 2010), weight (Yang, Mano­lio, et al., 2011), psy­chi­atric and other med­ical dis­or­ders (Lee et al., 2012; Lee et al., 2011; Lubke et al., 2012), and per­son­al­ity (Vinkhuyzen, Ped­er­sen, et al., 2012). GCTA was first applied to cog­ni­tive abil­ity in a study of 3,500 unre­lated adults, which yielded her­i­tabil­ity esti­mates of .40 and .51 for crys­tal­lized and fluid intel­li­gence, respec­tively (Davies et al., 2011). The GCTA esti­mate for gen­eral cog­ni­tive abil­ity was .47 in a meta-analy­sis across three stud­ies involv­ing nearly 10,000 adults (Chabris et al., 2012) and .48 in a study of nearly 2 thou­sand 11-year-old chil­dren (Deary et al., 2012).

GCTA has been used to esti­mate her­i­tabil­ity as cap­tured by geno­typ­ing arrays for height (Yang et al., 2010), weight (Yang, Mano­lio, et al., 2011), psy­chi­atric and other med­ical dis­or­ders (Lee et al., 2012; Lee et al., 2011; Lubke et al., 2012), and per­son­al­ity (Vinkhuyzen, Ped­er­sen, et al., 2012). GCTA was first applied to cog­ni­tive abil­ity in a study of 3,500 unre­lated adults, which yielded her­i­tabil­ity esti­mates of .40 and .51 for crys­tal­lized and fluid intel­li­gence, respec­tively (Davies et al., 2011). The GCTA esti­mate for gen­eral cog­ni­tive abil­ity was .47 in a meta-analy­sis across three stud­ies involv­ing nearly 10,000 adults (Chabris et al., 2012) and .48 in a study of nearly 2 thou­sand 11-year-old chil­dren (Deary et al., 2012)…This is the first study in which GCTA esti­mates of her­i­tabil­ity for diverse cog­ni­tive abil­i­ties were com­pared directly with twin-based esti­mates using the same mea­sures at the same age in the same sam­ple. The Affymetrix 6.0 DNA array yielded GCTA esti­mates that accounted on aver­age for .66 of the twin her­i­tabil­ity esti­mates for lan­guage, ver­bal, non­ver­bal, and gen­eral cog­ni­tive abil­i­ties. Note that the GCTA esti­mates accounted for a greater pro­por­tion of the twin her­i­tabil­ity esti­mates in the case of cog­ni­tive abil­i­ties than in the case of height (.44) and weight (.50).

…Why might these com­mon SNPs tag gen­eral cog­ni­tive abil­ity more than height and weight? Com­mon SNPs are likely to be com­mon because they are old, hav­ing spread through the pop­u­la­tion over many gen­er­a­tions, but there seems no obvi­ous rea­son why the evo­lu­tion­ary archi­tec­ture for gen­eral cog­ni­tive abil­ity should differ from height in this way. How­ev­er, there is one major genetic differ­ence between cog­ni­tive and phys­i­cal traits: Assor­ta­tive mat­ing (non­ran­dom mat­ing) is at least twice as great for gen­eral cog­ni­tive abil­ity (cor­re­la­tion between spous­es: ~.45) as for height and weight (~.20; Plomin et al., 2013). The effect of assor­ta­tive mat­ing is to increase addi­tive genetic vari­ance because chil­dren receive cor­re­lated genetic influ­ences from their par­ents, which spreads out the dis­tri­b­u­tion; more­over, the effects of assor­ta­tive mat­ing accu­mu­late gen­er­a­tion after gen­er­a­tion. If assor­ta­tive mat­ing is respon­si­ble for the fact that com­mon SNPs tag gen­eral cog­ni­tive abil­ity more than height and weight, then ver­bal abil­i­ties should show greater GCTA/twin her­i­tabil­ity ratios than non­ver­bal abil­i­ties do because ver­bal abil­i­ties show more assor­ta­tive mat­ing than non­ver­bal abil­i­ties (cor­re­la­tion between spous­es: ~.50 vs. .30). The results in Table 1 are con­sis­tent with this hypoth­e­sis: The GCTA/twin her­i­tabil­ity ratio is .65 for ver­bal abil­ity and .48 for non­ver­bal abil­i­ty.

Above a cer­tain lev­el, intel­li­gence does­n’t mat­ter. There was no sig­nifi­cant differ­ence in max­i­mum income earned by men with IQs in the 110-115 range and men with IQs higher than 150.

TODO: what’s going on there? weasel word­ing on ‘max­i­mum’? not a big enough sam­ple size to reach sta­tis­ti­cal-sig­nifi­cance

, Ham­brick et al 2014:

…Global mea­sures of intel­li­gence (IQ) have also been found to cor­re­late with per­for­mance in chess and music, con­sis­tent with the pos­si­bil­ity that a rel­a­tively high level of intel­li­gence is nec­es­sary for suc­cess in these domains. Fry­d­man and Lynn (1992) found that young chess play­ers had an aver­age per­for­mance IQ of 129, com­pared to a pop­u­la­tion aver­age of 100, and that the aver­age was higher for the best play­ers (top-third avg. = 131) in the sam­ple than the weak­est play­ers (bot­tom-third avg. = 124). Fur­ther­more, Grab­n­er, Neubauer, and Stern (2006) found that, even in highly rated play­ers, IQ pos­i­tively pre­dicted per­for­mance on rep­re­sen­ta­tive chess tasks (e.g., next best move). Bilalić et al. (2007) found that IQ was not a sig­nifi­cant pre­dic­tor of chess rat­ing in the sam­ple of elite young chess play­ers listed in Table 1 after sta­tis­ti­cally con­trol­ling for prac­tice. How­ev­er, the sam­ple size for the elite group was only 23, and mean IQ was sig­nifi­cantly higher for the elite group (M = 133) than for the rest of the sam­ple (M = 114).

  • Fry­d­man, M., & Lynn, R. (1992). The gen­eral intel­li­gence and spa­tial abil­i­ties of gifted young Bel­gian chess play­ers. British Jour­nal of Psy­chol­o­gy, 83, 233-235.
  • Grab­n­er, R. H., Neubauer, A. C., & Stern, E. (2006). Supe­rior per­for­mance and neural effi­cien­cy: The impact of intel­li­gence and exper­tise. Brain Research Bul­let­in, 69, 422-439. j.brain­res­bul­l.2006.02.009.
  • Bilal­ić, M., McLeod, P., & Gob­et, F. (2007). Does chess need intel­li­gence? A study with young chess play­ers. Intel­li­gence, 35, 457-470. 10.1016/j.intell.2006.09.005.

IQ cor­re­lates pos­i­tively with music per­for­mance, as well. Luce (1965) found a cor­re­la­tion of .53 (p b .01) between IQ and sight-read­ing per­for­mance in high school band mem­bers, and Salis (1977) reported a cor­re­la­tion of .58 between these vari­ables in a uni­ver­sity sam­ple. Gromko (2004) found pos­i­tive cor­re­la­tions between both ver­bal abil­ity and spa­tial abil­ity (rs = .35-.49) and sight-read­ing per­for­mance in high school wind play­ers, and Hay­ward and Gromko (2009) found a sig­nifi­cant pos­i­tive cor­re­la­tion (r = .24) between a mea­sure of spa­tial abil­ity based on three ETS tests and sight-read­ing per­for­mance in uni­ver­sity wind play­ers. Ruth­satz et al. (2008) found that Raven’s scores cor­re­lated pos­i­tively and sig­nifi­cantly with musi­cal achieve­ment in high school band mem­bers (r = .25). This cor­re­la­tion was not sta­tis­ti­cally sig­nifi­cant in a sam­ple of more highly accom­plished con­ser­va­tory stu­dents and music majors, but this could have been due to a ceil­ing effect on Raven’s, as these par­tic­i­pants had been heav­ily selected for cog­ni­tive abil­i­ty.

  • Luce, J. R. (1965). Sight-read­ing and ear-play­ing abil­i­ties as related to instru­men­tal music stu­dents. Jour­nal of Research in Music Edu­ca­tion, 13, 101-109.
  • Sal­is, D. L. (1977). The iden­ti­fi­ca­tion and assess­ment of cog­ni­tive vari­ables asso­ci­ated with read­ing of advanced music at the piano (un­pub­lished doc­toral dis­ser­ta­tion). Pitts­burgh, PA: Uni­ver­sity of Pitts­burgh.
  • Gromko, J. E. (2004). Pre­dic­tors of music sight-read­ing abil­ity in high school wind play­ers. Jour­nal of Research in Music Edu­ca­tion, 52, 6-15.
  • Hay­ward, C. M., & Gromko, J. E. (2009). Rela­tion­ships among music sight-read­ing and tech­ni­cal pro­fi­cien­cy, spa­tial visu­al­iza­tion, and aural dis­crim­i­na­tion. Jour­nal of Research in Music Edu­ca­tion, 57, 29-36.
  • Ruth­satz, J., Det­ter­man, D. K., Griscom, W. S., & Cir­ul­lo, B. A. (2008). Becom­ing an expert in the musi­cal domain: It takes more than just prac­tice. Intel­li­gence, 36, 330-338.

…Ruth­satz and Urbach (2012) admin­is­tered a stan­dard­ized IQ test (the Stan­ford-Bi­net) to eight child prodigies, six of whom were musi­cal prodi­gies. Despite ful­l-s­cale IQs that ranged from 108 to 147-just above aver­age to above the con­ven­tional cut­off for “genius”-all of the prodi­gies were at or above the 99th per­centile for work­ing mem­ory (in­deed, six scored at the 99.9th per­centile).

  • Ruth­satz, J., & Urbach, J. B. (2012). Child prodi­gy: A novel cog­ni­tive pro­file places ele­vated gen­eral intel­li­gence, excep­tional work­ing mem­ory and atten­tion to detail at the root of prodi­gious­ness. Intel­li­gence, 40, 419-426.

…Gen­eral intel­li­gence does not always pre­dict per­for­mance. In a study of foot­ball play­ers, Lyons, Hoff­man, and Michel (2009) found that scores on the Won­der­lic Per­son­nel Test, a widely admin­is­tered group intel­li­gence test, cor­re­lated essen­tially zero with suc­cess in the National Foot­ball League, even in the quar­ter­back posi­tion, which is believed to place the high­est demand on infor­ma­tion pro­cess­ing. Fur­ther­more, Ham­brick et al. (2012) found that spa­tial abil­ity pos­i­tively pre­dicted suc­cess in a com­plex geo­log­i­cal prob­lem solv­ing task in novice geol­o­gists, but not in experts.

  • Lyons, B., Hoff­man, B., & Michel, J. (2009). Not much more than g? An exam­i­na­tion of the impact of intel­li­gence on NFL per­for­mance. Human Per­for­mance, 22, 225-245.
  • Ham­brick, D. Z., & Meinz, E. J. (2012). Work­ing mem­ory capac­ity and musi­cal skill. In T. P. Alloway, & R. G. Alloway (Ed­s.), Work­ing mem­o­ry: The con­nected intel­li­gence (pp. 137-155). New York: Psy­chol­ogy Press.

“Intel­li­gence and semen qual­ity are pos­i­tively cor­re­lated”, Arden et al 2009

If the cor­re­la­tions among cog­ni­tive abil­i­ties are part of a larger matrix of pos­i­tive asso­ci­a­tions among fit­ness-re­lated traits, then intel­li­gence ought to cor­re­late with seem­ingly unre­lated traits that affect fit­ness-such as semen qual­i­ty. We found sig­nifi­cant pos­i­tive cor­re­la­tions between intel­li­gence and 3 key indices of semen qual­i­ty: log sperm con­cen­tra­tion (r = .15, p = .002), log sperm count (r = .19, p b .001), and sperm motil­ity (r = .14, p = .002) in a large sam­ple of US Army Vet­er­ans. None was medi­ated by age, body mass index, days of sex­ual absti­nence, ser­vice in Viet­nam, or use of alco­hol, tobac­co, mar­i­jua­na, or hard drugs.

…in­tel­li­gence cor­re­lates with many impor­tant health out­comes, even longevity (Bat­ty, Deary, & Got­tfred­son, 2007).

  • Bat­ty, G. D., Deary, I. J., & Got­tfred­son, L. S. (2007). Pre­mor­bid (early life) IQ and later mor­tal­ity risk: Sys­tem­atic review. Annals of Epi­demi­ol­o­gy, 17 (4), 278−288.

…the effect size is con­gru­ent with phe­no­typic cor­re­la­tions observed for other bod­ily cor­re­lates of intel­li­gence such as height (r = .14, r = .15) (Sil­ven­toinen, Posthu­ma, van Bei­jster­veldt, Bar­tels, & Booms­ma, 2006; Sun­det, Tambs, Har­ris, Mag­nus, & Tor­jussen, 2005).

  • Sil­ven­toinen, K., Posthu­ma, D., van Bei­jster­veldt, T., Bar­tels, M., & Booms­ma, D. I. (2006). Genetic con­tri­bu­tions to the asso­ci­a­tion between height and intel­li­gence: Evi­dence from Dutch twin data from child­hood to mid­dle age. Genes, Brain and Behav­ior, 5(8), 585−595.
  • Sun­det, J. M., Tambs, K., Har­ris, J. R., Mag­nus, P., & Tor­jussen, T. M. (2005). Resolv­ing the genetic and envi­ron­men­tal sources of the cor­re­la­tion between height and intel­li­gence: A study of nearly 2600 Nor­we­gian male twin pairs. Twin Research and Human Genet­ics, 8(4), 307−311.


  1. Deary IJ, Strand S, Smith P, Fer­nan­des C (2007) Intel­li­gence and edu­ca­tional achieve­ment. Intel­li­gence 35: 13-21 doi:10.1016/j.intell.2006.02.001. . doi: 10.1016/j.intell.2006.02.001. IQ scores are often used as an index of gen­eral cog­ni­tive abil­i­ties. Such IQ mea­sures exhibit sub­stan­tial cor­re­la­tions from late child­hood through adult­hood (e.g., IQ scores were esti­mated to cor­re­late 0.73 from ages 11 through 77 in a lon­gi­tu­di­nal study [2]).
  2. Deary IJ, Whal­ley LJ, Lem­mon H, Craw­ford J, Starr JM (2000) The Sta­bil­ity of Indi­vid­ual Differ­ences in Men­tal Abil­ity from Child­hood to Old Age: Fol­low-up of the 1932 Scot­tish Men­tal Sur­vey. Intel­li­gence 28: 49-55 doi:10.1016/S0160-2896(99)00031-8. . doi: 10.1016/S0160-2896(99)00031-8.

“Child­hood cog­ni­tive abil­ity accounts for asso­ci­a­tions between cog­ni­tive abil­ity and brain cor­ti­cal thick­ness in old age”, Karama et al 2013

We ana­lyzed data on 588 sub­jects from the Loth­ian Birth Cohort 1936 who had intel­li­gence quo­tient (IQ) scores from the same cog­ni­tive test avail­able at both 11 and 70 years of age as well as high­-res­o­lu­tion brain mag­netic res­o­nance imag­ing data obtained at approx­i­mately 73 years of age. Cor­ti­cal thick­ness was esti­mated at 81,924 sam­pling points across the cor­tex for each sub­ject using an auto­mated pipeline. Mul­ti­ple regres­sion was used to assess asso­ci­a­tions between cor­ti­cal thick­ness and the IQ mea­sures at 11 and 70 years. Child­hood IQ accounted for more than two-third of the asso­ci­a­tion between IQ at 70 years and cor­ti­cal thick­ness mea­sured at age 73 years. This warns against ascrib­ing a causal inter­pre­ta­tion to the asso­ci­a­tion between cog­ni­tive abil­ity and cor­ti­cal tis­sue in old age based on assump­tions about, and exclu­sive ref­er­ence to, the aging process and any asso­ci­ated dis­ease.

, Chi­ang et al 2011

We devel­oped an analy­sis pipeline enabling pop­u­la­tion stud­ies of HARDI data, and applied it to map genetic influ­ences on fiber archi­tec­ture in 90 twin sub­jects. We applied ten­sor-driven 3D fluid reg­is­tra­tion to HARDI, re-sam­pling the spher­i­cal fiber ori­en­ta­tion dis­tri­b­u­tion func­tions (ODFs) in appro­pri­ate Rie­mann­ian man­i­folds, after ODF reg­u­lar­iza­tion and sharp­en­ing. Fit­ting struc­tural equa­tion mod­els (SEM) from quan­ti­ta­tive genet­ics, we eval­u­ated genetic influ­ences on the Jensen-Shan­non diver­gence (JSD), a novel mea­sure of fiber spa­tial coher­ence, and on the gen­er­al­ized fiber anisotropy (GFA) a mea­sure of fiber integri­ty. With ran­dom-effects regres­sion, we mapped regions where diffu­sion pro­files were highly cor­re­lated with sub­jects’ intel­li­gence quo­tient (IQ). Fiber com­plex­ity was pre­dom­i­nantly under genetic con­trol, and higher in more highly anisotropic regions; the pro­por­tion of genetic ver­sus envi­ron­men­tal con­trol var­ied spa­tial­ly. Our meth­ods show promise for dis­cov­er­ing genes affect­ing fiber con­nec­tiv­ity in the brain.

“Inves­ti­gat­ing Amer­i­ca’s elite: Cog­ni­tive abil­i­ty, edu­ca­tion, and sex differ­ences”, Wai 2013

Are the Amer­i­can elite drawn from the cog­ni­tive elite?…How­ev­er, whether the elite are pri­mar­ily com­posed of indi­vid­u­als in the top per­centiles of the abil­ity dis­tri­b­u­tion who have attended the most pres­ti­gious col­leges and uni­ver­si­ties has not yet been empir­i­cally exam­ined…To address this, five groups of Amer­i­ca’s elite (to­tal N = 2254) were exam­ined: For­tune 500 CEOs, fed­eral judges, bil­lion­aires, Sen­a­tors, and mem­bers of the House of Rep­re­sen­ta­tives. Within each of these groups, nearly all had attended col­lege with the major­ity hav­ing attended either a highly selec­tive under­grad­u­ate insti­tu­tion or grad­u­ate school of some kind. High aver­age test scores required for admis­sion to these insti­tu­tions indi­cated those who rise to or are selected for these posi­tions are highly fil­tered for abil­i­ty…Fe­males were under­rep­re­sented among all groups, but to a lesser degree among fed­eral judges and Democ­rats and to a larger degree among Repub­li­cans and CEOs. Amer­i­ca’s elite are largely drawn from the intel­lec­tu­ally gift­ed, with many in the top 1% of abil­i­ty…­Mur­ray (2008) was cor­rect that a large por­tion of Amer­i­ca’s elite are drawn from the intel­lec­tu­ally gift­ed. This held for every group except the House of Rep­re­sen­ta­tives, which had a lower per­cent­age hav­ing attended an Elite School. If the defi­n­i­tion of elite is broad­ened to include either atten­dance at an Elite School or Grad­u­ate School then the major­ity met these cri­te­ria and are likely in the top per­centiles of abil­i­ty. This would include 56.6% of the bil­lion­aires, 67.0% of the CEOs, 68.1% of the House, 83.0% of the Sen­ate, and all of the judges. All the fed­eral judges and Sen­a­tors and nearly all the other groups attended col­lege…This study used aver­age SAT or ACT scores of a col­lege or uni­ver­sity (Amer­i­ca’s Best Col­leges, 2013) as an approx­i­ma­tion for abil­ity level (Frey & Det­ter­man, 2004; Koenig et al., 2008), which may not hold for each indi­vid­ual case. It would have been opti­mal to have access to indi­vid­ual test scores, but unfor­tu­nately this data was not pub­licly avail­able. How­ev­er, using aver­age SAT and ACT scores as an approx­i­ma­tion for abil­ity level may give an under­es­ti­mate because extremely smart peo­ple may not have cho­sen to attend a top school for mul­ti­ple rea­sons (e.g., finan­cial, schol­ar­ship, stay­ing close to home). Alter­na­tive­ly, using this method may also give an over­es­ti­mate because there are many lega­cies and ath­letic admits to elite insti­tu­tions who do not usu­ally meet the typ­i­cal test score cri­te­ria (Espen­shade & Rad­ford, 2009). …This study demon­strates that in Amer­i­ca, Democ­rats were more likely than Repub­li­cans to have a higher per­cent­age of Sen­ate and House mem­bers who attended an Elite School which places these indi­vid­u­als in the top 1% in abil­ity (see Fig. 1 panel B and Appen­dix A). There­fore, among the elected elite, Democ­rats had a higher abil­ity and edu­ca­tion lev­el, on aver­age, than Repub­li­can­s…This also shows that Bill Gates and Mark Zucker­berg (in­cluded in the Tech­nol­ogy sec­tor), who are often used as promi­nent exam­ples in the media as to why going to col­lege is not nec­es­sary for suc­cess (e.g., Lin, 2010: “Top 10 col­lege dropouts”; Williams, 2012: “Say­ing no to col­lege”), are actu­ally excep­tions to the rule. Within the bil­lion­aire sam­ple, 37 (8.7%) were clearly marked as a col­lege drop out by the Forbes staff who com­piled the data. The major­ity of the bil­lion­aires (88%) went to col­lege and grad­u­at­ed. …Even within a group in the top 0.0000001% of wealth and a group of CEOs who were com­pen­sated quite highly (well within the top 1% of wealth), there were differ­ences in the edu­ca­tion and abil­ity level between those who earned more money com­pared to those who earned less. The analy­ses in Table 3a and b demon­strate that even within bil­lion­aires and CEOs, higher edu­ca­tion and abil­ity level is related to higher net worth and com­pen­sa­tion. Prior research demon­strated that even within a group in the top 1% in abil­i­ty, higher abil­ity is asso­ci­ated with higher income (Wai et al., 2005). The analy­ses in Table 3c demon­strated that even within the top 1% of abil­i­ty, higher abil­ity is asso­ci­ated with higher net worth and com­pen­sa­tion. There­fore, this study adds to, expands, and strength­ens the lit­er­a­ture link­ing edu­ca­tion, abil­i­ty, and wealth (Mur­ray, 1998; Nyborg & Jensen, 2001; Zax & Rees, 2002), and pro­vides fur­ther evi­dence that does not sup­port an abil­ity thresh­old hypoth­e­sis (Kun­cel & Hezlett, 2010; Park et al., 2007; Wai et al., 2005) - or the idea that more abil­ity does not mat­ter beyond a cer­tain point in pre­dict­ing real world out­comes.

  • Kuncel, N. R., & Hezlett, S. A. (2010). “Fact and fic­tion in cog­ni­tive abil­ity test­ing for admis­sions and hir­ing deci­sions”. Cur­rent Direc­tions in Psy­cho­log­i­cal Sci­ence, 19, 339-345
  • Mur­ray, C. (1998). Income inequal­ity and IQ. Wash­ing­ton, D.C.: AEI Press.
  • Nyborg, H., & Jensen, A. R. (2001). “Occu­pa­tion and income related to psy­cho­me­t­ric g”. Intel­li­gence, 29, 45-55
  • Park, G., Lubin­ski, D., & Ben­bow, C. P. (2007). “Con­trast­ing intel­lec­tual pat­terns pre­dict cre­ativ­ity in the arts and sci­ences”. Psy­cho­log­i­cal Sci­ence, 18, 948-952
  • Wai, J., Lubin­ski, D., & Ben­bow, C. P. (2005). “Cre­ativ­ity and occu­pa­tional accom­plish­ments among intel­lec­tu­ally pre­co­cious youths: An age 13 to age 33 lon­gi­tu­di­nal study”. Jour­nal of Edu­ca­tional Psy­chol­o­gy, 97, 484-492
  • Zax, J. S., & Rees, D. L. (2002). “IQ, aca­d­e­mic per­for­mance, envi­ron­ment, and earn­ings”. The Review of Eco­nom­ics and Sta­tis­tics, 84, 600-614.

This might seem obvi­ous (“elite schools pro­duce the elites”) but is worth ver­i­fy­ing. It’s also inter­est­ing that being elected does­n’t sub­stan­tially affect the edu­ca­tional cre­den­tials & inferred IQ (even Rep­re­sen­ta­tives are still 20x more likely to be from an elite school), since given their behavior/policies/statements one might assume elected politi­cians are medi­oc­ri­ties or oth­er­wise not espe­cially intel­li­gent. I can give a per­sonal exam­ple: I grew up in New York, where one of the Sen­a­tors has been for as long as I can remem­ber, , who never struck me as very sub­tle or intel­li­gent, an impres­sion fur­thered when he made his ill-fated pub­lic call years ago for the - still oper­at­ing - Silk Road to be shut down; so I was some­what shocked to learn via Steve Sailer that he claimed a per­fect 1600 on the pre-cen­ter­ing SAT (and then went to Har­vard where he was Phi Beta Kap­pa) which implies that he was more intel­li­gent than myself or most of Less­Wrong, and com­bined with his flaw­less elec­tion record, fur­ther sug­gests that I have badly mis­un­der­stood him and he is actu­ally a bril­liant polit­i­cal mas­ter­mind.

, McIn­tosh et al 2013:

Back­ground: Genome-wide asso­ci­a­tion stud­ies (GWAS) have shown a poly­genic com­po­nent to the risk of schiz­o­phre­nia. The dis­or­der is asso­ci­ated with impair­ments in gen­eral cog­ni­tive abil­ity that also have a sub­stan­tial genetic con­tri­bu­tion. No study has deter­mined whether cog­ni­tive impair­ments can be attrib­uted to schiz­o­phre­ni­a’s poly­genic archi­tec­ture using data from GWAS.

Meth­ods: Mem­bers of the Loth­ian Birth Cohort 1936 (LBC1936, n = 14,937) were assessed using the Moray House Test at age 11 and with the Moray House Test and a fur­ther cog­ni­tive bat­tery at age 70. To cre­ate poly­genic risk scores for schiz­o­phre­nia, we obtained data from the lat­est GWAS of the Psy­chi­atric GWAS Con­sor­tium on Schiz­o­phre­nia. Schiz­o­phre­nia poly­genic risk pro­file scores were cal­cu­lated using infor­ma­tion from the Psy­chi­atric GWAS Con­sor­tium on Schiz­o­phre­nia GWAS.

Results: In LBC1936, poly­genic risk for schiz­o­phre­nia was neg­a­tively asso­ci­ated with IQ at age 70 but not at age 11. Greater poly­genic risk for schiz­o­phre­nia was asso­ci­ated with more rel­a­tive decline in IQ between these ages. These find­ings were main­tained when the results of LBC1936 were com­bined with that of the inde­pen­dent Loth­ian Birth Cohort 1921 (n 14,517) in a meta-analy­sis.

Con­clu­sions: Increased poly­genic risk of schiz­o­phre­nia is asso­ci­ated with lower cog­ni­tive abil­ity at age 70 and greater rel­a­tive decline in gen­eral cog­ni­tive abil­ity between the ages of 11 and 70. Com­mon genetic vari­ants may under­lie both cog­ni­tive aging and risk of schiz­o­phre­nia.

fat/obesity; “Child­hood Intel­li­gence and Adult Obe­sity”, Kanazawa 2013

Design and Meth­ods: A pop­u­la­tion (n=17,419) of British babies has been fol­lowed since birth in 1958 in a prospec­tively lon­gi­tu­di­nal study. Child­hood gen­eral intel­li­gence is mea­sured at 7, 11, and 16, and adult BMI and obe­sity are mea­sured at 51. Results: Child­hood gen­eral intel­li­gence has a direct effect on adult BMI, obe­si­ty, and weight gain, net of edu­ca­tion, earn­ings, moth­er’s BMI, father’s BMI, child­hood social class, and sex. More intel­li­gent chil­dren grow up to eat more healthy foods and exer­cise more fre­quently as adults. Con­clu­sion: Child­hood intel­li­gence has a direct effect on adult obe­sity unmedi­ated by edu­ca­tion or earn­ings. Gen­eral intel­li­gence decreases BMI only in adult­hood when indi­vid­u­als have com­plete con­trol over what they eat.

…Obe­sity is just one of a large num­ber of health prob­lems that afflict less intel­li­gent indi­vid­u­als, increase their mor­tal­i­ty, and decrease their life expectancy (7-9)…Thus, rel­a­tive to their less intel­li­gent coun­ter­parts, more intel­li­gent chil­dren are more likely to grow up to espouse the evo­lu­tion­ar­ily novel val­ues of left­-wing lib­er­al­ism or athe­ism (13); to be noc­tur­nal (15); to con­sume the evo­lu­tion­ar­ily novel sub­stances of alco­hol, tobac­co, and psy­choac­tive drugs (16); to pre­fer evo­lu­tion­ar­ily novel instru­men­tal music such as clas­si­cal and light music (17); and regard­less of their genetic and hor­monal pre­dis­po­si­tion, to engage in evo­lu­tion­ar­ily novel homo­sex­ual behav­ior (18)…The avail­able evi­dence sug­gests that more intel­li­gent indi­vid­u­als are more likely to exer­cise more fre­quently than less intel­li­gent indi­vid­u­als (21,22)…­Con­sis­tent with the pre­dic­tion of the Hypoth­e­sis, Teas­dale et al. (23) report that, among a sam­ple of 26,274 young Dan­ish men, intel­li­gence and body-mass index (BMI) are sig­nifi­cantly neg­a­tively cor­re­lated even net of edu­ca­tion.

  • Batty GD, Deary IJ, Got­tfred­son LS. Pre­mor­bid (early life) IQ and later mor­tal­ity risk: sys­tem­atic review. Ann Epi­demiol 2007;17:278-288.
  • Got­tfred­son LS, Deary IJ. Intel­li­gence pre­dicts health and longevi­ty, but why? Curr Direct Psy­chol Sci 2004;13:1-4
  • Kanazawa S. Mind the gap…in intel­li­gence: reex­am­in­ing the rela­tion­ship between inequal­ity and health. Br J Health Psy­chol 2006;11:623-642
  • Kanazawa S. Why lib­er­als and athe­ists are more intel­li­gent. Soc Psy­chol Q 2010; 73:33-57
  • Kanazawa S, Perina K. Why night owls are more intel­li­gent. Pers Indi­vid Differ­ences 2009;47:685-690
  • Kanazawa S, Hell­berg JEEU. Intel­li­gence and sub­stance use. Rev Gen Psy­chol 2010;14:382-396
  • Kanazawa S, Perina K. Why more intel­li­gent indi­vid­u­als like clas­si­cal music. J Behav­ior Decis Mak­ing 2012;25:264-275.
  • Kanazawa S. Intel­li­gence and homo­sex­u­al­i­ty. J Biosoc Sci 2012;44:595-623.
  • Hall PA, Elias LJ, Fong GT, Har­ri­son AH, Borowsky R, Sarty GE. A social neu­ro-
  • sci­ence per­spec­tive on phys­i­cal activ­i­ty. J Sport Exer­cise Psy­chol 2008;30:432-449.
  • Kingma EM, Tak LM, Huis­man M, Ros­malen JGM. Intel­li­gence is neg­a­tively asso­ci­ated with the num­ber of func­tional somatic symp­toms. J Epi­demiol Com­mun Health 2009;63:900-906.
  • Teas­dale TW, Sørensen TIA, Stunkard AJ. Intel­li­gence and edu­ca­tional level in rela­tion to body mass index of adult males. Hum Biol 1992;64:99-106.

, Yeo et al 2011

Phe­no­typic vari­a­tion in human intel­lec­tual func­tion­ing shows sub­stan­tial her­i­tabil­i­ty, as demon­strated by a long his­tory of behav­ior genetic stud­ies. Many recent mol­e­c­u­lar genetic stud­ies have attempted to uncover spe­cific genetic vari­a­tions respon­si­ble for this her­i­tabil­i­ty, but iden­ti­fied effects cap­ture lit­tle vari­ance and have proven diffi­cult to repli­cate. The present study, moti­vated an inter­est in “muta­tion load” emerg­ing from evo­lu­tion­ary per­spec­tives, exam­ined the impor­tance of the num­ber of rare (or infre­quent) copy num­ber vari­a­tions (CNVs), and the total num­ber of base pairs included in such dele­tions, for psy­cho­me­t­ric intel­li­gence. Genetic data was col­lected using the Illu­mina 1MDuoBead­Chip Array from a sam­ple of 202 adult indi­vid­u­als with alco­hol depen­dence, and a sub­set of these (N = 77) had been admin­is­tered the Wech­sler Abbre­vi­ated Scale of Intel­li­gence (WASI). After remov­ing CNV out­liers, the impact of rare genetic dele­tions on psy­cho­me­t­ric intel­li­gence was inves­ti­gated in 74 indi­vid­u­als. The total length of the rare dele­tions sig­nifi­cantly and neg­a­tively pre­dicted intel­li­gence (r = −.30, p = .01). As prior stud­ies have indi­cated greater her­i­tabil­ity in indi­vid­u­als with rel­a­tively higher parental socioe­co­nomic sta­tus (SES), we also exam­ined the impact of eth­nic­ity (Anglo/White vs. Oth­er), as a proxy mea­sure of SES; these groups did not differ on any genetic vari­able. This cat­e­gor­i­cal vari­able sig­nifi­cantly mod­er­ated the effect of length of dele­tions on intel­li­gence, with larger effects being noted in the Anglo/White group. Over­all, these results sug­gest that rare dele­tions (be­tween 5% and 1% pop­u­la­tion fre­quency or less) adversely affect intel­lec­tual func­tion­ing, and that pleiotropic effects might partly account for the asso­ci­a­tion of intel­li­gence with health and men­tal health sta­tus. Sig­nifi­cant lim­i­ta­tions of this research, includ­ing issues of gen­er­al­iz­abil­ity and CNV mea­sure­ment, are dis­cussed.

“Effi­ciency of func­tional brain net­works and intel­lec­tual per­for­mance”, van den Heuvel 2009

Our brain is a com­plex net­work in which infor­ma­tion is con­tin­u­ously processed and trans­ported between spa­tially dis­trib­uted but func­tion­ally linked regions. Recent stud­ies have shown that the func­tional con­nec­tions of the brain net­work are orga­nized in a highly effi­cient smal­l­-world man­ner, indi­cat­ing a high level of local neigh­bor­hood clus­ter­ing, together with the exis­tence of more long-dis­tance con­nec­tions that ensure a high level of global com­mu­ni­ca­tion effi­ciency within the over­all net­work. Such an effi­cient net­work archi­tec­ture of our func­tional brain raises the ques­tion of a pos­si­ble asso­ci­a­tion between how effi­ciently the regions of our brain are func­tion­ally con­nected and our level of intel­li­gence. Exam­in­ing the over­all orga­ni­za­tion of the brain net­work using graph analy­sis, we show a strong neg­a­tive asso­ci­a­tion between the nor­mal­ized char­ac­ter­is­tic path length lambda of the rest­ing-s­tate brain net­work and intel­li­gence quo­tient (IQ). This sug­gests that human intel­lec­tual per­for­mance is likely to be related to how effi­ciently our brain inte­grates infor­ma­tion between mul­ti­ple brain regions. Most pro­nounced effects between nor­mal­ized path length and IQ were found in frontal and pari­etal regions. Our find­ings indi­cate a strong pos­i­tive asso­ci­a­tion between the global effi­ciency of func­tional brain net­works and intel­lec­tual per­for­mance.

“Genet­ics of Brain Fiber Archi­tec­ture and Intel­lec­tual Per­for­mance”, Chi­ang et al 2009

The study is the first to ana­lyze genetic and envi­ron­men­tal fac­tors that affect brain fiber archi­tec­ture and its genetic link­age with cog­ni­tive func­tion. We assessed white mat­ter integrity vox­el-wise using diffu­sion ten­sor imag­ing at high mag­netic field (4 Tes­la), in 92 iden­ti­cal and fra­ter­nal twins. White mat­ter integri­ty, quan­ti­fied using frac­tional anisotropy (FA), was used to fit struc­tural equa­tion mod­els (SEM) at each point in the brain, gen­er­at­ing three­-di­men­sional maps of her­i­tabil­i­ty. We visu­al­ized the anatom­i­cal pro­file of cor­re­la­tions between white mat­ter integrity and ful­l-s­cale, ver­bal, and per­for­mance intel­li­gence quo­tients (FIQ, VIQ, and PIQ). White mat­ter integrity (FA) was under strong genetic con­trol and was highly her­i­ta­ble in bilat­eral frontal (a2 = 0.55, p = 0.04, left; a2 = 0.74, p = 0.006, right), bilat­eral pari­etal (a2 = 0.85, p < 0.001, left; a2 = 0.84, p < 0.001, right), and left occip­i­tal (a2 = 0.76, p = 0.003) lobes, and was cor­re­lated with FIQ and PIQ in the cin­gu­lum, optic radi­a­tions, supe­rior fron­to-oc­cip­i­tal fas­ci­cu­lus, inter­nal cap­sule, cal­losal isth­mus, and the corona radi­ata (p = 0.04 for FIQ and p = 0.01 for PIQ, cor­rected for mul­ti­ple com­par­ison­s). In a cross-trait map­ping approach, com­mon genetic fac­tors medi­ated the cor­re­la­tion between IQ and white mat­ter integri­ty, sug­gest­ing a com­mon phys­i­o­log­i­cal mech­a­nism for both, and com­mon genetic deter­mi­na­tion. These genetic brain maps reveal her­i­ta­ble aspects of white mat­ter integrity and should expe­dite the dis­cov­ery of sin­gle-nu­cleotide poly­mor­phisms affect­ing fiber con­nec­tiv­ity and cog­ni­tion.

“Smarter peo­ple are (a bit) more sym­met­ri­cal: A meta-analy­sis of the rela­tion­ship between intel­li­gence and fluc­tu­at­ing asym­me­try”, Banks et al 2010

Indi­vid­ual differ­ences in gen­eral men­tal abil­ity (g) have impor­tant impli­ca­tions across mul­ti­ple dis­ci­plines. Research sug­gests that the vari­ance in g may be due to a gen­eral fit­ness fac­tor. If this is the case, a rela­tion­ship should exist between g and other reli­able indi­ca­tors of fit­ness. Some empir­i­cal results indi­cate a rela­tion­ship between g and fluc­tu­at­ing asym­me­try. How­ev­er, there have been incon­sis­ten­cies in the results, some of which may be due to ran­dom sam­pling error, and some of which may be due to mod­er­at­ing vari­ables, pub­li­ca­tion bias, and method­olog­i­cal issues. To help clar­ify the lit­er­a­ture, a meta-analy­sis was con­ducted on the rela­tion­ship between g and fluc­tu­at­ing asym­me­try. Based on 14 sam­ples across 1871 peo­ple, esti­mates of the pop­u­la­tion cor­re­la­tion ranged from −.12 to −.20. There was a differ­ence in the mag­ni­tude of the cor­re­la­tion between pub­lished stud­ies and unpub­lished stud­ies with pub­lished stud­ies show­ing larger mag­ni­tude neg­a­tive cor­re­la­tions and unpub­lished stud­ies yield­ing results closer to zero. The impli­ca­tions for our under­stand­ing of g and its rela­tion­ship with fluc­tu­at­ing asym­me­try are dis­cussed.

“The Inher­i­tance of Inequal­ity”, Bowles & Gin­tis 2002

Cor­re­la­tions of IQ between par­ents and off­spring range from 0.42 to 0.72, where the higher fig­ure refers to mea­sures of aver­age parental and aver­age off­spring IQ (Bouchard and McGue, 1981; Plomin et al., 2000).

We have located 65 esti­mates of the nor­mal­ized regres­sion coeffi­cient of a test score in an earn­ings equa­tion in 24 differ­ent stud­ies of U.S. data over a period of three decades. Our meta-analy­sis of these stud­ies is pre­sented in Bowles, Gin­tis and Osborne (2002). The mean of these esti­mates is 0.15, indi­cat­ing that a stan­dard devi­a­tion change in the cog­ni­tive score, hold­ing con­stant the remain­ing vari­ables (in­clud­ing school­ing), changes the nat­ural log­a­rithm of earn­ings by about one-sev­enth of a stan­dard devi­a­tion. By con­trast, the mean value of the nor­mal­ized regres­sion coeffi­cient of years of school­ing in the same equa­tion pre­dict­ing the nat­ural log of earn­ings in these stud­ies is 0.22, sug­gest­ing a some­what larger inde­pen­dent effect of school­ing. We checked to see if these results were depen­dent on the weight of over­rep­re­sented authors, the type of cog­ni­tive test used, at what age the test was taken and other differ­ences among the stud­ies and found no sig­nifi­cant effects. An esti­mate of the causal impact of child­hood IQ on years of school­ing (also nor­mal­ized) is 0.53 (Win­ship and Koren­man, 1999). A rough esti­mate of the direct and indi­rect effect of IQ on earn­ings, call it b, is then b ϭ 0.15 ϩ (0.53)(0.22) ϭ 0.266.

  • Bowles, Gin­tis and Osborne (2002) “The Deter­mi­nants of Indi­vid­ual Earn­ings: Skills, Pref­er­ences, and School­ing.” Jour­nal of Eco­nomic Lit­er­a­ture. Decem­ber, 39:4, pp. 1137-176
  • Win­ship and Koren­man, 1999 “Eco­nomic Suc­cess and the Evo­lu­tion of School­ing with Men­tal Abil­ity”, in Earn­ing and Learn­ing: How Schools Mat­ter. Susan Mayer and Paul Peter­son, eds. Wash­ing­ton, D.C.: Brook­ings Insti­tu­tion, pp. 49-78

A meta-analy­sis of 63 stud­ies showed a sig­nifi­cant neg­a­tive asso­ci­a­tion between intel­li­gence and reli­gios­i­ty. The asso­ci­a­tion was stronger for col­lege stu­dents and the gen­eral pop­u­la­tion than for par­tic­i­pants younger than col­lege age; it was also stronger for reli­gious beliefs than reli­gious behav­ior. For col­lege stu­dents and the gen­eral pop­u­la­tion, means of weighted and unweighted cor­re­la­tions between intel­li­gence and the strength of reli­gious beliefs ranged from −.20 to −.25 (mean r = −.24). Three pos­si­ble inter­pre­ta­tions were dis­cussed. First, intel­li­gent peo­ple are less likely to con­form and, thus, are more likely to resist reli­gious dog­ma. Sec­ond, intel­li­gent peo­ple tend to adopt an ana­lytic (as opposed to intu­itive) think­ing style, which has been shown to under­mine reli­gious beliefs. Third, sev­eral func­tions of reli­gios­i­ty, includ­ing com­pen­satory con­trol, self­-reg­u­la­tion, self­-en­hance­ment, and secure attach­ment, are also con­ferred by intel­li­gence. Intel­li­gent peo­ple may there­fore have less need for reli­gious beliefs and prac­tices.

“The Rela­tion Between Intel­li­gence and Reli­gios­i­ty: A Meta-Analy­sis and Some Pro­posed Expla­na­tions” Zuck­er­man et al 2013:

A meta-analy­sis of 63 stud­ies showed a sig­nifi­cant neg­a­tive asso­ci­a­tion between intel­li­gence and reli­gios­i­ty. The asso­ci­a­tion was stronger for col­lege stu­dents and the gen­eral pop­u­la­tion than for par­tic­i­pants younger than col­lege age; it was also stronger for reli­gious beliefs than reli­gious behav­ior. For col­lege stu­dents and the gen­eral pop­u­la­tion, means of weighted and unweighted cor­re­la­tions between intel­li­gence and the strength of reli­gious beliefs ranged from −.20 to −.25 (mean r = −.24). Three pos­si­ble inter­pre­ta­tions were dis­cussed. First, intel­li­gent peo­ple are less likely to con­form and, thus, are more likely to resist reli­gious dog­ma. Sec­ond, intel­li­gent peo­ple tend to adopt an ana­lytic (as opposed to intu­itive) think­ing style, which has been shown to under­mine reli­gious beliefs. Third, sev­eral func­tions of reli­gios­i­ty, includ­ing com­pen­satory con­trol, self­-reg­u­la­tion, self­-en­hance­ment, and secure attach­ment, are also con­ferred by intel­li­gence. Intel­li­gent peo­ple may there­fore have less need for reli­gious beliefs and prac­tices.

…To our knowl­edge, the first stud­ies on intel­li­gence and reli­gios­ity appeared in 1928, in the Uni­ver­sity of Iowa Stud­ies series, Stud­ies in Char­ac­ter (How­ells, 1928; Sin­clair, 1928). These stud­ies exam­ined sen­so­ry, motor, and cog­ni­tive cor­re­lates of reli­gios­i­ty. Intel­li­gence tests were included in the bat­tery of admin­is­tered tasks. Both How­ells (1928) and Sin­clair (1928) found that higher lev­els of intel­li­gence were related to lower lev­els of reli­gios­i­ty. Accu­mu­la­tion of addi­tional research dur­ing the sub­se­quent three decades prompted Argyle (1958) to review the avail­able evi­dence. He con­cluded that “intel­li­gent stu­dents are much less likely to accept ortho­dox beliefs, and rather less likely to have pro-re­li­gious atti­tudes” (p. 96). Argyle also noted that, as of 1958, all avail­able evi­dence was based on chil­dren or col­lege stu­dent sam­ples. He spec­u­lat­ed, how­ev­er, that the same results might be observed for adults of post-col­lege age. In the sub­se­quent decade, the pen­du­lum swung in the oppo­site direc­tion. Kosa and Schom­mer (1961) and Hoge (1969) drew con­clu­sions from their data that were incon­sis­tent with those of Argyle (1958). Accord­ing to Kosa and Schom­mer, “social envi­ron­ment reg­u­lates the rela­tion­ship of men­tal abil­i­ties and reli­gious atti­tudes by chan­nel­ing the intel­li­gence into cer­tain approved direc­tions: a sec­u­lar-ori­ented envi­ron­ment may direct it toward skep­ti­cism, a church-ori­ented envi­ron­ment may direct it toward increased reli­gious inter­est” (p. 90). They found that in a Catholic col­lege, more intel­li­gent stu­dents knew more about reli­gious doc­trine and par­tic­i­pated more in strictly reli­gious orga­ni­za­tions.

…Hoge (1969, 1974) tracked changes in reli­gious atti­tudes on 13 Amer­i­can cam­pus­es. He com­pared sur­vey data, most of which were col­lected between 1930 and 1948, with data that he col­lected him­self in 1967 and 1968. On four cam­pus­es, Hoge also exam­ined the rela­tion between SAT scores and reli­gious atti­tudes. Cor­re­la­tions were small and mostly neg­a­tive. Hoge (1969) con­cluded that “no organic or psy­chic rela­tion­ship exists between intel­li­gence and reli­gious atti­tudes and . . . the rela­tion­ships found by researchers are either due to edu­ca­tional influ­ences or biases in the intel­li­gence tests” (p. 215). Hoge acknowl­edged that range restric­tions of col­lege stu­dents’ intel­li­gence scores may decrease cor­re­la­tions between intel­li­gence and other vari­ables. Nev­er­the­less, he con­cluded that the low neg­a­tive-in­tel­li­gence-re­li­gios­ity cor­re­la­tions implied that there is no rela­tion between intel­li­gence and reli­gios­i­ty.

…As if in response to Beit-Hal­lahmi and Argyle’s (1997) call, the last decade has seen a num­ber of large-s­cale stud­ies that exam­ined the rela­tion between intel­li­gence and reli­gios­ity (Kanaza­wa, 2010a; Lewis, Ritchie, & Bates, 2011; Nyborg, 2009; Sherkat, 2010). Kanazawa (2010a), Sherkat (2010), and Lewis et al. (2011) all found neg­a­tive rela­tions between intel­li­gence and reli­gios­ity in post-col­lege adults. Nyborg (2009) found that young athe­ists (age 12 to 17) scored sig­nifi­cantly higher on an intel­li­gence test than reli­gious youth. The last decade also saw stud­ies on the rela­tion between intel­li­gence and reli­gios­ity at the group lev­el. Using data from 137 nations, Lynn, Har­vey, and Nyborg (2009) found a neg­a­tive rela­tion between mean intel­li­gence scores (com­puted for each nation) and mean reli­gios­ity scores. How­ev­er, IQ scores from unde­vel­oped and/or non-West­ern­ized coun­tries might have lim­ited valid­ity because most tests were devel­oped for West­ern cul­tures. Low lev­els of lit­er­acy and prob­lems in obtain­ing rep­re­sen­ta­tive sam­ples in some coun­tries may also under­mine the valid­ity of these find­ings (Hunt, 2011; Richards, 2002; Volken, 2003). In response to these cri­tiques, Reeve (2009) repeated the analy­sis but set all national IQ scores lower than 90 to 90. The result­ing IQ-re­li­gios­ity cor­re­la­tion was not lower than what had been reported in prior stud­ies (see Reeve, 2009, for a dis­cus­sion of his trun­cat­ing pro­ce­dure). In the same vein, Pes­ta, McDaniel, and Bertsch (2010) found a neg­a­tive rela­tion between intel­li­gence and reli­gios­ity scores that were com­puted for all 50 states in the United States. These results are less sus­cep­ti­ble to the prob­lems (e.g., cul­tural differ­ences) that plagued stud­ies at the coun­try lev­el.

…Stud­ies in this area have found that, rel­a­tive to the gen­eral pub­lic, sci­en­tists are less likely to believe in God. For exam­ple, Leuba (1916) reported that 58% of ran­domly selected sci­en­tists in the United States expressed dis­be­lief in, or doubt regard­ing the exis­tence of God; this pro­por­tion rose to nearly 70% for the most emi­nent sci­en­tists. Lar­son and Witham (1998) reported sim­i­lar results, as evi­denced by the title of their arti­cle-“Lead­ing sci­en­tists still reject God.” Of course, higher intel­li­gence is only one of a num­ber of fac­tors that can account for these results.

…Out­side of aca­d­e­mic jour­nals, how­ev­er, there have been at least two reviews (Beck­with, 1986; Bell, 2002). Beck­with (1986) con­cluded that 39 of the 43 stud­ies that he sum­ma­rized sup­ported a neg­a­tive rela­tion between intel­li­gence and reli­gios­i­ty, and Bell (2002) sim­ply repeated this tal­ly. How­ev­er, some of the stud­ies reviewed by Beck­with were only indi­rectly rel­e­vant (e.g., com­par­isons between more and less pres­ti­gious uni­ver­si­ties), and some rel­e­vant stud­ies were exclud­ed.

…The first row of Table 2 presents basic sta­tis­tics describ­ing the rela­tion between intel­li­gence and reli­gios­ity for all 63 stud­ies. Results are pre­sented for ran­dom-effects analy­ses (un­weighted mean cor­re­la­tions) and fixed-effects analy­ses (weighted mean cor­re­la­tion­s). Fifty-three stud­ies showed neg­a­tive cor­re­la­tions while 10 stud­ies showed pos­i­tive cor­re­la­tions. Thir­ty-seven stud­ies showed sig­nifi­cant cor­re­la­tions; of the­se, 35 were neg­a­tive and 2 were pos­i­tive. The unweighted mean cor­re­la­tion (r) between intel­li­gence and reli­gios­ity was −.16, the median r was −.14, and the weighted mean r was −.13. The sim­i­lar­ity of these three indi­ca­tors of cen­tral ten­dency indi­cates that the dis­tri­b­u­tion was approx­i­mately sym­met­ri­cal and was not skewed by sev­eral very large stud­ies that were in the data­base. Ran­dom- and fixed-effects mod­els yielded sig­nifi­cant evi­dence that the higher a per­son’s intel­li­gence, the lower the per­son scored on the reli­gios­ity mea­sures.

…When GPA[grade-point-average]-religiosity cor­re­la­tions from the five stud­ies using only GPA are com­bined with GPA-religiosity cor­re­la­tions from the four stud­ies using GPA as well as other intel­li­gence mea­sures, the mean GPA-religiosity cor­re­la­tion was not sig­nifi­cantly differ­ent from zero, MGPA = −.027, p = .33. It was con­cluded that GPA had no mean­ing­ful rela­tion to reli­gios­ity and, accord­ing­ly, all sub­se­quent analy­ses omit­ted the five stud­ies that used only GPA.

…As expect­ed, the extreme groups effect size (M = −.43) that was sig­nifi­cantly more neg­a­tive than that of the unbi­ased stud­ies (p < .001 by post hoc least sig­nifi­cant differ­ence [LSD] test).

…The fixed-effects trim and fill method for detect­ing pos­si­ble pub­li­ca­tion bias yielded neg­li­gi­ble impact for the pre-col­lege and non-col­lege groups. For the col­lege group, how­ev­er, there was evi­dence of pub­li­ca­tion bias, such that nine neg­a­tive effect sizes would need to be added to yield a sym­met­ri­cal dis­tri­b­u­tion. The impu­ta­tion of these effects resulted in an adjusted mean effect of −.21, notice­ably quite differ­ent from the observed weighted mean effect of −.15. Because the adjusted effect size is hypo­thet­i­cal, it will not be incor­po­rated into sub­se­quent analy­ses. How­ev­er, this result and the range restric­tion in intel­li­gence scores in this group sug­gest that the true intel­li­gence-re­li­gios­ity rela­tion in the col­lege pop­u­la­tion may be more neg­a­tive than the lit­er­a­ture indi­cates.

…As an exploratory analy­sis, we exam­ined the rela­tion between per­cent­age of males in each study and effect size of the intel­li­gence-re­li­gios­ity rela­tion. In the 34 stud­ies in which it could be deter­mined, per­cent­age of males was pos­i­tively cor­re­lated with unweighted effect sizes, r(32) = .50, p < .01. This cor­re­la­tion indi­cates that the neg­a­tive intel­li­gence-re­li­gios­ity rela­tion was less neg­a­tive in stud­ies with more males. This rela­tion held in terms of mag­ni­tude for the pre-col­lege and col­lege groups, r(6) = .48, ns, and r(12) = .51, p = .06, but was weaker at the non-col­lege lev­el, r(10) = .19, ns. When ana­lyzed as a fixed-effects regres­sion, the rela­tion between per­cent­age of males and effect size was also markedly pos­i­tive, p < .001. A more direct test of the pos­si­bil­ity that the intel­li­gence-re­li­gios­ity rela­tion is less neg­a­tive for males is a with­in-s­tudy com­par­i­son between males and females. Kanaza­wa1 con­ducted this test for two stud­ies (Kanaza­wa, 2010a; com­bined N = 21,437). If any­thing, the results pointed in the oppo­site direc­tion. The intel­li­gence-re­li­gios­ity cor­re­la­tions for females and males, respec­tive­ly, were −.11 and −.12 in Study 1, and −.14 and −.16 in Study 2. Although the differ­ence between females and males was not sig­nifi­cant, even when com­bined meta-an­a­lyt­i­cally across stud­ies (Z = 1.39, p = .16), the direc­tion of this differ­ence is incon­sis­tent with the between-s­tud­ies find­ing of the meta-analy­sis.

…As pre­vi­ously not­ed, some inves­ti­ga­tors sug­gested that edu­ca­tion medi­ates the rela­tion between intel­li­gence and reli­gios­ity (Hoge, 1974; Reeve & Basa­lik, 2011). Inter­est­ing­ly, Kanazawa (S. Kanaza­wa, per­sonal com­mu­ni­ca­tion, Jan­u­ary 2012) espouses an oppos­ing view, namely that intel­li­gence accounts for any neg­a­tive rela­tion between edu­ca­tion and reli­gios­i­ty. Table 8 presents results that address the two com­pet­ing hypothe­ses. The analy­ses are based on seven stud­ies from three sources. Results from the stu­dent sam­ple stud­ied by Blan­chard-Fields, Hert­zog, Stein, and Pak (2001; first row in Table 8) can be excluded because of range restric­tion for intel­li­gence and edu­ca­tion (in­deed, all cor­re­la­tions for that study were weak). The results of the remain­ing six stud­ies indi­cate that edu­ca­tion does not medi­ate the intel­li­gence-re­li­gios­ity rela­tion. To begin with, intel­li­gence was more neg­a­tively related to reli­gios­ity than was edu­ca­tion (un­weighted mean cor­re­la­tions were −.18 and −.06, respec­tive­ly). We tested the sig­nifi­cance of this differ­ence sep­a­rately for each study, using a pro­ce­dure for com­par­ing non­in­de­pen­dent cor­re­la­tions (Meng, Rosen­thal, & Rubin, 1992); the com­bined differ­ence across the six stud­ies was highly sig­nifi­cant, Z = 9.32, p < .001. Fur­ther­more, con­trol­ling for edu­ca­tion did not have much of an effect on the intel­li­gence-re­li­gios­ity rela­tion-un­weighted means of the six zero-order and par­tial cor­re­la­tions were −.18 and −.17, respec­tive­ly. In con­trast, con­trol­ling for intel­li­gence led to a some­what greater change in the edu­ca­tion-re­li­gios­ity rela­tion; the unweighted means for the six zero-order and par­tial cor­re­la­tions were −.06 and .00, respec­tive­ly. This find­ing is con­sis­tent with S. Kanaza­wa’s (per­sonal com­mu­ni­ca­tion, Jan­u­ary 2010) view that intel­li­gence accounts for the edu­ca­tion-re­li­gios­ity rela­tion. How­ev­er, given that the analy­sis is based on only six stud­ies, our con­clu­sions are ten­ta­tive.

…Table 9 presents the find­ings. In all four com­par­isons, Ter­man’s sam­ple scored sig­nifi­cantly lower on reli­gios­ity than the gen­eral pub­lic (the aver­age of these effects was used in the meta-analy­sis as one of the extreme groups’ stud­ies). Admit­ted­ly, the years of data col­lec­tion and ages of the two groups do not match per­fect­ly. How­ev­er, the results are so strong that it is diffi­cult to imag­ine that more exact match­ing would make a differ­ence. These results are even more strik­ing if the Ter­mites’ reli­gious upbring­ing is con­sid­ered. Ter­man and Oden (1959) reported that close to 60% of Ter­mites reported that they received “very strict” or “con­sid­er­able” reli­gious train­ing; approx­i­mately 33% reported receiv­ing lit­tle train­ing, and about 6% reported no reli­gious train­ing. This sug­gests that the Ter­mites under­went changes in their reli­gios­ity after their child­hood. …In Ter­man’s sam­ple (N = 410), 1.2% checked the reli­gious option, com­pared with .4% in the Hunter group (N = 139), Z < 1. These results sug­gest that on an absolute lev­el, reli­gion was rel­a­tively unim­por­tant to mid­dle-aged adults who were iden­ti­fied as gifted in child­hood in both sam­ples. In addi­tion, we spec­u­late that if the Hunter sam­ple is sim­i­lar to the Ter­man sam­ple with respect to reli­gios­i­ty, it too may be less reli­gious than the gen­eral pop­u­la­tion. In the Ter­man and the Hunter sam­ples, a high intel­li­gence level at an early age pre­ceded lower reli­gios­ity many years lat­er. How­ev­er, our analy­ses of these results nei­ther con­trolled for pos­si­ble rel­e­vant fac­tors at an early age (e.g., socioe­co­nomic sta­tus) nor exam­ined pos­si­ble medi­a­tors (e.g., occu­pa­tion) of this rela­tion.

…In­tel­li­gence can be reli­ably mea­sured at a very early age while reli­gios­ity can­not (e.g., Jensen, 1998; Larsen, Hart­mann, & Nyborg, 2008). In their clas­sic study, for exam­ple, H. E. Jones and Bay­ley (1941) showed that the mean of intel­li­gence scores assessed at ages 17 and 18 (a) cor­re­lated .86 with the mean scores assessed at ages 5, 6, and 7; and (b) cor­re­lated .96 with the mean of intel­li­gence scores assessed at ages 11, 12, and 13. Because intel­li­gence can be mea­sured at an early age, it can be used to pre­dict out­comes observed years lat­er. For exam­ple, Deary, Strand, Smith, and Fer­nan­des (2007) reported a .69 cor­re­la­tion between intel­li­gence mea­sured at age 11 and edu­ca­tional achieve­ment at age 16. Unlike intel­li­gence, reli­gios­ity assessed at an early age is a weak pre­dic­tor of reli­gios­ity assessed years lat­er. For exam­ple, Willits and Crider (1989) found only small to mod­er­ate cor­re­la­tions between reli­gios­ity at age 16 and that at 27 (.28 for church atten­dance and .36 for belief­s). O’Con­nor, Hoge, and Alexan­der (2002) found no rela­tion­ship between mea­sures of church involve­ment at ages 16 and 38.

…First, although the preva­lence of reli­gios­ity varies widely among coun­tries and cul­tures, more than 50% of the world pop­u­la­tion con­sider them­selves reli­gious. Using sur­vey data col­lected by P. Zuck­er­man (2007) from 137 coun­tries, Lynn et al. (2009) and Reeve (2009) observed a preva­lence of 89.9% believ­ers in the world and 89.5% believ­ers in the United States. How­ev­er, a recent Win-Gallup Inter­na­tional (2012) poll of 59,927 per­sons in 57 coun­tries found that only 59% of the respon­dents (60% in the United States) con­sider them­selves reli­gious, a decline of 9% (13% in the United States) from a sim­i­lar 2005 poll. Athe­ism might be con­sid­ered a case of non­con­for­mity in soci­eties where the major­ity is reli­gious. This is not so, how­ev­er, if one grows up in largely athe­ist soci­eties, such as those that exist in Scan­di­navia (P. Zuck­er­man, 2008).

…There is also empir­i­cal evi­dence sug­gest­ing that reli­gios­ity may be an in-group phe­nom­e­non, rein­forc­ing proso­cial ten­den­cies within the group (see a review by Noren­za­yan & Ger­vais, 2012), but also pre­dis­pos­ing believ­ers to reject out­-groups mem­bers (see meta-analy­sis by D. L. Hall, Matz, & Wood, 2010). To become an athe­ist, there­fore, it may be nec­es­sary to resist the in-group dogma of reli­gious beliefs. Not sur­pris­ing­ly, there is evi­dence of anti-athe­ist dis­trust and prej­u­dice (Ger­vais, Shar­iff, & Noren­za­yan, 2011; Ger­vais & Noren­za­yan, 2012b; for a review, see Noren­za­yan & Ger­vais, 2012).

…In­tel­li­gence also con­fers a sense of per­sonal con­trol. We iden­ti­fied eight stud­ies that reported cor­re­la­tions between intel­li­gence and belief in per­sonal con­trol (Grover & Hert­zog, 1991; Lach­man, 1983; Lach­man, Bal­tes, Nes­sel­roade, & Willis, 1982; Martel, McK­elvie, & Stand­ing, 1987; Miller & Lach­man, 2000; Prenda & Lach­man, 2001; Tolor & Reznikoff, 1967; P. Wood & Englert, 2009). All eight cor­re­la­tions were pos­i­tive, with a mean cor­re­la­tion (weighted by df of each study) of .29. In addi­tion, higher intel­li­gence is asso­ci­ated with greater self­-effi­ca­cy-the belief in one’s own abil­ity to achieve val­ued goals (Ban­dura, 1997). This con­struct is sim­i­lar to per­sonal con­trol beliefs but has been exam­ined sep­a­rately in the lit­er­a­ture. In a meta-analy­sis of 26 stud­ies, the mean cor­re­la­tion between intel­li­gence and self­-effi­cacy was .20 (Judge, Jack­son, Shaw, Scott, & Rich, 2007).

…Choos­ing the large delayed reward serves as an indi­ca­tor of self­-con­trol. Shamosh and Gray (2008) meta-an­a­lyzed the rela­tion between intel­li­gence and delay dis­count­ing (the lat­ter con­struct is iden­ti­cal to delay of grat­i­fi­ca­tion except that high delay dis­count­ing indi­cates poor self­-con­trol). Their analy­sis, based on 26 stud­ies, yielded a mean r of −.23. This sug­gests that intel­li­gent peo­ple are more likely to delay grat­i­fi­ca­tion (i.e., less likely to engage in delay dis­count­ing).

…On the other hand and in line with Kanaza­wa’s (2010a) mod­el, genetic influ­ences have been impli­cated not only in intel­li­gence (cf., Nis­bett et al., 2012b), but also in reli­gios­ity (D’Onofrio, Eaves, Mur­relle, Maes, & Spilka, 1999; Koenig, McGue, & Iacono, 2008). Fur­ther­more, the model was used to pre­dict other cor­re­lates of intel­li­gence (e.g., polit­i­cal lib­er­al­ism and, for men, monogamy), and those pre­dic­tions received empir­i­cal sup­port. In con­clu­sion, Kanaza­wa’s (2010a) inter­pre­ta­tion remains an intrigu­ing pos­si­bil­i­ty.

…This func­tion was not included in our dis­cus­sion of func­tional equiv­a­lence because, to the best of our knowl­edge, there is no evi­dence per­tain­ing to the rela­tion between intel­li­gence and death anx­i­ety. Although this logic sug­gests that the neg­a­tive rela­tion between intel­li­gence and reli­gios­ity might decline at the end of life, the rel­e­vant evi­dence we have indi­cates oth­er­wise. The highly intel­li­gent mem­bers of Ter­man’s sam­ple retained lower reli­gios­ity scores (rel­a­tive to the gen­eral pop­u­la­tion) even at 75 to 91 years of age (Table 9). Addi­tional research is needed to resolve this issue.

“Intel­li­gence (IQ) as a Pre­dic­tor of Life Suc­cess”, Firkowska-Mankiewicz 2002

“Lead­ing sci­en­tists still reject God”, Lar­son & Witham 1998

Research on this topic began with the emi­nent US psy­chol­o­gist James H. Leuba and his land­mark sur­vey of 1914. He found that 58% of 1,000 ran­domly selected US sci­en­tists expressed dis­be­lief or doubt in the exis­tence of God, and that this fig­ure rose to near 70% among the 400 “greater” sci­en­tists within his sam­ple^1. Leuba repeated his sur­vey in some­what differ­ent form 20 years lat­er, and found that these per­cent­ages had increased to 67 and 85, respec­tive­ly^2.

In 1996, we repeated Leuba’s 1914 sur­vey and reported our results in Nature^3. We found lit­tle change from 1914 for Amer­i­can sci­en­tists gen­er­al­ly, with 60.7% express­ing dis­be­lief or doubt. This year, we closely imi­tated the sec­ond phase of Leuba’s 1914 sur­vey to gauge belief among “greater” sci­en­tists, and find the rate of belief lower than ever - a mere 7% of respon­dents.

…Our cho­sen group of “greater” sci­en­tists were mem­bers of the National Acad­emy of Sci­ences (NAS). Our sur­vey found near uni­ver­sal rejec­tion of the tran­scen­dent by NAS nat­ural sci­en­tists. Dis­be­lief in God and immor­tal­ity among NAS bio­log­i­cal sci­en­tists was 65.2% and 69.0%, respec­tive­ly, and among NAS phys­i­cal sci­en­tists it was 79.0% and 76.3%. Most of the rest were agnos­tics on both issues, with few believ­ers. We found the high­est per­cent­age of belief among NAS math­e­mati­cians (14.3% in God, 15.0% in immor­tal­i­ty). Bio­log­i­cal sci­en­tists had the low­est rate of belief (5.5% in God, 7.1% in immor­tal­i­ty), with physi­cists and astronomers slightly higher (7.5% in God, 7.5% in immor­tal­i­ty). Over­all com­par­i­son fig­ures for the 1914, 1933 and 1998 sur­veys appear in Table 1.

The sig­nifi­cance of vari­a­tions in intel­li­gence has also been exam­ined among indi­vid­u­als with alco­hol depen­dence, as lower intel­li­gence as assessed in child­hood or in early adult­hood pre­dicts greater comor­bid­ity [14], a greater propen­sity for hang­overs [15], greater mor­tal­ity from alco­hol-re­lated health prob­lems [16], and poor treat­ment out­comes [17].

“Ear­ly-Life Intel­li­gence Pre­dicts Midlife Bio­log­i­cal Age”, Schae­fer et al 2015:

Objec­tives. Ear­ly-life intel­li­gence has been shown to pre­dict mul­ti­ple causes of death in pop­u­la­tions around the world. This find­ing sug­gests that intel­li­gence might influ­ence mor­tal­ity through its effects on a gen­eral process of phys­i­o­log­i­cal dete­ri­o­ra­tion (i.e., indi­vid­ual vari­a­tion in “bio­log­i­cal age”). We exam­ined whether intel­li­gence could pre­dict mea­sures of aging at midlife before the onset of most age-re­lated dis­ease.

Meth­ods. We tested whether intel­li­gence assessed in early child­hood, mid­dle child­hood, and midlife pre­dicted midlife bio­log­i­cal age in mem­bers of the Dunedin Study, a pop­u­la­tion- rep­re­sen­ta­tive birth cohort.

Results. Lower intel­li­gence pre­dicted more advanced bio­log­i­cal age at midlife as cap­tured by per­ceived facial age, a 10-bio­marker algo­rithm based on data from the National Health and Nutri­tion Exam­i­na­tion Sur­vey (NHANES), and Fram­ing­ham heart age (r = 0.1-0.2). Cor­re­la­tions between intel­li­gence and telom­ere length were less con­sis­tent. The asso­ci­a­tions between intel­li­gence and bio­log­i­cal age were not explained by differ­ences in child­hood health or parental socioe­co­nomic sta­tus, and intel­li­gence remained a sig­nifi­cant pre­dic­tor of bio­log­i­cal age even when intel­li­gence was assessed before Study mem­bers began their for­mal school­ing.

Dis­cus­sion. These results sug­gest that accel­er­ated aging may serve as one of the fac­tors link­ing low ear­ly-life intel­li­gence to increased rates of mor­bid­ity and mor­tal­i­ty.

“Intel­li­gence Pre­dicts Health and Longevi­ty, but Why?”, Got­tfred­son & Deary 2004:

Large epi­demi­o­log­i­cal stud­ies of almost an entire pop­u­la­tion in Scot­land have found that intel­li­gence (as mea­sured by an IQ-type test) in child­hood pre­dicts sub­stan­tial differ­ences in adult mor­bid­ity and mor­tal­i­ty, includ­ing deaths from can­cers and car­dio­vas­cu­lar dis­eases. These rela­tions remain sig­nifi­cant after con­trol­ling for socioe­co­nomic vari­ables. One pos­si­ble, par­tial expla­na­tion of these results is that intel­li­gence enhances indi­vid­u­als’ care of their own health because it rep­re­sents learn­ing, rea­son­ing, and prob­lem-solv­ing skills use­ful in pre­vent­ing chronic dis­ease and acci­den­tal injury and in adher­ing to com­plex treat­ment reg­i­mens.

…O’­Toole and Stankov (1992) used IQ at induc­tion into the mil­i­tary, along with 56 other psy­cho­log­i­cal, behav­ioral, health, and demo­graphic vari­ables, to pre­dict non­com­bat deaths by age 40 among 2,309 Aus­tralian vet­er­ans. When all other vari­ables were sta­tis­ti­cally con­trolled, each addi­tional IQ point pre­dicted a 1% decrease in risk of death. Also, IQ was the best pre­dic­tor of the major cause of death, motor vehi­cle acci­dents. Vehic­u­lar death rates dou­bled and then tripled at suc­ces­sively lower IQ ranges (100-115, 85-100, 80-85; O’Toole, 1990).

…To date, Scot­land is the only coun­try to have con­ducted IQ test­ing on almost a whole year-of-birth cohort. This took place in the remark­able Scot­tish Men­tal Sur­vey of 1932 (SMS1932). Using these pro­ce­dures, the researchers traced 2,230 (79.9%) of those chil­dren who took the MHT in Aberdeen: 1,084 were dead, 1,101 were alive, and 45 had moved away from Scot­land. In addi­tion, 562 were untraced… IQ at age 11 had a sig­nifi­cant asso­ci­a­tion with sur­vival to about age 76. On aver­age, indi­vid­u­als who were at a 1stan­dard­-de­vi­a­tion (15-point) dis­ad­van­tage in IQ rel­a­tive to other par­tic­i­pants were only 79% as likely to live to age 76. The effect of IQ was stronger for women (71%) than for men (83%), partly because men who died in active ser­vice dur­ing World War II had rel­a­tively high mean IQ scores. Fur­ther analy­ses of the Aberdeen sub­jects found that a drop of 1 stan­dard devi­a­tion in IQ was asso­ci­ated with a 27% increase in can­cer deaths among men and a 40% increase in can­cer deaths among women (Deary, Whal­ley, & Starr, 2003). The effect was espe­cially pro­nounced for stom­ach and lung can­cers, which are specifi­cally asso­ci­ated with low socioe­co­nomic sta­tus (SES) in child­hood.

…Higher intel­li­gence might lower mor­tal­ity from all causes and from spe­cific causes partly by affect­ing known risk fac­tors for dis­ease, such as smok­ing. In the com­bined SMS1932-Midspan data­base, there was no sig­nifi­cant child­hood IQ differ­ence between par­tic­i­pants who had ever smoked and those who had never smoked (Tay­lor, Hart, et al., 2003). How­ev­er, at the time of the Midspan stud­ies, par­tic­i­pants who were cur­rent smok­ers had sig­nifi­cantly lower child­hood IQs than ex-smok­ers. For each stan­dard devi­a­tion increase in IQ, there was a 33% increased rate of quit­ting smok­ing. Adjust­ing for social class reduced this rate only mild­ly, to 25%. Thus, child­hood IQ was not asso­ci­ated with start­ing smok­ing (mostly in the 1930s, when the pub­lic were not aware of health risks), but was asso­ci­ated with giv­ing up smok­ing as health risks became evi­dent.

…How­ev­er, health inequal­i­ties tend to increase when health resources become more avail­able to every­one (Got­tfred­son, in press). That is, increased avail­abil­ity of health resources improves health over­all, but the improve­ments are smaller for peo­ple who are poorly edu­cated and have low incomes than for peo­ple with more edu­ca­tion and bet­ter incomes. Com­pared with peo­ple in high-SES groups, peo­ple with low SES seek more but not nec­es­sar­ily appro­pri­ate care when cost is no bar­ri­er; adhere less often to treat­ment reg­i­mens; learn and under­stand less about how to pro­tect their health; seek less pre­ven­tive care, even when it is free; and less often prac­tice the healthy behav­iors so impor­tant for pre­vent­ing or slow­ing the pro­gres­sion of chronic dis­eases, the major killers and dis­ablers in devel­oped nations today.

Yet social class cor­re­lates with vir­tu­ally every indi­ca­tor of health, health behav­ior, and health knowl­edge. The link between SES and health tran­scends the par­tic­u­lars of mate­r­ial advan­tage, decade, nation, health sys­tem, social change, or dis­ease, regard­less of its treata­bil­i­ty. Health sci­en­tists view the per­va­sive­ness and finely graded nature of this rela­tion­ship between SES and health as a para­dox, lead­ing them to spec­u­late that SES cre­ates health inequal­ity via some yet-to-be-i­den­ti­fied, highly gen­er­al­iz­able “fun­da­men­tal cause” (Got­tfred­son, in press). The socioe­co­nomic mea­sures that best pre­dict health inequal­ity also cor­re­late most with intel­li­gence (ed­u­ca­tion best, then occu­pa­tion, then income). This means that instead of IQ being a proxy for SES in health mat­ters, SES mea­sures might be oper­at­ing pri­mar­ily as rough prox­ies for social-class differ­ences in men­tal rather than mate­r­ial resources.

…Health work­ers can diag­nose and treat incu­bat­ing prob­lems, such as high blood pres­sure or dia­betes, but only when peo­ple seek pre­ven­tive screen­ing and fol­low treat­ment reg­i­mens. Many do not. In fact, per­haps a third of all pre­scrip­tion med­ica­tions are taken in a man­ner that jeop­ar­dizes the patien­t’s health. Non-ad­her­ence to pre­scribed treat­ment reg­i­mens dou­bles the risk of death among heart patients (Gal­lagher, Vis­coli, & Hor­witz, 1993). For bet­ter or worse, peo­ple are sub­stan­tially their own pri­mary health care providers.

For instance, one study (Williams et al., 1995) found that, over­all, 26% of the out­pa­tients at two urban hos­pi­tals were unable to deter­mine from an appoint­ment slip when their next appoint­ment was sched­uled, and 42% did not under­stand direc­tions for tak­ing med­i­cine on an empty stom­ach. The per­cent­ages specifi­cally among out­pa­tients with “inad­e­quate” lit­er­acy were worse: 40% and 65%, respec­tive­ly. In com­par­ison, the per­cent­ages were 5% and 24% among out­pa­tients with “ade­quate” lit­er­a­cy. In another study (Williams, Bak­er, Park­er, & Nurss, 1998), many insulin-de­pen­dent dia­bet­ics did not under­stand fun­da­men­tal facts for main­tain­ing daily con­trol of their dis­ease: Among those clas­si­fied as hav­ing inad­e­quate lit­er­a­cy, about half did not know the signs of very low or very high blood sug­ar, and 60% did not know the cor­rec­tive actions they needed to take if their blood sugar was too low or too high. Among dia­bet­ics, intel­li­gence at time of diag­no­sis cor­re­lates sig­nifi­cantly (.36) with dia­betes knowl­edge mea­sured 1 year later (Tay­lor, Frier, et al., 2003).

Duar­te, Craw­ford, Stern, Haidt, Jus­sim, and Tet­lock:

[T]he observed rela­tion­ship between intel­li­gence and con­ser­vatism largely depends on how con­ser­vatism is oper­a­tional­ized. Social con­ser­vatism cor­re­lates with lower cog­ni­tive abil­ity test scores, but eco­nomic con­ser­vatism cor­re­lates with higher scores (Iy­er, Kol­e­va, Gra­ham, Dit­to, & Haidt, 2012; Kem­melmeier 2008). Sim­i­lar­ly, Feld­man and John­ston (2014) find in mul­ti­ple nation­ally rep­re­sen­ta­tive sam­ples that social con­ser­vatism neg­a­tively pre­dicted edu­ca­tional attain­ment, whereas eco­nomic con­ser­vatism pos­i­tively pre­dicted edu­ca­tional attain­ment. Togeth­er, these results likely explain why both Heaven et al. (2011) and Hod­son and Busseri (2012) found a neg­a­tive cor­re­la­tion between IQ and con­ser­vatism–be­cause “con­ser­vatism” was oper­a­tional­ized as Right-Wing Author­i­tar­i­an­ism, which is more strongly related to social than eco­nomic con­ser­vatism (van Hiel et al., 2004). In fact, Carl (2014) found that Repub­li­cans have higher mean ver­bal intel­li­gence (up to 5.48 IQ points equiv­a­lent, when covari­ates are exclud­ed), and this effect is dri­ven by eco­nomic con­ser­vatism (which, as a Euro­pean, he called eco­nomic lib­er­al­ism, because of its empha­sis on free mar­ket­s). Carl sug­gests that lib­er­tar­ian Repub­li­cans over­power the neg­a­tive cor­re­la­tion between social con­ser­vatism and ver­bal intel­li­gence, to yield the aggre­gate mean advan­tage for Repub­li­cans. More­over, the largest polit­i­cal effect in Kem­melmeier’s (2008) study was the pos­i­tive cor­re­la­tion between anti-reg­u­la­tion views and SAT-V scores, where β = .117, p < .001 (by com­par­ison, the regres­sion coeffi­cient for con­ser­vatism was β = −.088, p < .01, and for being African Amer­i­can, β = −.169, p < .001).


Many key deter­mi­nants of well-be­ing cor­re­late highly with the results of IQ tests, and other mea­sures of intel­li­gence. Many spe­cific life out­comes have been shown to cor­re­late highly with intel­li­gence (Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993). While no causal link has been demon­strated between higher lev­els of cog­ni­tion and hap­pi­ness (Gow et al. 2005), numer­ous stud­ies have high­lighted that increased cog­ni­tion improves the like­li­hood of spe­cific mark­ers of well­be­ing, while lower lev­els of gen­eral intel­li­gence pre­dis­poses an indi­vid­ual to var­i­ous forms of social dis­ad­van­tage (Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993).

Sev­eral promi­nent intel­li­gence stud­ies have demon­strated that higher gen­eral intel­li­gence cor­re­lates with such life out­comes as increased income (Rowe et al. 1998; Zagorsky 2007; Got­tfred­son 2003), improved qual­ity of health and reduced mor­tal­ity (Batty et al. 2007; Whal­ley and Deary 2001; Got­tfred­son and Deary 2004) and over­all increased life chances (Mur­ray 2002; Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993; Got­tfred­son 2011). Intel­li­gence appears to have a promi­nent effect over a broad range of social and eco­nomic life out­comes. Life is diffi­cult for indi­vid­u­als with bor­der­line intel­lec­tual dis­abil­i­ty; an IQ below 75. This group is at a high risk of fail­ing ele­men­tary edu­ca­tion, being unable to mas­ter sim­ple daily tasks, being clas­si­fied as unem­ploy­able, and are at an increased risk of being socially iso­lated (Edger­ton 1993; Koegel and Edger­ton 1984; Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993). Indi­vid­u­als with bor­der­line intel­lec­tual dis­abil­ity are at a great risk of liv­ing in poverty (30 %), hav­ing ille­git­i­mate chil­dren (32 %), being a chronic wel­fare depen­dent (31 %) and have very poor employ­ment oppor­tu­ni­ties (Got­tfred­son 1997; Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993). While indi­vid­u­als within this group are capa­ble of lead­ing sat­is­fy­ing lives, they will most likely require sig­nifi­cant social sup­port in order to do so. Those in this group who do live inde­pen­dently have a ten­dency to live volatile and unpre­dictable lives due to the lack of sta­bil­is­ing resources that come with increased intel­lec­tual com­pe­tence (Edger­ton 1993; Got­tfred­son 1997).

…Over half of these indi­vid­u­als fail to reach the min­i­mum recruit­ment stan­dards for the US mil­i­tary (Hunter and Schmidt 1996). Indi­vid­u­als with minor to mod­er­ate low­er, nor­mal intel­li­gence are still at sig­nifi­cant risk of liv­ing in poverty (16 %), being a chronic wel­fare depen­dent (17 %) and are much more likely to drop out of school (35 %) com­pared to indi­vid­u­als with aver­age intel­li­gence (Her­rn­stein and Mur­ray 1994; Kirsch et al. 1993; Got­tfred­son 2011). The odds of incar­cer­a­tion remain steady for all low­er, nor­mal intel­li­gence groups (7 %) but reduce by more than half for aver­age intel­li­gence lev­els (3 %), indi­cat­ing a par­tic­u­lar sus­cep­ti­bil­ity to incar­cer­a­tion at lower intel­li­gence lev­els (Got­tfred­son 1997; Her­rn­stein and Mur­ray 1994).

  • Bat­ty, G.D., I.J. Deary, and L.S. Got­tfred­son. 2007. Pre-mor­bid (early life) IQ and later mor­tal­ity risk: Sys­tem­atic review. Annals of Epi­demi­ol­ogy 17(4): 278-288.
  • Edger­ton, R.B. 1993. The cloak of com­pe­tence, revised and updated edi­tion. Berke­ley: Uni­ver­sity of Cal­i­for­nia Press
  • Got­tfred­son, L.S. 1997. Why g mat­ters: The com­plex­ity of every­day life. Intel­li­gence 24: 79-132.
  • Got­tfred­son, L.S. 2003. g, jobs, and life. In The sci­en­tific study of gen­eral intel­li­gence: Trib­ute to Arthur R. Jensen, ed. H. Nyborg. New York: Perg­a­mon.
  • Got­tfred­son, L.S. 2011. Intel­li­gence and social inequal­i­ty: Why the bio­log­i­cal link? In Hand­book of indi­vid­ual differ­ences, ed. T. Chamor­ro-Pre­muz­ic, A. Furhnam, and S. von strumm. New York: Wiley.
  • Got­tfred­son, L.S., and I.J. Deary. 2004. Intel­li­gence pre­dicts health and longevi­ty, but why? Cur­rent Direc­tions in Psy­cho­log­i­cal Sci­ence 13: 1-4.
  • Gow, A.J., M.C. White­man, A. Pat­tie, L. Whal­ley, J. Starr, and I.J. Deary. 2005. Life­time intel­lec­tual func­tion and sat­is­fac­tion with life in old age: Lon­gi­tu­di­nal cohort study. BMJ 331: 141-142.
  • Her­rn­stein, R.J., and C.A. Mur­ray. 1994. The bell curve: Intel­li­gence and class struc­ture in Amer­i­can life. Salt Lake: Free Press.
  • Hunter, J.E., and F.L. Schmidt. 1996. Intel­li­gence and job per­for­mance: Eco­nomic and social impli­ca­tions. Psy­chol­o­gy, Pub­lic Pol­i­cy, and Law 2: 447-472.
  • Kirsch, I.S., A. Junge­blut, L. Jenk­ins, and A. Kol­stad. 1993. Adult lit­er­acy in Amer­i­ca: A first look at the results of the national adult lit­er­acy sur­vey. Prince­ton: Edu­ca­tional Test­ing Ser­vice.
  • Koegel, P., and R.B. Edger­ton. 1984. “Black ‘six-hour retarded chil­dren’ as young adults”. Mono­graphs of the Amer­i­can Asso­ci­a­tion on Men­tal Defi­ciency 6: 145-171.
  • Mur­ray, C. 2002. IQ and income inequal­ity in a sam­ple of sib­ling pairs from advan­taged fam­ily back­grounds. The Amer­i­can Eco­nomic Review 92: 339-343.
  • Rowe, D.C., W.J. Ves­terdal, and J.L. Rodgers. 1998. Her­rn­stein’s syl­lo­gism: genetic and shared envi­ron­men­tal influ­ences on IQ, edu­ca­tion, and income. Intel­li­gence 26: 405-423
  • Whal­ley, L.J., and I.J. Deary. 2001. Lon­gi­tu­di­nal cohort study of child­hood IQ and sur­vival up to age 76. BMJ 322: 819.
  • Zagorsky, J.L. 2007. Do you have to be smart to be rich? The impact of IQ on wealth, income and finan­cial dis­tress. Intel­li­gence 35: 489-501.

“Clever Enough to Tell the Truth”, Ruffle & Tobol 2017:

We con­duct a field exper­i­ment on 427 Israeli sol­diers who each rolled a six-sided die in pri­vate and reported the out­come. For every point report­ed, the sol­dier received an addi­tional half-hour early release from the army base on Thurs­day after­noon. We find that the higher a sol­dier’s mil­i­tary entrance score, the more hon­est he is on aver­age. We repli­cate this find­ing on a sam­ple of 156 civil­ians paid in cash for their die reports. Fur­ther­more, the civil­ian exper­i­ments reveal that two mea­sures of cog­ni­tive abil­ity pre­dict hon­esty, whereas self­-re­port hon­esty ques­tions and a con­sis­tency check among them are of no val­ue. We pro­vide a ratio­nale for the rela­tion­ship between cog­ni­tive abil­ity and hon­esty and dis­cuss the gen­er­al­iz­abil­ity of this result.

“Intel­li­gence and socioe­co­nomic suc­cess: A meta-an­a­lytic review of lon­gi­tu­di­nal research”, Strenze 2007

The rela­tion­ship between intel­li­gence and socioe­co­nomic suc­cess has been the source of numer­ous con­tro­ver­sies. The present paper con­ducted a meta-analy­sis of the lon­gi­tu­di­nal stud­ies that have inves­ti­gated intel­li­gence as a pre­dic­tor of suc­cess (as mea­sured by edu­ca­tion, occu­pa­tion, and income). In order to bet­ter eval­u­ate the pre­dic­tive power of intel­li­gence, the paper also includes meta­analy­ses of parental socioe­co­nomic sta­tus (SES) and aca­d­e­mic per­for­mance (school grades) as pre­dic­tors of suc­cess. The results demon­strate that intel­li­gence is a pow­er­ful pre­dic­tor of suc­cess but, on the whole, not an over­whelm­ingly bet­ter pre­dic­tor than parental SES or grades. Mod­er­a­tor analy­ses showed that the rela­tion­ship between intel­li­gence and suc­cess is depen­dent on the age of the sam­ple but there is lit­tle evi­dence of any his­tor­i­cal trend in the rela­tion­ship.

“Does intel­li­gence fos­ter gen­er­al­ized trust? An empir­i­cal test using the UK birth cohort stud­ies”, Stur­gis et al 2010

Social, or ‘gen­er­al­ized’ trust is often char­ac­terised as the ‘atti­tu­di­nal dimen­sion’ of social cap­i­tal. It has been posited as key to a host of nor­ma­tively desir­able out­comes at the soci­etal and indi­vid­ual lev­els. Yet the ori­gins of indi­vid­ual vari­a­tion in trust remain some­thing of a mys­tery and con­tinue to be a source of dis­sensus amongst researchers across and within aca­d­e­mic dis­ci­plines. In this paper we use data from two British birth cohort stud­ies to test the hypoth­e­sis that a propen­sity to express gen­er­al­ized trust varies sys­tem­at­i­cally as a func­tion of indi­vid­ual intel­li­gence. Intel­li­gence, we argue, fos­ters greater trust in one’s fel­low cit­i­zens because more intel­li­gent indi­vid­u­als are more accu­rate in their assess­ments of the trust­wor­thi­ness of oth­ers. This means that, over the life-course, their trust is less often betrayed and they are able to accrue the ben­e­fits of norms of rec­i­proc­i­ty. Our results show that stan­dard mea­sures of intel­li­gence admin­is­tered when cohort mem­bers were aged 10 and 11 can explain vari­abil­ity in expressed trust in early mid­dle age, net of a broad range of the­o­ret­i­cally related covari­ates.

“Asso­ci­a­tions between IQ and cig­a­rette smok­ing among Swedish male twins”, Wen­ner­stad et al 2010

It has been sug­gested that cer­tain health behav­iours, such as smok­ing, may oper­ate as medi­a­tors of the well-estab­lished inverse asso­ci­a­tion between IQ and mor­tal­ity risk. Pre­vi­ous research may be afflicted by unad­justed con­found­ing by socioe­co­nomic or psy­choso­cial fac­tors. Twin designs offer a unique pos­si­bil­ity to take genetic and shared envi­ron­men­tal fac­tors into account. The aim of the present national twin study was to deter­mine the inter­re­la­tions between IQ at age 18, child­hood and attained social fac­tors and smok­ing sta­tus in young adult­hood and mid-life. We stud­ied the asso­ci­a­tion between IQ at age 18 and smok­ing in later life in a pop­u­la­tion of 11 589 male Swedish twins. IQ was mea­sured at mil­i­tary con­scrip­tion, and data on smok­ing and zygos­ity was obtained from the Swedish Twin Reg­is­ter. Infor­ma­tion on social fac­tors was extracted from cen­sus­es. Data on smok­ing was self­-re­ported by the twins at the age of 22-47 years. Logis­tic regres­sion mod­els esti­mated with gen­er­alised esti­mat­ing equa­tions were used to explore pos­si­ble asso­ci­a­tions between IQ and smok­ing among the twins as indi­vid­u­als as well as between-and within twin-pairs.

A strong inverse asso­ci­a­tion between IQ and smok­ing sta­tus emerged in unmatched analy­ses over the entire range of IQ dis­tri­b­u­tion. In with­in-pair and between-pair analy­ses it tran­spired that shared envi­ron­men­tal fac­tors explained most of the inverse IQ-smok­ing rela­tion­ship. In addi­tion, these analy­ses indi­cated that non-shared and genetic fac­tors con­tributed only slightly (and non-sig­nifi­cant­ly) to the IQ-smok­ing asso­ci­a­tion. Analy­sis of twin pairs dis­cor­dant for IQ and smok­ing sta­tus dis­played no evi­dence that non-shared fac­tors con­tribute sub­stan­tially to the asso­ci­a­tion. The ques­tion of which shared envi­ron­men­tal fac­tors might explain the IQ-smok­ing asso­ci­a­tion is an intrigu­ing one for future research.

“The asso­ci­a­tion between coun­ty-level IQ and coun­ty-level crime rates”, Beaver & Wright 2011:

An impres­sive body of research has revealed that indi­vid­u­al-level IQ scores are neg­a­tively asso­ci­ated with crim­i­nal and delin­quent involve­ment. Recent­ly, this line of research has been extended to show that state-level IQ scores are asso­ci­ated with state-level crime rates. The cur­rent study uses this lit­er­a­ture as a spring­board to exam­ine the poten­tial asso­ci­a­tion between coun­ty-level IQ and coun­ty-level crime rates. Analy­sis of data drawn from the National Lon­gi­tu­di­nal Study of Ado­les­cent Health revealed sta­tis­ti­cally sig­nifi­cant and neg­a­tive asso­ci­a­tions between coun­ty-level IQ and the prop­erty crime rate, the bur­glary rate, the lar­ceny rate, the motor vehi­cle theft rate, the vio­lent crime rate, the rob­bery rate, and the aggra­vated assault rate. Addi­tional analy­ses revealed that these asso­ci­a­tions were not con­founded by a mea­sure of con­cen­trated dis­ad­van­tage that cap­tures the effects of race, pover­ty, and other social dis­ad­van­tages of the coun­ty. We dis­cuss the impli­ca­tions of the results and note the lim­i­ta­tions of the study.

tat­toos and pierc­ings:

pg406, Stren­ze, “Intel­li­gence and Suc­cess”, ch25 of Hand­book of Intel­li­gence Evo­lu­tion­ary The­o­ry, His­tor­i­cal Per­spec­tive, and Cur­rent Con­cepts, ed Gold­stein et al 2015:

Table 25.1 Rela­tion­ship between intel­li­gence and mea­sures of suc­cess (Re­sults from meta-analy­ses)
Mea­sure of suc­cess r k N Source
Aca­d­e­mic per­for­mance in pri­mary edu­ca­tion 0.58 4 1791 Poropat (2009)
Edu­ca­tional attain­ment 0.56 59 84828 Strenze (2007)
Job per­for­mance (su­per­vi­sory rat­ing) 0.53 425 32124
Occu­pa­tional attain­ment 0.43 45 72290 Strenze (2007)
Job per­for­mance (work sam­ple) 0.38 36 16480 Roth et al. (2005)
Skill acqui­si­tion in work train­ing 0.38 17 6713
Degree attain­ment speed in grad­u­ate school 0.35 5 1700 Kun­cel et al. (2004)
Group lead­er­ship suc­cess (group pro­duc­tiv­i­ty) 0.33 14 Judge et al. (2004)
Pro­mo­tions at work 0.28 9 21290 Schmitt et al. (1984)
Inter­view suc­cess (in­ter­viewer rat­ing of appli­cant) 0.27 40 11317
Read­ing per­for­mance among prob­lem chil­dren 0.26 8 944 Nel­son et al. (2003)
Becom­ing a leader in group 0.25 65 Judge et al. (2004)
Aca­d­e­mic per­for­mance in sec­ondary edu­ca­tion 0.24 17 12606 Poropat (2009)
Aca­d­e­mic per­for­mance in ter­tiary edu­ca­tion 0.23 26 17588 Poropat (2009)
Income 0.20 31 58758 Strenze (2007)
Hav­ing anorexia ner­vosa 0.20 16 484 Lopez et al. (2010)
Research pro­duc­tiv­ity in grad­u­ate school 0.19 4 314 Kun­cel et al. (2004)
Par­tic­i­pa­tion in group activ­i­ties 0.18 36 Mann (1959)
Group lead­er­ship suc­cess (group mem­ber rat­ing) 0.17 64 Judge et al. (2004)
Cre­ativ­ity 0.17 447 Kim (2005)
Pop­u­lar­ity among group mem­bers 0.10 38 Mann (1959)
Hap­pi­ness 0.05 19 2546 DeN­eve & Cooper (1998)
Pro­cras­ti­na­tion (need­less delay of action) 0.03 14 2151 Steel (2007)
Chang­ing jobs 0.01 7 6062 Griffeth et al. (2000)
Phys­i­cal attrac­tive­ness -0.04 31 3497 Fein­gold (1992)
Recidi­vism (re­peated crim­i­nal behav­ior) -0.07 32 21369 Gen­dreau et al. (1996)
Num­ber of chil­dren -0.11 3 Lynn (1996)
Traffic acci­dent involve­ment -0.12 10 1020 Arthur et al. (1991)
Con­for­mity to per­sua­sion -0.12 7 Rhodes and Wood (1992)
Com­mu­ni­ca­tion anx­i­ety -0.13 8 2548 Bourhis and Allen (1992)
Hav­ing schiz­o­phre­nia -0.26 18

r cor­re­la­tion between intel­li­gence and the mea­sure of suc­cess, k num­ber of stud­ies included in the meta-analy­sis, N num­ber of indi­vid­u­als included in the meta-analy­sis

“Intel­li­gence in young adult­hood and cause-spe­cific mor­tal­ity in the Dan­ish Con­scrip­tion Data­base – A cohort study of 728,160 men”, Chris­tensen et al 2016:

An inverse asso­ci­a­tion has been reported between early life intel­li­gence and all-cause mor­tal­i­ty. The aim of this study was to inves­ti­gate whether this well-estab­lished asso­ci­a­tion differed accord­ing to the under­ly­ing cause of death and across differ­ent birth cohorts. The asso­ci­a­tions between young adult intel­li­gence and mor­tal­ity from nat­ural and exter­nal causes were inves­ti­gated in the Dan­ish Con­scrip­tion Data­base (DCD), which is a cohort of more than 700,000 men born 1939–1959 and fol­lowed in Dan­ish reg­is­ters from young adult­hood until late mid-life. Young adult intel­li­gence was inversely related to all-cause mor­tal­ity with a 28% higher risk of dying dur­ing the study period per 1 stan­dard devi­a­tion (SD) decrease in intel­li­gence test score (HR = 1.28 95% CI = 1.27–1.29). The strength of the observed inverse asso­ci­a­tions did not vary much across main groups of nat­ural and exter­nal causes with the excep­tion of the asso­ci­a­tions for mor­tal­ity from res­pi­ra­tory dis­eases (HR = 1.61 95% CI = 1.55–1.67) and homi­cide (HR = 1.65 95% CI = 1.46–1.87) which were more pro­nounced com­pared to the rest. More­over, for skin can­cer mor­tal­i­ty, each SD increase in intel­li­gence test score was asso­ci­ated with a small increase in mor­tal­ity risk (HR = 1.03 95% CI = 1.01–1.15). Fur­ther­more, the asso­ci­a­tion between intel­li­gence and mor­tal­ity was stronger for those born 1950–1959 com­pared to those born 1939–1949 for almost all nat­ural and exter­nal causes of death.

“Asso­ci­a­tion of Fluid Intel­li­gence and Psy­chi­atric Dis­or­ders in a Pop­u­la­tion-Rep­re­sen­ta­tive Sam­ple of US Ado­les­cents”, Keyes et al 2017:

Objec­tive: To inves­ti­gate the asso­ci­a­tion of fluid intel­li­gence with past-year and life­time psy­chi­atric dis­or­ders, dis­or­der age at onset, and dis­or­der sever­ity in a nation­ally rep­re­sen­ta­tive sam­ple of US ado­les­cents.

Design, Set­ting, and Par­tic­i­pants: National sam­ple of ado­les­cents ascer­tained from schools and house­holds from the National Comor­bid­ity Sur­vey Repli­ca­tion-Ado­les­cent Sup­ple­ment, col­lected 2001 through 2004. Face-to-face house­hold inter­views with ado­les­cents and ques­tion­naires from par­ents were obtained. The data were ana­lyzed from Feb­ru­ary to Decem­ber 2016. DSM-IV men­tal dis­or­ders were assessed with the World Health Orga­ni­za­tion Com­pos­ite Inter­na­tional Diag­nos­tic Inter­view, and included a broad range of fear, dis­tress, behav­ior, sub­stance use, and other dis­or­ders. Dis­or­der sever­ity was mea­sured with the Shee­han Dis­abil­ity Scale.

Main Out­comes and Mea­sures: Fluid IQ mea­sured with the Kauf­man Brief Intel­li­gence Test, normed within the sam­ple by 6-month age groups.

Results: The sam­ple included 10 073 ado­les­cents (mean [SD] age, 15.2 [1.50] years; 49.0% female) with valid data on fluid intel­li­gence. Lower mean (SE) IQ was observed among ado­les­cents with past-year bipo­lar dis­or­der (94.2 [1.69]; P = .004), attention-deficit/hyperactivity dis­or­der (96.3 [0.91]; P = .002), oppo­si­tional defi­ant dis­or­der (97.3 [0.66]; P = .007), con­duct dis­or­der (97.1 [0.82]; P = .02), sub­stance use dis­or­ders (al­co­hol abuse, 96.5 [0.67]; P < .001; drug abuse, 97.6 [0.64]; P = .02), and spe­cific pho­bia (97.1 [0.39]; P = .001) after adjust­ment for a wide range of poten­tial con­founders. Intel­li­gence was not asso­ci­ated with post-trau­matic stress dis­or­der, eat­ing dis­or­ders, and anx­i­ety dis­or­ders other than spe­cific pho­bia, and was pos­i­tively asso­ci­ated with past-year major depres­sion (mean [SE], 100 [0.5]; P = .01). Asso­ci­a­tions of fluid intel­li­gence with life­time dis­or­ders that had remit­ted were atten­u­ated com­pared with past-year dis­or­ders, with the excep­tion of sep­a­ra­tion anx­i­ety dis­or­der. Mul­ti­ple past-year dis­or­ders had a larger pro­por­tion of ado­les­cents less than 1 SD below the mean IQ range than those with­out a dis­or­der. Across dis­or­ders, higher dis­or­der sever­ity was asso­ci­ated with lower fluid intel­li­gence. For exam­ple, among ado­les­cents with spe­cific pho­bia, those with severe dis­or­der had a mean (SE) of 4.4 (0.72) points lower IQ than those with­out severe dis­or­der (P < .001), and those with alco­hol abuse had a mean (SE) of 5.6 (1.2) points lower IQ than those with­out severe dis­or­der (P < .001).

Con­clu­sions and Rel­e­vance: Numer­ous psy­chi­atric dis­or­ders were asso­ci­ated with reduc­tions in fluid intel­li­gence; asso­ci­a­tions were gen­er­ally small in mag­ni­tude. Stronger asso­ci­a­tions of cur­rent than past dis­or­ders with intel­li­gence sug­gest that active symp­toms of psy­chi­atric dis­or­ders inter­fere with cog­ni­tive func­tion­ing. Early iden­ti­fi­ca­tion and treat­ment of chil­dren with men­tal dis­or­ders in school set­tings is crit­i­cal to pro­mote aca­d­e­mic achieve­ment and long-term suc­cess.


  • Lead:
  • :

These exam­ples show that, con­trary to Shal­iz­i’s claims, all cog­ni­tive abil­i­ties are inter-cor­re­lat­ed. We can be con­fi­dent about this because the best evi­dence for it comes not from the pro­po­nents of g but from numer­ous com­pe­tent researchers who were hel­l-bent on dis­prov­ing the gen­er­al­ity of the pos­i­tive man­i­fold, only to be refuted by their own work .

Depend­ing type of job and how per­for­mance is mea­sured GMA explains between 30% and 70% of the vari­a­tion in peo­ple’s work per­for­mance (i.e. cor­re­la­tions of between .56 and .84), which is larger than any other known pre­dic­tor.4 [“When per­for­mance is mea­sured objec­tively using care­fully con­structed work sam­ple tests (sam­ples of actual job tasks), the cor­re­la­tion (va­lid­i­ty) with intel­li­gence mea­sures is about .84 - 84% as large as the max­i­mum pos­si­ble value of 1.00, which rep­re­sents per­fect pre­dic­tion. When per­for­mance is mea­sured using rat­ings of job per­for­mance by super­vi­sors, the cor­re­la­tion with intel­li­gence mea­sures is .66 for medium com­plex­ity jobs (over 60% of all job­s). For more com­plex jobs, this value is larger (e.g. .74 for pro­fes­sional and man­age­r­ial job­s), and for sim­pler jobs this value is not as high (e.g. .56 for semi­-skilled job­s). Another per­for­mance mea­sure that is impor­tant is the amount learned in job train­ing pro­grams (Hunter et al., 2006). Regard­less of job lev­el, intel­li­gence mea­sures pre­dict amount learned in train­ing with valid­ity of about .74 (Schmidt, Shaffer, and Oh, 2008).” From: Schmidt, Frank L, and John E Hunter. “Select on intel­li­gence.” Hand­book of prin­ci­ples of orga­ni­za­tional behav­ior(2000): 3-14.]

Evi­dence from sev­eral meta-s­tud­ies shows that when per­for­mance is mea­sured using work-sam­ple tests, the cor­re­la­tion between GMA and per­for­mance is 0.84. When super­vi­sor rat­ings are used, the cor­re­la­tion is low­er, at 0.74 for high­-com­plex­ity job­s.5[Schmidt, Frank L, and John E Hunter. “Select on intel­li­gence.” Hand­book of prin­ci­ples of orga­ni­za­tional behav­ior(2000): 3-14.]

GMA also pre­dicts how high up you get in the job hier­ar­chy - i.e. your occu­pa­tional lev­el.6 US Employ­ment Ser­vice data shows a strong cor­re­la­tion (0.72) between GMA and occu­pa­tional level and US mil­i­tary data shows that mean GMA scores are higher at higher occu­pa­tional lev­els. Also, there is a wider vari­ety of GMA scores at lower occu­pa­tional lev­els than at higher ones. It seems that there are high­-s­cor­ing peo­ple in low-level occu­pa­tions, but low-s­cor­ing peo­ple are unlikely to get pro­moted to higher lev­el­s.7[Schmidt, Frank L, and John Hunter. “Gen­eral men­tal abil­ity in the world of work: occu­pa­tional attain­ment and job per­for­mance.” Jour­nal of per­son­al­ity and social psy­chol­ogy 86.1 (2004): 162.]

But to fully show the link we need to track peo­ple with known GMA over time to see if high GMA indi­vid­u­als end up being more suc­cess­ful. This has been done.8[Schmidt, Frank L, and John Hunter. “Gen­eral men­tal abil­ity in the world of work: occu­pa­tional attain­ment and job per­for­mance.” Jour­nal of per­son­al­ity and social psy­chol­ogy 86.1 (2004): 162.] In a lon­gi­tu­di­nal study of 3,887 young adults, GMA pre­dicted move­ment in the job hier­ar­chy 5 years lat­er. Another study found that if peo­ple were in a job that was less com­plex than their GMA would pre­dict, they moved up to a more com­plex job and vice ver­sa. The pre­dic­tiv­ity of GMA even holds when con­trol­ling for socioe­co­nomic sta­tus by com­par­ing bio­log­i­cal sib­lings. “When the sib­lings were in their late 20s (in 1993), a per­son with aver­age GMA was earn­ing on aver­age almost $18,000 less per year than his brighter sib­ling who had an IQ of 120 or higher and was earn­ing more than $9,000 more than his duller sib­ling who had an IQ of less than 80.”9[Schmidt, Frank L, and John Hunter. “Gen­eral men­tal abil­ity in the world of work: occu­pa­tional attain­ment and job per­for­mance.” Jour­nal of per­son­al­ity and social psy­chol­ogy 86.1 (2004): 162.]

The link has also been con­firmed by two nat­ural exper­i­ments.[Schmidt, Frank L, and John E Hunter. “Select on intel­li­gence.” Hand­book of prin­ci­ples of orga­ni­za­tional behav­ior(2000): 3-14.]

With high GMA, peo­ple are more able to go beyond exist­ing job knowl­edge and make judge­ments in unfa­mil­iar sit­u­a­tion­s.12[Schmidt, Frank L, and John E Hunter. “Select on intel­li­gence.” Hand­book of prin­ci­ples of orga­ni­za­tional behav­ior(2000): 3-14.]

Although GMA pre­dicts per­for­mance in all jobs the more com­plex the job is13, the stronger the rela­tion­ship between GMA and per­for­mance.14[Hunter, John E. “Cog­ni­tive abil­i­ty, cog­ni­tive apti­tudes, job knowl­edge, and job per­for­mance.” Jour­nal of voca­tional behav­ior 29.3 (1986): 340-362.] And the more com­plex the job, the more vari­a­tion there is between top per­form­ers and bot­tom per­form­er­s.15[Hunter, John E, Frank L Schmidt, and Michael K Jud­i­esch. “Indi­vid­ual differ­ences in out­put vari­abil­ity as a func­tion of job com­plex­i­ty.” Jour­nal of Applied Psy­chol­ogy 75.1 (1990): 28.] So if you have one of the high­est lev­els of GMA in a highly com­plex job, you’ll have a high out­put com­pared to the aver­age per­former.

“Child­hood intel­li­gence in rela­tion to major causes of death in 68 year fol­low-up: prospec­tive pop­u­la­tion study”, Deary et al 2017:

“Intel­li­gence and per­sist­ing with med­ica­tion for two years: Analy­sis in a ran­domised con­trolled trial”, Deary et al 2009: