Develop men tui Psychology 1977, Vol. 13, No. 3, 184-191 Cumulative Deficit in IQ of Blacks in the Rural South

ARTHUR R.JENSEN University of , Berkeley

The cumulative deficit hypothesis with respect to age decrement in IQ between the ages of 5 and 18 was investigated in large samples of white and black school children in rural Georgia. Age decrement in verbal and nonverbal IQ was measured by the average IQ difference between younger and older siblings. It was found that blacks (but not whites) showed significant and substantial decrements in both verbal and nonverbal IQs as a linear function of age in the rank from about 5 to 16 years of age. An environmental interpretation of the age decrement in IQ seems reasonable in view of the comparative lack of such a decrement in a parallel study of California blacks whose environmental circumstances are markedly better than those of the black sample from rural Georgia.

The cumulative deficit hypothesis is in- in a California school district and found a tended to explain the increasing decrement slight but significant age decrement in verbal in mental test scores, relative to population IQ in the black sample, but no evidence norms, as a function of age in groups consid- whatever of a decrement in nonverbal IQ, ered environmentally deprived. According although the black sample scored equally far to the hypothesis, the decrement is a result below (about one standard deviation) the of the cumulative effects of environmental white sample in nonverbal as in verbal IQ. disadvantages on mental development. Jensen suggested, however, that the sib- The of the cumulative deficit hy- ling method might reveal an age decrement pothesis and its theoretical and methodolog- in the IQ of blacks in other regions of the ical problems have been reviewed by Jensen country where blacks have experienced (1974a). It was concluded that most of the greater environmental disadvantages. Age studies of the phenomenon are seriously decrement in verbal and scholastic abilities flawed by methodological deficiencies. The in Southern blacks was suggested in the Col- majority of studies have found no evidence eman report, (Coleman, Cambell, Hobson, of an age-related IQ decrement in blacks. McPartland, Mood, Weinfeld, & York, Jensen (1974a) proposed investigating IQ 1966, p. 274) but is not proven by the cross- decrement by the sibling method; that is, sectional IQ x Age data which could reflect using the difference in selective migration of abler pupils out of the scores between younger and older siblings rural South, causing an increasing accumu- within the same family as an indicator of IQ lation of poorer students in the higher grades decrement. If there is a true IQ decrement, in school. older siblings should obtain lower test Although the cumulative deficit hypothe- scores than their younger siblings, and there sis applies to scholastic achievement as well should be a positive correlation between as to IQ, it is clear from the literature on sibling age difference and IQ difference. this topic that the core hypothesis concerns Jensen applied the sibling method to large measured (Jensen, 1974a). The samples of whites and blacks of ages 5 to 12 author has argued elsewhere that standard- ized IQ tests measure essentially the same Requests for reprints should be sent to Arthur R. general factor of mental ability equally well Jensen, Institute of Human , University of in both whites and blacks. Although IQ tests California, Berkeley, California 94720. are culturally loaded in varying degrees, 184 CUMULATIVE DEFICIT IN IQ 185 there is virtually no evidence, in terms of a population is mostly rural-agricultural, with a very low number of statistical and psychometric cri- median family income compared to the national aver- age. The black group as a whole would be classified as teria (e.g., predictive validity, reliability, very low socioeconomic status (SES) on any index of item analysis, Race x Items interaction, fac- SES. The white population is predominantly low and tor structure, etc.) that the tests are lower-middle SES. Some 1,300 school children, ap- biased with respect to the present white and proximately 49% whites and 51% blacks, were tested. black populations in the (Jen- sen, 1974b, 1976). The black IQ deficit, whatever its causes, appears to be a quite Tests general cognitive deficit rather than nar- Subjects were tested on the California Test of Mental rowly culture specific. Maturity (1963 Revision), a standardized test of general intelligence, which yields deviation IQs for verbal and If a cumulative deficit in mental devel- nonverbal abilities at every grade level from kindergar- opment as indexed by IQ actually exists at ten through Grade 12 (see Euros, 1972, pp. 631-636). all in any segment of the United States popu- The California Test of Mental Maturity is factorially lation, it should probably be expected most very comparable to other standardized group tests of verbal and nonverbal IQ such as the Lorge-Thorndike in blacks of the rural South. Their environ- Intelligence Tests, which Jensen (1974a) used in the mental circumstances would seem much California study. (The California Test of Mental Matur- more likely to contribute to the cumulative ity was used instead of the Lorge-Thorndike in the deficit effect than would the relatively good Georgia study, since the testing was done as part of the environmental conditions of the California school's state-mandated testing program, which re- quired the California Test of Mental Maturity.) The school sample involved in Jensen's first sib- California Test of Mental Maturity IQs are standard- ling study. The aim of the present study, ized scores (M — 100, SD = 15) based on large samples therefore, is to apply the sibling method to of school children from 49 states. the investigation oflQ decrement in samples of whites and blacks in the rural South. The sampled populations, particularly the black Results and Discussion group, are not intended to be representative of the total white and black populations in Sample Statistics on Age and IQ the United States. Blacks in the locality The total white sample (N = 653) has a under study are probably as severely disad- mean age of 12 years 4 months, SD = 3 years vantaged, educationally and economically, 7 months. The mean age of the total black as can be found anywhere in the United sample (N = 826) is 11 years8 months,SD - States today. If an age decrement does not 3 years 3 months. The white mean total IQ is exist in this group, it would seem most 102, SD = 16.7; the black mean total IQ is doubtful that it could be found in any sub- 71, SD = 15.1. population within our borders. Unlike the California study, in which children from kindergarten through sixth grade were used, Sibling Analyses the present study includes children from kindergarten through twelfth grade, thereby All of the analyses are based on siblings increasing the chances of detecting IQ dec- from families with two or more children who rement by the method of differences be- are in school and who have been tested. tween younger and older siblings. (The mean number of children per family with two or more children is: white = 2.42, black = 3.29.) Method An age decrement in IQ should be indi- cated by a positive difference between Subjects younger and older siblings (i.e., Y-O). The subjects in this study were all of the white and black children enrolled in the public schools of a small rural town in the southeastern part of Georgia.1 The Test of IQ as an Interval Scale. The sibling method must assume an interval scale of the I am indebted to R. T. Osborne for securing these measurements in question. This becomes an data. especially important consideration in com- 186 ARTHUR R. JENSEN paring sibling differences across white and having different scale properties in various black groups whose IQs are predominantly parts of the full range. distributed in different ranges of the IQ scale. Are IQ differences in the lower and Younger - Older Sibling IQ Difference. If IQ upper parts of the scale really equivalent? declines with age, there should be a positive The most appropriate method of deter- mean difference between the IQs of younger mining this for the present study is to deter- minus older siblings (i.e., Y-O). Table 1 mine if there is any systematic relationship shows the mean Y-O sibling IQ differences between absolute sibling IQ differences and for all sibling pairs of the same birth order the mean IQ of sibling pairs. If there is no within each family of a given size. This significant correlation between absolute dif- method thus does not confound the mean ferences and means of sibling IQs, it could sibling difference with family size as would not be argued that the results of the sibling be the case if we simply averaged all possible method used to determine IQ decrement sibling differences within each family. Doing (based on the IQ difference between the latter tends to exaggerate the magnitude younger and older siblings) are a scale ar- of Y - O sibling differences, should they tifact. There is no theoretical basis for ex- exist, in whichever group (in this case the pecting that sibling absolute differences in black) that has the larger number of siblings IQ should differ as a function of their general per family. Families with more than five sib- location on the IQ scale. The finding of a lings were excluded from the analysis, since significant correlation between absolute dif- the Ns are too small to permit reliable statis- ferences and means, therefore, would sug- tical treatment. gest that the full range of IQs in the two racial samples do not fall on an interval Positive Y-O sibling differences which scale. are significantly greater than zero are indi- To insure that possible nonlinear as well cated by asterisks. One-tailed / tests are as the linear components of the correlation used since only positive Y-O sibling differ- between IQ differences and means could be ences are indicative of an age decrement in detected, a multiple regression analysis was IQ. (Negative differences could, of course, used and the squared multiple correlation also be significant and interesting in their (R2) was tested for significance. First, pow- own right, but they would not indicate an age ers of the sibling pair absolute difference decrement in IQ and so would not be rele- (Idl, Id/2, IdP, Idl4, /d/5) in IQ were used as vant to testing the present hypothesis.) the predictor variables, with sibling pair It is clear from Table 1 that there are larger mean IQ as the dependent variable. Second, and more significant Y-O sibling IQ differ- the first five powers of the sibling mean IQ ences for blacks than for whites. This holds were used as predictors, with the sibling pair to about the same degree for both verbal and absolute differences as the dependent vari- nonverbal IQ. In contrast, the direction and able. This was done separately for whites magnitude of the sibling differences in the and blacks, as well as for the two racial white sample are inconsistent and small. groups combined. All possible paired com- This finding is made more impressive by the parisons of siblings were used in these fact that the white sibling pairs show a sig- analyses (white = 364 pairs; black = 1004 nificantly (p < .05) greater age separation pairs). than do the black siblings. In every analysis R2 was negligible (.002 Thus, overall there is a significant age dec- < R2 < .005) and nonsignificant (.17

= 16.22, and for whites it absolute differences and mean IQs of sibling is .12, SD = 16.80. The difference of 3.19 pairs. Therefore, any signed sibling mean between these means is significant, /(641) = differences, such as the mean of Y-O, can- 2.38,p < .02, when the standard error of the not be interpreted as an artifact of the IQ difference is based on the mean variance CUMULATIVE DEFICIT IN IQ 187

within groups (i.e., family sizes and sibling

g3v?s7?vvv B- W - pairs).

o differenc e I Age Difference and IQ Difference £ u '•$ I OO«N

6 B- W is quite small (n = 20) one cannot give much differenc e Blac k importance to this sharp break in the trend. OooOrjoo — Ovv-. QOv Q (Ni^oo^r-i^i — ^o^5 Over the range of 1 to 7 years apart, the '£ regression coefficent of sibling IQ difference RS*=SRR88a on age differences is 1.62 for verbal IQ, 1.19 3 differe n

III 1 "I hi for nonverbal IQ, and 1.42 for total IQ. That a S O •o a* o^trO^r-'ocai^f^od is to say, for every year's difference in age, c 0 O over the age range from 6 to 16, verbal IQ Z S JS decreases on the average 1.62 points per S S-Nl; £ 3 year, nonverbal IQ decreases 1.19 points per u V year, and total IQ decreases 1.42 points per •§ year. This rate of decline could account for a

U 1 6 -1•a total cumulative decrement of some 14 to 16 1 V IQ points between the ages of 6 and 16 years. >? o y r-tN^1^^^^, — c\Tf The ages at which the gradual decrement V c in IQ begins and ends cannot be determined b £ i •f KSoc^o.j^.jgs fc from the present data, which include only •8 a 1 1 f 1 "~ 1 1 •5 subjects ranging in age from 6 to 16 years. It g Q would be important to know if Y-O sibling > u 00 IQ differences persist beyond the age where i ^_ slls&pssls I! the younger sibling is 18. Since mental « 1 5 2 " —4 growth stabilizes at about age 18, one should ^ ^ ^j u ? C S expect from the cumulative deficit hypothe- & 1 1 2 fflo g sis that by age 18 the average deficit of the CO 1)£ 1 6 L' CO ° i younger sibling should become equal to that i of the older, so that the younger-older sib- * ling IQ difference should disappear after age 1 S 18 or so. Information regarding this predic- 1 III" 1 tion would seem to be crucial for the viabil- I Q differenc e f erence s < reate r tha n z e f = Olde r - 1 He d A ) t o yo u w ity of the cumulative deficit hypothesis. is a vCr^r^ — -c

S"t OQCQUCQUOCQUQul test , * p < .05 V ) = Younge r - O k ndicat e mea n i s sig r the siblings who are 3 years apart. They .£ <" < 03 < CO' O < Co' U Q (7 I show a significantly decreasing IQ differ- <** > ence going from the younger to the older part S 1" fi I (5 < 3 j Z. « « of the age scale. This result is so markedly 8 u. t out of line with the results for 2 years apart oo oo

Table 2: Mean Difference in Verbal (V), and Nonverbal (NV), and Total (T) IQ Between Younger and Older Black Siblings as a Function of Age and Age Difference, with F Test of Linear Trend in Each Column Younger - Older IQ Difference \ Year apart 2 Years apart 3 Years apart 4 Years apart Ages V NV T Ages V NV T Ages V NV T Ages V NV T 6-7 -3.2 1.4 -1.1 6-8 -5.3 -2.0 -3.3 6-9 5.7 12.0 10.4 6-10 10.3 7.3 11.1 7-8 -3.4 2.5 .1 7-9 -4.0 -.5 -1.3 7-10 10.2 8.1 11.4 7-11 3.3 5.4 3.9 8-9 -5.1 6.3 -.3 8-10 5.6 0.9 3.9 8-11 7.2 3.4 4.3 8-12 2.4 1.5 .9 9-10 -2.8 -8.2 -5.6 9-11 6.3 -3.3 1.2 9-12 7.6 -.4 4.6 9-13 8.1 12.2 9.7 10-11 6.1 8.8 6.1 10-12 .1 1.8 .3 10-13 5.7 11.0 7.5 10-14 5.5 1.7 2.7 11-12 -4.6 1.9 — 1 I 11-13 5.4 7.5 7.0 11-14 -.6 6.6 2.6 11-15 2.3 .9 1.1 12-13 9.4 10.5 10.3 12-14 3.2 3.5 2.3 12-15 .5 .1 -1.0 12-16 10.7 12.1 10.6 13-14 -6.6 -7.9 -8.6 13-15 -2.3 -6.7 -5.6 13-16 -2.2 -3.8 -4.0 yo 14-15 7.5 10.6 9.6 14-16 4.2 8.3 6.4 H 15-16 10.2 11.0 11.5 C Mean" -.03 3.13 1.33 2.23 1.60 1.74 4.35 4.77 4.53 6.26 6.36 5.97 F» 6.02 .87 3.43 1.57 1.54 .83 7.16 4.16 9.43 .24 .27 .01 116 116 116 186 186 186 144 144 144 114 114 114 dF cn P < .02 .35 .07 .21 .22 .36 .01 .05 .01 .63 .60 .90 Z C/! tn 5 Years apart 6 Years apart 7 Years apart 8 Years apart Z 6-11 8.2 9.2 9.5 6-12 3.6 4.5 4.8 6-13 10.3 9.4 10.8 6-14 6.9 2.1 5.6 7-12 3.7 4.5 4.9 7-13 9.3 22.8 17.1 7-14 2.8 4.3 2.6 7-15 4.0 13.8 8.4 8-13 11.0 7.4 9.4 8-14 14.2 6.3 9.5 8-15 11.3 13.2 11.4 8-16 3.3 2.7 .7 9-14 7.1 2.4 3.8 9-15 6.3 .2 2.4 9-16 15.7 9.3 12.3 10-15 .1 9.3 2.3 10-16 7.5 3.5 5.5 11-16 6.9 9.1 8.3 Mean 6.58 6.76 6.45 8.30 7.64 7.97 10.29 9.52 9.75 4.74 5.27 4.34 F .39 .08 .25 .05 2.73" 1.34" 1.53 .12 .24 .29 .00 .27 df 90 90 90 55 55 55 55 55 55 17 17 17 P < .53 .78 .62 .81 .10 .25 .22 .73 .63 .59 .99 .61

a Each value was weighted by N in obtaining the mean of the Y-O differences. b F for linear trend only. Other trend components (quadratic, cubic, quartic). if significant beyond p < .04 are indicated in footnotes. c Degrees of freedom for the donomi naton the numerator always has

and 4 years apart that it is regarded as anom- 2, 3, etc.) of the younger sibling, (d) Y-O alous and theoretically uninterpretable difference in birth order, and (e) family size within the present set of data. No other (1 to 5). trends in sibling IQ differences as a function Stepwise regressions were done, with a of absolute age even approach significance. predetermined order of entering the inde- (The significant cubic component for 6 years pendent variables so that one can determine apart, noted in footnotes d and e of Table 2, the independent contribution of the racial can only be regarded as flukes within the classification after the four other sources of overall context of these results.) variance have all been accounted for. Table In general, the trend analysis indicates 3 shows the results in terms of the squared that the magnitude of the Y-0 sibling IQ multiple correlation (R2), which is the difference, for any given age difference, is cumulative proportion of the total variance not significantly a function of absolute ages in the Y-O sibling differences associated of the siblings in the range of ages sampled in with each additional independent vari- this analysis (viz., ages y to xy years, with able. 1 to 8 years differences in siblings' ages). Also shown is the simple correlation (zero-orderr) between the Y-O sibling IQ Multiple Regression Analysis of Family differences and each of the independent Size, Age Spacing, and Birth Order variables. Effects on IQ Decrement The analyses in Table 3 indicate that the first four independent variables account for Zajonc and Markus (1975) have presented some 2 to 3 percent of the variance in Y-O evidence that IQ is related to family size, sibling differences, which is significant be- birth order, and spacing (i.e., age difference yond the .001 level. The race variable inde- between siblings adjacent in birth order). To pendently contributes an additional 2.1 per- what extent is the IQ age decrement in cent of the variance of sibling differences on blacks related to the above variables? And the verbal, nonverbal, and total IQ. For each what proportion of the variance in Y-O test the independent contribution of race is sibling differences is associated with the ra- significant beyond the .001 level. This fact cial classification (blacks vs. white) inde- clearly establishes the significance of the pendently of the above variables? race difference in age decrement in IQ as To find out, multiple regression analyses indicated by the Y-O sibling difference. were done. The dependent variable is every possible Y-O sibling difference (N = 1031) Multiple Regression Analysis Within Racial in IQ (verbal, nonverbal, and total IQ sepa- Groups. To determine the contribution of rately). The five independent variables are: each of the independent variables (except (a) race (white or black), (b) sibling age dif- race) listed in Table 3 to the sibling differ- ference (0- Y in months), (c) birth order, (1, ences within each racial group, stepwise

Table 3: Multiple Regression Analysis with Y-O Sibling IQ Difference" as the Dependent Variable, (A' = 1031) and Simple r Between Sibling Difference and Independent Variables Independent Verbal IQ Nonverbal IQ Total IQ variable" R2 rc R2 rc R2 rc Family size .001 .040 .007*' .085 .005* .071 Birth order of Y sibling .003 -.019 .008* .021 .007* -.013 Y-O difference in birth order .008* .092 .017*** .124 .014** .117 O-Y age difference .035*** .181 .023*** .121 .028*** .155 Race .056*** .131 .044*'* .157 .049*** .146 Nole. Y-O = Younger - Older; O - Y = Older - Younger. » The overall mean Y-O sibling IQ difference is: Verbal IQ = 3.39, SD = 16.68; Nonverbal [Q = 3.79, 20. 12; Total IQ = 3.54, 5£> = 17.36. " Variables listed in the forged order of entering the stepwise regression. v With 1030 df. r greater than .026 is significant at the .05 level; r greater than .081 is significant at the .01 level. * p < .05. ** p <.01. *** p <.OOI. 190 ARTHUR R. JENSEN multiple regression analyses were per- form of the mental growth curves of blacks formed on the white and black samples sepa- and whites, with blacks having a more nega- rately. The results are shown in Table 4. tively accelerated growth curve, which None of the values of/?2 is significant in the would result in younger-older sibling dif- white sample. The fact that birth order and ferences in black IQ when the IQs are nor- age difference jointly do not show a sig- malized on a predominantly white sample. nificant R2 means there is no evidence for an The existing data do not permit a defini- age decrement in IQ in the white sample. tive rejection of one or the other of these In the black sample, however, both birth alternative hypotheses. Moreover, these order and age difference independently con- two hypotheses are not mutally exclusive; tribute a significant (p < .001) increment to both genetic and environmental factors the variance of sibling IQ differences, for could be involved in the progressive decre- verbal, nonverbal, and total IQ. This fact ment phenomenon. However, the present clearly establishes a significant age decre- results on Georgia blacks, when viewed in ment in IQ in the black sample. connection with the contrasting results for California blacks, would seem to favor an Summary and Conclusions environmental interpretation of the progres- sive IQ decrement. If the progressive IQ These sibling comparisons for poor black decrement were a genetic racial effect per children in rural Georgia clearly show a sig- se, it should have shown up in the California nificant and substantial decrement in verbal blacks as well as in the Georgia blacks, even and nonverbal IQ between kindergarten and if one granted that California blacks have a Grade 12. The IQ decrement is a fairly linear somewhat larger admixture of Caucasian function of age within this range. The phe- ancestry than do blacks in Georgia (Reed, nomenon predicted by the cumulative deficit 1969). But the California blacks showed a hypothesis is thus demonstrated at a high slight, though significant, decrement only in level of significance. verbal IQ, which one might expect to be According to the cumulative deficit hy- more susceptible to environmental or cul- pothesis, the age decrement in IQ is a result tural effects than nonverbal IQ. The blacks of the cumulative effects of environmental of rural Georgia, whose environmental dis- disadvantages in factors related to mental advantages are markedly greater than in the development. A counter hypothesis would California sample, show considerable dec- be that there are genetic differences in the rements in both verbal and nonverbal IQ,

Table 4: Multiple Regression Analysis with Y - O Sibling IQ Difference" as the Dependent Variable, and Simple r Between Sibling Difference and Independent Variable, Separately by Race White (N = 349 Sibling Pairs) Black (N = 682 Sibling Pairs) Verbal IQ Nonverbal IQ Total IQ Verbal IQ Nonverbal IQ Total IQ independent 2 2 2 2 variable" R2 rc R r0 R rc R2 r" R rd R r"

Family size .005 -.069 .001 -.027 .001 -.028 .001 .026 .004 .060 .002 .044 Birth order of Y sibling .014 -.119 .002 -.040 .013 -.108 .001 -.006 .004 .012 .033 -.008 Y - O difference in birth order .019 -.044 .003 -.024 .017 -.017 .108** .124 .028*** .153 .022*** .140 0-Y age dif- ference .025 .073 .007 .050 .021 .061 .079*** .265 .039*** .181 053»*» .226 Note. Y-O = Younger - Older. " The overall mean Y-O sibling IQdifference for whites is: Verbal IQ= .33,50= 17.53: Nonverbal IQ= -. 64. SD = 20.27; TotallQ = -.01. SB- 17.23. For blacks: Verbal IQ = 4.96, SD = 16.01; Nonverbal IQ = 6.05. SD = 19.68: Total IQ = 5.35. SD = 17.16. " Variables listed in the forced order of entering the stepwise regression. f With 300 ilf, r greater than .113 is significant beyond the .05 level; r greater than . 148 is significant beyond the .01 level. " With 700 rf/, r greater than .088 is significant beyond the .05 level; r greater than . 115 is is significant beyond the .01 level. 1 p < .05. " p < .01 *' p < .001 CUMULATIVE DEFICIT IN 1Q 191 but again the decrement is larger for verbal York, R. L. Equality of educational opportunity. IQ. (Despite this, the verbal IQ still remains Washington, D. C.: U.S. Government Printing Office, 1966. slightly higher than nonverbal IQ for the Jensen, A, R. Cumulative deficit: A testable hypothe- Georgia blacks.) Thus it appears that a sis? Developmental Psychology, 1974,10, 996-1019. cumulative deficit due to poor environment (a) has contributed, at least in part, to the rela- Jensen, A. R. How biased are culture-loaded tests? tively low average IQ in the present sample Genetic Psychology Monographs, 1974,90, 185-244. (b) of blacks in rural Georgia. Jensen, A. R. Culture bias and construct validity. Phi Delta Kappan, 1976,57, 340-346. REFERENCES Reed, T. E. Caucasian genes in American Negroes. Science, 1969, 165, 762-768. Euros, O. K. (Ed.). Seventh mental measurements Zajonc, R. B., & Markus, G. B. Birth order and intel- yearbook (Vol. 1). Highland Park, N.J.: Gryphon lectual development. Psychological Review, 1975, Press, 1972. 82, 74-88. Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, J., Mood, A. M., Weinfeld, F. D., & (Received July 6, 1976)

Call for Guest Reviewers

In an effort to increase the number of individuals involved in the review of manuscripts submitted to Developmental Psy- chology, the Editor invites those who are interested in being considered as guest reviewers to write to him, indicating those research and topic areas in which they would be most qualified to review manuscripts.