The results for both the treadle andhopper groups can be explained by asingle underlying mechanism, that is,the development of a set to respond(industriousness) or of a set to notrespond (laziness). Presumably thetreadle group learned that reinforce-ment is contingent on some responseand they entered the autoshaping situa-tion with a set or expectation that asimilar contingency still applied. Thehopper birds responded less becausethey anticipated a continuation of thenoncontingency that existed in theirfirst stage of training.Our interpretation is entirely parallel

to that offered by Maier et al. (2)for the learned helplessness phenome-non. If that interpretation is valid, thenthe significance of subjects' ability tocontrol their environment is not re-stricted to commerce with aversivestimuli or to any single motivational orreward system. Thus, speculations in-volving presumed physiological or bio-chemical correlates of helplessness interms of stress reactions (3) are likely

Wilson (1) presented evidence thathe interprets as demonstrating that thesimilarity in overall level and develop-mental profile contour of scores on theBayley infant test is greater for mono-zygotic (MZ) than for dizygotic (DZ)twins for assessments made during thefirst year (3, 6, 9, and 12 months) andthe second year (that is, 12, 18, and 24months) of life. The intraclass correla-tions expressing similarity within pairswere very high, approaching and in somecases equaling the reliability of the test.

Wilson cites the fact that the intra-class correlations for DZ twins werehigh for both overall level (r=.75 and.79, for the first and second years, re-spectively) and profile contour (r = .52and .50), and that the size of theseconcordances signifies "that the differ-ences within DZ pairs produced by genesegregation and different life experi-ences are comparatively small in rela-tion to the sizable differences betweenpairs" (1, p. 917). He concludes thatgenetic factors are paramount in suchdevelopment. However, circumstancessurrounding the analysis of these datamay modify this interpretation.The degree of heritability for a trait

depends on the difference between the

1004

to be of limited use. Further researchwill be necessary to validate our in-terpretation as well as to explore thelimits of its applicabijity.

LARRY A. ENGBERGGARY HANSEN

ROBERT L. WELKERDAVID R. THOMAS

Department of Psychology, Universityof Colorado, Boulder 80302

References and Notes

1. M. E. P. Seligman and S. F. Maier, I. E.xp.Psychol. 74, 1 (1967).

2. S. F. Maier, M. E. P. Seligman, R. L.Solomon, in Punishment and Aversive Behlar-ior, B. A. Campbell and R. M. Church, Eds.(Appleton-Century-Crofts, New York, 1969),pp. 299-342.

3. M. E. P. Seligman, S. F. Maier, R. L.Solomon, in Av-ersive Conditioning and Learni-ing, F. R. Brush, Ed. (Academic Press, NewYork, 1971), pp. 347-400.

4. P. L. Brown and H. M. Jenkins, 1. Exp. Anal.Behar. 11, 1 (1968).

5. E. Gamzu and D. R. Williams, Science 171,923 (1971).

6. D. R. Williams and H. Williams, J. Erp.Anal. Behas. 12, 511 (1969).

7. One subject in the control group died duringthe experiment and was not replaced.

8. J. B. Overmier and M. E. P. Seligman, J.Coamp. Physiol. Psychol. 63, 28 (1967).

9. Supported by NIH research grant HD-03486and training grant MH-10427.

2 August 1972

correlations -for MZ and DZ twins, noton the absolute size of the DZ correla-tion. For example, "broadsense herita-bilities" can be calculated by takingtwice the difference between rN1z andrl;z (2). Mating is assumed to be ran-dom in this heritability estimate, an as-sumption that is probably tenable sinceinfant test scores do not relate verystrongly to adult mental or personalitycharacteristics that might influence mateselection (3). Wilson's data treated inthis manner give heritabilities of .30and .20 for overall level (first and sec-ond years, respectively) and .50 and.30 for profile contour. These valuesare lower than those reported for IQin later childhood (4), and do not seemto warrant the conclusion that "infantmental development was primarily de-termined by the twins' genetic blue-print and that . . . other factors servedmainly a supportive function" (1, p.914).The high correlations for DZ twins

may derive from common nongeneticas well as genetic circumstances. Infact, since the genetic correlation forDZ twins averages .50 (they have halftheir genes in common), correlationsbetween DZ twins substantially higher

than .50 must reflect common environ-mental circumstances or assortive mat-ing, or both these factors.When these same analyses were per-

formed on 142 sibling pairs from theFels Longitudinal Study, the intraclasscorrelations for Gesell developmentalscores were .24 and .44 for overalllevel (at 6 and 12 months and at 12,18, and 24 months, respectively) and.a9 and .14 for profile contour. Thetwin correlations reported by Wilsonare two to six times these values, de-spite the fact that the degree of geneticoverlap is the same for DZ and siblingpairs. Twins may be more similar thansiblings because they share environ-mental circumstances (prenatal environ-ment, stimulating familial experiences,and so forth) and because those en-vironmental factors have their effects atthe same age for twins but at differentages for siblings.The method of analysis used by Wil-

son also raises some issues of interpreta-tion. His intraclass correlations expresswithin-pair variability relative to theappropriate total variability separatelyfor MZ's and DZ's, before these groupsare compared. Such a procedure maybe justified when the variability betweenindividuals in one group is differentthan that in another but this is report-edly not the case for the data presented.A more direct approach is to contrast

the within-pair variability for MZ's andDZ's by-calculating the ratio of the meansquare within pairs for DZ's divided bythe mean square within pairs for MZ's.For Wilson's data, this procedure yieldssignificant differences between zygositygroups for overall level during the firstyear (F= 3.44, d.f. = 51/45, P<.000 )and second year (F = 1.74, d.f. =46/51, P < .03) and for profile con-tour during the first year (F= 1.94,d.f.=153/135, P<.0001) but notduring the second year (F = 1.1 8,d.f.=92/102, P = .25). Thus, whendifferences in similarity within pairs areconsidered directly, MZ's are moresimilar than .DZ's in profile contouronly during the first year of life.The entire univariate analysis-of-

variance model involving repeated mea-sures, which was used by Wilson toestimate profile similarity (as well as

the above-mentioned analyses), tends tohave a positive bias toward significanteffects (in this case favoring within-pair similarity) if the covariances (thatis, correlations) between all pairs ofscores measured on the same individ-uals are not equal (5). Such hetero-geneity is almost always present in

SCIENCE, VOL. 178

Similarity in Developmental Profile amongRelated Pairs of Human Infants

on August 12, 2021

http://science.sciencem

ag.org/D

ownloaded from

longitudinal data of this sort becauseadjacent-age assessments are morehighly correlated than chronologicallydistant ones. The data in Wilson's re-port do not permit an estimation of thedegree of positive bias with respect tosignificance levels or the size of intra-class correlations. A multivariate ap-proach to the repeated variable presentsa statistical alternative that does notmake these covariance assumptions.

Finally, in contrast to MZ pairswhich are always same-sexed, Wilson'sDZ group was composed of cross-sexedas well as same-sexed pairs. Sincefemales had higher scores than males atall ages (statistically significant at 18months), the reported results are fur-ther qualified to an unknown extent byintroducing this additional variabilitywithin DZ pairs.From the Fels sample, we compared

(i) 142 pairs of siblings, (ii) unrelatedcontrols matched for year of birth, sex,and parental education, and (iii) unre-lated pairs matched for year of birthand sex but not parental education.While siblings were more similar in thegeneral level of their performance onGesell tests administered at 6, 12, 18,and 24 months, there were no differ-ences in within-pair similarity betweenthe three groups for profile contour asassessed either by comparison of thesquare root of the sum of squared devi-ations between corresponding points orby multivariate profile analysis (6).When all the data on similarity among

related individuals in profile contour ofmental test performance (1, 6) is con-sidered, it would appear that the evi-dence for non-zero heritability of pro-file contour after the first year of lifeis ambiguous at best.

ROBERT B. MCCALLFels Research Institute,Yellow Springs, Ohio 45387

References and Notes

1. R. S. Wilson, Science 175, 914 (1972).2. D. S. Falconer, Introduction to Quantitative

Genetics (Ronald, New York, 1960), p. 185.3. R. B. McCall, P. S. Hogarty, N. Huriburt,

Amer. Psychol., in press.4. M. P. Honzik [Child Develop. 28, 215 (1957)]

reports increasing correlations for general levelof IQ between biological parent and adoptedchild as the child's age increases.

5. 0. E. P. Box, Ann. Math. Statist. 25, 484(1954).

6. R. B. McCall, paper presented at the meetingof the American Psychological Association,Honolulu, September 1972.

7. -, Science 170, 644 (1970).8. Research supported in part by NIH grantsHD 04160 to the author, NIH grants FR-04437,HD-00868, and FR-00222 to the Fels ResearchInstitute, the Grant Foundation of New York,and by the Fels Fund of Philadelphia. Ithank M. Appelbaum, V. Crandall, F. Falk-ner, A. Jensen, and J. Loehlin for their com-ments on an earlier draft of this technicalcomment.

1 May 1972

1 DECEMBER 1972

Wilson (1 ) states that early intel-lectual development is "primarily deter-mined" by genetic factors. As a majorsource of support for his argument,Wilson presents data indicating a highdegree of within-pair concordance foroverall intelligence of dizygotic twins.These data are used to illustrate theimportance of between-pair variance,which is postulated to be a function ofboth environmental and genetic factors.Wilson then proceeds to rule out en-vironmental variance as important bypresenting results indicating insignifi-cant correlations between socioeconom-ic status (SES) and early IQ. However,in ruling out environmental varianceby the use of SES, Wilson ignores cer-tain characteristics of SES which makethis variable unsuitable for this purpose.

First, there is a question of the utilityof SES as a valid measure of distinctenvironmental differences betweengroups. Wellman (2) indicated that therange of environments within anygiven SES level was so great as to makesuspect any statements using SES as ameasure of environment. Pavenstedt(3), studying the types of environmentsencountered within a given socioeco-nomic group, and Tulkin (4), investi-gating the role of socioeconomic status,race, and family environment in schoolachievement, pointed out that control-ling for SES does not at all control fortypes of experiences encountered.Rather than measure distinct homoge-nous environments, SES appears tomeasure overlapping heterogenous en-vironments. This overlap between sup-posed categories would minimize anycorrelation such as that between SESand intelligence. Evidence on this pointcan be seen in the data presented byWilson. Because dizygotic twins aregenetically similar to full sibs, the maxi-mum expected genetic correlation be-tween dizygotic twins should be .50(5). However, the dizygotic twins' cor-relations for mean and overall level,presented by Wilson, are greater thanthis expected value (of the six meancorrelations, three are .61 and threeare 1 .72; both of the two overall levelcorrelations are .75). This degree ofcorrelation indicates the operation ofnongenetic factors in the dizygotictwins' mental test performance. Sincethe correlations between SES and IQreported by Wilson are essentially zero-order, it seems that whatever the non-genetic factors affecting the twins' men-tal test performance, they are not beingadequately reflected by SES indices.The heterogenous nature of the SES

variable reflects a second, more seriousproblem for the use of this variable asan experiential measure. As human be-havioral ecologists (6) have pointedout, SES differences tell us nothing atall about the specific, proximal experi-ences being undergone by a particularchild. It is these specific experiencesthat will affect the development of thechild's intelligence-not distal labelssuch as SES, which may or may notreflect these experiences. This distinc-tion between distal as opposed toproximal experiences is reflected by theavailable evidence on early intelligenceand experience. Thus, significant rela-tions between the distal variable SESand early intelligence level are notfound. However, significant correla-tions are found between proximal vari-ables, such as linguistic stimulation inthe home or overstimulation of infants,and their performance on a Piaget-based scale of infant intelligence (7);between responsivity of the environ-ment or stimulus variety of the homeand infants' performance on the Bayleymental development scale (8); and be-tween tactual stimulation of infantsand their performance on the Bayleyscale (9).The above argument is not to be

considered as denying the importanceof genetic factors in the developmentof intelligence. The monozygotic-dizy-gotic comparisons reported by Wilsonprovide clear evidence for the opera-tion of genetic factors in the processof intellectual development. However,before one can assign the overwhelm-ing share of control to genetic factors,as Wilson proposes in this thresholdmodel, one must first produce resultsshowing that either (i) there are nodifferences in the specific environmentof children who are discrepant in in-telligence, or (ii) any differences thatoccur in the specific environment areunrelated to intellectual development.As of yet, evidence of this type is notavailable.

THEODORE D. WACHSDepartment of Psychological Sciences,Purdue University,Lafayette, Indiana 47907

References and Notes

1. R. Wilson, Science 175, 914 (1972).2. B. Wellman, in 39th Yearbook of the National

Society for the Study of Education (PublicSchool Publishing, Bloomington, Ill., 1940),pp. 1-21.

3. E. Pavenstedt, Amer. J. Orthopsychiat. 35, 89(1965).

4. S. Tulkin, J. Pers. Soc. Psychol. 9, 31 (1968).5. D. Falconer, Introduction to Quantitative

Genetics (Ronald, New York, 1960), pp. 1-184.

6. B. Caldwell, Merrill-Palmer Quart. 16, 260

1005

on August 12, 2021

http://science.sciencem

ag.org/D

ownloaded from

(1970); M. Schoggen and P. Schoggen,DARCEE Papers and Reports (Demonstra-tion and Research Center for Early Educa-tion, John F. Kennedy Center, George Pea-body College, Nashville, Tenn., 1971), vol.5, No. 2.

7. T. Wachs, I. Uzgiris, J. Hunt, Merrill-PalmerQuiart. 17, 283 (1971).

8. L. Yarrow, J. Rubenstein, F. Pedersen, paperpresented at the 1971 meeting of the Societyfor Research in Child Development.

9. N. Solkoff, S. Yaffee, D. Weintraub, C.Blaise, Develop. Psychol. 1, 765 (1969).

10. Supported by NIH research grant HD-04514.I thank F. Weizmann for his criticisms of apreliminary draft of this comment.

8 May 1972; revised 28 July 1972

McCall refers to the substantial sizeof the dizygotic (DZ) twin correlationsfor overall developmental level andprofile contour, and he quotes the smallportion from my report in which therelation between genetic influence andthe size of the DZ twin correlations isdiscussed. Then he states, "The degreeof heritability for a trait depends onthe difference between the correlationsfor MZ [monozygotic] and DZ twins,not on the absolute size of the DZ cor-relations."

All the analyses in my report wereexplicitly performed to provide statisti-cal tests of the differences between thecorrelations for MZ and DZ twins, asexplained in the text and tables preced-ing the mention of the DZ correlations.The conclusions about genetic influencewere based on the results of these tests,which showed significantly higher con-cordance for MZ twins, and the DZcorrelations were interpreted in thecontext of these results. The analysis ofthe twin data was in no sense deficientof comparisons of MZ and DZ data.

McCall's introduction of the heritabil-ity ratio brings up a chronic sourceof confusion and misinterpretation inthis research area. The ratio came origi-nally from plant and animal breedingstudies in which parentage was con-trolled and a planned mixture of geno-types was achieved through inbreedingor crossbreeding. When the appropriateassumptions were met, the ratio ex-pressed the proportion of genetic vari-ance in whatever phenotypic charac-teristic was being measured.

But as an expression of the geneticcontribution to test-score variance ona psychological variable, the heritabilityratio can be seriously misleading. Thereare several different formulas for theratio which use different combinationsof variance estimates and yield some-what different results (1). Further, thesame heritability ratio may be obtainedfrom two sets of within-pair correla-tions which have quite different implica-

1006

tions. For example, by McCall's methodof doubling the difference between MZand DZ correlations, a heritability esti-mate of .30 is obtained with R5,lz = .90and Rl,z = .75; and it is also obtainedfor R.1,z = .75 and RI>% = .60; for R,1z- .50 and Rl,z = .35; and for R^5y =.20 and RI)z = .05. The sampling errorof the difference gets larger as the cor-relations get smaller, so that while thefirst difference is significant, the next ismarginal (P = .08), and the other differ-ences are clearly nonsignificant. In fact,neither of the correlations in the lastexample is significantly different fromzero. The genetic influence is hardlythe same in each example, yet a herita-bility ratio obscures these essential dif-ferences and draws attention away fromthe central data, which are the within-pair correlations, their sampling error,and the test of significance for the dif-ference between the correlations.The limitations are serious enough

that the heritability ratio should bepermanently retired. Where the datawarrant it, there are other more com-prehensive models for estimating ge-netic and environmental variance com-ponents in psychological variables (2, 3).A comparison of sibling pairs with

DZ twins is a desirable method forevaluating the common environmentaleffects of "twinness" as such, if otherfactors such as age at testing and thetests administered are adequately con-trolled. The correlations reported byMcCall for the Fels sibling pairs are solow, however, as to raise questionsabout the reliability of the scores, par-ticularly if the tests were administeredby different examiners over a long pe-riod. The Gesell does not lend itself toformal scoring as readily as the Bayley,and the reliability of the scoring methoditself has been questioned (4).An analysis based on intraclass cor-

relations is preferable to an analysis ofwithin-pair variances, because the in-traclass correlation takes into accountall variance between subjects, and itexpresses the proportion of variance ac-counted for by pair membership. Thestatistic flows directly from the analysis-of-variance model, and it gives a mea-sure of covariation for the twins, or therelative homogeneity of scores withinpairs (3, 5). It also compensates for anysampling fluctuation in the distributionof scores within each zygosity group.The within-pair variability, however,

is based solely on the discrepancieswithin pairs, and it gives no indicationof the similarity of scores within pairs.

The F-tests proposed by McCall permitonly the limited conclusion that thediscrepancies are significantly larger forDZ pairs; they provide no measure (ortest) of -the concordance level for eithergroup. While gene segregation shouldenhance the differences within DZ pairs,the effect will be limited by assortativemating; and in any event, a comparisonof the two intraclass correlations willbe more reliable and informative thana comparison of the two within-pairvariances.The objection to the univariate

analysis-of-variance model is in errorwith respect to the presumed biasfavoring within-pair similarity. Noncon-stant covariances may affect the com-parisons involving the repeated mea-sures, in this case the various ages atwhich the test scores were obtained. Butthere is no test of ages as a main effect-the scores are standardized separatelyat each age, which abolishes the be-tween-age variance. The analysis testsfor the degree of homogeneity withinpairs in the total score summed acrossages, which is an unbiased test underany model; and it tests for the homo-geneity of the score profile at differentages. The actual size of this within-pair correlation for score profile is notaffected by the covariance structure be-tween tests, nor is there an effect on thetest of differences between -MZ andDZ pairs. The only possible effect is onthe probability that the within-pair cor-relation for score profile is differentfrom zero. If the covariance structureis deviant enough, the actual P valuemay '(for example) be .025 instead ofthe nominal P value of .01.Where necessary, the bias can be

corrected by an adjustment in the de-grees of freedom, as noted originallyby Box (6) and as confirmed by recentMonte Carlo studies (7). In the twindata, every within-pair correlation wassignificantly different from zero atP < .001 under the most extreme andconservative correction possible, so theresults are not being inflated by a hid-den bias.

Finally, McCall reports that he foundno evidence of greater similarity inprofile contour on the Gesell tests forsibling pairs than for unrelated pairs.The study is as yet unpublished so theresults are not available for evaluation.McCall has, however, published asimilar study of older siblings from theFels sample in which he averaged adja-cent-age scores to compensate for miss-ing data (9). This method of moving

SCIENCE, VOL. 178

on August 12, 2021

http://science.sciencem

ag.org/D

ownloaded from

averages obscures any age-to-agechanges that should be reflcCted in theprofile contour, and it eliminates thevariance component that is supposedlybeing tested. Further, his statisticalanalysis dealt with within-pair differ-ences in profile contour for sibs andunrelated pairs; as noted above, thisdoes not furnish a measure of con-cordance for either group or a meansof testing whether the concordancelevel is (i) greater than zero or (ii)significantly different for the two groups.Therefore, the methods and data of theunpublished study will need carefulexamination before the results can beinterpreted.On balance, none of the results or

conclusions from the original reportare changed by McCall's comments.Wachs in his first paragraph substi-

tutes the terms intellectual development,intelligence, and IQ for the originalterm infant mental development, whichwas specifically chosen to designate thecapabilities being manifested during in-fancy. The point was made clearly thatthere are pronounced changes in de-velopmental precocity from one age tothe next, which reflect the changingcapabilities being measured by the Bay-ley mental scale plus the idiosyncraticspurts and lags in development ex-hibited by each infant. The strength ofthe twin data is in showing that thechanges in precocity occur concordant-ly for genetically related pairs. Butthere is no direct correspondence be-tween precocity in infancy and latermeasures of intelligence, and Wachs'substitution of terms serves to confusean essential distinction between thecapabilities being measured at differentages.As to Wachs' assertion that I ruled

out the importance of environmentalvariance, my specific conclusion wasthat "the caretaking and stimulationneeded to support infant mental devel-opment are sufficiently supplied by mosthome environments that fall above thelevel of impoverished. In all likelihood.however, there may be a cumulativelatent influence absorbed from thehome environment during infancy thatcombines with genetic predispositionand gradually becomes manifest asschool age approaches; since the child'smeasured IQ becomes increasinglyrelated to his parents' IQ, educationallevel, and socioeconomic status as hegets older" (10, p. 917).

Wachs objects to socioeconomic sta-

tus '(SES) as a measure of environ-mental variance because it does notcontrol for the types of experience en-countered. No assertion was made inmy report that SES was a completemeasure of environmental variance; itdoes, however, furnish a reliable methodof designating differences between fam-ily environments, many aspects of whichdo relate immediately to the infant'sexperience. The problem with Wachs'argument is that while SES is a poorpredictor of infant developmental status,it becomes an increasingly better pre-dictor of intelligence in later childhood.If there is such a heterogeneous mix-ture of environments within each SEScategory, and if SES has so little rela-tion to the types of experience whichWachs believes to be important for thedevelopment of intelligence, then whereis this predictive power coming from?The fact is that as intelligence graduallystabilizes with age, the predictive powerof SES improves, and one cannot dismissSES as an unsuitable variable withoutacknowledging this relationship.

Wachs' reference to an expected max-imum genetic correlation for DZ twinshighlights another source of confusionin this area. There is a massive concep-tual gap between .50 as the average pro-portion of shared genes for pairs of off-spring drawn from randomly matchedparents, and .50 as the correlation co-efficient expressing the ratio of within-pair covariance to the total variance ina distribution of tesit scores. There areso many assumptions needed to bridgethe gap that one cannot arbitrarilyestablish a maximum value expected forgenetically related pairs, especiallywhere assortative mating is involved.The values will ultimately be determinedby large-scale, carefully controlled de-velopmental studies including groups ofdiffering genetic composition.As to the distinction between specific,

proximal experience as opposed to SESas a distal label or variable, the distinc-tion is arbitrary in that any environ-mental variable must register as a proxi-mal experience if it is to affect behavior.An SES category is a representation ofa particular life-style, the events ofwhich impinge directly on the child andfurnish a primary dinmension of hisproximal experience. The proximal vari-ables cited by Wachs appear in differ-ential degree along the SES continuum,and while the exceptions are always amatter of interest, they do not abolishthe relation between SES categories and

the proximal experience dimension thatWachs favors.

It would be highly desirable to havea separate measurement scale of proxi-mal experience to use in conjunctionwith SES ratings, and this would permita more complete test of the relation be-tween early proximal experience andschool-age intelligence. The supportingresults cited by Wachs are tentative anddeal with changes in infant develop-mental status, whereas the more basicquestion concerns whether the effects ofearly experience will independentlycontribute to a sustained upward (ordownward) shift in school-age IQ. Theissue is not whether a supportive andappropriately stimulating environmentshould be supplied for each infant-itseems to me this answer must be af-firmative under the most elementaryconsiderations of human care-butwhether the attained level of intelli-gence at school age is directly related tothe range of variation found in familyenvironments above the level of im-poverished. This, plus carefully con-trolled follow-up studies of enrichmentprograms with impoverished infants,will identify the differential weights tobe assigned to early experience in thedevelopment of intelligence.

In the interim, the original conclul-sion about infant mental developmentis reaffirmed: For the great majority ofpairs, their life circumstances fallwithin the broad limits of sufficiencythat permit the genetic blueprint tocontrol the course of infant mentaldevelopment.

RONALD S. WILSONChild Development Unit,University of LoutisvilleSchool of Medicine,Louiisville, Kentucky 40202

References

1. S. G. Vandenberg, Psychol. Bull. 66, 327(1966).

2. C. Burt, Brit. J. Math. Statist. Psychol. 24,1 (1971); R. B. Cattell, Psychol. Rev. 67, 353(1960); J. L. Jinks and D. W. Fulker, Psychol.Buill. 73, 311 (1970).

3. 0. Kempthorne and R. H. Osborne, Amer.J. Hum. Genet. 13, 320 (1961).

4. E. E. Werner, in Sixth Mental MeasuirementsYearbook, 0. K. Buros, Ed. (Gryphon, High-land Park, N.J., 1965), pp. 807-809.5. E. A. Haggard, Intraclass Correlation and

the Analysis of Variance (Dryden, New York,1958); D. S. Falconer, Introduction to Quan-titative Genetics (Ronald, New York, 1960).6. G. E. P. Box, Ann. Math. Statist. 25, 484(1954).

7. R. 0. Collier, F. B. Baker, G. K. Mandeville,T. F. Hayes, Psychometrika 32, 339 (1967).8. D. W. Fulker, J. Wilcock, P. L. Broadhurst.Behav. Genet. 2, 261 (1972).

9. R. B. McCall, Science 170, 644 (1970).10. R. S. Wilson, ibid. 175, 914 (1972).20 September 1972

I DECEMBER 19721007

on August 12, 2021

http://science.sciencem

ag.org/D

ownloaded from

Similarity in Developmental Profile among Related Pairs of Human InfantsRobert B. McCall, Theodore D. Wachs and Ronald S. Wilson

DOI: 10.1126/science.178.4064.1004 (4064), 1004-1007.178Science 

ARTICLE TOOLS http://science.sciencemag.org/content/178/4064/1004

REFERENCES

http://science.sciencemag.org/content/178/4064/1004#BIBLThis article cites 4 articles, 2 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

trademark of AAAS. is a registeredScienceAdvancement of Science, 1200 New York Avenue NW, Washington, DC 20005. The title

(print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for theScience

1972 by the American Association for the Advancement of Science

on August 12, 2021

http://science.sciencem

ag.org/D

ownloaded from