Multiple and partial correlation methods
The literature on social class and racial differences in intelligence is replete with reports of correlations between innumerable environmental, attitudinal, and personal variables and IQ. Correlations have been reportedl between IQ and variables such as family size, absence of father, l)rematurity, child-rearing practices, birth weight, family income, parental education, protein intake, books in the home, mother's age, position in family - the list is almost endless. Each variable naay yield some non-zero correlation with IQ, however small, and since some racial groups differ on many of these variables in the: same direction as they differ in IQ and in the same direction that the variables are correlated with IQ, it creates the impression tilat the racial IQ difference must be easily explainable in terms of the racial differences on all of these environmental factors. T'he fallacy in this, of course, is that the IQ variance (either within or between racial groups) accounted for by all these environmenLtal factors is not equal to the sum of all the various environme:ntal influences. In accounting for IQ variance, environmental measures, not being independent sources of variance, must be adde:d up in the fashion of a multiple regression equation. That is to say, the contribution to IQ variance of each environmental factor mllst be added up after removing whatever IQ variance it has in coInmon with all the previous factors added in. The greater the correlation between the environmental factors, the smaller is the contribution to IQ variance of each successive factor which is added to the composite of environmental variables.