We need not limit ourselves, however, to the argument by exclusion, for it is not difficult to present direct evidence that the differences between men are actually inherited by children from parents. The problem, formally stated, is to measure the amount by which the likeness of individuals of like ancestry surpasses the likeness of individuals of different ancestry. After subtraction of the necessary amount for the greater likeness in training, that the individuals of like ancestry will have, whatever amount is left will necessarily, represent the actual inheritance of the child from its ancestors--parents, grandparents, and so on. Obviously, the subtraction for environmental effects is the point at which a mistake is most probable. We may safely start, therefore, with a problem in which no subtraction whatever need be made for this cause. Eye color is a stock example, and a good one, for it is not conceivable that home environment or training would cause a change in the color of brothers' eyes. The correlation[30] between brothers, or sisters, or brothers and sisters--briefly, the fraternal resemblance--for eye-color was found by Karl Pearson, using the method described in Chapter I, to be .52. We are in no danger of contradiction if we state with positiveness that this figure represents the influence of ancestry, or direct inheritance, in respect to this particular trait. Suppose the resemblance between brothers be measured for stature--it is .51; for cephalic index, that is, the ratio of width of skull to length of skull--it is .49; for hair color--it is .59. In all of these points, it will be admitted that no home training, or any other influence except heredity, can conceivably play an important part. We could go on with a long list of such measurements, which biometrists have made; and if they were all summed up it would be found that the fraternal correlation in these traits as to the heritability of which there can be no dispute, is about .52. Here is a good measure, albeit a technical one, of the influence of heredity from the near ancestry. It is possible, too, to measure the direct correlation between a trait in parent and the same trait in offspring; the average of many cases where only heredity can be thought to have had any effect in producing the result, is .49. By the two methods of measurement, therefore, quite comparable results are obtained.