152   NATURAL INHERITANCE.   [ciwAr.

them 0.50 of the peculiarity of S, the four grandparents together with what they imply of the previous ancestry contribute 0.32, being an aggregate of 0'82, leaving a residue of 0.18 to be rateably assigned as 0.12 to light, and 0'6 to dark. A hazel-eyed Parent is here reckoned as contributing 0'16 to light and 0.9 to dark; a hazel-eyed Grand-Parent as contributing 0'5 to light and 0'3* to dark. All this is tabulated in Table 17, and its working explained by an example in the columns headed III. of Table 18.

Results.-A mere glance at Tables 19 and 20 will show how surprisingly accurate the predictions are, and therefore how true the basis of the calculations must be. Their average correctness is shown. best by the totals in Table 19, which give an aggregate of calculated numbers of light-eyed children under Groups I., II., and III. as 623, 601, and 614 respectively, when the observed numbers were 629 ; that is to say, they are correct in the ratios of 99, 96, and 98 to 100. .

Their trustworthiness when applied to individual families is shown as strongly in Table 20 whose results are conveniently summarised in Table 21. I have there classified the amounts of error in the several calculations thus if the estimate in any one family was 3 lighteyed children, and the observed number was 4, I should count the error as 1.0. I have worked to one place of decimals in this table, in order to bring out the different shades of trustworthiness in the three sets of calculations, which thus become very apparent. It will be