1886.] 



Family Likeness in Stature. 



53 



returns are undoubtedly careful and thoroughly trustworthy, so that 

 I have reason to place confidence in mean results. They bear those 

 internal tests that T apply to them better than I should have expected, 

 and when taken in connexion with and checked by the special data, 

 and used with statistical caution, they have proved very valuable to 

 me. 



I have discussed these materials in a great variety of ways to guard 

 myself against rash conclusions, but I shall not present more than 

 three primary tables, which contain sufficient materials for determin- 

 ing the constants of the formulas to be used. 



The first of them (Table III) refers to the children of what I call 

 u mid-parents " of various statures. A mid-parent is the imaginary 

 mean of the two parents, after the female measurements have been 

 transmuted to their male equivalents, so that a mid-parent of 70 

 inches in height refers to a couple whose mean stature under the above 

 reservations is 70 inches. I have given data in the " Jo urn. Anthrop. 

 Inst." (loc. cit.) to show that we need not regard differences in stature 

 between the parents, inasmuch as the distribution of heights among 

 the children proves to be statistically the same, so long as the mid- 

 parentages are alike, whether the two parents are the same or of 

 different statures. This blending of paternal and maternal qualities 

 in the stature of the offspring is one great advantage in selecting" 

 stature as a subject for the present inquiry. 



General Population. — (1.) Its variability. The value of the quartile 

 deviate in the population ogive (that is to say, the probable error) 

 may be deduced from the bottom lines of any one of the three Tables 

 III, IV, and V. Those in III and IV refer to data that are in part 

 but by no means wholly the same, that of V refers to almost totally 

 distinct data. The work is shown in Tables VI and VII; in the 

 former the ordinates are calculated whence the ogive is drawn, in the 

 latter I have given the values of the measured ordinates at the same 

 points along its axis as those to which the ordinates given in Table I 

 refer. The values of the quartile that I obtain in this way from the 

 three cases are 1*65, 1'7, and 1*7. I should say that the more careful 

 treatment that I originally adopted happened to make the first of 

 these values also 1*7, so I have no hesitation in accepting 17 as the 

 proper value of jp for all my data. 



(2.) Variability of system of mid-parents. I have published data in 

 the memoir already alluded to, to show that marriage selection takes 

 small account of stature, which is another great merit in stature as a 

 subject for this inquiry. Some further proof of this may be got by 

 comparing the variability of the system of mid-parents with that of 

 the general population. If the married couples had paired together 

 regardless of stature, their mean heights would be elements of a 

 statistical system identical with one in which the pairs had been 



