196 H. S. JENNINGS AND K. S. LASHLEY 



If the progeny show a hkeness in size not greater than that of 

 the, parents, then of course there is no indication of biparental 

 inheritance — such similarity as exists being fully accounted for 

 by the assortative mating between the parents. If there is actual 

 inheritance from both parents, the coefficient of correlation between 

 the progeny will he greater than that between the parents. 



The parents at the time of conjugation are adults, of a medium 

 size; they do not include specimens that are small because young, 

 nor specimens that are large in preparation for fission; this is fully 

 set forth in the paper by the senior author ('11). The progeny 

 of these conjugants, on the other hand, include, as we find them, 

 young and old. If, therefore, we take a single specimen of each 

 line at random, we shall often obtain a young small specimen of 

 a, an old large specimen of b ; the similarity in size that may actu- 

 ally exist between adults of average size being thus quite lost. 

 Such random selection would, therefore, give a much reduced 

 coefficient of correlation, as compared with that between adults 

 of average size. 



How can this difficulty be met? Evidently, the proper method 

 of procedure is to obtain by measurement of a considerable num- 

 ber of specimens the approximate mean for each line; then to 

 measure the correlation between these means, for the two lines 

 (a and h) of each pair. 



This is what we have done for the 86 lines with which the pres- 

 ent paper deals. The means for each line, together with the 

 number of specimens measured, are given in table 3. 



The coefficient of correlation between the means of the lines 

 belonging to a single pair was computed from this table, employ- 

 ing each mean to the second decimal place. The coefficient in 

 such a case is best computed without the formation of a table. 

 It is to be remembered that we are in the present case dealing with 

 like variates as members of the pairs; hence the methods employed 

 with symmetrical tables are to be used. The coefficient was com- 

 puted both by the product method and by the difference method. 

 These give the same result, a coefficient of 0.5703. Making use 

 of Shepard's correction, in connection with the product method. 



