500 Miscellanea 



tixity of a litio of ancestry is asscrtod whcii tlic ^iiicestral imrity is far less tlian that involved 

 in Professor Johannsen's "reine Linie," wliicli he defines as consisting of " Individuen, welclie 

 " von einem einzelnen selbst befruchtenden Individuum abstammen " (p. 9). 



Again, "with a vicw to reducing the absolute variability of a species it is idle to select beyond 

 " grandparents, and hardly protitiible to select beyond parents. The ratio <>f the variability of 

 "pedigree stock to the gencral i)oi)ulation dccreases 10 per cent. on the selection of parents, and 

 "only 11 i>er cent. on the additional selection of grandimrents. Beyond this no sensible change 

 "is niade."* 



In these two passagcs the tixity of the type and the higli variability of individuals abont the 

 type are asserted as absolutely ;vs they are iisserted by de Vries in the passagc of the Mntatioiis- 

 theorie quoted by Professor Johannsen. With cur present knowledge of the coefficients of inherit- 

 ance in man, horse, and dog, it would seem that from 2 to 4 generations of selection suffice to 

 form a line varying greatly about it-s type, yct rcmaining true to that type. AVhen we are told 

 that a bciin breeds true to its line we are told something which has beon shown to l« a nece.-5.s;iry 

 consequence of the law of ance.stral heredity ; if it were not true, the whole law woiUd be upset. 



The difFerence between the view put forward by Professor Johannsen, and that expressed in 

 1898 in the papor "On the Law of Anc&stral Inheritance" is therefore not a difFerence which 

 concerns the focus of regression in the oft'spring of selceted ancestors ; it is .simply a difJ'erence 

 as to the relatiou between successive generations of individuals within the line of ancestry. 

 Professor Johannsen believes the tables on pp. 21 — 24 of his work to show that "die persönliche 

 '■'■Beschaffenheit der Eltern, Grosseltern, oder irgend eines Aknen Aat— soweit meine Erfahmng 

 "reicht — keinen Einflms auf den diircLichiiittlichen Charakter der Nachkommen. Es ist aber 



"der Typus der ZjwiV, welche den durchschnittlichen Character der Individuen bestimmt " 



(pp. 61—62). Within the same line of ancestry, whatever individuals of a generation be chosen 

 as parents, the character of the resultant tilial generation will be the .same according to Profes.sor 

 Johannsen, or the coefficients of correlation and regres.sion between parents and oifspring, within 

 the .same line of ance.stry, will each = 0. 



The cxperimental results do not seem to us cousistent with this view. If the oä'spring of 

 every generation, within a given line of ancestry, breed tnie to the type of their line, subject to 

 such se;i.sonal and climatic influenccs as aft'ect the whole generation of their year, tlien when the 

 whole filial generation is compared with the whole parenhd generation, correlation l>et\veon the 

 two must nece.s.sari!y be perfect, and the coefficient of regression must be simply the ratio 

 between the stixndard deviations of the two generations. Professor Johannsen should, we thiuk, 

 first have shown that jierfect correlation does in fact exist between parents^ and oflspring in two 

 successive generations of plants ; and this he has failed to do ; in the c;ise of his maternal and 

 filial generations he has, however, published data which enable us to determine the required 

 correlation, and the txblc below givcs the result. Witli such data, the only method available is 

 the first method uscd for Shirley l'oppicst. It consi.sts in detcrmining the coiTclation between 

 evei-y individual bean of the filial generation, and the ynean character of its parent. Tlie absolute 

 value of the correlation so obt<xined will not be signifioint, but the coefficient of regression will 

 closely represent the true parentjvl correlation, — being a relation between nican filial character 

 and mean parcntal character. Taking Professor Johaiuisen's Vebersichtstabelle I. (p. 25), and 

 the maternal mcans given on pp. 21 — 24, the following table has been constructed, giving a 

 coefficient of coiTelation = 0'531 and of rcgres.siün = 0-591 +012.'). Tliis latter value represents 

 the coefficient of con-elation, .so far as the data allow it to be detcrmined, between filial and 

 maternal plants; and cousidering the paucity of maternal plants (only 19) the result is not in 

 bad accord with previous results for jiarontal correlation}:. The value 0591 +0'125 is not very 

 divergent from 05, but it cannot be hcld to appro.Kiniate to unity ! 



* Pearf^on : lac. eil. t Bionutrika, Vol. ii. Part i. p. (il). 



t Biovietrika, Vol. ii. p. 379. 



