Miscellanea 



601 



TABLE A. 



Correlatioii between Individual Filial Jkavs und Meaii Maternal Beans. 



Filial Beans. 



The letters A to T are used by Professor Johannsen to denote bis niueteen "pure lines." 



If Professor Johannseu's hypothesis were true, the only way in wbioh be oould account for tbe 

 regression bere observed, — a regression wbose existence be himself admits, — would be by assuming 

 tbat the characters, by which he has described tbe plants of bis pure bnes, are imperfectly corre- 

 lated with the actual mcaii characters of bis plaiits which lio supi)oses to represent the types of 

 bis lines; but tbis assmnption, while leaving bis hypothesis logically unsbaken, would destroy 

 tbe wbole value of bis experiments as evideiice in its favour, by destroying the vahie of tbe 

 measure by whicb he bas determined tbe characters studied. 



Some of the difficulties we have feit in following Professor Johannsen are uudoubtedly due to 

 tbe imperfect way in whicb he has nieasured the characters of ancestral plants*. The grand- 

 maternal plants, for example, are determined each by the character of a Single bean ; tbe small 

 value of such a determination may be judged from the variability among the beans of a Single 

 plant produoed by the last generation; the mean nunilier of such beans was 84'ö, and the mean 

 Standard deviation of all arrays due each to a singlo plant was 75-37 luilligi'anis, so tbat tbe mean 

 character of a niother-plant, infcrred from a siugle bean chosen at random among the offspring, 

 would be equally likely to lie iuside or outside the limits (true maternal mean + 50 mgrm.) and 

 (true maternal mean -50 mgrm.). There is thus, we venture to say, no strong probability that 

 the numbers by whicb Profes.sor Johannsen describes bis grandparental beans represent the mean 

 character of the seeds of the corrcsponding plants within + 100 nigrni. 



Again, tbe wbole evideuce, that the coeflBcient of filial regression within tbe line is zero, rests 

 on the tables on pp. 21 — 24; but in these tables we are only told (1) the mean weight of seeds 



" A preliniinary study of homotypoRis in the bean must of nece.ssity precede any attempt to measure 

 plant character by a s-inyle seed. Professor Johannsen would have to show that the honiotypic corre- 

 lation was perfect to justify his measure. This is very far from the fact not ouly iu I'haseolus 

 vulyaris, but in all beans liitherto exaniined from this standpoiut. 



Biometrika 11 64 



