INTERPRET A TION 567 



this difference is due to a genetic difference and that therefore the 

 hybrid will fail to show it. But there is another restriction to this type of 

 inference, viz., the different effects of the same treatment on chromo- 

 somes of related species. Such different effects are probably responsible 

 for the differences in the size of chromosomes observed by Farmer and 

 Digby (19 1 4) in diploid and tetraploid Primula kewensis, suggesting 

 that the tetraploid was derived by transverse splitting of the chromo- 

 somes, and by Tischler (1928) between species of Rihes, suggesting that 

 autosyndesis occurred in the hybrid {cf. Meurman, 1928 ; D., 1929 d), 

 because large and small pairs of chromosomes were found there {cf. 

 Upcott, 1936 b on Msculus ; v. Berg, 1931 b on Mgilops). 



In meiosis, erroneous assumptions have been made and still are made 

 with regard to the pairing of chromosomes, such as the following : — 



(i) Pairing has been held to be determined by " affinity " between 

 chromosomes {cf. Dobzhansky, 193 1). It is therefore constant (under 

 constant external conditions), and a few observations are an accurate 

 measure of this " affinity." Chromosomes which do not pair are not 

 " homologous." In this way a simple rule is made for analysing the 

 relationships and even phylogeny of species. The conclusions derived 

 from these imperfect assumptions are of doubtful value {e.g., in Nico- 

 tiana). Even with quantitative analysis {e.g., in Triticum-Aigilops 

 hybrids) it is doubtful whether relationships can be accurately assessed 

 because {inter alia) the differences in the chromosomes which reduce their 

 pairing are not necessarily directly related to, and therefore proportional 

 to, the differences which determine genetic differences in the organism 

 (f. Ch. V). 



(ii) Ring-formation has been held (by adherents of the theory of 

 telosynapsis) to be determined by an undefined " hybridity," a reduced 

 " homology," a genetic factor, and so on. These inferences were loose 

 and unjustifiable and disagreed with genetic evidence. They have 

 been quietly buried (Ch. IX). 



Summing up : The mechanical and genetical interpretations put on 

 chromosome behaviour have suffered from two chief sources of error. 

 First, the relative values of direct and comparative inference have not 

 been correctly assessed. Direct inferences have been admitted as 

 " facts " and " observations " or denied as " artefacts " and " mis- 

 interpretations " when actually their whole weight depended upon their 

 use for comparative inference. Comparative inference, on the other 

 hand, has too often been put aside as speculative and irrelevant when 

 indeed it has been the only means for any general advance in the 

 understanding of the processes concerned. Hypothesis based on it has 

 often proved more reliable than the " facts " of direct observation. 



Secondly, synthesis has preceded complete analysis. " The method 

 pf discovery and proof according to which the most general principles 



