NEW SPECIES 233 



drawn from this : first, that the double numbers owe their origin 

 chiefly to polyploidy ; secondly, that change in the basic number 

 is, relative to change by doubling, somewhat rare ; and thirdly, 

 that no differential elimination of forms with certain numbers has 

 occurred, as it might, owing to some numbers being more suitable 

 mechanically than others. These conclusions are already probable 

 on experimental grounds, for polyploidy is the most frequent of all 

 changes observed in plant species (cf. Crepis, Table 8). 



\\Tien the species of smaller systematic groups are considered, 

 differences are found both in the amount and in the kind of variation 

 in chromosome number. Some large groups are found to show no 

 variation (Ch. III). Others show the influence of the polyploid 

 series in various degrees. The graph (Fig. 64) of variation in 

 number in Crepis (Babcock and Navashin, 1930), and Crocus (Mather, 

 1932) illustrates types where polyploidy has played little part. 

 Crepis is more constant than Crocus, for it has a high proportion 

 with the number four which must be considered the original number 

 of the genus. In Crocus the distribution of numbers is almost at 

 random. How far the changes responsible are merely numerical, 

 i.e., due to changes in the numbers of certain chromosomes, and 

 how far they are structural, i.e., due to fragmentation and fusion, 

 can only be said after a detailed study of meiosis in hybrids. But 

 there is clearly a similar variety of species in the two genera with 

 less variation in chromosome number in Crepis than in Crocus, 

 This is probably due to the greater importance of sexual reproduction 

 (and therefore of regularity at meiosis) in the one than in the other. 



Crepis and Crocus approach the distribution of numbers found in 

 animals where there is no evidence of polyploidy. The opposite 

 extreme is found in genera such as Chrysanthemum, Triticum, Rosa, 

 Prunus, Rubus, Solanum and Aconitum, where all species have 

 multiples of a common basic number (Fig. 79). 



(ii) New Polyploid Species. Clearly it can be possible to recon- 

 struct the method of origin of a species only when it has arisen 

 recently. Suitable conditions could probably be found to allow 

 of several new polyploids that have arisen in experiment becoming 

 wild species. But amongst wild species there are a few the con- 

 ditions of whose origin can be stated with precision. First there 



