74 CROSSING. 



roof-rats. The hybrids had grey -bellies, but produced some, 25% , 

 white-bellied off-spring. When a few yellow rats had been produ- 

 ced, we took over the work. Most of the rats died in the first 

 attempt to pass them through the French custom office, and we 

 finally received only a few white-bellied rats, of which only two 

 ever bred. One female was successfully mated to a French 

 black Mus rattus male. The hybrids were black, of a colour 

 somewhat different from that of the pure French house-rats. 

 From these hybrids, bred together and bred to one of Bonhote's 

 males, we obtained several new colours.We obtained some yel- 

 lows, as was to be expected from Bonhote's work, but we also 

 obtained a cinnamon agouti, and three chocolates, several rats 

 of a bright silver colour, like that of a lilac mouse, rats with 

 a bright yellow belly and an agouti coat, animals with a white 

 tip to the tail, and waltzers. 



If we add together the young from two Fl sisters, which 

 gave waltzers, we find that they had thirty-eight young. 

 Among these there were as yet no chocolates, and none with 

 yellow bellies ; these came in later generations. But among the 

 thirty-eight there were four yellows, four waltzers, one silver, 

 and two with white tail- tips. These numbers show that this was 

 not a case of mutation, but of the production of new charact- 

 ers through recombination. If in the production of every new 

 recessive character, we would have been dealing with a case 

 of mutation, we would have expected a proportion of one in 

 four. If, however, we look to the cross itself for the production 

 of the new varietal characters, we can see how each novelty 

 may have originated where two genes, each present in one 

 species, were both absent from a zygote. In that case we would 

 expect the novel characters to be present in one animal among 

 sixteen. Now the frequency of the different new characters is 

 differently great, but if we calculate the average frequency, we 

 get a proportion of 2.7 to 38, which is sufficiently close to one 

 in sixteen to be significant. If we remember that chocolates 

 and yellow-bellied rats cropped out in the next two gener- 

 ations, it becomes clear that each of the five new recessive char- 



