ON ATAVISTIC VARIATION IN OENOTHERA CRUCIATA. 335 



petals seem to be two alternating types only united by rare inter- 

 mediate steps. 



In case my variety originated by a mutation, we would there- 

 fore have an imperfect one, producing the new type only in part of 

 the individuals and remaining true or returning to the pure form of 

 Oenothera cruciata in the others. In many points this case would 

 be analogous to that of Oenothera scintillans, which originated in 

 my garden from 0. Lamarckiana, but which, though artificially 

 pollinated with its own pollen, returns in each generation in often 

 a relatively large part of the individuals to the original type. 



In case my variety originated by a cross, it should first be re- 

 marked that the supposed hybrid is not intermediate between 0. 

 cruciata and some allied species, but that it has, as far as I can 

 judge, all the characters of the former, the constancy in the form 

 of the petals excepted. I have endeavored to show in my "Mu- 

 tations-theorie" (Part II, p. 100) that 0. muricata is probably the 

 only species which could have given such a hybrid with the 0. 

 cruciata. Now it is clear that having once obtained seeds from the 

 original and constant 0. cruciata, I possess the materials to bring 

 about this crossing and also those with other allied species, and to 

 try whether it will be possible to get an inconstant hybrid in this 

 way. I propose to do so, but from what I know of my 0. cruciata 

 varia and of the hybrids I have made of it with more than one other 

 species, it is very improbable that the hybrid 0. cruciata x 0. muri- 

 cata will be as a rule inconstant. It is far more probable that it will 

 exhibit the dominant character, which must be the broad petals, 

 and either give a constant progeny with this feature or split up 

 according to Mendel's laws. I take the broad petals to be phylo- 

 genetically older than the linear ones, because they are the rule in 

 the whole genus. But some authors suppose that from time to time 

 there arise exceptions to the ordinary laws governing hybrids, and 

 that such exceptions might be the source of many inconstant varieties 

 and even of inconstancy in general. In this case the chance of getting 

 such an exception by crossing would scarcely be greater than that 

 of getting a mutation from the pure type. 



However slight these chances may be, I propose to try both ways 

 and to see whether it would be possible to repeat the as yet doubtful 

 and unknown original production of my 0. cruciata varia. For there 

 are not a great many cases in which the starting-points are to be 

 had in so pure and so thoroughly controlled a state as in this instance. 

 As to the question of the rule of the constancy of 0. cruciata under 



