lSy6.] BOTANICAL LETTERS. 343 



vary, and no experiments as yet tried have thrown any light 

 on the subject. Now for the last ten years I have been 

 experimenting in crossing and self-fertilising plants ; and one 

 indirect result has surprised me much ; namely, that by taking 

 pains to cultivate plants in pots under glass during several 

 successive generations, under nearly similar conditions, and by 

 self-fertilising them in each generation, the colour of the 

 flowers often changes, and, what is very remarkable, they 

 became in some of the most variable species, such as Mimulus, 

 Carnation, &c, quite constant, like those of a wild species. 



This fact and several others have led me to the suspicion 

 that the cause of variation must be in different substances 

 absorbed from the soil by these plants when their powers of 

 absorption are not interfered with by other plants with which 

 they grow mingled in a state of nature. Therefore my son 

 and I wish to grow plants in pots in soil entirely, or as nearly 

 entirely as is possible, destitute of all matter which plants 

 absorb, and then to give during several successive generations 

 to several plants of the same species as different solutions as 

 may be compatible with their life and health. And now, can 

 you advise me how to make soil approximately free of all the 

 substances which plants naturally absorb ? I suppose white 

 silver sand, sold for cleaning harness, &c, is nearly pure silica, 

 but what am I to do for alumina ? Without some alumina I 

 imagine that it would be impossible to keep the soil damp 

 and fit for the growth of plants. I presume that clay washed 

 over and over again in water would still yield mineral matter 

 to the carbonic acid secreted by the roots. I should want a 

 good deal of soil, for it would be useless to experimentise 

 unless we could fill from twenty to thirty moderately sized 

 flower-pots every year. Can you suggest any plan ? for unless 

 you can it would, I fear, be useless for us to commence an 

 attempt to discover whether variability depends at all on 

 matter absorbed from the soil. After obtaining the requisite 

 kind of soil, my notion is to water one set of plants with 



