( 648 ) 
experiment (3) produces, more readily than the rose variant (2), as 
well red normal forms (J) as white ones (4). For the rest, this 
more variable variant has also proved to remain constant when 
quickly transplanted. 
Gases of atavism are frequently observed in these experiments. 
Thus, for example, the production of the normal form from viscosus 
(6) may easily be seen if the latter grows for a fortnight without 
transport on a bouillonagarplate; along the margin of the streaks 
some few normal colonies (1) will then become perceptible. 
The albus-v&ria,nts, also have a disposition to throw off a few 
red normal forms, but they do so only after growing for weeks or 
months on bouillon-agar; at first they are very constant. 
The to a certain extent completely regular production of the same 
variants of Bacillus prodigiosus, suggests the existence of variability 
in a special and determined direction, of orthogenesis, as Eimer 
expressed it. 
As under different nutritive conditions the same variant may 
appear, the food itself cannot be the stimulus; there must be, as 
said above, another cause in the interior of the cells, which, for B. 
prodigiosus , seems only active in an alkaline environment. 
On the other hand, the food, in a wider sense, has certainly a 
decisive influence on the variability, albeit indirectly. So we considered 
already the influence of the alkaline reaction of the medium if 
this alkali is produced by the microbes themselves. Another example 
is the following. As well in malt-wort as in bouillon the viscosus 
variant is regularly produced; but from malt-wort the auratus 
variant, which so readily takes rise in bouillon, is not obtained 
at all. Indeed, every culture condition gives a peculiar but con¬ 
stantly returning mixture of variants, differing both quantitatively 
and qualitatively from that found under any other conditions. But 
the real factors here active could not as yet be detected. 
From the foregoing the following results may be derived. 
1. Bacillus prodigiosus produces as well qualitative, as gain- and 
loss-variants, all obtained with certainty by determined experiments; 
the stock-form is always found unchanged in the same culture with 
the variants. 
£!* Ihe variants are from their origin as constant as their stock. 
The true factors which govern the variability in these experiments 
are still unknown 
2. By rapidly repeated re-inoculations and by other methods, nor- 
