In the case of _Oenothera muricata x biennis_ the differentiating units reduce the fertility to a low degree, threatening the offspring with almost complete infertility and extinction. But then we do not know whether these characters are really units, or perhaps only seemingly so and are in reality composed of smaller entities which as yet we are not able to segregate. And as long as we are devoid of empirical means of deciding such questions, it seems useless to go farther into the details of the question of the sterility. It should be stated here however, that pure varietal crosses, when not accompanied by unbalanced characters, have never showed any tendency to diminished fertility. Hence there can be little doubt that the unpaired units are the cause of this decrease in reproductive power. The genus _Oenothera_ is to a large degree devoid of varietal characteristics, especially in the subgenus _Onagra_, to which _biennis_, _muricata_, _lamarckiana_ and some others belong. On the other hand it seems to be rich in elementary species, but an adequate study of [263] them has as yet not been made. Unfortunately many of the better systematists are in the habit of throwing all these interesting forms together, and of omitting their descriptive study. I have made a large number of crosses between such undescribed types and as a rule got constant hybrid races. Only one or two exceptions could be quoted, as for instance the _Oenothera brevistylis_, which in its crosses always behaves as a pure retrogressive variety. Instead of giving an exhaustive survey of hybrids, I simply cite my crosses between _lamarckiana_ and _biennis_, as having nearly the aspect of the last named species, and remaining true to this in the second generation without any sign of reversion or of splitting. I have crossed another elementary species, the _Oenothera hirtella_ with some of my new and with some older Linnean species, and got several constant hybrid races. Among these the offspring of a cross between _muricata_ and _hirtella_ is still in cultivation. The cross was made in the summer of 1897 and last year (1903) I grew the fourth generation of the hybrids. These had the characters of the _muricata_ in their narrow leaves, but the elongated spikes and relatively large flowers of the _hirtella_ parent, and remained true to this type, showing only slight fluctuations and never reverting or segregating [264] the mixed characters. Both parents bear large capsules with an abundance of seed, but in the hybrids the capsules remain narrow and weak, ripening not more than one-tenth the usual quantity of seed. Both parents are easily cultivated in annual generations and the same holds good for the hybrid. But whereas the hybrid of muricata and biennis is a stout plant, this type is weak with badly developed foliage, and very long strict spikes. Perhaps it was not able to withstand the bad weather of the last few years.