I have been frequently and persistently credited with maintaining that the germ-plasm is invariable--a misunderstanding of my position, due perhaps to a somewhat too brief and terse statement which I made at an earlier period (1886). I had described the germ-plasm as 'a substance of great power of persistence,' and as varying with difficulty and slowly, basing this statement upon the age-long persistence of many species in which the specific constitution of the germ-plasm must have remained unchanged. The idea of 'germinal selection,' of a ceaseless struggle between the 'primary constituents' of the germ, and of the resulting continual slight and invisible rising and falling of individual characters, had not yet dawned upon me, nor had I at that time formulated the conception of 'determinants.' I was even doubtful at that time whether development, heredity, and variation were not interpretable on the assumption of an undifferentiated substance without primary constituents. But at no time was I unaware that the whole phyletic evolution of the organic world is only conceivable on the assumption of continual variation of the germ-plasm, that it actually depends upon this, even if these variations come about with exceeding slowness, and are thus in a certain sense 'difficult.' Now that I understand these processes more clearly, my opinion is that the roots of all heritable variation lie in the germ-plasm, and furthermore, that the determinants are continually oscillating hither and thither in response to very minute nutritive changes, and are readily compelled to _variation in a definite direction_, which may ultimately lead to considerable variations in the structure of the species, if they are favoured by personal selection, or at least if they are not suppressed by it as prejudicial. But selection is continually keeping watch over both kinds of variation, and if the conditions of life do not further the variation or do not even allow it to persist, selection eliminates everything that lessens the purity of the specific type, everything that transgresses the limits of utility, or that might endanger the existence of the species. Thus we understand how the germ-plasm may be variable, and yet at the same time remain unvaried for thousands of years, how it is ready and able to furnish any variation that is possible in a species if that is required by external circumstances, and yet is able to preserve the characters of the species in almost absolute constancy through whole geological ages; in short, how it can be at once readily variable and yet slow to vary.