Mutation Theory 159-164 (1909)
Prof. Hugo de Vries

In the preceding discussion I have had occasion to draw attention not merely to the splendid achievements of my predecessors but also to the numerous gaps in our knowledge.

The study of variability as opposed to mutability is a branch of human knowledge which has developed with great rapidity in the last few years. The statistical method of dealing with this phenomenon is, as we have already said, firmly established: comparative and experimental methods are just coming on.

I propose, therefore, to suggest a series of problems the solution of which would in my opinion throw much light on the essential difference between mutability and variability.

1. More examples of QUETELET'S law are wanted: their number can never become too great.

2. The curves in question should be plotted from the same individuals or from the same batch of individuals in successive years. The constancy of their means and their amplitude (GALTON'S Q and Q') should be determined. Changes in these values, and changes in the symmetry of the curves, should if possible be traced to their causes.

3. Polymorphic curves should be looked for and analysed. These may point to the existence of mixtures of perfectly distinct elementary species growing together or to the existence of antagonistic characters within the limits of a single species (for examples annual and biennial forms in Daucus, Beta, etc.) They may also be due to diseases. Finally they may be the so‑called "double curves" in which the several apices are to be regarded as ordinates in a curve of a higher order, and not as indications of mutation.

1J. H. BURKILL, Variation in the Number of Stamens and Carpels, Journ. Linn. Soc. Bot., Vol. 31.
2DUNCKER, ROUX'S Archiv, Vol. VIII, p. 163.
3GALTON, Natural Inheritance, and Proceedings Royal Society, Vols. 40 and 45; and ED. VERSCHAFFELT, Correlatieve Variatie by planten. Botan. Jaarboek, VIII, p. 92.

4. Correlative variation is a phenomenon of the highest importance.1 For example man presents many instances of correlation between mental and physical characters. Correlations fall in two categories. In the one are those cases in which the two characters are dependent in the same way although not to the same degree on external conditions. In the other are those cases in which variation in one character is the cause of variation in another,2 as for example the various phenomena of growth which are correlated with differences in photosynthetic activity. It is superfluous to refer the reader to GALTON'S method of studying correlation.3

4Variability can also be influenced by grafting and inoculation. See L. DANIEL. Compt. Rend., 1894, T. CXVIII, p. 992.

5. The relation between external conditions of life and variability ought to be investigated. Are there variations which are independent of such, or are there not? If there are, what are their causes? Do the individual external factors exert a separate influence or not? Is there a definite relation between the extent of this influence and the magnitude of the variation? Do all characters under the influence of high nutrition vary in a plus direction, and under a poor one in a minus direction?4

6. The sensitive period in the development of characters should be determined. When the rudiments of organs are visible under the microscope it is usually too late to exert any restraining influence on their development. But there may be exceptions to this rule. During the time which a character takes to develop there is probably one short period of extreme susceptibility; and this may be gradually attained and gradually lost. Here is matter for much interesting inquiry.

7. GALTON'S regression is very important. Suppose we sow seeds of a self‑fertilizing plant: and suppose that we know the amount by which it deviates from the mean of its ancestors in respect of certain characters. Then we determine the curve describing the result of our sowing. As a general rule, the mean of any character in the filial generation departs less from the normal, than the character in question borne by the parent plant does. According to GALTON, the relation between these two deviations is a constant one: the mean deviation of the children amounts to about a third of that of their parents. The question whether this is a universal principle naturally suggests itself; the experiments which I have made hitherto seem to point to the conclusion that it probably is.

8. Does this regression remain the same even when selection is continued for several generations? In other words, does the mean of a race never amount to more than a third of the value attained by the seed bearers chosen in every generation? Does the race in spite of its improvement persist in this relation to its progenitors, that is to say, does it lag at every generation relatively further behind the selected individuals which produced it? It seems to do; at any rate the decision of this point dominates the theory of the origin of species by the natural selection of individual variations.

5See DUNCKER, Biolog. Centralbl., 1898, p. 571. For each additional woo individuals the range of variation only increases as from 1 to 1.049.

9. QUETELET'S law enables us to calculate from a curve of variation the number of individuals that will exhibit a desired degree of deviation from the mean. It seems that this chance even in the case of small differences is a very remote one demanding as it does millions of individuals. At any rate it is desirable to make such calculations for as many cases as possible.5

10. Artificial selection is a device for reaching a certain magnitude of deviation from the average, with a minimum expenditure of trouble. Is this its only significance? Does the number of individuals with the undesired qualities diminish exactly at such a rate as we can calculate beforehand? That is to say, is regression independent of the ancestry of a given parent; in other words, does it make any difference whether the seed parent is the result of repeated selection, or is picked from a single sowing on a much larger scale?

11. In such experiments attention should be paid to one character and one only; although interesting results may often be obtained by measuring a second or even a third character as a sort of collateral inquiry. The selections carried out by breeders involve as many characters as possible; on account of correlations the improvement of the chief features can be carried further in this way, than would otherwise be possible. Such experiments should he made with a purely scientific end in view.

12. In starting an experiment attention must always be paid to the individual vigor of the seed‑parents. If this does not happen to coincide with the desired deviation, it is advisable to take both the strongest individuals and those exhibiting the greatest deviation, as seedbearers and to compare the posterity of the two.

13. There is a particular kind of selection experiment which should be carried out a great deal more than it is. I mean one which would start by choosing as seedparents plants with the smallest petals, the smallest fruits, those with the least degree of hoariness or the least number of spines, with the palest color in their petals, with the smallest number of stamens and carpels and so forth. According to the theory of natural selection such an experiment should result in the origin of apetalous, fruitless, glabrous, spineless, white‑flowered, unisexual or sterile plants and so forth. Whereas of course on the mutation theory this would not happen; provided that crossing was rigidly excluded from the experiment.

14. What we must aim at is a complete control of variation. We must become so thoroughly acquainted with the underlying factors that we can predict the results of our experiments.