d that there is a value which occurs
most commonly, the average or medium value, from which the larger and
smaller deviations, the so-called plus and minus variations fall away in
a continuous series and end in a limiting value. In the simpler cases
a falling off occurs equally on both sides of the curve; the curve
constructed from such data agrees very closely with the Gaussian curve
of error. In more complicated cases irregular curves of different kinds
are obtained which may be calculated on certain suppositions.

The regular fluctuations about a mean according to the rule of
probability is often attributed to some law underlying variability. (de
Vries, "Mutationstheorie", Vol. I. page 35, Leipzig, 1901.) But there is
no such law which compels a plant to vary in a particular manner. Every
experimental investigation shows, as we have already remarked, that
the fluctuation of characters depends on fluctuation in the external
factors. The applicability of the method of probability follows from
the fact that the numerous individuals of a species are influenced by
a limited number of variable conditions. (Klebs, "Willkurl. Ent." Jena,
1903, page 141.) As each of these conditions includes within certain
limits all possible values and exhibits all possible combinations, it
follows that, according to the rules of probability, there must be
a mean value, about which the larger and smaller deviations are
distributed. Any character will be found to have the mean value which
corresponds with that combination of determining factors which occurs
most frequently. Deviations towards plus and minus values will be
correspondingly produced by rarer conditions.

A conclusion of fundamental importance may be drawn from this
conception, which is, to a certain extent, supported by experimental
investigation. (Klebs, "Studien uber Variation", "Arch. fur Entw."
1907.) There is no normal curve for a particular CHARACTER, there is
only a curve for the varying combinations of conditions occurring in
nature or under cultivation. Under other conditions entirely different
curves may be obtained with other variants as a mean value. If, for
example, under ordinary conditions the number 10 is the most frequent
variant for the stamens of Sedum spectabile, in special circumstances
(red light) this is replaced by the number 5. The more accurately we
know the conditions for a particular form or number, and are able to
reproduce it by experiment, the nea