9

Logic II 9

if one were to [form?] discriminte the classes by a sort of weighted mean of all the
observable characters it would hit right; and even if one
were only to do this by estimation, we should probably do better
than by risking all upon a single definition. Another singular
error which naturalists make quite systematically is to
think that the intermediate forms connect two types almost
continuously (what they call quite continuously, but that is inaccurate
language) those two classes cannot be "founded in
nature" but must be purely artificial. There is not only no logical
foundation for such an opinion, but it so happens that
there is an instance ready at my hand to prove its falsity. Namely
if there is any cause in nature tending to produce forms of a certain
type, subject to accidental variations, and if from a great number of
such forms subject to the same general conditions we plot a
curve whose abscissa shall measure any variable character
of that type while the ordinates are proportional to the number of
instances of each quantity of the variable measured by the abscissa,