ch are exactly the same as those
he has had before. These methods will be better understood from the
account now to be given of the way they were carried out on a large
number of students.

[Footnote 7: Prof. H. C. Warren, Mr. W. J. Shaw, and the writer.]

The first experiments were made by Messrs. S. and B. in the University
of Toronto on a class of students numbering nearly three hundred, of
whom about one third were women. The instructors showed to the class
certain squares of cardboard of suitable size, and asked them to do
the following three things on different days: First, to reproduce from
memory, with pencil on paper, squares of the same size as those shown,
after intervals of one, ten, twenty, and forty minutes (this gives
results by the method of Reproduction); second, to say whether a new
set of squares, which were shown to them after the same intervals,
were the same in size as those which they had originally seen,
smaller, or larger (illustrating the method of Identification); third,
they were shown a number of squares of slightly different sizes, again
at the same intervals, and asked to select from them the ones which
they found to be the same size as those originally seen (method of
Selection).

The results from all these experiments were combined with those of
another series, secured from a large class of Princeton students; and
the figure (Fig. 8) shows by curves something of the result. The
figure is given in order that the reader may understand by its
explanation the "graphic method" of plotting statistical results,
which, with various complications, is now employed in psychology as
well as in the other positive sciences.

[Illustration: FIG. 8.--Memory curves: I. Method of Selection. II.
Method of Identification.]

Briefly described in words, it was found that the three methods agreed
(the curves are parallel)[8] in showing that during the first ten
minutes there was a great falling off in the accuracy of memory (slant
in the curves from 0 to 10); that then, between ten and twenty
minutes, memory remained relatively faithful (the curves are nearly
level from 10 to 20), and that a rapid falling off in accuracy
occurred after twenty minutes (shown by the slant in the lines from 20
to 40).

[Footnote 8: This figure shows curves for two of the methods only,
Selection and Identification.]

Further, the different positions of the curves show certain things
when properly understood. The curve se