--not in gross
wholes. They would express each of the "elemental" powers in the plural
and not in the singular.

The basis of this change of view within the last fifteen or twenty years
is to be found in experimental work. The question has definitely been
put to the test as to how far training in one line did influence others.
For a full description of the various types of experiments performed the
reader is referred to Thorndike's "Psychology of Learning," Chapter 12.
Only an indication of the type of work done and the general character of
the results can be given here. Experiments in the effect of cross
education, in memorizing, in observing and judging sensory and
perceptual data, and in forming sensori-motor association habits have
been conducted in considerable numbers. A few experiments in special
school functions have also been carried out. Investigations in the
correlation between various parts of the same subject and between
different subjects supposed to be closely allied also throw light upon
this subject. The results from these different lines of experiment,
although confusing and sometimes contradictory, seem to warrant the
belief stated above. They have made it very clear that the question of
transfer is not a simple one, but, on the contrary, that it is extremely
complex. They make plain that in some cases where large transfer was
confidently expected, that little resulted, while, on the other hand, in
some cases when little was expected, much more occurred. It is evident
that the old idea of a large transfer in some subtle and unexplained way
of special improvements to a general faculty is false. But, on the other
hand, it would be equally false to say that no transfer occurred. The
general principle seems to be that transfer occurs when the same bonds
are used in the second situation to the extent that the alteration in
these particular connections affects the second response. Both the
knowledge of what bonds are used in various responses and to what extent
alteration in them will affect different total responses is lacking.
Therefore, all that is at present possible is a statement of conditions
under which transfer is probable.

In general, then, transfer of training will occur to the extent that the
two responses use the same bonds--to the extent, then, that there is
identity of some sort. This identity which makes transfer possible may
be of all degrees of generality and of several different types.