is in need of new data from which an hypothesis
can be formed.

It is true that the actual working elements with which the scientist
operates are always complex in the sense that they are always something
more than elements in any specific investigation. They have other
connections and alliances. And this complexity is at once the despair
and the hope of the scientist; his despair, because he cannot be sure
when these other connections will interfere with the allegiance of his
elements to his particular undertaking; his hope, because when these
alliances are revealed they often make the elements more efficient or
exhibit capacities which will make them elements in some other
undertaking for which elements have not been found. A general resolves
his army into so many marching, eating, shooting units; but these
elements are something more than marching, shooting units. They are
husbands and fathers, brothers and lovers, protestants and catholics,
artists and artisans, etc. And the militarist can never be sure at what
point these other activities--I do not say merely external
relationships--may upset his calculations. If he could find units whose
whole and sole nature is to march and shoot, his problem would be, in
some respects, simpler, though in others more complex. As it is, he is
constantly required to ask how far these other functions will support
and at what point they will rebel at the marching and shooting.

Such, in principle, is the situation in every scientific inquiry. When
the failure of the old elements occurs it is common to say that
"simpler" elements are needed. And doubtless in his perplexity the
scientist may long for elements which have no entangling alliances,
whose sole nature and character is to be elements. But what in fact he
actually seeks in every specific investigation are elements whose nature
and functions _will not interfere_ with their serving as units in the
enterprise in hand. But from some other standpoint these new elements
may be vastly more complex than the old, as is the case with the modern
as compared with the ancient atom. When the elements are secured which
operate successfully, the non-interfering connections can be ignored and
the elements can be treated as if they did not have them,--as if they
were metaphysically simple. But there is no criterion for metaphysical
simplicity except operative simplicity. To be simple is to serve as an
element, and to serve as an element is to