Psychology_, vol. ii. p. 383.

The fact that impulse is not always the result of conscious foresight
is most clearly seen in the case of children. The first impulses of a
baby to suck, or to grasp, are obviously 'instinctive.' But even when
the unconscious or unremembered condition of infancy has been succeeded
by the connected consciousness of childhood, the child will fly to his
mother and hide his face in her skirts when he sees a harmless stranger.
Later on he will torture small beasts and run away from big beasts, or
steal fruit, or climb trees, though no one has suggested such actions to
him, and though he may expect disagreeable results from them.

We generally think of 'instinct' as consisting of a number of such
separate tendencies, each towards some distinct act or series of acts.
But there is no reason to suppose that the whole body of inherited
impulse even among non-human animals has ever been divisible in that
way. The evolutionary history of impulse must have been very
complicated. An impulse which survived because it produced one result
may have persisted with modifications because it produced another
result; and side by side with impulses towards specific acts we can
detect in all animals vague and generalised tendencies, often
overlapping and contradictory, like curiosity and shyness, sympathy and
cruelty, imitation and restless activity. It is possible, therefore, to
avoid the ingenious dilemma by which Mr. Balfour argues that we must
either demonstrate that the desire, _e.g._ for scientific truth, is
lineally descended from some one of the specific instincts which teach
us 'to fight, to eat, and to bring up children,' or must admit the
supernatural authority of the Shorter Catechism.[5]

[5] _Reflections suggested by the New Theory of Matter_, 1904, p. 21.