roducing past elements of the manifold, and, if knowledge is to
arise, of reproducing them according to rules. This faculty of
reproduction by which, on perceiving the element A, we are led to
think of or reproduce a past element B--B being reproduced according
to some rule--rather than C or D is called the faculty of association;
and since the rules according to which it works depend on empirical
conditions, and therefore cannot be anticipated _a priori_, it may be
called the subjective ground of reproduction.'

'But if the image produced by association is to play a part in
knowledge, the empirical faculty of reproduction is not a sufficient
condition or ground of it. A further condition is implied, which may
be called objective in the sense that it is _a priori_ and prior to
all empirical laws of imagination. This condition is that the act by
which the data of sense enter the mind or are apprehended, i. e. the
act by which the imagination _apprehends and combines_ the data of
sense into a sensuous image, must _make_ the elements such that they
have affinity, and therefore such that they can subsequently be
recognized as parts of a necessarily related whole.[17] Unless this
condition is satisfied, even if we possessed the faculty of
association, our experience would be a chaos of disconnected elements,
and we could not be self-conscious, which is impossible. Starting,
therefore, with the principle that we must be capable of being
self-conscious with respect to all the elements of the manifold, we
can lay down _a priori_ that this condition is a fact.'

[17] On this interpretation 'entering the mind' or 'being
apprehended' in the fourth paragraph does not refer merely to
the apprehension of elements one by one, which is preliminary
to the act of combining them, but includes the act by which
they are combined. If so, Kant's argument formally involves a
circle. For in the second and third paragraphs he argues that
the synthesis of perceptions involves reproduction according
to rules, and then, in the fourth paragraph, he argues that
this reproduction presupposes a synthesis of perceptions. We
may, however, perhaps regard his argument as being in
substance that knowledge involves _re_production by the
imagination of elements capable of connexion, and that this
reproduction involves _pro_duction by the imagination of the
data of sense, which are to be reproduced, int