Thus we may perhaps most easily comprehend that to which a thing
is related, when a name does not exist, if, from that which has a
name, we derive a new name, and apply it to that with which the
first is reciprocally connected, as in the aforesaid instances, when
we derived the word 'winged' from 'wing' and from 'rudder'.

  All relatives, then, if properly defined, have a correlative. I
add this condition because, if that to which they are related is
stated as haphazard and not accurately, the two are not found to be
interdependent. Let me state what I mean more clearly. Even in the
case of acknowledged correlatives, and where names exist for each,
there will be no interdependence if one of the two is denoted, not
by that name which expresses the correlative notion, but by one of
irrelevant significance. The term 'slave,' if defined as related,
not to a master, but to a man, or a biped, or anything of that sort,
is not reciprocally connected with that in relation to which it is
defined, for the statement is not exact. Further, if one thing is said
to be correlative with another, and the terminology used is correct,
then, though all irrelevant attributes should be removed, and only
that one attribute left in virtue of which it was correctly stated
to be correlative with that other, the stated correlation will still
exist. If the correlative of 'the slave' is said to be 'the master',
then, though all irrelevant attributes of the said 'master', such as
'biped', 'receptive of knowledge', 'human', should be removed, and the
attribute 'master' alone left, the stated correlation existing between
him and the slave will remain the same, for it is of a master that a
slave is said to be the slave. On the other hand, if, of two
correlatives, one is not correctly termed, then, when all other
attributes are removed and that alone is left in virtue of which it
was stated to be correlative, the stated correlation will be found
to have disappeared.

  For suppose the correlative of 'the slave' should be said to be 'the
man', or the correlative of 'the wing"the bird'; if the attribute
'master' be withdrawn from' the man', the correlation between 'the
man' and 'the slave' will cease to exist, for if the man is not a
master, the slave is not a slave. Similarly, if the attribute 'winged'
be withdrawn from 'the bird', 'the wing' will no longer be relative;
for if the so-called correlative is not winged, it follows that 'the
wing' has no correlative.

  Thus it is essential that the correlated terms should be exactly
designated; if there is a name existing, the statement will be easy;
if not, it is doubtless our duty to construct names. When the
terminology is thus correct, it is evident that all correlatives are
interdependent.

  Correlatives are thought to come into existence simultaneously. This
is for the most part true, as in the case of the double and the
half. The existence of the half necessitates the existence of that
of which it is a half. Similarly the existence of a master
necessitates the existence of a slave, and that of a slave implies
that of a master; these are merely instances of a general rule.
Moreover, they cancel one another; for if there is no double it
follows that there is no half, and vice versa; this rule also
applies to all such correlatives. Yet it does not appear to be true in
all cases that correlatives come into existence simultaneously. The
object of knowledge would appear to exist before knowledge itself, for
it is usually the case that we acquire knowledge of objects already
existing; it would be difficult, if not impossible, to find a branch
of knowledge the beginning of the existence of which was
contemporaneous with that of its object.

  Again, while the object of knowledge, if it ceases to exist, cancels
at the same time the knowledge which was its correlative, the converse
of this is not true. It is true that if the object of knowledge does
not exist there can be no knowledge: for there will no longer be
anything to know. Yet it is equally true that, if knowledge of a
certain object does not exist, the object may nevertheless quite
well exist. Thus, in the case of the squaring of the circle, if indeed
that process is an object of knowledge, though it itself exists as
an object of knowledge, yet the knowledge of it has not yet come
into existence. Again, if all animals ceased to exist, there would
be no knowledge, but there might yet be many objects of knowledge.

  This is likewise the case with regard to perception: for the
object of perception is, it appears, prior to the act of perception.
If the perceptible is annihilated, perception also will cease to
exist; but the annihilation of perception does not cancel the
existence of the perceptible. For perception implies a body
perceived and a body in which perception takes place. Now if that
which is perceptible is annihilated, it follows that the body is
annihilated, for the body is a perceptible thing; and if the body does
not exist, it follows that perception also ceases to exist. Thus the
annihilation of the perceptible involves that of perception.

  But the annihilation of perception does not involve that of the
perceptible. For if the animal is annihilated, it follows that
perception also is annihilated, but perceptibles such as body, heat,
sweetness, bitterness, and so on, will remain.

  Again, perception is generated at the same time as the perceiving
subject, for it comes into existence at the same time as the animal.
But the perceptible surely exists before perception; for fire and
water and such elements, out of which the animal is itself composed,
exist before the animal is an animal at all, and before perception.
Thus it would seem that the perceptible exists before perception.

  It may be questioned whether it is true that no substance is