A Transcendental Argument by Strawson (Andrew Pickin)
In this essay I analyse and evaluate the first of two transcendental arguments against
forms of scepticism given by Strawson in his Individuals.
The argument
Strawson (1959) argues against the Hume-like position that all we have, in the case of
non-continuous observation, is different kinds of qualitative identity, and never numerical
identity. The argument runs as follows:
There is no doubt that we have the idea of a single spatio-temporal system of
material things; the idea of every material thing at any time being spatially related,
in various ways at various times, to every other at every time. There is no doubt at
all that this is our conceptual scheme. Now I say that a condition of our having this
conceptual scheme is the unquestioning acceptance of particular-identity [i.e.
numerical identity] in at least some cases of non-continuous observation. Let us
suppose for a moment that we were never willing to ascribe particular-identity in
such cases. Then we should, as it were, have the idea of a new, a different, spatial
system for each new continuous stretch of observation … Each new system would
be wholly independent of every other. There would be no question of doubt about
the identity of an item in one system with an item in another. For such a doubt
makes sense only if the two systems are not independent, if they are parts, in some
way related, of a single system which includes them both. But the condition of
having such a system is precisely the condition that there should be satisfiable and
commonly satisfied criteria for the identity of at least some items in one sub-system
with some items in the other.1
The crux of the argument may be broken down:
1
Strawson (1959:35) (Strawson’s italics).
1) The Hume-like position is a position of doubt; it entails that in cases of non-continuous
observation one should always doubt that there is numerical identity.
2) Suppose that object A was observed in continuous stretch of observation α and that
object B was observed in continuous stretch of observation β. Then doubting that A and B
are numerically identical, which is forced upon us by the Hume-like position, does not
make sense unless the two systems containing A and B respectively (call these two
systems A* and B*) are not independent, and are parts of a single system which includes
them both.
3) If A* and B* are not independent, and are parts of a single system which includes them
both, then there must be satisfiable and commonly satisfied criteria for the identity of at
least some items in A* with some items in B*.
4) By steps 2) and 3), if doubting that A and B are numerically identical is to make sense,
there must be satisfiable and commonly satisfied criteria for the identity of at least some
items in A* with some items in B*. For these items we cannot doubt their numerical
identity. Yet, by 1), the Hume-like position demands that we doubt their numerical
identity. Therefore, the Hume-like position is untenable; if all of the doubts which
comprise the Hume-like position are to make sense, then we cannot maintain all of them.
Criticising the argument
Steps 1) and 2)
Step 1) is sound; it merely states the position that is to be attacked. Step 2) makes a
single claim which derives, I think, from the idea that what we can doubt depends on the
conceptual scheme that we possess, insofar as possessing a conceptual scheme entails that
there are certain possibilities that it does not make sense to entertain, and if it does not
make sense to entertain these possibilities then it does not make sense to doubt them.
(Call this Principle P.) In step 2), Principle P is applied to the conceptual scheme that
assigns a different spatial system for each new continuous stretch of observation
(henceforth we shall call this conceptual scheme “conceptual scheme X”), and doubting
that object A and object B are numerically identical. To evaluate step 2) we should do
two things: a) consider the plausibility of Principle P and b) consider the validity of using
it in step 2).
a) Principle P is trivially true. For example, within our conceptual scheme it does not
make sense to entertain the possibility that the number thirty-three is numerically
identical to the Eiffel Tower (and therefore, it does not make sense to doubt that the
number thirty-three is numerically identical to the Eiffel Tower). It may be objected that it
is not our conceptual scheme that ensures that this is so, but to sustain this objection the
objector would have to argue that the rules governing which possibilities it makes sense
to entertain are not part of our conceptual scheme. Since a conceptual scheme is usually
thought to be a kind of all-encompassing theory of the world, it is difficult to see how he
might do this. In any case, that question need not concern us here as it is clear that
Strawson understands a conceptual scheme as stipulating, amongst other things, rules that
govern which possibilities it makes sense to entertain, and to quibble with this
understanding would merely be a digression.
b) The application of Principle P in step 2) is valid just in case the conceptual scheme X
entails that object A and object B are incomparable by numerical identity. Is this the case?
Let us first recall the example of the number thirty-three and the Eiffel Tower. The easiest
way to explain the fact that our conceptual scheme entails that it does not make sense to
entertain the possibility that the former is numerically identical to the latter, is to observe
that our conceptual scheme assigns them to two distinct domains (the domain of natural
numbers and the domain of physical objects), and stipulates that it does not make sense to
compare by numerical identity an object from the first domain with an object from the
second domain. Therefore, if conceptual scheme X assigns objects A and B to two
distinct domains and stipulates that objects from the first domain cannot be compared by
numerical identity to objects in the second domain, then the application of Principle P in
step 2) is a valid one. Well, it is obvious that this is exactly what conceptual scheme X
does. It assigns A to one spatial system (this is the first domain) and object B to a distinct
spatial system (this is the second domain). It further treats the two systems as
independent, which is to say that none of the objects in one bear any spatial or temporal
relations to objects in the other. This entails that objects from the first domain cannot be
compared by numerical identity to objects in the second domain.
I conclude that step 2) of Strawson’s argument is valid. For clarity, I summarise my
reasoning: I have shown that Principle P is true and that its application to conceptual
scheme X and doubting that object A and object B are numerically identical is valid,
which is just to say that it is true that if we possess conceptual scheme X then we cannot
make sense of doubting that object A and object B are numerically identical. As we have
seen, the reason why this is so is that conceptual scheme X treats systems A* and B* as
independent. Thus, if we are to make sense of doubting that A and B are numerically
identical, then we must treat A* and B* as dependent systems.
Steps 3) and 4)
To evaluate step 3) it is necessary to know what Strawson means by stating that “there
must be satisfiable and commonly satisfied criteria for the identity of at least some items
in A* with some items in B*.” Well, we know that irrespective of what he means, he
must mean something that is sufficient to make it unreasonable to doubt the numerical
identity of the items in A* with the items in B* for which the criteria for identity are
satisfied, because if he does not, then step 4) is invalid. What makes it unreasonable to
doubt some claim p? I take it that a minimum condition for it being unreasonable to doubt
p is that there is a good reason to believe p. Does the existence of satisfied criteria for the
identity of a pair of objects give us a good reason to believe that the objects are identical?
It does not, unless we have a good reason to believe that the criteria are genuine criteria
for identification, as well as a good reason to believe that the criteria are satisfied.2
We may now state step 3) in the minimum form that it must take if step 4) is to go
through:
3) If A* and B* are not independent, and are parts of a single system which includes them
both, then there must be criteria such that i) the satisfaction of these criteria implies the
identity of at least some items in A* with some items in B*; ii) these criteria are in fact
satisfied; iii) we believe these criteria to be criteria such that their satisfaction implies the
identity of at least some items in A* with some items in B* and have a good reason for
this belief; iv) we believe these criteria to be satisfied and have a good reason for this
belief.
We can now see that step 3) is plainly false, by considering any number of examples.
Suppose, for instance, that during the night I awaken twice. The first time I awake, I
briefly look around the room that I am sleeping in (this is the first continuous stretch of
observation, A*). The second time I awake, I again briefly look around the room, and
notice that this is my friend’s room (this is the second continuous stretch of observation,
B*). I assume that I have been moved for some reason, and I go back to sleep. Naturally, I
treat A* and B* as two dependent systems, both as parts of a single system including
them; which is just to say that I take there to be spatial and temporal relations between
objects observed in A* and objects observed in B*. (I think, for example, that objects in
the first room are x number of miles from objects in the second room, and that objects in
the first room were observed x number of hours before objects in the second room.) But
this does not entail that either condition ii) or iv) is true in the above formulation of step
3). Condition ii) does not have to be true as it does not have to be the case that some
2
Stroud (2000:14-15) fills in the gaps in Strawson’s argument in essentially the same
way. He adds the following pair of claims to Strawson’s argument: 1) If we know that the
best criteria we have for the reidentification of particulars have been satisfied, then we
know that objects to exist unperceived; 2) We sometimes know that the best criteria we
have for the reidentification of particulars have been satisfied.
object belonging to system A* is numerically identical to some object belonging to
system B*. Condition iv) does not have to be true as it does not have to be the case that I
believe some object in A* to be numerically identical to some object in B*. In fact, it is
likely that I will believe that no object in A* is numerically identical to some object in
B*.
We have seen the minimum form that step 3) must take, and by counter-example, we
have seen that in its minimum form step 3) is invalid. I conclude that Strawson’s
argument against the Hume-like position fails.
Bibliography
Strawson (1959): Individuals: An Essay in Descriptive Metaphysics, Methhuen, London.
Stroud (2000): Understanding Human Knowledge, OUP, Oxford.