[hist-analytic] Carnap on Philosophy
danny.frederick at tiscali.co.uk
Wed Aug 26 11:01:22 EDT 2009
I was not maintaining that the claim that there is singular causation can be
empirically tested. I was maintaining that it might become testable if our
theories developed in a particular direction. Here is what I had in mind: it
is no doubt an off-the-wall possibility, and it will be very simplistically
stated, but it might be sufficient to give my claim some plausibility.
Suppose someone conjectures that causes, no matter how different they may
be, have an identifying mark, something which signifies 'the causality of
the cause' (to use Kant's expression). Let's suppose that he analyses our
current stock of well-tested causal laws, 'events of type A1 cause events of
type B1,' 'events of type A2 cause events of type B2,' etc., and he finds
that events of types A1-An share a common attribute, F (not necessarily an
observable one), it being a part of accepted knowledge for each of the Ai up
to An that it has F. He then investigates, empirically, events of type An+1
onwards, to test his hypothesis that they all have F. Suppose his hypothesis
survives these tests: not only is it consistent with what we previously
knew, but it has also predicted novel facts. So we then end up with an
accepted universal law about causes: an event is a cause if and only if it
has feature F. Of course, WHAT it causes depends on its other features.
Suppose we know that it is not a law that events of type D cause events of
type E. In fact events of these two types rarely occur together. But on one
occasion when we find an event of type D followed by an event of type E,
facts about the circumstances make it plausible that the former caused the
latter. We form the hypothesis that this particular D-type event caused this
particular E-type event. The hypothesis is falsifiable because we have a
standard test for the presence or absence of feature F. We apply the test
and discover that the event of type D had feature F. This is the only time
an event of type D is known to have had feature F. Isn't this evidence for
This doubtless raises a lot of questions, in particular, about the
acceptability of the test result given that the experiment is not repeatable
(by hypotheses the causal relationship between this D-type and E-type event
is singular). But perhaps further ingenuity could answer these questions;
or, perhaps more likely, maybe this half-assed example may inspire someone
else to come up with a better one.
That's all for now!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the hist-analytic