268
PARTISAN REVIEW
several times before someone pointed out that it was green. And it
was
green - it didn't even
look
blue; it's just that I persisted in calling
it blue. I didn't even notice that I was using "blue" for green (or
whatever was really going on) until another person corrected me.
According to Ayer such events don't matter; I still "recognized" the
quale
green
even if I referred to it as "blue." What is this act of
"primary recognition" that connects my mind to a universal?
According to Berkeley and Hume I do not have such a thing as
an "abstract idea" or a "general idea" of green. When a particular
token - be it a green color patch or a token of the word "green" - oc–
curs in my mind, and is used as a symbol for the whole class of green
sense data, all that happens is that the token is associated with a cer–
tain class of other tokens to which it is similar or which are similar to
one another. Ayer and Russell depart from Berkeley and Hume on
this point - and with good reason . For they see that if! can think of a
particular
relation of "similarity," then I am able to recognize at least
one universal. Thus universals cannot really be avoided in the way
Berkeley and Hume wanted to do .
But a naturalistic theory of the mind must try to analyze
"primary recognition" into something scientifically more intelli–
gible - say, into straightforward causal processes. Here is where the
trouble starts.
If
a class A of events is highly statistically correlated with an–
other class B of events (with, say, a correlation coefficient of .97),
then any class
A'
of events which has almost the same members as A
will also be correlated very highly with B. Thus there is no such
thing as
the
class A of events with which a given class B is correlated.
If
the relation between occurrences of a sign, say the words "green
sense-datum," and events (occurrences of a green patch in my visual
field) were merely statistical correlation, then those words would be
correlated with many different - at least slightly different - classes of
events. There would be no such thing as
the
class of events associated
with "green sense-datum," and no basis for saying that a particular
event (imagine I utter the words "green sense-datum" when the sense–
datum is really blue, and I fail to notice the slip) wasn't
really
associated with the words .
If
one believes in non-Humean causation, then one can get
around the problem by saying that the "right" class of events A is the
class of events which exhibits whatever property objectively
brings
about
utterances of the form "this is a green sense datum" in the stan-