If meaning is not in the head, then we are, in a certain sense,

110 J. A. FODOR
If meaning is not in the head, then we are, in a certain sense, not
responsible for what what we say means. In particular, we are not responsible
for the consistency of what we say in the way that we are, I suppose,
responsible for the consistency of our de dicto beliefs. Suppose that "water2"
means XYZ. Then, presumably, there is some sense in which the form of
words "water2 is not XYZ" is self-contradictory in virtue of the meanings of
its constituent expressions. But it does not follow that " water 2 is not XYZ"
expresses a self-contradictory de dicto belief; no doubt one contradicts something
when one uses that form of words, but one need not be said to contradict
oneself. That would follow only on the assumption that you can infer, Gricewise,
from the meaning of what someone says to the content of the propositional
attitudes he entertains. But, as we have repeatedly had cause to remark,
given Putnam's lexicographic views, that assumption cannot be taken as selfevident.
On the contrary, the principle of charity forbids us to make it in this
case. To put the point in a nutshell, if meaning is not in the head, then talking a
language you know is a lot like talking a language you do not know; in neither
case is there a direct inference from what you utter to what you believe.
I am not, of course, recommending Putnam's lexicographic intuitions;
I am only saying that he has the technical option of holding onto the principle
of charity for beliefs and to "water2 means XYZ" by, in effect, giving up
Grice's principle and refusing to permit direct inferences from what people
say to what they de dicto believe. On the other hand, while this position is
coherent it is surely unattractive since the question what it is that someone
believes when he believes that water2 is wet is still unanswered. And we are
running out of candidates.
Other Options From here on, the argument will go like this. I am going
to accept the intuition that Putnam's account of the meaning of "water2"
is primarily intended to explain: viz., that utterances of "water2 is wet" on
Earth2 have different truth conditions from the homophonic utterances on
Earth. But I am going to claim that this difference in truth conditions has
nothing to do with the meaning of "water2" (or of "water", or of "H2O",
or of "XYZ",. . . etc.). Indeed, I shall claim that it has nothing to do with
any fact of lexicography. Here is how I propose to show this: I shall assume,
contrary to the spirit of Putnam's proposals, that the belief that "water2 is
wet" expresses is something of the order of: the transparent, drinkable . . .
stuff people sail on is wet. (I shall refer to this as the "phenomenological
belief.) And I shall argue that, even on that assumption, you would expect
tokens of "water2 is wet" to have different truth conditions from tokens of
the corresponding English expression. Notice that I am not claiming that
"water2 is wet" does express the phenomenological belief; I propose to
remain totally agnostic on that issue. My argument will be just that the right
explanation of the facts about truth conditions survives that assumption;
hence that, so far as those facts are concerned, we can hold that "water2 is
wet" expresses the phenomenological belief if we're inclined to do so.14
To begin with: If "water2 is wet" expresses the phenomenological
belief, so too, presumably, does "water is wet". So far, then, the present
analysis raises no problems for cognitive science since we no longer have