Let "ф" (and "d'"') range over properties of persons (possible au-diences); appropriate substituends for "¢" (and "ф'") will include such diverse expressions as "is a passerby," "is a passerby who sees this notice," "is a native English speaker," and is identical with Jones." As will be seen, for U to mean something it will have to be possible to identify the value of "ø" (which may be fairly indetermi-nate) which U has in mind; but we do not have to determine the range from which U makes a selection.

The fifth redefinition is as follows:

"U meant by uttering × that •up" is true iff

(3ф) (3/) (3c):

1. U uttered x intending x to be such that anyone who has ф would think that

1. x has /
2. fis correlated in way c with 4-ing that p
3. (3ф'): U intends x to be such that anyone who has ' would think, via thinking (1) and (2), that U dis that p
4. in view of (3), Us that p; and

Il. (operative only for certain substituends for "•

U uttered x intending that, should there actually be anyone who has ф, he would via thinking (4), himself & that p;

and

Ill. It is not the case that, for some inference-element E, U intends x

to be such that anyone who has o will both (1°) rely on E in coming to er that p

(2") think that (3ф"): U intends x to be such that anyone who has ' will come to ur that p without relying on E.

Notes: (1) "f+" is to be read as "4" if clause (II) is operative, and as "think that U g's" if clause (II) is nonoperative.

(2) We need to use both "ф" and "ф'," since we do not wish to require that U should intend his possible audi-ence to think of U's possible audience under the same description as U does himself.