Remembering, Imagining, and PRO

It is widely appreciated that certain uses of anaphora, and in English

especially of PRO and the emphatic reflexive forms, can give rise to

peculiarly "first-personal" interpretations, which it has become customary

following David Lewis to call de se. There are then two local questions,

concerning English and some more familiar languages: (i) what is the nature

of these interpretations? and (ii) why are they triggered by the particular

items that trigger them? There are also general cross-linguistic questions,

for instance whether these effects are universal, and what relation, if any,

they bear to logophoric phenomena in languages having special logophoric

forms. Here I consider only questions (i) and (ii), hoping that the

answers, if correct, may help to classify the cross-linguistic phenomena

more precisely.

Contexts such as those in (1)-(3) below have been widely studied:

(1) John/Each man expects (that) he will be a hero.

(2) John/Each man expects (that) he himself will be a hero.

(3) John/Each man expects himself/PRO to be a hero.

The consensus is that there is something first-personal about the contents

of the complements in (2) and (3) that need not obtain in (1); that is, that

there are contexts in which the assertion (1) is true (with he bound to the

subject, John or each man) that are not first-personal. The contrast, then,

is between the ways in which (2) and (3) must be understood, and (1) need

not be understood; so (1) possesses an ambiguity that (2) and (3) lack.

There would appear to be three, and only three, modes of explanation of the

observed contrasts.

First (as in Higginbotham (1991), for example) it may be that the contexts

in which (1) is not first-personal are concealed de dicto contexts, where

some ³mode of presentation² attends the bound pronoun he. In that case, the

interpretations of (2) and (3) must be, and that of (1) can be,

first-personal simply because it is de re. This view is rejected in Lewis

(1979) for substantially theoretical reasons. It may also be questioned on

the ground that it doesn¹t reveal what is first-personal about, say, John¹s

or each man¹s expectations in (3).

Second, the road taken by Lewis, and in a bargain-basement form by

Chierchia, it may be proposed that the complements of the de se

constructions are of a higher logical type, properties instead of

propositions. Stalnaker (1981) responded to Lewis, effectively (I will

argue) in such a way as to defend the first view, albeit from a different

angle --- the modes of presentation do not figure in the logical form of the

examples, but rather in setting up the possible worlds in which John has

expectations about himself, but doesn¹t realize that it is himself he has

expectations about.

Third, a possibility endorsed in one form by Perry, and notoriously having

roots in Frege, it may be that there is a special, first-personal

interpretation of PRO, and perhaps of the reflexive forms as well. The

problem then is to elucidate what this interpretation is.

In this presentation I will argue that the third view is correct,

differing however from Perry on what the special interpretation of PRO is.

My view will, if correct, cash out some passing remarks in Higginbotham

(1995) and (forthcoming).

I examine some of the issues surrounding the first person within a

domain that has not, so far as I am aware, received any substantial

discussion, that of contexts of remembering, imagining and the like, where

PRO is the subject of a verbal gerundive complement, as in (4):

(4) John remembered/imagined [PRO going to the movies].

The first-personal character of such reports is foreshadowed in the

notorious example (5), which arose some years ago in discussion between

Jerry Fodor and Judith Thomson:

(5) Only Churchill remembers giving the speech (about blood, toil, tears,

and sweat).

As Fodor remarks, (5) is true provided that (a) only Churchill gave the

speech, and (b) he remembers doing so. (Fodor used this observation (in

conjunction with the Katz-Postal hypothesis, that optional grammatical

transformations did not change meaning) to argue against

"pronominalization;² that is, the view that the subject is merely

understood, rather than expressed, because a pronominal or reflexive element

has been deleted in the course of the derivation.) Evidently, the truth

conditions of the possibly true (5) are not the same as those of the

obviously (on at least one interpretation) false (6):

(6) Only Churchill remembers his giving the speech.

Could the difference between (5) and (6) be put down to the fact that the

understood subject, unlike the pronoun, is obligatorily interpreted as a

bound variable? If so, we should expect the content of the memory in (6) to

differ from one that is fully conceptualized only in that an object (John,

or Churchill) figures there directly. But this move seems to fail to do

justice to the distinction between (5) and (6). For consider those people

listening to Churchill¹s speech. Don¹t they remember the very same thing

that Churchill remembers? If so, how can (6) be true under the weak

circumstances that Fodor describes?

The gerundive complements as in (4)-(6) are easily shown to have

event-like, rather than proposition-like reference. Since some of the

higher verbs are factive, others not, uniformity suggests that their

reference is to properties of events. Moreover, these complements, when and

only when their subject is PRO, exhibit ³immunity to error through

misidentification² in the sense of Shoemaker. There is an intuitive

difference between (7) and (8):

(7) Mary imagined herself flying through space (although she didn't

realize it was she herself who was flying through space).

(8) Mary imagined [PRO flying through space] (although she didn't

realize it was she herself who was flying through space).

(7) is possibly true; (8) is contradictory, or so it would appear. If so,

then these and similar examples show that PRO is, so to speak, more

first-personal than the first-person pronoun itself. This consequence, in

the realm of remembering and imagining, points up the reality of the de se

phenomenon more radically than the ordinary contexts of propositional

attitude, or epistemic states. A solution to the problem of logical form

for the cases considered will have to have the following properties:

(a) It must apply equally to remembering (factive) and to imagining

(nonfactive);

(b) It must identify the source of immunity to error through

misidentification, and pin it

on the properties of PRO, as opposed to the first-person pronoun, or

reflexive form;

and

(c) It must do this in such a way that immunity to error through

misidentification in the

classification of ordinary perceptual experience follows along as a

special case.

To have outlined these desiderata, I think, is also to have outlined a

solution. What we have omitted so far from the discussion is that both

remembering and imagining involve the classification of experiential

elements, just as perceptual identification does. I shall call these

elements by the deliberately Kantian name of intuitions, merely as a term of

art, and I shall not inquire into their nature. But I will assume that

remembering and imagining involve both intuitions, and the ascription of

those intuitions to contents. An event of remembering, imagining, or

perceiving is a complex bundle e, that involves both intuition and

ascription. And, finally, the PRO subject of the gerundive complement is

simply given as: the subject of e; and it is for this reason that errors of

misidentification are impossible.

The reason for immunity to error through misidentification in the case of

the classification of a present perception is that the subject given in the

classification is not identified at all, except as being the subject having

the perception. That, I suggest, is just what is happening in memory and

imagination. In general, if x remembers or imagines PRO being j, the logical

form is as in (9)

(9) (Ee) Remember/Imagine[x,e, ^le¹ (j(the subject of e, e¹))]

the complement (a property of events) is different from that in (16):

(10) (Ee) Remember/Imagine[x,e, ^le¹ (j(x, e¹))]

even though x is (necessarily) the subject of e.

Our account allows for immunity to error through misidentification in the

case of ordinary perceptual experience. When x thinks x hears trumpets,

there is an event e (a complex, involving both intuition and ascription)

such that

think[x, e, ^le¹ hear(the subject of e, trumpets, e¹)]

In this context, in English, the distinction between PRO and the

first-person pronoun is neutralized: you can¹t say, ³I believe to hear

trumpets² (although the comparable construction exists, for example, in

Italian).

We have, not incidentally, a basis for showing why Fodor was right to

suggest that it is true that only Churchill remembers giving the speech,

provided only that only he gave it, and he remembers doing so. The trick

here is in the second premise, which uses PRO. Others than Churchill can

have memorial experiences ascribable to Churchill¹s giving the speech; but

since none of them ever gave the speech, they cannot have memorial

intuitions such that the subject of the experience is the giver of the

speech --- though of course they can imagine giving the speech. Returning,

finally, to the examples of propositional attitude, as in (1)-(3), we can

see what is special in the de se contexts, namely that the controller of

PRO, or the antecedent of the anaphor, is given as the subject of the state

that the reference of the controller is in.

What is special about the first person I cannot be revealed in direct

contexts, in that what is first-personal about me when I say or otherwise

affirm, ³I was born in Tennessee² is not expressed by what I affirm, either

by its content (which is just about me, that I was born in Tennessee) or by

what Kaplan calls its character (which is the same for all speakers of

English). To see why, in normal circumstances, such an affirmation is

special, and immune to errors through misidentification, we need to observe

that it is the affirmation of the proposition, if one the content of whose

subject is: the subject of that very affirmation, that makes it de se, and

distinguishes it as an affirmation about oneself, not just about a person

who happens to be oneself.

James Higginbotham

14 February 2002