. 3
( 13)


including one™s own ¬rsthand version “ corresponds to anything that actually
exists in one™s head
At ¬rst sight this might not seem reasonable. Granted that whatever is inside
the head might be dif¬cult to observe, and granted that it might also be a mistake
The Self as Narrator 61

to talk about a “ghostly supervisor,” nonetheless there surely has to be some kind
of a supervisor in there: a supervisory brain program, a central controller, or
whatever. How else could anybody function “ as most people clearly do function “
as a purposeful and relatively well-integrated agent?
The answer that is emerging from both biology and Arti¬cial Intelligence is
that complex systems can in fact function in what seems to be a thoroughly “pur-
poseful and integrated” way simply by having lots of subsystems doing their own thing
without any central supervision. Indeed most systems on earth that appear to
have central controllers (and are usefully described as having them) do not. The
behavior of a termite colony provides a wonderful example of it. The colony as
a whole builds elaborate mounds, gets to know its territory, organizes foraging
expeditions, sends out raiding parties against other colonies, and so on. [. . .] Yet,
in fact, all this group wisdom results from nothing other than myriads of individ-
ual termites, specialized as several different castes, going about their individual
business “ in¬‚uenced by each other, but quite unin¬‚uenced by any master-plan.
[SO, 39“40]17

Dennett illustrates the unreality of central supervision in humans with
the phenomenon of Multiple Personality Disorder (MPD). Writing with a
collaborator, Nicholas Humphrey, he hypothesizes that a child subjected
to severe abuse may be forced to invent more than one ¬ctional self,
whereupon the child is obliged to elect one of these ¬ctional characters as
“Head of Mind,” who can then be occasionally deposed by competitors.18
The currently active personality purports to be in control, but we who
observe the succession of pretended controllers know that, in reality,
nobody is home.
There is no doubt but that Dennett™s ¬ctionalism about the self pro-
vides an attractive explanation for the phenomenon diagnosed as MPD.
According to Dennett, the self is like an imaginary friend from our child-
hood “ an especially close imaginary friend who became not merely our
alter ego but, so to speak, our auto ego. Just as some of us may have de-
veloped more than one imaginary friend, if we had unusual emotional
needs, so others may have developed more than one self, in response to
unusual circumstances, such as sexual abuse. What could be easier for a
child already engaged in populating an imaginary world? And just as our
imaginary playmates vied for the status of being our “best friend,” so our
imaginary selves may vie for the status of being our “true self.” If so, then
we suffer from MPD. Different selves take control at different times, but
only in the same way as different imaginary friends succeed one another
as favorite.
At this point, however, there is a gap in Dennett and Humphrey™s ac-
count. When one imaginary friend supplants another as favorite, nothing
J. David Velleman

much changes in the real world. But when one self supplants another in a
patient diagnosed with MPD, the patient™s behavior changes dramatically:
he walks a different walk, talks a different talk, and expresses different
states of mind. Surely, something has changed in the processes control-
ling his behavior.
Here is how Dennett and Humphrey explain changes of personality:

The language-producing systems of the brain have to get their instructions from
somewhere, and the very demands of pragmatics and grammar would conspire
to confer something like Head of Mind authority on whatever subsystem cur-
rently controls their input. [. . .] Suppose, at different times, different subsystems
within the brain produce “clusters” of speech that simply cannot easily be in-
terpreted as the output of a single self. Then “ as a Bible scholar may discover
when working on the authorship of what is putatively a single-authored text “ it
may turn out that the cluster makes best sense when attributed to different selves.
[so, 42“43]

According to this explanation, different modules in the brain take control
of the language-producing systems, yielding output whose interpretation
calls for postulation of different Heads of Mind. Different selves thus
correspond to different actual centers of control, but the selves are still
¬ctional personi¬cations of those centers, different abstracta postulated
for the sake of interpreting a narrative containing severe discontinuities.
The problem with this explanation is that it accounts only for changes
in the patient™s verbal behavior, whereas multiples are reported to change
their posture, gait, handwriting, and their projects and pursuits as well.
Why should discontinuities in the patient™s autobiography be accompa-
nied by corresponding changes in the patient™s course and manner of
action? If a human being just contains “lots of subsystems doing their
own thing,” then why can™t one of them do its thing with his feet even as
another does its thing with his mouth, so that he walks the walk of one
personality while telling the story of the other?

An answer to this question is implicit in some of Dennett™s descriptions
of self-narration, but it attributes more sophistication to the self-inventor
than Dennett acknowledges. The answer is that an autobiography and
the behavior that it narrates are mutually determining.
In the case of the self-narrating robot, Dennett imagines a strict order
of determination in one direction. He observes that “[t]he adventures of
Gilbert, the ¬ctional character, [. . .] bear a striking and presumably non-
coincidental relationship to the adventures of this robot rolling around
in the world.”19 And he explains this relationship between story and life
The Self as Narrator 63

by suggesting that the one is determined by the other: “If you hit the robot
with a baseball bat, very shortly thereafter the story of Gilbert includes
being hit by a baseball bat by somebody who looks like you.” Presumably,
the robot is designed to tell a story that corresponds to the life of that
very robot.
What Dennett doesn™t seem to imagine, in the case of this robot, is
that he might also be designed to make his life correspond to his story.
As Dennett tells it, the robot gets locked in a closet, calls out “Help me,”
and later sends us a thank-you note for letting him out. But surely a robot
smart enough to thank us for letting him out of the closet would also be
smart enough to tell us before he went back in. “I™m going into the closet,”
he would say, “Don™t lock the door.” And then he™d go into the closet,
just as he had said he would. (If he didn™t do what he had said, he might
get stuck somewhere else and have to wait for help while we went looking
for him in the closet.) A robot that can maintain correspondence in one
direction, by saying that he™s locked in the closet when he is, should be
able to maintain correspondence in the other direction, by going into
the closet when he has said that he will. Thus, whereas the robot will
sometimes update his story to re¬‚ect recent events in his career, at other
times he will narrate ahead of himself and then follow a career that re¬‚ects
his story.
Although Dennett doesn™t attribute this sort of sophistication to the
robot, he does implicitly attribute it to a patient with MPD:

Consider the putatively true case histories recorded in The Three Faces of Eve
(Thigpen & Cleckley, 1957) and Sybil (Schreiber, 1973). Eve™s three faces were
the faces of three distinct personalities, it seems, and the woman portrayed in
Sybil had many different selves, or so it seems. How can we make sense of this?
Here is one way, a solemn, skeptical way favored by the psychotherapists with
whom I have talked about the case: When Sybil went in to see her therapist for
the ¬rst time, she was not several different people rolled into one body. Sybil
was a novel-writing machine that fell in with a very ingenious questioner, a very
eager reader. And together they collaborated to write many, many chapters of a
new novel. And, of course, since Sybil was a sort of living novel, she went out and
engaged the world with these new selves, more or less created on demand, under
the eager suggestion of a therapist. [CNG, 111]

What does Dennett mean when he says that Sybil “engaged the world
with these new selves”? Surely, he means that Sybil acted out the stories
that she and her therapist had composed. She was a “living novel” in the
sense that she not only narrated the roles she played but also played the
roles that she narrated.
J. David Velleman

That™s why Sybil™s behavior always manifested the personality whose
story she was telling at the moment. Her life shaped her story, and her
story shaped her life, all because she was designed to maintain correspon-
dence between the two. Hence the control of her speech and the control
of her movements were not entirely independent. They were in fact inter-
dependent, since the controller of her speech must have been responsive
to her movements, and the controller of her movements must have been
responsive to her speech.

Yet if a self-narrator works in both directions, then the self he invents is
not just an idle ¬ction, a useful abstraction for interpreting his behavior.
It “ or, more precisely, his representation of it “ is a determinant of the
very behavior that it™s useful for interpreting.20 Indeed, the reason why
the narrator™s representation of a centrally controlling self is so useful for
interpreting his behavior is that it, the representation, really does control
his behavior to some extent.
Of course, the central controller he has may not be much like the one
he represents himself as having. After all, a self-narrator doesn™t represent
himself as being centrally controlled by his own story.
Or does he?
In order to answer this question, we must consider some prior ques-
tions that Dennett overlooks. First, consider whether the behaviors at-
tributed to Gilbert by the robot™s novel-writing computer include the
behavior of writing the novel. When the robot gets locked in a closet, he
tells about Gilbert™s being locked in a closet; but when he tells the story
of Gilbert, does he also tell about Gilbert™s telling that story? He says “Call
me Gilbert”; but does he ever say, “I™m Gilbert and this is my story”? He
writes a note that says “Thank you,” but can he also write a note that says
“I™m writing to say thanks”? I can™t imagine why not.
Nor can I imagine how the robot would tell the story of Gilbert without
including information about the causes and effects of the events therein.
When he calls for help, he might well elaborate, “I™ve gotten myself locked
in the closet,” thus attributing his current predicament to what he did a
moment ago. And when he writes his thank-you note, he might well begin,
“I™m writing because you let me out of the closet,” thereby attributing his
present behavior to an earlier cause. A story that merely described one
event after another, without mentioning any causal connections, would
hardly qualify as a narrative.
Thus, the features of himself that the robot can ascribe to Gilbert
ought to include this very activity of self-description; and he should also
The Self as Narrator 65

be able to describe the causes and effects of his activities, including this
one. Hence in ascribing his activities to Gilbert, the robot should be able
to describe the causes and effects of his doing so.
Now, what causal role might the robot attribute to his own remark
“I™m going into the closet”? He might say, “I™m telling you this because
I™m on my way into the closet,” thereby casting his speech as an effect
of his movements. But this remark would be strictly accurate only if the
robot was going into the closet anyway and was merely reporting on his
current trajectory. What I have imagined, however, is that the robot goes
into the closet partly because of having said so, in order to maintain
correspondence between his story and his life. Insofar as the robot can
report on the causes and effects of his behavior, then, he ought to say,
“I™m going into the closet partly because I™ve just said so” “ or, perhaps,
“I™m hereby heading for the closet,” a remark that implicitly ascribes this
causal role to itself.
I think that human self-narrators make such remarks frequently, when-
ever they make promises or other verbal commitments, which may be as
trivial as “I™m heading for the closet.” As you putter around the of¬ce at
the end of the day, you ¬nally say, “I™m going home,” not because you were
already about to leave, but because saying so will prompt you to leave. As
your hand hovers indecisively over the candy dish, you say, “No, I won™t,”
not because you weren™t about to take a candy, but because saying so may
stop you from taking one.21 These utterances are issued as commitments,
in the understanding that they will feed back into your behavior. Hence
you do understand that your running autobiography not only re¬‚ects but
is also re¬‚ected in what you do.

These observations suggest that the “central controller” of a person may
indeed be a ¬ction, not in the sense that it is a ¬ctional character in the
person™s autobiography, but in the sense that it is the person™s autobi-
ography “ the re¬‚ective representation that feeds back into the person™s
behavior.22 This central controller is in fact what social psychologists call
the self. In the social-psychology literature, the word “self” denotes a per-
son™s self-conception rather than the entity, real or imagined, that this
conception represents. And the same literature reports evidence for the
feedback loop I have posited.
Researchers have found, for example, that subjects tend to predict that
they will vote in the next election at a far higher rate than the average
turnout; but that the turnout among those who have predicted that they
will vote is also higher than the average.23 Many who wouldn™t otherwise
J. David Velleman

have voted, it seems, end up voting because of having predicted that
they would, thus conforming their lives to their stories.24 Like Sybil, who
“lived out” the novels that she composed with her therapist, these sub-
jects lived out the predictions that they were prompted to make by the
Similar research has documented a slightly different phenomenon,
known as the attribution effect. Subjects can be led to act annoyed or
euphoric depending on whether they are led to believe, of arti¬cially
induced feelings of arousal, that they are symptoms of annoyance or
euphoria.25 Subjects can be prevented from acting shyly in unfamiliar
company by being led to attribute their feelings of anxiety to something
other than shyness.26 And researchers can modify the degree of retali-
ation that a subject carries out against putative aggressors by modifying
the degree of anger that he believes himself to be feeling toward them.27
All of these experiments suggest that people tend to manifest not just
what they™re feeling but also what they represent themselves as feeling.
Whether they behave angrily depends, not just on whether they are an-
gry, but on whether they interpret their feelings by updating their auto-
biographies with the attribution “I™m angry.” Whether they behave shyly
depends on whether the current episode of their autobiography says “I™m
feeling shy.”
Here the subjects are “living out” their self-conceptions in a more holis-
tic sense. Unlike the self-predicting voters, they aren™t doing things that
they have described themselves as doing. Rather, they are doing things
that would accord with what they have described themselves as feeling.
But this process, too, is implicit in Dennett™s account of self-narration. For
as we have seen, Dennett says that “[w]e try to make all of our material
cohere into a single good story.”28 And acting in accordance with our self-
ascribed emotions is a way of ensuring that our story-material will cohere.
Consider how this process might be implemented in the robot who
calls himself Gilbert. If the robot is locked in the closet, his internal state
may include the initiation of a subroutine that searches for avenues of
escape from danger and quickly selects the one most readily available.
This subroutine will have a name “ say, “fear” “ and so the robot will
report “I™m locked in the closet and I™m starting to get frightened.” And
now two different modules in the robot will dispose him to take action.
One is the fear module, which may recommend breaking down the door
as one of several preferred alternative avenues of escape; the other is the
narrative module, which will recommend “I™m breaking down the door”
as one of several preferred continuations the story. If after he said “I™m
The Self as Narrator 67

getting frightened,” the robot continued his story with “I think I™ll back
up my hard disk,” then he would no longer be writing a passable novel,
since his “material” wouldn™t cohere. His narrative module will therefore
favor “I™m breaking down the door” as a more coherent way to continue
the story. And the narrative module can go ahead with this continuation
of the story, con¬dent of being borne out by the robot™s behavior, since
the robot is sure to break down the door once his preexisting fear is
reinforced, in motivating that behavior, by his disposition to maintain
correspondence between his story and his life.
Thus, having attributed an internal state to himself (“I™m getting fright-
ened”), the robot is in¬‚uenced to act in accordance with that attribution.
Like a human being, he tends to manifest fear not only because he™s “feel-
ing” it but also because he “thinks” it™s what he™s feeling.

I have now introduced the idea of the robot™s having a “narrative module”
that produces Gilbert™s autobiography. This module must incorporate,
¬rst, the function of ensuring that the robot™s story corresponds to its
life and, second, the function of maintaining the internal coherence of
the story itself. The module must be designed to produce a text that is
both consonant with the facts and suf¬ciently consonant with itself to
qualify as a story.
Moreover, I have suggested that the robot can maintain correspon-
dence between its story and its life in either direction, by narrating its
actions or by acting out its narrative. Hence in pursuit of narrative co-
herence, the module can sometimes choose, among possible turns in its
story, the one that would best ¬t the story thus far, precisely because it
can then in¬‚uence the robot™s life to take the corresponding turn. The
narrative module needn™t always depend on the robot™s career to provide
material for a coherent story; it can sometimes tell a coherent story and
induce the robot™s career to follow.
In previous work, I have argued that a creature equipped with such a
module would amount to an autonomous agent.29 I won™t repeat those
arguments here, but let me brie¬‚y illustrate some of them with the help
of Dennett™s self-narrating robot.

As Gilbert rolls down the hall, he may autobiographically announce where
he is going. But he needn™t just report where he is already programmed
to go, since his disposition to maintain correspondence between story
and life will dispose him go wherever he says he™s going. Suppose that
he is in the middle of his Fetch New Batteries subroutine, which sends
J. David Velleman

him to the supply closet (where he sometimes get locked in). The fact
remains that if he said “I™m on my way to the library,” his disposition
to maintain correspondence would dispose him to head for the library
instead. So if another, concurrently running subroutine can get Gilbert™s
speech-producing module to emit “I™m on my way to the library,” then it
may be able bring about a change of course.
Now, Gilbert™s disposition to maintain correspondence wouldn™t be
suf¬cient to make him head for the library if no other subroutines in-
clined him in that direction. Even if he said “I™m on my way to the li-
brary,” his Fetch New Batteries routine would still favor heading for the
supply closet, and his disposition to bear out his story would be unlikely
to override a routine for obtaining essential resources. But I imagine his
inner workings to be in the following, rather complicated state. Various
task-speci¬c subroutines are running concurrently, and some of them are
making bids for control of his locomotive unit, to propel him toward one
destination or another. His Fetch New Batteries subroutine is bidding
for a trip to the supply closet, while his Departmental Service subroutine
may be bidding for a trip to the library, in order to ¬ll a faculty mem-
ber™s request for a book. Meanwhile, the narrative-composing module
is busy updating the story of Gilbert™s most recent adventures and the
ongoing evolution of his inner states, including which task-speci¬c sub-
routines are running and where they are bidding him to go. And the
disposition of this module to maintain correspondence between his story
and his life, though not suf¬cient by itself to override other demands for
locomotion, is suf¬cient to tip the balance in favor of one or another
of those demands. So if Gilbert says “I™m heading for the supply closet,”
his disposition to bear out his story will reinforce the battery-fetching
demands, and he™ll head for the supply closet; whereas if he says “I™m
heading for the library,” his disposition to bear out his story will rein-
force the demands of departmental service, and he™ll head for the library
instead. As long as the competition among those subroutines is not too
lopsided, the narrative module is in a position to decide where Gilbert
When I say that the narrative module can “decide” where Gilbert goes, I
mean it can literally decide. For as we have seen, this module is in a position
to have Gilbert speak the truth in naming any one of several destinations,
each of which he would thereby head for, if he said so. The novelist in
Gilbert can therefore make up where Gilbert is headed, choosing among
different available turns in his story, none of which is privileged as the
turn that the story must take in order to be true. As a self-narrator, then,
The Self as Narrator 69

Gilbert faces an epistemically open future “ which gives him, in my view,
as much free will as a human being.30

On what basis will the narrative-composing module make its decision?
It can declare a winner in the contest among demands for locomotion,
but on what basis will it adjudicate among those demands? The answer,
already implicit in Dennett™s theory, is that it will adjudicate on the basis
of how best to continue the story “ how to “make [its] material cohere.”31
In many cases, acting on one demand will already make more narrative
sense than acting on another, and the narrative-composing module will
therefore declare a winner simply by telling the more coherent continua-
tion of the story. But if neither continuation would make more narrative
sense at this point, then the module can ¬ll in more detail about its cur-
rent situation, by recording which demand is stronger than the other or
by recording more of the circumstances “ which may arouse more inter-
nal states, which can in turn be recorded. At some point, the story will
become more amenable to one continuation or other, and the narrative
module can go ahead with the better continuation, thereby making its
In this way, I believe, the module will decide on the basis of consid-
erations that serve as reasons for acting. In canvassing Gilbert™s outer
circumstances and inner states, it will weigh them as considerations in
light of which various possible actions would make sense. It will thus
weigh Gilbert™s circumstances and states as providing a potential rationale
for his next action “ that is, an account that would make the action intel-
ligible, a coherent development in his story. When the novelist in Gilbert
writes in the action with the best rationale, he will in effect be deciding
for reasons.

Note that this claim places signi¬cant constraints on the conception of
narrative coherence on which I can rely. One might have thought that
whether an action would make for a coherent continuation of Gilbert™s
story ultimately depends on whether he has reason for taking it. My claim,
however, is that whether Gilbert has reason for taking an action ultimately
depends on whether it would make for a coherent continuation of his
story. Because I make the latter claim, I cannot adopt the former in
order to explicate narrative coherence, since my account would then
become viciously circular: narrative coherence cannot ultimately depend
on rational justi¬cation if rational justi¬cation ultimately depends on
narrative coherence.
J. David Velleman

Of course, we can tell a story about Gilbert that makes sense because it
portrays him as taking actions for which he has reasons; for we can portray
him as taking actions because they cohere with his story. Indeed, I have
already claimed that self-narration takes account of its own effect on the
subject™s behavior, by portraying him as hereby heading for the supply
closet or the library. To this extent, self-narration already relies for some
of its coherence on the fact that the subject is doing what coheres with
this very story “ hence on the fact that he is doing something for which
he has reasons, as I conceive them. But this fact cannot be the sole basis
for the narrative coherence involved. There must be some prior basis on
which the subject™s action makes sense in light of his story before it can
also make sense in light of his tendency to do what makes sense.
The nature of narrative coherence is a topic that lies beyond the scope
of this chapter.32 But I have already indicated one basis on which Gilbert
can regard actions as cohering with his story independently of his having
reasons for taking them. I have supposed that Gilbert understands his
own inner workings, in the form of the various subroutines that are vying
to control his behavior. Gilbert understands that whatever he does will
be controlled by one of these subroutines and will consequently make
sense by virtue of having a causal explanation, which cites the relevant
subroutine as the controlling cause. In considering which action would
make for a coherent continuation of his story, Gilbert can look for an
action that would have the most satisfying causal explanation in light of
the subroutines vying for control.
Of course, where Gilbert has subroutines vying for control, human be-
ings have con¬‚icting motives, which serve as controlling causes of their
behavior. Where Gilbert looks for an action that would best be explained
by his subroutines, humans look for an action that would best be ex-
plained by their motives. That™s why humans look to their motives “ that
is, to their desires and beliefs “ as reasons for acting.

In deciding for reasons, the inner novelist plays the role that is ordi-
narily attributed to the self. A third conception of the self has there-
fore emerged. According to Dennett™s conception of the self, with which
I began, the self is the merely ¬ctional protagonist of a self-narrator™s
autobiography. According to the second conception, the self is the
autobiographer™s re¬‚ective representation, which guides his actions as
well as his speech. What has now emerged, however, is that control
rests with the narrative module “ the inner novelist, recording the sub-
ject™s last step and declaring his next step, in a way that amounts to
The Self as Narrator 71

deciding for reasons. According to the third conception, then, the self is
the narrator.
This third conception of the self no longer supports the skepticism of
Dennett™s initial conception. The protagonist of Gilbert™s autobiography
is no longer, as Dennett believes, a merely ¬ctional character whose shoes
cannot be ¬lled by the actual author. Now that the robot has a central
controller that makes decisions for reasons, he has a self, and so his story
has come true.
Note that what ¬lls the shoes of the protagonist in the story of Gilbert
is the robot, not the robot™s self. “Gilbert” is not the name of a self; it™s the
name of a uni¬ed agent who has a self, in the form of an inner locus of
agential control. My current claim is that the self-narrating robot really
is endowed with a self in this sense and can therefore live up to the
portrait of the protagonist in his autobiography. He is endowed with a
self because his inner narrator is a locus of control that uni¬es him as an
agent by making decisions on the basis of reasons.

The self-narrating agent is a bit like an improvisational actor, enacting
a role that he invents as he goes. The difference is that an improvisa-
tional actor usually invents and enacts a role that he is not playing in fact.
His actions represent what they are not “ actions other than themselves,
performed out of motives other than his. By contrast, the self-narrator is
an ingenuous improviser, inventing a role that expresses his actual mo-
tives in response to real events. He can improvise his actual role in these
events because his motives take shape and produce behavior under the
in¬‚uence of his self-descriptions, which are therefore underdetermined
by antecedent facts, so that he partly invents what he enacts.
Yet how can an agent act out invented self-descriptions without some-
how falsifying them, by being or doing something other than is therein
described? How can enacting a role fail to involve fakery or bad faith?
The answer is that when the agent invents descriptions to be enacted,
he describes himself as the inventor-enactor of those descriptions. He
describes himself as hereby heading for the supply closet or the library, thus
describing his actions as ¬‚owing from these descriptions, as realizations
thereof. The protagonist in his autobiography is therefore both ¬ctive
and factual “ ¬ctive, because his role is invented by the one who enacts
it; factual, because it is the role of one inventing and enacting that role.
To be sure, a self-narrator can go beyond what is factual, if he ap-
plies self-descriptions whose autobiographical application won™t make
them true. Although he can sometimes tip the balance of his antecedent
J. David Velleman

motives in favor of leaving the of¬ce by saying “I™m leaving,” at other
times he can™t, and then a declaration of departure would be ineffectual “
an instance of weakness of will. Alternatively, his motives for going home
may already be suf¬cient to make him go home no matter what he says “
in which case, “I™m leaving” is the only true thing for him to say. Within
these constraints, however, the self-narrator retains considerable latitude
for invention. Even if he is already determined to leave the of¬ce, he is
probably capable of going home or going out for a drink, or perhaps just
taking a walk, depending on what he writes into his story.
To this extent, I can endorse Dennett™s claim that the self is a ¬ctive
character. Where I disagree with Dennett is over the claim that, being
¬ctive, this character doesn™t exist in fact. Dennett thinks the real-life
author of an autobiography is signi¬cantly different from the character
portrayed as the protagonist. I think that the author of an autobiography
is just like the protagonist, since the protagonist is portrayed as a self-
improvising character, the inventor-enactor of his own story “ or, as I
prefer to say, an autonomous agent.

My disagreement with Dennett over the truth-value of a human being™s
autobiography results from two subsidiary disagreements. On the one
hand, Dennett believes that a human being has no central controller,
whereas I believe that Dennett himself is committed to crediting a hu-
man being with a central controller, in the form of a narrative intelligence.
On the other hand, Dennett believes that a human being™s autobiogra-
phy portrays his central controller as a “brain pearl” or Cartesian ego,
whereas I believe that this autobiography portrays the central controller
as the narrative intelligence that it is. We live up to our aspirations with
respect to selfhood, then, partly because we have more of a self than
Dennett expressly allows, and partly because we aspire to less than he
I have overlooked another disagreement with Dennett, which I should
mention before closing. Although Dennett tries to deny the unity of the
self-narrating agent, he commits himself expressly to the unity of the
narrative “ to the proposition that “We try to make all of our material co-
here into a single good story.”33 Indeed, the unity of this narrative seems
to account for the temporal unity of the purely ¬ctional self in which
Dennett believes. This ¬ctional character remains one and the same
self because he is the protagonist in one and the same continuing story.34
In my view, however, we tell many small, disconnected stories
about ourselves “ short episodes that do not get incorporated into our
The Self as Narrator 73

life-stories. The process of self-narration shapes our day-to-day lives in
units as small as the eating of a meal, the answering of a phone, or even
the scratching of an itch; but our life stories do not record every meal
eaten, every phone answered, or every itch scratched. Because the narra-
tives of these minor episodes are never uni¬ed into a single story, their
protagonist cannot derive his unity from theirs. The agent who types this
letter ˜a™ is the same person who cut his fore¬nger with that pocketknife
in the summer of 1959, but not because there is any single narrative in
which he ¬gures as the protagonist of both episodes.
So when I describe the inner narrator as a uni¬ed self, I am not speak-
ing of the temporal unity that joins a person to his past and future selves; I
am speaking of agential unity, in virtue of which a person is self-governed,
or autonomous. In my view, autonomy is not related to personal identity
in such a way that a single entity plays the role of self in both phenomena:
that which makes us self-governed is not that which makes us self-same
through time.35

The material in this chapter was ¬rst presented to a seminar on the self,
taught in the fall of 1999 at the University of Michigan. Versions of the chap-
ter have been presented to the Philosophy Departments of the University of
Pittsburgh, the University of Maryland, the University of Chicago, and the Uni-
versity of G¨ ttingen; to a conference on Morality and the Arts at the University
of California, Riverside, with John Martin Fischer serving as commentator;
and as one of the Jerome Simon Lectures at the University of Toronto. I have
received helpful comments from the audiences on these occasions as well as
from Linda Brakel and Dan Dennett.
A list of philosophers who have held this view would include Charles Taylor
(Sources of the Self: The Making of the Modern Identity [Cambridge, MA: Harvard
University Press, 1989]; Human Agency and Language [Cambridge: Cambridge
University Press, 1985]); Harry Frankfurt (The Importance of What We Care
About [Cambridge: Cambridge University Press, 1987]); Christine Korsgaard
(The Sources of Normativity [Cambridge: Cambridge University Press, 1996];
“Self-Constitution in the Ethics of Plato and Kant,” Journal of Ethics 3 [1999]:
1“29); Tamar Schapiro (“What is a Child?” Ethics 109 [1999]: 715“38); and
Michael Bratman (“Re¬‚ection, Planning, and Temporally Extended Agency,”
Philosophical Review 109 [2000]: 35“61).
See, especially, Schapiro.
See Daniel Dennett: “The Origins of Selves,” Cogito 3 (1989): 163“73 [here-
inafter OS]; “The Reality of Selves,” in Consciousness Explained (Boston: Little,
Brown and Company, 1991), Chapter 13 [RS]; “The Self as a Center of Narra-
tive Gravity,” in Self and Consciousness: Multiple Perspectives, eds., Frank S. Kessel,
Pamela M. Cole, and Dale L. Johnson (Hillsdale, NJ: Erlbaum Associates,
J. David Velleman

1992), 103“115 [CNG]; with Nicholas Humphrey, “Speaking for Ourselves,”
reprinted in Brainchildren: Essays on Designing Minds (Cambridge, MA: MIT
Press, 1998), 31“58 [SO].
Dennett describes his view as a “middle-ground position” on the question
“whether there really are selves” (RS, 413).
CNG, 114“15.
I use the term “¬ctive” because, to my ear, it shares with “¬ctional” the sense
of “invented” or “made up”, but not the sense of “untrue.” Those who do
not already share these linguistic intuitions should take them as stipulated
RS, 424. See p. 423: “independently existing soul-pearls.”
RS, 420.
RS, 420.
RS, 428.
RS, 422.
OS, 163.
RS, 429.
If the objection here is merely that writing a passable novel is an activity
that is most perspicuously interpreted as the product of a conscious thinker,
then Dennett can of course agree, since he believes that positing a conscious
thinker, Gilbert, is the most perspicuous way of interpreting the novel-writing
robot. What he denies is that writing a novel requires a real, conscious thinker
of the sort that would be postulated by such an interpretation.
Quoted at notes 9“11.
RS, 418, quoted after n. 13.
See also OS, 167“68, and RS, 416, where Dennett remarks, “There is [. . .]
no Oval Of¬ce in the anthill,” just as he subsequently remarks that “there is
no Oval Of¬ce in the brain” [RS, 429].
SO, 41. For another narrative-based analysis of MPD, see Valerie Gray Hard-
castle and Owen Flanagan, “Mupltiplex vs. Multiple Selves: Distinguishing
Dissociative Disorders,” The Monist 82 (1999) 645“57.
CNG, 108. Note, then, that Dennett does not conceive of autobiographies as
“entirely confabulated” narratives in which “anything goes” (Hardcastle and
Flanagan, 650, 653).
Flanagan says, “[T]he self as represented has motivational bearing and behav-
ioral effects. Often this motivational bearing is congruent with motivational
tendencies that the entire system already has. In such cases, placing one™s
conception of the self into the motivational circuits enables certain gains
in ongoing conscious control and in the ¬ne-tuning of action” (“Multiple
Identity,” p. 140).
I discuss cases like these in “How to Share an Intention,” Philosophy and
Phenomenological Research 57 (1997): 29“50; and in The Possibility of Practical
Reason (New York: Oxford University Press, 2000a).
Dennett almost strays into this second conception of the self. For example:

A self, according to my theory, is not any old mathematical point, but an abstraction
de¬ned by the myriads of attributions and interpretations (including self-attributions
The Self as Narrator 75

and self-interpretations) that have composed the biography of the living body whose
Center of Narrative Gravity it is. As such, it plays a singularly important role in the
ongoing cognitive economy of that living body, because, of all the things in the envi-
ronment an active body must make mental models of, none is more crucial than the
model the agent has of itself. [RS 426“27]

Dennett begins this passage by speaking of the self as an abstract object
posited by the host™s autobiography. But then he speaks of the self as playing
“a singularly important role in the ongoing cognitive economy” of the host,
and ¬nally he describes it as “the model that the agent has of itself.” At this
point, it is unclear whether he is speaking of an abstract object or of the
host™s representation of it, which is a real element in the host™s psychology,
positioned to play a causal role in his mental economy.
Greenwald, A. G., Carnot, C. G., Beach, R., and Young, B., “Increasing Vot-
ing Behavior by Asking People if They Expect to Vote,” Journal of Applied
Psychology, 72 (1987): 315“18.
I explore this literature in “From Self-Psychology to Moral Philosophy,” in
Action Theory and Freedom, Philosophical Perspectives 14 (2000), 349“77. For
a more recent philosophical discussion of this phenomenon, see Richard
Moran, Authority and Estrangement: An Essay on Self-Knowledge (Princeton:
Princeton University Press, 2001), 38 ff.
Schachter, S., and Singer, J. E., “Cognitive, Social and Physiological Determi-
nants of Emotional States,” Psychological Review 69 (1962): 379“99.
Brodt, S. E., and Zimbardo, P., “Modifying Shyness-Related Social Behavior
Through Symptom Misattribution,” Journal of Personality and Social Psychology
41 (1981): 437“49.
Berkowitz, L., and Turner, C., “Perceived Anger Level, Instigating Agent,
and Aggression,” in Cognitive Alteration of Feeling States, eds. H. London and
R. E. Nisbett (Chicago: Aldine, 1972), 174“89; Zillman, E., Johnson. R. C.,
and Day, K. D., “Attribution of apparent arousal and pro¬ciency of recovery
for sympathetic activation affecting excitation transfer to aggressive behav-
ior,” Journal of Experimental Social Psychology 10 (1974): 503“15; Zillman, D.,
“Attribution and Misattribution of Excitatory Reactions,” New Directions in
Attribution Research, vol. 2, eds. John H. Harvey, William Ickes, and Robert F.
Kidd (Hillsdale, NJ: Erlbaum, 1978), 335“68.
CNG, 114, quoted on p. 57.
See Practical Re¬‚ection (Princeton: Princeton University Press, 1989);
http://www-personal.umich.edu∼velleman/Practical Re¬‚ection; and The
Possibility of Practical Reason (Oxford: Oxford University Press, 2001).
For a detailed defense of this claim, see my “Epistemic Freedom,” Paci¬c
Philosophical Quarterly 70 (1989): 73“97; reprinted in The Possibility of Practical
CNG, 114.
But see my “Narrative Explanation,” The Philosophical Review 112 (2003):
CNG, 114 (quoted on p. 57), emphasis added.
This view is endorsed by Flanagan, “Multiple Identity,” p. 136: “Augustine™s
Confessions is an autobiography. It is the story of a single self. This is established
J. David Velleman

in part because Augustine is able to produce an account that narratively
links up the multifarious episodes of his life from the ¬rst-person point of
35. I argue for this view in “Identi¬cation and Identity,” in Contours of Agency,
a Festschrift for Harry Frankfurt, edited by Sarah Buss and Lee Overton
(Cambridge, MA: MIT Press, 2002), 91“123.

Autonomy and Self-Identity

Marina A. L. Oshana

In discussions of autonomous agency, much attention is paid to the psy-
chological, social, and historical conditions the autonomous person must
satisfy, and to the various epistemic and metaphysical phenomena that
might jeopardize these conditions. Discussants assume, in ascribing au-
tonomy to individuals, a “self” that is capable of acting, that this self
has a coherent and sustained identity over time, and that the actor is
“truly” or “deeply” herself in acting. A capacity for unimpaired critical
self-re¬‚ection is included in standard accounts of autonomy as well. The
task of self-re¬‚ection is to appraise aspects of a person™s self, such as cog-
nitive, affective, valuational, and dispositional states, as well as personal
commitments, social roles, and ideals, to determine if these are compo-
nents of the person™s life with which the person “wholeheartedly identi-
¬es” or embraces without reservation so as to render them “authentic” to
Accounts of autonomous agency vary in the details. For example, de-
fenders of a liberal conception of autonomy might disagree about the
nature of authenticity. Other philosophers repudiate all such depictions
of the autonomous self on the grounds that they falsify the nature of the
self, and the conditions of its identity and authenticity. Among postmod-
ernists, for example, the charge arises that the assumption of a coherent
self misrepresents persons in presupposing a permanency of identity,
where in fact the identity of persons is pliant.2 Communitarians charge
that the ideal of authenticity, “of being true to myself and my particular
way of being”3 by an “inwardly generated” set of criteria is inaccurate

Marina A. L. Oshana

because it overlooks the fundamentally dialogical character of human
development. We become who we are through our interaction and con-
versations with others; “we de¬ne our identity always in dialogue with,
sometimes in struggle against, the things our signi¬cant others want to
see in us.”4
In what follows, I am going to overlook concerns about whether tradi-
tional accounts of autonomy portray persons in an accurate fashion. My
interest here is not in the question of whether a coherent and sustained
identity is metaphysically or socially plausible, or even desirable. (I be-
lieve it is all three, but that is beside the point.) Rather, the task of this
chapter is to assess carefully the role played by an agent™s conception of
herself, or her “self-identity,” in accounts of autonomy, her conception
being rooted in a pliant or stable identity notwithstanding.
I will explore four questions: One, what constitutes self-identity or
an agent™s conception of herself? Two, in what fashion is autonomous
agency dependent upon and characterized in terms of the person™s con-
ception of herself? Three, insofar as a person™s conception of herself
is a component of her autonomy, must the agent endorse, or at least
fail to repudiate, the elements constitutive of her self-conception? Four,
when do the elements constitutive of a person™s conception of herself
impair autonomy? My objective in answering these questions is to show
that having a self-conception is an essential component of being au-
tonomous and, moreover, that an agent™s self-conception need not be
authentic in the manner traditional accounts describe if the agent is to be

I The Concept of Self-Identity
Who am I and how do I conceive of myself? Since a person™s self-
conception may fail to re¬‚ect who she in fact is, these are different ques-
tions. The ¬rst question addresses the person™s identity, while the second
cites her self-identity or self-conception. An answer to the ¬rst question
is found in what Amelie Rorty and David Wong call one™s “central iden-
tity traits” “ the characteristics and relationships that are integral to a
person™s nature, motives, and life-plans. Identity traits may consist of be-
liefs, preferences, values, articles of faith, dispositions of temperament,
habits, commitments and ideals, as well as relationships, social roles, and
biology. Those integral to a person™s identity can be demarcated in sev-
eral ways. Rorty and Wong note, ¬rst, that these tend to be characteristics
upon which other aspects of oneself, such as one™s predilection to certain
Autonomy and Self-Identity 79

beliefs, desires, attitudes, relationships, social roles, and actions depend.
Second, inasmuch as one™s identity encompasses a set of general ends
and values, one™s central identity traits guide practical deliberation, af-
fecting a person™s motives for action as well as the acts she performs and
the manner of performance. Third, central traits such as gender and
ethnicity animate social interaction by in¬‚uencing the way a person is
categorized and dealt with by others. Fourth, central traits include those
dispositional and affective characteristics that dominate in situations that
require coping with stress or con¬‚ict, characteristics that are, importantly,
“the focus of self-evaluation and self-esteem.”5
While these traits are not exhaustive of identity, certain of them pro-
vide the raw material out of which a conception of self emerges. One
becomes aware of oneself “ one™s self-conception is thrown into relief “
in the course of an intellectually, emotionally, and experientially oriented
investigation of oneself.6 This sort of inquiry “tend[s] to arise when there
are problems of action and policy, when an “identity crisis” triggers an
attempt to articulate an individual . . . identity, particularly when there
is disagreement about [its] characterization and importance.”7 Not ev-
ery stage of this process occurs at a level of deliberative and conscious
investigation; self-awareness can confront one uninvited. And some as-
pects central to one™s self-conception or self-identity may be hidden from
scrutiny. But I will use the phenomenon of deliberate investigation as
an illustration of the manner in which one™s conception of self emerges
most explicitly and lucidly.
For example, one investigates the beliefs that could move one to act.
Or one notices the sensations or emotions certain activities, relation-
ships, and states of affairs elicit in oneself. One takes stock of the set
of one™s experiences, and appraises their value. Thus the task of becom-
ing aware of oneself has a cognitive and a non-cognitive dimension. What
emerges from this process of exploration is a picture of how one identi¬es
oneself “ that is, a self-conception.
Not every aspect of a person™s self-conception draws from, or is con-
stituted out of, one™s central identity traits. One™s self-conception may
be inaccurate or confused, even if one™s identity cannot be inaccurate.
For example, my self-conception might include the belief that I am the
present King of France. What unites some of the principal identity traits
and provides them with their centrality to self-identity, as opposed to iden-
tity of a self more generally, is that each holds us in a peculiarly tenacious
way. Their centrality to a person™s self-conception is established by the
fact that a person will regard herself as radically changed if the trait is
Marina A. L. Oshana

lost or strongly modi¬ed. At root, how we conceive of ourselves is typi¬ed
most readily in what we consider to be the ineliminable and intractable
aspects of ourselves.

Ia Volitional Necessity
The concepts of the ineliminable, the intractable, and the unthinkable
have been developed by Harry Frankfurt in a series of thoughtful and in-
novative essays. Frankfurt argues that a person™s essential nature or iden-
tity as an agent is constituted by his necessary personal characteristics “
certain ineliminable beliefs, desires, values, articles of faith, personal
relationships, and so forth without which the person cannot be what
he is “ and these are characteristics of a person™s will. Frankfurt states:

To the extent that a person is controlled by his volitional necessities, there are
certain things that he cannot help willing or cannot bring himself to do. These
necessities substantially affect the actual course and character of his life. But
they affect not only what he does: they limit the possibilities that are open to
his will, that is, they determine what he cannot will and what he cannot help
willing. Now the character of a person™s will constitutes what he most centrally is.
Accordingly, the volitional necessities that bind a person identify what he cannot
help being . . . Just as the essence of a triangle consists in what it must be, so the
essential nature of a person consists in what he must will. The boundaries of his
will de¬ne his shape as a person.8

Frankfurt™s discussion speaks to what I have called a person™s self-
conception or self-identity. He locates the core of self-identity in the will “
that is, in the desires, preferences, and attachments a person wants to be
motivated by. Speci¬cally, self-identity is fashioned out of, and delineated
by, certain types of higher-order desires “ namely, those that we make in-
eliminable because of our evaluative commitment to them.9 Frankfurt™s
idea is that we cannot help but will certain states of affairs because we care
deeply “ inextricably “ about them. To do or to be otherwise is simply
Assuming this is right, we each have a distinct volitional character by
virtue of which we can make choices. Three questions arise. One, how
is this volitional character disclosed to us? Two, what distinguishes the
volitionally necessary aspects that constitute a person™s self-conception,
and not just her identity? Three, what is the nature of this necessitation?
A person™s volitional character can be identi¬ed by employing the
following thought experiment. The person asks herself: “What would I
do if confronted with circumstances that tested my values, or required
Autonomy and Self-Identity 81

me to adjust my values? Which of those characteristics and attachments
seemingly vital to my identity would I be willing to abandon even were I
deeply con¬‚icted about doing so? Finally, what would I not repudiate in-
sofar as I remain the person I want to be?” By this test, one selects from the
motivations that constrain her will, and thus her identity, those motives
for action that constrain her because she cares about them. One arrives
at a point where it is impossible, given one™s self-conception, to be even
weakly responsive to subjective reasons to alter those aspects central to
who one considers oneself to be. Such aspects become subjectively inelim-
inable. To paraphrase Gerald Dworkin, it is by ¬rst raising the question of
whether one will accept or reject certain characteristics of one™s identity
that a person™s self-conception or self-identity is revealed; in discovering
what she cannot help but accept or reject, a person de¬nes her nature
and takes responsibility for the kind of person she is.10
The variety of necessitation here is speci¬ed counterfactually. Voli-
tional necessity is contextualized against the circumstances in which a
person ¬nds her will tested. The volitionally necessary aspects of identity
typically become aspects of self-identity because they are authenticated, or
embraced, or cared about. One™s self-conception is conspicuous in those
characteristics and attachments a person could not bring himself to part
with even were he able to do so, in those actions he ¬nds “inconceivable”
to perform, and in those choices he considers “unthinkable,” insofar as he
remains the person he is. But it is a genuine form of necessitation nonethe-
less, given that the characteristics a person ¬nds himself left with are ones
without which he cannot be true to himself.
As a case in point, I know I could never abandon my mother to a life
of great poverty, which would surely be her fate were I not to provide
for her in a modest ¬nancial way. The locution “I could not live with
myself ” is apt, as is “I would not recognize myself. ” Suppose it was ap-
parent that my mother was squandering her income on frivolous and
impulsive purchases. This might constitute good reason to stop provid-
ing for her ¬nancially, since it would be obvious that her income was
not being used in a way that alleviated her impoverishment. In this case,
however, I would be more likely to seek the reason behind her conduct,
and attempt to address it. If the behavior were out of character, I might
insist that my mother seek counseling or medical assistance to determine
whether the behavior resulted from a psychological or physiological im-
pairment. Or if boredom and loneliness were factors, I might encourage
her to seek the diversion of companionship and of pastimes I know her to
Marina A. L. Oshana

One might object that while frivolous spending does not count as a
strong enough reason to motivate me to stop providing for my mother,
other reasons might do so. If I were presented with irrefutable evidence
that my mother was using the income I sent her to supply her grandchild
with dangerous and illicit drugs, I would surely be moved not only to deny
her further assistance but to threaten to report her to the authorities
as well. But my self-conception might well require “ make volitionally
necessary “ that I act on the youngster™s behalf in this case. No doubt I
would feel some measure of guilt “ I might even feel con¬‚icted “ and my
decision would only come after I had exhausted all conceivable alternative
avenues for altering my mother™s behavior. The very fact that I experience
this struggle signals the depth of commitment I feel toward the welfare
of my mother and the depth of my concern that her needs are met. But
I could not live with myself were I to continue to support her ¬nancially.
The circumstances have changed, and mandate a change in response,
but this change re¬‚ects who I am in a deep way.
Of course, should the situation remain as I originally described it, I
could not remain true to myself “ I would not recognize myself as the
person I am “ were I to withhold assistance from my mother. That action
simply is not among my options in the original situation, even though I
may have the motive, the occasion, and the ability to perform it: “Here I
stand; I can do no other.”11
I value the attachment to my parent in a way that contributes meaning
to my life, and my valuing the attachment plays a role in the realization
and sustenance of the conception of myself I embrace. Not providing
¬nancial assistance to my mother is an action to which I am averse, and
happily so; this aversion is something I would not want to lose.12 What is
noteworthy for my autonomy is that in a manner of speaking, the aversion
is irresistible: there are very few circumstances where I could overcome
the aversion consistently with my self-conception. The responsiveness-to-
reasons test establishes that where I do overcome the aversion, it is only
because I cannot remain true to my self-conception by continuing to act
in a bene¬cent way. To do so would signal a greater loss or a marked
revision of my self-conception. But given the circumstances that typically
obtain, the test establishes that I cannot gather the will to perform this
action; I am volitionally limited by the things I care about in this special
way “ people, beliefs, values, affective states “ and these limitations de¬ne,
in part, who I consider myself to be.
More deeply, it may be impossible, given my self-conception, that
I should form an intention to become the kind of person who could
Autonomy and Self-Identity 83

abandon her parent to poverty. Even if it were possible for me to con-
sider acting in this way, I could not want myself to do so. Here I both
“resist the effort to do what I remain deeply averse to doing” and resist
the idea of being de¬ned in any other way.13 As Thomas Nagel claims in
The Possibility of Altruism:

There is nothing regrettable about ¬nding oneself, in the last analysis, left with
something which one cannot choose to accept or reject. What one is left with is
probably just oneself, a core without which there could be no choice belonging to
a person at all. Some unchosen restrictions on choice are among the conditions
of its possibility.14

Ib Circumstantial Necessity
Certain components central to a person™s nature may be inescapable,
not because she cares about them and cannot imagine this being other-
wise consistently with her self-conception, but because they are factors
over whose presence in the person™s life and effect on her life she has
no say. Such factors may be described as ones acquired as the result of
constitutive and circumstantial luck. These include certain idiosyncratic
physical and mental abilities (strengths as well as af¬‚ictions), and the
talents, temperament, gender, sex, ethnicity, and familial relationships
that distinguish a person. For example, the attachment I have to my
mother is one that depends on my biography. These aspects are doubly
inescapable since one does not cultivate them “ they are acquired by
birth, biology, gender, and the like “ and they invariably shape a person™s
The elements of identity that are inescapable in this sense interest me
especially for two reasons. First, they force the question of the extent to
which a person™s self-identity is wedded to whatever happens to be her
identity. For example, to what extent is a person™s self-conception bound
up with the inescapable fact that she is of mixed ethnicity yet demon-
strably “African-American”? To what extent is a person™s self-conception
bound up with the inescapable fact that she is a woman? These aspects
announce to the world who a person is whether or not she accepts these
factors as ¬xing her self-identity. The phenomenon here is not that of
volitional necessity. Most people cannot “forget themselves” with respect
to the racial and gender classi¬cations that bind them, but this is not
because race and gender do constrain a person™s will (though they can
constrain a person™s will).
Second, one may feel alienated from these traits, just as one may feel
empowered by virtue of them. If a person would prefer that some of
Marina A. L. Oshana

these factors did not contribute to her identity in the ways they do, the
problem of alienation with respect to aspects of her self-conception arises.
The question of interest, then, is this: If a person does not want her
identity to include any number of the inescapable aspects of herself, is
her self-identity undermined and in a way that vitiates her autonomy? This
question brings into focus a point of disagreement between the views of
self-identity and autonomy to which I subscribe, and the requirements of
autonomy defended by mainstream theorists. It also points to a possible
asymmetry between endorsement as an element of the unthinkable, or the
volitionally necessary, and endorsement as an element of the inescapable,
at the heart of which are the concepts of authenticity and its antithesis,
alienation. I will address this issue shortly.
The point of this section has been to show that the unthinkable and
the inescapable coalesce into something constitutive of a person™s self-
identity or self-conception. Some of these properties form a person™s
de¬nite volitional character; they provide the limits that anchor her judg-
ment and specify the requirements of her integrity.15 Together, they form
the basis for the self-conception she seeks to express. Some ineliminable
factors are so essential to a person™s self-conception that at a certain level,
authenticity with respect to them ceases to be an issue.

II Why Autonomous Agency Depends Upon
Having a Self-Conception
The concept of autonomy requires a ground, parameters that give the
notion of self-directed choice and action plausibility and coherence. The
practice of autonomy also calls for a ground, something that enables
persons to guide their actions and choices. A person cannot embark
upon a life of autonomy, and autonomous choice and action cannot
commence nor be sustained, where she lacks a de¬nite, if not fully ar-
ticulate, set of objectives, preferences, or principles that enable her life-
plans to be unequivocally his own. Without an antecedent moral, cogni-
tive, and conative structure, we are “vacant of identi¬able tendencies and
constraint . . . unable to deliberate or to make conscientious decisions”16
Being autonomous requires ¬rst and foremost that a person have the
capacity and the disposition to know her will and know which of her
beliefs, desires, affective states, relationships, and so on are distinctive of
and essential to her self-conception. Most importantly, the autonomous
agent knows the aspects of her self-identity on which she can rely, or which
she is con¬dent will manifest and be effective, for better or for worse, as
Autonomy and Self-Identity 85

the circumstances mandate. The autonomous agent must recognize that
these characteristics and attachments are central to who she is, and to
how she perceives herself, and she must be familiar with the role they
occupy in her world.
I am not claiming that an individual need carry her self-identity in
stark relief at every moment, or even most of the time. Indeed, persons
who are constantly attentive to themselves are not thought to possess a
healthy consciousness of themselves but rather are self-conscious, or in
a state that tends to disable self-motivated action. I am claiming that an
absence of self-re¬‚ection, and an indifference to one™s self-conception,
eclipse autonomous agency. Agent autonomy consists in taking control
of “ or, better, ownership of “ one™s life.17 Someone who does not, as a
rule, acknowledge some cognitive, affective, attitudinal, and behavioral
characteristics and attachments as part of her self-conception, nor con-
cede the absence of others, and who lacks a desire for self-understanding,
if not a capacity for self-evaluation, is not in a position to assume an active
and authoritative voice in the direction of her life. This is because guid-
ing one™s life calls not just for a self, or an identi¬able entity, but for an
agent alive to herself as someone with a particular vision, with plans and
expectations, concerns, values, and commitments that merit and invite a
range of treatment on the part of others and that can be more or less suc-
cessfully realized.18 Having a self-conception provides some assurance “
certainly not complete, but essential “ that a person™s governance over
her life is her own.
It would appear, then, that a ground of autonomy would be one™s
self-conception “ the goals, preferences, or principles of choice “
aspects of the will “ it would be unthinkable to abandon or to repudi-
ate, as well as those constitutive and circumstantial elements it is impos-
sible to escape. Without an antecedent self-conception, even one that is
a work-in-progress, one cannot be autonomous. This is not because the
characteristics constitutive of a person™s self-conception mandate that her
life-choices assume a certain shape or take a particular direction. It is be-
cause these supply a compass for ¬nding the direction of action that best
comports with what is emotionally, imaginatively, and cognitively mean-
ingful to the agent.

III Autonomy and Alienation
The prototypical account of autonomy requires that a person conceive of
herself as someone who can affect the world in light of a perspective and
Marina A. L. Oshana

plan for life that is of her making. A corollary is that an autonomous
agent takes responsibility for her self-identity. But what does taking
responsibility involve, and how does this occur? Speci¬cally, must the
elements constitutive of a person™s self-conception be authentic if she is
to take responsibility for her self-identity and thus for her autonomy?
“Authenticity” is a term employed widely among discussants of agent
autonomy (and responsibility) to refer to a property of the constituents
of personal autonomy. Authenticity is standardly taken to be a function
of the structure of a person™s cognitive states, conative states, or values,
and, more recently, of the attitude of acceptance a person adopts toward
the genesis of these cognitive states, conative states, or values. The idea is
that a person is autonomous if she is moved by values, desires, beliefs, and
attitudes that would withstand unimpaired self-scrutiny. Presumably, the
legitimacy of a person™s attachments, partnerships, ethnic and cultural
identity, and social roles can also be authenticated for autonomy by similar
tests. One simply asks how the person regards these phenomena when
they are examined in an unblemished, critical light, or how the person
would regard them were she to re¬‚ect upon their development and their
effect upon her. Are they seen as aspects of herself to which she feels
an af¬nity, or as roles she wants to occupy? Or does she feel disaffected
and estranged from them, to the extent that she repudiates them? If the
former, the story goes, the criterion of authenticity is met, and we can be
secure in the thought that the elements that in¬‚uence the direction of an
agent™s choices and actions are de¬nitively the agent™s own. The key for
autonomy, thus, is whether or not the person feels alienated from those
aspects of herself that affect her choices and actions.19
Making authenticity the hallmark of autonomy forces us to examine
the status of those factors that contribute to our identity and to our self-
conception in ways we cannot escape but that we wish fervently would play
a less essential and focal role. Must a person endorse, be satis¬ed with, or
at least fail to feel alienated from what is volitionally necessary and what is
inescapable if these are to be included within her self-conception? Must
a person endorse, be satis¬ed with, or at least fail to feel alienated from
what is volitionally necessary and what is inescapable if the person is to
be autonomous?
If, as I have suggested, volitional constraints upon a person™s will re-
¬‚ect her self-conception because they survive a counterfactual thought-
experiment test, then some version of authenticity would appear to be
entailed. But in fact it is misleading to claim that the volitionally necessary
aspects of identity typically become aspects of self-identity when and only
Autonomy and Self-Identity 87

when they are authenticated by a process of intentional, critical intro-
spection and self-scrutiny. Frankfurt is vague on this point, but at times
indicates that what provides volitional necessity with the anchoring or
terminating stature it has (one is no longer free to raise the question “is
this what I most want to do?”) is precisely that volitional necessity does
not presuppose the agent™s scrutiny and active endorsement. (Indeed, a
person might even ¬nd herself at odds with some of her deepest volitional
commitments. I might wish, for example, that I were less constrained by
my attachment to my mother. I do not think, however, that the knowledge
that one is dissatis¬ed with the centrality certain volitional constraints oc-
cupy in one™s self-concept need challenge autonomy. I will return to this
point momentarily.)
Additionally, it may be preferable for the view of autonomy I defend
that I make no use of authenticity in grounding volitional necessity and
self-identity, given that I want ultimately to deny authenticity as a condi-
tion of autonomy. Thus John Santiago has questioned the necessity (not
to mention the advisability) of relying on any aspect of Frankfurt™s model
of psychological autonomy in delineating the conditions for agent au-
tonomy, preferring a more thoroughly social or “narrative” anchor for
self-awareness than endorsement or authenticity provide. But I see no
need to throw out the baby with the bath water “ Frankfurtian accounts
contribute a fruitful explanatory analysis of the psychological element of
agent autonomy. The question to ask is not why allow volitional necessity
a central role in self-identity and agential autonomy, but whether that
role, as detailed here, relies on endorsement (explicitly or otherwise). I
deny authenticity qua critically re¬‚ective endorsement or satisfaction.
Others might worry whether authenticity is ever a requirement of the
inescapable circumstantial aspects of a person™s self-conception. It is with
respect to my views on this matter that Jennifer Hawkins has questioned
whether the view of the self that I attribute to Frankfurt is correct.20
Hawkins contends that Frankfurt only includes among the components
of self-identity evaluative commitments and other motivational states.
Mine is a wider notion of self-identity in that I include important de-
scriptive items such as cultural aspects of the self, satisfaction with which
or wholehearted commitment to which are irrelevant for autonomy, and
so not, Hawkins contends, required by Frankfurt. If Hawkins is correct,
then the target view of autonomy and authenticity I am criticizing is not
Frankfurt™s (and not Frankfurtian). But I believe Hawkins is incorrect:
In the ¬rst place, Frankfurt includes among the elements of a person™s
self the non-optional characteristics and commitments that a person
Marina A. L. Oshana

simply ¬nds himself with, or discovers about himself; these are founda-
tional to his higher-order evaluations rather than the product of such
evaluations. An individual™s more general evaluative commitments are
invariably premised on aspects of the self such as race, gender, and sexual
orientation.21 Second, such aspects of the self are not merely descriptive,
since the individual so described cannot avoid evaluating himself under
these very descriptions. In both respects, descriptive elements such as race
and gender are things a person must embrace, repudiate, or take some
adjudicative stance toward. If the authenticity condition of autonomy
is central to and exhaustive of Frankfurt™s (and others™) view of auton-
omy, and authenticity is not needed for autonomy, then the view must be
In short, what is at issue here is whether the standard picture of authen-
ticity can be of use for autonomy in cases where a person™s self-identity
is bound up with facts both inescapable and unwanted. Is a person™s au-
tonomy circumscribed, and her self-identity inauthentic, to the degree
that the boundaries that de¬ne her (to herself and to the world) include
factors she would prefer not ¬x her self-conception? In such cases, the
person cannot help but identify herself via factors that she wishes did
not occupy so central and essential a role in her self-conception. As Rorty
and Wong rightly note, a person can acknowledge the centrality of charac-
teristics and relations to her self-identity, and the tremendous social and
psychological force of these, while valuing neither the characteristics and
relations nor their centrality.22
Consider the phenomenon of race-consciousness, or awareness of the
societal signi¬cance of one™s race. For all the talk of color-blindness,
and despite the fact that race is at best only a quasi-scienti¬c concept,
one™s race is tremendously signi¬cant in so racially strati¬ed a society as
the United States. For persons who are not white, racial identity is so
ingrained that one cannot, some have said, “forget oneself ” or fail to be
appreciative of one™s racialized identity.23 Forgetting one™s race involves
a lapse of self-awareness.
It would seem paradoxical, then, that a person could be capable of
such lapses. An African-American woman such as myself might even be
chronically guilty of such breaches if she were not alive to many of the
norms of the African-American female sub-culture, though others were,
and although she were reminded of these norms in the expectations of
others. Arguably, being alive to one™s racialized identity means accepting
certain norms as appropriate standards for choice and action. But a per-
son might appreciate her heritage and take pride in it (just as she might
Autonomy and Self-Identity 89

experience shame or embarrassment when members of the sub-culture
behave badly), even while failing to be fully part of it. Suppose that being
culturally black is not an aspect of a person™s self-identity, although being
African-American remains an inescapable part of that person™s self-
conception. In my own life, for example, race does not just anchor my
identity “ who I am. Race also anchors my self-conception, and how I
regard myself at the same time it interferes with my self-conception. It
anchors what I stand for, and what I stand behind, at the same time as it
presents obstacles to my realization of these.
It may be impossible for an African-American woman who pursues
a career in (say) philosophy to avoid seeing herself as an anomaly in a
predominantly white (and male) profession. She will, like it or not, be
confronted with the suspicion that she fails to adhere to certain norms
of conduct that are expected of her. And her “failure” to appropriate
the expected accoutrements of her racialized identity may come at sub-
stantial cost “ suspicion or hostility from some members of the black
community, curiosity or patronization from some members of the liberal
white community, anger from those threatened by “uppity” behavior, by
the fact that as a Black woman, she has apparently forgotten what she is
supposed to be like. The Black female academic might feel pressed to
explain herself, and she might experience a con¬‚icted sense of self. But
none of this necessarily comes at a cost of her autonomy, as we shall see.
A member of a racial minority whose professional and personal rela-
tionships are “out of place” will often assume a stance vis-` -vis her racial-
ized identity that is marked by restlessness, if not by outright resistance.
Following Frankfurt, the racial minority might not be satis¬ed with the
central place this ineliminable and unchangeable aspect of herself oc-
cupies in her self-identity. For example, I would prefer that race did not
so essentially inform my self-conception. If I am not actively alienated
from this central identity trait, at least I fail to be wholehearted with re-
spect to its being part of my self-conception. Persons like myself might
not wish to be identi¬ed by some of the norms of the African-American
female sub-culture, nor wish their self-conception to be bound up with
these norms because of how they bind a person. Such norms bind vis-` -visa
the false belief that African American women behave in an identi¬able
way. They bind by what K. Anthony Appiah calls the “scripted” or non-
optional components of collective identity, the point of which is to supply
a person™s life with a certain “narrative unity.”24 These components con-
sist of, I assume, certain values, behavioral norms, practices, and social
Marina A. L. Oshana

One might resist this script, but not because the script is inescapable.
Rather, one might resist it if one does not “choose to make these collective
identities central to individual identities.”26 While such identity casting
is socially attributed, it is not an identity every individual subjectively
accepts.27 I can say that while my racialized identity may be of imper-
sonal value, any personal meaning I invest in it is not that of the collec-
tive identity. In terms suggested by Joseph Raz, my racialized identity is
frequently a tie I ¬nd myself burdened with against my will, and which
I would rather be without but from which I cannot shake myself free.28
That I do not value this scripted identity does not mean that it is of a type
that lacks value, since the value of a type of attachment “does not depend
exclusively on the fact that those whose attachments they are embrace
them willingly or with approval.”29 But that I do not choose to accept this
scripted identity reveals an important component of my self-conception.
It is unthinkable of me to embrace any script that would call upon me to
de¬ne myself against a standard I disavow and wish to divest of personal
I have claimed that a person™s autonomy is grounded in her self-identity
or self-conception, in those components of her identity she cannot re-
pudiate without doing violence to the person she is. And race is very
much an inescapable component in this sense. We cannot escape the
racialized norms that de¬ne us, and that inform our self-concept, even
where we regard these norms as alien. Consciously or not, welcome or
not, one™s racialized identity contravenes upon most aspects of one™s self-
conception. The concern is whether one can reconcile the person one
takes herself to be with social expectations of who one is (and who one
ought to be). Can a person, for example, be autonomous despite the fact
that she does not endorse, or wholeheartedly identify with, an aspect of
her character that is essential to her self-identity? Does the person who
forgets her race, or gender, lack a “healthy, authentic psychology”?31 Is
such a person condemned to a diminished state of autonomy as a result
of this forgetfulness?

IV How One™s Self-Conception of the Ineliminable
Might Impair Autonomy.
IVa Threats to Autonomy
I am not convinced that every instance of forgetting oneself, even where
the object of one™s lapse is as integral to a person™s self-conception as is
race, robs a person of a “healthy, authentic psychology.” Suppose that I
Autonomy and Self-Identity 91

regarded neither my African-American heritage nor any accompanying
norms as subjectively ineliminable aspects of my self-conception. This is
perhaps an uncommon phenomenon, one that is dif¬cult to maintain for
precisely the reason that factors such as race are so entrenched in our so-
ciety. But there are cases of persons for whom this is true, and one would
be foolish to claim that this alone yields diminished autonomy. What
would yield diminished autonomy would be to deny my African-American
heritage and any associated norms their centrality in my public life and
within the relational positions I occupy. By denying an essential identity-
forming aspect, I would not only fail to attend to the manner in which
racial narratives play out in social interaction but would falsify myself.
This would signal a kind of self-betrayal or self-deception, an attempt to
defeat my identity.32 And the effect would be a kind of practical disabil-
ity, making self-management a more complicated endeavor. But while
self-betrayal and self-deception are disabling, they are very different phe-
nomena from that of acknowledging, with eyes wide open, the experi-
ence of disaffectedness from certain aspects of one™s self-conception.33
The former injures autonomy, whereas the latter does not.
Acknowledgment of race aside, the fact that a person might wish to
escape the grip of these scripted identities does not of itself gainsay
autonomy any more than the fact that they are inescapable does so.
Insofar as ineliminable and scripted characteristics of a person™s self-
conception such as racial identity and race consciousness impugn auton-
omy, they do so for reasons quite different from threats to psychological
To understand this, we need to note that certain traits of character
that are intractable aspects of one™s self-conception “ either because their
absence is unthinkable, or because they are inescapable “ are only part
of what grounds our autonomy. For a person™s self-conception might fail
to represent accurately her social and psychological circumstances. And
it is these circumstances and the type of life they permit that are germane
to autonomy, not the fact that aspects such as racial identity and race
consciousness are scripted, ineliminable, or unwanted.
To be autonomous is to stand in a certain position of authority over
one™s life with respect to others. Thus, if a person is to be autonomous,
the circumstances to which he authentically assents must grant him the
latitude to choose and to live in a self-directed fashion. Racial identity
doesn™t always allow this, as we all know. K. Anthony Appiah charges
that in the context of a racist society, “it will not even be enough to
require being treated with equal dignity despite being Black for that will
Marina A. L. Oshana

require a concession that being Black counts naturally or to some degree
against one™s dignity. And so one will end up asking to be respected as a
Black.”34 No matter how successful a person might be in liberating herself
from the psychological appurtenances of race, or in maintaining race-
consciousness free of racial self-consciousness, or in appropriating her
racially scripted identity so as to give it the stamp of authenticity, race can
encumber a person in a fashion antithetical to autonomy. Being Black in
a racist society situates one in a position that narrows the range of one™s
autonomy even if being Black is not in itself antithetical to autonomy.
Autonomy requires that equilibrium of power be effected by the agent
between herself and society. The possibility of effecting such equilibrium
and the ease with which this is achieved depends largely on the energy
that social navigation requires. The invasive quality of racial scripting to
self-management stems from the fact that racial scripting more often than
not is disabling in practice. It is not enough for autonomy that a person au-
thentically embrace the social constraints mandated by the inescapable
aspects of her life, for the fact that she ¬nds these constraints accept-
able does not mean they are acceptable or adequate for self-governance.
One™s self-conception as a member of a marginalized group, and the
very grounds that nurture this self-conception, can frustrate autonomy,
in part because autonomy calls for social recognition and respect of a sort
“scripting” often impedes, even where one™s self-conception is authenti-
cally her own as mainstream accounts require.
Accordingly, navigating the de¬ned contours of one™s racial identity
may require that a person forget herself in order to be herself “ in order
to keep to the self-conception that af¬rms and sustains her autonomy.
The African-American academic, for example, might be obliged to break
away from the decorum expected of a black woman, if this decorum in-
cludes, say, eshewing scholarship on the work of colonial and pre-colonial
European men, or forgoing intimate associations with white people. For-
getting oneself may be the way the African-American female academic
must live if her autonomy is to ¬‚ourish. Autonomy requires a person
having the freedom to distance herself, or to step back, from the socially
given roles and practices that contribute to her identity.
The lesson is not that the self-conception of the autonomous agent
must be free of the effect of ineliminable forces upon identity such as race.
Rather, the lesson is that because being autonomous requires, in typical
cases, that a person be in a certain kind of social network, what can decide
autonomy is the effect factors such as social roles and characteristics such
as allegiance to members of one™s racial group and a commitment to
Autonomy and Self-Identity 93

notions of correct racial behavior have on one™s life. Being subject to racial
pro¬ling, for example, both by those in one™s own race and by those in a
dominant race, frustrates autonomy, even for the person who subjectively
identi¬es with and values the centrality of race and of racialized norms
in her self-conception.

IVb Why Authenticity Is Not Needed for Autonomy
I noted earlier that I part company with recent accounts of autonomy over
the question of how the phenomena of authenticity and alienation affect


. 3
( 13)