1.

One difficulty for theories of identity rests in the common confusion between ideals of human flourishing and criteria of personhood. We would like to have integrity, for instance, and so some insist that integrity is an essential part of being a person. Such mistakes have framed debates between E-theorists and D-theorists, including the one between Derek Parfit and Marya Schechtman from a few decades ago. Parfit (Reference Parfit1984) defended an E-theory, and in doing so he gave a moral injunction to exercise detachment from our past (or future). His more focused arguments mounted a compelling challenge to the inference from psychological continuity to metaphysical identity: the fact that my conscious life is continuous with a certain past does not imply that I am uniquely attached to that past or that I am in any deep way the same person who experienced that past. Continuity is to be distinguished from ‘connectedness’, just as a conversation can continue without any of the same participants or topics (and so without a definitive answer to the question whether it is the same conversation). One can appreciate these points without Parfit's otherworldly thought-experiments about fission and fusion or the suggestion that we may have memories (quasi-memories or ‘q-memories’) not connected to our past. But even granting that one cannot derive personal identity from memory, this says nothing about how particular persons should regard their histories.

In her early career Schechtman (Reference Schechtman1990, Reference Schechtman1996) made an industry of uncovering some of the confusions of Parfit's argumentative machinery. She dismissed the fission and fusion experiments by insisting, rightly, that memories are individualized. We cannot reasonably imagine anything like two people sharing a memory (and hence sharing a past through fission), since differences in personal history before or after the remembered event will be reflected in how the event is remembered. Memory content cannot be abstracted in the simplistic way that Parfit demanded, as if it were a still in a film sequence. But this point about the individuation of memories does not imply anything about what Parfit called connectedness to the past, and Schechtman (Reference Schechtman2001) later supplied the missing step by arguing that memory allows for ‘empathic access’. Her assumption was that in order for my past actions to be really mine, I must have some direct relationship to them. In Parfit's better moments, on the other hand, he merely requested of his readers that they take a more detached stance toward their past and future. The q-memory arguments can be taken as a dispensable propaedeutic to this point. The dispute between Schechtman and Parfit hinged ultimately on which approach, D-theory or E-theory, provides a better picture of human flourishing.

A later polemic by Galen Strawson (Reference Strawson2004) highlighted the elitist implications of this whole debate. Schechtman and Parfit were really arguing about whether persons should feel connected to their remembered pasts, and there is no easy answer to this question. Psychologists have since confirmed that some people, whom Strawson labels ‘episodics’, look to the past with a sense of detachment, whereas others find a deep connection in their memories (Klein and Nichols [Reference Klein and Nichols2012] summarize this research). Episodics, for instance, ask not ‘which actions of my past are really mine?’ but rather ‘how could I have done such things?’ Perhaps they answer ‘it doesn't even seem like me at all’, or they simply shrug in indifference as they inhale the scent of the fleeting present. Strawson concluded, quite reasonably, that an episodic is not for that matter less a person than is the self-constituting, authentic practitioner of Schechtman-style existentialism; hence, Strawson (Reference Strawson2004: 447) writes, ‘On the strong form of Schechtman's view, I am not really a person’. But it is also the case that some diachronics have good reasons to invest in the past. A liberated child soldier or a victim of sexual abuse is not always best served by living in the moment and exercising an easy detachment from remembered events. These considerations should discourage inferences from ethical ideals to metaphysical conclusions (or vice versa), if for no other reason than that ethical ideals—including Frankfurt's (Reference Frankfurt and Schoeman1987) ‘wholeheartedness’ but also Strawson's happy-go-lucky episodic—reflect parochial preferences.

Distinguishing between two senses of the verb ‘to identify’ helps to separate these issues of personal identity, human flourishing, and self-knowledge. In this essay I employ ‘identify*’ for the intransitive sense of ‘identify with’ as in ‘he identifies with how you feel’ or ‘she identifies with the downtrodden’. Many theorists assume that our ability to identify* in this sense has something to do with personhood, but it is fairly clear from the examples that it does not. We identify* with all sorts of things, and even when we identify*, as D-theorists wish, with our own past, that past is sometimes imagined or falsified. Some psychologists even encourage counterfactual autobiography, so that people who identify* with a nonexistent past are not all delusional (Campbell Reference Campbell1997). Identification* is a mental act, albeit one that often comes with ethical requirements (in Schechtman's [Reference Schechtman2001: 102] definition ‘empathic access’ is a sort of ‘affective remembering of the past, together with its behavioral implications’). In identifying* with my own history I summon the passions of earlier pursuits, do what I should have done then, and generally feel one with my earlier apparitions. Identification*, ‘empathic access’, and ‘psychological connectedness’ all refer to a sort of solidarity, the likes of which can be experienced in regard to someone or something that is historically and biologically distinct from the self or even nonexistent. D-theorists tend to believe that such solidarity with our actual past is crucial to human flourishing. But even if that thesis is true, it does not license the further conclusion that this is what defines a person as a person. There are many persons and types of person for whom dissociation from the remembered or actual past is of great importance.

‘Identify’ without an asterisk will be used in the transitive sense of ‘to specify an object for description’, as in ‘we have identified the problem’. My point of calling a theory of self-knowledge externalist is to stress that it requires that we identify our past only in this sense. Knowledge of the past does not require identification* with it, and knowledge of our own case provides no exception. Self-knowledge without identification* can satisfy the strictest criteria of meaningfulness, and it can underwrite moral obligations. The American historian who studies the slave trade does not identify* with seventeenth-century plantation owners, but she does learn to appropriate moral responsibility for racial injustice. Some episodics are likewise storytellers about their own past, and some storytelling (e.g., confession of long past criminal activity) has a structural requirement that the storyteller does not identify* with any character, even where one character is the same person as the storyteller. The issue is only whether and how persons give reliable descriptions of their past in narrative form, not whether the act of remembering implies any deep connection to it. The big question that narrative self-examination can answer is ‘how did I ever get here?’ And the answer to this may or may not tell us what the meaning of it all is.

2.

One outcome of these conceptual distinctions is that storytelling is separable from identification*, so that we have the possibility of what Strawson would label the storytelling episodic. Strawson distinguished episodics from diachronics in regard to styles of self-experience, since some people identify* with their memories more strongly than others do. This distinction, however, seemed to run together with the observations that some people tell stories about their past whereas others do not, and that some people may have more self-knowledge than others. But there is no prima facie reason why an episodic should tell fewer stories, just as there are no obvious reasons to think that one type of person should have more self-knowledge. For example, our reformed criminal and our recovered drug addict may each confess stories about the past with a high level of dissociation, and these confessions may express sophisticated forms of self-knowledge. There are thus people who know their past without identifying* with it, just as there are probably people who identify* with a past but have mostly false beliefs about it.

There are also people who opt not to tell stories of the past, and this is to be held distinct from the question of whether those people identify* with their past selves. In the original text of ‘Against Narrativity’ Strawson recited a striking argument to the effect that non-storytellers have a better understanding of themselves than do storytellers.Footnote ¹ He considers what I call the skeptical argument from revision:

1. 1. Memory in human beings is deeply revisionary.

2. 2. If we tell stories, we remember more.

3. 3. Therefore, if we tell stories we revise more.

4. 4. If we revise more (by telling stories), then our sense of self is less accurate.

5. 5. Therefore, to know ourselves it is better not to tell stories about ourselves.

The conclusion to this argument has the appearance, but only the appearance, of paradox. It would be unfortunate if the effort to know about ourselves through sharing stories of the past were a futile endeavor, but some cognitive endeavors are self-defeating in similar ways. Correcting a language-learner's grammar, it has long been believed, inhibits rather than aids language acquisition (see Hendrickson Reference Hendrickson1978); psychologists have claimed that some classes of decision are better executed when they are not consciously deliberated beforehand (Dijksterhuis, Bos, Nordgren, and van Baaren Reference Dijksterhuis, Bos, Nordgren and van2006); and common experience provides cognate phenomena for some morally questionable activities, such as the liar who complicates his situation by telling more lies or the overly defensive person who moves further from a just defense the more she advocates her cause. Strawson merely called attention to the fact that personal narrative is potentially an activity of this sort due to the fact that memory is in some ways epistemically unreliable.

The premise that ‘memory in human beings is deeply revisionary’ is both harmless and hard to dispute. It admits of elaborations in a variety of contexts. Each instance of recalling an event, as Strawson related, brings about physiological alterations. Many studies also show that sharing a memory leads to alterations of its semantic content. Your storytelling friend will likewise change the details slightly each time she tells the story. Even if she is not especially prone to embellishment, she will accumulate changes that eventually (1) make later retellings markedly different from earlier ones and (2) cause her to become less clearly aware of how she originally experienced the event. She will not be sure which features she experienced the first time and which were added on through later retellings.

The second premise of the proposed argument is nearly trivially true, and the third follows with some qualification from the first two. By frequently retelling her story, your friend changes her perspective on it more significantly than she would if she never retold it or retold it only a few times. Some psychologists (Freeman Reference Freeman2010:70) have speculated that there is a threshold of story repetition, after which the story tends to become heavily stylized or conventionalized; we should be wary of any story told very many times. Storytellers are therefore in one respect all revisionary, even though some storytellers are classed by Strawson as ‘revisionary’ and others are not. Some people embellish or justify their past in order to impress their friends or conceal their wrongdoings. Perhaps others are not revisionary in these morally questionable senses, but if they dare to story-tell, then they are nonetheless revisionary in the more basic sense.

It is the fourth step of Strawson's argument that is subject to debate. He argues that

1. 4. If we revise more (by storytelling), then our sense of self is less accurate.

This thesis, however, does not follow from premises 1–3 above, nor is it supported by additional psychological literature. To be sure, it follows from 1–3 that storytellers have a sense of past actions or events that differs markedly from how they previously viewed those actions and events, and this threatens the possibility of identifying* with the past self about whom we tell stories. But this is where the argument begins, not where it ends. Most theorists of narrative have considered our experiences of the present to be incomplete or epistemically limited. Strawson, on the other hand, takes over an assumption from the earlier empiricist tradition in epistemology concerning how perception plays a foundational role in knowledge. He adopts the Lockean view according to which memory is simply a ‘revived’ perception (Locke Reference Locke1689: X.2). In that case, the farther we depart from any given experience in terms of content or time, the more problematic our situation becomes. Our past experience of past events (the original perception) marks the standard against which the memory is to be measured, and on this test it indeed scores poorly, and more poorly, it seems, the more that psychologists learn about how memory works. But there is no deep philosophical injunction to assess retrospection on the model of observation, and some research in the philosophy of psychology has discouraged the assimilation of memory to perception.Footnote ²

Do storytellers have a less accurate sense of self? Nothing of this sort follows. But if premises 1–3 are valid, then storytelling is revisionary. And if storytelling is revisionary, then storytellers cannot reliably recount stories from the standpoint of past agents, even if the storyteller is the same person as one of the agents. In that case, the argument from revision imposes a serious condition on diachronic self-knowledge, implying that the storyteller resist the temptation to identify* with her past self.

3.

One rhetorical flaw in narrative theories results from the fact that common parlance employs the word ‘narrative’ in a dismissive way, as when we discover that a person misunderstands herself by adopting an all-too-common trope. A friend of mine, for instance, is currently struggling with the self-understanding of his younger brother. A law student at an Ivy League university, the young man is on the brink of considerable worldly success. But an unpleasant experience as a volunteer has led him to overuse the concept of desert in regard to material wealth, and he has concluded that the plight of the poor is a product of their own laziness. My friend expresses this point by telling me that his brother ‘has adopted the narrative of the self-made man’. In indicating his brother's self-understanding, my friend seems to mean both that the adoption of this narrative structure is a contingent matter and that its application to his brother's case is questionable.

Recent psychological work has supported this suspicious turn of common parlance. One of the more interesting theses in the current literature is that memories, and so the narratives based upon them, are self-fulfilling in relation to the self-conception of the person in question. This motivates a skeptical argument that is in some respects even more dangerous than the one based on revision, which we can call the skeptical argument from selection:

1. 1. The selection of memory episodes is biased in regard to our self-concept.

2. 2. Therefore, such episodes cannot serve as sufficient evidence of that concept.

3. 3. We often have independent evidence for the inaccuracy of someone's self-image.

4. 4. Therefore, we have good reason to doubt the stories that people tell about themselves.

Psychologists (Singer, Blagov, and Berry Reference Singer, Blagov and Berry2013) have sought to explain the first premise by postulating distinct memory systems for particular events and for self-images or concepts. Memories of particular events are called ‘episodic memories’, and it seems that even children possess these long before the development of autobiographical memory. When autobiographical memory develops, probably in early adolescence (Habermas Reference Habermas2011), persons acquire more or less malleable self-concepts that serve as filters determining which particular events are apt to be recalled. These concepts are supposed to form a different system because they possess a degree of independence from the episodic memories. The conclusion of this interpretation of the research, then, is that the autobiographical narratives we adopt have a prohibitive effect on memory recall, excluding many episodes that would bring our narrative into question. A woman who has adopted the narrative of an individualist, or of an independent woman (perhaps one who identifies* with Beyoncé's music), will be more prone to recall moments when she rejected men, gained financial independence, or other such fitting memories. Her recall is under the grip of her narrative, just as in more general historiography an investigator can look for and discover true descriptions of past events that best match the initial research hypothesis.

No one should find the argument, to this point, either surprising or distressing. We learned when we were very young not to trust naively the self-serving stories that people tell, and it is common for people to receive each other's tales with a degree of skepticism. The surprising element enters only when we consider that we have no reason to restrict the argument to other people:

1. 5. If my self-concept has selected memories in a manner over which I have no control, then my own memories would provide no evidence of that concept.

2. 6. But there is no apparent basis on which I could, with warrant, exempt myself from the rule that persons filter memories in accordance with their self-concepts.

3. 7. Therefore, I should not treat my stories of my past as evidence of the kind of person I am.

The proposed argument concludes that we have good reason to be suspicious of our own stories, at least in terms of their role as evidence if not also (as the argument from revision has it) in terms of their actual content. I should not treat my memories as support for my self-concept because the research tells me that my mind is going around my back, as it were, selecting confirmatory memories and ignoring disconfirmatory ones. This conclusion more clearly poses a dilemma for diachronic self-knowledge. Note that the argument does not require that we have specific evidence, in our own case, that our sense of self is inaccurate or flawed. My stories about my humble beginnings and subsequent rise to prominence cannot overturn my belief that I am self-made. But nor can these stories serve as evidence for the belief, since the very belief is partly responsible for their accessibility as remembered stories.

Since the argument does not imply that our self-concepts are always or usually false, it says nothing about whether we in fact have self-knowledge. It implies only that we do not attain such knowledge directly by means of recalling stories. The fact that I am able to summon a particular version of my past actions does not, in isolation, provide evidence of the truth of that version. Autobiographical storytelling is radically fallible, and the memories that appear to support it lack the characteristics that philosophers typically have sought in introspection: storytelling about our own past, no matter how vivid that past feels and no matter how much we incline toward identification* with it, is neither transparent, immediate, nor certain. To whatever extent we wish to know about our past, then, we should become as skeptical toward our own stories as we tend to be about those of other people.

The question still stands, however, whether telling more stories complicates rather than improves our situation. Defenders of the argument from revision will likely see the argument from selection as supporting their case. Perhaps storytellers, to exacerbate Strawson's initial worry, pack lies on top of lies; they revise and distort memories that are subject to selection bias in the first place. But this worry overlooks one of the premises of selection skepticism: we can sometimes tell the difference between true and false, or supporting and nonsupporting, narratives. This is why we can assert the premise that ‘we often have independent evidence for the inaccuracy of someone's self-image’. People have the capacity to discriminate true narratives from false ones, and so the skepticism implied by the argument from selection is not global. It suggests rather that aspects of our past are like the fronts of our noses, invisible from our own perspective but clearly seen by those around us.

The argument from selection presents an attack on first-personal bias without threatening narrative self-knowledge as such. There is perhaps a form of asymmetric and even privileged access in self-knowledge since no one else can make all the same statements I can make about my past with the same warrant (not having been there in my past, not having direct access to my memory, etc.). But my stories about my past are corrigible and amendable, so that others can and do correct me in my storytelling. Our self-concepts motivate biased, perhaps even misleading, recollections of the past. Nonetheless, it follows from this only that narratives based on memory serve as evidence in self-knowledge only if other people broadly corroborate what we tell—the other people not being subject to the same selection bias in regard to our case. We attain self-knowledge mainly by sharing stories with others and allowing the stories to be corrected by them. Hence my friend can remind his brother that the three generations of lawyers in their family have contributed to his academic success, and your therapist (or your friends) can assist you in uncovering episodes that your self-narratives exclude.

The deeper skepticism proposed by the argument from selection rather motivates the investigation of our otherwise lost memories and so turns the argument from revision on its head (insofar as the latter appeared to recommend less storytelling). The important point is that people are able, in the best cases, to overturn false narratives. The public record of this point lies in the contemporary genre of memoir and in other instances of the therapeutic investigation of forgotten episodes. But the practice is much more common than these references suggest, and we have evidence of it on every appropriately critical discussion of our past. Public and social activities of storytelling require rather than preclude the kind of suspicion of our narratives that the argument from selection motivates: when I tell a story of my past, it is as if I am offering a hypothesis for scientific debate and inviting my interlocutors to correct my position. The argument also suggests that I should approach all autobiographical stories, those of others as well as my own, with a healthy but suspicious attitude toward the authority of the storyteller.

4.

Our skeptical arguments instruct the storyteller to refrain from identification* and corroborate her tales, but they say little about which stories are candidates for corroboration in the first place. In this section I argue that there are important classes of event descriptions that preclude identification*, quite independently of any considerations about how memory contents may be distorted. Such event descriptions become assertible only retrospectively (retrospective descriptions) and so do not permit the kind of comparisons with past experiences at which the argument from revision takes aim. Retrospective descriptions differ importantly from descriptions that are available contemporaneously to the event (contemporaneous descriptions). This distinction provides a basic stock of diachronic or retrospective concepts, and my contention will be that narrative self-knowledge consists in the appropriate application of these concepts. Diachronic self-knowledge concerns those things about which persons tell stories: friendships, affairs, overcoming obstacles, bouts of depression, personal triumphs, etc. The assertibility of descriptions of these significant items, I argue, floats free from the ever-receding contemporaneous descriptions at which the argument from revision takes aim.

The distinction between retrospective and contemporaneous descriptions follows an old observation by philosophers of action. Many in the tradition of Elizabeth Anscombe (Reference Anscombe and Black1950) and Donald Davidson (Reference Davidson1967a, Reference Davidson and Rescherb) argued that there is a class of retrospective descriptions in that some actions are described in terms of temporally variant effects. If A stabs B, after which B lives for a month, the action is called a stabbing in that interim. If the victim then dies and the coroner declares the stabbing to be the cause, then the stabbing becomes a killing, and we say that ‘A killed B’. ‘Killing’, however, is not a linguistic item that has intrinsically diachronic identity conditions, but one rather that depends only on identifying an action by a certain effect. It can, but need not, denote an action that took a long time to complete. Most important, this example presumes that the fact of the stabbing is not subject to question.

‘Killing’ denotes an action that is observable under the condition that the effect is apparent; hence, we speak of witnesses to murders. Some retrospective descriptions differ in that they are not observable in the same sense: there can be a witness to a friendship, a recovery, or a war, but the people who witness these are engaged in a very different activity. Consider, following Arthur Danto (Reference Danto1965), the notion of an ideal chronicler who is able to record precise descriptions of every event in the universe as it occurs but who has no knowledge of future events. That sapient individual would not know, in 1618 or even 1628, that ‘The Thirty Years War began in 1618’. It is arguable, but not important for my exposition here, that the chronicler could not even determine whether such a cumulative and collective event as a ‘war’ took place. (This depends, of course, on what counts as an observable phenomenon.) The chronicle would include rather statements such as ‘X defenestrates Y’, ‘Y falls’, ‘Y bleeds’, ‘Y dies’, etc. No one could have described these as episodes of the Thirty Years War until at least 1648, because no one could have identified the set of events as a set that composes that object. In this case we are not, as in the case of Davidson's killer, describing an event by its effects but rather as an episode in a much larger event.

Retrospective descriptions that escape the argument from revision, however, are neither observable in the manner of ‘killing’ nor trivial in the manner of Danto's examples. ‘The Thirty Years War’ forces the point by making it seem as if historical knowledge is somehow contingent on there being an explicit temporal quantifier in the subject of the sentence. Historical terms (e.g., ‘the Second World War’, ‘the Renaissance’, or ‘pre-Socratic Philosophy’) frequently have a kind of retrospective taxonomy built into them, a fact that does not by itself establish a deep philosophical point about the nature of retrospection. A similar example is the Cambridge-style sentence that describes one event as having preceded another, but with little else of importance resting on this chronology: ‘The Battle of Liege was earlier than the attack on Normandy’. One may also place items in a sequence and then describe one event in regard to its place in that list (‘Ronald Reagan was the fortieth of our forty-four presidents’) without it being the case that we thereby know anything significant about the described item. These descriptions do float free of minor variations in the truth-value of contemporaneous descriptions: you can get a lot wrong about Reagan without that affecting the fact that his presidency was the fortieth. But the strictly retrospective aspect of the description (there have since been four presidencies) is insignificant for most who would inquire into Reagan's tenure.

In the above cases it is clear that the past was better known in the past, even if it was not known under our particular comparative descriptions. The participants in the First World War could not have referred to it as such, but their knowledge of its events was probably superior to our knowledge of it except in regard to either its temporally removed effects or some Cambridge-style taxonomical properties. There are more important kinds of retrospective description, some of which appeal to the significance an event has as a result of its place in a sequence. Historians of soccer will assert that ‘Thierry Henry played his last match in 2014’. This description concerns the place of the entire sequence of events (all the games he played) in relation to factors outside that sequence, such as Henry's place in the hearts of his fans or the place of his career in his own life. Other descriptions will situate an event in a sequence, and in doing so attribute to it a causal role not only for other items in the sequence, but for a range of other events. ‘The night of the dance was our first kiss’ can, in context, imply that the kiss bears a causal relationship not only to other kisses but also to marriage, divorce, reproduction, or many other (cross-sortal) items. These aspects of the description do not rely directly on quantifying the subsentential terms, and they plausibly, but only by implication, describe significant aspects of the events in question.

For some cases of retrospective description the described aspect is significant, explicit, and (relatively) epistemically independent of contemporaneous descriptions: ‘I met my best friend at the playground’ describes one event by reference to a range of interrelated events of diverse sorts, and descriptions of those events acquire a precise meaning under the conceptual umbrella ‘friendship’. Bundling a range of events and actions over time in turn identifies the friendship as a friendship. The question of narrative self-knowledge is simply the question of how we arrive reliably at descriptions of this sort, where the corresponding contemporaneous description was of the type ‘I sat on the swing next to the curly-haired girl’. The latter is the claim of naïve experience, assertible in the immediate context, whereas the former is a thoughtful reflection informed by wise experience. It requires only that we order past events or actions in accordance with a conceptual apparatus (‘friendship’) already present in our language. Coming to formulate, reliably, sentences of this sort requires that we track long conjunctive claims about friendships, drug addictions, bouts of depression, periods of recovery, traumatic events, learning moments, etc. Such retrospective descriptions tend not to be brought into question by minor variations in episodic memory content: we can forget or revise many of the details of sitting on the swing without falsifying the assertion that ‘I met my best friend on the playground’.

There are some true stories that are not undermined by minor falsifications in the details, and some of these have retrospective assertibility conditions. Some of these stories concern ourselves, as when we say that we met our best friend on the playground. There are thus some items of descriptive self-knowledge that are immune to memory distortion. This provides us with a useful reply to the argument from revision, one that has the benefit of resolving some further debates between narrativists and E-theorists. These philosophers have disputed whether episodic persons can have friendships. Anthony Rudd (Reference Rudd2007), following Kathleen Wilkes (Reference Wilkes1998), argued that friendship requires loyalty, and that loyalty in turn requires a strong level of identification* with the past. This was exactly the move that Strawson (Reference Strawson2004) had classified as elitist, and he argued from literary evidence that some episodic persons have a gift for friendship. He then proposed (2007) a theory of ‘episodic ethics’ in response to the narrativists. But the whole argument rests on the fact that both parties overstress the mental act of identification* when it suffices that people can offer true retrospective descriptions. It is not a condition of friendship that each of the friends identify* with their past selves, but rather that the friendship be identified diachronically and described retrospectively. If your friend experiences alienation from her past, you do not thereby have grounds for ending the friendship. What matters is only that the right kinds of actions and events have accumulated such that the relationship between the persons can be properly identified, by any interested party, as a friendship.Footnote ³ So, pace Strawson, the fact that episodics can be friends does not imply that ‘friendship’ is not intrinsically diachronic.

5.

The antirealist connotations of ‘narrative’ present a further rhetorical obstacle that has led many theorists astray: a narrative is a thing that people make or invent; it thus seems as if narrativists wish to inform the world that persons are not real, but rather inventions in the sense of fictive stories such as we find in literature and on television. The metaphor works for those, such as Daniel Dennett, who believe that persons are indeed fictions invented by biological organisms. Dennett (Reference Dennett1991) concluded that we, in a certain sense of ‘we’, do not exist. But this works less well when narrativism is not a defense of eliminativism or when it does not insist upon some biological criterion of existential commitment. The person who achieves a specifically narrative form of diachronic self-knowledge does not invent herself in the way that a storyteller invents a story, but rather she specifies sequences of significant actions or events that cast the past under new and previously unavailable descriptions. When she tells stories about herself, she nonetheless tells of things that really exist and whose existence precedes the storytelling: I really had the friendship and did not first construct it by relating the story. My knowing that there was a friendship is perhaps a condition of my accepting certain responsibilities, and my telling stories about the relevant past is perhaps a condition of both. But neither the knowledge nor the storytelling is a condition of the friendship existing in the first place. Sometimes people even inquire together into these matters, as when someone asks ‘are we friends?’—a question I take as a real one about personal history requiring shared investigation, rather than as a disguised form of the question ‘will you be my friend?’

This antirealist mistake has a further consequence in that it disguises the fact that diachronic self-knowledge is needed precisely to the extent that persons undergo change: new retrospective descriptions arise only because change has taken place. Meeting a stranger, in my example, became meeting my best friend after a subsequent sequence of events was completed. And a consequence of this is that stories are as revisionary as Strawson feared, except that the fact does not pose a problem for their epistemic value. The revisionary aspect of storytelling has often, but wrongly, served as reason to infer that histories (whether or not they are about ourselves) are constructs and even fictional. The locus classicus of this debate is a passage by Louis Mink ([Reference Mink, Fay, Golob and Vann1970] 1987: 60):

The examples in this passage illustrate the retrospective character of stories rather than their fictive character. The historian does not invent the fact that Columbus discovered America. Rather, she identifies a geopolitical object called ‘America’, and she locates Columbus's voyage within the chain of causally significant events in its history. One could not have said of Columbus in 1493 (if one had been in his hometown), ‘he's not around, because he's busy discovering America’, and this is not merely a matter of ignorance on the part of fifteenth-century Genoans. One could have described the voyage in 1493, but in order to call it ‘discovering America’ one needed to frame it in the context of a vast sequence of significant events (events not otherwise falling under a type) available only to interested persons with several centuries of retrospect. In 1493 the voyage likewise was not yet a candidate for explanations of world politics whereas subsequent changes in the geopolitical item ‘America’ brought it about that the event acquired such explanatory power.

Having mistaken retrospective description for fiction, some narrativists then argued that persons really are stories. This led in turn to the false suggestion that (assuming also the unity of the person) there is a single, correct narrative of my life that exists regardless of whether anyone ever tells it (this is the train of thought MacIntyre (Reference MacIntyre1981) follows in After Virtue, subsequent to his quotation of Mink). Such a view is impracticable as a philosophy of history and needlessly romantic as a theory of identity. At best one could claim that there is a unique history to be told for each object at each time (from each perspective): a correct history of America today, but a different one tomorrow. It could even turn out that alterations in America's race relations remove Columbus's voyage from its role in our past, just as a person who experiences a traumatic event can make that event less significant by recovering from its effects. Self-constitution theories of identity, which claim that the stories we tell (or some other acts we perform) somehow constitute the objects about which we tell them, are impracticable precisely because they are incompatible with the fact that objects—and hence which retrospective descriptions are true—change over time.

This emphasis on the change that the subject of history undergoes has implications for the attempt to identify* with our past selves in stories. My ability to tell reliably about my past is not only consistent with change, but rather requires change. There is a sense, then, in which I reliably tell about my past only to the extent that I do not immediately identify* with it. When I write about my past, for instance, I do not identify* qua writer with the character that represents me in the story. The memoirist's situation is similar to that of any historian: it requires a careful distinction between the writer, the narrator, and the main character (Birkets Reference Birkets2008). And whether this distinction leaves room for further acts of empathic access or psychological connectedness is a separate question that depends partly on how much the person in question has changed. A person who stayed the same her entire life and who possessed a maximal degree of psychological connectedness probably would have very poor material for a memoir, though she may indeed have integrity and excel at the ethical characteristic that Schechtman and others misleadingly call ‘self-constitution’. Likewise the theory of historiography requires that the historian distinguish her perspective from that of the persons investigated (Ankersmit Reference Ankersmit2012). The fact that, in the case of self-knowledge, these are the same person does not reduce the need for this central element of dissociation.

6.

These conclusions may seem to present a pessimistic picture regarding our ability to know our own past. They suggest that we attain diachronic self-knowledge only under the conditions that:

1. 1. we not immediately identify* with the past character under consideration,

2. 2. we have changed significantly since the events remembered, and

3. 3. we corroborate our stories with others.

Condition 1 rules out making any inference to personal identity: our ability to remember and recount past events cannot serve as a criterion for what makes a person a person. Condition 2 rules out the possibility of the object of knowledge being something, call it a self, that persists through time. ‘Self-knowledge’ is a reflexive noun (compare: ‘self-study’ or ‘self-cleaning oven’) indicating that the object of knowledge happens to be the same entity as the subject. The person with self-knowledge does not know some peculiar object called the self any more than a self-cleaning oven cleans selves. Rather, the person knows about friendships started and affairs ended by the same person that she happens to be. And condition 3 rules out privileging our own memories over the testimony of others. But there is nothing pessimistic in any of this. As in the sciences, knowledge is a hard-won achievement of human discourse, not the default condition of human nature. There is no reason to think of our own case as an exception to this rule, nor is this fact anything to lament.