Why Narratives Can't Create Selves
Three arguments against the (Dennett-style) narrative conception of the self
Let’s assume, along with a number of thinkers and traditions, the view that there are no objectively-existing selves. At any given time, there is no unified self, and even if there were, selves wouldn’t persist across time. Humans fail to meet psychological unity and continuity conditions for selfhood.
Is there a way we can nonetheless give a positive account of the self that salvages something like our ordinary sense of personal identity? According to Daniel Dennett and others, narratives create selves.
Here, I will give three reasons to doubt the narrative conception of the self. I’ll argue that under the assumption that real selves don’t exist, narratives can’t bring them into existence.
The paper containing these arguments got a revise & resubmit at the Journal of Consciousness Studies about 12 years ago. I never got around to resubmitting it. If you’re interested in seeing more details of this argument, let me know and I can send you the larger paper.
Self-portrait in front of Julie Oppermann, TH1223, Marjorie Barrick Museum of Art (UNLV), September 2025.
What is the Narrative Conception of the Self?
Dennett of course accepts that there are humans. And humans tell life-narratives. For Dennett, the “center of narrative gravity” of these stories - the main character - is the self. He writes: “We try to make all of our material cohere into a single good story. And that story is our autobiography” (1992, p. 114). Fireman, McVay, and Flanagan summarize: “[n]arrative does not merely capture aspects of the self for description, communication, and examination; narrative constructs the self” (2003, 5).
Advocates of the narrative conception believe that the narrative must at least loosely map onto the life of the human being who is telling the narrative. Part of the point is that the narrative won’t precisely and accurately describe the individual’s life is, because the narrative will make the life more coherent. But there is a limit: if Helen is a professor in Oregon in 2025, but their self-narrative is about a prince in medieval Spain, then the narrative does not create the self. Thus for the view to work, there must be some non-perfect degree of matching between the narrative and the life lived.
So here is what I will call the narrative conception of the self (NCS): there is a self insofar as there is a narrative told by a human being which coheres around a unified character (and which bears a reasonable resemblance to the corresponding real human being). So even if there are no real narrative-independent selves, one can still make true claims using the notion of a “self” or an “I”. The self is the main character in the story.
Many philosophers employ the notion of a narrative self, and not all accept this version of the narrative view. I myself tend to think that many philosophers overrate the power of narratives, as Galen Strawson has argued. But I won’t try to argue here against other ways in which philosophers connect narrativity and identity.
Argument 1: What is contained within a text?
Consider the following structure for a novel. In the first half, there is the development of a certain character named “Eve”. We learn some things about her upbringing on a farm, her life in the city during her education, her travels abroad after graduating from college. However, in the second half of the book, “Eve” comes apart. Not in a strange metaphysical sense, or in a tragic sense where Eve’s life-plans fail. Rather, it becomes apparent in the second half that the first half is a trick on the reader: “Eve” does not name one person, but instead two fully distinct human individuals. One lived most of her life on a farm, the other lived most of her life in a city. Perhaps, in a stirring finale, they meet, and have to work together, or against each other. It’s like Fight Club, but the opposite.
Or consider another book that seems to follow the exploits of a character named “Derek”. In a surprising afterword, the author says the following:
You, the reader, may have believed that this book is a novel about a certain character named “Derek”. But you are wrong. In fact, this is a book of short stories, where each “chapter” is a short story about an individual who is completely distinct from the individuals in other “chapters”, who all just happen to share the same name.
What these two possible books highlight is that sameness of character is not contained within the intrinsic features of a narrative; rather, the unity of the character is contained in the intentions of the storyteller (or perhaps the reader). The worry for Dennett’s NCS is that it might turn out, as one is seemingly narrating one’s own life, that there was not a unified character in the narration in the repeated use of the name. What might determine that there is one? That’s the concern I’d like to flesh out.
Do characters in narratives meet unity and continuity conditions?
Let’s imagine that the advocate of NCS can respond to this particular worry. Perhaps we can just say that the authorial intention in the narrative is for the name to refer to the same human individual.
However, there is a similar worry that is more pressing. It is a difficult question how it is that human characters in narratives can be said to meet psychological unity and continuity conditions. In a typical book, some of the main character’s psychology is revealed to the reader in the text, but for the most part narratives require of the reader/hearer to constantly fill in missing details of what is going on inside the mind of the character. There is little within a narrative itself which guarantees that the human in the narrative is a coherent and continuous self.
This problem can be seen as distinctly troubling for NCS by considering a point made by David Lewis:
[C]onsider the following question… Did Sherlock Holmes have three nostrils? The answer of course is no, but not because Conan Doyle ever says that he doesn’t, or that he has two, but because we’re entitled to make that extrapolation. In the absence of evidence to the contrary, Sherlock Holmes’ nose can be supposed to be normal (Lewis, 1978).
Lewis’s idea is that since narratives cannot be explicit about everything, we may assume certain things about characters. But now the worry for the revised NCS is that, just as we may assume that Sherlock Holmes has two nostrils, we may also assume that the human beings in life-narratives have the same quirks that real humans have – the same quirks that make real humans fail to meet whatever unity and continuity conditions which are required for identity!
Unless there is some explicit statement in the life-narrative that the human in the narrative is indeed a coherent self, we may simply assume that the human in the narrative has all the same identity problems as a real human being. So a proponent of this form of NCS would be back to square one in showing why appealing to narratives helps in providing truth conditions for claims involving the existence of selves.1
What are characters in narratives?
In literary theory, one of the fundamental issues is the very question of identity conditions for a character in a text. Is the cowardly teenager we meet in the first chapter of a novel indeed the same person as the hero in the last? Is the story about the person becoming a braver person, or is it a story about a new person, a braver person, rising out of the ashes of a weaker person? For any narrative with the richness of a typical human life, the very same question of identity of character arises for the character in the narrative as for any real self.
In fact, literary theory is quite disparaging of attempts to base the unity of the self on the unity of a human in a narrative. As Peter Lamarque summarizes,
Under the literary gaze fictional characters – or those with literary stature – assume an increasingly fragile constitution. They come to look more and more like the postmodernist version of the self. Indeterminate, deeply implicated in textual and narrative strategies, the product of interpretation at nearly every level…” (2003, p. 49).
If literary critics are correct that characters in narratives have the same identity problems as alleged real selves, then there is no salvaging Dennett’s NCS. Just as the previous point from Lewis indicates, if there are no real, unified, coherent, selves, there are no unified, coherent characters in narratives. As Lamarque puts it, literary theorists “have come to see human beings as more like fictional characters: diffuse, uncentered, lacking unity…” (2003, p. 45). It is perhaps ironic that the critique of the allegedly real self from literary theorists, with which Dennett is presumably sympathetic, may be expressed using the very claim that real selves are no better off than characters in narratives.
What does this leave us with?
My suspicion is that when ordinary people (including philosophers) hear or read a narrative about a human, we naturally assume that it is about a self. But it is thus the readers/hearers of narratives who give narrative characters their unity. Thus narratives do not, on their own, provide identity conditions for selves, abstract or otherwise. Perhaps we can just stipulate that our narratives are about coherent selves; but if we did that, we might as well just go ahead and stipulate that we ourselves in the extra-narrative world are coherent selves, and be done with it, without having to invoke narratives to do the work.
This is not a mere technical problem. Marya Schechtman, for instance, is careful in noting that in some cases there may be unarticulated aspects of the narrative that do undermine its role in self-creation - “some cases where unacknowledged influences direct a person’s narrative in ways that make it internally troubled and divided, and these interfere with personhood and identity” (1996, p. 129). Here Schechtman has in mind Freudian-type worries about an individual’s misunderstood motives. She also notes that self-narratives are often censored so that they omit painful aspects of the individual’s life (1996, pp. 117-118). But if the negative aspect of Dennett’s form of NCS is right – that the typical human’s mental life is so internally disunified that it fails to meet conditions for personhood – then this casts serious doubt on whether any humans in narratives should count as meeting those conditions. This is so even if the individual in the narrative is said to make autonomous life-decisions on the basis of reasons. What ensures that the human in the narrative isn’t disunified like the rest of us in the real world?