Forgetting Oneself: Tsongkhapa and Severance

Abstract

This paper explores philosophical issues of personal identity and its connection to forgetting through the famed Tibetan Buddhist thinker Tsongkhapa (1357–1419). Tsongkhapa, in turn, follows in the Middle Way (madhyamaka) tradition of Nāgārjuna (c. 150–250 CE) and Candrakīrti (c. 600–650 CE). Specifically, Tsongkhapa demonstrates that we can make sense of a consistent personal continuity despite the disruptions of forgetting and remembering. In so doing, he nuances the notion of personhood, revealing that it does not exist the way we think. I rely on a thought experiment derived from the hit TV show Severance to demonstrate the ramifications of his theory. By way of conclusion, I explore how Tsongkhapa’s analysis constitutes a notion of “positive forgetting”.

1. Introduction

What makes you you? That you are you seems obvious. However, what is responsible for this sense? Initially, this question might seem absurd. Our identity appears primitive and inscrutable. We are ourselves in the same way that anything is itself. A rock is a rock because, well, it’s a rock. But on closer analysis, this facile notion of personal identity dissolves. We’re not like rocks. Rocks seem to be made of the same stuff over time. People are not. We are constantly changing, both in mind and body, from moment to moment. Indeed, the rock of 100 years ago is likely little different from that rock today. But what makes you the same person today as the baby who shares your birthday, with all the accompanying changes from birth to adulthood?

Memory would seem to be key to the answer. The memories that I preserve over a lifetime give my experience continuity. Even if my mind and body are constantly changing, the memories I hold appear relatively stable. And this seems sufficient to maintain my sense of self: Only insofar that I remember who I am, with all the attendant memories that that entails, do I have a sense of being me.

The role of memory in identity formation prompts its relevance to the theme of this special issue: forgetting. For if memory is closely related to identity, then what are the consequences that forgetting has for our sense of self? If to recognize oneself is to remember, then must the self change when one forgets something about themself?

Tibetan Buddhists were acutely aware of this issue. And although Buddhists famously reject the self (anātman), they were deeply invested in explaining our experience of having a self. Indeed, according to Tsongkhapa (1357–1419), founder of the Geluk (dge lugs) school of Tibetan Buddhism, that experience is not, in and of itself, erroneous. Our sense of self, he argues, is perfectly reasonable for negotiating quotidian interactions in the world. We only hit trouble when we assume that this helpful heuristic of a self belies some metaphysical heft, when we assume that just because a sense of self is useful in daily transactions, that therefore it must be real.

Here, the issue of remembering and forgetting comes to the fore. If memory is responsible for our sense of self, then its consistency must be an illusion, given that the causes of this sense are inconsistent. We remember and forget. Our memories are not stable. This supports the Buddhist conclusion that the self cannot exist as it seems. Our sense of being personally stable actually reduces to fluttering memories, which are not—much like flickering pictures create the illusion of enduring objects in a film.

Still, Tsongkhapa cannot just leave his argument here. He must provide a cogent account of how this weak self—the pragmatic sense of self—arises despite its having no metaphysical correlate. Why do we feel like we have an enduring self if no such self exists? If memories are not stable, how can they alone produce a sense of stability? Here, forgetting becomes as much a foil as a friend, for if memory creates our sense of self, then we would expect this sense to change as we remember and forget. But this is not the case—it is always me whose memory changes; that me-ness does not change with changes in my memory. This challenges the idea that the sense of self is merely an artifact of memory. Tsongkhapa thus needs a positive, phenomenological account of how the sense of self survives the metaphysical deconstruction of any real self.

In the following, I offer an analysis of Tsongkhapa’s view of forgetting and its relation to personal continuity. I proceed in four sections. The first acquaints the reader with the philosophical context from which Tsongkhapa writes. This is the Root Verses on the Middle Way (Mūlamadhyamakakārikā) by Nāgārjuna (c. 150–250 CE) and its commentary, In Clear Words (Prasannapadā) by Candrakīrti (c. 600–650 CE). The next session explores Tsongkhapa’s own commentary on the Root Verses, An Ocean of Reasoning (Rigs pa’i rgya mtsho).

Next, in the third section, I extend Tsongkhapa’s analysis by relying on a thought experiment based on the hit Apple TV series, Severance, along with some additional help from the Buddhist thinker Śāntideva (c. late 7th–mid 8th centuries CE). I rationally reconstruct Tsongkhapa as critiquing the “myth of the given”—specifically, he denies that experience unilaterally provides some particular sense of self. Instead, this sense is highly mutable relative to the conventions of an epistemic community. Lastly, in the conclusion, I demonstrate how Tsongkhapa’s theory provides a positive account of forgetting, one that rejects any naive account of forgetting as an absence of memory simpliciter.

2. Nāgārjuna and Candrakīrti

Nāgārjuna was the founder of what later came to be known as the “Middle Way” (madhyamaka) tradition. And in Tibet, at least, Candrakīrti eventually held sway as the most authoritative interpreter of Nāgārjuna’s intention. But even what Candrakīrti intended became a matter of debate. Indeed, philosophical differences between Tibetan Buddhist schools often cut at his diverging interpretations.

Tsongkhapa takes the essence of Middle-Way insight to be the following. Our world is populated with various objects, everything from tables and chairs to institutions and systems, to theories and abstractions. We have certain conventions to designate these entities. We call a chair “chair” and some religion “a religion.” Most of us naively think these names are mere pointers; they are just tags to indicate something “out there” in the world, independent of that name. But according to Tsongkhapa’s interpretation of the Middle Way, belief in this “something” wholly extrinsic to naming is a mark of ignorance. More than a belief, we have a deeply held conviction in this something, a visceral sense that what we name subsists on its own outside of its being named. This conviction, Tsongkhapa states, is not just an error, but the cause of all our suffering.

Importantly, however, (and this is where he departs from other Tibetan interpreters of Candrakīrti), this is not to say that naming itself is erroneous. On the contrary, things can only exist insofar that they are named. Nevertheless, ignorance creates an extra layer of superimposition, the sense of something existing outside of that naming qua the independent referent of that name. Tsongkhapa finds this conceit incoherent, since it would self-subvertingly posit some entity that both exists independently of but in relation to its name. Nothing exists in this ultimate fashion, floating out on its own, as the unnamable referent of the names we use to describe the world.

In sum, then, Tsongkhapa accepts that objects exist conventionally, that is, through the power of names. But he rejects that they exist ultimately, that is, independent of that naming. The Middle-Way project, then, is to deconstruct into oblivion any entity that could exist ultimately, independently of our conventional designations. To apply this to the pertinent example to come, we can think about this point temporally: We might think that “birth” designates some inherent feature of our temporal history. On this conceit, there is a moment that, on the strength of its own, inherent features, warrants its designation as the moment of our “birth.” This would imply that birth exists ultimately, comprising undeniable features qualifying that moment as “birth” before its appellation as such.

But the Middle-Way philosopher contends that this is an impossibility. The only way that we can make sense of the concept of birth is in relation to a whole network of conventions that buttress it. There is nothing that demands that emerging from the womb constitutes birth, in and of itself, without the accompanying concepts “life,” “mother,” “child,” etc. And these conventions have no meaning without the convention of “birth” in turn. In other words, there is no raw, transparent data that forces its designation without an entire host of conventions, always already in play, that make those data intelligible. Or, better said, it is only conventions that render anything data at all.1

One of the most important necessary conventions from which “birth” cannot be disentangled is “death.” If something could be birthed without the end of something else, then birth would become a first cause2 and exactly that sort of thing which exists on its own, ultimately, and independent of the network of relations. It would be first in the metaphysical sense, a birth with no cause, and therefore demand its identification as “birth” by the power of its inherent primacy rather than the force of convention.

However, this type of birth is impossible, since then it would be a causeless cause. It would arise from nothing, yet somehow initiate its own effects. This would undermine causality, for if things could arise from nothing, then there would be no need for them to be caused. Things would just pop in and out of existence of their own accord. Therefore, if we understand birth as a “first” in this metaphysical sense, it becomes conceptually incoherent. Nāgārjuna articulates this position in verse 11.3 of his Root Verses:

If birth came first, while old age and death came after

Then birth would be ageless and deathless, and one would be born without having died.

Candrakīrti explains the verse in more detail:

If birth came first, then it would occur without3 death. But we cannot say that birth occurs without old age and death, since then it would not be accretive (asaṃsṛtatva). If the birth of a being, so conceived (parikalpyamāna), could occur without that being having gotten old and dying, then Theodore’s birth, so conceived, could occur in this life as a first cause without his having died. And thus, [his] samsaric existence would have a beginning, which would render it hopelessly causeless. We would have to remain silent concerning what we were prior. Nor could we conceive of the end of anything that came prior. Things would just arise after having not existed at all before—[i.e., causlessly]

(Nāgārjuna and Candrakīrti 1960, p. 96.15–20).4

Thus, no entity can have a true, metaphysical first cause. Theodore, as an example, can only be born into a new life based on “his” death in a past life. Admittedly, premising a critique of first causes toward a proof of past lives—and, by extension, reincarnation—seems to be a non sequitur. The analysis of causality tracks; there cannot be a first cause. But this does not entail that there must be some personal continuity between Theodore’s precursors and Theodore himself. That Theodore’s birth had etiological priors does not demand that his personal continuity extends infinitely into the past.

3. Tsongkhapa

Although Nāgārjuna and Candrakīrti’s argument may fail as a proof for reincarnation in the traditional sense, it does have ramifications for how we understand personal continuity within this life. If we bracket past lives and focus instead on previous moments of this life, Nāgārjuna and Candrakīrti’s deconstruction has more recognizable purchase. Specifically, how do we account for Theodore’s personal continuity throughout his life while recognizing that each moment of that continuum is temporally discrete?

This seems to constitute an antinomy. How can the moments of Theodore be numerically distinct and yet share in Theodore’s identity? When one says, “Theodore was x,” this seems to suggest that the current Theodore predicates some quality x indexed to the past by “was.” In other words, the Theodore of now partakes of some quality obtaining of Theodore then. This would absurdly render the past and present contemporaneous. On the other hand, we may think that the scope of “was” includes the subject and the predicate. In that case, “Theodore was x” implies that “Theodore now is no longer x,” absolving the problem. But it persists. How can the same Theodore hold contradictory predicates, both “was x” and “is no longer x”? If Theodore now is no longer x, then that is to say that he does not share identity with anything that predicates x, whether now or in the past.

What then is the relation between these two moments of “Theodore”? They cannot be exactly the same, since absurdities would ensue. But they cannot be completely different either, since then the past Theodore would be as different from the current Theodore as Dolores. This defies common sense. On this point, Tsongkhapa offers some clarification, appealing to memory and forgetting to do so. He begins in the voice of an interlocutor who raises this very objection. Let us say that Theodore does not remember his previous life; he has forgotten who he was. Then, through some spiritual practice and subsequent clairvoyance, he recalls his past life.

[Objection]: Let’s accept that Theodore remembers “In my previous lives, I was born as so and so.” Even so, to accept that this is as he remembers it is not tenable, since, if we were to accept that, then the Theodore of this life would exist as the same person as these previous lives and thus would be permanent

(Tsong kha pa 2019, p. 195.5–7).5

In other words, if we accept that Theodore’s realization that “I was this person in the past” is accurate, then that past life must, in some way, currently exist in order to be predicatable of Theodore now. If that past Theodore did not still exist, then there is no way that the current Theodore could be that person. As I suggested, however, this is only a problem with the implicit reification of Theodore across time. Outside of this assumption, “Theodore was x” does not refer to the presently existing Theodore. The scope of “was” includes the subject. And so, “Theodore was x” implies a past Theodore, preventing the absurdity.

However, the interlocutor may (though not definitively) assume a metaphysical presentism indicative of Buddhist traditions outside of the Middle Way, according to which only the present moment exists. If this is the case, then there is still a problem, since, on such a view, statements about the past cannot have a referent. So, if past Theodore does not exist, then saying he “was” anything is absurd. Or, if “Theodore was x” holds true even now, then the past still exists now, undermining Buddhist views of momentariness (at least according to the presentist).

The Middle-Way philosopher’s counter is to reject metaphysical presentism—all metaphysical positions, in fact, inevitably devolve into absurdities. The fact that presentism cannot give a cogent account of conventional, common-sense statements about the past evinces its error.6 Tsongkhapa, then, faults the interlocutor for their failed account of how memory works in everyday conversation, noting such discourse simply does not carry the metaphysical baggage that the interlocutor contends. And so, remembering the past necessitates nothing about the rememberer and remembered bearing some metaphysically robust relation.

[Reply:] This argument reflects a failure to understand how memory works. For example, during a debate, someone might accept, at first, that something like sound is permanent. But later, not remembering what they accepted beforehand, they may accept that it is impermanent. Then, when the opponent accuses them of a contradiction, they will remember [their previous assertion]. Thus, at that moment, they will remember insofar as they think, “Aha, I previously accepted that it was permanent.” But they will not remember with the thought, “This ’I’ that exists at this later point previously accepted that it was permanent”

(Tsong kha pa 2019, p. 95.7–11).7

Tsongkhapa uses the example of a debate to illustrate the point. Two individuals, person A and B, are engaged in a debate. Person A believes, at first, that sound is permanent and states this position in the debate. Classically, person A holds some Hindu assumptions about sacred sounds like “Om” being eternal (nitya). Person B is a Buddhist who argues that no produced phenomenon like sound can be eternal and permanent.

Inevitably (at least according to the Buddhist), the adept debater, Person B, will be able to convince Person A that sound is impermanent and not eternal. Despite being convinced, Person A may have forgotten that they previously accepted that sound is permanent. But Person B has not forgotten, and so accuses Person A of contradicting themself. Prompted by the accusation of a contradiction, Person A will then remember that they previously accepted that sound was permanent, but now agree that it is impermanent.

This remembering of what was forgotten is essential for any debate to proceed successfully; the Middle-Way debater’s main goal is to demonstrate their opponent’s inconsistency and catch their opponent in a contradiction through reductio ad absurdum (prasaṅga, thal ’gyur). This contradiction, in turn, depends on some mismatch between what the opponent said previously and what they state currently, often as a product of having forgotten their previous commitment. And without subsequent remembering of what was forgotten, that contradiction would have no argumentative force.

As Tsongkhapa points out, however, nothing in this story requires some metaphysical absurdity that Person A qua rememberer must be numerically identical with Person A qua remembered. In other words, to say that “I previously thought x” is not to say that “I now thought x.” Indeed, that current I does not think x at all since its opinion has changed. Tsongkhapa then explains how this bears on Theodore’s personhood.

Likewise, when the human Theodore of this life remembers his previous birth, although he remembers with the thought, “Before, I was born as this in that place,” he does not remember with the thought, “I qua this life was born before as this in that place.” Thus, the self, only insofar as it is the intentional object of the thought “I,” also exists across those previous lives. Therefore, there is no contradiction in Theodore’s remembering such a self, having made it the object of his intention. And so, that the Theodore of this life existed in previous lives does not render him permanent

(Tsong kha pa 2019, p. 195.13–18).8

Put simply, the fact that the “I” in the present can think about itself as temporally indexed in the past—the fact that I can intend (dmigs) my past self as an object—in no way necessitates that those two moments of myself are identical. But nor, importantly, does this preclude that these two moments both belong to “me.” This “mere I” (nga tsam)—“mere” as in a conventional “I” without some definitive metaphysical correlate—is all that constitutes personal continuity. Tsongkhapa adds, “Likewise, although this very person in this life does not continue on in a future birth, the intentional object of the thought ‘I’ continues on as just that mere I” (Tsong kha pa 2019, p. 195.18–20).9

In essence, Tsongkhapa’s position (in concert with standard Middle-Way strategy) trades metaphysics for ordinary language.10 “Theodore was x” neither demands that the past and present are metaphysically identical. Nor does it render discrete moments of Theodore disparate to the point of their bearing no relation. Nor does it entail that the present is real and the past unreal. This is because talk about the past carries no temporal metaphysical baggage. Indeed, conventional discourse demands the elision of metaphysics, since metaphysical speculation invariably fails to preserve the cogency of everyday discourse.

4. Severance and Śāntideva

Tsongkhapa’s appeal to linguistic convention as the final arbiter of the bounds of the self appears, prima facie, plausible. It placates concerns raised earlier about the intermittence of memory—remembering and forgetting—and its incongruity with a seemingly stable self. Specifically, the fluctuating nature of memory does not dictate unilaterally our self-conception. Only in concert with conventions, from which those memories themselves are inextricable, do memories inform how we identify. And so, those conventions govern how experiential data are marshaled toward a sense of self.

Let’s take a mundane example. As so often happens, let’s suppose that Theodore has forgotten some childhood memory. Suddenly, in a flash—perhaps when he returns to his hometown and chances upon a familiar hill—he remembers careening down its slope some past winter with a childhood friend in his sled. Now, how does Theodore remember which child he was? After all, there were two children in the sled. We might think that Theodore was the child from whose perspective the memory is revisited. However, on a little reflection, anyone can appreciate that memories do not always work like this. They are often disembodied, almost from a third-person point of view.

Even if one insists that memories are always first-person, this will not absolve the problem. If the memory is purely from the first-person perspective, then the person to whom that perspective belongs is not part of the content of the memory. I am myself the only thing that categorically cannot be within my direct field of view. So, there is no clear content with which Theodore can identify—there is just the world of the memory.

This seems almost such a silly question as to be unworthy of inquiry. Nevertheless, we cannot find anything within the raw content of the memory that demands, unilaterally, that Theodore identify with it. If it is third-person, then which boy? If it is just a sense-field of experience, then there is no Theodore there to identify with. Still, that such memories are “ours” or about “us” poses no confusion, except on reflection. Theodore would claim with ease, “Ah, I just remembered sledding down this hill with Jimmy!”

Notice how forgetting is instrumental in constructing this argument. Let’s say, for example, that Theodore did not previously forget this memory of sledding and held it continuously from when the memory happened until seeing the hill. In that case, we could explain his identification with the memory as derivative of his sense that the current experience is always “mine.” The sense of being the sledder while doing the action might carry forward in an unobstructed continuum. Here, too, how and with whom Theodore would identify is not necessarily straightforward. But we could at least explain the memory’s “mine–ness” as an extension of how this occurs in the present.

In the case of forgetting and re-remembering, however, this continuity is not available. Theodore has no accessible, prior association with this memory that could explain his identification. Indeed, we can very easily identify with memories that never happened at all, since memory is known to be incredibly fallible Arce et al. (2023); Howe and Knott (2015); Riesthuis et al. (2023). In this regard, re-remembering forgotten memories and being fooled by confabulated new ones are phenomenologically identical. And yet in both cases, it seems obvious to the rememberer that these events happened to “me.”

Tsongkhapa’s theory offers an account of how this occurs. It is not the experience per se that grants us our sense of self, but a laundry list of conventions that dictate what elements of our experience we identify with. For example, Theodore might recognize the boy in the sled as “me” by his favorite pair of snow pants. Or, if from the first person, he knows, by convention, that “he” must be the person behind the sense field constituting that memory. These conventions might fail to identify true memories in all cases; we might still identify with memories that never happened. However, we have a whole arsenal of other conventions that determine a false memory, such as its failure to be corroborated by other witnesses, etc.

Still, something feels unsatisfying here, as if we are putting the cart before the horse. We have an ineluctable feeling that memories are “mine” not because we designate them as such, but we so designate them because they feel to be mine. That the naming comes first, and the sense second, feels backward. Tsongkhapa and his Buddhist company would contend that this feeling is just a product of deep-seated ignorance. However, perhaps we need more evidence to fortify Tsongkhapa’s case. This comes from thinking through a modern-day thought experiment.

4.1. Severance

This thought experiment hails from Apple TV’s hit show Severance. Here’s the premise. Mark Scout, dealing with the tragic death of his wife, has struggled to keep a job. He decides to undergo the severance procedure, offered by Lumon Industries. By placing a small chip in Mark’s brain, his memories become bifurcated. When he is at work on the “severed” floor, he only remembers his time at work. And when he is home, he only remembers his time outside of work. He is thus effectively split into two persons: his “innie,” who only has access to experiences at work, and his “outie,” who never remembers what happens at work, but maintains the memories from before the procedure and continues to lead his life outside the office. Because his innie has no recollection of his outie’s past traumas, the innie can work effectively.

As the show progresses (spoiler alert!) and because of various intrigues that we need not detail here, outie Mark decides he wants to reintegrate with his innie. This would reverse the severance procedure such that both innie and outie memories would be equally accessible in “Mark’s” continuum going forward. This precipitates the central philosophical question of the show. Who will Mark be once he is reintegrated? Mark’s innie and outie also love different women, Helly and Gemma, respectively. So, who the new Mark will love would seem to reflect his personal continuity. If he loves Gemma, is it the outie’s personal continuity that continues? If Helly, then is it the innie’s? What if he loves both? What if neither?

And similar questions arise concerning how the new Mark would experience “forgetting.” Which set of experiences, the innie’s or the outie’s, would the new Mark feel as if he was “remembering,” having forgotten them before, and which would feel mundane, as part of his everyday set of memories? Perhaps he would not experience either as “remembered” at all. Being aware of his past, he would realize that neither set of memories was really “his,” given that the amalgamated Mark is effectively a new person.11

Several permutations appear equally possible. Indeed, nothing intrinsic to the experience demands one over the other. Instead, details about the innie’s and outie’s knowledge of each other, as well as their shared awareness of the reintegration procedure, would be informative of how the new Mark would interpret “his” newfound memories. For example, when the new Mark “remembered” being on the severed floor, he would know this was the past experience of the innie. Or if that memory was of being at home, then the new Mark would know “That was my outie’s memory.”

But imagine instead that Mark has no awareness that he is severed. Let’s say that his innie and outie both think they are the only “Mark” and that each one is “awake” when the other is “asleep.” Then, without either’s knowledge, the two are reintegrated, and the new Mark wakes up in his bed. This Mark would seem to “remember” forgotten experiences. But which set of experiences would present themselves as the recovery of forgotten memories and which would appear part and parcel of his quotidian existence? Perhaps, in this case, where he woke up would be decisive. E.g., “Oh I’m home just as always. But—how strange!—I forgot that I have been working in an office all this time!”

I do not intend to resolve any of these questions here. Indeed, Tsongkhapa’s point is that none of these can be resolved by experience alone—the experience of memory does not simply give our personal continuity naively. The thought experiment of Severance demonstrates this clearly, since, intuitively, we would think that not just the experience, but Mark’s knowledge of his innie and outie, as well as the reintegration procedure, will affect his perception of his memories.

More pointedly, there is no experiential difference between remembering something that was forgotten, such as remembering having sled down a hill, and the hypothetical experience of someone else’s memories being implanted in our mind, such as in Severance. In both cases, we have novel access to memories that we did not have prior. What differentiates whether that new memory is deemed “remembering something forgotten” or “someone else’s memory” is a host of conventions dictating our interpretation, including contextual knowledge and contingent definitions of what makes me me.

This is Tsongkhapa’s exact point about forgetting. My claim of having remembered something that I forgot does not demand some numerical identity between a past and present me, no more than the new Mark is “really” his innie or his outie. Likewise, our sense of when forgetting occurs is underdetermined by “experience” per se, instead governed by a host of conventions, including how we define the boundaries of the self.12 Tsongkhapa’s theory, then, reveals forgetting and sense of self as highly intertwined. Remembering and forgetting are not primitives, but dictated by a sense of self, which is in turn determined by a web of conventional contingencies.

This all suggests a sense of self that, despite our feelings to the contrary, is highly mutable, affected much more by how we talk than we might initially assume. And this feature is essential to the Buddhist project, since liberation and ethical development are predicated on the malleability of our sense of self. Śāntideva gives a thorough discussion of this point, which we cover in the next subsection.

4.2. Śāntideva

So far, our analysis reveals that, according to the Middle Way, personal continuity is neither some objective fact of the matter nor a feeling given in experience. Both mark failed attempts to determine what makes you you, independent of conventional determinations. Instead, the Middle Way contends that both the feeling of personal continuity and its “objective” criteria mutually cohere. Our language can describe why we feel like a memory is “ours,” and it feels that way as a result of habits of discourse. There is no independent state of affairs, be they feelings or facts, that unilaterally determines personal identity. Rather, these imbricate.

And not these two alone. For example, we have further conventions that determine when the feeling of this continuity is erroneous. For example, when individuals do not identify with aspects that they should, we call this “dissociation.” And there are also other pathologies that arise when one identifies with aspects that they shouldn’t, an array of identity delusions. These diagnoses, in turn, have ramifications for experience.

But this is not a vicious circularity, where the experience and definition of self are so tightly bound as to become ad hoc. Rather, personal identity is enmeshed in a network of mutually coherent conventions. And any change in that network must occur from within, like Jenga blocks in a tower, piece by piece, cohering with the tower as a whole. It is not like an unanchored raft drifting hither and thither without any bearing. And so, personal continuity is not self-referential in a manner that leaves it unmoored, undefined, and haphazard. This does mean, however, that our conventions for personal identity are changeable and defeasible. How you identify you as you might change.

Indeed, much of the Buddhist soteriological project is predicated on the mutability of personal identity. This is nicely illustrated by another Indian Middle-Way thinker, Śāntideva (685–763 CE).

We must maintain that the body is a singular entity

Comprised of many types of things, like the hands and the rest.

So too must we maintain that everyone

Has joy and suffering just like me

(Śāntideva and Prajñākaramati 1960, v. 8.91).13

Śāntideva contends that our self is not singular simpliciter, but a composite, including the various parts of the body. Śāntideva here gives the spatial equivalent of Tsongkhapa’s temporal claim about the self. Just as temporally discrete moments of the self are not numerically identical, but are deemed a “mere I” en masse, Śāntideva argues that spatially discrete parts of the body help compose the self.

He goes on to explain that this carries moral force. The suffering that makes up a part of “me” is mine by convention only. Ultimately, nothing differentiates my suffering from yours. So, to the extent that we think ridding ourselves of suffering is a good thing, so too must we alleviate suffering wherever it arises, since its determination as “mine” is just conventional.14

You might think, “Someone else’s suffering doesn’t hurt me,

So there is no reason to protect them from it.”

But you also are not hurt by the suffering of your future body,

So you why do you protect that?

You might think, “In that case, I’m the same person.”

But this just a false conceit.

One person dies

From which another is born.

So, if you think that one should protect only that person

Whose suffering is one’s own,

Then since the foot’s suffering is not the hand’s,

Why should the hand protect the foot?

(Śāntideva and Prajñākaramati 1960, vv. 8.97–99).15

If we think that only “our” suffering demands prevention, then why should the hand protect the foot? We might think that the hand and foot are part of the same body, and so are part of the same person. But then why should the Buddhist care about what happens to “their” body in the next life, since, by virtue of having a different body, that should be a different person? Thus, grasping this suffering as mine, and therefore attending to it, while ignoring the suffering of others, implies an inconsistent view of personal identity.

This would seem to counter Tsongkhapa’s original claim that it is perfectly cogent to think of all lives as part of Theodore. But keep in mind Tsongkhapa’s nuanced point about the “mere I.” That is, although there is no metaphysical continuity between lives, those various lives can be similarly labeled as “me.” This labeling is consistent despite, ultimately, being a “false conceit.” And Tsongkhapa makes this explicit in the case of future lives:

Likewise, although the person of this very life does not continue on in a future birth [emphasis added], the intentional object of the thought “I” continues on as just that mere I. Therefore, given that being birthed [in this life] one is apprehensive about future suffering, it makes sense to remove non-virtue in this life, which is the cause of subsequent suffering, and make efforts toward virtue in this life for the sake of what is to be enjoyed

(Tsong kha pa 2019, p. 195.18–21).16

Rather than at ends, Tsongkhapa’s “mere I” gives a conceptual framework to understand Śāntideva’s analysis. Although there are conventional criteria by which we differentiate “mere I”s, delineating groupings of mental and physical parts over space and time as distinct individuals, these groupings are not metaphysically determined. This does not render them arbitrary either. We are always already in a network of conventions that determine how we carve up selves. But that does not mean that those carvings were cut at natural joints. “Naturally,” that is, independent of our conventional determinations, there are no joints at all. Thus, if we think that how we happen to identify belies a metaphysical basis for ethical conduct—such that I am justified in prioritizing the relief of my own suffering—then we have failed to recognize the conventional nature of personal identity.

5. Conclusions: Positive Forgetting

The Buddhist authors that we have covered suggest a (perhaps) counterintuitive approach to the question of personal identity. They do not offer a normative epistemology for correct or accurate identifications of self—at least not “accurate” in the sense of correspondence based on how the self really exists. Exactly because there is no real self, they describe our sense of personal identity as a function of an epistemic community. And whatever standards some community may use to determine the bounds of the self, they will not derive from matters of fact transparently—nothing demands me-ness independently of how we conventionally expect selves to appear. This is the Middle-Way thrust. Because all such standards are conventional through and through, none are absolute and indefeasible.

However, this does not devolve into an all-out relativism, where identity becomes ad hoc and haphazard. We inherit our notion of self from these communities, with those standards baked in. Our sense of self may be pliable, but this malleability begins from somewhere, an episteme within which standards are already set. The Buddhist project, then, is to shift these standards of self in a direction that will promote greater compassion towards others. As Śāntideva suggested, this could be achieved through expanding identification beyond one individual body and mind. But, again, these changes are pragmatic. Though rare, someone may be overly altruistic at the expense of their own person. Here, compassion might mean increasing identification with “one’s own” mind and body. Whatever the situation, the Buddhist project of identification appears aimed at maximizing its scope, which, as a result of ignorance, is adventitiously limited and overly circumspect.

The mutability of the sense of self is therefore central to the Buddhist project. And, crucially, only because there is no fixed, real self do conventional distinctions have the power to dictate that sense. If that self were fixed, it would subsist independent of our discourse, and could bear no relation to conventions like “self.” Middle-Way reasoning here reflects their general ontology of relations, that no relatum can subsist independently of its co-relatum. Ergo, there is no self that is the independent referent of “self” and the vast semantic network that informs its usage.

Tsongkhapa substantiates this characterization of the self through an exploration of forgetting. If the person who forgot and subsequently remembers referred to a fixed self, absurdities would ensue—the past would become the present, and one person would hold contradictory predicates, both remembering and forgetting. Tsongkhapa’s argument here equally relies on an analysis of linguistic convention. Even though we say “I forgot that I said that, but now, I remember,” this does not demand a strict, metaphysical identity between these three “I”s. Coherent language usage does not require its conventions to map some metaphysical terrain.

In addition to these linguistic points, we can also garner a phenomenological argument from this Middle-Way analysis. This allows for a rational reconstruction of Tsongkhapa into a theory of “positive forgetting.” The gist of the argument is this: While we naively think of forgetting as the mere absence of memory, there is actually no such thing as forgetting outside of some positive, remembered content.

Notice in Tsongkhapa’s analysis that forgetting is only discussed in the context of someone having remembered what they forgot. What would it mean, on the other hand, to forget something without remembering? In this case, forgetting would not be a phenomenon at all since it would be a simple absence of awareness. But as Candrakīrti contends, the conceit of something existing completely outside of awareness is a contradiction, for awareness is requisite for establishing anything at all (Candrakīrti 1982, fols. 271b7–272a3).

Forgetting, then, only obtains insofar as it is an object of experience, and that experience demands some type of remembering. This is the Middle-Way point: there can be no free-floating forgetting that obtains independent of our ability to designate it. And that designation requires a signified, a “positive” correlate—the remembered.

We can think of other examples of forgetting, however, besides full-blown remembering of something forgotten. For example, we can have a vague sense that we have forgotten something without recalling exactly what was forgotten. Or, we can be aware of having forgotten where we put our keys without remembering where they are. Or, we might simply infer that something has been forgotten, such as when we discover an old note we wrote but have no recollection of its contents. These are three different types of forgetting: the first as a vague sense, the second as a gap in content, and the third as an inference. All three imply forgetting without total recall, such as in Tsongkhapa’s example of fully remembering what was claimed prior.

Nevertheless, the fundamental point remains: forgetting can only occur as a phenomenon, since only phenomena exist. And this requires phenomenal characteristics, which depend on recollection. The first type of forgetting requires recall of some undefined thing being known in the past, e.g., “I recall being aware of something, but now have forgotten.” The second requires a partial memory, though incomplete, e.g., “I recall putting my keys somewhere, but cannot remember where.” The third case is trickier. There is no positive mnestic content.

Still, some experience is required to serve as an indication of the forgotten memory. Even without any access to its mnestic content, some present phenomenon is necessary to provoke its inference. In this case, it would be the reading of the note. And, again, a host of conventions govern when those experiences evince a memory. In the case of the note, we might ask how we know that this note is written by our own hand. Only based on those conventions do the contents of that note become “my” forgotten memory or a record of someone else’s experience.

Śāntideva goes so far as to claim that inference of this sort is the primary means by which memory occurs. He argues against an opponent17 who proposes something intrinsic to experience that gives memories their “mine-ness.” Śāntideva replies with an analogy, arguing that memory functions in the same way that some hibernating creature during deep slumber may be bitten, only to awake later and infer that the now-infected bite was from a rat. In this way, they “remember” that they were so bitten (Śāntideva and Prajñākaramati 1960, v. 9.24). Śāntideva, therefore, concludes that memory can function without any past mnesic content whatsoever.

We can extend this argument to forgetting. Just as the conclusion of having been bitten only becomes present for that creature thanks to the bite, there must be some positive sign for us to infer having forgotten something. If, for example, the bite was benign and left no trace, that creature would have no sense of having been bitten—at least from the creature’s perspective. Perhaps from our outside perspective, aware of the bite, we can speak of the creature being bitten, even if the creature does not subsequently know. But here too, we cannot speak of their failure to be aware qua a mere absence of awareness. That ignorance relies on the presence of our knowledge of the bite. This all holds for forgetting: Whoever infers that something was forgotten, some phenomenal content is necessary.

In every case, then, forgetting is, counterintuitively, only cogent when juxtaposed against something known or remembered. In Tsongkhapa’s original example, the debater remembers exactly what they forgot. Likewise, in the case of partial memory, a part of a memory persists. Lastly, even in inference, forgetting only obtains relative to some probative conclusion based on some content-laden, positive experience substantiating it.

The Middle-Way analysis of forgetting thus reflects their analysis of self. Even more so, the two analyses inform each other. Because forgetting and remembering are bound together in a shared conventional network, the me who forgets and remembers is equally conventionally delineated. The interdependence between forgetting and remembering also substantiates that neither has any inherent phenomenal quality demanding its designation as such. So too, Tsongkhapa argues, there is no phenomenal quality inherently identifiable as “me” who undergoes the processes of forgetting and remembering. All these designations are determined from a much more complex web of interrelated meanings that cannot be obtained from phenomenal qualities alone.

Nevertheless, phenomenal qualities are necessary as a basis for these designations—even when that designation denotes an absence. The same holds for forgetting, even qua the absence of a memory, since without this phenomenal content, nothing is present to mind that could support its attribution. In this sense, Tsongkhapa gives a “positive” account of forgetting. Forgetting is not, as we naively think, a vacuous nothing—it is a presence as much as an absence.

Middle-Way arguments about this positive element of forgetting are not mere philosophical carping—they are central to the final goal of Buddhist practice. By demonstrating the interdependence of forgetting and remembering, the Buddhist aims to disrupt our naive sense of self, a sense which the Buddhist argues leads to suffering, our own and others. This same reasoning about the interdependence of forgetting and remembering precipitates a similar understanding of the relation between self and other. Tsongkahapa has thus effectively used forgetting to advance practitioners toward greater compassion and possibly enlightenment. Just as forgetting itself cannot exist without positive remembering, Tsongkhapa reveals forgetting as a substantive contemplative object, one that is highly generative toward the loftiest spiritual realizations.