Memory and Self-Reference

ABSTRACT

Our memories elicit, in us, both beliefs about what the external world was like in the past, and beliefs about what our own past experience of it was like in the past. What explains the power of memories to do that? I tackle this question by offering an account of the content of our memories. According to this account, our memories are ‘token-reflexives’, in that they represent their own causal origin. My main contention will be that our memories are able to provide us with evidence for the two types of beliefs due to the self-referential nature of their content. First, I will put forward a series of thought-experiments intended to raise several intuitions about the veridicality of memories. Next, I will introduce the view that memories are token-reflexives, and I will motivate it by pointing out that the view accommodates the relevant intuitions. And, then, I will return to the two types of beliefs prompted by our memories, and argue that conceiving memories as token-reflexives allows us to explain why memory has the power to elicit the two types of beliefs in us.

KEYWORDS:

• Memory
• token-reflexives
• intentionality
• truth-conditions
• self-reference
• belief

1. Introduction

The relation between perception and belief has been explored in considerable depth in epistemology and the philosophy of mind. The relation between memory and belief, by contrast, has been explored to a lesser degree. This project is an attempt to clarify the latter relation by addressing a particular question about memory and belief. The question is why our memories have the power to elicit two types of beliefs in us; beliefs about what the external world was like in the past, and beliefs about what our own experience of it was like in the past. In what follows, I will tackle this question by offering an account of the content of our memories. According to this account, our memories are ‘token-reflexives’, in that they represent themselves. Specifically, they represent their own causal origin. The main tenet of this paper will be that our memories are able to provide us with evidence for the two types of beliefs mentioned above thanks to the self-referential nature of their content.

I will proceed as follows. In section 2, I will first highlight the two aspects of the relation between memory and belief that require explanation. Then, I will introduce the methodology that will be used to pursue such an explanation. In section 3, I will put forward a series of thought-experiments about memory. These thought-experiments are meant to raise several intuitions about the veridicality of our memories. I will argue that any view about the content of memories must accommodate those intuitions. In section 4, I will introduce the view that memories are token-reflexives, and I will motivate it by pointing out that the view accommodates the intuitions raised in section 3. Finally, in section 5, I will return to the relation between memory and belief, and argue that conceiving memories as token-reflexives allows us to explain the two interesting aspects of that relation.

2. The Project: Scope, Explanandum and Methodology

To delineate the scope of our project, let us, first of all, specify the type of memory that will concern us in this discussion. Let us start with the broad idea that a subject can form beliefs about the past thanks to memory (or, equivalently, that memory allows a subject to form beliefs about the past). This idea is rather weak and, thus, quite plausible. It is just the idea that memory provides a subject with the necessary information to form beliefs about the past. Whether the subject actually forms those beliefs or not is a different matter, a matter that will depend on a number of factors. A subject needs to, for example, pay attention to, and be interested in, the issue regarding which their memory is providing them information. They also need to have the appropriate conceptual repertoire to form the relevant beliefs. And, importantly, the subject’s ability to form those beliefs needs to be unencumbered by the presence of defeaters, whether those are defeaters of the undermining or of the rebutting kind.¹

The broad idea that a subject can form beliefs about the past thanks to memory can be spelled out in two different ways, depending on what we mean by ‘providing information’ exactly. Suppose, for example, that I attended a party last week and my friend Mary was there, wearing a blue dress. Days later, I could believe that Mary was at the party thanks to memory in two different senses.² I could believe it thanks to memory in the sense that, in the past, someone told me that Mary was there, I believed them, and that belief has been preserved up to the present time by memory. This could happen even if I never saw Mary while we both were at the party. Alternatively, I could believe it thanks to memory in the sense that, at the present time, I have a memory experience that originates in a past perceptual experience of mine, a perceptual experience of Mary being there.³ And, on the basis of that memory experience, I form the belief that Mary was at the party. This could happen even if I never believed, before now, that Mary was at the party. These two ways in which the faculty of memory allows us to form beliefs about the past are quite different. Being able to form a belief thanks to memory in the former sense does not require having, in the present, a memory experience. Being able to form a belief thanks to memory in the latter sense does. Being able to form a belief thanks to memory in the latter sense does not require having had that belief at some point in the past. Being able to form a belief thanks to memory in the former sense does. We may call the former kind of epistemic position ‘non-experiential remembering,’ and we may call the latter kind of epistemic position ‘experiential remembering.’ The focus of this paper will be on experiential remembering.⁴

Experiential remembering is special in that our memories seem to perform, from an epistemic point of view, a sort of double duty. We can form two kinds of beliefs about the past on the basis of our memories. On the one hand, we can form beliefs about objective facts out there, in the world. These facts do not require being experienced by anyone for them to be the case. The presence of Mary at the party last week, for example, is an objective fact. We may abbreviate the claim that we can form beliefs about objective facts in the past on the basis of our memories as ‘Objective Memory Thought’ (or ‘OMT’, for short):

Objective Memory Thought (OMT)

For every subject S and memory experience E:If S has E, then there is a fact p such that

1. S can form the belief that p on the basis of E.

2. p is an objective fact in the past.

If we focus on OMT, experiential remembering seems to be analogous to perception from an epistemic point of view. Just like we may form beliefs about objective facts in the present on the basis of our perceptual experiences, we may form beliefs about objective facts in the past on the basis of our memories. Now, it seems natural to think that an explanation of OMT must be found in the content of our memories. After all, one would think that if a subject is forming a belief about an objective fact in the past based on one of their memories, it is because the memory at issue is providing the subject with evidence for their belief. And it is difficult to see how the subject’s memory could provide them with evidence for their belief unless the content of the relevant belief was somehow included in the content of the subject’s memory. It is difficult to see, in other words, how the subject’s memory could provide them with evidence for their belief unless the memory itself was representing the objective fact in the past.

On the other hand, we can form, based on our memories, beliefs about what it was like for us to perceptually experience, in the past, some objective facts. Let us call facts about what it is like for us to have some mental states (mental states such as our past perceptual experiences of objective facts), ‘phenomenal facts.’ Clearly, phenomenal facts are not objective facts. Consider, for example, the phenomenal fact that, when I seemed to see Mary at the party, the colour of Mary’s dress appeared to be blue to me. This is not a fact which would have obtained if I had not perceptually experienced Mary’s presence at the party. We can abbreviate the claim that we may form, on the basis of our memories, beliefs about phenomenal facts regarding our own past mental states as ‘Phenomenal Memory Thought’ (or ‘PMT’, for short):

Phenomenal Memory Thought (PMT)

For every subject S and memory experience E:If S has E, then there is a fact q such that

1. S can form the belief that q on the basis of E.

2. q is a phenomenal fact about a past mental state of S.

If we focus on PMT, experiential remembering appears to be analogous to introspection from an epistemic point of view.⁵ Just like our introspective states allow us to form beliefs about what it is like for us to have mental states in the present, our memories allow us to form beliefs about what it was like for us to have mental states in the past. Once again, it seems natural to think that an explanation of PMT must be found in the content of our memories. For one would think that if a subject is forming a belief about what it was like for them to have, for example, some perceptual experience in the past on the basis of one of their memories, it is because the memory at issue is providing the subject with evidence for their belief. And it is hard to see how the subject’s memory could provide them with evidence for their belief unless the content of the relevant belief was somehow included in the content of the subject’s memory. That is, it is hard to see how the subject’s memory could provide them with evidence for their belief unless the memory itself was representing the phenomenal fact about the subject’s past perceptual experience.

There is, however, a certain property of experiential remembering which raises a prima facie difficulty for PMT. A subject can remember a fact, experientially, from two kinds of visual perspectives. One is the perspective from which the fact would have been presented to the subject if the subject themselves had experienced it, or been a part of it, in the past. The other one is the perspective from which the fact would have been presented to a different observer if, in the past, they had witnessed the fact; a presentation which includes the remembering subject as a participant in the relevant fact. Thus, when a subject remembers a fact from the first kind of perspective, they do not visualise themselves, as it were, from the outside whereas, when they remember a fact from the second kind of perspective, they do visualise themselves from the outside. Let us call memories which present facts to a subject from the first type of perspective ‘first-person’ or ‘field’ memories, and let us call memories which present facts to a subject from the second kind of perspective, ‘third-person’ or ‘observer’ memories.⁶ One might think that the following scenario constitutes a counter-example to PMT. Suppose that I have an observer memory of Mary being at the party, an experience wherein I visualise, not only Mary being at the party, but also myself looking at her across the room. In this scenario, my observer memory presents to me a phenomenal fact regarding how the scene was perceived in the past, namely, the fact that it was perceived from a perspective which allowed the perceiver to see me across the room. But, knowing full well that I could not have occupied such a perspective, I am unable to form the belief that, in the past, the scene at the party was perceived in such a way. Is this not a counter-example to PMT?

In my view, it is not. Notice, first of all, that, in the above-mentioned scenario, the fact that Mary appeared to be at the party from a perspective which allowed the perceiver to see me across the room is a fact about a past perceptual experience of mine. It is, to be sure, a fact that never obtained. And yet, it remains the case that one of things which my observer memory is informing me of is that I perceived Mary at the party from a perspective which allowed me to see myself across the room.⁷ It seems, then, that the fact that Mary appeared to be at the party from a perspective which allowed the perceiver to see me across the room satisfies clause (2) in PMT. Secondly, notice that, in the scenario in which I have the observer memory described above, I am in a position to believe, on the basis of my memory, that Mary was perceived, in the past, from a perspective which allowed the perceiver to see me across the room. For my memory is providing me with evidence for a belief, not only about what Mary’s appearance was like at the party, but also about what my own appearance was like. After all, my memory is not silent, or neutral, on the issue of what we both looked like at the time. What happens is that the evidence in question is being rebutted by my reasons for thinking that, in the past, I could not have seen myself across the room. In other words, my reasons for thinking that, in the past, I could not have seen myself across the room are acting as rebutting defeaters for the evidence provided by my memory. It seems, then, that the fact that Mary appeared to be at the party from a perspective which allowed the perceiver to see me across the room satisfies clause (1) as well. Thus, the phenomenon of observer memory does not seem to provide us with a counter-example to PMT after all.

Interestingly, Objective Memory Thought and Phenomenal Memory Thought go, in a certain sense, hand in hand. In what sense? In the sense that any instance of OMT is an instance of PMT, and vice versa. We may refer to this fact as the ‘intertwinement’ of objective and phenomenal thought in memory:

Intertwinement

For any subject S and memory experience E: If S has E, then(There is a fact p such that

1. S can form the belief that p on the basis of E, and

2. p is an objective fact in the past.)

if and only if(There is a fact q such that

1. S can form the belief that q on the basis of E, and

2. q is a phenomenal fact about a past mental state of S.)

The case of Mary being at the party, once again, illustrates Intertwinement quite nicely. Consider the memory that originates in my perceptual experience of Mary being at the party. It seems that if that memory allows me to form the belief that Mary was at the party, then it also allows me to form a belief about what it was like for me to seem to see that Mary was at the party. It allows me to form, for example, the belief that the colour of Mary’s dress appeared to be blue to me. Conversely, it seems that if my memory allows me to form the belief that the colour of Mary’s dress appeared to be blue to me, then my memory also allows me to form the belief that Mary was at the party.

There is, however, a certain attitude that we can have towards our own past cognitive capacities which, on the face of it, seems to raise a prima facie difficulty for Intertwinement. This is the attitude of mistrust. One might think that the following scenario constitutes a counter-example to Intertwinement. I could have reasons for doubting the reliability of my perceptual faculties while I attended the party. (Perhaps I think that I was drunk, or took some drugs, while I was there.) In that case, I could, on the one hand, refrain from forming the belief that Mary was in fact at the party and, on the other hand, be willing to form some beliefs about what it was like for me to seem to perceive that she was there. Is this not a counter-example to Intertwinement? In my view, it is not. Notice that, in this scenario, I still have the capacity to judge, on the basis of my memory, that Mary was at the party. My memory is still providing me with evidence for the belief that Mary was there. After all, my memory is not silent, or neutral, on the issue of whether she was there or not. What happens is that the evidence in question is being undermined by my reasons for doubting the reliability of my perceptual faculties in the past. In other words, my reasons for doubting the reliability of my perceptual faculties in the past are acting as undermining defeaters for the evidence provided by my memory. Thus, the possibility of mistrusting one’s own past cognitive capacities does not seem to raise a counter-example to Intertwinement after all.

Experiential remembering seems to have, then, a special kind of scope. Perception only seems to allow us to form beliefs about what happens in the world, and introspection only seems to allow us to form beliefs about what happens in our own minds. Experiential remembering, by contrast, seems to be able to have those two kinds of epistemic powers at once. Why is that? The issue that will concern us in this discussion is what it is about our memories which can elicit both kinds of beliefs in us. I will approach this issue by focusing on the content of our memories. Thus, the main question that we will try to answer in what follows is this: What must our memories represent (or, equivalently, what must their content be) for them to be able to elicit, in us, both beliefs about the past state of the world, and beliefs about the past state of our own minds?

Now, if the project is framed as that of building an account of the content of our memories that illuminates the relation between memory and belief, then it seems that, before that project can get off the ground, we will need be explicit about what notion of content is going to be operative in such an account, and what methodology is going to be used for building it. Let us start by noticing that our memories can be evaluated as either correct or incorrect, true or false. This suggests that our memories have truth-conditions, that is, conditions relative to which they are true, and conditions relative to which they are false. Admittedly, memories can also be evaluated in other ways. Memories can, for example, be evaluated as ‘authentic’ if they correspond to, or match, the past perceptual experiences in which they originate.⁸ They can be evaluated, too, as having some combination of these two features. One might, for example, evaluate memories as ‘alethic’ if they are either true and authentic, or true but not authentic.⁹ It seems reasonable to think, then, that if memories have truth-conditions, then they must also have conditions which relate to their being authentic, or to their being alethic. Such conditions are certainly interesting. They are interesting, for example, for the issue of whether the faculty of memory has malfunctioned while producing a certain memory experience.¹⁰ But if our aim is to clarify how memory can inform, or provide evidence for, belief, then it seems that the useful notion of content for our purposes here cannot be a notion of content which equates the content of our memories with their authenticity conditions. And, to the extent that they are partly defined by using the notion of authenticity, it cannot be a notion of content which equates the content of our memories with their alethicity conditions either. For what a subject can come to believe, on the basis of one of their memories (or, equivalently, what that memory provides evidence for) is what it would take for the memory to be true, and not what it would take for the memory to be authentic. In what follows, then, we will think of the content of our memories as their truth-conditions.¹¹

The notion of content as truth-conditions suggests, in turn, a certain test for evaluating any hypothesis about the content of a memory. Suppose that we attribute some content to a memory, and we entertain some possible situation regarding that memory; a situation in which some of the facts are different from the actual facts. Then, the content that we attributed to our memory needs to capture our intuitions on whether the memory represents that possible situation correctly or not. If, intuitively enough, the memory that we are considering does not represent the possible situation correctly, but the content that we attributed to it commits us to the view that it does, then it seems that the content that we attributed to the memory was too thin, or not specific enough. For it makes it too easy for the memory to represent a possible situation correctly. Conversely, if the memory that we are considering represents, intuitively enough, the possible situation that we are entertaining correctly, but the content that we attributed to it commits us to the view that it does not, then it seems that the content that we attributed to it is too thick, or too detailed. For it makes it too hard for the memory to represent a possible situation correctly. Let us proceed, now, to the task of building a proposal about the content of our memories by using this methodology.

3. The Veridicality of Memories

The proposed methodology for building a proposal about the content of our memories relies on our intuitions about the veridicality of memories. To pinpoint those intuitions, it will be helpful to keep some particular example in mind throughout our discussion in this section. Let us remind ourselves, then, of the different elements in the Mary example once more. In that example, I attended a party last week and, while I was there, I seemed to see that Mary was at the party, wearing a blue dress. Let us call my perceptual experience ‘P.’ Mary was indeed at the party wearing a blue dress, so I did not misperceive Mary by having P. Days later, I have a memory that originates in my past perceptual experience of Mary being at the party. Let us call it ‘M.’ M is a memory on the basis of which I am able to form the belief that Mary was at the party. M is also a memory on the basis of which I am able to form the belief that seeming to see that Mary was at the party was, for me, like such-and-such.¹² I am able to form, for example, the belief that her dress appeared to be blue to me at the time. Let us refer to this possible situation as ‘W₀.’ The question, now, is what the content of memory M is in W₀.

Intuitively enough, it seems that M in W₀ is true of the possible situation in which it is happening, namely, W₀. But we can also evaluate M in W₀ relative to other possible situations in which things are slightly different. That is, we can imagine other possible situations in which things are slightly different, and ask ourselves whether my memory M, occurring in W₀, represents those possible situations correctly or not.¹³ Which things could have been different? There seem to be three moving parts in the Mary example. There is, first of all, the fact that Mary is at the party wearing a blue dress. Secondly, there is the fact that I seem to see her there, wearing a blue dress, by having P. And, thirdly, there is the fact that I have M as a result of having had P. Consider, accordingly, three possible alternative situations:

W₁ Mary is not at the party. And yet, I seem to see her there, wearing a blue dress, by having P. Days later, I have M as a result of having had P.

W₂ Mary is at the party, wearing a blue dress. But I seem to see her there by having a different perceptual experience P*, a perceptual experience wherein her dress appears, incorrectly, to be yellow to me. Days later, I have M as a result of having had P* .

W₃ Mary is at the party wearing a blue dress, and I seem to see her there, wearing her blue dress, by having P. Days later, I have M. But my having M is unrelated to my having had P at the party. My having M is the product of my wishfully thinking that Mary had been there.

What are our intuitions about the veridicality of M in W₀ with regards to these possible situations? First of all, M in W₀ does not seem to represent W₁ correctly. This intuition points us in the direction of building, into the content of M in W₀, the fact that Mary was at the party. Committing ourselves to building the fact that Mary was at the party into the content of M in W₀ may seem like a minimal commitment. And yet, some views of mnemonic content dispute this commitment. There is a view according to which what matters for the veridicality of a memory with regards to a possible situation is whether, in that situation, we seemed to perceive a certain fact in the past, and not whether the fact actually obtained.¹⁴ On this view, the presence or absence of Mary at the party in W₁ is irrelevant for whether M in W₀ represents W₁ correctly or not. What matters is whether, in W₁, I seemed to perceive that Mary was at the party, wearing a blue dress. And, in W₁, I did seem to perceive that. Thus, the view predicts that my memory M in W₀ represents W₁ correctly, which seems counter-intuitive.

Memory M in W₀ does not seem to represent W₂ correctly either. This intuition points us towards building, into the content of M in W₀, the fact that I seemed to perceive that Mary was at the party by having, not just any perceptual experience, but a perceptual experience of a certain phenomenal type. Committing ourselves to building the fact that I had a perceptual experience of a certain phenomenal type into the content of M in W₀ may seem like a minimal commitment too. And yet, some views of mnemonic content dispute this commitment as well. There is a view according to which our memories inherit their contents from the perceptual experiences in which they originate. As a result, memories may fail to represent all which was represented by those perceptual experiences. But memories cannot represent something which was not represented by those perceptual experiences. We may call this view, the ‘inheritance’ view.¹⁵ The inheritance view predicts that M in W₀ represents W₂ correctly. For M in W₀ cannot represent anything which was not represented by P in W₀. And the fact that I seemed to perceive Mary by having a perceptual experience which is phenomenally similar to M is not something which was originally represented by P in W₀. Thus, the fact that I seemed to perceive Mary by having a perceptual experience which is phenomenally similar to M is not, on the inheritance view, something which can be represented by M in W₀. This means that M in W₀ must be true of W₂. After all, M in W₀ is true of W₀, and the only difference between W₀ and W₂ is that, in W₂, I do not seem to perceive Mary by having a perceptual experience which is phenomenally similar to M. But that difference cannot, on the inheritance view, make M in W₀ false with regards to W₂, since that is not a fact which can be represented by M. Thus, if M in W₀ represents W₀ correctly, then it must represent W₂ correctly as well, which seems counter-intuitive.¹⁶

Finally, memory M in W₀ seems to fail to represent W₃ correctly too. This intuition suggests that, furthermore, we should build, into the content of M in W₀, the fact that M causally originates in my perceptual experience P. Otherwise, the content that we attribute to M in W₀ will fail to capture the intuition that M in W₀ is not true of W₃, since the causal connection between P and M appears to be the salient difference between W₀ (which seems to be correctly represented by M in W₀) and W₃ (which does not).

What lesson should we draw, then, from this series of thought-experiments, about the type of content that we should attribute to a memory? The lesson seems to be that the content in question should involve three elements. It should involve an objective fact. It should involve the fact that the subject has experienced it by having a perceptual experience of a certain phenomenal type. And it should involve the fact that the subject is having the memory at issue as a result of having perceptually experienced the objective fact. This seems to be a highly complex type of content. It may seem challenging, then, to come up with a template which delivers, for each memory, a content of this type that we can plausibly attribute to the memory. However, we will be able to build such a template, I suggest, if we turn our attention to a certain notion in semantics, namely, the notion of token-reflexivity.

4. Memories As Token-Reflexives

The notion of token-reflexivity has its origin in the semantics of indexical expressions, expressions such as ‘here,’ ‘this’ or ‘I.’ There is an approach to the semantics of indexicals according to which the semantic properties of indexical expressions are properties of them as tokens, and not types. As a result, the contribution that an indexical expression makes to the truth-conditions of the sentence in which it occurs can only be specified by mentioning some properties of the very token of the indexical expression occurring in that sentence.¹⁷ Roughly, the idea is that an utterance of, let us say, ‘it is cold here’ is true just in case it is cold at the place where the token of ‘here’ occurring in that sentence is being uttered. Similarly, an utterance of ‘it is noon now’ is true just in case it is noon at the time at which the token of ‘now’ occurring in that sentence is being uttered, and so on. It seems natural, then, to think of token-reflexivity as the reference of a token to a token, that is, to think of token-reflexivity as self-reference.¹⁸ Thus conceived, token-reflexivity turns out not to be a new idea in the philosophy of mind. There is a long tradition of thinking about a particular kind of mental states, namely, conscious mental states, as referring to themselves.¹⁹ However, the idea that memories, in particular, might be token-reflexives has received relatively little attention in the literature.²⁰ My proposal is to appeal to the idea that memories refer to themselves in order to frame a notion of mnemonic content which is able to incorporate, into the content of our memories, the three elements suggested by our intuitions about the veridicality of memories. The proposed notion of mnemonic content can be formulated as follows:

Token-Reflexive View (TRV)

For any subject S, memory experience E and fact p:

If S has E, and S would report E by claiming to remember that p, then the content of E is that E originates in a perception of p which was, for S, like having E.

Notice an interesting feature of this proposal about the content of memories. For any subject S and memory experience E, if S has E, and E has the content that TRV attributes to it, then E is token-reflexive, in the sense that the content of E concerns E itself. But, interestingly, if TRV is correct, then E is, not only token-reflexive, but doubly token-reflexive. For the content that TRV attributes to E makes reference to E twice. On the one hand, the content attributed to E makes reference to its causal origin. For it mentions that E originates in a perception of a certain fact. Thus, on TRV, memory E represents, among other things, that a perception of a certain fact is at the causal origin of memory E itself. On the other hand, the content attributed to E makes reference to what it was like for the subject to have that perception. Specifically, it makes reference to the fact that perceiving the relevant fact was, for S, like having E. Thus, on the token-reflexive view, memory experience E represents, among other things, that the experience of perceiving a certain fact in the past was, for the subject, like that of having memory E itself.²¹

What kinds of considerations can be produced in support of the token-reflexive view? The main consideration in support of TRV is that it can accommodate our intuitions about the series of thought-experiments discussed in section 3. Consider, once again, my memory M in situation W₀. The content that TRV attributes to M is that M originates in a perception of Mary being at the party which was, for me, like having M. Now, suppose that this is indeed the content of my memory M in W₀. Then, we can accommodate our intuition that M in W₀ does not represent W₁ correctly. After all, in W₁, Mary is not at the party. It is no wonder, then, that we have the intuition that M in W₀ does not correctly represent W₁.

We can also accommodate our intuition that M in W₀ does not represent W₂ correctly. After all, in W₂, I do not seem to perceive Mary by having an experience that phenomenally resembles M. It is no wonder, then, that we have the intuition that M in W₀ does not correctly represent W₂. Notice that what matters for whether M in W₀ represents W₂ correctly is whether, in W₂, my past perceptual experience phenomenally resembles my current experience of having M and not whether, in W₂, I have P specifically. Requiring the latter would be too demanding. After all, consider a possible situation W₄ in which Mary is at the party, wearing a blue dress, but I seem to see her there by having a perceptual experience P** wherein her dress appears to be blue to me, but a slightly different shade of blue from the shade of blue presented to me by P in W₀. And suppose that, days later, in W₄, I have M as a result of having had P**. Possible situation W₄ is similar to W₂ in that, in both situations, I do not have P in the past. And yet, intuitively enough, M in W₀ does not seem to represent W₂ correctly, but it does seem to represent W₄ correctly. This is a difference in our intuitions that TRV can accommodate. By not imposing such a strict requirement as the requirement that, in the past, I must have had experience P in particular, TRV leaves room for some variation between the phenomenal character of my past perceptual experience of Mary and the phenomenal character of my current memory, which seems a virtue of the view.²²

Finally, if TRV is correct, then we can accommodate our intuition that M in W₀ does not correctly represent W_3. After all, in W₃, M does not originate in my having perceived Mary at the party. And the content attributed to M in W₀ by TRV requires, for any possible situation to be correctly represented by M in W₀, that I have M as a result of having perceived Mary in that situation. It seems natural, then, for us to have the intuition that M in W₀ does not correctly represent W₃ either. It seems, therefore, that the token-reflexive view of memory can be motivated by appealing to our intuitions about the veridicality of memory through a series of thought-experiments.²³ Let us return, now, to the features of the relation between memory and belief which were highlighted in section 2, and consider whether the token-reflexive view might be able to throw some light on them.

5. Memory and Belief

The token-reflexive view of memory can help us explain OMT, PMT and Intertwinement. Let us take OMT first. OMT is to be expected if TRV is right, since the content attributed to memories by TRV is, in part, about objective facts. If my memory M, for example, represents that it originates in a perception of Mary being at the party which was, for me, like having M, then it does not seem surprising that, on the basis of M, I can form a belief about an objective fact, namely, the presence of Mary at the party. After all, if TRV is right, then the only thing that I need to do, in order to form such a belief, is to take the content of my memory M at face value.

Similarly, PMT is also to be expected if TRV is correct, since the content attributed to memories by TRV is, in part, about what the perception of an objective fact was like for the subject in the past. If M represents that M originates in a perception of Mary being at the party which was, for me, like having M, then it is not surprising that, on the basis of M, I can form a belief about what it was like, for me, to seem to perceive that Mary was at the party. After all, if TRV is right, then the only thing that I need to do, in order to form such a belief, is, once again, to trust the content of my memory M.

Finally, if TRV is correct, then it makes sense that Intertwinement takes place as well. For, according to TRV, the content of a memory is not about an objective fact, not exclusively. And it is not about what the subject’s perception of an objective fact was like, not exclusively. It is about both of those facts at the same time, and about the relation between them. It is no wonder, then, that every time that we can form a belief about an objective fact based on one of our memories, we can also form a belief about what our perceiving that fact was like, and vice versa. After all, the content of our memories concerns both objective facts and phenomenal facts if TRV is right. Thus, it seems that the token-reflexive view can indeed shed some light on the intricate relation between memory and belief which is characterised by OMT, PMT and Intertwinement.

Let us take stock. We started our discussion by noticing that memory and belief are related in a curious way. Their relation makes memory analogous to perception, but also analogous to introspection. To explain the two aspects of this relation, a view about the content of memories was put forward. This is the view that memories are token-reflexives, in that they refer to themselves. Specifically, they represent their own causal histories. We have seen that this view can explain the complex ways in which memory is related to belief. And we have seen that the view can do this while, at the same time, accommodating our intuitions about the veridicality of memories in a variety of scenarios. I conclude, therefore, that the token-reflexive view is the correct view about the content of our memories.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Australian Research Council [FT160100313].

Notes

1. Undermining defeaters cast doubt on the evidential support for a belief while rebutting defeaters provide evidential support for the opposite belief. On undermining, versus rebutting, defeaters, see (Pollock 1974).

2. Notice that this example involves a fact. Sometimes we report our episodes of remembering by claiming to remember objects (as in ‘I remember Mary’), by claiming to remember events (as in ‘I remember Mary’s entrance’), by claiming to remember facts (as in ‘I remember that Mary was at the party’) and by claiming to remember properties (as in ‘I remember being happy that Mary was at the party’). An analogous point can be made about the way in which we report our episodes of perceiving. For the purposes of this discussion, I will concentrate on both memory and perception for facts, or states of affairs. I intend to remain neutral on whether memory for either events, properties or objects is reducible to memory for facts, or vice versa. On this issue, see (Openshaw 2022). Likewise, I intend to remain neutral on whether perception for either events, properties or objects is reducible to perception for facts, or vice versa. On this issue, see (Vernazzani 2022).

3. We can think of memory experiences (or, for short, ‘memories’) as quasi-perceptual experiences with two types of features. A memory enjoys intentional features in virtue of which, by having the memory, a subject represents some fact. A memory also enjoys phenomenal features in virtue of which there is such a thing as what it is like for the subject to have the memory. Given that memories are experiences, in what follows, I will individuate memories by their phenomenal features. Also, while discussing both memory and perception, I will talk about experiences being ‘of’ some facts. This kind of talk is only meant to abbreviate that, if prompted, the subject would report their experience by saying that they remember, or that they perceive, the relevant fact. The precise nature of what a subject represents by having a memory will be the topic of our discussion in sections 3 and 4.

4. In the psychological literature, it is common to distinguish between an ‘episodic’ and a ‘semantic’ declarative memory system, a terminology introduced in (Tulving 1972). As far as I can see, the instances of remembering covered by the semantic memory system are, in the terminology used above, instances of non-experiential remembering, whereas the instances of remembering covered by the episodic memory system are, in the terminology used above, instances of experiential remembering. Nothing in the discussion that follows, however, hinges on this point.

5. This analogy has not gone unnoticed in the literature. Matthew Soteriou, for example, conceives experiential remembering as involving self-knowledge of the conscious character of one’s past experience in (Soteriou 2013, 181).

6. For seminal work on this distinction, see (Nigro and Neisser 1983).

7. Admittedly, this point has been challenged in the literature. For a defence of the view that observer memories present past facts from an unoccupied point of view, see (McCarroll 2018).

8. For details on the notion of authentic memory, see (Bernecker 2009, 38–40).

9. For discussion of alethic memory, see (Michaelian and Sant’anna 2022).

10. This issue does not hinge on whether the relevant memory experience is false. Otherwise, our faculty of memory would be at fault every time that we misremember something experientially because we misperceived it in the past, which seems counter-intuitive. It seems much more reasonable to think that the issue of whether our faculty of memory has malfunctioned while producing a certain memory experience hinges on whether the relevant memory experience is authentic, or perhaps alethic.

11. To equate content with truth-conditions is, however, not an uncontroversial assumption. For discussion of some of the issues which hang on this assumption, see, for example (Glüer 2012).

12. To be clear, ‘M’ is a name for a memory experience which is introduced through a description of the causal origin of that experience; a description which fixes the reference of ‘M’, but does not give its meaning. Thus, the claim ‘M originates in my past perceptual experience of Mary being at the party’ turns out to be true a priori, but contingently. After all, M could have had a different causal origin, since memory experiences are individuated, for the purposes of this discussion, by their phenomenal features and not by their relational properties. (Analogous considerations apply to the name ‘P’.).

13. There does not seem to be anything mysterious about our ability to do this. This is just what we do, for example, when we ask ourselves whether a photograph, or a portrait, of some person happens to represent a different person correctly; a person who could have posed, but did not actually pose, for the photograph or portrait in question. Whether our views about pictorial representation commit us to answer this question in the negative or not, the question is surely intelligible.

14. For some defences of this position, see (Von Leyden 1961, 61) and (Meinong 1973, 256).

15. The inheritance view captures, for example, Thomas Reid’s position when he claims that ‘things remembered must be things formerly perceived or known’ (Reid 2002, 254). More recently, the inheritance view has been labelled ‘preservationism’ in the philosophy of memory literature (Michaelian and Robins 2018, 23). A closely related view to the inheritance view is the so-called ‘causal theory of memory’, according to which a subject’s mental state qualifies as a memory of some fact only if it causally originates in one of the subject’s past perceptual experiences of that fact (C. B. Martin and Deutscher 1966). Even though the causal theory of memory was originally devised as an answer to a metaphysical question (the question of what it takes for a mental state to qualify as a memory), the causal theory of memory entails the core idea in the inheritance view. This is the idea that a memory cannot have more content than the experience in which it originates. For a contemporary defence of the causal theory of memory, see (Bernecker 2009, 104–127).

16. The view that memories inherit their contents from the perceptual experiences in which they originate faces challenges from a different angle as well. It is in tension with a certain picture of memory which has become dominant in psychology (Dale, Loftus, and Rathburn 1978; Loftus and Palmer 1974; Loftus and Pickrell 1995) and is becoming increasingly popular in philosophy (De Brigard 2014; Michaelian 2016). This is the view that memory is ‘reconstructive,’ in the sense that the content of our memories integrates content that we have acquired through our past perceptual experiences with content from other sources, such as inference, testimony and the imagination. The reconstructive view of memory, however, does not offer a template for determining, for each particular memory, what the content of that memory is. Thus, it is hard to tell whether the reconstructive view ultimately fares better, or worse, than the inheritance view when it comes to explaining OMT, PMT and Intertwinement.

17. The idea originates in Hans Reichenbach’s theory of indexicals in (Reichenbach 1947, 284). For a defence of indexicals as token-reflexives, see (García-Carpintero 1998).

18. For the purposes of this discussion, I will be identifying token-reflexivity with self-reference. There are reasons, however, for doubting this identification. On this issue, see (Simchen 2013).

19. In (Kriegel 2009b), Uriah Kriegel traces this tradition, from Franz Brentano’s view that every mental state is intentionally directed at itself (Brentano 1973), through Reid’s view that sensations are identical with the introspective knowledge of their own instantiation (Reid 2002), back to, arguably, Aristotle’s view that all states of perceptual consciousness are reflexive (Caston 2002). Kriegel inserts himself in this tradition when he argues, in (Kriegel 2009a), that mental states which are phenomenally conscious are so in virtue of representing themselves.

20. There are some exceptions. See, for example (M. G. F. Martin 2001, 278; Searle 1983, 48–52) and (Fernández 2006).

21. The doubly token-reflexive nature of mnemonic content is what makes TRV a novel view. It distinguishes TRV from other self-referential views, such as the ‘reflexive’ view of mnemonic content introduced in (Fernández 2006) and developed in (Fernández 2019). On the reflexive view, if a subject has a memory experience, and they would report it by claiming to remember a certain fact, then, strictly speaking, the content of their memory experience is that it originates in a particular perception of that fact (Fernández 2019, 79). The reflexive view, however, is neutral on whether or not the past perception at issue needs to have been phenomenally similar to the subject’s current experience of having their memory. TRV, by contrast, is not neutral on this issue. As we will see below, this difference turns out to be relevant for evaluating which view best captures some of our intuitions about the veridicality of memories.

22. Other self-referential views have trouble managing this difference in our intuitions. On John Searle’s ‘causally self-referential’ view, for example, a memory of seeing a flower represents that the memory was caused by the visual experience at issue, which in turn was caused by the presence of the flower (Searle 1983, 95). Similarly, on the reflexive view discussed above, if a subject has a memory experience, and they would report it by claiming to remember a certain fact, then the content of their memory experience is that it originates in a particular perception of that fact (Fernández 2019, 79). Both views commit us, then, to the position that the content of my memory M in W₀ must make reference, specifically, to perceptual experience P. This is not problematic when it comes to accommodating our intuitions about W₂. Given that I do not have P in W₂, both views correctly predict that M in W₀ is not true of W₂. But, since I do not have P in W₄, both views also predict, incorrectly, that M in W₀ is not true of W₄.

23. However, one might be concerned that, even if TRV captures our intuitions about the veridicality of memory, it makes the contents of our memories implausibly complex. After all, subjects would not express the contents of their memories in the way in which TRV spells them out. This is a reasonable concern that has been raised for other versions of the self-referential approach to memory. David Armstrong, for example, raises this kind of concern, in (Armstrong 1991, 154), against John Searle’s view that memories are causally self-referential. The response to this concern is that, for the purposes of our discussion here, the contents of memories are only meant to constitute the truth-conditions of those memories. And such conditions are not always epistemically available to the subject through introspection. For discussion, see (Searle 1991, 184).