1 Introduction

This paper explores some surprising effects of trauma on memory, and a puzzle they generate: the puzzle of observer memory for trauma. I articulate the puzzle and argue for a novel set of views that offers a way out.

My focus is on psychological trauma rather than physical trauma (such as brain damage). But I mostly just say “trauma”. Everyday talk uses this term and its cognates loosely. We might describe any highly unpleasant experience as traumatic. The U.S. Department of Veteran Affairs is stricter: a trauma is roughly an experience of a “shocking and dangerous event” in which you think “your life or others’ lives are in danger.”¹ And in the most recent Diagnostic and Statistical Manual of Mental Disorders (DSM-V):

Trauma is now defined as exposure to actual or threatened death, serious injury or sexual violence in one or more of four ways: (a) directly experiencing the event; (b) witnessing, in person, the event occurring to others; (c) learning that such an event happened to a close family member or friend; and (d) experiencing repeated or extreme exposure to aversive details of such events, such as with first responders. Actual or threatened death must have occurred in a violent or accidental manner; and experiencing cannot include exposure through electronic media, television, movies or pictures, unless it is work-related. (Jones & Cureton, 2014, pp. 261–262)

I will not defend a theory of trauma. The above helps provide an adequate understanding of trauma and the types of events that cause it.

Trauma is perhaps surprisingly common. Roughly 51% of women and 61% of men experience a trauma at some point (Leskin et al., 1998, p. 984). For women, this trauma more commonly involves “sexual assault and child sexual abuse”; for men, it more commonly involves “accidents, physical assault, combat, disaster, or to witness death or injury”.² Although over half of the U.S. population experiences a trauma at some point, it is still plausible that trauma is or results in an atypical cognitive condition. Experiencing or continuing to suffer from trauma is not a normal state. It’s a disruption of normalcy.

This disruption can lead to posttraumatic stress disorder (PTSD), but this is not typical. Only 7-8% of the population develops PTSD in their life. Although women are less likely to experience trauma than men, they are more than twice as likely to ever develop PTSD (10% of women, compared to 4% of men).³ This may be due to the kinds or severity of the trauma women more commonly experience.

There is little discussion of how trauma matters in epistemology, yet much to discuss. To forgo some breadth for depth, I look at how trauma matters primarily in the epistemology of memory.⁴ For the most part, I focus on episodic memory, which is our memory for past events that tends to be imagistic and autobiographical, sometimes called “what/where/when” memory (Tulving, 1972). Episodic memory is responsible for you recalling yourself eating breakfast this morning, and picturing what you were eating and wearing. Contrast this with semantic memory, which is memory for information in propositional form. Semantic memory is responsible for you recalling that you yet again ate a sugary cereal for breakfast this morning.

It’s worth looking at how trauma matters in memory and epistemology. Work here might help us understand, identify with, and respect individuals who might have seemed unlike us. Perhaps we are inclined to incorrectly and unreflectively downgrade the credibility of testifiers who report on their own trauma, assuming these individuals are unlikely to report objectively about the trauma or the event that caused it, perhaps especially when it has led to PTSD. Work on the epistemic significance of trauma may help us weigh this testimony properly. And labor here can help us see how to respect the evidence from our own trauma.

2 The puzzle of observer memory for trauma

Trauma tends to affect memory in several epistemologically significant ways. Here I focus on just imagery and perspective, and here emerges the puzzle. In the section after I evaluate support for two pieces of the puzzle.

Experience is perspectival. To no reader’s surprise, we by default experience from a first-person perspective. And much that we recall by way of episodic memory is from the first-person perspective or field perspective. In recollecting some event you witnessed, it’s as though you are seeing it with your eyes again. But not all recollection is like this. Think of an instance when you gave a talk or a speech you were nervous about, or when you were self-conscious while entering a full room, or when you went swimming.⁵ This memory may not have the field perspective. Instead it might have a third-person or observer perspective. It’s not as though you see it with your own eyes again. It’s as though you’re watching yourself act out the event. Since our default experience is from the first-person, it’s odd that we sometimes remember things from the third-person. Ordinarily this is a visual perspective we never had. The different perspectives in memory matter, and not just because one is the original perspective and the other isn’t. The perspectives concentrate on different details. Memory in the field perspective tends to come with greater sensory and affective information. Memory from the observer or third-person perspective tends to come with more information about the remembered event.⁶

Trauma is importantly related to perspective in memory. Memory for a traumatic experience tends to be from the observer rather than field perspective.⁷ It may not be so hard to explain this. As noted, field memories are richer in sensory and affective detail. Some interpret our remembering traumatic experiences from the observer perspective rather than from the original perspective as a cognitive avoidance strategy. The change in perspective minimizes “emotional arousal” and spares one of the “horror of reliving” the traumatic event.⁸ The observer perspective, with its emphasis on different types of information, can make recollection less painful.

The first piece of the puzzle is beginning to emerge. Observer memory, the kind of episodic memory we tend to have for a traumatic event, ordinarily takes on a perspective other than the original. So it seems that:

1. Observer memory tends to misrepresent.⁹

That is, observer memory for an event tends not to fully accurately represent the event, or at least not depict it as it was originally experienced. Advocates of (1) think this tendency to misrepresent is special. Observer memory distorts the past in a way that, say, field memory does not.

And a tendency to misrepresent is an epistemically important defect in any kind of memory. Many will find it plausible that:

2. If a memory type tends to misrepresent, it cannot yield knowledge of or justification for believing details of past events.¹⁰

There are significant limits to the epistemic powers of memory types that tend to misrepresent. These memory types can confer neither knowledge of, nor justification for believing, propositions of a certain class. These are propositions about the specifics of the event represented by the memory type. There is no support for believing details about an event from a form of memory that tends to misrepresents events. At best, this form of memory could support beliefs about the event’s gist – its general features. Suppose observer memory tends to misrepresent. Then it according to (2) at best supports beliefs about an event’s gist. It’s not entirely clear what counts as a detail or as part of the gist. But examples can adequately guide us here. You remember from the observer perspective an event in which you heard some tragic news over the phone. By this observer memory you may now know that news, and may now know that you learned about it over the phone. But by it you perhaps cannot know what you were wearing, whether you were sitting or standing, what your first thoughts were in response, etc. In order to have this knowledge from observer memory, it must not tend to misrepresent.

A final, intuitively credible claim completes the puzzle. It’s plausible that observer memory does support believing details about the past. More precisely:

3. Observer memory can yield knowledge of or justification for believing details about past trauma.¹¹

According to (3), observer memory is not limited to supporting beliefs about a traumatic event’s gist. It can support beliefs about the details too.

(1), (2), and (3) are jointly incompatible. At least one must be false. At first glance, none is obviously false, and so we have the puzzle of observer memory for trauma. We escape the puzzle by discrediting (1), (2), or (3). The defended ways out of the puzzle are to deny either (1) (McCarroll, 2017, 2018) or (3) (Fernández, 2015). The latter solution appears especially troubling. Denying (3) may threaten much of our apparent knowledge of the details of traumatic events, since much of this apparent knowledge is from observer memory.

In the next section I defend a novel solution to the puzzle: I deny (2). I think (1) and (3) are compatible. What’s more, I think both are true. In addition to challenging (2), I present a new reason for accepting (1). I don’t say much to support (3), although I explain why its denial is not so disconcerting after all. I also develop further points about the epistemic powers of observer memory. My overall position, then, is that observer memory does tend to misrepresent, but it still has epistemic power regarding details of the past, although with special limits; but it has other epistemic powers too. This position consists of four theses. I defend each one at a time.

3 Solving the puzzle

3.1 Observer memory misrepresents

I claim (1) is true – observer memory tends to misrepresent. Before I defend this, a caveat is in order: the exact contents of episodic memory are underexplored. When you have an episodic memory of some event, which properties, or kinds of properties, could be represented as instantiated by way of this recollection? The answer is not obvious. A liberal and possibly naïve answer allows that many varieties of both low-level and high-level property types are represented. In episodic memory it may not simply be such low-level properties as colors (e.g., being red) and shapes (e.g., being round) that are represented as instantiated. High-level properties (e.g., being a hat; belonging to Jose) may also be represented.

The literature on episodic memory generally accepts the liberal answer, and without much defense. This answer may be mistaken. It could be that far fewer property types are represented in the contents of episodic memory, and this may dissolve the puzzle we face here. Maybe episodic memory is much more schematic, and we simply interpret objects in episodically recalled events as having such properties as being a hat or belonging to someone. There is much work to do in cataloging the contents of episodic memory. For simplicity, this paper works within the framework of the received view, according to which we can identify the contents of field and observer memories in a simple if naïve way. It’s the work of another day to scrutinize the received view.

Now, on to (1). The most obvious support for it is that observer memory represents from a new visual perspective. In observer memory, including most observer memories for trauma, you observe yourself. You are represented as appearing various ways. But this was not part of your original experience. Some philosophers (e.g., Fernández, 2015, p. 541) argue observer memory thus fails to carry out memory’s function of preserving information from the original experience. Observer memory is instead including new content. That content has some other source – other memories, imagination, experiences after the original event, and so on. Observer memory is thus distortive and not likely accurate.

This seems to be the initially most tempting reason to accept (1). But it may not survive scrutiny. For one, granting that observer memory includes new information, there isn’t an empirical case that observer memory is thereby less accurate. As Chris McCarroll (2017, pp. 328 n.11) points out, “The empirical evidence that does exist on visual perspective and accuracy in memory is scant and inconclusive.” Research here leaves open that observer memory is as accurate as field memory.

Additionally, as McCarroll (2017, p. 326) points out, this tempting reason falls short of supporting (1). Rather, it mainly illustrates that observer memory is reconstructed. Its content has been cobbled together from parts of various memories, experiences, and perhaps imaginings. But evidence of reconstruction isn’t by itself evidence of misrepresentation. After all, field memory is also reconstructive. It too synthesizes information from multiple sources. Yet we don’t therefore reasonably doubt its accuracy. More generally, episodic memory (which, again, includes memories from either the field or observer perspective) is reconstructive. It typically synthesizes. We shouldn’t suppose reconstruction guarantees distortion. Observer memory reconstructs, but it does not thereby misrepresent. Observer memory isn’t purely preservative, but it doesn’t follow that it is less likely accurate. Reconstruction could simply result in memory with different accurate content, content not identical to that of the original experience or to that of field memory for the original event, but content still faithful to that event.

Still, it may seem (1) is plausible because some of the new content of observer memory is clearly inaccurate. You are recalling from a visual perspective that was not your own. That itself renders the observer memory inaccurate. Or, since in observer memory you view yourself from this new visual perspective, the memory is inaccurate because it misrepresents you as having seen yourself during the original event.

But this concern isn’t clearly correct. It could be, as McCarroll (2017, p. 328) thinks, that this visual perspective itself is “not part of the content of the memory: the content of observer perspective memories is the past event, not having seen oneself at the time of the past event.” The idea is that observer memory is representing an event from earlier. It’s not representing you as having witnessed yourself during that event, or as having visually appeared to yourself then in a certain way. And the same event can be accurately represented from more than one perspective. The different perspectives are simply different “modes of presentation” of the same event (McCarroll, 2018, Chapter 6). So, it looks like the different visual perspective by itself isn’t evidence of misrepresentation. McCarroll’s reply needs to be ruled out before the change in visual perspective compellingly supports (1).

The initially tempting reason to accept (1) is not as good as it appears. Still, I think (1) is correct. And I can support this while remaining neutral on whether observer memory inaccurately represents you as seeing yourself, or as having a visual perspective that you in fact didn’t have, at the time of the event. I think observer memory typically involves not just new information, but new information that isn’t quite right, such that observer memory tends to misrepresent. I’ll build up support for this point.

Consider a case from Mark Rowlands (2016, pp. 189–190). Rowlands remembers an event from childhood, in which his father left the room briefly, shortly before a televised boxing match with Cassius Clay (Muhammad Ali) began. When Rowlands’s father returned, he saw on the TV that the match was already over. Rowlands (2016, p. 190) remembers his father’s expression going “from confusion to suspicion to acceptance to joy,” vividly. But Rowlands can tell this memory is not quite right in one regard. It’s not depicting his father’s age properly. Rowlands isn’t picturing his father as young as he was during the event, but instead as he looked toward the end of his life. Rowlands is inaccurately remembering aspects of his father’s face.

Now, Rowlands does not say this memory is from the observer perspective. But even if it is, there is no special indictment against observer memory here. After all, Rowlands could tend to inaccurately episodically recollect his father’s appearance from either perspective, field or observer. But Rowlands’ example suggests others. I invite you to think of an observer memory of your own, of an event from a decade or two ago. What does your face look like? Are you visualizing yourself at the right age? I suspect that by default you’re not. You’re instead visualizing your face in the recalled event overly resembling your current face, just as Rowlands pictured his father not as a young man but as he looked much later. You’re not adjusting the age of your face properly. This helps explain why we are often surprised by how young we look in old photos. Prior to seeing a photo, your observer memory for the photographed event would have inclined you toward visualizing yourself with an older, or otherwise different, face. Indeed, you might have even forgotten what you looked like at the time. Without the assistance of a photo, you might have even been unable to produce accurate imagery of yourself in the event.

So, there is reason to accept (1): observer memory tends to misrepresent. It tends to misrepresent the face of the remembering subject. Crucially, my support for (1) is distinct from any support McCarroll addresses. This is because I’m not suggesting your observer memory inaccurately represents you as appearing to yourself at the time of the event. Instead I’m suggesting that it inaccurately represents your appearance at the time of the event.¹² More specifically, it misrepresents your face, and likely more. Field and observer memory might both tend to misrepresent the faces of others, but observer memory has a special flaw. Field memory does not have a tendency to misrepresent your face, as it does not normally depict your face, while observer memory normally does.

You might worry about my support for (1). Suppose observer memory and field memory are just different modes of presentation for the same past event. It may then seem that observer and field memory misrepresent together or not at all. After all, they reconstructively represent the same thing. If one perspective misrepresents, the other must also, albeit from a different perspective. So, there is no difference in how the perspectives tend to misrepresent. And since field memory does not tend to misrepresent, observer memory mustn’t either. There is no case for (1) here.

However, the modes of presentation allegedly focus on different aspects of the event (McCarroll, 2017, p. 330). It’s plausible that this different focus involves a representational difference; the modes of presentation for the same event have distinct content. And a difference in representation allows a difference in misrepresentation. So, even if field and observer memory are just different modes of presentation for the same event, and field memory tends not to misrepresent, observer memory could still tend to misrepresent.

I have defended the claim that observer memory tends to misrepresent. It is disposed to err in a way that field memory is not. This is not because of how it portrays your past visual perspective or past observations of yourself. This is because of how it represents your past appearance.

3.2 Observer memory has epistemic power regarding past details

So far I have accepted the first piece of the puzzle: observer memory tends to misrepresent. Still, it is plausible that observer memory can yield either knowledge or justification for believing details of past events. That is, I reject (2), the second piece of the puzzle, a piece taken for granted by others who have attempted to respond to this puzzle.

According to (2), if a memory type tends to misrepresent, it cannot yield knowledge of or justification for believing details of past events. Initially (2) is attractive. A tendency to misrepresent is an epistemically relevant defect. How the tendency matters will vary on different theories of justification or knowledge. But here’s one view of its relevance. Memory is supposed to do certain things and not others. Memory appears to have one or more proper functions. One of its proper functions is epistemically pertinent: its preservative function (Fernández, 2015, p. 541). Memory is supposed to retain past information. And observer memories don’t carry out this preservative function. Observer memories contain altered rather than purely preserved information. They are not appropriately produced. Presumably this distortive production compromises likely accuracy, at least with respect to the recollected event’s details. Since observer memory is not carrying out the proper function of memory, and is not likely accurate about details, it cannot confer justification or knowledge concerning anything other than an event’s gist. So, observer memory, if it tends to misrepresent, has no epistemic power regarding the details of the past event. In other words, (2).

It would be disappointing to have to accept both (1) and (2), given that they together entail that (3) is false. We would have no knowledge of or justification for believing details of past events from the type of episodic memory we tend to have from trauma, if (1) and (2) are true. Still, not all would be lost – and not just because observer memory could nonetheless support beliefs about the event’s gist. Observer memories, including those from trauma, could still have non-epistemic benefits. They could still be evolutionarily adaptive, or could carry out a proper function of memory other than the preservation of past information. Observer memory could undertake memory’s function of organizing personal past events into a coherent narrative. In the case of trauma, observer memory could do this by allowing a kind of “‘phenomenal distancing’ from the traumatic event” (Fernández, 2015, p. 542). (1) and (2) together do not fully impugn observer memory.

Despite (2)’s appeal, there is no need to accept it. (2) concerns a misrepresenting memory type’s power to confer knowledge and justification. As I will show, (2) is false on what might be the leading theories of justification – namely, evidentialism and process reliabilism. I need not show a misrepresenting memory type can confer knowledge; (2) is in peril if a misrepresenting memory type has either of the epistemic powers (2) denies. I will also show the above support for (2) is inconclusive. So, there is overall reason to doubt (2). This provides a way out of the puzzle.

It is more readily apparent that (2) is incorrect on evidentialism than on process reliabilism. According to evidentialism, approximately, a subject is justified in having a doxastic attitude toward a proposition if and only if that attitude fits the subject’s evidence (Feldman & Conee, 1985). That is, believing, disbelieving, or withholding belief in a proposition is justified for a subject just when the subject’s evidence supports having that attitude toward that proposition.

Here is why (2) is false on at least one plausible version of evidentialism. Suppose evidentialism is true. And suppose a given memory type tends to misrepresent. Still, on some credible versions of evidentialism, having an experience is justifying evidence.¹³ A given memory experience may for a subject have such phenomenology or content that it provides for that subject evidence that prima facie justifies the subject in believing its content. If the subject lacks a defeater for this justification, believing is for the subject overall justified. (A defeater is a reason to not believe, or to doubt some justification.) Evidentialism, then, is compatible with a misrepresenting memory type providing prima facie justification. The memory type may have justifying features despite its tendency to err.

What’s more, a subject with this justification may lack defeaters for it. The subject may lack evidence that the given memory type tends to misrepresent, or may otherwise lack reason to doubt the deliverances of observer memory. Perhaps as far as the subject can tell, the memory type tends to get things right. The subject would have to learn something, perhaps about how the memory type misrepresents, in order for its overall justification to weaken. But even then, the memory type provides prima facie justification. In short, a mere tendency to misrepresent is no obstacle to providing prima facie or overall justification on evidentialism. But then (2) is false. If evidentialism is true, a memory type that tends to misrepresent can still have justificatory power.

A simple revision might improve (2). Given the distinction between prima facie and overall justification, and given the significance of defeaters, it is more plausible that:

2*. If it is reasonable for S to believe that a memory type tends to misrepresent, then for S it cannot yield knowledge of or overall justification for believing details of past events.

If it’s reasonable for a subject to believe a memory type misrepresents, that subject appears to have a defeater for any justification from that memory type (at least regarding the details of an event recollected by that memory type). In such a case, that memory type does not overall justify believing for the subject.

(2*) is a step up. It is likelier true than (2), and perhaps also likely true. However, it has implications just about overall justification. It is compatible with (2*) that a misrepresenting memory type provides prima facie justification. But then, while (2*) is more plausible, it does not contribute to a puzzle. There is no inconsistency between (1), (2*), and (3). Despite observer memory tending to misrepresent, it could still yield justification for believing details about a past event, and perhaps yield knowledge too when the subject has no relevant defeaters. I conclude that (2) is false on evidentialism, and that a promising revision to (2) is innocuous for our purposes.

On process reliabilism, (2) is no more plausible, although this may be less clear at the outset. Oversimplified, process reliabilism says that a belief is justified just in case it results from a reliable belief formation (or retention) process (Goldman, 1979). A reliable belief formation process is one that tends to yield true beliefs rather than false ones.¹⁴ I claim that a memory type that tends to misrepresent can still be reliable, and so the beliefs it forms can still be justified, on reliabilism.¹⁵ So on reliabilism, observer memory could still provide justification even if it tends to misrepresent; (2) is false.

My claim may seem incoherent. How could a process type that tends to misrepresent still be reliable, where reliability is a tendency to yield truth? I will explain. Whether a belief formation process type is reliable depends on whether it tends to produce true beliefs. Observer memory’s reliability, then, will depend on whether it tends to produce true beliefs. To check its reliability, we should look at just which beliefs we form from observer memory. And reflection will show these beliefs don’t tend to be false, or needn’t tend to be false, even if observer memory tends to misrepresent. At any rate, they aren’t false notably more often than beliefs from other forms of memory, such as field memory, which we commonly assume to be reliable.

Let’s see why this is so. I have suggested that observer memories, unlike field memories, tend to predictably inaccurately represent certain things, such as the subject’s own face. Others have suggested observer memory inaccurately represents you as having a certain perspective (one you did not have), or as having observed yourself (when you did not). It initially seems that a misrepresenting memory type will be unreliable because we suppose we will form too many beliefs that share the contents of the misrepresenting memories of that type. But notice that we don’t tend to form beliefs based on the predictably inaccurately represented portions of observer memories. For example, suppose it is the case that an observer memory inaccurately represents you as seeing yourself in the past event, or as having experienced the event from a particular point of view. Still, you do not in fact form the false belief that you experienced the event from that other point of view. And you don’t believe you saw yourself during the event! You instead tend to form beliefs that share just the accurate content of the observer memory.

The moral here is that a memory type may misrepresent in ways that do not tend to affect belief. This appears to be the case with observer memory. Processes of belief formation involving observer memory do not appear to yield beliefs that are false especially often. Even though observer memory misrepresents more often than field memory, they could be equally reliable. As it is plausible that field memory justifies, reliabilists will regard it as reliable enough to be justifying. If reliabilists are correct here, observer memory could also be reliable enough to be justifying.

A process of belief formation involving a misrepresenting memory type can still be reliable and thus, on reliabilism, still confer justification. Even if observer memory tends to misrepresent, it does not follow that observer memory cannot confer justification or knowledge. On reliabilism, (2) is false.

I have argued that on evidentialism and process reliabilism, the two most promising theories of justification, (2) is mistaken. It is also worth showing why the initial support for (2) is in fact inconclusive. The initial reason for accepting (2) had to do with memory having a proper preservative function. Observer memory does not carry out this function, and so is not justifying. But a compelling case for (2) requires more. There is no support for (2) here unless a proper functionalist theory of justification is true. This sort of theory says, at a minimum, that justification results from the proper functioning of cognitive faculties. But we need not accept such a theory. There is no presumption in favor of this sort of theory, and there are promising alternatives. On many alternatives – including evidentialism and process reliabilism – an improperly functioning faculty can still have justificatory power. Additionally, even if there is reason to accept a proper functionalist theory, it is a further question whether memory’s proper epistemic function is purely preservative. Proper functionalism is compatible with memory having an at least partially generative proper epistemic function. A case for (2) requires further development. The particular support for (2) considered above is inconclusive in the absence of reason to accept an appropriate proper functionalist theory of justification.

On what might be the two most popular theories of justification, (2) is false. And a natural revision to (2) improves its credentials but generates no puzzle. The initial support for (2) is inadequate. So there’s reason not to accept (2). Trauma affects episodic memory in an important way, but it doesn’t follow trauma eliminates memory’s epistemic power. A way out of the puzzle opens before us.

3.3 Observer memory has less epistemic power than field memory

I’ve argued that observer memory has epistemic power. Although I think observer memory misrepresents, on a few leading theories it can still be a source of justification. I’ve accepted (1) and denied (2). Another way of clearing room for the view that observer memory has epistemic power is to accept (2) but to deny (1). Some who think this is how observer memory could have epistemic power also posit that this power is on a par with that of field memory. Both are sources of knowledge. For instance, McCarroll (2017, p. 331) writes, “Because observer perspectives are just (non-distorted) memories, they provide the same epistemic benefit as field perspectives, albeit with a focus on different information about the past.” Elsewhere he (McCarroll, 2018, pp. 184–185) adds:

But if memory is a source of knowledge—in the sense that it provides us with ways of thinking about the past, and we obtain knowledge about the past through memory—then observer perspectives as (potentially) genuine and non-distorted memories can be a source of knowledge. Remembering from-the-outside [i.e., observer memory] can be epistemically beneficial.

If observer memory is like field memory, then it not only can be a source of knowledge, but indeed frequently results in knowledge. When it does not, something out of the ordinary is happening – for example, the subject is confabulating, and so the subject’s belief from the memory is false.

However, observer memory is not on a par with field memory, and it rarely confers knowledge. Observer memory does prima facie justify. But, typically, a subject who has an observer memory has evidence that this memory misrepresents. This evidence weakens the overall justification for the subject from the observer memory. The subject has a partial defeater for the justification that observer memory provides. The subject has this partial defeater because she has evidence that the observer memory misrepresents. And the scope of its inaccuracy is usually unclear. But, clearly enough, observer memory is less accurate than closely-related memory types, such as field memory. Of course, observer memory still might not be a worse overall justifier than field memory if observer memory lacks some flaw field memory has. But we have no reason to suppose this is the case. So, there is less overall justification from observer memory than from field memory.

To be sure, a subject can have evidence that her episodic memory is inaccurate, where the inaccuracy is not peculiar to the memory’s perspective. The subject would misremember the event whichever way memory presented it, from the field or observer perspective. Mere misremembering is no special problem for observer memories.

But there are other, common enough, distinctive inaccuracies of observer memory that the subject will have evidence of. For example, as I suggest above, observer memory typically inaccurately represents the subject’s face (although confirming this is in part an empirical matter). And a typical subject has evidence of this inaccuracy. A typical subject has evidence that she didn’t look as she appears in the remembered event. The typical subject has evidence that her face in the remembered event overly resembles her current face, or that her face is otherwise wrongly depicted. The subject may overlook this evidence, but she still has it.

This evidence helps defeat the subject’s justification from observer memory for believing this inaccurate content. But all else being equal the evidence also partially defeats the justification for believing other content of the observer memory. Evidence that some content of a memory is incorrect is evidence that some of its other content is incorrect too, when there is no reason to suppose the inaccuracy is quarantined. Explanations about the memory as a whole being defectively generated or preserved have become more plausible. So, observer memory looks like a worse overall justifier than field memory. By representing the subject’s face, observer memory typically has an additional way in which it could err, and typically errs in that way, and the subject has evidence of this error. Perhaps this defeater fully defeats the justification that observer memory provides for believing any of its content, but I need not defend that claim. The level of justification required for knowledge is high, so even a partial defeater could prevent knowledge. There typically is an additional obstacle to observer memory conferring knowledge. At any rate, observer memory still confers less overall justification than field memory does. There are special limits to the epistemic powers of observer memory.

But here’s a worry for this claim. I may be right about most observer memories, but perhaps observer memory for trauma is exceptional. Observer memory for trauma might be exceptional if it, unlike other observer memories, has epistemic goods that compensate for its shortcomings. For example, observer memory for trauma can hyper-focus on certain details, thus yielding evidence that supports particularly specific beliefs about the traumatic event. A victim, for example, may have paid special attention to a mugger’s weapon during a traumatizing robbery. Observer memory for this traumatic event may provide more evidence about the appearance of this weapon, and so may have powers that other observer memories lack (Schacter, 1996, p. 210). So, even if the epistemic power of observer memory has special limits, observer memory for trauma has unique assets.

This worry is mistaken. Observer memory for trauma might have these assets, and so the epistemic power of observer memory for trauma may rival that of ordinary field memory. But observer memory for trauma is not yet on a par with field memory for trauma. That’s because field memory for trauma could also involve hyper-focus, delivering the same special evidence. Yet observer memory for trauma still has shortcomings that field memory for trauma lacks, such as errors about the subject’s own appearance. So observer memory for trauma is not yet on a par with field memory for trauma.

Although observer memory typically misrepresents, it does provide prima facie justification. But its patterns of misrepresentation leave its justification prone to greater partial defeat than justification from other forms of memory, such that it may not typically provide knowledge. Trauma, then, often affects memory, in a way where memory provides justification but perhaps not knowledge of details of past events.

This may seem to have the worrisomely skeptical implication that a subject cannot have knowledge about a past traumatic event in her life, even an event she remembers. Indeed, denying (3) – which says observer memory can yield knowledge of or justification for believing details about past trauma – also may seem to have this very implication. So, although I accept (3), I may still seem to incur the main costs of denying it. But that is not the case. Other sorts of memory, such as field memory or semantic memory, could provide a subject with knowledge of a past traumatic event, even if observer memory cannot.¹⁶

3.4 Observer memory has epistemic power regarding the present

I’ve argued that the kind of memory we tend to have for traumatic events misrepresents, yet still has epistemic power, but with special limits. The rest of this section explains how remembering from-the-outside has additional upshots that are of epistemological significance. Its having these upshots is compatible with it not typically providing knowledge of the remembered event’s details.

Notice that the main puzzle I set out to solve simply concerned the relation between, on the one hand, observer memory and, on the other, knowledge of and justification for believing the details of a past event. The puzzle did not address observer memory’s epistemic power with respect to information on other matters, or events at other times. And it is not hard to see why. It seems commonsensical that any sort of episodic memory could inform us only or mostly about the past. After all, episodic memory is of past events.

The main puzzle may give a false impression here. Observer memory also helps provide knowledge of or justification for believing propositions about the present, or about other events. Recall that some who support the conjunction of (1) and (2) counterbalance this denial of observer memory’s epistemic power by proposing that observer memory still serves a narrative function. Memory that serves this function is “engaged in an inventive project wherein we build representations of our past by integrating content that we have acquired through our own experience with content from other sources, such as testimony, inference and imagination,” resulting in “a smooth and robust narrative of our lives” (Fernández, 2015, p. 540).

I propose that memory, including observer memory for trauma, that serves this narrative function can provide evidence. Memory serving this function can give a subject evidence about what her current narrative is, and evidence about how a recalled event fits in her current narrative. An observer memory for trauma that serves this narrative end may, for example, provide the subject with evidence that she currently interprets her personal past as involving great undeserved suffering, but also a triumph over it.

Observer memory for trauma that is not serving a narrative function can also be evidence. This memory can give a subject evidence about her current emotions toward a past traumatic event. The observer memory could, for instance, be evidence that the traumatic event is currently quite painful for the subject (otherwise she might recall it from the field perspective).¹⁷ Also, a subject might currently remember an event from a perspective, but also be aware of which perspective she remembered the event from in the past. Detecting that the perspective has not changed can give the subject evidence that her emotions toward or acceptance of the event has not changed. Detecting that the perspective has changed can give a subject evidence that her emotions toward or acceptance of the event has changed. If, for example, she is aware she previously remembered the event from the observer perspective, but now remembers from the field perspective, the subject may have evidence that the event has become less painful to her. This change in emotion is itself an event, one that observer memory is evidence for, but it is not itself the event remembered from an observer perspective.

In short, memory from the observer perspective can provide evidence that helps the subject be self-aware in a way. Insofar as trauma affects perspective, it helps provide this self-awareness. The overall justification from this evidence can be strong enough for knowledge. There is no partial defeater here, as there is with evidence from observer memory about the remembered event’s details. The subject’s awareness of how the observer memory misrepresents past details does not defeat the justification from the evidence I have described in this section. Observer memory for trauma can provide the subject with justification, typically undefeated, for believing: that she has a certain narrative, that the traumatic event has a certain place in that narrative, that she has certain feelings toward that event, and that these feelings have (or have not) changed.

It is not only observer memory for trauma that provides justification for believing along these lines. Other observer memories and field memories sometimes provide this self-awareness too. There is no glory in glorifying trauma or its effects. But we do well to see what we can learn by both.

4 Conclusion

I have suggested that trauma affects memory in a way that contributes to the puzzle of observer memory for trauma. I have devised an escape from the puzzle by showing how even a misrepresenting memory type can support beliefs about the details of past events. I have outlined special limits to this support. But I have shown that trauma’s effects on memory nonetheless put us in a position to be self-aware.

Acknowledgments

For helpful comments and conversations, I thank Earl Conee, Kevin McCain, Chris McCarroll, two anonymous referees, and audiences at Marquette University, the 2019 Southeastern Epistemology Conference, the 2020 Philosophy of Memory in the Bay Conference, and a 2020 Université Grenoble Alpes Virtual Philosophy of Memory Seminar.

References

---

1. https://www.ptsd.va.gov/understand/common/common_adults.asp↩︎

2. https://www.ptsd.va.gov/understand/common/common_adults.asp.↩︎

3. https://www.ptsd.va.gov/understand/common/common_adults.asp.↩︎

4. Trauma is of course not the only atypical cognitive condition that matters for both memory and epistemology. Consider for example dementia. Interestingly, in cases of dementia, “remembering how to do various tasks does often persist longer” than “explicit knowledge of one’s own past” (Sutton & Williamson, 2014, p. 321). Dementia might therefore provide evidence against intellectualism, according to which knowledge-how reduces to knowledge-that. Remembering how to do tasks appears to involve knowledge-how, and explicit knowledge of one’s own past might involve knowledge-that. And if intellectualism is true, knowledge-how cannot persist longer than knowledge-that. But in dementia knowledge-how may persist longer. I leave this potential problem for intellectualism for other papers to explore.↩︎

5. I borrow most of these examples from McCarroll (2017), who attributes them to Sutton (2010).↩︎

6. See Sutton (2010, p. 29).↩︎

7. Cf. Fernandez (2015, p. 541), Porter & Birt (2001) and Schacter (2001, p. 174). I leave open whether it is possible to have memory for a traumatic event that did not occur.↩︎

8. See Sutton (2010, p. 34), who quotes McIsaac and Eich (2004, p. 252).↩︎

9. See Fernandez (2015).↩︎

10. See Fernandez (2015). McCarroll (2017) also seems to accept (2), but unlike Fernandez he denies that observer memory satisfies the antecedent. For discussion of epistemic upshots of distortive memory, see Bortolotti and Sullivan-Bissett (2018), Puddifoot and Bortolotti (2019), and the final subsection of this paper.↩︎

11. See McCarroll (2017, 2018).↩︎

12. McCarroll (2018, Chapters 2–3) explores how memory can constructively encode, such that memory for an event has an observer perspective from the outset. The observer perspective is not, in other words, always added during retrieval. I’ve argued that observer memory tends to misrepresent the face of the remembering subject. My claim is compatible with memory constructively encoding, and does not imply that only during retrieval can memory take on an observer perspective. Still, observer memory will tend to misrepresent even experiences encoded from the observer perspective. This is not because of the encoded perspective itself. It is because observer memory, even in these cases, tends to misrepresent the appearance of the rememberer, such that she overly resembles herself at the time of the remembering.↩︎

13. See, e.g., Conee and Feldman (2008). The experience can be a justifier without itself standing in need of justification. In this way, the version of evidentialism above is, like most, foundationalist.↩︎

14. This too is an oversimplication of a fraught issue. For discussion of reliability, see Frise (2018).↩︎

15. An observer memory process, like other memory processes, may have many belief formation subprocesses, and may itself be a subprocess of many complex processes of belief formation. Identifying which of these processes is ultimately relevant to justification is famously difficult, and not a problem I will solve here. For discussion of this generality problem, see Feldman (1985) and Conee and Feldman (1998).↩︎

16. Cf. Frise (2022), where I argue more generally that a typical subject does not have knowledge of the past from episodic memory alone, but I grant that other forms of memory could provide this knowledge. Strange and Takarangi (2015) suggest episodic memory for trauma can be especially distortive, due to the subject committing source monitoring errors – mistaking a memory of merely imagined details of a past event as experienced details. So perhaps even field memory for trauma is often too flawed, leaving semantic memory to largely account for our having the knowledge of our trauma we reasonably think we have. Exploring this further is beyond the scope of this paper.↩︎

17. This fits well with Debus’s (2007) view that a subject’s emotion that accompanies an episodic memory is not an emotion remembered from the remembered event, but rather is the subject’s current emotional response to the event. Cf. Goldie (2012, pp. 51–52).↩︎