The literature on observer perspective memory typically holds that it is a phenomenon that is dependent on reconstructive processes at the moment of retrieval. On such an understanding all visual memory imagery would be encoded from a field perspective, and the change to an observer perspective would occur at retrieval. This line of thought may explain why older memories are often recalled from an observer perspective. Information that was initially encoded as a field perspective becomes semanticised over time, where contextual details are lost, and is eventually reconstructed from an observer perspective. But if one acknowledges the wholly (re)constructive nature of memory one must also consider the context of encoding. One must also consider the possibility that some experiences are encoded into observer memories.
The constructive processes involved in memory encoding are thought to involve selection (where only certain stimuli are encoded), abstraction (where meaning is abstracted from the information selected and some content is lost), interpretation (where relevant prior knowledge is invoked to interpret the event), and integration (in which a holistic representation is formed from the products of the selection, abstraction, and interpretation processes). Importantly, these same constructive processes will be employed in the encoding of both field and observer perspectives, but may select for the salient information in both cases. It is not the case that observer perspectives necessarily involve more construction. Perhaps most experiences will unfold while we are attending to information that is apposite for the construction of field perspective memories. But sometimes, in some circumstances, the information that has been selected, abstracted, and interpreted from an event will be integrated and encoded into an observer memory.
In such cases, observer perspectives are neither false nor distorted memories. Remembering from-the-outside may accurately reflect the content of the original experience. In essence, there may be observer perspective experiences, even though such experiences do not involve seeing oneself from-the-outside. The information used in the construction of observer memories may be non-visual, and in observer memories nothing need have been added to the content of memory.
I suggest that observer perspective memories may be constructed from information that was available at the time of the original experience. But what are these external perspectives at encoding? Observer perspectives differ from field perspectives in virtue of their spatial perspectival characteristics. In observer perspectives one sees oneself in the remembered scene from a vantage point that one did not occupy at the time of the original event. These two elements of observer perspectives—the external point of view and the self-presence of such images—are things that need to be explained.
Episodic memory is inherently spatial. The events that we remember occurred in a particular spatial context. Yet there are a number of ways in which spatial information can be used by perceptual and cognitive processes in order for the cognising agent to represent external space and navigate through it. Broadly speaking, there are two primary types of spatial representation. They may be either egocentric: locating objects in the environment from a frame of reference relative to (part of) the body. For the visual modality, the egocentric reference frame involves viewpoint-dependent representations of the visual scene. These viewpoint-dependent visual representations can be seen as analogous to field perspectives. However, there is now a wealth of evidence that spatial cognition in humans (and other animals) is based in large part on non-egocentric spatial representations. Such spatial representations are allocentric: locating objects in a frame of reference centred on some feature or object in the external environment. In the visual modality these spatial representations will be viewpoint-independent.
There are different ways of construing allocentric spatial representations, different notions of allocentric space. One way of understanding allocentric spatial representation is to think of it as involving a virtual point of view, where space would be centred from a location where there is no actual object or person. This notion of a virtual point of view is developed in an influential theory of spatial representation. On this view, the “allocentric spatial system … represents the environment from any location and includes within itself a representation of the subject-as- object” (O’Keefe 1993/1999: 44-45). In effect, there are two points of view: the current perceived one, to which a self-representation would be assigned, functioning as something like a place marker, and one which would be free to move to imaginary locations and view the scene from a different perspective. Importantly, even though the allocentric spatial system can be understood as a memory system, it can be used online as one engages in occurrent cognising of the spatial environment.
I propose that such allocentric representations may be involved in the construction of observer perspectives. In both observer perspective memories and allocentric representations there are two points of view: a representation of the self from-the-outside and a detached point of view from which one visualises oneself. Allocentric information available during perceptual experience can account for the detached point of view and the representation of the self we find in remembering from-the-outside.
But if the self-presence of observer perspectives does not involve visually perceiving oneself, how does such a representation of the self arise? How is the self-presence of observer perspective experiences constructed? Again, the answer can be found in how we process spatial information. One key idea is that spatial representations based on one sensory modality can be translated or transformed into a different modality. For example, tactile or kinaesthetic information may be translated into visual imagery. In other words, such cross-modal transformation of information can lead one to generate a visual image even without the input of visual perception: a nonvisual source of information may be translated into a visual idea of that information.
I suggest that precisely this kind of multimodal integration of information occurs in the construction of observer perspective memories. Contra the objections from the perceptual impossibility and perceptual preservation arguments (outlined in the previous post), one need not perceive oneself from-the-outside in order to have a memory in which one ‘sees’ oneself from-the-outside, and nothing need be added to the content of such memories. The external perspective and the self-presence of remembering from-the-outside can be available at the moment of encoding. Of course, this is a hypothesis, and whether observer perspectives are actually constructed in this way is ultimately an empirical question. Nonetheless, this picture fits with some influential and empirically supported theories of episodic memory, which emphasise the importance of the multimodal integration of information, including spatial information, for the construction of representations of the personal past. My claim is that this spatial information involves allocentric representations, and that this information is available at encoding.
The picture I develop is also supported from a quite different literature―work on social phobia and anxiety disorders. One model of social phobia posits that a distorted image of one’s public self plays an important role in this anxiety disorder: “while in social situations, patients experience spontaneously occurring images in which they ‘see’ themselves as if from an observer’s perspective” (Hackmann et al. 2000: 602). Crucially, though, ‘see’ in this context is not taken to refer to a direct visual perception of oneself. Rather, it is thought that “the observer perspective in social phobia is problematic because the perspective is constructed from interoceptive information such as the subjective intensity of symptoms” (Wells & Papageorgiou 1999: 658). Even though such observer perspective experiences may involve images that are inaccurate and problematic in the context of anxiety disorders and social phobias, more benign observer perspective experiences may, under the right circumstances, be available to us all.
Indeed, this way understanding observer memories, as involving, at least in some cases, observer perspective experiences, helps explain some of the empirical evidence on point of view in memory imagery. Recall, for example, that observer perspectives are more common for events that involve a high degree of emotional self-awareness, such as giving a talk in public. And some studies show that women are more likely than men to remember events from an observer perspective in general, but especially for situations in which they feel sexually objectified (Huebner & Fredrickson 1999; cf. Rice 2010). In situations like these, in which one is emotionally self-aware, the context of encoding seems to be heavily influencing the content of memory. The understanding of memory I develop helps account for this by showing the senses in which one can adopt an external perspective on oneself at the time of the original experience. For a more complete account of observer memory we must pay attention to the context of encoding as well as the context of retrieval.
I have provided a way of understanding the spatial perspectival characteristics of observer perspectives such that one can adopt an external perspective on oneself both in memory and during perceptual experience. This still leaves us with questions as to the nature of this external perspective itself. The self in observer perspectives is viewed from an external point of view. But by whom? And in what way? I provide answers to these questions in my final post.
Thank you for this new post. I think you make a strong argument.
One thing this makes me think about is how we use episodic memory in way-finding. The egocentric view, at any point along our journey, is just a pattern of shades and tones, unless placed within an ongoing model of external reality. Object models are elicited in the neocortex (e.g. the inferotemporal lobe), but the model is only mapped onto allocentric space within the hippocampus / parahippocampal gyrus / retrosplenial system. In that latter system, the egocentric view is mapped onto a 3D spatial model. This model allows us to keep track of aspects of our environment (like where the exits are) that may not be in our current view.
Memory allows the view from time a to be compared to time b, and conclusions drawn, thereby, about what the change in view says about the world. Every movement through the space helps map new information onto the model of that space, until it is familiar enough, that one can accurately imagine the view from any location or angle in the space, even a location that one hasn’t been in yet.
This ability to imagine the space shows that our brains are constantly trying to map those views onto an allocentric perspective. Presumably, this is to help us be able to pinpoint where a memory took place, so we can learn from that information, and later avoid or return to that spot. It’s almost like one of those GPS markers, a pin in the map, which says: this memory happened here.
Since constructing that allocentric information is a continual part of memory encoding, it makes sense to me that the memory can thereafter migrate from the subjective perspective to whichever perspective best encapsulates the gist and narrative of the memory. After all, memories need to be useful more than they need to be accurate. Those two adjectives (useful + accurate) usually occupy the same Venn diagram space, but sometimes, the clearest lesson, that will most efficiently drive behavior, may come from an altered memory perspective.
Hi Matt,
Thanks so much for sharing these thoughts. Great stuff, very interesting.
Allocentric representations are thought to be more enduring, which would make the comparisons you talk about at different times available over longer time scales. Of course, I would want to suggest that, at least in some cases, the perspective in observer memories is not an ‘altered’ one.
That is also a fascinating autobiographical fragment that you’ve shared―the case of your ‘invisible observer’. Sartre provides a famous example which is similar, where someone who is peeping through a keyhole, and then hears the footsteps of another person, becomes aware of being an object to the other person’s look. The look of the ‘Other’ can, for Sartre, make us objects to ourselves. Importantly, we can still experience this objectivising look of the Other even if they are not really there and we only imagine them. Interestingly, you might expect that shameful events would more likely be recalled from an observer perspective, but there is evidence to show that self-evaluative emotions like shame were not associated with more observer perspectives compared to other emotions (e.g., Berntsen & Rubin 2006).
Thanks again!
Your words about the imagined objective view in social situations reminds me of an autobiographical phenomenon, what I used to think of as ‘the invisible observer’. By which I mean, when I was by myself, but doing something that was considered socially verboten (e.g. farting or picking my nose), I would still suffer some pang of being ashamed, as if being observed. It wasn’t like I believed there was someone there; but it still kinda felt that way. Usually, it felt like someone just over my right shoulder was watching my every move. That felt presence only faded in my 40’s.
I have wondered if that felt observer originated in childhood, because I was afraid of my parents observing me at shameful moments. So I modeled their presence behind me, just in case they were there, and tried to manage my behavior accordingly.
That artificial perspective of the imagined observer might also help explain the migration in perspective in memories, especially childhood ones.