A New Model of Memory: Representational Change and Representational Success
Rebecca Copenhaver - Washington University in St. Louis
In the beginning of Descartes’ Meditations, the meditator searches for a scenario that would make them doubt their own senses.
The meditator admits that when it comes to things that are very small, or very far away, perceptual experiences might deceive.
But the fact that we’re bad at seeing, or hearing, or smelling things that are minute or distant isn’t enough to make us doubt our senses—it isn’t enough to give up on perception altogether.
You can forgive Mnemosyne (Memory) for feeling like the sibling who is forever judged under a different standard than her sister, and fellow Titan, Theia, the goddess of sight and vision. Maybe this is why Rossetti gives her a meme-able unimpressed face.
If we’re to believe the tabloids, Memory is notoriously unreliable. She leaves you hanging when you need her, she shows up when she’s not invited, and when she does show up, she seems so, I don’t know, vague and contrived.
But the tabloids are wrong. Just as perception is not impeached by the fact that we can’t see tardigrades or hear frequencies at 20 kHz, so too, the fact that you can never remember the name of your mail carrier, or that you ruminate on something your colleague did in a meeting five years ago, or that your memory of last night’s baseball game isn’t a play-by-play doesn’t make memory defective.
How did we get here? We made a mistake, but it wasn’t in holding Memory to a different standard than her sibling. It was in holding her to the same standard.
Memory and perception both represent the world to us as being or having been a certain way, a way that the world can fail to be or have been, to some degree, or altogether.
It makes sense to think of perceptual experiences in terms of accuracy: our perceptual experiences can be more or less precise, detailed, comprehensive, and true to the world before our eyes, ears, noses, and so on.
And we thought: well, memory is like perception, but not perception of the world in front of us. Memory perceives the world as it was, either by perceiving the past itself or by storing our past perceptions of the world.
So, we asked Memory the same questions we asked of Perception: Are memories accurate? How precise are they? How much detail do they capture or preserve? how comprehensive are they? Are they true to the world as it was? Are they high-fidelity? Memory does not fare well under this line of questioning.
Here are some of the ways we’ve tried to get Memory to answer to the responsibilities of Perception:1
I propose a new model. I propose to think about ways in which a memory can be a successful representation as a memory. This includes ways in which a representation can succeed in being a memory, and ways in which a representation that is a memory can succeed.
In other words, I propose to replace the notion of memory accuracy with the notion of representational success. More to come on this.
But there is one requirement in our list that seems to apply to memory uniquely, one that doesn’t fit well with perception: Bernecker’s authenticity.
Bernecker’s truth is an accuracy requirement: memories must be faithful to the past just as perceptions must be faithful to what is present. But authenticity doesn’t concern what we experienced in the past; it concerns the past experiences themselves.
Authenticity requires that we remember not only the past – and be faithful to it – but that we remember our experiences of the past as well, and that our memories be faithful to our experiences of the past.
Authenticity may be an outlier, but it descends as much from our original error as do accuracy and truth: the error of thinking of memory is a kind of perception, or a kind of doubling of perception. We perceive, and a memory is a perception of a perception that returns us to what we perceived, or to our past perception itself.
I propose we replace authenticity with the broader category of representational change.
Unlike authenticity, representational change is not a success condition, it is not an evaluative standard, it is evaluatively neutral. Observer memories are intended to be a classic illustration of why we need an authenticity requirement.
When we perceive, we do so from a perspective: typically, an egocentric perspective. When we remember, we also do so from a perspective: again, often an egocentric perspective. In observer memories, we remember not from the perspective from which we originally perceived (egocentric) but from the perspective of an observer. Anyone who has played video games understands these shifts in perspective.
On the authenticity model, observer memories can be true, or accurate – they can be faithful to past events – but they can’t be authentic. They’re not faithful to past experience.
In other words, authenticity treats observer memories as a kind of memory failure. In doing so, authenticity reveals its origins in the myth that memory perceives the world as it was, either by perceiving the past itself or by storing past perceptions.
On the authenticity model, an ideal successful memory reproduces a perceptual experience save but in some crucial temporal and spatial details: representing as there and then, what had been here and now.
But this notion of an ideal memory as a reproduction sets the highest possible standard for success: any change from an original perception is a failure. In other words, the authenticity model puts Memory in a classic double-bind (as so many standards do): for her to succeed, she must change, and if she changes, she fails.
Representational change is not the result of something going wrong in memory. It’s not even a lamentable side-effect of memory processes. Memory processes are processes of representational change. Memory makes information available by changing it, changes that allow us to use information for thought and action.
From working memory, to short and long-term memory, from semantic to episodic to autobiographical memory, the processes of encoding, storage, and retrieval make information available by changing it, changes that allow us to use information for thought and action.
Observer memories are only one example of representational change. Here are some others: newly formed memories become long-term memories through consolidation, and consolidation often involves integration, by which information is bound and chunked, combined, and altered, in relation to stored information, schemas, concepts, frames, and scripts. Consolidation and integration reactivate and replay memories, preparing them for use in thought and action.
Representational change includes semanticization, through which episodic memories change into semantic memories over time. It includes fading effects, as when the affective intensity of a memory lessens over time. And it includes spatial and temporal sequence changes in which an event is represented and re-represented in changing spatial or temporal contexts.
So much for representational change, but what about representational success?Here again I suspect that many of us (including myself) have thought about memory errors as memory versions perceptual errors.
Or we think of memory errors in terms set by Martin & Deutscher’s conceptual analysis, which gives rise to increasingly baroque possibilities such as relearning, veridical relearning, and veridical confabulation.
I propose to think about memory in terms of representational success and I propose pluralism about representational success.
There is no such thing as representational success as such, only success along dimensions specified by principles generated by the nature and function of the states, systems, processes, mechanisms, and dynamics under evaluation.
For example, the format of a memory might be iconic or discursive.2
The two formats obey different principles and thus succeed along different dimensions. Broadly speaking, iconic representations represent by way of isomorphic principles while discursive representations represent by way of compositional principles. Each set of principles generates dimensions of representational success distinctive of their format type.
We should expect representational success to have many dimensions: precision, detail, vivacity, fineness of grain, context, target, valence, arousal, familiarity, fittingness, and so on.
The list is open-ended; it’s intended to invite further work on the variety and complexity of successful and unsuccessful remembering in a way that is informed by, and suited to, the memory sciences. The list is also heterogenous: it is intended to include dimensions and degrees of success for states and processes that the memory sciences identify as potentially implicated in successful representation.
And thinking about memory in terms of representational success allows us to study actual memory phenomena from the wild rather than consider cut-and-paste possibilities from perception.
Theia, goddess of sight and vision looks after us well. By her, we respond to and engage with our present environment. She makes us differentially, often minutely, sensitive to features and changes central to our lives as practical agents. Psychologists put this less poetically by describing perception as stimulus dependent.
Because Perception enables us to engage causally with a complex, present environment in a rapid, often automated way, accuracy is an appropriate standard for her representational success.
But Memory is wild and heterogenous in a way that Perception is not. Perception has many modalities (sight, hearing, touch, etc.) but there are many kinds of memory.
And memory is not differentially sensitive to the present environment. It’s not stimulus dependent. Memory is environmentally scaffolded, and many memory cues are environmental cues. But Memory is not counter-factually coupled with the environment in a way that makes accuracy an appropriate standard for her representational success.3
So let Mnemosyne be Mnemosyne. Stop asking her to answer to what is properly asked of her sister Theia.
If we look at her, in a new, clear light, she may reveal fresh discoveries. She is, after all, the mother of the Muses.
References
Bernecker, S. 2010. Memory: A philosophical study. Oxford: Oxford University Press.
Bernecker, S. 2015. “Visual Memory and the Bounds of Authenticity”. In D. Moyal-Sharrock, V. Munz & A. Coliva (eds.), Mind, Language and Action: Proceedings of the 36th International Wittgenstein Symposium (pp. 445-464). Boston: De Gruyter.
Michaelian K. 2016. Confabulating, Misremembering, Relearning: The Simulation Theory of Memory and Unsuccessful Remembering. Frontiers in Psychology, 7: 1857.
Michaelian, K. 2020. Confabulating as Unreliable Imagining: In Defence of the Simulationist Account of Unsuccessful Remembering. Topoi, 39: 133-148.
Quilty-Dunn, J. 2016. Iconicity and the Format of Perception. Journal of Consciousness Studies, 23(3-4): 255-263.
Quilty-Dunn, J. 2019. Is Iconic Memory Iconic?Philosophy and Phenomenological Research, 101(3): 660-682.
Robins S. 2016. Misremembering. Philosophical Psychology, 29(3): 432-447.
Bernecker, 2010, 2015. Michaelian, 2016. See Michaelian 2020 for an updated theory. Robins, 2016.
See Qulty-Dunn, 2016, 2019.
I’m grateful to my colleague Casey O’Callaghan for these insights about perception.