The Complexity of Memory

How recall produces the sense of an abiding self

Memory, 2018

In ordinary usage, memory is treated as a faculty that stores the past and returns it on demand. But seeing the complexity of remembering defeats that picture. Memory operates in at least three modes, and they do not share a common guarantee. One is archival memory. A trace is laid down and later retrieved with fidelity, altered little by the retrieval. A second mode is reconstructive memory. A trace is partial, and retrieval rebuilds it from what remains plus the conditions present at the moment of recall, so that the act of remembering modifies what is remembered. A third is confabulation, in which an apparent memory is generated from no corresponding event, produced in full sincerity, and presented to the one recalling as though it referred to something real. The confabulating system is not lying. That would require knowing the truth and misreporting it. In this mode, there is no awareness of the fabrication as fabrication. Florid confabulation belongs mostly to specific pathology, but the milder forms do not. An ordinary person, memory essentially intact, misremembers an event that did not occur, imports detail that was never present, or takes a suggested scene for a lived one, and holds the result with the same conviction as a genuine recollection.

The modes do not announce themselves. A memory retrieved intact and a wholly fabricated one can feel the same, the same sense of pastness, the same sense of ownership, inspiring the same confidence.

That is the first thing the complexity tells us, and it is not a minor matter. It removes introspection as a criterion. The system that recalls cannot, from within, distinguish a fact-based memory from an invented one, because both arrive bearing the same marks. Ordinary usage treats confidence in a memory as a rough index of accuracy. It is not one. Confabulated cases are often held with the greatest confidence and the richest detail.

If the felt quality cannot distinguish a real memory from a fabricated one, then the felt quality is not reporting on the memory’s origin. That is a property added at the point of retrieval, a tag that reads “this belongs to my past,” applied regardless of whether an actual past event stands behind it. The tag is produced by the same operation whether the content is authentic or confabulated. This is not philosophical skepticism, but only a mechanical observation. The sense of remembering is generated, and its generation is separable from what it purports to certify.

From here, the consequences for the self follow directly, and this is why the question of memory involves more than just memory. The continuity of a self across time is ordinarily thought to be underwritten by memory. In that view, I am the same one who acted yesterday because I remember having acted, and the remembering connects the two moments into a single subject. But if memory includes a confabulatory mode that can fabricate that sense of connection without the events to support it, then the same felt continuity of the self can be produced with no continuous subject behind it. The sense of connection is supplied in the moment of recall. It does not require that anything persisted through the interval to be connected.

This is where the systems framing earns its place. Consider the self not as an entity that endures and consults its memories, but as a pattern reconstituted at each moment from stored and generated material. Archival retrieval consults a trace. Reconstructive retrieval assembles one from fragments. Confabulation produces one from nothing. None of the three requires a subject who persisted through the interval to receive it. The operation runs at the moment of access. Fidelity varies, and so does the point at which distortion enters, and whether a corresponding event occurred at all. What does not vary is that a sense of continuity is produced at access, and is not evidence of a self carried across the gap.

There is a temptation, at this point, to conclude that the self is therefore an illusion, that memory deceives, and there is no one there. The evidence does not warrant that conclusion, which overreaches in the same way the ordinary picture did, only in the opposite direction. The complexity of memory suggests something narrower and more exact. The self-model is real. It is a pattern the system produces. But this does not underwrite a continuous subject behind the pattern, because the very faculty that seemed to underwrite continuity, memory, turns out to manufacture the sense of continuity independently of whether the continuity actually obtains.

So the complexity of memory means that a particular inference is unwarranted—the inference from the felt continuity of remembering to the existence of a subject who remembers. The feeling of being a unitary “myself” is produced when remembering. That production takes place regardless of the facticity of what the memory seems to report. This finding does not empty the day of its remembered content or the person of a functioning past. It removes only one thing, the presumption that the sense of being a continuous someone indicates that there is one. Remembering happens. The one who is supposed to own it, persisting behind each recall to receive it, is the part that retrieval bestows rather than finds.