The word forgetting refers to an outcome, not a particular process. The outcome is that we cannot remember something we once were able to remember. This can be temporarily so, or permanently. There has been a long-standing and yet unresolved debate whether forgetting reflects the actual loss of a memory, or impeded access to an existing memory. This is more of a principal problem, as Nietzsche pointed out remarking that “The existence of forgetting has never been proved: we only know that some things do not come to our mind when we want them to.”.

According to Nietzsche, it seems that the only theoretical position defendable is that forgetting reflects a problem to retrieve memories, and this view has also been a prominent position in memory research. But this view is not without problems, as it can never be falsified, or disproven. Imagine a thought experiment, in which you make sure that an organism acquires a memory. Then you apply some intervention that for unknown reasons leads to amnesia. Will it be possible to devise a test so that there would be one possible outcome of it that shows that the reason for amnesia is not a retrieval problem? Let's assume that whatever you try, the original memory does not return. Can you now reject the explanation that impaired retrieval is to blame? Because one could always argue that you simply have not yet found the correct method to get the memory back, it is you can never conclude that the memory is actually still available and simply at the moment inaccessible. A theory that cannot be disproven is therefore not accessible to fair empirical tests, and it is always true. The theory resembles then more a religion than a scientific model.

But what about the other option, that you forget because a memory has been destroyed? This positions has similar problems. Let's get back to our thought experiment, but this time, our intervention brings back the memory, reversing the amnesia. Does this outcome allow us to reject the explanation that memory was genuinely lost because why would it return otherwise? This is also not straightforward because one could argue that the intervention that brought back the memory provided enough information about the memory itself so that it could be reconstructed based on whatever remained of it.

None of these positions seem to offer theoretical comfort, and some researchers have suggested compromises. For example, Bjork & Bjork suggested that forgetting reflects both a storage and a retrieval failure. In their model, what we call a memory has two basic components, a retrieval cue linking to the actual memory contents. While the memory contents is never lost, the retrieval cue may degrade. We have similarly argued that every explanation of forgetting as compromised retrieval must first explain how retrieval is organized in the brain. At the most basic level, retrieval will reflect some form of change to the connectivity of the brain, and these changes, we know, are modifications to the structure of neurons that are actively maintained. In order to impair retrieval, therefore, these modifications need to be impacted, and, on the most fundamental level, this amounts to altering connections, perhaps erasing them or weakening them, which are all storage phenomena. Thus, at the basic level, impairing retrieval means erasing modifications that enable retrieval.

It may therefore be a matter of perspective whether forgetting reflects a retrieval or a storage problem. From a more basic, cellular view, forgetting goes along with loss of connectivity among neurons. From a higher level, like a brain systems view, these basic changes impair the memory function of retrieval. In the end, both positions may be right at the same time.

As all compromises, this explanation is a bit unsatisfying because it assumes that there is no such thing as the loss of established memory contents – memories are forever. Humans may prefer this position instinctively because we distill our identity from our autobiographical memories, such that a loss of them directly threatens our sense of self. The slow and painful dissolution of a person along with these memories in neurodegenerative conditions such as Alzheimer's disease is a reminder how precious our knowledge of our own past is for us. But is absolute permanence a realistic assumption for long-term memories?

We prefer an alternative, pluralistic view, in which different processes lead to forgetting so that sometimes forgetting reflects genuine memory loss, and sometimes a retrieval problem. This is for three main reasons. First, while we retain a lot of the long-term memories formed during a day, most of these memories are mundane and essentially superfluous. Keeping them forever will eventually produce an existence that Funes had to endure, and it is unlikely that our inability to recall everything that happened last week, means always that we cannot retrieve these events. Second, we know from decades of research that our life memories are elaborate reconstructions, not simply read-outs from a faithful record of the past, suggesting that what we remember is a plausible guess based on what memories of the past are still available or accessible at that point. Third, we know that the neurobiological structures and processes associated with memory are highly dynamic and are constantly in flux. If we believe that there is a link between these neurobiological factors and memory, it is difficult to assume that all memories remain as they were, and forgetting then reflects to a certain degree loss of information.

Instead of trying to decide whether all forgetting can be explained in terms of memory availability or accessibility, we believe it may be more informative to focus on the actual neurobiological processes in place that cause forgetting, in a theoretical neutral way.