| by Alexandra Vilceanu
Every passing moment, this one included, is nothing more than a fleeting instance of reality. That is not, however, the way we experience it. What we know best is progression, gradients, a flow. We stitch together this moment with the one before it and the one before that until we create an experience. We then preserve this experience for what may be a very long while or even a lifetime. We may return to it as we wish, compare it with other experiences, assess it, sometimes even dream it. Some argue that this is, ultimately, all we are: the resulting product of a given number of experiences, conjured by an arbitrary set of circumstances.
None of this would be possible without the ability to convert information from the environment into something we can store for long-term use. Much like a computer, we need to repurpose a picture, a sound, or a text into some format our brain can not only understand but preserve and later on re-create.
In the history of memory formation research, two models emerge as prevalent.
First along came the Multi-Store Model, proposed by R.C. Atkinson and R.M. Shiffrin in 19681. In this instance, memory is created by passing information from one store to another. The stores include a sensory register, short-term memory, and, ultimately, long-term memory. According to it, all environmental stimuli pass into the sensory store, regardless of whether we pay attention to them or not. Paying attention to a particular stimulus is what ensures its passage into the short-term store. Transitioning from the short-term store to the long-term store is supposed only to be acquired through repetition.1,2 Key differences between the three leading stores are highlighted below:
|Encoding||No attention required||Attention required||Repetition required|
|Maintenance||Not possible||Constant repetition||Repetition|
|Capacity||Large||Small||No known limit|
|Duration||1-2 seconds||Up to 30 seconds||Minutes to years|
|Retrieval||Readout||Probably automatic||Retrieval cues prompting a search for information|
Key differences between stores of the Multi Store Model – adapted from Craik, F. I. M. & Lockhart, R. S. Levels of processing: A framework for memory research. J. Verbal Learning Verbal Behav. 11, 671–684 (1972).
We can distinguish four main steps involved in turning any form of external stimulation into a memory: Encoding – Storage – Consolidation – Retrieval.3 The Multi-Store Model provides a cohesive framework for at least the first three.
Encoding is the step that prepares information for entering the memory. It is that “paying attention” that selects some of the many sensory stimulations we receive over a day to turn them into valuable experiences. The first step towards turning just another passing image or sound into data that can be retrieved even once the stimulation is gone.3,4
Storage involves the structural framework – the mechanisms and brain regions involved in actually maintaining the information. The stores are worthy predecessors of our current understanding of Short and Long-term Memory.
Short-term memory is limited, both in capacity and duration. In one of the most highly-cited papers in psychology - The magical number seven, plus or minus two5 – George Miller argues that the average capacity of short-term memory is somewhere between 5 and 9 “chunks” of information at a time, regardless of their nature. Regarding duration, something persists within the Short term memory for 15-30 seconds, although this time can be prolonged by continually keeping our attention directed at it.1,6 Consolidation is what ensures the transition towards the Long-Term store and this, in Atkinson’s view, is acquired through repetition.1
Long-Term memory, once reached, is virtually limitless – there is no known maximum amount of information we can hold, and we hang on to specific pieces of information for years.1 Our brains don’t run out of space like a hard drive would because memories don’t take up space as files would. Memory does not occupy a specific location or set of neurons, but rather a pattern, a specific neural circuit, an engram that needs to be fired for the experience to be re-lived.7 Since the brain is estimated to hold at minimum 100 trillion single synapses, the possibilities are virtually endless.8
To complete the structural framework provided by the stores, another concept was put in place – control processes. Atkinson suggests that:
“A computer analogy might help illustrate the distinction between memory structure and control processes. If the memory system is viewed as a computer under the direction of a programmer at a remote console, then both the computer hardware and those programs built into the system that cannot be modified by the programmer are analogous to our structural features; those programs and instruction sequences which the programmer can write at his console and which determine the operation of the computer, are analogous to our control processes. In the sense that the computer’s method of processing a given batch of data depends on the operating program, so the way a stimulus input is processed depends on the particular control processes the subject brings into play. The structural components include the basic memory stores; examples of control processes are coding procedures, rehearsal operations, and search strategies.”1
The Multi-Store Model soon came to be heavily debated and ultimately countered (or completed) by the Levels of Processing Model, a paradigm proposed by Fergus I. M. Craik and Robert S. Lockhart in 1972.6 According to it, repetition is not the only way to ensure persistence. New information is best remembered when linked adequately to existing knowledge. In other words, a memory is nothing more than “a by-product of perceptual analysis”6. When the information is linked to essential, well-established existing knowledge and the subject is highly motivated to memorize it; this process can prove particularly efficient – a so-called deep encoding, obtained through profound meditation upon this new data.3 This is set to contrast the shallow processing of information, based on how a particular text sounds, for example, which is said only to produce short-lasting memories.6,9
Craik and Tulving set out to prove this matter three years later. In an experimental set-up, participants were presented with 60 words which they were supposed to process in three different manners: see if the word has capital letters (visual processing), evaluate whether the word rhymes with anything (phonemic processing) and, ultimately, try to think of what it means and fit it in a sentence (semantic processing). They then had to pick the original words from a list of 180. As we now guess intuitively, participants recalled more words that were semantically processed, meaning those that were elaborated and deeply encoded.9,10
This is, essentially, the foundation for that ‘deep understanding’ and ‘focus’ we now encourage during learning, rather than constant repetition.
Remarkable memories, for whom the system we described seems to function without fail, do exist. People with hyperthymesia, around 60 documented cases around the world, recall an extraordinary amount of information, mainly autobiographical.11,12 Some can pinpoint events that took place in roughly any day of the past. In a documentary called Boy Who Can’t Forget, then-student Aurelien tries to describe his unusual recall capacity:
“It’s odd to communicate that I can just remember these sorts of things without even trying, without them having any importance. I just remember them, and I think that would be alien to a lot of people. […] It’s complicated to put across how you can remember something like that even though it means nothing. People always think you remember important things, but I think it’s an example where I remember things which are completely unimportant and have absolutely no value at all. I just happen to remember them.”
The very fact that he and others like him seem to make none of the efforts we previously described to preserve information may suggest that his impressive ability does not, in fact, only reside in memorizing. He is impressively good at retrieving memories. We all know what we had for breakfast today, yesterday and the day before that. As weeks and years go by, however, these memories become nearly impossible to retrieve, just another breakfast among tens of thousands, in what we account for as memory decay.
The retrieval of memories may well be the most delicate of steps. It involves bringing back into consciousness different pieces of information scattered across the brain, to make a cohesive experience that emulates the original one.
There are several ways in which we retrieve memories. Perhaps the easiest one is recognition – associating an experience with a previous similar encounter, such as recognizing faces or solving multiple-choice questions. Recalling something is slightly more complicated, as we remember something which currently has no physical correspondent. This relies on our ability to uncover specific information from memory.
Fascinatingly enough, it is thought that we remember either through direct retrieval (a specific cue immediately fires the required network) or hierarchical inference (a cue is immediately linked to the appropriate class of information, which may then be scanned), more efficiently than the sequential scans computers would perform. Alternatively, there is a two-stage theory of memory stating that, given a cue, we retrieve specific candidate memories, then choose the one we are most familiar with. Several different strategies are, in fact, at play here – including comparison, inference, supposition – until we achieve a memory that is not only similar to what we first saw, but it is also coherent, within itself, and with other associated memories. It matters, sometimes more than it is technically accurate, that it makes sense to us.3,13
During recall, the brain fires in a specific pattern, close to that originally generated as a result of the real event. Retrieving memories is, in essence, a constructive process – much like a perception, except one that is built within the brain. This, in turn, makes it susceptible to errors, so that the product is not entirely identical to the original. We are aware it is a memory and not a perception.3,13
In his famous experiment, “War of the Ghosts,” Frederic Bartlett set out to prove the constructive nature of memory retrieval. He assigned his participants to read a Canadian Indian Folklore titled “War of the Ghosts”, and then re-tell it at longer and longer intervals of time. As more time passed, participants progressively forgot information. Their stories were shorter, condensed, and where their memory failed them, they tended to fill out the gaps with elements that were particular to each individual and their cultural background. Canoes were, for example, remembered by the English participants as boats.3,14
We may not all remember what we had for breakfast on February 29th, 2004. Nevertheless, the very fact that we can remember anything is still a remarkable feat, highly responsible for our identities and behaviors. All that is left to do now is make good use of it, be it through deeply learning new information, memory tricks, memory training, or whatever other methods we see fit.