A Memory Menagerie

What we call “memory” isn’t just one process – or even one type of process. In fact, neuropsychologists classify memories using a system that can be a little bewildering – which is why I’m going to do my best to clear up some distinctions.

Buckle up - this is gonna be a blog post to remember!

So, without further ado, let’s take a tour of this memory zoo.

Part I. Memory time ranges
Scientists today usually divide memory into three basic ranges of time: working memory, short-term memory, and long-term memory. However, these distinctions are hotly debated, as I’ll explain below.

1. Working memory
Strange as it might seem, we need a certain kind of memory to assemble a coherent sense of what we think of as “the present.” The exact nature of working memory is still a argued quite a bit, but it’s generally agreed that our overall sense of “the present” refreshes about every 20 to 30 seconds, and contains a very limited number of “items” on which our attention can simultaneously focus. These items can be auditory/phonological, visual/spatial, or conceptual/abstract – and they can be triggered by external stimuli or synthesized from stored data.

Working memory is crucial for tasks like math and reading, which require us to “hold” a representation of an item in our focus of attention for several seconds or more. In short, working memory is roughly synonymous with “what you’re thinking about” in the current moment.

2. Instantaneous memory
Unlike working memory, which by its nature forms a part of our conscious attention, immediate (or “instantaneous”) memories are sensory memories that form almost instantly, fade in a few hundred milliseconds, and may or may not enter our conscious awareness. In the visual domain, this is sometimes known as “iconic memory;” while it’s sometimes called “echoic memory” in the world of sound.

Because instantaneous memory allows informational patterns to persist in our brains after an actual stimulus is no longer present, it helps our brains assemble a cohesive flow of subjective experience. Thus, it allows us to detect changes in visual or auditory information, which lets us perceive movement in animation and film, and detect musical harmonies in sets of tones.

A brief aside: Priming
A topic that often comes up in connection with working memory and instantaneous memory is that of priming – when exposure to a certain stimulus influences our perceptions of later stimuli. One simple example of priming is that when people are shown a word – say, “table” – then asked to complete a word starting in “tab__,” they’re more likely to answer “table” than people who weren’t primed with that word.

A more intriguing example is that when people are shown a set of dots moving in a clockwise direction, and then look at a set of dots moving in a counterclockwise direction, they’re likely to report that the second set of dots was moving more dramatically away from the direction of the first set that it actually way.

The general idea here is that the present moment isn’t discrete from the past – our recent events and ideas leave impressions on our senses, and those impressions are constantly influencing what we perceive at any given instant.

3. Short-term memory
It’s important to clear up a bit of confusion right from the start – many laypeople (and even some scientists) use the terms “working memory” and “short-term memory” interchangeably – today’s neuroscientists and psychologists typically classify working memory as just one subset or type of short-term memory.

Keep in mind that the term “working memory” refers to attention processes – such as reading and math – that temporarily store and manipulate the information that composes our sense of “the present moment.” Though some scientific papers do refer to working memory as short-term memory, the phrase “short-term memory” is more widely used to describe any information that’s kept “on hand” for instant recall – such as a sentence you read a minute ago, or what room you just left.

A lot of debate surrounds the exact capacity of short-term memory, but the range of 7±2 items has been highly influential. More recent research points to a number around 3 or 4. The rate at which short-term memories “decay” is also debated, but it’s widely agreed that without repetition, most of these memories are eventually lost for good.

4. Long-term memory
A lot of scientists think these memories are encoded (written into long-term storage) from short-term ones during sleep. They often stick around for a person’s entire life – at the very least, they’re much harder to forget than short-term or working memories are. They can also be harder to recall, but once they’re loaded into working memory, it’s easier to recall them in the future. One interesting feature is that the mere act of recalling a long-term memory seems to change it.

Part II: Two memory models
Before we go any further, there’s a very important point to make: the exact boundary between short-term and long-term memory is (you guessed it) the cause of a lot of debate. In fact, not all scientists agree that there’s a clear distinction between the three memory systems at all. The two main competing theories about this distinction are known as the dual-store and single-store memory models:

1. The Dual-store memory model
This is the more conventional one – it draws a distinction between short-term and long-term memories (without making any particular statement about the distinction between working memory and short-term memory). One argument in favor of the dual-store model is that long-term memory’s capacity seems to be much larger than that of the short-term and working memory systems.

Objections against the dual-store model mainly center around two arguments:
a) even if experimental subjects are distracted from recalling a recently-performed task, they’ll still perceive the task’s various steps as “recent” and “contiguous” with one another;
b) the length of time an item spends in short-term memory isn’t a direct predictor of its strength in long-term memory.

2. The Single-store memory model
This one posits that there’s only one type of memory, in which context provides the sense of recency. In the single-store model, short-term memory and long-term memory (and, presumably, working memory) are all just different ways of perceiving memories encoded in a single system.

Objections against the single-store model point out that although this theory helps explain some features of memory – such as the fact that recall fades gradually until about 10 minutes have passed, and then fades much more gradually over the next few months – it still doesn’t provide a very clear explanation for some brain phenomena, such as why we apparently need sleep to organize our short-term memories into long-term ones, and why more permanent memories appear to be encoded in separate synaptic maps from short-term ones.

These arguments help to demonstrate that even if short-term and long-term memory are different systems, we may not be defining their boundaries correctly. And so, the discussion continues.

Part III. Memory data types
Within the ranges listed above, many scientists also divide memories along another set of lines. They mainly have to do with what type of information the memories primarily focus on:

1. Explicit memory
In general, it’s easiest to think of explicit memories as “memories you’re conscious of having.” Facts you memorize, events you recall, and numbers you manipulate to solve a math problem are all explicit memories.

Some scientists classify explicit memories into several sub-types:

a) Visuo-spatial memoriesworking memory items dealing with visual images, real or imagined; they seem to refresh about every 10 seconds
b) Phonological memoriesworking memory items dealing with sound and speech, real or imagined; they seem to refresh on a loop that’s about 3 to 4 seconds long
c) Declarative memories – memories for specific facts and events. In the dual-store model, declarative memories are typically considered part of the long-term memory system, and they fall into two sub-sub-categories (yeah, I know, I know…)
     i) Semantic memories – memories of facts/understandings
     ii) Episodic memories – memories of occurrences/events

2. Implicit memory
It’s easiest to think of these memories as “memories for how to do things,” or “how things happened.” They’re stored in a very different way from explicit ones – instead of being consciously learned and recalled, they’re stored through experience and/or practice, and come into play when we involuntarily remember how to ride a bike, how to swim, how falling in love feels, or how embarrassing it was to spill a drink. Instead of focusing on specific facts, they’re focused on associations and environmental stimuli. In the dual-store model, they’re considered part of the long-term memory system.

Studies of implicit memory are mainly concerned with procedural memories – memories for how to perform a task. A procedural memory could be something as simple as how to tie a shoe, or as complex as how to play a sonata. It might even be a cognitive skill. The unifying characteristic is that it’s learned through practice.

The wild card: Emotional memory
These are memories about how a past event felt, or about emotional associations with an explicit memory. Scientists haven’t resolved the question of whether these memories are actually part of the implicit memory system, or represent a system of their own.

Emotional memories are an unusual breed – unlike procedural memories, they seem to form almost instantly; but unlike explicit memories, recalling or suppressing them isn’t under our conscious control. They’re also very hard (but not impossible) to forget. Some scientists have proposed that emotional memories represent an entirely separate – and more primitive – memory system: one that involves the amygdala. The amygdala also helps strengthen explicit and procedural memories, though, so its exact overall role in memory remains unsure.

Part IV: Memory time directions
There’s one last important memory distinction that needs to be mentioned. Like the ones described in Part I above, they’re also related to time – but instead of pertaining to a time range, they’re related to time’s direction.

1. Retrospective memory
This just means a memory for any event, fact, or procedure encountered in the past, after some delay. It can be explicit or implicit, episodic or semantic, long-term or short-term, or even emotional. This term doesn’t apply to working memory, though, because it implies an interruption between the experience that triggered the memory and the act of recalling it.

2. Prospective memory
These are memories involving the timing of events and actions that haven’t happened yet, such as an appointment that’s coming up, a chore that needs to be performed, or a shower you’re about to take. These memories are sometimes called memories for the future (which I think is insanely confusing), but the term basically just refers to our ability to think and plan about events we haven’t actually experienced.

Well, there you have it. I hope I’ve helped make things clearer instead of thoroughly confusing you. If you’ve got any questions, or would like me to tidy any of this up, feel free to drop me a line and I’ll do what I can. But I hope this has piqued your interest in memory research, and shown you how far the field still has to go before anyone agrees on…well, on much of anything.

In all honesty, it seems silly to me that the different types of memory are so hard to remember – if that’s not evidence that the universe is absurd, I don’t know what is!

16 thoughts on “A Memory Menagerie

  1. While I mostly agree with the temporal classification, I would slightly disagree with you on a few definitions.

    First, I would categorize working memory as a form of short-term memory (it is short term after all) and I would also say that short-term memories can also tap into the long-term memory systems and mechanisms that do not participate in working memories. In fact I would argue that non-WM STM are “made of the same stuff” as LTM. There is some sort of continuity in these definitions which makes them imperfect and difficult to juggle with. The way I understand the division of explicit memories is by thinking about patient HM (http://en.wikipedia.org/wiki/Patient_HM). I would say that in terms of explicit memories, he had an intact ability to form working memories whereas his ability to generate other types of explicit memories ( LTM and non-WM STM) was null. [That being said it seems that even temporal lobe patients can develop some form of long-term declarative memory (see here: http://www.ncbi.nlm.nih.gov/pubmed/16049487).]

    I think, the point of contention about different types of explicit memories when focusing on their duration comes from the fact that their temporal realm is stretchable when tested experimentally. HM could remember things for 15 min for instance or more, provided he wasn’t distracted.

    I have to say I do like your description of STM as “at hand”.

    Second, regarding the number of memory stores, I think there is now very strong evidence that more than 2 types of stores exist and this is supported by the legion of lesion studies that provide evidence for the physical location of these many stores.

    Third, I completely agree with your definition of working memory. I would perhaps add the term online to describe it, and resistant to interference/distraction. An NIMH workshop is currently discussing the concept of working memory here: http://www.nimh.nih.gov/research-funding/rdoc/working-memory-workshop-proceedings.shtml and it is very informative.

    And finally fourth,it seems that at least one species is missing from your menagerie: immediate or instantaneous memory as defined by Sperling, don’t you agree? This is an extremely brief sensory memory, fading in less than a few hundred milliseconds called sometimes iconic or echoic memory, depending on its modality (see for instance here: http://en.wikipedia.org/wiki/Iconic_memory). Such memories will only enter the true, flexible stores with Attention.
    Perhaps also, a few words on priming and where it fits in your classification could be a good idea?

    This exercise is a good idea anyway, thank you!

    1. First off, thank you so much for providing such a detailed commentary (complete with citations!). Let me address your points in order.

      I struggled quite a bit with the decision about whether to classify working memory under short-term memory, or under its own heading. As you say, distinctions between the time ranges tend to blur, and seem to indicate that the semantic categories we’re using may not be entirely appropriate at all – for instance, like you mentioned, short-term recall can also utilize data from long-term storage, and there’s definitely data to support the idea that short-term and long-term memories are components of a single “system.” In regard to the patient HM, I’m not entirely clear on whether the distinction was between working and short-term memories, or between declarative and procedural ones. Would you say it’s possible for working memory to include declarative/explicit content?

      You raise a very good point about the durations of explicit memories. Though plenty of studies measure recall duration differences with vs. without distractions, I’m not sure how applicable this data is to non-lab environments, when many types of declarative and procedural recall tasks are often being processed in quick succession. This may be a contributing reason why the single-store vs. dual-store debate has been so prolonged, and many of the participants have seemed to be “arguing past” each other.

      Which brings me to your second main point – though I’ve read some studies that correlate different functional connectivity patterns with different types of storage and recall, I honestly wasn’t aware of lesion studies that pinpointed different storage types to specific locations. Would you mind linking to a few? I’d love to read them.

      Thanks for the link on your third point – I’ll definitely dive into that workshop write-up as soon as I get the chance.

      And as for your fourth point – yes; I can’t believe I didn’t at least point out the distinction between working and iconic memory. I’ll work on a draft with that revision, as well as your ideas for the working memory description.

      I’m glad you enjoyed the post; as I hope you can tell, you’ve been a huge help. It’s always a pleasure to talk with another lover of neuroscience (even if it’s just in the comments section!). I’ve added your blog to my “links” wall. Thanks for stopping by!

  2. I would say that working memory is entirely explicit since it is by definition conscious.

    Regarding the number of stores and their localization, as early as Shallice and Warrington and their patient KF, it has been thought that multiple stores exist. KF only had a deficit in the retention of phonological information (inner voice) not visuo-spatial information (inner eye). This patient could identify semantic stimuli but could not store them, which was interpreted as evidence for the existence of a phonological store and by extension of other stores for other types of information. In addition, it is well accepted that visual information when it enters the cortex gets split into object and spatial location streams, the “what” and the “where” and one may agree that these different cortical modules enable the storage (at least for the short-term) of distinct types of information.
    In 2003, Allan Baddeley reviewed some of the evidence, based on lesion and functional imaging studies, that tends to demonstrate the existence of different stores (http://www.ncbi.nlm.nih.gov/pubmed/14523382). I am sure that this anatomical description has progressed since this original review.

    What I am describing above is sort of restricted to working memory. I don’t know enough to say much about LTM. I think that one of the consensus is that explicit memory systems are highly distributed in the brain but it does not rule out the existence of multiple discrete storage places. That being said, I would just emphasize that long-term storage is initially thought to occur via and within the hippocampus (temporal lobe) at least temporarily. Storage is then progressively transfered to the neocortex. Although I don’t have any example in mind, there are plenty of animal studies showing such transfer, and you could consider HM a clinical illustration of this to some extent. This observation tends to support the existence of multiple long-term storage systems.

    Maybe someone else would like to add something to these few thoughts and correct me if I am wrong.

    Anyway, thanks for the link. I wish I had more time to write about Neuroscience and I have to say that I am quite impressed at how prolific you are in that respect. Keep it going!

  3. Pingback: The Connectome

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s