The traditional model of memory has guided research for decades, but may now be due for a change.

Multiple Memory Systems theory proposes that the brain stores different types of LTM in separate modules.

New research shows that activation of brain regions for different types of LTM information overlap.

Building a model of the phenomenon you’re studying can help you understand how that thing works. Roadmaps are models of the highways and roads between here and there. We use them to predict how we might get to our desired destination even if we encounter roadblocks, weather, and bad traffic. Scientists use models a lot in their work. Good models should simplify what the scientist is studying, making it easier to describe, explain what you’re studying, and to predict how it will function in a new situation. Scientists test their models to see if they’re accurate. If they encounter a problem with the model, then the model is changed to accommodate the problem, and our understanding moves forward.

Human memory has been the subject of study in psychology almost since the beginning of the discipline in 1879. Over the years, several models of memory have been proposed. One particularly influential model of memory is called the Atkinson-Shiffrin model (Atkinson & Shiffrin, 1968), aimed at describing how memory is structured and how it works. This model breaks memory up into three essential components.

First, information enters Sensory Memory (lasting only a few seconds before being passed along into the system). From there, information is passed to Short Term Memory (STM) which has a limited capacity of 7 to 9 items or chunks of information by itself, unless that information is actively rehearsed. STM holds on to information, as the name suggests, for a short period of time before it is passed on to the final stage, called Long Term Memory (LTM), a semi-permanent storehouse for all memory that has made it through the system.

In the 1970’s and 80’s, two general types of LTM were added. “Implicit” (aka non-declarative) memories, like memory for how to ride a bicycle, or how to type without looking at the keyboard, were unconscious and unintentionally formed, while “explicit” (aka declarative) memories were intentionally formed memories for facts, words, and past events. We can put declarative memories into words. We can perform non-declarative memories but probably have difficulty putting them into words. The type of information contained within LTM was also described. For example, “episodic” memory for previous experiences and the context and emotions associated with them, and “semantic” memory for words, numbers, concepts and ideas, were added.

As neuroscience came into prominence in psychology in the latter part of the 1980’s, specific regions of the brain were incorporated into the model. The most influential of these neural models is Multiple Memory Systems (MMS). Based originally on the observation that damage to specific brain regions led to specific problems with memory, researchers proposed that “the brain stores information based on the independent and parallel activity of a number of modules, each with distinct properties, dynamics, and neural basis” (Ferbinteanu, 2019, page 61).

So, each type of LTM had its own neural system, independent of the others, yet operating in parallel with the others. Suppose you were watching someone ride a unicycle, a previously unknown experience. All that new information comes into the memory modules in the brain, with information about how one might balance on a unicycle flowing into non-declarative, procedural modules, while information about your own personal interactions with bicycles and unicycles and what the word means, etc., entering the declarative memory modules, all at the same time. These memory modules are separated within the brain, each working on a different aspect of the incoming information.

The MMS theory or model of memory has effectively guided research on memory since its inception. However, that may all be on the cusp of change with new research questioning how actually independent and how parallel these modules are.

Tibon, Greve, Humphreys, Quent and Henson (2025) examined whole brain fMRI images taken from participants asked to use their semantic and episodic memories on a relatively simple task. On two separate days, participants were asked to retrieve semantic and episodic memories for logos (pictures) and brand names (words). Some of these stimuli were quite familiar and real world (like the brand name “Google” or the Nike swoosh logo), requiring participants to access previously stored semantic memory. Others were new, previously unknown brands and logos taught to the participants prior to the memory test, requiring them to access their episodic memory.

They hypothesized that the regions of the brain used for episodic memory would be distinct from those used in semantic memory. However, that is not what they saw. They found that the regions of the brain activated by episodic and semantic recall overlapped; there was no difference in activation of the brain regions known to be involved in these two different types of LTM. They concluded that “our study shows very little evidence for distinct processing of and no evidence for (or evidence against) distinct representation of semantic and episodic memories” (page 8).

This information is particularly important for the treatment of memory disorders like Alzheimer’s disease. Better understanding of the neural basis of memory, and that it is likely a whole brain event, might help researchers create better interventions and treatment for the kinds of memory loss so devastating in dementias.

Atkinson, R.C.; Shiffrin, R.M. (1968). "Chapter: Human memory: A proposed system and its control processes". In Spence, K.W.; Spence, J.T. (eds.). The psychology of learning and motivation. Vol. 2. New York: Academic Press. pp. 89–195.

Ferbinteanu, J. (2019). Memory systems 2018—toward a new paradigm. Neurobiology of Learning and Memory. 157, 61–78. doi:10.1016/j.nlm.2018.11.005.

Tibon, R., Greve, A., Humphreys, G., Quent, J.A., and Henson, R. (2025). Neural activations and representations during episodic versus semantic memory retrieval. Nature Human Behavior, https://doi.org/10.1038/s41562-025-02390-4.

© Psychology Today