Learning in the brain
Introducing long-term memory and working memory:
o Key features and limitations
o How do they interact during learning to build (and then use) the knowledge base.
o Guidelines to aligning teaching and learning strategies with the learning stages.
Basic brain function
The function of the brain, in very general terms, is to receive signals from the environment, process them and respond effectively to support the well-being and survival of the organism. The information about what may be effective in a given situation is retrieved from the long-term memory system that stores knowledge gained from past experiences.
How is the knowledge stored in long-term memory?
The network model
It is widely accepted on the basis of scientific findings that knowledge is stored in neuronal patterns. Neurons are the basic functional unit of the brain, they communicate by sending and receiving electrochemical signals. By activating each other they form pathways and networks of synchronous activity. If new connections are formed to create new networks, they can be potentially reactivated in the future. Our knowledge is stored in specific patterns of neuronal activity - as represented in this highly simplified model of nodes and connectors.
The dynamic nature of the network
The specific patterns are initially created during learning, then potentially stabilized and stored in long-term memory. In the future, they may be reactivated upon recollection or retrieval. To illustrate using the model: let's say that the concept "Tiger" is represented as a small network (turquoise rectangle in the figure). This network is associated through connections to other networks (representing for example “Zoo”, "Predator", "Jungle", “Mammal”, etc.) that together encompass what we know about tigers. To execute an effective response, the relevant knowledge should be accessed and retrieved, combined with stimuli coming from the environment, and then manipulated and connected. The mental function that maintains this sequence of actions in "real-time is called working memory.
Working memory is the cognitive function that enables conscious mental processing or real-time thinking: combining incoming information with retrieved knowledge from long-term memory, manipulating them to create an output (that may or may not be associated with a response).
The most notable feature of this function is its limited capacity: it is possible to handle and manipulate only a small number of items simultaneously. Many are familiar with the 7±2 items limit (i.e. the normal range of items held in working memory is 5 to 9 items), however, the capacity may be even lower and closer to just 4 isolated and unfamiliar items. The exact number is not so important, but the notion that this function is highly limited is a major factor in the basis of our cognitive abilities.
Working memory - key features:
1. It is the "live" thinking function: connecting incoming new information with prior knowledge, and manipulating them.
2. It has a limited capacity.
3. Overload leads to information loss – either incoming information will not be processed, or an “in-process” item will be dropped for a new one.
4. Processing in working memory is essential for long-term storage: it is the information’s “entry ticket” to the long-term memory storage.
This last point emphasizes a crucial part in the interaction between working memory and long-term memory, Daniel Willingham depicted it in the famous quote: “Memory is the residue of the thought”1. In other words, long-term representations of knowledge (=memory) may be formed only as a consequence of manipulation in working memory (=thought).
So processing in working memory is essential for long-term storage, but do we remember everything that is manipulated in working memory?
The answer is obviously no, not in the long term. So, what else is important to consider?
What kind of processing in working memory supports storage in long-term memory?
if "memory is the residue of the thought", then the 'thought' should process the information in a way that can be "read" by the memory store.
Think of working memory as the reception counter to a huge archive. The incoming information must be labeled and stored according to the standard system of organization, otherwise, it cannot be stored systematically and it would never be found.
It follows that it would be useful to understand the principle of long-term memory organization in order to process the information in a similar way, so:
How is knowledge organized in the long-term memory store?
According to "Levels of Processing Model of Memory"2, when information is processed deeply and meaningfully, it will be better remembered than if processed on the basis of its surface characteristics like shape, colour or, sound.
Deep processing involves meaning or semantics: the ability to understand a new concept in terms of already familiar concepts and already familiar connections between them. Hence, deep, content-related, meaningful processing in working memory at the time of learning would be the first necessary step for the new information to be connected to existing knowledge in ways that will allow it to be accessible for future use.
To summarize, working memory processing resources are highly limited, and yet meaningful processing is essential for storage in long-term memory. It is therefore important to use these resources effectively when learning. There are many tested and proven effective teaching strategies, but a question that often comes up is when to apply each strategy for the best results?
Understanding the principal characteristics of the two major memory systems that are involved, their limitations, and importantly the interaction between them can support decisions regarding the use of strategies.
To that end, let's complete the picture by describing a model of knowledge representation in long-term memory over time. No knowledge base is built in a singular encounter with the information, Rather, a key element of our learning system is that it is shaped over time and on the basis of what we do with the knowledge each time we retrieve it from long-term memory and manipulate it in workingmemory.
The model includes four conceptual stages: know, understand, use, and practice to achieve mastery.
A model: evolving interactions between working memory and long-term memory.
Part 1: focus on long-term memory representations:
The first stage of learning any concept is the initial encounter: when a learner first sees or hears a completely new concept (word, object, etc.) that was never encountered before, and is therefore not represented in any way (left panel). Following the first encounter, the new concept may be represented (right panel): connections (edges in turquoise) are formed among activated nodes to create a small network. If this happens, the learner may be able to recognize this concept as familiar in the near future if it is explicitly presented again. It is possible to say that the learner knows this concept, but since the concept is not yet connected it may allow potential recognition but not much more.
For example: when you notice a new word in a foreign language.
Next, the now recognizable concept is explained and may start to become meaningful. Meaning is generated when the new concept (turquoise) is associated with other concepts (words, objects, procedures, etc.) that the learner is already familiar with (dark grey), in a way that makes sense to the learner. The learner now understands the meaning of the new concept. When we understand, we start to have a sense of how the concept may be useful, but this doesn't mean that the concept is truly helpful, yet. To make it happen we need to go through the next stage.
Back to the example: when the new foreign-language word is defined and/or translated.
Next, in order for something to be meaningful, it has to be functional: it is crucial to check if we can access the new concept and use it. The goal is to build pathways that will enable recalling the concept when needed - and we need to start constructing these pathways. The image on the right in comparison to the previous one demonstrates the difference between a concept that "makes sense" (upper) and a useful one (right), or between availability and accessibility (respectively). This highlights a point that is often hidden from learners: acquiring information by taking it in is not the same action as trying to "take it out", or actively retrieve it. We tend to be overconfident about functionality when something merely "makes sense". However, only when those retrieval pathways are established, that we can USE the learned concept11.
Back to the example: when we are able to use the new vocabulary word in a simple sentence to convey meaning.
4. PRACTISE to achieve MASTERY
When we practice using the new concept repeatedly over time, in various contexts, with connections to a range of cues, we gradually build more pathways, and then we establish and consolidate them, we eventually get to a state of MASTERY. The images on the right demonstrate how it takes a lot of effort to pave new pathways (left), but with repetitions, we manage to make those pathways more efficient (right): stronger, and more robust (omitting the necessary). When mastered, concepts, even complex ones, can be retrieved easily and quickly, even automatically. The concept can now be described as part of a “Schema” (a very well-connected and practiced network of concepts).
Back to the example: when we practice the new vocabulary world over and over again, in many different sentences, tenses, grammatical forms, and in relation to many other words and concepts until it becomes fully embedded in our active vocabulary and used fluently.
Part 2: How much effort does it take? focus on working-memory function:
Let's look at how each level of knowledge representation influences the requirements from working memory: how much mental effort is invested in manipulating the item in focus and what is left for other items and cognitive manipulations?
When a learner knows the concept, the only possible manipulation is recognizing it when presented. This is not very demanding (but also not very productive), and it also doesn't interfere much with other items.
When a concept is understood (e.g. when it is connected to prior knowledge), one can attempt to manipulate and use it, however, as the concept is not well-practiced at this stage, this seemingly simple action may be demanding. The attempt to hold it in working memory, along with its meaning in order to perform any type of action may result in cognitive overload. For example, when we try to understand a complex sentence or set of instructions that include a new concept that we have just grasped, sometimes the cognitive load forces us to drop something or to revert to sequential processing - decipher the new concept first, then the instructions.
With some initial practice (e.g. practicing the use of a new vocabulary word in sentences) it becomes easier to use the concept in practiced situations, and the load on working memory is gradually reduced. The concept becomes useful.
As we continue to practice over time, making connections to more concepts, and in different contexts, we are paving more effective pathways, and with additional repetitions, we consolidate the effective ones. While each step requires cognitive effort, the result is worth it:
Once the concept is mastered (e.g. when a word is fully integrated into functional vocabulary), using it becomes effortless and automatic. It requires much less of our working memory resources. In such a situation, the available resources can be used for deciphering incoming information, even when it is complex. The load is relieved.
Long-term memory and working memory interactions
It becomes clear how long-term memory and working memory interact when learning. The depth of processing in working memory influences the future potential representations in long-term memory and the levels of representation in long-term memory affects the working-memory capacity to handle additional information. When learners encounter a new concept, they must process it meaningfully in relation to already familiar concepts. However, the level of representation of these familiar concepts determines greatly how much resources are left for deep processing. The more established the knowledge is, the less processing resources are required, and the easier it is to process deeply. Research shows that learning new information based on a well-established schema is quicker and easier 3.
Choosing strategies with learning and memory in mind
The illustration below attempts to depict the relevant LTM and WM considerations along a continuum of learning something new over time and repetitions. The long-term memory representation in the middle, the working-memory consideration above (illustrated as a meter rather than as a box, but the idea is the same). Underneath, there are lists of examplar strategies that are generally appropriate to the stage of learning. It is often argued that novices and experts benefit from different learning strategies, this idea is depicted here. The working memory bar demonstrates that transitioning to a new phase may require much cognitive effort that is reduced with practice.
Creating connections :
In the initial learning phase, we should focus on the main new items or ideas and on creating meaningful connections to prior knowledge: well-crafted explanations, concrete examples, models, illustrations4, and analogies are many ways to harness existing concrete knowledge and help learners to make the intended connections. In this stage, holding the just learned information in mind may require most of WM resources (as the connections are not made yet), so it is wise to follow the guidelines from the Cognitive Load Theory5 and Mayer's principles of multimedia learning6, that emphasize methods to reduce undesirable load, selecting for and focusing on making a connection and integrating new and prior knowledge. However, creating connections is an initial necessary but insufficient step:
In order to make these connections truly meaningful, we have to make sure they are functional. In the practice phase, we are gradually building up this ability. We begin with activities that require the learners to use what they have just learned, emphasizing techniques that check for understanding7 and foster generative learning8. Next, we move to emphasize strategies that help maintain the long-term access pathways like retrieval practice (trying to answer questions, rather than reviewing answers), and gradually, as learners advance we can introduce desirable difficulties9 by increasing the level of difficulty by distributing the repetitions over time and varying the cues and the contexts (to practice transfer). These strategies create (eventually) robust pathways - that can be used effortlessly in various situations, this is when the learners demonstrate mastery.
This illustration stands for every level of learning: from simple concepts like words to more complex concepts that comprise any domain of knowledge. So this is actually a learning spiral: the process is repeated for every level, and if done effectively, then this level serves as a solid basis (well interconnected and easily retrieved) for higher-order learning. In this blog by science teacher Damian Benney you can see a practical example of how these ideas come to life in the high-school chemistry classroom, and here is the interpretation by Physics teacher Ben Rogers.
Willingham, D. T. (2009). Why don't students like school?: A cognitive scientist answers questions about how the mind works and what it means for the classroom. John Wiley & Sons.
2. McLeod, S. A. (2007, December 14). Levels of processing. Simply Psychology. www.simplypsychology.org/levelsofprocessing.html and:
Craik, F.I.M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268-294.
Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal behavior, 11, 671-684.
4. Caviglioli, O. (2019). Dual Coding for Teachers. John Catt Educational.
5. Cognitive load theory: Research that teachers really need to understand (2017) by the Centre for Education Statistics and Evaluation, State of New South Wales (Department of Education) [PDF]
6. Mayer, R. E. (2021). Evidence-based principles for how to design effective instructional videos. Journal of Applied Research in Memory and Cognition, 10(2), 229-240.
8. Fiorella, L., & Mayer, R. E. (2016). Eight ways to promote generative learning. Educational Psychology Review, 28(4), 717-741.
Caviglioli, O, Learning As A Generative Activity. A visual summary of Logan Fiorella and Richard E Mayer’s 2015 book Learning As A Generative Activity. download here: https://www.olicav.com/#/posters/
9. Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. Psychology and the real world: Essays illustrating fundamental contributions to society, 2(59-68). [PDF]
Brown, Roediger and McDaniel (2014) “Make it Stick: The Science of successful Learning” Harvard University Press
First published: April 2018, updated: Feb2022