The value of a simple theory of working memory

You are walking through a busy town centre, and a campaigner stops you to chat. While she explains her cause, someone carrying a new door to be fitted passes through the space between you, briefly obscuring your view. When you can see each other again, your conversation with the campaigner resumes. How do you remember where you left off? Should you listen or speak? What would be proper to say? Is the person now standing before you even the same person you had been speaking with?

That last question may seem daft, but well-controlled experiments have shown that you probably wouldn't notice if the unfamiliar person you were speaking with were magically swapped with some other unfamiliar person. When such a test is carried out, with the campaigner moving away with the passing door and another person surreptitiously replacing them, about half do not notice the change, and simply carry on with their conversation. 

These studies of 'change blindness' are powerful demonstrations that our ability to take in details around us is starkly limited – much more so than we feel or notice. In any given moment, we take in a small amount of information. Exactly which information we absorb changes rapidly in line with our goal. Your goal was planning what to say or predicting what the campaigner might say next, not remembering the campaigner. As the door passed, any details about what the campaigner looked like vanished in a blink.

The mental system we use to parse new information and decide what to do with that information is known as working memory. If scientists know one thing about working memory, it is that working memory is severely limited. Until we're so familiar with something that it is unforgettable, information is subject to the limitations of working memory. Even adults at their peak mental health can only hold a small number of independent thoughts in working memory at once. 

What is that small number? Perhaps if you studied Psychology, you'll remember from your undergraduate days the rule of thumb from a classic 1956 paper from George Miller, 'The Magical Number 7 plus or minus 2'. But in November, we gathered as members of the organising committee for CowanFest, the festschrift to celebrate a distinguished researcher whose theories of memory and attention have suggested that regardless of what kind of information we're considering, adults can actually hold only 3-5 discrete things in mind at once. 

The one-day meeting in November 2024 included talks from invited speakers and a poster session for early-career researchers to present new ideas, all inspired by Professor Nelson Cowan, currently Curators' Distinguished Professor in the Department of Psychological Sciences at the University of Missouri. Cowan's 'magical number four' draws upon a wide range of literature including memory tests of visual information like abstract drawings or pictures of familiar objects, as well as memory tests of verbal information like lists of words or numbers. And it has implications for how we teach and learn.

A constant cycling of ideas

The multiple component working memory model has strongly influenced how psychologists view cognitive functioning. A key feature of the multiple component model is the assumption that different parts of the mind handle temporary storage of distinct kinds of information. Although there is some truth to this, if practice can be better served by a simpler concept, namely that our capacity at any given moment is extremely limited, then psychologists applying working memory theory in practice might benefit from Cowan's insight. 

So how is Cowan's approach different from thinking of working memory as a system for holding some verbal and some visual thoughts? Cowan thinks of working memory as whatever ideas are currently the most active in long-term memory. Colloquially, you might consider that the things at the 'top of your mind' are in Cowan's 'focus of attention', which is limited to 3 to 5 thoughts regardless of whether they are in verbal or visual form. When you are solving a problem, the bits of information you are using are in this focus of attention. Other related ideas might be quickly called upon from long-term memory and swapped into this focus of attention. Very unrelated things would take more time to activate if called upon. Information that we need to fulfill incremental goals is quickly cycled into and then out of the focus of attention as we identify what information is needed, evaluate it, and choose a response. If we are familiar with a task, the next step follows fluently, and we hardly notice any difficulty in shifting from one part of a task to the next. 

According to Cowan's model of working memory, this fluency arises because working memory is part of long-term memory, and activation spreads between connected ideas, with different ideas becoming more or less active quickly over time. However, this processing can break down when we are completing tasks where the steps are not well known or when we are using information that is new to us and not strongly linked to more familiar concepts: we forget more of the incoming information, activate fewer relevant and helpful pieces of long-term knowledge, and may ultimately lose track of the broader goal we're working towards before we achieve it.

How is the embedded processes model different to thinking of working memory in terms of separate verbal and visual components? Fortunately, the two models are similar in many respects, so it is not necessary to forget everything you know about the utility of verbal rehearsal or avoiding interference. But Cowan's focus on this constant cycling of ideas as activation spreads and emphasis on the very limited capacity 'at the top of your mind' leads to somewhat different practical applications. Thinking in terms of verbal and visual components may lead practitioners to assume that capacity limits are best dealt with by presenting information in a mixture of formats. Even worse, they may be tempted to assume that it is fine to present too much information, so long as some of it is words and some of it is pictures. While mixing information formats can be beneficial, that is not because the amount we can focus on is the sum of some verbal limit plus some visual limit. Cowan's framework asserts that the overall limit still applies. 

So, if there is a way to use limited working memory capacity more effectively, it lies in building up long-term knowledge – the discrete concept you are thinking of becomes a rich 'chunk' connecting to multiple ideas. Recalling the intervening door study – this trick only works with an unfamiliar stranger because the stranger's features are new to you. Therefore, they are not quickly integrated with long-term knowledge. You would not be fooled if a close friend tried to play this trick on you. You would not need the focus of attention to maintain lots of new details about your friend, so you would easily notice if they swapped themselves for a new conversation partner.

Communicating a new concept

Cowan's framework integrating immediate attention with long-term memory leads to two conceptually simple recommendations for communicating a new concept. These recommendations should be particularly useful for audiences who struggle most to learn, such as children under 10, people with attention disorders, or people with a lot on their minds. 

Importantly, working memory appears similarly limited for all types of information once you restrict the influence of familiarity and the ability to use specialised techniques like verbal rehearsal to boost memory. Across childhood, children learn a lot about the world that allows their long-term memory to better support their working memory. While this information network is developing, you can expect younger children to hold even less information before it is – poof! – gone. 

So, here's what we can do:

1. Don't overestimate how much information can be taken in. Break your overall goal into small sub-steps. Ensure that completing each step is supported before moving on. 

2. Do find ways to connect new information to concepts the audience is likely to know well. Help them recognise these similarities so that they may scaffold novel information onto their existing knowledge. 

Professor Nelson Cowan recently delivered the keynote address at the annual meeting of the Psychonomic Society in New York. His address is available here.