Lab Talk

working memory

The Puzzle of Working Memory

Working memory, the capacity to keep things ‘in mind’ such as a phone number, is a phenomenon poorly understood in Neuroscience. What do we know about it so far?

Try to memorize a nine-digit phone number from a single read and then type it into a phone without looking back. Can you do it without a mistake? If you can, you have a high working memory capacity. What determines this capacity to store numbers and why is it so limited? Neuroscience does not yet understand it.

A Brief History of Working Memory

Research on on this phenomenon started many decades ago. In the 50’s one important paper was written by George Miller, who summarized results of many studies made in the past. He wrote a paper titled “The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information“ (Miller 1956). Miller noticed a repeating pattern across a variety of studies that at first appeared to have nothing to do one with another. When he reviewed how many different levels of tones or tastes people can distinguish (“An already heard tone or a new tone?”), the same number showed up. When reviewing the short-term memory capacities (then called immediate memory), again the same number jumped out. When looking at the literature studying the span of attention, the same number came out of the experiments. What tied these together? The paper he wrote became one of the most influential publications ever written in psychology.

It took quite some years after that until enough additional knowledge had been accumulated so that Allan Baddeley had ground to re-brand short-term memory into something else. Baddeley noticed that we ‘work’ with our short-term memory, and suggested that we think about it as working memory (Baddeley 1992). We use working memory to comprehend text, think, solve problems, and reach insights. In other words, Baddeley told us that is not just about storage of information. It is about our full cognitive capacity – about our intellect. Hundreds of studies ensued and confirmed and further enlightened Baddeley’s ideas.

The Curious Features of Working Memory

Throughout these studies, a few fundamentally important things about working memory have been discovered. One is that it is possible to distinguish a ‘pure’ capacity of working memory from the capacity one gains from practice and the use of  mental tricks. For example, when memorizing a phone number, it will help if you memorize them with a special rhythm that organizes them in groups of three. When we take out all of the tricks, it seems like the ‘pure’ capacity of working memory, at least in vision, lays, on average, at about four items (Cowan 2001).

Another discovery is that we can store four items only if we have categories for things that are being stored (Olsson & Poom 2005). If you get multiple items that are not categorizable — items that don’t appear to have any relationship to anything that you have seen before so you have no separate drawers for item one and item two–your working memory capacity drops down to one. Yes, one. You can remember only one thing if you do not already have a unique category for that thing. It follows that our working memory effectively juggles only mentally meaningful items. Working memory has a hard time working with raw information to which no semantic interpretation can be assigned. This result goes hand in hand with what George Miller discussed in his original paper when he noted that we store information better if we are able to organize it into familiar chunks (“IBM FBI KGB” is easier to remember than “DJUGBEGMT).

Yet, another surprising finding is that working memory capacity is tightly related to attention. The more items you are capable of storing in working memory, the more you are able to simultaneously spread your attention across those stimuli and detect outliers. For example, westerners have a much harder time storing Chinese character than Latin letters, and they are exactly that much slower in finding a different Chinese character among same ones (Awh & Jonides 2001).

Thus working memory is the place where we think, juggle ideas, solve problems, and understand the world – the epicenter of our conscious mental life. It is thus also no surprise that the capacity of working memory (short-term memory) is highly correlated to IQ (Cohen & Sandberg 1977). The more you can trick your basic working memory capacity i.e., the more you can extend it over four or seven items, the higher your IQ will be. If it is no problem for you to read a nine-digit phone number and type it in confidently with no mistakes, you will probably score well on IQ tests.

see related post Intelligence and Neuroscience

In my own studies I was surprised to discover that the speed of information storage in long-term memory depended on a person’s capacity of working memory. Some people learn faster than others. I was able to predict how quickly someone will store into long-term memory certain visual patterns simply by looking how effectively they could store this type of patterns in their working memory (Nikolić & Singer 2007).

The Missing Brain Theory of Working Memory

While research in cognition regularly reaches new insights into the functioning of working memory, brain science is struggling to figure out how such mental phenomena emerge from neural activity. We don’t know what it is in the brain that limits the capacity of working memory. How is working memory implemented in the brain on the first place? Why is working memory so tightly related to attention and semantics? Where is semantics in the brain anyway?

In neuroscience, at least one type of studies are easy: locating things. Thanks to fMRI it is now relatively straightforward to establish which brain areas “light up” with more blood-oxygen flow during working memory tasks. We now know, for example, that the contents of working memory are mostly stored in the posterior sensory areas of the brain, while prefrontal areas are involved in control of that storage (e.g., Mayer et al. 2007).

See related post Stimulation, Sensation and Localization in the Cortex

But the true mystery is how the whole thing works. How do billions of neurons end up with a limitation of four items stored in working memory? And how is this simultaneously related to attention, semantics and long-term memory?

There have been a few attempts to explain the limited capacity of short-term memory, but these studies focused solely on the capacity (e.g., Siegel et al. 2009). They asked: what is it that could possibly be limited to four in the brain? But they did not ask: how does semantics play a role? How does this affect attention? Why does working memory only store meaningful categorical information? How is this information carried over to long-term memory?

We need a brain theory that explains all that. Only if all these aspects are reasonably well accounted for by a single theory, then we are likely on our way of having a satisfactory answer to understanding working memory.


The author, Danko Nikolić, is affiliated with savedroid AG, Frankfurt Institute for Advanced Studies and Max Planck Institute for Brain Research



Awh, E., & Jonides, J. (2001). Overlapping mechanisms of attention and spatial working memory. Trends in Cognitive Sciences, 5, 119-126

Baddeley, A. (1992). Working memory. Science255(5044), 556-559.

Cohen, R. L., & Sandberg, T. (1977). Relation between intelligence and short-term memory. Cognitive Psychology, 9(4), 534-554.

Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral & Brain Sciences, 24, 87-114.

Mayer, J. S., Bittner, R. A., Nikolić, D., Bledowski, C., Goebel, R., & Linden, D. E. (2007). Common neural substrates for visual working memory and attention. Neuroimage36(2), 441-453.

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological review63(2), 81.

Nikolić, D., & Singer, W. (2007). Creation of visual long-term memory. Perception & psychophysics69(6), 904-912.

Olsson, H., & Poom, L. (2005). Visual memory needs categories. Proceedings of the National Academy of Sciences, 102, 8776-8780.

Siegel, M., Warden, M. R., & Miller, E. K. (2009). Phase-dependent neuronal coding of objects in short-term memory. Proceedings of the National Academy of Sciences106(50), 21341-21346.



3 thoughts on “The Puzzle of Working Memory

  1. You review the work of Joaquin Fuster, who showed frontal lobe units activate during working memory to keep the memorandum active.

Leave a Reply