how fuzzy memories improve decision-making

You’ve only been in the shopping centre for a few minutes, but back in the car park, you suddenly freeze. Where did I park? The memory feels gone. You guess and start to head left. Then you see the sign – “Blue Zone 1” – and realise your guess was correct.

This everyday experience is at the heart of new research colleagues and I have published in the journal Computational Brain & Behavior.

It shows that even “wrong” short-term memories may not be empty guesses.

What is short-term memory?

When we compare options, debate opinions, and recall very recent events, we hold this information in our “working memory”, also known as our short-term memory.

Researchers agree short-term memory has its limits – most people can only keep four or fewer items in mind at any one time. But the nature of these limitations is debated.

One popular “slot” theory suggests information is either stored fully or not at all. Think of this like a light switch: on or off. If the light is off, any resulting decision is a stab in the dark.

A competing “resource” theory suggests information is stored with different levels of accuracy. Think of this like a lighthouse shining through fog. Some memories are lit up, clear and strong. Others are just out of sight, fuzzy and only touched by the faintest of light.

If true, these fuzzy signals may still help us recognise important information and make good decisions, even when it feels like the memory has disappeared entirely.

Searching for the signal

My colleagues and I used advanced models of decision speed and accuracy to test whether “guessed” memories contained useful information for a subsequent recognition decision.

Put simply, if memory works like a light switch, then guesses will contain no useful information when assessed in a later recognition task.

However, if guesses come from fuzzy memories, then even wrong recollections will contain some useful information for later recognition.

To test these theories, we showed participants up to six colourful dots. After hiding the dots, we then asked them to recall the colour of one of the previous dots on a colour wheel.

Once they’d chosen their colour, we gave them a second shot.

After calculating their response error, we presented the correct colour alongside an alternative one that was located an equal yet opposite distance from the response location as the correct colour on the colour wheel. This ensured the target and alternative (or “distractor”) colours were similarly difficult to recognise, relative to the participant’s initial recollection.

We then asked participants to recognise the correct colour from among the pair.

Fuzzy or gone?

As expected, people recalled colours less precisely when they had more items to remember. But what happened when their answers looked like guesses?

We first separated “guessed” memories (for example, responses made on the wrong side of the colour wheel) from memories centred on the correct colour.

The rate at which participants accurately recognised the correct colour far exceeded chance following guesses for four or fewer items (more than 70%) and on average, was above chance for six items (55%).

Our modelling shows why: even when recall looked very wrong, people were still drawing on the same fuzzy memory trace to achieve better than chance recognition.

We didn’t take this finding at face value.

Using the latest computational methods, we considered alternative accounts: swap errors (where the wrong item was initially recalled and then corrected for during recognition), complex guessing patterns, varying memory limits, and combinations thereof.

No other account explained our findings.

Why ‘fuzzy’ short-term memories matter

Daily life rarely gives us the time or mental space to rely on perfect recall, especially when the information we’re retrieving isn’t all that crucial: the exact location of our parked car or the exact details of an acquaintance’s face.

Our work shows that even when short-term memories feel like they’re gone, they may still hold information that’s useful for making correct decisions.

Understanding this is important.

We often treat short-term memory like a light switch. But that might be leaving people in the dark.

For example, comprehensive short-term memory tests need to assess not only recall, but how recognition memory helps people achieve their daily tasks.

Our work encourages people to reevaluate “guesses” as low-precision memories.

For example, instead of dismissing mistakes as errors, educational tools might use recognition prompts to probe memory lapses and encourage learning.

Our work also supports the use of redundant information when navigating settings using short-term memories. “Blue Zone 1” is a great memory prompt to navigate a car park, but it’s also incredibly helpful in other settings such as hospitals.

Finding ways to better support short-term memory for people’s everyday decision-making, as in the above examples, is a matter for future research.

For now, when you next feel a memory is lost, know that it might not be gone. With any luck, it’ll be there, catching just enough light to lead you in the right direction.