Sensory Memory

Written and Reviewed by the Cognitive Train Team
← Types of Memory

Every other memory system is judged by what it keeps. This one is judged by how fast it lets go.

Wave a sparkler in the dark and you don't see a moving dot. You see a line of light that isn't there.

Ask someone to repeat themselves, then answer before they can — pulling the question out of the air a second after it was asked.

Both moments expose the same machinery. Something briefly held on to sensory information after the world stopped supplying it, then let that information go. This is sensory memory, the stage every sight and sound passes through before it becomes anything you'd call an experience, and it is the only memory system whose success depends on losing almost everything it receives.

The System That Holds Too Much

Start with what makes it strange. Short-term memory is famously cramped — a handful of items, held for a few seconds by effort. Sensory memory is the opposite on both counts: it appears to capture a great deal of the sensory field at once, requires no effort whatsoever, and then expires almost immediately.

High capacity, no control, near-zero duration. That doesn't look like a memory system. It looks like a design flaw.

It also raises an awkward question. If the store really holds more than people can report, how would anyone know? Whatever goes unreported leaves no trace to point at.

Proving It by Not Asking for Everything

George Sperling's 1960 solution was to stop asking people what they saw.

He flashed a grid of twelve letters for a twentieth of a second. Asked to name the whole grid, people managed only four or five — which looked like a hard ceiling on how much the eye takes in. Then Sperling changed one thing. After the grid vanished, a tone told the viewer which single row to report.

Whichever row was cued, most of it came back.

The reasoning is what makes the experiment famous. The cue arrived after the display was gone, so nobody could have prepared for it — meaning far more than the reportable handful must have still been available in some rapidly fading form. The ceiling wasn't on seeing. It was on getting the information out before it decayed.

Sperling then delayed the tone, and watched the advantage shrink. By around a second, the cue bought nothing at all.

Sperling's Partial-Report Method A grid of twelve letters is flashed for a twentieth of a second. Asked to report everything, people recall about four or five letters. Cued after the grid vanishes to report one row, they recall most of that row, showing much more was briefly available. Sperling's Partial-Report Method Flashed for 1/20 of a second RBKL QXNP SDFZ ↑ a tone cues this row, after it vanishes Report everything ≈ 4–5 letters Report only the cued row most of any row → far more was there

Different Senses, Different Clocks

The visual version Sperling isolated is called iconic memory, and it is brutally short — the sparkler trail, gone in a fraction of a second. Hearing plays by different rules.

When Darwin, Turvey and Crowder rebuilt the partial-report design for the ear in 1972, playing items from separate spatial locations and cueing one afterward, the advantage persisted for up to about four seconds. Several times longer than the visual store.

That gap isn't a ranking of the senses. It follows from what each sense is looking at. A page of text sits still and can be re-read, so vision can afford a memory that expires instantly. A spoken word cannot be re-heard — by the time it ends, its beginning is already past, and nothing in the world will hand it back. Echoic memory holds longer because hearing has nothing else to hold on to.

Touch has a store of its own, usually called haptic memory. It's the reason you can register the shape and texture of something your fingers brushed against a moment ago, and it's thought to last a couple of seconds — between the two better-known channels, though it has been studied far less than either. Sensory memory is best understood as a general property of perception rather than a pair of special cases.

Why the Losing Is the Point

Now the design flaw starts to look like the design.

Your senses deliver vastly more per second than any mind could interpret, and almost none of it will matter a moment later. A system that stored it all wouldn't be a better brain — it would be a brain choked by its own input, still working through the previous second while the current one arrived. Something has to discard most of it, fast enough to make room for what's next.

Sensory memory is that filter. It keeps much more of the incoming stream available than you could ever report, just long enough for attention to reach in and pull out what's worth keeping, then clears itself for the next sample. The loss isn't the cost of the mechanism. The loss is the mechanism.

Which means the interesting question was never how much sensory memory holds. It's what governs the handoff.

Attention Decides What Survives

Attention is the gate, and it is narrow. Whatever you attend to in that fleeting window gets read out into short-term and working memory, where it can be rehearsed and manipulated and eventually stored. Everything not selected is overwritten by the next arrival — not deliberately erased, simply replaced. The sequence is less tidy than store-then-select, too: there's evidence that attention shapes how these fleeting representations form in the first place, rather than only picking through them afterwards.

This is where the everyday complaints come from. Someone gives you directions while you're thinking about something else and the words evaporate. You look right at your keys and later have no idea where they are. In neither case did perception fail. The sensory stage worked; the selection never happened, so nothing reached the systems that could have held it.

Only after that handoff does anything begin the process of encoding into memory you can actually retrieve. And the reason forgetting so often feels like a memory problem is that a great deal of it was never a memory problem at all — it was an attention problem, several stages earlier.

Everywhere, Invisible

The strangest thing about sensory memory is how thoroughly it hides.

It's in the streak behind a moving flashlight, and in the way a dropped plate seems to keep ringing for a beat after it lands. It's in the question you answer a second after you failed to listen to it, and in the texture of a fabric your fingers have already left. Each is the same thing: information still briefly usable after the world stopped supplying it.

Not every trick of perception belongs here, though. A bright flash leaves a floating patch because of an afterimage, a different process altogether, and the smoothness of film owes more to how the visual system computes motion than to anything lingering in a sensory store.

You never notice any of it, because a filter that worked perfectly would be undetectable by definition.

What's Actually Worth Training

No established method has been shown to extend the raw duration of the sensory store, and chasing it would be aiming at the wrong stage anyway.

What varies enormously between people is the next link: how much of what survives the filter your short-term and working memory can hold and use. That's measurable and it's consequential — it's the capacity that determines whether a set of spoken instructions makes it to the end. The Short Term Memory Test covers it across five formats, the Digit Span Test and N-Back Test push on working memory directly, the Memory & Recall section gathers the rest, and the wider brain tests collection maps the abilities feeding into it.

So the sparkler trail isn't a trick of the eye or a small failure of perception. It's the visible edge of a system quietly discarding the world at high speed so that a fraction of it can become a thought. Cognitive Train's free brain training exercises work on that surviving fraction, and the types of memory guide follows it through every system it reaches next.