A full screen briefly registers whole in sensory memory, then almost all of it is gone
Aliases: sensory memory · iconic memory · echoic memory · sensory register · Sperling partial report
What it is
After the eyes sweep across a full screen of text, even though attention only lands on a few of the characters, the entire scene is briefly and completely registered in an earlier stage of memory called sensory memory. The visual portion of it is iconic memory; the auditory portion is echoic memory. This is a different thing from working memory, discussed elsewhere: capacity is nearly unlimited — an entire visual field's worth of detail can be registered — but duration is extremely short. Visual sensory memory typically fades within a few hundred milliseconds; auditory sensory memory lasts somewhat longer, but still only on the order of a few seconds.
Why it happens
Sensory memory is thought to be a raw, unprocessed trace-like registration of incoming stimulation by the sensory system — almost an afterimage the stimulus itself leaves in the nervous system, without any involvement of understanding or filtering the content. That's exactly why its capacity can approach near-panoramic coverage: registration doesn't require time-consuming semantic processing. But precisely because this record is raw and unconsolidated, it is extremely easy for incoming new information to overwrite, and it also decays rapidly on its own over time. Only if attention lands on some portion of the content within this very brief window does that portion get transferred to a more stable later stage of memory; whatever wasn't attended to disappears along with the sensory trace itself, leaving no residue.
Studying it
The classic method for demonstrating that sensory memory's capacity far exceeds what people can actually report is Sperling's partial report technique: a grid of letters is flashed briefly, and if participants are asked afterward to report every letter they saw, the number they can produce is quite limited. But if, right after the display ends, a cue signals that participants only need to report one specific row of the grid, their accuracy on that row is far higher than the proportion implied by the whole-report accuracy — showing that at the instant the display ends, information about the entire grid was in fact still present; there just wasn't enough time to report it all before it started fading. By systematically delaying when this cue appears, researchers can also measure the time course of sensory memory's decay.
Common independent variables: the delay before the cue signal appears, and the amount of information displayed. Common dependent variables: partial-report accuracy, whole-report accuracy, and the gap between the two.
Methodological caveat: the capacity and decay speed measured with the partial-report technique are highly sensitive to presentation parameters (brightness, contrast, display duration); comparing specific numbers across studies requires checking that presentation parameters actually match. This paradigm also mainly validates the visual channel — the auditory sensory memory needs a separate paradigm of its own to establish.
Where it stops holding
- This entry covers the basic property that sensory memory has large capacity and short duration; it does not cover how this kind of memory interacts with subsequent attentional allocation — that is a separate claim.
- Sensory memory's duration varies with the physical properties of the stimulus (brightness, duration, sensory channel), and the visual and auditory channels differ noticeably in magnitude. A single blanket number does not accurately cover every case.
- This entry applies only to the raw sensory trace before attentional processing. Once some portion of content is selected by attention, how it persists and for how long follows the rules of the subsequent memory stage, no longer bound by the decay rate of sensory memory.
Applying it
- Any design assumption relying on "the user glances at it once and gets all the information" needs scrutiny: even though everything on screen may be momentarily and fully registered at the sensory-memory level, only the portion that falls within the attentional window can actually be extracted and used — content outside that window is not actually acquired by the user.
- For scenarios requiring a user to react to a briefly flashed screen (rapid carousels, momentary prompts), don't assume users can recall every detail afterward the way they might replay a recording — sensory memory decays far faster than most people's intuition expects.
- Verification: design a simplified partial-report task, comparing users' overall recall accuracy for briefly displayed content against their immediate recall accuracy for one local portion. If local accuracy is noticeably higher, it shows that at the current display duration, a substantial amount of information is only sitting in sensory memory and never actually captured.