The Multi-Store Model of Memory

Every sound and sight you encounter first lands in sensory memory, a brief, high-capacity holding area that preserves a near-exact copy of raw input for less than a second to a few seconds. Iconic memory holds visual snapshots for about half a second; echoic memory holds sound for a few seconds longer, which is why you can still "hear" the last words of a sentence even after someone stops talking. Almost all of this fades fast because the system discards it unless attention selects it for the next stage.

Attention moves selected information into short-term memory, a small, fragile store that holds roughly seven items for up to about 30 seconds. This is where the model's key mechanism lives: rehearsal, the repeating or active working of information, is what pushes it from short-term into long-term memory, a vast and durable store. Attention gets information in the door, but only rehearsal moves it further. Interrupt that rehearsal, say, with a counting task, and the short-term trace decays or gets bumped before it transfers, so recall for that material drops sharply.

This staged flow explains the serial position curve from memory research: words from early in a list have had time for repeated rehearsal and are more likely to have reached long-term storage, producing the primacy effect, while the last few words are still sitting in short-term memory at the moment of recall, producing the recency effect. Words in the middle get squeezed out of both stores and recalled worst. The model's divisions set up later refinements of short-term memory, like working memory, which examine what happens inside that stage in more detail.