The multi-store model of memory is a psychological framework proposing that memory flows through three distinct stages: a sensory register, a short-term store, and a long-term store. Proposed by Richard Atkinson and Richard Shiffrin in 1968, it was one of the first models to describe memory as a sequence of separate systems rather than a single process. Despite being over 50 years old, it remains one of the most widely taught models in psychology and provides the foundation for how we talk about memory today.
How the Three Stores Work Together
The core idea is simple: information enters through your senses, gets briefly held in a sensory register, and either fades away or moves into short-term memory if you pay attention to it. From short-term memory, information can be transferred into long-term memory through rehearsal. At each stage, information that isn’t processed further is lost. Think of it like a series of filters, where each store has different rules for how much it can hold and how long it keeps things.
This linear flow is what makes the model both powerful and easy to understand. It gave researchers a shared vocabulary and a testable structure. But it also became the model’s biggest point of criticism, since real memory doesn’t always follow such a neat path.
The Sensory Register
The first stage captures everything your senses take in, holding a near-complete snapshot of what you just saw, heard, or felt. This store has a virtually unlimited capacity, but it fades almost immediately. Visual sensory memory (called iconic memory) lasts only a few hundred milliseconds after you perceive something. In experiments, performance on recall tasks drops to chance levels after about 2 seconds, meaning the sensory trace is effectively gone by then.
Auditory sensory memory (echoic memory) lasts a bit longer, generally a few seconds. This is why you can “replay” something someone just said even if you weren’t fully paying attention. The sensory register operates automatically. You don’t choose to store things there. What determines whether information survives past this stage is attention: if you focus on a piece of sensory input, it gets passed along to short-term memory. Everything else disappears.
Short-Term Memory
Short-term memory is where you actively hold and work with information. Its defining features are a tight capacity limit and rapid decay. George Miller’s classic 1956 paper suggested the limit is roughly seven items, give or take two. Miller also introduced the concept of “chunking,” the idea that you can expand effective capacity by grouping items into meaningful units. The letter string FBICIAUSA is nine letters but only three chunks if you recognize FBI, CIA, and USA as familiar acronyms. Miller himself considered chunking, not the number seven, to be the more important insight.
Without active rehearsal, information in short-term memory fades quickly. A landmark experiment by Peterson and Peterson tested recall of simple items after delays of 3, 6, 9, 12, 15, and 18 seconds, during which participants were prevented from rehearsing by being asked to count backwards. Recall dropped sharply across those intervals, showing that short-term memories can vanish in under 20 seconds if you can’t keep them active.
In the Atkinson-Shiffrin model, rehearsal is the bridge between short-term and long-term memory. The more you repeat or work with a piece of information, the more likely it is to make the transfer. Later research refined this idea by distinguishing between two types of rehearsal: maintenance rehearsal (simple repetition, like saying a phone number over and over) and elaborative rehearsal (connecting new information to things you already know). Elaborative rehearsal produces stronger, more retrievable long-term memories, while maintenance rehearsal mainly keeps things alive in short-term memory without necessarily creating a lasting trace.
Long-Term Memory
Long-term memory is the final store, and it differs from the other two in both capacity and duration. There is no known upper limit to how much it can hold, and memories stored here can persist for a lifetime. The trade-off is that getting information into long-term memory requires more effort, and retrieving it isn’t always reliable. Memories can be incomplete, distorted, or temporarily inaccessible even when they haven’t been truly lost.
According to the original model, long-term memory is primarily encoded semantically, meaning based on the meaning of information rather than how it looks or sounds. This contrasts with short-term memory, which tends to rely more on acoustic coding (the sound of words). That said, long-term memory clearly stores visual, emotional, and procedural information as well, which is one area where later models expanded on Atkinson and Shiffrin’s framework.
Evidence Supporting Separate Stores
One of the strongest pieces of evidence comes from the serial position effect. When people are asked to memorize a list of words and then recall them in any order, they tend to remember the first few items (the primacy effect) and the last few items (the recency effect) better than those in the middle. This produces a distinctive U-shaped curve. The primacy effect is thought to reflect long-term memory: early items receive the most rehearsal and are more likely to be encoded deeply. The recency effect reflects short-term memory: the last few items are still being held in the short-term store at the time of recall.
The most compelling clinical evidence came from patients with specific brain injuries. The most famous case is Henry Molaison (known in the literature as H.M.), who had portions of his medial temporal lobes removed to treat severe epilepsy. After surgery, H.M. could still hold information in short-term memory normally. He could carry on a conversation, repeat back a string of six or seven digits, and even retain a three-digit number for as long as 15 minutes through continuous rehearsal using an elaborate memory strategy. But the moment his attention was diverted to something new, he forgot the entire event. He could not form new long-term memories.
This pattern demonstrated that short-term and long-term memory rely on different brain structures. Damage to the medial temporal lobes (including the hippocampus) severely impairs long-term memory formation while leaving short-term memory intact. Conversely, damage to parts of the frontal lobes disrupts short-term memory and the recency effect but leaves long-term memory largely unaffected. The fact that one system can be damaged without impairing the other is strong evidence that they are genuinely separate.
Key Criticisms of the Model
The multi-store model’s biggest weakness is that it treats each store as a single, uniform system. Short-term memory, in particular, appears to be more complex than one box. Atkinson and Shiffrin themselves actually described verbal and visual short-term stores, and even used the phrase “working memory,” but their published model became widely known in its simplified three-box form. In 1974, Baddeley and Hitch proposed the Working Memory Model, which replaced the single short-term store with multiple components: a system for processing language, a system for visual and spatial information, and a central executive that coordinates them. This better explained how people can do two different kinds of tasks simultaneously (like navigating while listening to directions) without one overwhelming the other.
A second major criticism targets rehearsal as the primary way information enters long-term memory. People clearly form lasting memories for events they never deliberately rehearsed, like a surprising moment or an emotionally charged experience. Rehearsal matters, but it isn’t the only mechanism, and the model overstates its role.
The third critique concerns what happens when rehearsal stops. The original model assumes information simply decays from short-term memory without rehearsal, but some researchers argue that forgetting in short-term memory is caused by interference from new information rather than passive time-based decay. This debate remains active.
Why the Model Still Matters
Despite its limitations, the multi-store model established the foundational vocabulary that all later memory research builds on. The distinction between short-term and long-term memory is now supported by decades of neurological evidence, even if the details of each store are more nuanced than Atkinson and Shiffrin’s original framework suggested. When psychologists talk about working memory, episodic memory, or memory consolidation, they are refining categories that the multi-store model first clearly defined. It remains the starting point for understanding how memory is organized, and newer models are best understood as expansions of its core structure rather than replacements.