How Human Memory Works, and Why the Brain Forgets

From Sensation to Long-Term Storage

Memory is not one process but several distinct stages — encoding, storage and retrieval — according to the Cleveland Clinic. Sensory memory holds raw input from the eyes, ears, skin, nose and tongue for only a couple of seconds before it fades or passes into short-term memory, the clinic explains.

Short-term memory, which involves the prefrontal cortex, has a limited capacity and duration — lasting from a few seconds to minutes unless information is rehearsed or grouped into chunks, the Cleveland Clinic states. More specific estimates come from classic psychology research summarized by Simply Psychology: George Miller proposed in 1956 that it holds roughly seven items, plus or minus two, and experiments using the Brown-Peterson task found unrehearsed information fades within about 15 to 30 seconds. Information judged important can then move into long-term memory, described by the Cleveland Clinic as a store with no known limit that can hold facts, events, skills and habits for years, split broadly into declarative memory (facts and episodes) and procedural memory (skills and habits).

How a Memory Is Physically Built

According to the Queensland Brain Institute at the University of Queensland, a memory is the reactivation of a specific group of neurons produced by lasting changes in the strength of connections between them. When one neuron repeatedly triggers another, their connection strengthens through a process called long-term potentiation, while rarely used connections weaken or disappear, the institute states.

The hippocampus, in the brain's temporal lobes, is the main memory-formation hub for episodic and declarative memories, both the Cleveland Clinic and the Queensland Brain Institute note. QBI adds that the hippocampus supports neurogenesis — the creation of new neurons — which studies in mice link to better memory performance, and that in humans, exercise linked to greater hippocampal volume is associated with stronger performance on memory tasks. Sleep plays a direct role as well: during slow-wave sleep, QBI explains, hippocampal neurons replay the day's experiences in compressed sequences in a dialogue with the neocortex that helps decide which details get stored long-term, and insufficient sleep interrupts that process.

Why the Brain Forgets

One of the oldest explanations, trace decay theory, holds that a memory trace fades automatically over time if it goes unused — an idea traced to early experiments by Brown in 1958 and by Peterson and Peterson in 1959, according to Simply Psychology. Critics of the theory point out that it struggles to explain why people can vividly recall distant events they have not thought about in years.

A second limit is simple capacity. Psychologist George Miller proposed in 1956 that short-term memory holds roughly seven items, plus or minus two; once that capacity fills, new information can displace what is already there, Simply Psychology notes, citing Waugh and Norman's 1965 experiments showing recall worsened as more digits intervened between learning and testing. In long-term memory, interference theory holds that similar memories compete with each other: proactive interference occurs when old learning disrupts new learning, and retroactive interference occurs when new learning disrupts recall of earlier material — a pattern Underwood and Postman demonstrated in 1960 by showing that learning a second word list impaired recall of an earlier one.

A third explanation is retrieval failure, also called cue-dependent forgetting: information stays stored but cannot be accessed without the right cue. Tulving and Pearlstone showed in 1966 that supplying category names sharply improved recall of words that had seemed forgotten. Context matters too — Baddeley's 1975 study of divers found they recalled about 40% more words when tested in the same environment, on land or underwater, where they had first learned them, and Goodwin and colleagues reported in 1969 that recall can also depend on physiological state, such as alcohol intoxication at the time of learning versus testing.

A fourth cause is a breakdown in consolidation itself. Simply Psychology cites the case of patient H.M., who retained normal short-term memory after 1953 surgery that damaged his hippocampus but lost the ability to form new long-term memories — evidence that the hippocampus is required to convert short-term memories into lasting ones.

Measuring Forgetting: The Ebbinghaus Curve

In 1885, German psychologist Hermann Ebbinghaus published the first systematic study of forgetting, testing only himself on thousands of lists of meaningless three-letter "nonsense syllables" designed to rule out any effect of prior knowledge. He measured forgetting as "savings" — how much faster he could relearn a list compared with learning it the first time — at intervals from 20 minutes to 31 days.

These figures come from a 2015 replication published in PLOS One by Jaap Murre and Joeri Dros of the University of Amsterdam, who repeated Ebbinghaus's method with a single subject learning 70 lists of nonsense syllables over 75 days. The authors reported a "remarkable resemblance" to the 1885 original, though with lower retention at the 31-day mark, a difference they attributed partly to longer learning times used during the modern experiment.