Memories are not stored as exact replicas of experiences; instead, they are modified and reconstructed during retrieval and recall. Memory storage is achieved through the process of encoding, through either short- or long-term memory. During the process of memory encoding, information is filtered and modified for storage in short-term memory. Information in short-term memory deteriorates constantly; however, if the information is deemed important or useful, it is transferred to long-term memory for extended storage. Because long-term memories must be held for indefinite periods of time, they are stored, or consolidated, in a way that optimizes space for other memories. As a result, long-term memory can hold much more information than short-term memory, but it may not be immediately accessible.

Sensory memory allows individuals to retain impressions of sensory information for a brief time after the original stimulus has ceased. It allows individuals to remember great sensory detail about a complex stimulus immediately following its presentation. Sensory memory is an automatic response considered to be outside of cognitive control. Information from sensory memory has the shortest retention time, ranging from mere milliseconds to five seconds. It is retained just long enough for it to be transferred to short-term (working) memory. There are two types of sensory memory, echoic memory which is the memory of sound and iconic memory which is the memory of an image.

Short-term memory (working memory) is the ability to hold information for a short duration of time (on the order of seconds). In the process of encoding, information enters the brain and can be quickly forgotten if it is not stored further in short-term memory. When several elements (such as digits, words, or pictures) are held in short-term memory simultaneously, their representations compete with each other for recall or degrade each other.

In contrast to short-term memory, long-term memory is the ability to hold information for a prolonged period of time. There are two kinds of long-term memory, explicit (or declarative) and implicit memory. Explicit memories are those we can consciously remember or recall, and include episodic memories of events/experiences as well as semantic memories of common knowledge or concepts. Implicit memories are less conscious, and include procedural memories like how to ride a bike and emotional conditioning.

Items stored in short-term memory move to long-term memory through rehearsal, processing, and use. The capacity of long-term memory storage is much greater than that of short-term memory, and perhaps unlimited. However, the duration of long-term memories is not permanent; unless a memory is occasionally recalled, it may fail to be recalled on later occasions. This is known as forgetting.

Long-term memory storage can be affected by traumatic brain injury or lesions. Amnesia, a deficit in memory, can be caused by brain damage. Anterograde amnesia is the inability to store new memories; retrograde amnesia is the inability to retrieve old memories. These types of amnesia indicate that memory does have a storage process. Retrograde memories refer to memories that occur prior to a specific point in time – generally an injury or illness that affects long-term memory. 

A variety of different memory models have been proposed to account for different types of recall. In order to explain the recall process, however, a memory model must identify how an encoded memory can reside in memory storage for a prolonged period of time until the memory is accessed again, during the recall process. Note that all models use the terminology of short-term and long-term memory to explain memory storage.

The multi-trace distributed memory model suggests that the memories being encoded are converted to vectors (lists of values), with each value or “feature” in the vector representing a different attribute of the item to be encoded. These vectors are called memory traces. A single memory is distributed to multiple attributes so that each attribute represents one aspect of the memory being encoded. These vectors are then added into the memory array or matrix (a list of vectors). In order to retrieve the memory for the recall process, one must cue the memory matrix with a specific probe. The memory matrix is constantly growing, with new traces being added in.

The multi-trace model has two key limitations: the notion of an ever-growing matrix within human memory sounds implausible, and the idea of computational searches for specific memories among millions of traces that would be present within the memory matrix sounds far beyond the scope of the human-recalling process. The neural network model is the ideal model in this case, as it overcomes the limitations posed by the multi-trace model and maintains the useful features of the model as well.

The neural network model assumes that neurons form a complex network with other neurons, forming a highly interconnected network; each neuron is characterized by the activation value (how much energy it takes to activate that neuron), and the connection between two neurons is characterized by the weight value (how strong the connection between those neurons is). In this model, connections are formed in the process of memory storage, strengthened through use, and weakened through disuse.

The dual-store memory search model, now referred to as the search-of-associative-memory (SAM) model, remains one of the most influential computational models of memory. Two types of memory storage, short-term store and long-term store, are utilized in the SAM model. In the recall process, items residing in the short-term memory store will be recalled first, followed by items residing in the long-term store, where the probability of being recalled is proportional to the strength of the association present within the long-term store. Another type of memory storage, the semantic matrix, is used to explain the semantic effect associated with memory recall.

Spreading activation is a theory proposed that relates storage of memories to the activation of a series of nodes. These nodes create an activation pattern of other related nodes and this is how singular events are remembered. For example at the mention of the colour red and the word vehicles two nodes that may be activated at the same time relate a person memory of a car they owned or a fire truck.