Long-term memory (LTM) is the memory system responsible for the storage and retrieval of information over extended periods. Unlike working memory, which is limited to a few items held for seconds, LTM has no known capacity limits and can retain information for a lifetime. It encompasses everything from personal experiences and factual knowledge to motor skills and conditioned responses, organized into multiple subsystems with distinct properties and neural substrates.
Taxonomy of Long-Term Memory
The most influential taxonomy distinguishes declarative (explicit) memory — consciously accessible memories that can be verbally reported — from nondeclarative (implicit) memory — memories expressed through performance rather than conscious recollection. Declarative memory further divides into episodic memory (personal experiences bound to specific times and places) and semantic memory (general knowledge independent of personal context). Nondeclarative memory includes procedural memory (skills and habits), priming, classical conditioning, and perceptual learning.
├── Declarative (Explicit)
│ ├── Episodic (events, personal experiences)
│ └── Semantic (facts, general knowledge)
└── Nondeclarative (Implicit)
├── Procedural (skills, habits)
├── Priming
├── Classical conditioning
└── Perceptual learning
Encoding into LTM
The transfer of information into LTM depends on the nature and depth of encoding. Craik and Lockhart's levels of processing framework demonstrated that deeper, more meaningful processing (semantic encoding) produces better memory than shallow processing (structural or phonological encoding). Elaborative encoding — connecting new information to existing knowledge — and distinctive processing — encoding what makes an item unique — both enhance LTM formation.
Consolidation
Newly formed memories are initially fragile and become more stable over time through consolidation. Systems consolidation theory proposes that memories are initially dependent on the hippocampus but gradually become independent of it as they are integrated into neocortical networks — a process that may take years. Sleep plays a critical role in consolidation: slow-wave sleep appears to support the replay and transfer of declarative memories, while REM sleep may consolidate procedural and emotional memories.
Henry Molaison (H.M.), who had his hippocampus and surrounding medial temporal lobe structures surgically removed in 1953 to treat epilepsy, demonstrated the critical role of the hippocampus in forming new declarative memories. After surgery, H.M. could no longer form new episodic or semantic memories (anterograde amnesia) but retained pre-surgical memories, working memory, and the ability to learn new motor skills. His case established the distinction between declarative and nondeclarative memory and launched decades of memory neuroscience research.
Retrieval
Retrieval from LTM depends on the availability of appropriate cues. The encoding specificity principle (Tulving) states that retrieval is most successful when the cues present at retrieval match those present at encoding. Context-dependent memory (better recall in the same environment as encoding) and state-dependent memory (better recall in the same physiological or emotional state) demonstrate the power of encoding-retrieval match. Retrieval is not a passive readout but an active reconstructive process that can modify the retrieved memory itself.
Neural Substrates
Different LTM systems depend on different brain structures. Episodic memory depends critically on the hippocampus and medial temporal lobe. Semantic memory involves distributed neocortical networks, particularly in the temporal and frontal lobes. Procedural memory depends on the basal ganglia and cerebellum. Priming involves changes in the same cortical areas that process the primed stimuli. This neural dissociation among memory systems provides some of the strongest evidence for the multiple memory systems framework.