Long-term memory comprises distinct systems: declarative (explicit) memory includes facts and episodes accessible to consciousness, depending on medial temporal lobe structures; procedural (implicit) memory involves motor skills and habits expressed through action, depending on basal ganglia and cerebellum. These systems operate independently, explaining why people can lose declarative knowledge while retaining motor skills.
From your prerequisite on memory storage and consolidation, you know that long-term memories are not stored as a single unified trace but involve structural changes distributed across neural systems. This topic extends that understanding by revealing that there are fundamentally different *kinds* of long-term memory that rely on different neural architectures, operate by different principles, and can be selectively damaged or preserved.
Declarative memory — sometimes called explicit memory — is memory that can be consciously retrieved and stated in words. It divides into two subtypes you encountered in cognitive neuroscience: episodic memory (specific personal events) and semantic memory (general facts and concepts). What unifies them is accessibility to consciousness: declarative memories can be deliberately retrieved, evaluated, and communicated. Neurologically, declarative memory depends critically on the medial temporal lobe (MTL), particularly the hippocampus and surrounding entorhinal, perirhinal, and parahippocampal cortices. The MTL binds together distributed cortical representations into a coherent retrievable memory trace during encoding and consolidation.
Procedural memory — sometimes called implicit memory — operates entirely differently. It is memory expressed through skilled performance rather than conscious recollection. Riding a bike, touch-typing, hitting a tennis backhand, tying shoes — these are procedural memories. Critically, you cannot typically explain in words exactly what your muscles are doing, and trying to consciously introspect on an automated skill often *degrades* it (the "centipede's dilemma" — think too hard about walking and you stumble). Procedural memory depends on the basal ganglia (for habit learning and stimulus-response associations) and the cerebellum (for motor timing and error correction) — structures with little direct connection to the MTL system.
The power of this distinction comes from the dissociations revealed in neurological patients. The patient H.M., after bilateral hippocampal removal, had profound anterograde amnesia — he could not form new declarative memories. Yet he showed normal procedural learning: when tested on mirror-drawing (tracing a star while viewing only a mirror reflection), he improved across sessions at the same rate as controls — and had no memory of ever having done the task before. Each session, he believed he was doing it for the first time. His procedural memory system was intact while his declarative system was destroyed. The opposite pattern also exists: patients with basal ganglia damage (Huntington's disease) show impaired procedural learning while declarative memory is relatively preserved.
These double dissociations confirm that declarative and procedural memory are not simply strong and weak versions of the same process — they are genuinely different systems. This has practical implications for rehabilitation: a patient with MTL damage may still be able to acquire new motor skills, routines, and habits even if they cannot consciously remember learning them. Therapy can bypass the damaged declarative system by targeting the intact procedural system, designing intervention around repetition and practice rather than explicit instruction and recall.