Table 1.

Models of memory, their predictions and related human intracranial studies.

Model class	Key predictions	Related studies
Hippocampus VTA/SN loop model	Dopamine neurons respond to novel stimuli, this novelty response is indicative of memory formation success, and exhibits single-trial learning.	[24, 27]
Sparse and separate encoding of semantic and episodic aspects of declarative memories	Highly visually selective cells (response sparsity), separate neurons encoding whether a stimulus has been seen before or not and other episodic aspects,	[35–39, 58]
Temporal context model, contextual reinstatement	Jump-back in time during retrieval, slowly drifting neural state, recollection/high confidence recognition predicted by reinstatement	[77, 78, 81] [79, 80]
Mnemonic evidence accumulation	Separate accumulators for “new” and “old” decisions. Accumulator output signal is modulated by memory strength. Accumulators predict choices rather than ground truth.	[97, 98]
Balance-of-evidence for metacognitive/confidence judgments	Magnitude of difference between the separate integrators for “new” and “old” choices is proportional to confidence, whereas the sign of the difference is indicative of the choice. Both decisions can be made at the same point of time.	[58]