Decoding intended recall

Reading out, from neural activity, the content a person is trying to retrieve — which memory, category, or item is being recalled — rather than what they are currently perceiving. Human evidence comes from category- and item-level decoding in the medial temporal lobe and cortex: reinstatement of encoding-time activity patterns at retrieval can be detected, and in some intracranial and fMRI studies the identity of a recalled item or its category can be classified above chance shortly before or during recall.

Genuine content decoding of specific episodic memories remains far more limited than motor or speech decoding. Memories are distributed, reconstructive, and idiosyncratically coded, retrieval is internally generated and hard to time-lock, and most demonstrations are coarse (category over exemplar) and rely on decoders trained on the same person's encoding activity. Claims of reading arbitrary private memories are, at present, unsupported.