Cellular code for mnemonic pattern separation in the human hippocampus is revealed by false memories

The study addresses whether pattern separation—a computation proposed to help the hippocampus distinguish familiar from novel items—actually operates in the human hippocampus, a question that has remained contested without an established single-cell correlate. To investigate, researchers recorded single-neuron activity from 3,506 neurons in the human brain while 97 patients performed a recognition memory task designed to elicit false memories when novel images were similar to previously seen images, producing associated behavioral errors.

They identified two types of memory-selective neurons distributed across the brain. One type responded differently to falsely recognized novel images versus correctly recognized familiar images, a pattern compatible with pattern separation; the other signaled the subject’s choice. At the population level, these cells predicted mnemonic ground truth in the hippocampus and the decision in the pre-supplementary motor area, illustrating a progression from mnemonic signals to decisions. Removing the pattern separation–signaling cells abolished the continuous memory strength gradient present in the hippocampus, suggesting these cells help separate memories of different strength.

The results establish a single-cell correlate for mnemonic pattern separation in the human hippocampus and demonstrate its behavioral relevance in episodic memory. The findings link cellular-level pattern separation to false-memory errors and to downstream decision signals, advancing understanding of how the human brain differentiates similar experiences.