Summary
A bioRxiv preprint reports a single-neuron correlate of mnemonic pattern separation in the human hippocampus. The finding came from recordings of 3,506 neurons in 97 patients performing a recognition-memory task involving false memories.
A bioRxiv preprint reports a single-neuron correlate of mnemonic pattern separation in the human hippocampus, a brain structure central to episodic memory. The researchers recorded activity from 3,506 neurons in 97 patients while they performed a recognition-memory task designed to produce false memories.
Pattern separation is the proposed ability to represent similar experiences as distinct memories. It could, for example, help distinguish between two similar images or events instead of blending them together. Although theoretical models have assigned this computation to the hippocampus, the preprint says its operation in the human hippocampus has remained contested because no established single-cell correlate had been identified.
The study is a preprint and has not yet undergone peer review.
How the false-memory task exposed memory signals
Participants viewed images and later had to recognise whether images were familiar. Some new images were similar to images seen earlier. These similar-but-new images sometimes produced false memories: participants incorrectly judged them to be familiar.
The researchers identified two types of memory-selective neurons distributed across the brain. One group responded differently to novel images that were falsely recognised and familiar images that were correctly recognised. According to the authors, this response pattern was compatible with pattern separation because it distinguished the underlying memory status of similar items even when the participant made an incorrect choice.
A second group of neurons signalled the participant’s choice. This distinction allowed the researchers to separate activity related to the memory representation from activity related to the decision made during the task.
From memory representation to decision
At the population level, the pattern-separation-signalling cells predicted the study’s mnemonic ground truth in the hippocampus. In this context, mnemonic ground truth refers to whether an image was actually novel or familiar, rather than whether the participant’s response was correct.
Signals in the pre-supplementary motor area, by contrast, predicted the participant’s decision. The authors interpret this as evidence for a progression from memory-related signals in the hippocampus to choice-related signals in a brain region involved in action selection and decision processes.
The researchers also tested the role of the pattern-separation-signalling cells by removing them from the population analysis. This abolished the continuous memory-strength gradient observed in the hippocampus. The result supports the authors’ interpretation that these cells help separate memories according to strength rather than encoding only a simple familiar-versus-new category.
The finding connects individual neuronal responses with a behavioural error—false recognition of a similar new image—and with the broader organisation of memory-guided decisions. It therefore offers a cellular-level account of how the human brain may preserve distinctions between overlapping experiences.
Evidence level and scope
The evidence comes from direct single-neuron recordings in the human brain during a controlled recognition-memory task. The sample comprised 97 patients and 3,506 recorded neurons, but the supplied source does not provide further participant-level details in its abstract.
The cell-removal result is a population-analysis test rather than a physical intervention in the brain. The study also examines pattern separation in the specific task used, so its findings describe neural processing during that experimental form of recognition memory. As a bioRxiv preprint, the report represents the authors’ current analysis pending peer review.