Fig. 1: Difference between indexing theory and concept-neuron-based hippocampal coding of episodic memories, and procedures for Experiments 1 and 2.

a, Left, the indexing theory proposes that hippocampal neurons represent a conjunctive code, acting as pointers to various elements of an episode (the smell of the coffee, your friend, the background music, the café and so on) rather than directly coding for the elements themselves¹⁴. Right, concept neurons in the hippocampus are thought to code for specific elements or concepts^7,8,9. Within this framework, a group of neurons collectively code an episodic memory, with each neuron representing a specific element involved in that episode (that is, a neuron coding for the coffee, another neuron coding for your friend and so on^11,12). It is important to note that one index or one concept is likely to be coded by an assembly of neurons, not a single neuron. b, Procedure for Experiment 1. The participants imagined a vivid episode with an animal cue and two associate images (two faces, two places or a face and a place), rated its plausibility, and later recalled the associated images when cued with the animal cue. The experiment was self-paced, and each episode was learned and tested once. This approach is suitable for investigating episodic memory as originally defined by Tulving in 1972²⁹. Following each encoding block of roughly 20 episodes, the participants performed a short distractor task. The pink areas represent the time windows used for subsequent analyses (Methods). c, Procedure for Experiment 2. Left, the memory task was largely the same as in Experiment 1 (b), but events had one cue and one associate image (both an animal, a face or a place). Right, after the memory task, the participants performed a memory-free visual tuning task where previously shown stimuli were presented multiple times in quick succession. This approach has traditionally been used to identify putative concept neurons. The original images have been replaced due to copyright issues.