A fixed hippocampal memory timing test failed in 25 patients with spike data and 21 fMRI adults

We tested whether late hippocampal activity relates more strongly to later memory for its own event than for adjacent events. Separate public studies supplied human single-neuron spikes and BOLD fMRI. Development used eight patients and six MRI adults. A local code lock preceded tests in 25 other patients and 21 other adults. Mean own-event correlations were 0.09245 and 0.03754. Mean own-minus-neighbor contrasts were 0.10852 and 0.06393, with 95% person t intervals [-0.06252,0.27956] and [-0.00004,0.12791]. The fixed shared positive rule failed, p=0.10138. All negative people and controls remain. Most spike sessions have memory reports for only half the encoding events. Original-neighbor joins leave 12-13 centers per reserved patient. A later exact-center nuisance test was not estimable in any of these patients. MRI remained estimable but uncertain. An elementary shared-state model gives D=(1-rho)/2 with independent unit readout errors. A state with rho=0.9 has a positive timing score without event-triggered binding. Under the source masks, feature counts and nuisance designs, this slow state passed the full rule in 1,913/2,000 Gaussian panels. This is a true artificial association, not proof of binding or a fitted explanation of human data. A later stage comparison failed a 56-test Holm inventory. The package gives code, permitted compact inputs, source versions and terms, all results, figures, plans, audits and repeat instructions. Clinical sampling, sparse memory reports, different tasks, source filters and vascular effects limit scope. No positive cross-modal neural generalization or causal mechanism was confirmed.

hippocampus · episodic memory · single neurons · fMRI · temporal specificity · latent state · negative result · reproducibility