A neural substrate in the human hippocampus for linking successive events

A neural substrate in the human hippocampus for linking successive events Paz Gelbard-Sagiv Mukamel Harel Malach Fried

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

A neural substrate in the human hippocampus for linking successive events 2010 Proceedings of the National Academy of Sciences 10.1073/pnas.0910834107 Paz2010-a-neural-substrate-in-the-human-hippocampus Paz2010-a-neural-substrate-in-the-human-hippocampus.txt, the project's staged copy 41b11779e66aa675fc0fc1f1f8403f76ec930ede not-redistributable neuroscience

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them how single human neurons come to signal the temporal relationship between successive events Quotation withheld because this record is not cleared for redistribution. how neuronal activity patterns evolve as memory for the temporal order of events is formed Quotation withheld because this record is not cleared for redistribution. shows hippocampal activity in successive time segments becomes mutually predictive across repetitions, specifically in hippocampus, and that this predicts each subject's later free recall Quotation withheld because this record is not cleared for redistribution. existence Quotation withheld because this record is not cleared for redistribution. memory for the temporal order of events Quotation withheld because this record is not cleared for redistribution. electrophysiology electrophysiology Quotation withheld because this record is not cleared for redistribution. mechanisms Because the hippocampus alone integrates successive events together with the context in which they occurred, it can build context-specific branching representations, which is why the developing temporal linkage appears there and not in amygdala or cingulate cortex Quotation withheld because this record is not cleared for redistribution. mechanisms Memory formation is the internal convergence of one network state onto the network state representing the stimulus that followed it in time, so that each state comes to lead reliably to the next Quotation withheld because this record is not cleared for redistribution. mutual information between successive time bins measures how much knowing activity now reduces uncertainty about activity next, which linear correlation cannot capture Quotation withheld because this record is not cleared for redistribution. the analysis deliberately avoids assuming a constant representation, looking at a neuron's relation to itself over time rather than to the stimulus Quotation withheld because this record is not cleared for redistribution. the increase comes from reduced conditional entropy rather than raised marginal entropy, so it is the relationship changing and not the firing statistics Quotation withheld because this record is not cleared for redistribution. the hippocampus is proposed to be the site because it alone can integrate successive events with the context in which they occurred, allowing branching representations Quotation withheld because this record is not cleared for redistribution. memories are framed as stable network states, with formation meaning one state leads more reliably to the next Quotation withheld because this record is not cleared for redistribution. Alternative accounts in terms of growing visual familiarity or improved retrieval cannot be ruled out completely. what this limits: the learning interpretation: WE CANNOT RULE OUT THESE POSSIBLE CONTRIBUTIONS COMPLETELY - growing visual familiarity and improving retrieval both remain available Quotation withheld because this record is not cleared for redistribution. Assemblies larger than three neurons could not be analysed with the data the paradigm yielded. what this limits: the assembly analysis: NEURAL ASSEMBLIES OF n > 3 REQUIRE MUCH LARGER AMOUNTS OF DATA AND COULD NOT BE MEASURED IN THE CURRENT PARADIGM Quotation withheld because this record is not cleared for redistribution. two alternative interpretations - growing familiarity and improving retrieval - cannot be completely ruled out, and the authors say so duplicate of: limitations[0] · what this limits: the same scope as limitations[0] - the two alternative accounts are argued against, not excluded Quotation withheld because this record is not cleared for redistribution. the familiarity account is argued against by an external finding rather than by a control in this dataset what this limits: the familiarity rebuttal specifically, whose evidence is external: IT WAS SHOWN RECENTLY THAT THE TOTAL NUMBER OF SPIKES ELICITED IN A HIPPOCAMPAL [response does not grow with familiarity] - a result from another paper, not a control here Quotation withheld because this record is not cleared for redistribution. the retrieval account would also fit the recall correlation, which the authors acknowledge what this limits: the correlation with subsequent memory: THIS INTERPRETATION ALSO WOULD FIT WITH THE REMARKABLE CORRELATION WE FOUND BETWEEN THE NEURONAL MEASURE AND THE SUBSEQUENT INDIVIDUAL MEMORY, so the headline result does not discriminate between the accounts Quotation withheld because this record is not cleared for redistribution. the data come from the same experiments already reported for a different finding what this limits: any count of independent studies: Quotation withheld Quotation withheld because this record is not cleared for redistribution. subjects are epilepsy patients implanted on purely clinical criteria what this limits: the population and the recording sites: SINGLE AND MULTIUNITS WERE RECORDED (27 SESSIONS IN 13 INDIVIDUALS) in epilepsy patients whose implants follow clinical need Quotation withheld because this record is not cleared for redistribution. mutual information and its decomposition into marginal and conditional entropy, used both as the measure and as the control on it Quotation withheld because this record is not cleared for redistribution. the paradigms requiring memory for temporal order, against which this free-viewing design is explicitly contrasted Quotation withheld because this record is not cleared for redistribution. the clips re-shuffled so that the six viewings came from ENTIRELY DIFFERENT movie clips while staying within the same repetition number eliminates: the experimental context - the room - and the mere passage of time as the source of the temporal relationships, both of which advance identically whichever clip is shown · outcome: held Quotation withheld because this record is not cleared for redistribution. temporal relationships is defined operationally as mutual information between spike counts in successive time windows, expressed as a percentage of the maximum possible information Quotation withheld because this record is not cleared for redistribution. two rival sources of the same correlation are named and separately estimated: context-independent drift, measured in blank periods, and pure-stimulus structure, measured by shuffling across repetitions Quotation withheld because this record is not cleared for redistribution. the free-viewing paradigm is defended as closer to real memory formation than animal paradigms, because it uses no reinforcement and needs no overtraining Quotation withheld because this record is not cleared for redistribution. memory formation is recast as a change in which network state reliably follows which, rather than as a change in what a neuron represents Quotation withheld because this record is not cleared for redistribution. method: single-unit recording in epilepsy patients freely viewing repeated audiovisual clips, with mutual information between successive time windows as the measure a second reading DIFFERING from the recorded value. Recorded: electrophysiology — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. phenomenon: temporal binding during episodic memory formation a second reading DIFFERING from the recorded value. Recorded: memory for the temporal order of events — note: a second reading, verbatim-verified, disagreeing with the value on the record. Not applied - a person decides. Quotation withheld because this record is not cleared for redistribution. builds-on → Hebb's original proposal (ref. 33) that neuronal assemblies can be linked temporally Quotation withheld because this record is not cleared for redistribution. extends → the same group's report of hippocampal reactivation during free recall, from the same experiments - a paper independently tagged in this corpus Quotation withheld because this record is not cleared for redistribution. supports → Hebb's proposal that neuronal assemblies can be linked temporally Quotation withheld because this record is not cleared for redistribution. supports → fMRI evidence that hippocampal activation tracks retrieval of temporal order in naturalistic scenes status: the paper's own words: we cannot rule out Quotation withheld because this record is not cleared for redistribution. qualifies → the interpretation that the effect is growing visual familiarity Quotation withheld because this record is not cleared for redistribution. temporal relationships —operationalised-as→ mutual information temporal relationships mutual information strength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span writes 'the relationship between subsequent activities', not 'temporal relationships'. Quotation withheld because this record is not cleared for redistribution. temporal relationships —causes→ free recall temporal relationships free recall strength of the relation: 0.7 · note: the paper says related directly to rather than causes; r = 0.87 across subjects. the weight carries the correlational status Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but temporal relationships are 'RELATED DIRECTLY TO' memory performance across subjects; causes overstates an association. Quotation withheld because this record is not cleared for redistribution. hippocampus —distinct-from→ amygdala hippocampus amygdala strength of the relation: 1 · note: regional specificity is itself the control: an effect present only in hippocampus cannot be general arousal Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span reports a gradual increase in the hippocampus but not other structures; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. temporal relationships —distinct-from→ context-independent temporal relationships context-independent strength of the relation: 0.95 · note: estimated separately in blank periods and subtracted, rather than assumed absent Quotation withheld because this record is not cleared for redistribution. temporal relationships —distinct-from→ pure-stimulus temporal relationships pure-stimulus strength of the relation: 0.95 · note: estimated by shuffling across repetitions of the same clip Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span describes stimulus correlations to be controlled; distinct-from is loose. Quotation withheld because this record is not cleared for redistribution. mutual information mutual information Quotation withheld because this record is not cleared for redistribution. correlation correlation Quotation withheld because this record is not cleared for redistribution. regression regression Quotation withheld because this record is not cleared for redistribution. t-test t-test Quotation withheld because this record is not cleared for redistribution. multiple-comparison correction multiple-comparison correction Quotation withheld because this record is not cleared for redistribution. chi-squared chi-squared Quotation withheld because this record is not cleared for redistribution. ANOVA ANOVA Quotation withheld because this record is not cleared for redistribution. Fisher's exact test Fisher's exact test Quotation withheld because this record is not cleared for redistribution. Fisher's combined probability test Fisher's combined probability test Quotation withheld because this record is not cleared for redistribution. permutation (shuffle) test permutation (shuffle) test Quotation withheld because this record is not cleared for redistribution. Mann-Whitney U test Mann-Whitney U test Quotation withheld because this record is not cleared for redistribution. bias correction of an estimator bias correction of an estimator Quotation withheld because this record is not cleared for redistribution. mutual information between successive time windows mutual information between successive time windows Quotation withheld because this record is not cleared for redistribution. shuffling across repetitions shuffling across repetitions Quotation withheld because this record is not cleared for redistribution. Fisher's combined test Fisher's combined test Quotation withheld because this record is not cleared for redistribution. marginal and conditional entropy decomposition marginal and conditional entropy decomposition Quotation withheld because this record is not cleared for redistribution. With repeated viewing, activity in successive time segments became gradually correlated, so activity in one window came to predict the activity that followed. Quotation withheld because this record is not cleared for redistribution. This linkage appeared after only two or three viewings of an episode and was stronger than both context-independent and stimulus-driven correlations. Quotation withheld because this record is not cleared for redistribution. The gradual growth of temporal relationships across clip repetitions occurred in the hippocampus only, not in amygdala, anterior cingulate or entorhinal cortex at the single-cell level. Quotation withheld because this record is not cleared for redistribution. The strength of hippocampal temporal relationships during viewing predicted how many clips a subject later recalled freely, with a correlation of 0.87 across subjects. Quotation withheld because this record is not cleared for redistribution. The effect was not an artifact of changing firing rates, because mean firing rate was stable across both clip repetition and clip duration. Quotation withheld because this record is not cleared for redistribution. Decomposing mutual information showed the rise came from falling conditional entropy (better prediction of the next window) rather than from rising marginal entropy (changed firing statistics). Quotation withheld because this record is not cleared for redistribution. When simultaneously recorded pairs and triplets of cells were analysed, a significant increase in temporal relationships appeared in entorhinal cortex as well. Quotation withheld because this record is not cleared for redistribution. Just over half of all recorded neurons (362 of 671) showed a significant increase in temporal relationships for at least one clip. Quotation withheld because this record is not cleared for redistribution. activity in successive time windows became mutually predictive across clip repetitions Quotation withheld because this record is not cleared for redistribution. the effect emerged within two to three presentations Quotation withheld because this record is not cleared for redistribution. the gradual increase appeared in hippocampus and not in amygdala, entorhinal cortex or anterior cingulate Quotation withheld because this record is not cleared for redistribution. only in hippocampus did the effect exceed both rival sources of correlation Quotation withheld because this record is not cleared for redistribution. the measure correlated with each subject's later free recall at r = 0.87, and in no other structure Quotation withheld because this record is not cleared for redistribution. the same result held with ordinary Pearson correlation instead of mutual information Quotation withheld because this record is not cleared for redistribution. the rise in mutual information came from falling conditional entropy, not from changing firing statistics Quotation withheld because this record is not cleared for redistribution. overall firing rate was stable across repetitions and along clip duration Quotation withheld because this record is not cleared for redistribution. no evidence of developing slow oscillations that could mimic the effect Quotation withheld because this record is not cleared for redistribution. temporal relationships temporal relationships Quotation withheld because this record is not cleared for redistribution. pure-stimulus relationships pure-stimulus relationships Quotation withheld because this record is not cleared for redistribution. context-independent correlations context-independent correlations Quotation withheld because this record is not cleared for redistribution. responsive neurons responsive neurons Quotation withheld because this record is not cleared for redistribution. temporal relationships temporal relationships Quotation withheld because this record is not cleared for redistribution. brain structure brain structure Quotation withheld because this record is not cleared for redistribution. mutual information mutual information Quotation withheld because this record is not cleared for redistribution. memory performance memory performance Quotation withheld because this record is not cleared for redistribution. number of spikes number of spikes Quotation withheld because this record is not cleared for redistribution. r = 0.87, P < 0.01 — across-subject correlation between hippocampal temporal relationships and free-recall success Quotation withheld because this record is not cleared for redistribution. 54% (362/671) — proportion of all recorded neurons showing significant temporal relationships for at least one clip Quotation withheld because this record is not cleared for redistribution. 52% (349/671) — proportion of recorded units classified as single units rather than multiunits Quotation withheld because this record is not cleared for redistribution. AH, n = 180; amygdala n = 160; EC n = 224; AC n = 107 — number of units recorded in each brain structure across 27 sessions in 13 patients Quotation withheld because this record is not cleared for redistribution. AH, 49%; EC, 34%; amygdala, 23%; AC, 25% — proportion of neurons per structure whose temporal relationships exceeded the pure-stimulus baseline Quotation withheld because this record is not cleared for redistribution. 22% — proportion of neurons whose during-clip temporal relationships exceeded their context-independent (blank-period) relationships Quotation withheld because this record is not cleared for redistribution. 250 ms — width of the time window into which spikes were binned for the main analysis Quotation withheld because this record is not cleared for redistribution. n = 4,884 pairs and n = 2,1061 triplets — number of simultaneously recorded cell pairs and triplets analysed for assembly-level temporal relationships Quotation withheld because this record is not cleared for redistribution. 671 units in 27 sessions from 13 individuals; hippocampus 180, amygdala 160, entorhinal 224, anterior cingulate 107 — the recorded sample Quotation withheld because this record is not cleared for redistribution. r = 0.87, P < 0.01 — correlation between the hippocampal measure and individual recall Quotation withheld because this record is not cleared for redistribution. 10 to 16 clips of 5 to 10 s, each repeated six times — the viewing design Quotation withheld because this record is not cleared for redistribution. within two to three presentations — how fast the effect emerged Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1073/pnas.0910834107.