A rapid, hippocampus-dependent, item-memory signal that initiates context memory in humans

Abstract

The hippocampus, a structure located in the temporal lobes of the brain, is critical for the ability to recollect contextual details of past episodes. It is still debated whether the hippocampus also enables recognition memory for previously encountered context-free items. Brain imaging and neuropsychological patient studies have both individually provided conflicting answers to this question. We overcame the individual limitations of imaging and behavioral patient studies by combining them and observed a novel relationship between item memory and the hippocampus. We show that interindividual variability of hippocampal volumes in a large patient population with graded levels of hippocampal volume loss and controls correlates with context, but not item-memory performance. Nevertheless, concurrent measures of brain activity using magnetoencephalography reveal an early (350 ms) but sustained hippocampus-dependent signal that evolves from an item signal into a context memory signal. This is temporally distinct from an item-memory signal that is not hippocampus dependent. Thus, we provide evidence for a hippocampus-dependent item-memory process that initiates context retrieval without making a substantial contribution to item recognition performance. Our results reconcile contradictory evidence concerning hippocampal involvement in item memory and show that hippocampus-dependent mnemonic processes are more rapid than previously believed.

Figure 1

Experimental Design, Hippocampal Volumes, and Neuropsychological Test Scores for Patients and Controls (A) Study and test phase trial sequences for the item and context memory task. At study, word-scene pairs were presented, and participants were required to judge whether the word denoted a living or nonliving object (see Supplemental Information). At test, old and new words were presented, and participants were required to judge the old/new status of the word. If they responded “new,” they rated their confidence with the options “not sure” or “sure” (see Supplemental Information for analyses of confidence judgments). If they responded “old,” they rated their confidence for the upcoming context decision with the options “not sure,” “sure,” or “very sure” and then chose the scene originally paired with the word from three alternatives (plus a blank square if they believed the scene was not presented). (B) Hippocampal volumes across the participant group showing controls (dark gray) and patients (white). (C) Literacy, numeracy, full-scale IQ (FSIQ), and memory quotient (MQ) neuropsychological test scores. Patients showed comparable performance to controls and to the standard population mean (i.e., 100 ± 15) in terms of literacy, numeracy, and FSIQ but showed a clear deficit in MQ. Error bars show ±1 SEM for each condition.

Figure 2

Correlating Hippocampal Volume with Memory Performance and Magnetoencephalographic Effects Correlation analyses between bilateral hippocampal volume and item memory (Pr) (A), context memory (conditional context hits) (B), the 300–350 ms occipitotemporal effect (hits – CRs) (C), and the 350–400 ms frontotemporal effect (hits – CRs) (D) across all participants. Solid lines represent the line of best fit, and dashed lines represent 95% confidence intervals. Patients are highlighted in red, controls in black.

Figure 3

Topographies, Event-Related Fields, and Time Windows of Interest for the Magnetoencephalographic Effects Topographies (A and D), event-related fields (ERFs) collapsed across patients and controls (B and E), and time-window analyses (C and F) for the clusters revealed in the 3D SPM analysis (see Figure S1). The occipitotemporal 300–350 ms effect (A–C) is shown; the frontotemporal 350–400 ms effect (see Supplemental Information for analysis of later 700–750 ms effect) (D–F) is shown. Topographies represent the difference in fT between hits and CRs (collapsed across controls and patients) at the mid time point in the 50 ms time window, with the black circles highlighting the sensors selected for further analyses. The ERF plots show the ERFs for hits and CRs averaged across the peak sensors highlighted in the topographies, with time windows for further analyses highlighted in gray. The time-window analyses represent the average (fT) within the 50 ms highlighted in gray in the ERF plots, plotted separately for controls and patients. Error bars show ±1 SEM for each condition; ^∗∗∗p < 0.001, ^∗∗p < 0.01, ^∗p < 0.05.

Figure 4

Event-Related Fields and Time Windows of Interest across Context Hits, Context Misses, and Item CRs for the Two Magnetoencephalographic Effects (A and D) ERFs collapsed across patients and controls and time-window analyses for (B and E) the early time windows (300–350 ms occipitotemporal effect and 350–400 frontotemporal effects) and (C and F) the 500–600 ms time window. The occipitotemporal effect (A–C) is shown; the frontotemporal effect (D–F) is shown. ERF plots show ERFs for context hits, context misses, and item CRs (collapsed across controls and patients), with time windows for further analyses highlighted in gray. The time-window analyses represent the average (fT) within the 50 ms or 100 ms highlighted in gray in the ERF plots, plotted separately for controls and patients. Error bars show ±1 SEM for each condition; ^∗p < 0.05.