Fig 1.
Semantic memory including spatial components.
(a) (Left) Through repetitive experiences (e.g., reading a book in a study room, etc.), ILAs (e.g., Statue of Liberty ↔ northeast in the US) can be stored in the allocentric frame. The stored memory, which does not accompany a unique spatiotemporal context from a single episode (e.g., watching TV in a living room), could be categorized into semantic memory. (Right) The recall process of the item-location associative memory may differ between the presence (i) and absence (ii) of a map on which the item’s location can be shown. (i) When the map is present, the item can be located on it no matter how it is tilted (0° to 360°), suggesting the semantic recall in the allocentric frame. (ii) In the absence of the map, the item can be located in an assumed spatial context (e.g., top for the north), providing mental imagery of the retrieved location from the first-person perspective. The outline map image of the US was made based on an image from public domain, “Pixabay” (https://pixabay.com/illustrations/us-map-outline-us-map-america-1674031/). (b) ILA pattern. Two items, 1 from set A (e.g., I-A) and the other from set B (e.g., I-B), were assigned to each location (e.g., co-location I) on the map image. Scale bars for both item and map stimuli, with a 5° visual angle. The background map image was made based on an image (EMU 13) from public domain, “USGS” (https://www.usgs.gov/media/images/emu-13). (c) Schematic diagram of the ILA task. An item-cue and a context-cue were sequentially presented in each trial. The monkeys’ gaze was fixated on the center until the end of the delay-2, then saccade to the target location (indicated by the red arrowhead and red dashed circle) according to the 2 cues. A successful trial was rewarded with juice paired with feedback showing the associated location of the item-cue on the context-cue. Relative sizes of the stimuli were magnified for display purposes. ILA, item-location associative.
Fig 2.
Item-cue selective activities in the MTL and area TE (TE).
(a) Recording regions. MRI images corresponding to the coronal planes anterior 4, 10, and 17 mm from the interaural line of monkey C (right hemisphere). The recording regions are the PRC, HPC, and PHC of the MTL and TE. A reference electrode implanted in the center of the chamber was observed as a vertical line of shadow in the coronal plane at A10. amts, anterior middle temporal sulcus. rs, rhinal sulcus. ots, occipital temporal sulcus. D, dorsal. V, ventral. L, lateral. M, medial. (b) Percentage of item-cue selective neurons out of recorded neurons in each area. Parentheses, number of recorded neurons in each area. Dashed line, 1.0% chance level. The number of item-cue selective neurons was significantly larger than the chance level (P < 0.0001 for each area, one-tailed binomial test). Asterisk indicates results of a χ-square test: PHC and TE, P < 0.0001***, χ2 = 23.3, d.f. = 1; PHC and PRC, P < 0.0001***, χ2 = 24.5, d.f. = 1; PHC and HPC, P < 0.0001***, χ2 = 36.0, d.f. = 1. Source data are available in S1 Data. HPC, hippocampus; MTL, medial temporal lobe; PHC, parahippocampal cortex; PRC, perirhinal cortex.
Fig 3.
A PRC neuron showing the co-location effect on item-cue selective activities.
(a) Solid lines and dashed lines indicate SDFs in trials with item-cues from the stimulus sets A and B, respectively. Dark and light gray shading, 90% confidence interval of 10,000 bootstraps (see Materials and methods) for the stimulus sets A and B, respectively. Black and gray dots, raster plots for the stimulus sets A and B, respectively. Brown bar, presentation of the item-cue. The background map image was made based on an image (EMU 13) from public domain, “USGS” (https://www.usgs.gov/media/images/emu-13). (b) Mean discharge rates of the neuron during the item-cue period for each item. r, correlation coefficient. Asterisk indicates the result of a two-tailed permutation test: P = 0.0024***. Source data are available in S1 Data. PRC, perirhinal cortex; SDF, spike density function.
Fig 4.
Memory retrieval signal in the MTL.
(a) Distributions of co-location indices for item-cue selective neurons in each area. Red, TE (n = 47); blue, PRC (n = 84); purple, HPC (n = 136); green, PHC (n = 22). Co-location indices in all MTL areas were significantly positive (P < 0.0001, Y = 0, for PRC, P < 0.0001, Y = 0, for HPC, P = 0.0009, Y = 0, for PHC, two-sided Wilcoxon signed-rank test) and greater than those in TE (P = 0.0043, KS = 0.31 for PRC; P < 0.0001, KS = 0.59 for HPC; P = 0.0056, KS = 0.43 for PHC; Kolmogorov–Smirnov test). (b) Cumulative frequency histograms of co-location indices for item-cue selective neurons in 2 monkeys. For both monkeys, the HPC showed the largest co-location index (median = 0.92 and 0.84 for monkeys B and C, respectively), followed by the other 2 MTL areas (median = 0.86 and 0.35 in the PRC, 0.78 and 0.59 in the PHC for monkeys B and C, respectively). The co-location index was close to the lowest in the TE for both monkeys B (0.17) and C (−0.04). (c) Time course of population-averaged co-location indices. Lines and shading, mean and SEM across the item-cue selective neurons with high co-location indices (r > 0.8) in MTL areas. The population-averaged co-location index increased earlier in the PRC than in the HPC (P = 0.0382*, two-tailed permutation test). Arrow, half-peak time. (d) Cumulative frequency histograms of half-peak time of co-location indices for individual neurons. Half-peak times were calculated for each item-cue selective neurons with high co-location indices (r > 0.8) in MTL areas. The half-peak times were shorter in the PRC than in the HPC (P = 0.0395*, KS = 0.29, Kolmogorov–Smirnov test) and PHC (P = 0.0201*, KS = 0.57). Source data are available in S1 Data. HPC, hippocampus; MTL, medial temporal lobe; PHC, parahippocampal cortex; PRC, perirhinal cortex; SEM, standard error of the mean.
Fig 5.
An HPC neuron exhibiting both item-cue selectivity and target selectivity.
(a) Lines, SDFs for each item-cue during the item-cue period (left) or SDF for each target location during the choice-fixation period (right). I-IV, co-location I-IV. Brown bar, presentation of the item-cue. TR, top-right; BR, bottom-right; BL, bottom-left; TL, top-left. Gray bar, fixation on the target location. (b) Mean responses during the item-cue period for each co-location (top) and during the choice-fixation period for each target location (bottom). Error bar, standard error. Significant selectivity for the co-locations (P < 0.0001***, F (3, 42) = 56.6, one-way ANOVA) and for the target locations (P < 0.0001***, F (3, 42) = 129.3, one-way ANOVA). (c) Combinations of item- and context-cues resulted in the bottom-left (BL) target location. Mean responses during the choice-fixation period for combinations resulted in the BL target location. Error bar, standard error. Responses were not significantly different among the combinations that resulted in the BL target location (P = 0.9942, F (2, 8) = 0.0058, one-way ANOVA). The background map image was made based on an image (EMU 13) from public domain, “USGS” (https://www.usgs.gov/media/images/emu-13). Source data are available in S1 Data. HPC, hippocampus; SDF, spike density function.
Fig 6.
Representational similarity between co-locations and target locations.
(a) Correlation matrices between population vectors for co-locations of item-cues and those for target locations in each MTL area. The population vectors consist of activity from the recorded neurons in each area (PRC, n = 319; HPC, n = 456; PHC, n = 232). TR, top-right; BR, bottom-right; BL, bottom-left; TL, top-left. (b) Representational similarity between co-locations and target locations was estimated as a mean value of correlation coefficients for 4 combinations of co-locations and target locations corresponding to each of −90°, 0°, and 90° map image conditions. Error bar, standard deviation. Asterisk indicates results of two-tailed permutation test: P < 0.0001***. Source data are available in S1 Data. HPC, hippocampus; MTL, medial temporal lobe; PHC, parahippocampal cortex; PRC, perirhinal cortex.
Fig 7.
Two-stage recall model of semantic memory.
Schematic diagram of neuronal signals in the MTL during the item-cue period in the ILA task. Perceptual signal of the item-cue transmits from the neocortex to the PRC, in which the item information would be transformed to the mnemonically linked location information, which is represented in an allocentric manner. The retrieval signal would transmit (via the ERC) to the HPC, in which the allocentric location would be combined with the default context to construct the self-reference representation of the retrieved location. The self-referenced location signal would spread from the HPC (via ERC) to PHC. HPC, hippocampus; ILA, item-location associative; MTL, medial temporal lobe; PHC, parahippocampal cortex; PRC, perirhinal cortex.