The hippocampus is thought to have a critical role in episodic memory by incorporating the sensory input of an experience onto a spatial framework embodied by place cells. Although the formation and stability of place fields requires exploration, the interaction between discrete exploratory behaviors and the specific, immediate and persistent modifications of neural representations required by episodic memory has not been established. We recorded place cells in rats and found that increased neural activity during exploratory head-scanning behaviors predicted the formation and potentiation of place fields on the next pass through that location, regardless of environmental familiarity and across multiple testing days. These results strongly suggest that, during the attentive behaviors that punctuate exploration, place cell activity mediates the one-trial encoding of ongoing experiences necessary for episodic memory.
At a glance
- The Hippocampus as a Cognitive Map (Clarendon Press, 1978). &
- The global record of memory in hippocampal neuronal activity. Nature 397, 613–616 (1999). , &
- Hippocampal place cells acquire location-specific responses to the conditioned stimulus during auditory fear conditioning. Neuron 37, 485–497 (2003). , , , &
- The medial temporal lobe. Annu. Rev. Neurosci. 27, 279–306 (2004). , &
- Evolution of declarative memory. Hippocampus 16, 795–808 (2006). &
- Hippocampal place cells: parallel input streams, subregional processing, and implications for episodic memory. Hippocampus 16, 755–764 (2006). , &
- The behavioral ecology of intermittent locomotion. Am. Zool. 41, 137–153 (2001). &
- Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr. Clin. Neurophysiol. 26, 407–418 (1969).
- Rats and mice share common ethologically relevant parameters of exploratory behavior. Behav. Brain Res. 125, 133–140 (2001). , , , &
- Stopping behavior: constraints on exploration in rats (Rattus norvegicus). Behav. Brain Res. 53, 21–33 (1993). , &
- Voles scale locomotion to the size of the open-field by adjusting the distance between stops: a possible link to path integration. Behav. Brain Res. 141, 73–81 (2003). , &
- “Short-stops” in rats with fimbria-fornix lesions: evidence for change in the mobility gradient. Hippocampus 4, 577–582 (1994). , , , &
- The point of entry contributes to the organization of exploratory behavior of rats on an open field: an example of spontaneous episodic memory. Behav. Brain Res. 182, 119–128 (2007). &
- Quantifying the buildup in extent and complexity of free exploration in mice. Proc. Natl. Acad. Sci. USA 108 (suppl. 3), 15580–15587 (2011). , , , &
- The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells. J. Neurosci. 7, 1951–1968 (1987). &
- Comparison of population coherence of place cells in hippocampal subfields CA1 and CA3. Nature 430, 456–459 (2004). , , &
- Head direction cell representations maintain internal coherence during conflicting proximal and distal cue rotations: comparison with hippocampal place cells. J. Neurosci. 26, 622–631 (2006). , &
- Functional differences in the backward shifts of CA1 and CA3 place fields in novel and familiar environments. PLoS ONE 7, e36035 (2012). , , &
- The hippocampal memory indexing theory. Behav. Neurosci. 100, 147–154 (1986). &
- The hippocampal indexing theory and episodic memory: updating the index. Hippocampus 17, 1158–1169 (2007). &
- Hippocampal synaptic plasticity: role in spatial learning or the automatic recording of attended experience? Phil. Trans. R. Soc. Lond. B 352, 1489–1503 (1997). &
- Relationship between hippocampal theta activity and running speed in the rat. J. Comp. Physiol. Psychol. 88, 324–328 (1975). , &
- First occurrence of hippocampal spatial firing in a new environment. Exp. Neurol. 62, 282–297 (1978).
- Brain aging: changes in the nature of information coding by the hippocampus. J. Neurosci. 17, 5155–5166 (1997). , , &
- Hippocampal plasticity across multiple days of exposure to novel environments. J. Neurosci. 24, 7681–7689 (2004). , &
- A sequence predicting CA3 is a flexible associator that learns and uses context to solve hippocampal-like tasks. Hippocampus 6, 579–590 (1996).
- Experience-dependent asymmetric shape of hippocampal receptive fields. Neuron 25, 707–715 (2000). , &
- Prediction, sequences and the hippocampus. Phil. Trans. R. Soc. Lond. B 364, 1193–1201 (2009). &
- Hippocampal-neocortical interactions in memory formation, consolidation, and reconsolidation. Annu. Rev. Psychol. 61, 49–79 (2010). &
- Hippocampal-cortical interactions and the dynamics of memory trace reactivation. Prog. Brain Res. 193, 163–177 (2011). &
- Long-term stability of the place-field activity of single units recorded from the dorsal hippocampus of freely behaving rats. Brain Res. 509, 299–308 (1990). &
- Experience-dependent, asymmetric expansion of hippocampal place fields. Proc. Natl. Acad. Sci. USA 94, 8918–8921 (1997). , &
- A double dissociation between hippocampal subfields: differential time course of CA3 and CA1 place cells for processing changed environments. Neuron 42, 803–815 (2004). , &
- Dynamics of the hippocampal ensemble code for space. Science 261, 1055–1058 (1993). &
- Long-term plasticity in hippocampal place-cell representation of environmental geometry. Nature 416, 90–94 (2002). , , , &
- Gradual changes in hippocampal activity support remembering the order of events. Neuron 56, 530–540 (2007). , &
- Neuronal code for extended time in the hippocampus. Proc. Natl. Acad. Sci. USA 109, 19462–19467 (2012). et al.
- Independent codes for spatial and episodic memory in hippocampal neuronal ensembles. Science 309, 619–623 (2005). et al.
- Hippocampal neurons encode information about different types of memory episodes occurring in the same location. Neuron 27, 623–633 (2000). , , &
- Independent rate and temporal coding in hippocampal pyramidal cells. Nature 425, 828–832 (2003). , &
- Prospective and retrospective memory coding in the hippocampus. Neuron 40, 1227–1239 (2003). &
- A cognitive map for object memory in the hippocampus. Learn. Mem. 16, 616–624 (2009). &
- Robust conjunctive item-place coding by hippocampal neurons parallels learning what happens where. J. Neurosci. 29, 9918–9929 (2009). , &
- Hippocampal sharp waves: their origin and significance. Brain Res. 398, 242–252 (1986).
- Place-selective firing of CA1 pyramidal cells during sharp wave/ripple network patterns in exploratory behavior. Neuron 49, 143–155 (2006). , &
- Neural ensembles in CA3 transiently encode paths forward of the animal at a decision point. J. Neurosci. 27, 12176–12189 (2007). &
- Representation of non-spatial and spatial information in the lateral entorhinal cortex. Front. Behav. Neurosci. 5, 69 (2011). &
- Functional correlates of the lateral and medial entorhinal cortex: objects, path integration and local-global reference frames. Phil. Trans. R. Soc. Lond. B 369, 20130369 (2014). , &
- Place representation within hippocampal networks is modified by long-term potentiation. Neuron 39, 843–853 (2003). , &
- Hippocampal place fields emerge upon single-cell manipulation of excitability during behavior. Science 337, 849–853 (2012). , &
- Dynamic interactions between local surface cues, distal landmarks, and intrinsic circuitry in hippocampal place cells. J. Neurosci. 22, 6254–6264 (2002).
- An information theoretic approach to deciphering the hippocampal code. Adv. Neural Inf. Process. Syst. 5, 1030–1037 (1993). , , &
- Quantifying location-specific information in the discharge of rat hippocampal place cells. J. Neurosci. Methods 127, 123–135 (2003). , , &
- Supplementary Figure 1: Illustrations of the signal processing steps used to detect head scanning. (150 KB)
This figure accompanies the methodological descriptions in Methods/Scanning events. The signal processing for isolating head-scan events is based on four head-motion quantities (a) derived from Cartesian (x, y) position data: track angle (α), track-angle velocity ( ), track-centered radius ( ), and radial velocity Track-angle velocity is used to characterize forward locomotion on the track (a, left), while radius and radial velocity are used to characterize lateral head movement (a, right). Gray rats: position at time t; black rats: position at time t+Δt for a short interval Δt. The signal processing flow (b) iterates across every position data frame (x, y) in a recording session. First, is checked against a running-velocity threshold (10 °CW/s, or ~6 cm/s) to determine if the animal is running forward around the track. If the current frame is a running frame, then ( and ) are added to respective 4 s buffers of lateral movement samples ( buffer and buffer) and the iteration continues to the next frame. If not running, then and are checked to see if the position is off-track or if either of the values are significantly outside of their respective buffered distributions of expected lateral movements (Eqns. 5 and 6). If so, the frame is marked as a putative scanning sample and the iteration continues. After the last position frame, the set of scanning samples are processed into head-scan events, which are then filtered (e.g., by minimum duration) to determine the final set of scans.
- Supplementary Figure 2: Example scanning events for double cue-rotation and novelty sessions. (544 KB)
Additional examples of detected head-scan events (Methods/Scanning events) are shown as highlighted segments (red) of the trajectory (black) across the session. We show examples from different rats performing the double cue-rotation (DR; a) and novelty (b) experiments.
- Supplementary Figure 3: Velocity modulation of local theta and interneuron activity during head scanning. (176 KB)
We showed power spectra of CA1 local field potentials (Fig. 1f) demonstrating theta power during head scanning as intermediate between non-scanning pauses and forward running. This result might be explained by velocity-modulation of the local theta rhythm combined with a mixture of slow (pause-like) and faster (running-like) movements of the head and body during scanning. a, Velocity-modulation curves for CA1 theta power (top; n = 21 rats with CA1 tetrodes) and frequency (bottom) show that increased head speed correlates with higher theta power (two-tailed Friedman test across 16 speed bins, χ2(15,320) = 133, P = 6.28×10–21) and higher theta frequency (χ2(15,320) = 108, P = 3.74×10–16). Black/gray: mean ± s.e.m. across rats. b, Average head speed sharply drops at the start of scanning and recovers on a time-course approximating the median scan duration (dotted line). c, Temporal cross-correlations (mean ± s.e.m. across rats) of theta power (top) and frequency (bottom) with scan start times show that theta power is sharply reduced, consistent with the pause from running that precedes a scan. Theta recovers strongly within the time-scale of the typical scan. d, To address whether interneuron activity was also modulated by scanning, we collected spike trains from 225 interneurons recorded from 19/24 rats with potentiation events. Putative interneurons were identified by high average firing rates and short spike widths. Cross-correlograms for individual cells were normalized by mean firing rate and within-rat averages were median-filtered with a 0.37 s kernel to reduce noise. Interneuron cross-correlations (n = 19 rats with interneuron data; mean ± s.e.m.) demonstrate a moderate reduction in firing rate centered on scan start times. The time-course of head speed is similar to that of reduction in both theta power and interneuron firing rates, suggesting that theta and interneuron activity during scanning are largely controlled by head speed, consistent with observations of other exploratory behaviors.
- Supplementary Figure 4: Illustration of place-field potentiation event detection procedure. (194 KB)
This figure is a supplement to the methodological descriptions in Methods/Field potentiation events. The procedure examines the firing within an isolated place field on every lap of the recording session that has at least 2 baseline laps (preceding the test lap) and at least 2 follow-up laps (after the test lap). a, The nominal amount of potentiation Cpotentiation that we search for is adjusted based on the lap number m of the test lap (i.e., the term (0.5 + m/N) in Eqn. 10) from 50–150% from the start to the end of the session. Illustrated potentiation events of the same size (double arrows) are shown occurring ‘Early’, ‘Middle’, and ‘Late’ within the session to demonstrate why this is necessary. The dot-product measure Cjm (Eqn. 9) uses the whole-session firing-rate map (‘Session’) as a stable reference for comparing field strength across laps. However, note that the strength of the session rate-map decreases with later potentiation events due to averaging place-field activity over the session. The lap adjustment in Eqn. 10 counteracts this effective decrease in potentiation threshold. b, A more realistic illustration of a place-field potentiation event with variable lap-to-lap activity (top) shows the test lap m, the 5 baseline laps B, and the 5 follow-up laps F. To test lap m for potentiation (bottom), the dot-product strength Cj is computed for every lap (blue circles) and the baseline and follow-up thresholds (magenta lines; Eqns. 10 and 11) are determined relative to the strength Cjm of the test lap. The baseline (cyan line) is the maximum field strength of all baseline laps in B. If the baseline is below the potentiation threshold (Eqn. 12) and at least 3 (or 2, if there are only 2) follow-up laps in F are above the persistence threshold (Eqn. 13), then lap m is a potentiation event.
- Supplementary Figure 5: Place-field potentiation and weakening events: Firing-rate maps and prevalence. (215 KB)
a–b, Linearized firing-rate maps of the whole-session, forward-running activity of potentiated place fields (a) and place fields with mid-session ‘weakening’ events (b), sorted on the ordinate by track-angle COM of each field. c, Track-unwrapped firing-rate maps (as in Fig. 2b) of 356 place-fields weakening events (Methods). d–g, Event prevalence (mean ± s.e.m. across rats) is measured as the ratio of potentiation (d–e; Supplementary Table 1) or weakening (f–g) events to the number of recorded place fields across the 5 DR and 3 novelty sessions. In DR experiments, Sessions 2 and 4 were cue-mismatch configurations; in novelty experiments, Session 2 was the novel room. For the first day of testing (d), potentiation prevalence varied significantly across sessions (DR: n = 14 rats, two-tailed Friedman test, χ2(4,65) = 12.3, P = 0.0156; novelty: n = 6 rats, χ2(2,15) = 9.48, P = 0.00875). While the DR experiment showed a general decline in potentiation events across sessions, the novelty experiment showed a sharp increase in the prevalence of potentiation events in the novel session (session 2), consistent with prior reports of abrupt place field formation in novel environments23-25. A similar trend is apparent in Session 2 of the DR experiment, which is the first time that the rats ever experienced the mismatch condition of the otherwise familiar sets of cues. Considering all testing days (e), prevalence varied across DR sessions (n = 20 rats; χ2(4,95) = 20.2, P = 0.000467) but showed only a trend in the novelty experiment (n = 10 rats; χ2(2,27) = 5.20, P = 0.0743). Weakening events (f–g) occurred less frequently overall than potentiation events, but also varied with session on the first day of the novelty experiment (though without the large increase in Session 2 prevalence observed for potentiation events; f, right; n = 6 rats, χ2(2,15) = 6.42, P = 0.040) and across all testing days of the DR experiment (g, left; n = 20 rats; χ2(4,95) = 16.5, P = 0.00239).
- Supplementary Figure 6: Categorized examples of scan-activated place-field potentiation events. (533 KB)
A large selection of significant scan–potentiation examples are presented to illustrate both the variability of the effect and consistency of the colocalization of scanning activity and subsequent place field changes. Examples are shown using the compact display scheme introduced in Figure 3b, where the left column is a polar firing-rate map of place-field activity on the laps before the predictive scan, the middle three columns show three lap trajectories centered on the potentiation event lap, and the right column is a polar firing-rate map of the place field(s) after the lap following the potentiation event. Peak firing rates (spikes/s) are displayed at the center of polar plots; lap numbers are displayed at the center of lap plots. In lap plots, spike color is behavior-coded: scan (blue), peri-scan (cyan), pause (orange), forward running (red), and other (black). Each example is labeled on the left with the cell identification: e.g., ‘Rat091-05-1 t7c1’ indicates rat subject 91, testing day 5, session 1, tetrode 7, and cell 1. a, Place fields that were silent or nearly silent prior to the predictive scan. b, Pre-existing but weak fields becoming much stronger after colocalized scanning activity. c, Place cells with a pre-existing field developing an additional field at another location. The pre-existing field sometimes weakens or becomes silent, but not always. d, Narrow fields that broaden counter-clockwise (CCW) around the track after colocalized scanning activity. e, Two simultaneously recorded cells on first exposure to the novel room on the hexagonal track. The top cell forms a new field on lap 3 after a scan on lap 2. Both cells are highly active on a complex scan on lap 4, after which the top cell potentiates and the bottom cell forms a new field. Both fields occur at the same track location. f, Additional examples in two other rats of de novo scan–activation of place fields on first exposure to the novel room.
- Supplementary Figure 7: Simultaneously recorded place cells form fields activated by scans in the novel room. (361 KB)
An ensemble of 5 place cells from the first day of testing in the novel room demonstrates formation of de novo place fields. This is the same ensemble shown in Figure 3d, but here we show the entire recording session (19 laps around the circular track) to provide context for the development of the place fields. We show every place cell for which we detected a potentiation event (out of 13 active place cells recorded during the session), as well as an additional cell (‘t2c6’) that failed to meet the persistence criteria for potentiation (Eqn. 13) but clearly develops spatially selective firing in the second half of the session. Each detected potentiation event was significantly predicted (Supplementary Fig. 8) by colocalized scanning activity on the prior lap. Dashed cyan boxes highlight the predictive scan for each event; dashed magenta boxes highlight the predicted events. Scan (blue), peri-scan (cyan), and forward running (red) spikes are represented with circles (a) or tick marks (b). The session is shown both as series of trajectory plots for each cell (a) as well as raster plots across unwrapped track-angle position from the beginning of the recording session (b). The trajectory plots (a) are vertically compressed to ensure that cell-lap trajectories are individually discernible. Note that two of the cells (‘t2c7’ and ‘t2c6’) both have two distinct scan–potentiation events, where the first event is the initial onset or formation of the field and the second event is a subsequent strengthening and/or broadening of the field. While initial novel-room exposure represents only a minority of our data, these examples (including Supplementary Fig. 6e–f) demonstrate that scan–potentiation occurs for new fields that develop as the animal becomes familiar with a novel environment.
- Supplementary Figure 8: Illustrations of the predictive scan testing window and test procedure. (236 KB)
This figure accompanies Figure 4a and the methodological descriptions in Methods/Scan Prediction Index analysis. a, Every potentiation event is defined by the track-angle positions of the first and last spikes (start/end; solid green lines) of the potentiation traversal and by the center-of-mass (COM; dotted green line) of all spikes in the traversal. b, Minimum (cyan) and maximum (magenta) track-angle windows (Methods, Eqns. 14 and 15) are defined relative to the COM and start/end, respectively. c, We take the outermost bounds of the two windows (asterisks) to form the predictive testing window (blue). d, Finally, the test window can be applied to other laps in the session to test whether the amount of scan activity significantly predicts the potentiation event. e, To test activity on laps (e.g., laps m and n here) for predictive scans, we first check whether the starting track-angle position of any scans fall within the test window on that lap. If not, no further testing is done, and the potentiation event is marked as non-predicted. Otherwise, we test for significantly greater firing during the scans within the window than could be expected by any place-field firing (i.e., firing during forward movement) within the window. We illustrate this process for the within-window scans labeled ‘Scan 1’, ‘Scan 2’, and ‘Scan 3’ on the two laps. First, we determine the time interval corresponding to the test window: the start of the interval is the latest crossing of the leading (CCW) edge of the window, and the end is the earliest crossing of the trailing (CW) edge. The test interval is divided into Δt = 20 ms bins. The time-lines to the right of laps m and n show bin edges (black ticks), spike times (red ticks), and behavioral epochs (blue/cyan highlights: scan/peri-scan epochs). Lap m only has spiking activity during Scan 1, but lap n has both scan and place-field activity. Considering the lap n activity, the test interval is filtered to remove scan and pause behaviors. The result is the ‘Observed testing interval’, which is a set of N time bins with corresponding spike counts ki for each bin i, producing a spike-count distribution (histogram). The firing rate f of the interval is computed as the sum of spike counts across bins divided by the total duration of the bins. Next we construct a bootstrap distribution of firing rates. For 1,000 iterations, we bootstrapped the test interval by randomly sampling with replacement the observed spike-count distribution for N time bins. Several bootstrap intervals are illustrated for example. The firing rates of these bootstrap intervals constitute the expected firing-rate distribution used to test the significance of scan firing. f, We show the cumulative density functions for the lap m (left) and lap n (right) expected firing-rate distributions. The upper bound of the expected 95% c.i. corresponds to the 97.5 percentile (cyan dotted line) and is used as the minimum threshold ( ; red dotted line) for significance of the scan firing (Eqn. 16). Lap m had no place-field activity, thus its expected distribution and significance threshold are both 0 (left). Scan 1 had a non-zero firing rate (blue dashed line) and would thus be counted as a predictive scan for the place-field potentiation event in question. Lap n had expected place-field firing rates centered ~7.5 spikes/s (black curve) and a corresponding threshold ~9.4 spikes/s (right). Scan 2 is below threshold and thus not significant, while Scan 3 is above threshold and would be counted as predictive. The Scan Prediction Index analysis examines the proportion of scan-predicted potentiation events by testing the lap prior to the event.
- Supplementary Figure 9: Procedure for generating expected distributions for the Scan Prediction Index. (416 KB)
This figure is a supplement to Figure 4b–c, Figure 5a–e, Supplementary Figure 10a–d, and the methodological descriptions in Methods/Scan Prediction Index analysis. We used a Monte Carlo randomization and resampling process to generate expected SPI distributions, which is illustrated here for an example DR dataset (rat subject 95, testing day 4). a, We show time-series for each session in the example dataset of radial distance from the track mid-line where scan events are highlighted (blue). Supposing that session 4 (dashed box, dimmed) contains the potentiation event for which we are creating expectations, we collect the start times (relative to the start of the session) and durations for all scan events in the other sessions (i.e., 1, 2, 3, and 5). b, We create a pool of randomized scans by adding each non-event-session scan to the pool 10 times with up to 10 s of uniform random jitter. Any randomized scans that extend beyond the duration of the event session are excluded (‘Session duration cutoff’). The resulting temporal distribution (black curve) can be sampled to generate randomized scan trains that preserve the scanning characteristics for that animal on that testing day. c, Each randomized scan in the pool (arrows: examples) is assigned the spike count from the corresponding scan interval within the event-session spike train (session 4, red circles). Randomized scans are allowed to overlap with actual scan intervals from the event session. d, The randomized scans are tested for predictiveness as usual (Supplementary Fig. 8), with the results illustrated as an array representing the randomized scan pool: scans test positive (orange) or negative (white). We randomly choose the number of scans observed during the event session (blue outlines), which is session 4 in this example. If no positive scans are chosen (as in Sample 1), then the random sample is considered non-predicted; if at least one positive scan is chosen (as in Sample n), then the sample is considered predicted. e, For all potentiation events for which we are computing a baseline, we illustrate the random sampling process as a 3-dimensional matrix (left) with potentiation events, the randomized scan pools, and random samples along the axes. The front face of this matrix shows a single random sample, where positive predictive scans (orange) can be randomly selected (blue outlines) for each potentiation event. Some potentiation events (red outlines) selected a positive scan and are considered to be randomly predicted. For each of 5,000 samples, we collected the set of events that are randomly predicted in a 2-dimensional matrix (middle). The fraction of randomly predicted events in each column is the expected SPI (Eqn. 17) for the corresponding random sample. We use the distribution of these SPIs (right) to calculate the mean and empirical 95% confidence interval (c.i., error bar). The c.i. constitutes the chance expectation for the scan–potentiation effect. We consider the observed predictiveness to be significant if it is greater than the upper bound of the chance expectation.
- Supplementary Figure 10: Marginal and post hoc analysis of the scan–potentiation effect. (196 KB)
This figure is a supplement to Figure 5 which presents the primary results of the SPI analysis. a–d, We show additional SPI data points (red circles) along with the corresponding expected random baselines (black bars: mean and 95% c.i.). a, SPIs across experimental session number (both experiments combined). b–d, Minimum thresholds are applied to select the potentiation events (b) or scans (c–d) that contribute to predictive tests. For these plots, the right panel shows the same data but normalized to the expected chance value. SPIs increase when restricting analysis to later laps, but not relative to expected values (b). While restricting analysis to larger (c) or longer (d) scans reduces overall predictiveness, the effect remains significantly above baseline up to 1.5–2× the expected values. e, Post hoc comparisons of predicted and nonpredicted events. Distributions are displayed as cumulative density functions. We constructed post hoc event distributions by dividing field potentiation events into positively predicted events and nonpredicted events. Predicted events (red; Δp) underwent stronger total potentiation (n = 195 events, 0.982 ± 0.521, mean ± s.d.) than nonpredicted events (black, Δnp; n = 594 events, 0.836 ± 0.442; two-tailed, two-sample Kolmogorov-Smirnov test, D = 0.163, P = 0.000685), where total potentiation was defined as the dot-product field strength of the potentiated traversal minus the maximum field strength of the event’s baseline laps. This difference resulted from a trend toward weaker baselines (thin lines; D = 0.109, P = 0.0549) and significantly stronger potentiated fields (thick lines; D = 0.128, P = 0.0152) for predicted events compared to nonpredicted events. f, Post hoc comparisons of predictive and nonpredictive scans. We constructed post hoc scan distributions by dividing all of the scans from the recording sessions of every potentiation event into those that tested positive for the event, those that tested negative but were within the event testing window, and the remaining non-predictive scans. Predictive scans (solid blue; n = 214) were larger than nonpredictive scans that occurred both within (dashed blue; n = 264 scans; D = 0.149, P = 0.00901) and outside of (black; n = 27,631 scans; D = 0.105, P = 0.0168) the testing window.
- Supplementary Text and Figures (9,241 KB)
Supplementary Figures 1–10 and Supplementary Table 1