Overview of athlete and training studies, and representational similarity analysis.

(a) We examined memory athletes (n = 17) and compared them to matched controls (n = 16) in a single MRI session (athlete study). (b) Individuals in the training study were assigned to one of three groups (memory training group, n = 17, active controls, n = 16, passive controls, n = 17). They completed two MRI sessions prior and after the 6-week-long interval (pre- and post-training sessions), and a behavioural re-test after 4 months. (c) General structure of all MRI sessions: baseline and post-task rest periods (not discussed further), word list encoding and temporal order recognition tasks (10 min each), immediate (20 min post-MRI) and delayed (24 hours post-MRI) free recall (dashed frame indicates that the delayed test was only included in the training study). (d) We conducted three types of representational similarity analysis (RSA). Between-subject RSA: this assessed the neural pattern similarity between individuals within the same experimental group. Trial-specific RSA: this examined neural pattern similarity across the studied words, but within an individual. Trainee-athlete RSA: this evaluated the degree of neural pattern similarity between trainees and memory athletes. Note that the schematic shows memory athletes and individuals of the memory training group (the “trainees”) only, but the same analysis was performed for all remaining groups as well.

Method of loci training yields distinct neural representations between experienced individuals during word list encoding.

(a) Detailed schematic of the between-subject RSA: We compared the individual-to-group neural pattern similarity between memory athletes (n = 17) and matched controls (n = 16) during word list encoding (relative to the implicit fixation cross baseline), and between pre- and post-training within the memory training group (n = 17, see Methods, and see supplementary materials for control analyses). (b) Schematic of the word list encoding task. A trial was defined as the presentation of one word. The 72 trials were organised into 12 blocks of 6 words each. A block began with an instruction screen presented for 5 seconds, followed by the 6 words, each presented for 3 seconds and separated by a jittered fixation cross of 2-5 seconds. After every block, a fixation cross was displayed for 25 seconds. (c) Regions of interest (ROIs) that were used for all analyses pertaining to word list encoding. (d, f) ROI results of the between-subject RSA (control analyses for the active and passive control groups are shown in Fig. S1). (e, g) Searchlight results of the between-subject RSA. Results were thresholded at p < 0.05, family-wise error (FWE)-corrected at cluster-level, using a cluster-defining threshold of p < 0.001 (Tables S2-3). (h) Detailed schematic of the trial-specific RSA: We calculated neural pattern similarity on a trial-by-trial basis (relative to the implicit fixation cross baseline) for each participant and compared memory athletes to matched controls, and pre- to post-training time points in the memory training group (same n as above). (i, j) ROI results of the trial-specific RSA (control analyses for the active and passive control groups are shown in Fig. S3). (d, f, i, j) Data points represent individual participants (athlete study, n = 33, memory training group, n = 17). Boxplots display the median and upper/lower quartiles, with whiskers extending to the most extreme data points within 1.5 interquartile ranges above/below the quartiles. The data points with error bars show the mean ± standard error of the mean (S.E.M.). Data distributions are based on the respective probability density function. All ROI-based results survived multiple comparison corrections using the false discovery rate (FDR) with the Benjamini-Hochberg procedure set to q < 0.05. ***p < 0.001; *p < 0.05; ns, not significant. Pattern similarity reflects Fisher z-transformed Pearson correlations (r).

Method of loci training yields distinct neural representations between trainees and memory athletes during word list encoding.

(a) Detailed schematic of the trainee-athlete RSA: We calculated the trainee-to-athletes (“trainees” are individuals of the memory training group) and trainee-to-controls neural pattern similarity during word list encoding (relative to the implicit fixation cross baseline) and compared between pre-(red) and post-training (blue) time points (see Methods). (b) ROI results (control analyses for the active and passive control groups are shown in Fig. S4). Data points represent individual participants (memory training group, n = 17). Boxplots display the median and upper/lower quartiles, with whiskers extending to the most extreme data points within 1.5 interquartile ranges above/below the quartiles. The data points with error bars show the mean ± S.E.M. Data distributions are based on the respective probability density function. (c) Searchlight results for the time point × group interaction. Results were thresholded at p < 0.05, FWE-corrected at cluster-level, using a cluster-defining threshold of p < 0.001 (Table S4). (d) The scatter plot depicts the correlation between memory performance at the 4-month reset (4 months > immediate free recall test during the pre-training session) and the predicted changes in neural pattern similarity between all individuals of the training study and memory athletes from pre- to post-training (n = 44). ***p < 0.001; **p< 0.01; ns, not significant. Pattern similarity reflects Fisher z-transformed Pearson correlations (r).

Method of loci training yields shared neural representations between experienced individuals during temporal order recognition.

(a) Schematic of the temporal order recognition task. Participants viewed 24 triplets taken from the previously encoded word list. During recognition trials, participants judged whether the order of words in each triplet matched their original presentation during word list encoding. Main task trials were interspersed with control trials (in an alternating ABAB sequence), during which triplets of novel words were presented. Participants were asked to indicate whether these words were arranged in ascending or descending order based on the number of syllables (syllable-counting task). All trials began with a cue that signalled the start of the trial (3 s), followed by the triplet presentation (9 s), and a response screen (3 s). (b) ROIs that were used for all analyses pertaining to temporal order recognition (all analyses were performed relative to the control task baseline). (c, e) ROI results of the between-subject RSA (control analyses for the active and passive control groups are shown in Fig. S5). (d, f) Searchlight results of the between-subject RSA. Results were thresholded at p < 0.05, FWE-corrected at cluster-level, using a cluster-defining threshold of p < 0.001 (Tables S5-6). (g, h) ROI results of the trial-specific RSA pertaining to the (g) athlete study and (h) the training study (control analyses for the active and passive control groups are shown in Fig. S7). (c, e, g, h) Data points represent individual participants (athlete study, n = 33, memory training group, n = 17). Boxplots display the median and upper/lower quartiles, with whiskers extending to the most extreme data points within 1.5 interquartile ranges above/below the quartiles. The data points with error bars show the mean ± standard error of the mean (S.E.M.). Data distributions are based on the respective probability density function. All ROI-based results survived multiple comparison corrections using the false discovery rate (FDR) with the Benjamini-Hochberg procedure set to q < 0.05. ***p < 0.001; ns, not significant. Pattern similarity reflects Fisher z-transformed Pearson correlations (r).

Method of loci training yields shared neural representations between trainees and memory athletes during temporal order recognition.

(a) ROI results of the trainee-athlete RSA (control analyses for the active and passive control groups are shown in Fig. S8; analysis was performed relative to the control task baseline). Data points represent individual participants (memory training group, n = 17). Boxplots display the median and upper/lower quartiles, with whiskers extending to the most extreme data points within 1.5 interquartile ranges above/below the quartiles. The data points with error bars show the mean ± S.E.M. Data distributions are based on the respective probability density function. (b) Searchlight results for the time point × group interaction. Results were thresholded at p < 0.05, FWE-corrected at cluster-level, using a cluster-defining threshold of p < 0.001 (Table S7). (c). The scatter plot depicts the correlation between memory performance at the 4-months reset (4 months > immediate free recall test during the pre-training session) and the predicted changes in neural pattern similarity between all individuals of the training study and athletes from post- to pre-training (n = 44, see Methods). **p< 0.01; ns, not significant. Pattern similarity reflects Fisher z-transformed Pearson correlations (r).

Results for between-subject ROI-based RSA during word list encoding, active and passive control groups.

Regions-of-interest (ROIs) used in the between-subject RSA during word list encoding (i.e., active controls × active controls, passive controls × passive controls, analyzed separately for pre- and post-training sessions) included the (a) lateral prefrontal cortex and (b) superior frontal gyrus. Figure legend: Data po ints represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci leading to distinct neural representation among individuals is stronger than active and passive control groups.

Searchlight RSA of the increased between-subject distinct neural representation (pre-minus post-training) in the memory training group after training during the word list encoding task, compared to (a) active or (b) passive control. Results are reported at p < 0.05, family-wise error (FWE)-corrected at the cluster level (cluster-defining threshold p < 0.001). Colour bar indicates t-values.

Results for trial-specific ROI-based RSA during word list encoding, active and passive control groups.

Regions-of-interest (ROIs) used in the trial-specific RSA during word list encoding (i.e., active controls x active controls, passive controls × passive controls, analyzed separately for pre- and post-training sessions) included the (a) lateral prefrontal cortex and (b) superior frontal gyrus. Figure legend: Data points represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Results for trainee-athlete ROI-based RSA during word list encoding, active and passive control groups.

Regions-of-interest (ROIs) are used in the trainee-athlete RSA during word list encoding, analyzed with repeat measures ANOVA of 2 × 2 factors: time point (pre-vs. post-training) and group (neural pattern similarity between active/passive controls × athletes vs. active/passive controls × matched controls). Figure legend: Data points represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Results for between-subject ROI-based RSA during temporal order recognition, active and passive control groups.

Regions-of-interest (ROIs) used in the between-subject RSA during temporal order recognition (i.e., active controls × active controls, passive controls × passive controls, analyzed separately for pre- and post-training sessions) included the (a) hippocampus and (b) precuneus. Figure legend: Data points represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant, * p<0.05, **p<0.01. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci increased pattern similarity between individuals stronger than active and passive control groups during temporal order recognition.

(a, b) control analysis, showing whole-brain searchlight of increased pattern similarity between individuals (post-minus pre-training) in the memory training group after training, compared to (B) active or (C) passive control. Whole-brain results are reported at p < 0.05, family-wise error (FWE)-corrected at the cluster level (cluster-defining threshold p < 0.001). Colour bar indicates t-values.

Results for trial-specific ROI-based RSA during temporal order recognition, active and passive control groups.

Regions-of-interest (ROIs) used in the trial-specific RSA during temporal order recognition (i.e., active controls × active controls, passive controls × passive controls, analyzed separately for pre- and post-training sessions) included the (a) hippocampus and (b) precuneus. Figure legend: Data points represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Results for trainee-athlete ROI-based RSA during temporal order recognition, active and passive control groups.

Regions-of-interest (ROIs) are used in the between-group RSA during temporal order recognition, analyzed with repeat measures ANOVA of 2 × 2 factors: time points (pre-vs. post-training) and group (neural pattern similarity between active/passive controls × athletes vs. active/passive controls × matched controls). Figure legend: Data points represent individual participants (active controls, n = 16; passive controls, n = 17); boxplots display the median and upper/lower quartiles, whiskers show 1.5 interquartile ranges above/below the quartiles; circles with error bars correspond to the mean ± standard error of the mean (S.E.M.); distributions illustrate the probability density function of individual data points. Abbreviations: ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Gradient analysis of the lateral prefrontal cortex and hippocampus.

Anatomical masks of (a) left lateral prefrontal cortex (LPFC) subregions based on dorsal (red), middle (green), and ventral (blue) partitions, and (b) bilateral hippocampus (HIP) subregions based on anterior (red), middle (red), and posterior (blue) partitions are shown on a template brain.

Method of loci training yields distinct neural representation between experienced individuals along the dorsal-ventral axis of the left lateral prefrontal cortex.

(a, b) ROI results of the between-subject RSA for the memory athlete study and memory training group, and (c, d) analyses for the active and passive control groups. Data points represent individual participants in all panels (athlete study: n = 33; memory training group: n = 17; active control group: n = 16; passive control group: n = 17). ***p < 0.001, **p < 0.01; *p < 0.017; results remained significant after false discovery rate (FDR) correction using the Benjamini– Hochberg procedure. ∼ denotes results that did not pass the correction but that exhibited a trend toward significance. ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training yields distinct neural representations between trials along the dorsal-to-ventral axis of the left lateral prefrontal cortex.

(a, b) ROI results of the trial-specific RSA for the memory athlete study and memory training group, and (c, d) analyses for the active and passive control groups. **p < 0.01; results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. ∼ denotes results that did not pass the correction but that exhibited a trend toward significance. ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training yields shared neural representations between experienced individuals along the posterior-to-anterior axis of the bilateral hippocampus.

(a, b) ROI results of the between-subject RSA for the memory athlete study and memory training group, and (c, d) analyses for the active and passive control groups. * p < 0.05, **p < 0.01, ns, not significant; results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training yields shared neural representations between trials along the posterior-to-anterior axis of the bilateral hippocampus.

(a, b) ROI results of the trial-specific RSA for the memory athlete study and memory training group, and (c, d) analyses for the active and passive control groups. **p < 0.01, ns, not significant; results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. ∼ denotes results that did not pass the correction but that exhibited a trend toward significance. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training yields shared neural representations between trainees and athletes in the bilateral anterior hippocampus during temporal order recognition.

ROI results of the comparison between trainees and memory athletes pertaining to the anterior hippocampus for (a) memory training group, (b) active control group, and (c) passive control group. Data points represent individual participants (athlete study, n = 33, memory training group, n = 17; active control group, n = 16; passive control group, n = 17). Results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. ∼ denotes results that did not pass the correction but that exhibited a trend toward significance. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training yields shared neural representations in the medial temporal lobe between experienced individuals during word list encoding.

ROI results of the between-subject RSA pertaining to the (a) athlete study and (b) the training study, and analyses for (c, d) the active and passive control groups. Data points represent individual participants (athlete study, n = 33, memory training group, n = 17, active control group, n = 16, passive control group, n = 17). **p < 0.01; Results remained significant after false discovery rate (FDR) correction using the Benjamini–Hochberg procedure. ∼ denotes results that exhibited a trend toward significance. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci training does not induce shared neural representations between encoded items in the medial temporal lobe during word list encoding.

ROI results of the trial-specific RSA pertaining to the (a) athlete study and (b) the training study, and analyses for (c, d) the active and passive control groups. Data points represent individual participants (athlete study, n = 33, memory training group, n = 17, active control group, n = 16, passive control group, n = 17). ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

No significant effect of method of loci training on neural representations in the medial temporal lobe between trainees and memory athletes during word list encoding.

ROI results of the comparison between trainees and memory athletes pertaining to the hippocampus and parahippocampal gyrus for (a) memory training group, (b) active control group, and (c) passive control group. Data points represent individual participants (athlete study, n = 33, memory training group, n = 17; active control group, n = 16; passive control group, n = 17). ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Training-induced increase in neural similarity to memory athletes was negatively associated with memory performance in the memory training group.

(a) Voxel-wise multiple regression analysis showing contrast difference maps (Post > Pre), with individual memory improvement (number of words freely recalled at the 4-months retest minus 20-minute free recall during the pre-training session) included as a covariate of interest. A significant cluster was observed specifically in the memory training group, located in the left lateral prefrontal cortex (peak MNI coordinates: x = −34, y = 26, z = 36; cluster size = 30 voxels; p < 0.05, FWE-corrected at the cluster level), using a cluster-defining threshold of p < 0.001 and a minimum cluster size of 27 voxels. (b-d) Scatter plots illustrating the correlation between training-induced increases in neural similarity to memory athletes and memory improvement per group, extracted from the lateral prefrontal cortex cluster shown in (a). Plots are shown for visualization purposes only. memory training group, n = 14, active control group, n = 14, passive control group, n = 16. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Trial-specific encoding-retrieval similarity analyses.

(a, b) Neural pattern similarity between encoding and retrieval activity patterns for paired and unpaired trials. (c, d) Neural pattern similarity of paired trials across pre- and post-training sessions in the memory training group and across groups in the athlete study (memory athletes vs. matched controls). (a-d) data points represent individual participants (memory athlete n = 17; matched control n = 16; memory training group: n = 17). ns, not significant. Pattern similarity reflects Fisher’s z-transformed Pearson’s correlations (r).

Method of loci enhances sequence-geometry of neural patterns during temporal order recognition.

(a-c) Workflow for sequence-geometry RSA. (a) Neural model construction with trial-by-trial pattern similarity based on each trials’ position in the sequence (Sub, subject). (b) Theoretical sequence-geometry model with reciprocal distance. (c) Illustration of the reciprocal distance function with decreasing similarity at increasing trial distance. (d) Neural pattern similarity (hippocampus) as a function of trial distance in the memory training group after training during word list encoding (left) and temporal order recognition (right). (e) Results of the sequence-geometry analysis pertaining to word list encoding (left) and temporal order recognition (right). Error bars show mean ± standard error of the mean (SEM). Sample sizes: memory training group, n = 17; active controls, n = 16; passive controls, n = 17. Reported p-values are uncorrected for multiple comparisons; none of the effects survived false discovery rate (FDR) correction (Benjamini–Hochberg). ROIs: HIP, hippocampus; anterior, middle, and posterior HIP subdivisions; Precun, precuneus; PHG, parahippocampal gyrus.

Method of loci enhances trial-wise neural pattern similarity between experienced individuals during temporal order recognition.

(a-c) Workflow for trial-wise Intersubject Similarity (ISC) RSA. (a) Trial-wise beta estimates for all trials within a given ROI. (b) For each trial, subject, and ROI, we constructed a voxels-by-subject matrix and computed the correlation between each subject’s neural pattern with the average of all other participants, and Fisher’s z-transformed it, yielding a trial-level ISC value. (c) Trial-level ISC values were averaged across trials to produce one ISC estimate per subject and ROI. (d) Results of the trial-wise ISC analysis pertaining to word list encoding (left) and temporal order recognition (right). Error bars show mean ± standard error of the mean (SEM). Reported p-values are uncorrected for multiple comparisons; *** indicates results that remained significant after false discovery rate (FDR) correction (Benjamini-Hochberg procedure). Error bars represent mean ± SEM. Sample sizes: memory athletes, n = 17; matched controls, n = 16. ROIs: HIP, hippocampus; anterior HIP, anterior hippocampus; middle HIP, middle hippocampus; posterior HIP, posterior hippocampus; Precun, precuneus; PHG, parahippocampal gyrus.

Demographic details.

Note that these data were presented previously (Dresler et al., 2017; Wagner et al., 2021).

Searchlight RSA, between-subject, athlete study, word list encoding.

Analysis comprised an independent t-test comparing memory athletes (n = 17) to matched controls (n = 16). Contrast: memory athletes < matched controls during the word list encoding > fixation baseline. Critical cluster size: 36 voxels.

Searchlight RSA, between-subject, memory training study, word list encoding.

Analysis comprised an independent t-test comparing the memory training group (n = 17) pre- to post-training. Contrast: post-training < pre-training during the word list encoding > fixation baseline. Critical cluster size: 30 voxels.

Searchlight RSA, trainee-athlete, memory training study, word list encoding.

The analysis used a repeated-measures ANOVA to compare neural pattern similarity between the memory training group (n = 17) and the memory athlete study, examining pre- and post-training effects across two factors: time point (pre-vs. post-training) and group (neural pattern similarity between trainees × athletes vs. trainees × matched controls). The interaction contrasts during word list encoding. Critical cluster size: 38 voxels.

Searchlight RSA, between-subject, athlete study, temporal order recognition.

Analysis comprised an independent t-test comparing memory athletes (n = 17) to matched controls (n = 16). Contrast: memory athletes > matched controls during the temporal order recognition > fixation baseline. Critical cluster size: 35 voxels.

Searchlight RSA, between-subject, memory training study, temporal order recognition.

Analysis comprised an independent t-test comparing the memory training group (n = 17) pre- to post-training. Contrast: post-training > pre-training during the temporal order recognition> fixation baseline. Critical cluster size: 31 voxels.

Searchlight RSA, trainee-athlete, memory training study, temporal order recognition.

The analysis used a repeated-measures ANOVA to compare neural pattern similarity between the memory training group (n = 17) and the memory athlete study, examining pre- and post-training effects across two factors: time point (pre-vs. post-training) and group (neural pattern similarity between trainees × athletes vs. trainees × matched controls). The interaction contrasts during temporal order recognition. Critical cluster size: 37 voxels.

Memory performance of pre-and post-training for training study.

Values represent the average number of freely recalled words ± standard deviation (SD). Note that these data were presented previously (Dresler et al., 2017; Wagner et al., 2021).

Training-related changes in neural representations are associated with better long-term memory performance after 24 hours and 4 months.

Correlations of observed memory performance improved (i.e., assessed in 24 hours (n = 49), and 4 months (n = 44)), with predicted memory outcome improved from machine learning prediction analysis based on ROIs (details see Methods), both on word list encoding and temporal order recognition tasks. Notes: ∼p < 0.10; *p<0.05; **p<0.01; ***p<0.001; ns not significant; n represents the sample size in corresponding calculation.

Summary of the results from the LPFC gradient analysis.

All results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. ∼ denotes results that did not pass the correction but exhibited a trend toward significance. ↑ indicates an increase in neural pattern similarity, ↓ indicates a decrease in neural pattern similarity. **p < 0.01, *p < 0.017.

Summary of the results for hippocampus gradient analysis.

All results remained significant after false discovery rate (FDR) correction using the Benjamini-Hochberg procedure. ∼ denotes results that did not pass the correction but exhibited a trend toward significance. ↑ indicates an increase in neural pattern similarity, ↓ indicates a decrease in neural pattern similarity. **p < 0.01, ***p < 0.001.

Per-group correlations with memory performance.

Regions-of-interest (ROIs): LPFC, lateral prefrontal cortex, SFG, superior frontal gyrus, HIP, hippocampus, Prec, Precuneus, neural pattern similarity extracted during word list encoding (LPFC, SFG) and temporal order recognition (HIP, Prec) tasks (see also Methods section). Pearson correlation (r) values are given for per-group correlations between neural pattern similarity and memory performance at the 4 months retest (4 months minus free recall performance at the 20-min test during the pre-training, PRE, session), along with p-values.

Recognition performance (d-prime) per group.

Performance values obtained from the temporal order recognition task during the pre- and post-training sessions, per group, showing the means (M) and standard errors of the mean (SEM).

Word lists used during the word list encoding task.