Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2026 Mar 12:2026.01.11.698919.
doi: 10.64898/2026.01.11.698919.

Brain organization of a memory champion

Affiliations

Brain organization of a memory champion

Roselyne J Chauvin et al. bioRxiv. .

Abstract

Memory athletes can achieve superior performance (e.g., memorizing 339 digits in 5 minutes) with extensive daily training, by converting abstract information into vivid scenes, and placing them along a mental path, that is later retraced (Method of Loci). Understanding the brain mechanisms underlying such training-derived mastery would increase our understanding of the brain's memory systems and could suggest novel approaches to improving cognition in other domains. As memory athletes use personalized training techniques, it has been challenging to study them with standard group paradigms. Fortunately, precision functional mapping (PFM) enables detailed investigation of individual brains through repeated sampling of resting-state functional connectivity and task fMRI. Here, we precisely mapped the brain organization of a 6-time U.S. Memory Champion (>13 hours fMRI). Relative to controls, the Memory Champion's network functional connectivity (FC) was strengthened with the retrosplenial, extrastriate visual, and dorsal frontal cortex (area 55b), as well as with the caudate nucleus. The Memory Champion had modules related to scene and semantic processing not seen in controls, alongside stronger connectivity between the caudate and classical memory networks. During rote memorization, the Champion's task fMRI patterns were typical, with the hippocampus active during encoding. This pattern was reversed when he used his Method of Loci technique, with greater hippocampal activity during recall than encoding. Hence, intense practice at converting abstract information into more memorable formats can develop a procedural memory skill that utilizes brain regions typically reserved for navigation, language, social cognition, and associative learning.

PubMed Disclaimer

Conflict of interest statement

E.M.G. may receive royalty income based on technology developed at Washington University School of Medicine and licensed to Turing Medical Inc. N.U.F.D. has a financial interest in Turing Medical Inc. and may benefit financially if the company is successful in marketing Framewise Integrated Real-Time Motion Monitoring (FIRMM) software products. N.U.F.D. may receive royalty income based on FIRMM technology developed at Washington University School of Medicine and Oregon Health and Sciences University and licensed to Turing Medical Inc. N.U.F.D. is a co-founder of Turing Medical Inc. TOL is a consultant for Turing Medical Inc. TOL holds a patent for taskless mapping of brain activity licensed to Sora Neurosciences and a patent for optimizing targets for neuromodulation, implant localization, and ablation is pending. These potential conflicts of interest have been reviewed and are managed by Washington University School of Medicine. The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Figures

Figure 1:
Figure 1:. Brain of a Memory Champion: Memorization technique, performance and functional connectivity.
a) Deck of cards memorization strategy used by the 6x US Memory Champion. The strategy involves a pre-encoded association system where each card is linked to three items (person, action, object), combined into a scene (‘Arnold Schwarzenegger (ace of spades) screaming at (ace of hearts) basketballs (two of diamonds)’) represented at a specific location within one of the Memory Champion’s memory palace loci, e.g. the bedroom of his house. b) 6x US Memory Champion’s US records over the years. The number of recalled items for three official US records: digits in 5 minutes, words in 15 minutes, and names and faces in 15 minutes plotted as a function of his age. c) Memory Champion’s brain regions with higher network centrality than the population. Regions with higher functional connectivity in the Memory Champion than 95% of individuals from the HCP population (n = 887), at both visits (2015, 2021), with one or more of the 17 functional networks (see Supplementary Fig. 1a) after mixture modeling normalization of each network FC map (see Methods). Darker brown colors indicate stronger FC with a higher number of functional networks (from 1 to 6), indicating higher centrality. Boundaries for the language (teal), lateral visual (dark blue), primary visual (ochre), and context-association (CAN, white) networks are shown in color.
Figure 2:
Figure 2:. Functional connectivity of regions with high centrality in the Memory Champion.
Whole-brain FC maps for representative brain regions identified in Fig. 1c, top 20% of strongest connections are shown. Left column shows the Memory Champion, middle column MSC02 control, and right column MSC06 control (2021 data; see Extended Data Fig. 2 for 2015 data, and Supplementary Fig. 3 for additional brain views). The top row shows correlation (r) for retrosplenial cortex (Brodmann areas 26, 29, and 30, part of context-association network -white border), middle row shows the left semantic region (area 55b; language network -teal border) and the bottom row shows the head of the caudate. The seed regions are shown in black. All other network borders indicate networks significantly more connected in the Memory Champion compared to the controls in 2021 (see Extended Data Fig. 2 for quantification). SCAN stands for somato-cognitive action network. Low signal-to-noise masked areas are displayed in dark grey.
Figure 3:
Figure 3:. Functional modules only found in the Memory Champion.
a) Functional network annotation of the Memory Champion’s high centrality regions. The high centrality regions from Fig. 1c, are shown on an inflated brain, colored by their functional network assignments. Regions are grouped into two sets of networks (network color label, bottom left) representing a scene (dark blue) and semantic (teal) module. Each region is named anatomically or functionally in dark blue for the scene module and teal for the semantic module. Abbreviations: parietal occipital (par. occ.), intra parietal sulcus (IPS), somato-cognitive action network (SCAN), Brodmann Area (BA). b) Top 15 meta-analytic task fMRI database terms (Neurosynth) associated with the Memory Champion’s high centrality regions. The font size of the word in the word clouds matches the relative weight of the term amongst the top 15.
Figure 4:
Figure 4:. Memory networks to subcortex functional connectivity.
a) Subcortical regions where the Memory Champion showed stronger FC than 95% of controls (HCP, n = 887), by network: top row parietal memory network (blue), second row default-mode network (red) and third row context-association network (white) FC maps in subcortex. The overlap across networks is shown in the fourth row, with pink indicating greater functional connectivity to all three memory networks. Results are displayed on the Memory Champion’s horizontal T1 structural slices with the network labelled color as gradient scale (right). White borders outline the Freesurfer segmented anatomical structures. See Extended Data Fig. 5; Supplementary Fig. 5 for replication in 2015 data. b) Subcortical functional connectivity of memory networks in controls (HCP). Left to right, row parietal memory network (blue), default-mode network (red) and context-association network (white) Z scored FC maps for regions where networks are competing for winner-take-all labels (i.e. network territories). Results are displayed on the MNI horizontal T1 structural slices. White borders outline anatomical structures. (See Extended Data Fig. 6 for winner-take-all maps, Extended Data Fig. 7 for additional slices of network territories in controls and quantification of pattern overlap with the stronger cortico-subcortical FC area in the Memory Champion. See Supplementary Fig. 6 for subcortical network territories of the Memory Champion).
Figure 5:
Figure 5:. Task activations during encoding and recall when using rote memorization or Method of Loci.
a) Reading span and Deck of cards memory task schemas for encoding and recall (proactive). Schema shows the encoding and proactive recall phases of the reading span task (top line) and the Deck of cards memory task (bottom line). Full task schemes are available in Supplementary Figs. 7 and 8. b) Task fMRI contrast of Encoding versus Recall, when using rote memorization in the Reading span task. Results are shown on each participant’s T1 horizontal slice (z = −15) for unthresholded Z-statistic maps of Encode > Recall (left) and Recall > Encode (right), for controls (MSC02, MSC06) and the Memory Champion (2021 data). White borders outline anatomical structures (Freesurfer segmentation). For additional slices and cortical results see Extended Data Fig. 8. c) Task fMRI activation of Encode versus Recall contrasts when the Memory Champion is using the Method of Loci strategy in the Deck of cards task. Controls are unable to perform this task. Results are shown on the Memory Champion’s T1 horizontal slices (z = 15, 5, −15) for unthresholded Z-statistic maps of Encode > Recall (left) and Recall > Encode (right), with white borders of anatomical structures. For cortical results see Extended Data Fig. 9.
Figure 6:
Figure 6:. Task fMRI activation when recalling digits of the number pi using Method of Loci.
a) Task schema for recalling the location of digit sequences within the number pi. Mental navigation (search; in blue) and retrieval (found; in red) of the locus where a prompted unique 5-digits sequence from the pre-learned 10,000 first digits of the number pi is stored. Contrasted events are ‘search’ (failed to find within 6 seconds) and ‘found’ (localized within 6 seconds) events. (see Supplementary Fig. 11 for full task schema). b) Memory Champion task activation contrast of Search > Found recall of 5-digit sequence locations within the 10,000 first digits of the number pi. Context-association network borders (white outlines) are displayed. c) Search > Found activations quantified by functional network. Average Z-statistic bar plot (*, P < 0.05, FDR-corrected). d) Memory Champion task activation contrast of Found > Search recall of 5-digit sequence locations within the 10,000 first digits of the number pi. Memory Champion high centrality region borders (black outlines) are displayed. e) Average Z statistics per high centrality regions grouped by scene and semantic modules and subcortical regions. Low signal-to-noise masked areas are displayed in grey. See additional views, network and region bar plots in Extended data Fig.10

References

    1. Forsberg A., Adams E. J. & Cowan N. Chapter One - The role of working memory in long-term learning: Implications for childhood development. in Psychology of Learning and Motivation (ed. Federmeier K. D.) vol. 74 1–45 (Academic Press, 2021).
    1. Miller G. A. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63, 81–97 (1956). - PubMed
    1. Cowan N. The Magical Mystery Four: How is Working Memory Capacity Limited, and Why? Curr Dir Psychol Sci 19, 51–57 (2010). - PMC - PubMed
    1. Dellis N. Memory Superpowers!: An Adventurous Guide to Remembering What You Don’t Want To Forget. (Abrams Books for Young Readers, 2020).
    1. Squire L. R. The Legacy of Patient H.M. for Neuroscience. Neuron 61, 6–9 (2009). - PMC - PubMed

Publication types

LinkOut - more resources