Memory variability, not memory level, tracks cingulum microstructure — but free water remains unmodelled

Question: How does within-person variability in memory performance relate to white matter microstructure (fractional anisotropy, free-water) along hippocampal-cingulum and fornix pathways in non-demented older adults?

Answer

The retrieved literature supports a compartment-specific dissociation rather than a general "white matter integrity supports memory" story: within-person variability in memory performance is associated with microstructure of the hippocampal-cingulum pathway while mean memory performance is associated with hippocampal volume, and the two predictors do not cross over [1]. Across these papers, this implies that variability indexes a distinct neural property — axonal/tract-level tissue organisation — rather than being a noisy proxy for low ability. However, the same papers make clear that this conclusion is provisional on two methodological points: FA in aged limbic tracts is heavily contaminated by extracellular free water, and free-water elimination materially changes both reliability and inference in older cohorts [9]; and no retrieved study has yet tested the variability–fornix relationship with a free-water-corrected metric in non-demented older adults. The fornix literature demonstrates behaviourally meaningful microstructure–memory associations, but for mean-level learning, retrieval speed and confidence rather than for variability [11] [14].

Memory variability and mean memory level map onto different tissue compartments

The only retrieved study to directly test this question found that higher memory intraindividual variability was associated with lower fractional anisotropy in the hippocampus-cingulum tract (β = −0.34, p < .01), whereas mean memory performance was not associated with that tract's FA (β = −0.07, p = .52) [1]. In the same 67-participant sample, higher mean performance was associated with greater hippocampal volume (β = 0.24, p = .03), while memory variability showed no volumetric association (β = 0.028, p = .84) [1]. The authors conclude that memory variability may be more closely linked to hippocampal white matter while mean performance better reflects hippocampal macrostructure [1].

This double dissociation is the strongest available argument that variability is not merely a psychometric artefact of lower ability: if it were, variability and mean performance would load on the same structural substrate [1]. Variability also carries independent clinical information — within-person variability, long-term forgetting and learning rate uniquely signified MCI irrespective of age in high-frequency smartphone assessment of 316 older adults [2].

FA in aged limbic tracts is not interpretable without free-water modelling

Free water accumulates in ageing white matter, and computational free-water elimination improved tractometry reliability by up to roughly 37% at every pipeline stage — fibre orientation estimation, tract delineation, and along-tract tissue property assessment — in 339 older adults aged 66–103 [9]. The same study showed that free-water elimination produced more accurate cross-validated predictions of clinician Fazekas scores, establishing that unmodelled free water degrades not just precision but validity of inference in aged brains [9].

This is not a peripheral concern for the cingulum finding, because older adults show simultaneously lower global FA and higher free water fraction alongside lower neurite density and higher orientation dispersion [4]. A lower cingulum FA in a high-variability older adult could therefore reflect increased extracellular water rather than reduced axonal coherence — the two interpretations that the question explicitly asks to separate [1] [4].

Free-water-corrected limbic metrics are feasible and biologically anchored: in 2,614 older adults across seven harmonised cohorts, free-water-corrected FA, axial, mean and radial diffusivity of the cingulum and fornix showed significant heritability (0.26–0.60 for 15 of 35 tract-by-metric combinations) [10]. This establishes cingulum and fornix free-water-corrected microstructure as a stable, genetically influenced individual-difference phenotype rather than measurement noise [10].

Fornix microstructure predicts memory level and speed, not variability

Greater fornix FA and younger age together explained 34% of the variance in memory retrieval speed, and higher fibre density and cross-section in posterior callosal and cingulum pathways accompanied slower responses [14]. Using free-water-corrected FA, inter-individual differences in pre-commissural — but not post-commissural — fornix microstructure correlated with spatial learning across repeated navigation attempts in healthy adults [11]. Fornix FA, specifically in the right mid-tract crus/body, was associated with holistic recollection in children, with no cluster-corrected associations for mnemonic discrimination or relational binding [8]. Fornix microstructure also negatively predicted confidence in internal-memory performance, and cingulum bundle FA predicted deviations from optimal reminder use, indicating tract-behaviour specificity extends beyond memory accuracy [3].

None of these studies operationalised within-person variability; all modelled central tendency [14] [11] [8]. The fornix–variability link therefore remains an inference by analogy from the cingulum finding, not an observed result [1].

Measurement precision of the cingulum-hippocampus segment bounds detectable effects

In a travelling-head tractometry study of 20 major tracts, the minimum detectable change differed by tract, with the cingulum hippocampus showing worse precision than the optic radiation, and intraclass correlations fell as scanner model and protocol diverged [5]. Fornix findings are additionally algorithm-dependent: probabilistic tractography produced stronger and more spatially extended effects than deterministic tracking in the fornix, a small and highly curved limbic pathway [12].

Consequently, a small but genuine correlation between free-water-corrected cingulum FA and memory variability could be undetectable in typical single-site samples, and the one positive report rests on 67 participants in a single cohort [1] [5].

Conflicts and gaps

Whether limbic tracts are the memory-relevant substrate in old age is contested. Age moderated white-matter–cognition associations tract-specifically such that younger adults showed stronger coupling of commissural and association tract microstructure with visual memory, while older adults showed greater dependence on projection and limbic tracts for executive and language function — not for memory [4]. This directly complicates the premise that hippocampal-cingulum microstructure is the aged memory substrate.

The direction of free-water effects is also unsettled. Free-water diffusion measures showed no direct effect on cognitive scores across the lifespan, with age instead modifying the free-water–cognition relationship [6]. In a paediatric cancer cohort, reduced extracellular free water and dispersion were linked to better cognition in patients while healthy controls showed mixed positive and negative associations across four of five metrics [7]. Whether higher free water in the cingulum should predict greater memory variability is therefore not established by any retrieved paper.

Three specific gaps remain: no study has jointly modelled mean performance and variability against free-water-corrected cingulum and fornix metrics; no study has estimated the variability–microstructure association longitudinally, leaving the direction of effect untested [1]; and hippocampal pathway variability has been linked to cognitive ability in a large healthy sample (N = 831) without decomposing free water or variability [13].

What would change the answer

A study that measures memory repeatedly across sessions (rather than relying on within-occasion dispersion across subtests) and estimates variability with hierarchical location-scale models — the approach that isolated variability as an MCI-specific marker [2] — would test whether the cingulum association survives once occasion-specific measurement error is separated from trait variability [1]. Fitting bi-tensor free-water models alongside FW-corrected FA, as done for limbic tracts in harmonised ageing cohorts [10] and for tractometry reliability in old age [9], and adding NODDI free water fraction and orientation dispersion to distinguish extracellular water from axonal coherence [4], would adjudicate whether the cingulum–variability effect is axonal or interstitial. Along-tract rather than tract-averaged sampling would localise the effect and mitigate the precision limits of the cingulum-hippocampus segment [5], and probabilistic reconstruction with cross-algorithm validation would stabilise fornix estimates [12]. Finally, a formal mediation model testing whether free-water-corrected cingulum microstructure mediates an age→variability pathway, with sample size powered to the tract's minimum detectable change, is the design most likely to overturn or confirm the current dissociation.

References

[1] Gonzalez, A. I., Tristao-Pereira, C., Martinez, L., Ramirez, N. J., Aquilino Silva, G., Alcina, J., Martinez, J. E., Giudicessi, A., Ku, V., Rowe, M. C., Ramirez-Gomez, L. A., Saldana, D. G., Quiroz, Y. T.. "Memory Intraindividual Variability and its Association with Hippocampal Integrity in Older Latino Adults Without Dementia". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.15.26365658

[2] Zita Oravecz, Joachim Vandekerckhove, Jonathan Hakun, Sharon Kim, Mindy Katz, Cuiling Wang, Richard Lipton, Carol Derby, Nelson Roque, Martin Sliwinski. "Computational phenotyping of cognitive decline with retest learning". psyarxiv. 2025-04-17T02:53:14.848Z. URL: https://osf.io/afe7m_v2

[3] Zheng, Y., Bo, B., Wang, D., Liu, Y., Gilbert, S. J., Li, Y., Kwok, S. C.. "Microstructural organization of Superior Longitudinal Fasciculus and Cingulum Bundle support metacognition driven cognitive offloading". bioRxiv. 2025-05-07T00:00:00.000Z. URL: https://doi.org/10.1101/2024.08.24.609527

[4] Chin, R., Hu, X., Ranieri, J., Kim, J.-H., Oh, H.. "Microstructural heterogeneity of white-matter aging and its age-dependent coupling to cognition". bioRxiv. 2026-10-09T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.09.757817

[5] Taguma, D., Yokoi, I., Kinjo, T., Tsuchida, S., Miyata, T., Matsuda, T., Lerma-Usabiaga, G., Takemura, H.. "Tractometry reproducibility and generalizability across scanners, scanner models, and acquisition protocols". bioRxiv. 2026-05-18T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.13.723388

[6] Lynch, K. M., Custer, R. M., Sepehrband, F., Toga, A. W., Choupan, J.. "Perivascular and parenchymal fluid characteristics are related to age and cognitive performance across the lifespan". bioRxiv. 2025-02-27T00:00:00.000Z. URL: https://doi.org/10.1101/2024.10.14.616343

[7] Nandi, S., McKenna, F., Henry, S., Nayak, S., Duong, T.. "White matter cognition mapping in pediatric cancer survivors: a two-cohort NODDI-Bingham study". bioRxiv. 2026-10-01T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.25.754456

[8] Naoya Tani, Linda Hoffman, Samantha Cohen, Nora Newcombe, Ingrid Olson. "Fornix microstructure is associated with holistic recollection in early childhood". psyarxiv. 2026-09-03T16:53:39.614Z. URL: https://osf.io/vzudt_v1

[9] Chang, K. H., Burke, L., LaPiana, N., Howlett, B., Hunt, D., Dezelar, M., Andre, J. B., Curl, P., Ralston, J. D., Rokem, A., MacDonald, C. L.. "Free water elimination tractometry for aging brains". bioRxiv. 2024-11-11T00:00:00.000Z. URL: https://doi.org/10.1101/2024.11.10.622861

[10] Lorenz, A., Sathe, A., Yang, Y., Durant, A., Wu, Y., Kim, M. E., Gao, C., Newlin, N. R., Ramadass, K., Kanakaraj, P., Mohd Khairi, N., Li, Z., Yao, T., Huo, Y., Dumitrescu, L., Shashikumar, N., Pechman, K. R., Risacher, S. L., Beason-Held, L. L., An, Y., Arfanakis, K., Erus, G., Davatzikos, C., Habes, M., Wang, D., Tosun, D., Toga, A. W., Thompson, P. M., Mormino, E. C., Zhang, P., Schilling, K., Albert, M., Kukull, W., Biber, S. A., Landman, B., Johnson, S. C., Bendlin, B., Schneider, J., Bennett, D. A., Jefferson, A. L., Resnick, S. M., Saykin, A. J., Hohman, T. J., Archer, D. B.. "Genetic architecture of the limbic white matter microstructure in aging and Alzheimer's Disease.". medRxiv. 2025-05-21T00:00:00.000Z. URL: https://doi.org/10.1101/2025.05.19.25327915

[11] Hodgetts, C. J., Postans, M., Williams, A. N., Graham, K. S., Lawrence, A. D.. "Fornix subdivisions and spatial learning: a diffusion MRI study". bioRxiv. 2025-05-16T00:00:00.000Z. URL: https://doi.org/10.1101/2025.05.15.654313

[12] Shuai, Y., Feng, Y., Villalon-Reina, J. E., Nir, T. M., Thomopoulos, S. I., Thompson, P. M., Chandio, B. Q.. "Assessing the Influence of Tractography Methods on Detected White Matter Microstructure in Alzheimer's disease". bioRxiv. 2026-03-11T00:00:00.000Z. URL: https://doi.org/10.1101/2025.11.11.687747

[13] Gumus, M., Bourganos, A., Mack, M. L.. "In vivo Quantification of White Matter Pathways in the Human Hippocampus". bioRxiv. 2025-01-13T00:00:00.000Z. URL: https://doi.org/10.1101/2025.01.13.632268

[14] Elder, M. A., Ellmore, T. M.. "Linking White Matter Integrity to Recognition Memory Speed: Fixel-Based and Fornix Analyses in Young to Middle Adulthood". bioRxiv. 2025-07-22T00:00:00.000Z. URL: https://doi.org/10.1101/2025.07.17.665387

How this survey was made

This survey answers one question: “How does within-person variability in memory performance relate to white matter microstructure (fractional anisotropy, free-water) along hippocampal-cingulum and fornix pathways in non-demented older adults?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 14 papers whose work this text draws on, chosen from the 126 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 22, 2026 and covered work from the previous two years.