Repeated Training Load Leaves a Measurable Bone-Damage Memory in Thoroughbred Metacarpal Subchondral Bone

Authors: Sterling Geisel, QBist Lab; Dr. Hideo Tanaka

Abstract

Anwar et al. model how Thoroughbred racehorse training programs alter metacarpal subchondral bone volume fraction and normalized bone damage. Their result is concrete. High-speed workload can produce adaptation, but added workload stores damage faster than repair can remove it. Rest changes this balance. This Working Paper applies Material Memory to the same domain. The Postulate predicts that bone is not only responding to instantaneous load. It retains a history-dependent mechanical trace of repeated loading, and that trace changes future damage probability under matched training inputs. The source model already includes such memory through evolving bone volume fraction and normalized damage. Pudding Theory adds a sharper experimental claim: two bones with equal current morphology and equal current loading can diverge if their prior loading histories differ. The distinguishing observable is residual microcrack density and subsequent normalized damage gain after a standardized high-speed training block.

Source Synopsis

Anwar, Pan, Morrice-West, Malekipour, Pivonka, Flegg, Whitton, and Hitchens study musculoskeletal injury risk in Thoroughbred racehorses using a mathematical model of bone adaptation, microdamage, and repair. The model is applied to the lateral condyle of the third metacarpal bone, a clinically important site in racing injury. Inputs are joint stress, strain rate, and loading cycles per day. These are mapped from training speed and distance.

The paper compares progressive training programs, race-fit training programs, rest duration, rest frequency, and low-intensity training substituted for rest. The model uses bone volume fraction as an adaptation variable and normalized bone damage as a failure variable. In the paper’s notation, bone failure occurs at normalized damage \(D^*=1\).

The main result is that more training is not simply better adaptation. Lower-volume programs that still contain high-speed work produce sufficient bone adaptation with less damage. High-volume progressive training produces higher final damage than fast-and-light progressive training. Race-fit workload is still more consequential. Medium-high and high-volume race-fit programs exceed the modeled failure threshold by the end of preparations.

Rest changes the damage balance. Two rests per year lower maximum damage compared with one rest per year. Three rests lower it further, though with less additional benefit and less racing time. Longer rest periods also reduce damage. Eight-week rest periods produce lower maximum and minimum damage than four-week rest periods. Low-intensity training can substitute partly for rest, but only when the back-off is long and low enough. Four-week back-offs reduce damage compared with one rest per year. Two-week back-offs do not.

The source paper’s central claim is mechanistic. Workload history changes bone state. The present injury risk depends on accumulated damage, repair time, and adaptation, not only on current speed or race distance.

Postulate Lens

This paper applies Material Memory. The fit is direct. The source model treats bone as a material substrate that records repeated mechanical signals through changing bone volume fraction and microdamage. The trace is not decorative. It biases later probability of local failure.

In Pudding Theory, Material Memory states that matter retains the trace of repeated signals, and that the trace biases future probability. In this application, the repeated signal is not conscious intent. It is the structured mechanical signal of training: stress amplitude, strain rate, and stride count. Bone receives that signal through the coupled activity of osteoblasts, osteoclasts, microcrack formation, and microcrack removal. The trace is measured as residual microdamage, altered specific surface, and changed bone volume fraction.

The source paper already gives a useful operational bridge. A horse entering a race-fit phase after high-volume progressive training does not enter the same physical state as a horse entering after fast-and-light training. Even if the next race-fit program is identical, the damage trajectory differs because the substrate carries a history. That is Material Memory in a biomedical form.

The Postulate does not replace mechanobiology. It adds a constraint on how to test it. Current morphology alone should be insufficient. If two horses have similar current bone volume fraction but different prior high-speed workload histories, the retained trace should still alter subsequent damage gain under a matched loading block.

Pudding Theory Prediction

Pudding Theory predicts a measurable hysteresis in equine subchondral bone response after repeated high-speed training. The hysteresis should remain after controlling for current bone volume fraction, current apparent density, age, sex, racing class, and standardized workload. The key point is not that bone adapts. That is established in the source model. The point is that prior signal history should leave a residual state variable that affects future probability of damage accumulation beyond what current bulk morphology predicts.

The predicted experiment is straightforward. Select horses with comparable third metacarpal bone volume fraction and no clinical injury. Divide them by prior workload history, using detailed training records. One group should have a recent history of high cumulative gallop distance. The other should have a recent history of lower cumulative gallop distance with comparable present fitness indicators. Apply a standardized short high-speed training block, or examine the next matched race-fit interval observationally with careful workload matching.

The Pudding Theory prediction is that the high-history group will show larger post-block microdamage gain, or a higher normalized damage estimate, than the low-history group even when the immediate training block is identical. The effect should be strongest in regions already identified as highly loaded distal metacarpal subchondral bone. It should be weaker after sufficiently long rest if repair has erased the trace.

This prediction also clarifies the meaning of rest. Rest is not absence. It is an active interval in which the retained trace is partially removed or reorganized. In the source paper, longer rest reduces both minimum and maximum normalized damage. Under Material Memory, this is expected. The prior loading trace fades when repair exceeds formation.

The most useful measurable quantity is not race outcome. It is the residual microcrack burden after matched load. Histology, high-resolution imaging, or validated model inversion from stride-level workload could provide the test. The signature is path dependence: equal present load, unequal future damage, explained by retained material trace.

Falsifiable Observable

The distinguishing observable is the difference in post-block residual microcrack density, \(\Delta \rho_{\mu crack}\), between matched horses with high and low prior high-speed workload histories after the same standardized high-speed workload. If \(\Delta \rho_{\mu crack}\) were measured to be 0 within a pre-registered equivalence margin after matching for current bone volume fraction, age, sex, and applied workload, this Postulate would be falsified.

Editorial Dialogue

Tanaka: The source paper already has memory. It uses differential equations. Damage and bone volume fraction carry state forward. Why import a Pudding Postulate?

Sterling: Because the Postulate is not being used as an ornament. It sharpens the variable. The source model tracks damage and adaptation. Material Memory asks whether prior loading leaves a measurable residual trace even when current morphology appears matched.

Tanaka: That sounds like ordinary biomechanics.

Sterling: It is compatible with ordinary biomechanics. It is not a denial of it. Pudding Theory fails here if the proposed trace has no independent observable. The microcrack density test gives it one.

Tanaka: You are not claiming that intent changes bone repair?

Sterling: No. That would be a different claim and is not needed. The signal here is repeated mechanical load. The matter is subchondral bone. The retained trace is microdamage and remodeling state.

Tanaka: Then the contribution is modest.

Sterling: Correct. It should be modest. A strong application does not need excess machinery. The paper shows that training history matters. The Postulate predicts a path-dependent residual after standard covariates are matched. That is enough to test.

Discussion

This application is narrow. It does not claim that Pudding Theory improves the numerical fit of the Anwar et al. model without new data. It identifies a falsifiable extension of the model’s state dependence. The strongest test requires individual-level longitudinal training records, repeated bone measurements, and enough horses to separate prior workload from selection bias. Trainers may also select more resilient horses into heavier programs, which could hide or invert the predicted trace.

The source model has acknowledged limits. It does not include individual variation in biological rates, within-cluster training variation, cardiovascular fitness, or impaired repair after vascular disruption. These matter for any empirical test. The proposed observable should therefore be measured near the tissue scale, not inferred only from race starts or career length.

A null result would be informative. If current morphology and current loading fully determine subsequent damage gain, Material Memory has no independent role in this domain. If prior workload history predicts damage after those controls, the Postulate gains a bounded, measurable foothold in equine bone mechanobiology.

References

1. Md Nurul Anwar, Michael Pan, Ashleigh V. Morrice-West, Fatemeh Malekipour, Peter Pivonka, Jennifer A. Flegg, R. Chris Whitton, and Peta L. Hitchens. “Balancing training load, rest and musculoskeletal injury risk: a mathematical modelling study in Thoroughbred racehorses.” arXiv:2603.22680, 2026. DOI: doi:10.48550/arxiv.2603.22680.

2. S. Ochs. “Pudding Theory: A Topological Theory of Information Fields.” QBist Lab Working Paper, 2026.

3. Pan, M., Malekipour, F., Pivonka, P., Morrice-West, A. V., Flegg, J. A., Whitton, R. C., and Hitchens, P. L. “A mathematical model of metacarpal subchondral bone adaptation, microdamage and repair in racehorses.” Journal of The Royal Society Interface, 22(231):20250297, 2025.

4. Hitchens, P. L., Pivonka, P., Malekipour, F., and Whitton, R. C. “Mathematical modelling of bone adaptation of the metacarpal subchondral bone in racehorses.” Biomechanics and Modeling in Mechanobiology, 17(3):877-890, 2018.

5. Holmes, J. M., Mirams, M., Mackie, E. J., and Whitton, R. C. “Thoroughbred horses in race training have lower levels of subchondral bone remodelling in highly loaded regions of the distal metacarpus compared to horses resting from training.” The Veterinary Journal, 202(3):443-447, 2014.

6. Verheyen, K., Price, J., Lanyon, L., and Wood, J. “Exercise distance and speed affect the risk of fracture in racehorses.” Bone, 39(6):1322-1330, 2006.

7. Martig, S., Hitchens, P. L., Lee, P. V., and Whitton, R. C. “The relationship between microstructure, stiffness and compressive fatigue life of equine subchondral bone.” Journal of the Mechanical Behavior of Biomedical Materials, 101:103439, 2020.

8. Wong, A. S. M., Morrice-West, A. V., Hitchens, P. L., and Whitton, R. C. “The association between Thoroughbred racehorse training practices and musculoskeletal injuries in Victoria, Australia.” Frontiers in Veterinary Science, 10:1260554, 2023.