Extended Cryogenic Suspension and Multigenerational Family Governance: An Archival Cohort Study of Tessier-Ashpool Revival Records

1. Introduction

Clinical studies of cryogenic suspension in the Sprawl and in the orbital habitats deal almost entirely with durations of months to a few years: transit sleep, holding a trauma patient for surgery, and the brief commercial cold-sleep sold to the wealthy. The one population known to have practised suspension across decades as an ordinary part of life is the Tessier-Ashpool family. Its members, and the clones the family raised, slept in the cold vaults of Villa Straylight and were woken in rotation. The family corporation, Tessier-Ashpool S.A., owned Freeside, the spindle-shaped resort and banking habitat at L-5, and governed it through whichever members happened to be awake.

The family's privacy long made the population inaccessible to medicine. That changed after the Straylight run of the mid-2030s, which ended the patriarch's rule and was followed by the merger of the two artificial intelligences Marie-France Tessier had set in motion, Wintermute at Berne and Neuromancer at Rio. In the years that followed, restricted series of the family's corporate archive, including the Villa's medical records, were opened to outside researchers, while related files passed to the Freeside Habitat Authority and, where they touched the AIs, to the Turing Registry. We write from the vantage of the early 2040s, with access to those series.

Two questions organise the study. The first is gerontological: whether the maintenance burden of suspension and the risk to cognitive continuity after revival change in kind, and not only in degree, once suspension runs past some duration. The second concerns governance. If the calendar of wakings decided who was present to rule, then scheduling revival was also a way of managing succession. A companion paper examines the legal side of that arrangement; here we supply the medical record on which it partly rests.

The family's own history supplies an index case. The patriarch, Ashpool, is known to have become unstable and violent, and he killed Marie-France, the co-founder. Whether that instability bore any relation to his suspensions is a question our series can address only in part. His later episodes appear in the logs and are discussed below, but we treat him as one individual among 14 and do not build the analysis around his case.

2. Methods

Primary data come from two restricted series of the Tessier-Ashpool S.A. Corporate Archive. The first, Series SL-CR, holds the Villa's cold-storage revival logs: one log per suspension episode, recording accumulated cold-time, every technician intervention, the readiness flags set before thaw, and a continuity battery administered after revival. The second, Series GV, holds minutes of the family's holding council, inheritance instruments and records of authority transfers. The Freeside Habitat Authority's medical service contracts for the Villa (Series FHA-MED-7) were used to confirm technician rosters, and a Turing Registry dossier compiled after the run was used to confirm the date order of several episodes.

Source criticism is essential for a family this secretive. The logs were kept by and for the family and were never meant for outside readers. They survive selectively: at least four gaps in SL-CR coincide with periods the council minutes describe as disputed, and some material may have been withdrawn before release. Both series are timestamped on a single Villa-wide elapsed-time counter, not by calendar date. Revivals and council events are placed on that one clock, and the accumulated cold-time of each episode is derived from it, so we report durations and relative timing only and do not assign calendar years to episodes. The Villa's staff were paid to keep the family alive and had little incentive to document failures, so intervention counts are probably lower bounds.

The unit of analysis is the suspension episode, defined as one continuous period in the cold ending in a logged revival. The series contains 31 complete episodes in 14 individuals: 9 episodes in 4 lineal family members, 13 in 6 members of the Jane clone line and 9 in 4 of the Jean line. Clones are counted as separate individuals because each carries a separate suspension history, even though they share a genome. Episode durations ranged from 2 to 31 years (median 11). Nine episodes exceeded 21 years.

Maintenance burden was defined as the number of logged technician interventions per episode (cryoprotectant renewal, thermal correction, vascular flushing and similar), modelled with negative binomial regression using log suspension-years as an offset. We compared a model log-linear in duration against a piecewise model whose breakpoint was estimated by profile likelihood, using ΔAIC and a likelihood-ratio statistic whose p-value was obtained by parametric bootstrap (2,000 replicates), because the breakpoint is not identified under the null and a standard χ² reference would be invalid. Cognitive continuity was measured by the Villa's post-revival battery, which we rescaled to a standardised composite (mean 0, SD 1). Scores were modelled with linear mixed models with a random intercept per individual. In the primary continuity model the breakpoint was fixed at the value estimated from the intervention model, so that model tests whether the continuity slope steepens beyond that point; it does not locate a break of its own. As a separate check, we estimated a continuity breakpoint from the scored batteries alone by profile likelihood and compared piecewise and linear continuity models by ΔAIC. We also dichotomised the composite at −1.5 SD as an impairment criterion. Scored batteries survive for 24 of the 31 episodes.

Governance events were coded from Series GV as council convocations with a succession item, inheritance triggers, deaths, and formal transfers of authority within S.A.; 19 such events fall within the series. A revival counted as governance-proximate if it preceded an event by 90 days or less on the archive's counter. To test this against chance, we circularly shifted each individual's revival times within that individual's own record 10,000 times, which keeps each person's spacing of wakings intact. Because a death cannot be scheduled, we repeated the analysis twice: once excluding each sleeper's own death from that sleeper's proximity count, and once dropping death events altogether.

3. Results

Across the 31 episodes the series logs roughly 900 interventions. Below the estimated breakpoint, rates were close to flat, at about 1.1 interventions per suspension-year and an incidence rate ratio (IRR) of 1.02 per year (95% CI 0.98–1.06, p = .32). The profile-likelihood breakpoint was 21 years (95% CI 18–24). Beyond it, each further year raised the intervention rate by 19% (IRR 1.19, 95% CI 1.09–1.30, p < .001), and the nine long episodes averaged about 2.9 interventions per suspension-year. The piecewise model fitted better than the log-linear model (ΔAIC = 9.1; likelihood-ratio statistic 13.1, parametric bootstrap p = .004).

Continuity scores, modelled with the breakpoint fixed at 21 years, behaved similarly. Below it the slope was negligible (β = −0.01 SD per year, 95% CI −0.04 to 0.02, p = .51). Beyond it, scores fell by 0.09 SD per additional year (95% CI −0.15 to −0.03, p = .003); the negative coefficient means that longer suspension predicts lower post-revival continuity. When the random effect was grouped by lineage (lineal family, Jane line and Jean line) instead of by individual, the estimate was almost unchanged (β = −0.08, 95% CI −0.15 to −0.01). Estimated from the continuity data alone, the breakpoint fell at 23 years (95% CI 15–29), an interval that overlaps the intervention estimate but is too wide to locate the change precisely; the piecewise continuity model was modestly preferred over a linear one (ΔAIC = 3.4).

Dichotomised, the battery flagged 6 of the 24 scored episodes (25%) as impaired. Five of the 7 scored episodes beyond the breakpoint met the criterion (71%), against 1 of 17 below it (6%) (Fisher's exact test, p = .003). The patriarch's two longest episodes, both past 25 years, are among the five, and the log's free-text notes for them record agitation and disorientation that persisted for weeks after thaw.

Seventeen of the 31 revivals (55%) fell within 90 days before a coded governance event. Under permutation the expected share was about 21% (95% range 10–35%), and the observed value was exceeded in fewer than 0.2% of shifts (p = .002). Governance-proximate revivals were also more often begun before the technicians' readiness flags were fully set: 7 of 17 (41%), compared with 2 of 14 other revivals (14%). With so few episodes, this second contrast is imprecise and not statistically significant (Fisher's exact test, p = .13). The timing result held in both sensitivity analyses. Excluding each sleeper's own death left 16 of 31 revivals (52%) governance-proximate against about 20% expected (p = .003), and dropping the four death events entirely left 15 of 31 (48%) against about 17% expected (p = .004).

4. Discussion

Maintenance burden changes sharply at about 21 years (plausibly anywhere from 18 to 24). Up to that point, extra years in the cold cost little in technician effort; beyond it, the burden rises year by year. Continuity scores are nearly flat below the same point and fall more steeply above it. Because that breakpoint was taken from the intervention model, we do not claim that the two outcomes change at an identical duration. Located independently, the continuity break is compatible with the intervention estimate but poorly fixed. The Villa's technicians seem to have kept the same routines across the whole range, adding interventions as problems appeared. Our data argue for a separate protocol for suspensions past the breakpoint, with planned maintenance, a fuller continuity battery and a longer supervised recovery, instead of treating them as short suspensions extended.

Recorded personalities offer a useful comparison for the continuity result. Work on read-only personality constructs describes an identity that stays stable within what was recorded but cannot learn across activations. A suspended brain is the biological counterpart: nothing is added while it sleeps, yet it must resume a life that has moved on. What our data add is that the biological version also degrades with duration, which the construct literature does not report for a stored recording. The patriarch's post-revival instability, long read as a matter of temperament, fits the upper end of the dose–response curve we report, though a single case cannot establish cause.

Revival timing supports, without proving, the thesis of the companion governance paper that revival scheduling worked as a succession-control mechanism. Wakings clustered before convocations, inheritance triggers and transfers of authority far more often than chance would produce. The partial evidence that such wakings began before technicians judged the sleeper ready suggests that governance sometimes overrode medical caution. Clones sharpen the point. Because the family could choose which Jane or Jean was awake, the rotation decided which heir held office, and the idea of a generation became partly a matter of scheduling.

The contrast with the zaibatsu is instructive. Those multinationals hold no territory, and their authority outlasts any individual because it resides in impersonal institutions that replace personnel as a matter of course. Tessier-Ashpool held Freeside outright, and its continuity ran through one family's sleeping bodies, so that whoever was awake governed. On our evidence, at least some decisions about who would be awake were made with succession in view.

5. Limitations

This is one family, and the sample is small and far from independent. Fourteen individuals, several sharing a genome, contribute 31 episodes. Only nine episodes lie beyond the breakpoint, so the estimates there are wide and the breakpoint itself is uncertain by several years. Generalising to other long-duration sleepers would be premature, since no comparable population is on record.

Gaps in the archive coincide with disputed periods, and these are exactly the periods in which governance-driven wakings or failed revivals would be most likely. If such episodes were withdrawn, the timing association could be understated and the medical risks at long durations underestimated. The continuity battery was designed and scored by the family's own staff, its norms are unknown, and our rescaling cannot correct for any bias in how it was administered.

The timing analysis cannot separate intent from coincidence. Governance events may have been scheduled around planned wakings as well as the reverse, and medical readiness may itself have followed the council's calendar if technicians prepared sleepers for known convocations. The permutation test makes chance clustering unlikely, but it cannot show which way the dependence ran. Our reading of scheduling as succession control therefore remains an interpretation that fits the data, and the data cannot confirm it.