Opening proposition
The idea in brief
Restoration is normally described as a return: a damaged file, language, ecosystem or collection is brought closer to an earlier state. But recovery requires selection. Conservators choose formats, specimens, pronunciations, baselines and error-correction rules.
Recovery canonization proposed that repeated restoration creates a one-way transfer of authority. Once later copies descend mainly from the recovered version, the recovery pipeline’s choices become harder to distinguish from properties of the original.
The process intended to preserve history becomes a hidden author of history. Variation that the pipeline cannot recognize gradually disappears, while whatever it reliably reconstructs gains the status of authenticity.
Full exposition
Restoration always contains a model
To recover something, a restorer must decide which features belong to the thing and which count as damage. A digital archive chooses acceptable formats; an ecologist selects a historical baseline; a language revival programme settles spellings and pronunciations; a conservator fills missing material according to an interpretation of the work.
These choices may be careful and well documented, but they are not neutral. They form a model of the original. When recovery occurs once, the model’s influence may remain visible as a repair. When recovery occurs repeatedly, later versions increasingly descend from earlier reconstructions rather than directly from the source.
How a repair becomes a canon
The process has three stages. First, damage or decay makes some features unavailable. Second, a recovery system reconstructs what its rules can recognise. Third, the reconstructed version becomes the input to future preservation. Features favoured by the recovery model are now more likely to survive the next cycle; features it cannot represent are less likely even to be noticed as missing.
After enough cycles, the distinction between original property and restoration convention becomes difficult to recover. The pipeline has acquired genealogical authority. Its outputs define the range of future possible outputs, so a practical canon emerges without any single declaration that one version is definitive.
The pattern across living and cultural systems
In a seed bank, regeneration grows a sample of stored seeds to replenish declining stocks. Germination conditions, population size and pollination control influence which genetic variation reaches the next stored generation. In an archive, repeated file migration preserves content that can be represented in destination formats while interaction, timing or visual behaviour may disappear.
The same structure can appear in oral tradition, ecological restoration and therapeutic memory work. A teaching curriculum selects one version of a story for reliable transmission. A restoration baseline privileges one historical ecosystem state. A repeatedly rehearsed narrative becomes easier to retrieve than the ambiguous material from which it was formed.
Preserving the plurality of possible originals
Recovery Canonization does not counsel against restoration. Without intervention, much would simply vanish. It argues for preserving the recovery process as part of the object’s history. Every reconstructed version should carry its assumptions, transformations, uncertainties and excluded alternatives with it.
Where possible, restorers can maintain multiple lineages rather than one canonical output: raw and migrated files, several ecological baselines, competing transcriptions, diverse seed samples and explicit confidence ranges. The aim is not to eliminate interpretation but to prevent interpretation from disguising itself as inheritance. A good recovery system preserves not only an answer, but evidence that other answers were once possible.
Every recovery system contains a theory
To recover something, a system must distinguish essential features from incidental ones. A backup restores files and permissions but may ignore the exact physical arrangement of a disk. A seed bank preserves genetic material but cannot preserve every ecological relationship surrounding the original plant. A conservator may restore a painting’s colours while accepting microscopic changes in material. These choices form an implicit theory of what the thing is.
That theory is unavoidable because no recovery process can preserve every causal detail. The danger begins when the recovery test is mistaken for a neutral window onto identity. What can be checked after restoration becomes what the system is organised to retain. Features outside the test may survive initially through luck, then disappear through repeated cycles without ever registering as losses.
From selective preservation to canon
Suppose each restoration accepts any output that matches a set of significant properties. The first restored version may differ from the original only slightly. When the restored version later becomes the input to another recovery, its differences are no longer measured against the original; they are part of the starting material. Repeated cycles allow variation within the accepted equivalence class to accumulate.
Meanwhile, maintainers become more familiar with the traits their tools can restore reliably. Documentation, expertise and quality assurance converge on those traits. The pipeline’s successful outputs become exemplars for future work. A practical test has turned into a canon: not merely a rule for judging copies, but the source from which people learn what the original was supposed to be.
Selection can be invisible
The process need not involve deliberate falsification. Each custodian may faithfully follow the best available standard, and every recovered version may pass all declared tests. Drift occurs because the standard creates a selective environment. Variants that are easy to encode, regenerate and verify survive more reliably than variants that depend on tacit knowledge, fragile context or relationships outside the archive.
This is why the final result can look unusually pure. Awkward, hybrid or weakly documented features disappear, leaving a version that expresses the preservation scheme’s categories with greater consistency than the original ever did. The recovered tradition, ecosystem or software system may appear canonical precisely because recovery has filtered away evidence of its former messiness.
What counts as evidence of drift
Comparing the latest restoration only with the immediately previous one will miss cumulative canonization. The relevant evidence includes early snapshots, independent descriptions, unused metadata, dissenting recollections and alternative restoration attempts. Disagreement among reconstructions is informative: it maps the dimensions the official pipeline has compressed or left underdetermined.
A particularly revealing test is cross-recovery. Give the same damaged source to teams using different assumptions and tools. If each produces a stable but systematically different lineage across repeated restorations, then the pipeline is acting as an author. Stability within one lineage should not be confused with fidelity to a unique original.
Plural recovery as preservation
The response is not to abandon standards, which would make recovery arbitrary. It is to preserve the contingency of the standard itself. Archives can retain raw captures alongside normalised versions, record transformations as first-class data, maintain more than one emulator or reconstruction, and mark which properties were chosen rather than discovered. Biological collections can store multiple lines and minimise the number of regeneration cycles.
At a cultural scale, plural recovery means allowing several communities to carry different accounts of what deserves restoration. This may feel less decisive than a single authorised version, but the plurality preserves information about the lost object’s possible identities. The deeper lesson is that recovery does not merely transport the past into the future. It establishes the tribunal before which future versions will prove that they are the past.
A thought experiment
A seed bank periodically grows stored seeds to produce fresh seed. Only plants that germinate under the regeneration protocol contribute to the next stored generation. After several cycles, the collection remains labelled as the original accession, but the protocol has selected which genes and traits count as preservable.
Consequences
What the idea changes
Preservation would be treated as an evolutionary or editorial process, not a neutral bridge through time. Every recovery system would need to disclose the traits it is capable of canonizing.
Ways to think with it
- It unified archive migration, ecological restoration, language revival and memory repair.
- It predicted convergence toward what a restoration pipeline can repeatedly recognize.
- It made provenance of repairs as important as provenance of originals.
Further reading
The intellectual neighbourhood
Notes on the idea’s provenance and editorial review are kept separately in the editorial appendix.