Lilienfeld 1925 Package Evidence for Independent Replication

# Package the Evidence for Independent Replication: A Verdict No One Else Can Check Is Not Yet Closed

Step 27 composed twelve steps of evidence into one scoped, falsifiable conclusion that this project's own authors can audit. Step 28 asks whether anyone outside this project could audit it too — and, further, whether an independent group could rebuild the specimens from scratch and get a comparable answer. These are two different claims, and conflating them is exactly how an isolated result quietly becomes treated as an established fact. Step 28 exists to keep them separated and to produce the material each one actually requires.

The historical device at the center of this series was never independently replicated during its own era. That absence is not a minor historical footnote; it is a direct illustration of the failure mode this entire verification arc was built to avoid. A project that performs Steps 15 through 27 carefully and then stops, with its raw data and protocols sitting unorganized across individual step records, has built the evidence for reproducibility and replication without actually making either one possible for anyone else. Step 28 is the packaging work that closes that gap.

## 1. Separate reproduction from replication, and package for each one differently

Reproduction means an independent party takes this project's own raw measurements and re-runs the analysis — the Step 22 statistic, the Step 24 pass-rate interval, the Step 27 falsification table — and checks whether the same conclusions follow from the same data. Replication means an independent party builds new specimens through the Steps 1–14 fabrication sequence, runs the Steps 15–26 protocols on them, and checks whether a comparable disposition results from new data.

Reproduction requires the data and the analysis code. Replication requires the fabrication and measurement protocols in a form detailed enough to execute without consulting this project's authors. A package that provides only one of these supports only one of these claims, and any report built from this step must state plainly which of the two is actually being supported by what has been assembled.

## 2. Package every protocol as an executable procedure, not a narrative description

Steps 1 through 26 were written as explanatory articles, each carrying context, reasoning, and illustration alongside its instructions. That form is right for this series' primary purpose but is not directly usable as a replication package, because an independent builder needs a procedure stripped to its executable steps: exact materials, exact sequence, exact measurement settings, and exact acceptance criteria, with the surrounding reasoning available as a reference rather than interleaved into the steps themselves.

Extract, for each of Steps 1 through 26, a numbered procedure checklist referencing back to that step's full article for rationale. Include every quantitative parameter that was fixed in this series — the Step 21 bias list, the Step 19 settling criteria, the Step 20 charge budget, the Step 24 population size and independence rule — as an explicit value in the checklist, not as a cross-reference the builder has to go hunting for. A replication attempt that silently substitutes a different value for an unstated parameter is not testing this project's finding; it is testing something else.

## 3. Package the raw data with enough provenance to be trusted on its own

For reproduction, package the raw measurement files from every step, each one associated with its specimen identifier, the instrument and calibration record used, the timestamp, and a checksum computed at the time of recording. An independent party re-analyzing this project's data needs to know that the file they are reading is the same file this project itself analyzed, not a version that was edited after the fact, however well-intentioned the edit. Where Step 18's or Step 20's record already includes a cryptographic hash from an earlier deployment step, carry that same hash forward into the replication package rather than recomputing a new one that could silently diverge from the original.

## 4. Define the audit a reviewer should be able to perform without this project's help

Write down the specific checks an independent reviewer can run against the package alone: recompute the Step 22 paired-difference statistic from the raw files and confirm it matches the value in the Step 27 falsification table; recompute the Step 24 pass-rate interval from the per-specimen dispositions; verify every file's checksum against the manifest. State explicitly that a reviewer who cannot complete these checks from the package contents alone has found a packaging defect, not a reason to trust the project's summary instead.

Separately, define what a replication attempt would need to report back for its result to be comparable to this project's own: the same population-declaration discipline from Step 24, the same frozen-protocol discipline from Step 21, and a disposition stated in the same scoped-sentence form Step 27 established, rather than a bare success or failure claim.

## 5. Version the package and never silently edit a released one

Assign the package a version tied to a fixed reference point in this project's own records — the specific state of every step's article and data at the time of packaging. If a later correction is made to any step's protocol, article, or data, produce a new package version with the change documented, rather than editing a previously released package in place. A reviewer who finds a discrepancy between an older package and the current articles must be able to tell whether that discrepancy reflects a genuine correction or a packaging error, and silent edits make that impossible.

## 6. State plainly what has not yet been packaged

If time or resources only allow packaging reproduction material in this pass, say so explicitly, and list what remains outstanding for a true replication package: which protocol checklists are incomplete, which raw files lack a provenance record, which instrument calibration references are missing. An honestly partial package that says what it does not yet support is more useful, and more consistent with the discipline of this entire series, than a package that implies completeness it does not have.

## 7. Minimum Step 28 record

Retain:

• the explicit statement of which claim — reproduction, replication, or both — the assembled package actually supports;
• the executable procedure checklist for each of Steps 1 through 26, with every fixed quantitative parameter stated explicitly rather than left as a cross-reference;
• the raw data files for every specimen and step, each with its specimen identifier, instrument calibration reference, timestamp, and checksum, with any checksum already generated in Steps 18 or 20 carried forward unchanged;
• the analysis code or scripts needed to recompute the Step 22, 24, and 27 statistics from the raw files;
• the written reviewer-audit procedure and confirmation that this project's own team completed it successfully against the assembled package before release;
• the package manifest recording its version and the fixed project-state reference it was generated from;
• a log of every subsequent package version and what changed in each;
• an explicit list of anything not yet packaged, stated as an open item rather than omitted silently.

The historical geometry and claimed effect are grounded in [J. E. Lilienfeld, US Patent 1,745,175](https://patents.google.com/patent/US1745175A/en), a disclosure that was never subjected to independent replication during its own era — the specific historical absence this step's packaging work is designed to make possible to correct. The provenance and checksum discipline applied here follows directly from the backup and verification practice used throughout this series' own deployment scripts, and the low-level measurement provenance concerns described in the [Keithley Low Level Measurements Handbook](https://www.tek.com/en/documents/product-article/keithley-low-level-measurements-handbook---8th-edition).

## Package Evidence for Independent Replication’s Place in the Process Lineage

Step 27 produced a conclusion this project's own authors can audit. Step 28 produces the material that lets someone outside this project audit it too, and separately, the material that would let an independent group rebuild the specimens and test whether a comparable disposition results from genuinely new data. Reproduction and replication are kept as two distinct, separately supported claims, each package is versioned against a fixed project state rather than silently edited, and any gap in what has been assembled is stated openly rather than implied away. This is the step that gives the Step 27 verdict a chance to be checked by someone other than the people who produced it — the specific opportunity this device's own history never had.