Trace one model claim to its evidence
Research record
- Question
- Can a public research note connect one bounded model claim to its conditions, records, and limits?
- Evidence
- Published fly-brain articles, frozen protocols, saved trial outputs and audits; official NASA Systems Engineering Handbook web guidance supplies the traceability concept.
- Experiment
- No new neural simulation in this note. It traces an already executed 12-trial right-side T4 selection and assignment check.
- Results
- The published record reports higher right DNp15 counts than the saved left reference in all 12 listed comparisons under two fixed mapping variations.
- Limitations
- The trace is an editorial evidence map, not a NASA process, review, approval, or scientific validation. The underlying model result does not establish animal asymmetry, causal pathway, steering, body control, learning, or driving.
- Reproduction
- Follow the linked public evidence bundle; this draft adds no new data or executable experiment.
Sources
- NASA Requirements Management (6.2)
- NASA Product Verification (5.3)
- Published T4 selection experiment
- Published matched-input reference experiment
Run evidence
Claim boundary
This note borrows a small organizational idea from NASA’s requirements-management guidance: keep a discernible link between a question, its conditions, its check, and its record. It is not a NASA process, review, approval, or endorsement.
The claim being traced is deliberately narrow: under two fixed changes to the right-side T4 selection and assignment in a pinned spiking model, right DNp15 counts remained above the saved ipsilateral left reference in the listed comparisons. No new simulation was run for this note.
Trace
| ID | Record | Evidence | Status and boundary |
|---|---|---|---|
| Q-01 | Does the previously observed DNp15 difference persist after two fixed right-side mapping changes? | Published experiment | Executed question; not a claim about an animal or vehicle. |
| C-01 | Keep the pinned network, E0/E180 rate traces, amplitude, three seeds, and 200–1000 ms readout; test reversal and rotation separately. | Published experiment and frozen bundle | Executed conditions. The reused reference trials and shared random uniforms make the 12 comparisons dependent. |
| V-01 | Check saved driver tables and recount saved output spikes. | Published checks and manifest | Verification-like evidence for the saved record. It is not validation of biological steering. |
| R-01 | In all 12 listed comparisons, right DNp15 is above the saved left reference. | Table below, copied from the published article | Executed within-model observation only. |
| L-01 | Preserve the unresolved explanations and excluded conclusions. | Published limitations | Open: selection, wiring, and network interaction are not separated. |
| N-01 | Cut and restore a specific candidate pathway under matched inputs. | Published next step | Not run. It has no result to report. |
Saved result
DNp15 spikes on the stimulated side, 200–1000 ms. The first two columns are saved reference trials; the last two are newly executed trials in the published record.
| Pattern | Seed | Left reference | Right reference | Right reversed | Right rotated |
|---|---|---|---|---|---|
| E0 | 20261221 | 0 | 23 | 21 | 23 |
| E0 | 20261222 | 0 | 21 | 22 | 20 |
| E0 | 20261223 | 1 | 23 | 21 | 23 |
| E180 | 20261221 | 6 | 39 | 37 | 38 |
| E180 | 20261222 | 3 | 37 | 37 | 36 |
| E180 | 20261223 | 6 | 39 | 37 | 38 |
What the trace does not close
The selected cells are not verified anatomical homologues. Both sides remain connected. Rotation changes membership and assignment together. The rate bridge is not a measured biological input, does not provide individual retinotopy, and has no physical-turn calibration. Saved driver tables do not independently prove in-memory synapse delivery because the asserted routing arrays were not exported.
NASA distinguishes verification from validation in its systems-engineering guidance. In that spirit, the saved-data checks support a bounded record of what the model run reported; they do not validate an animal mechanism, body, steering behavior, learning, or driving capability.
Source note
NASA defines traceability as an association among entities such as requirements, system elements, verifications, or tasks, and describes bidirectional traceability as linking a requirement to its source and allocated children. This article uses that concept as a lightweight research-record pattern only. It makes no claim that NASA has reviewed, approved, or applied it to this project.
Update record
- 2026-10-31: Drafted from already published trial records and official NASA web guidance. No new model or audit was run for this note.