From a signal to a supported claim
A photograph, radar return, witness account, or laboratory result can be authentic and still be open to several interpretations. The first question is therefore not What is it? but What was actually observed, how was it recorded, and what survived? Keep four layers separate: the observation; the measurements that can be recovered from it; an inference about behavior or performance; and a claim about origin. Evidence may support one layer without reaching the next.
NASA’s 2023 independent study emphasized calibrated instruments, multiple measurements, and complete sensor metadata because many existing UAP records were collected incidentally rather than for scientific study. That limitation is not a reason to ignore a report. It is a reason to match the strength of the conclusion to the quality of the data.
An evidence-type matrix
| Evidence type | What it can support | What it cannot establish by itself |
|---|---|---|
| Witness account | Where the observer was, what they perceived, sequence, duration, and immediate context | Exact range, size, speed, or origin when those values were not independently measured |
| Image or video | Recorded light, contrast, apparent motion, framing, and—when retained—time and camera settings | Distance, physical scale, or exceptional performance without geometry and a known sensor model |
| Radar or telemetry | Returns or tracks produced within a particular system’s coverage and processing rules | A unique identity merely because a track lacks a label; primary and cooperative secondary radar collect different information |
| Physical sample | Composition, structure, contamination, and test results for the item examined | Where the item came from without a documented chain of custody and representative sampling |
| Contemporaneous record | What an agency, operator, or investigator recorded at a stated time | That every statement inside the record was verified or that later retellings preserve its meaning |
Preserve provenance before interpretation
Begin with the earliest available generation of the evidence. Record who created it, on what device or system, when and where it was acquired, who held it, and every known transformation. Preserve the original file, a cryptographic hash when available, the full metadata, and a separate working copy. Note cropping, enhancement, transcoding, frame-rate changes, captions, and platform compression. NIST describes provenance as the documented history of a data asset—its origin, acquisition, processing, and alterations. A polished clip with no such history may be useful as a lead, but it is weaker than an ordinary-looking original with a traceable path.
For a physical item, use the same discipline: photograph packaging and seals, identify each custodian, document sampling locations, retain controls, and separate a laboratory’s measured result from a story about recovery. An unusual composition does not, by itself, identify a manufacturing process, owner, age, or origin.
Sensors are systems, not neutral windows
Ask what the instrument was designed to detect. A thermal camera displays differences in infrared energy through optics, detector behavior, software, gain, focus, and a chosen palette. A radar return depends on wavelength, geometry, coverage, target reflectivity, processing, and whether the system is primary or relies on a cooperating transponder. The FAA’s ASR-11 description, for example, distinguishes reflected primary-radar returns from secondary-radar replies that include aircraft information. Two displays can therefore appear to be independent while drawing from the same upstream track or fused data feed.
Geometry is often decisive. Apparent angular motion becomes linear speed only when range is known. Parallax can make a nearby or distant object appear to cross the background rapidly when the observing platform is moving. Zoom, autofocus, glare, stabilization, and dropped metadata can change apparent shape. Before calculating performance, recover platform position, sensor pointing angles, range, field of view, wind, time synchronization, and the uncertainty on each value. If those inputs are missing, publish a range of possible values or decline to calculate one.
Test independence, not repetition
Multiple witnesses strengthen a record when their viewing positions and accounts are genuinely independent. Multiple sensors strengthen it when they measure different physical properties and have synchronized clocks and known calibration. They do not automatically strengthen it when witnesses discussed the event before giving statements, outlets copied one report, or several screens displayed a single fused track. Build a source tree: identify the earliest account, every derivative copy, and each truly separate measurement.
Worked example: a fast-looking infrared dot
Observation: a moving aircraft’s infrared display shows a small target that seems to cross the background quickly. The recording is authentic. Measured: the display supplies pointing angles and a range value, while the aircraft’s complete track and atmospheric conditions are unavailable publicly. Initial inference: the target may be moving at extraordinary speed. Alternative test: reconstruct the line of sight across plausible aircraft paths and winds, then compare apparent and intrinsic motion.
In its GoFast analysis, AARO reported high confidence that the object did not display anomalous or exceptional behavior, while also warning that missing track and atmospheric data prevented an exact speed and heading solution. The useful conclusion is calibrated: the video is real; the apparent motion is real; the extraordinary-performance interpretation is not supported by the available geometry. The result would change if reliable synchronized data established a different range, platform path, or target motion. AARO’s Puerto Rico reconstruction offers the same lesson from another case: apparent high speed and apparent entry into water were tested against platform motion, wind, and line of sight rather than accepted from the video alone.
Use confidence language that exposes the limits
- Observed: directly present in the retained record.
- Supported: consistent with several independent measurements and no stronger contradiction.
- Probable: the best current explanation, with material uncertainty stated.
- Possible: compatible with the data but not uniquely supported.
- Unresolved: the available evidence does not discriminate among explanations.
Unresolved is a data state, not a category of origin. For a full case workflow, use Evaluate a UAP Claim. Review the site’s sources and methodology, check technical language in the glossary, and locate original records through Primary Sources. Then compare the strongest ordinary explanation in Critical Analysis. Applied examples include the site’s briefing on the Navy FLIR, Gimbal, and GoFast recordings and its review of the Rhodes photographs and their provenance gaps.