Forty-one independent accounts collected between the second and sixth field seasons, of which twenty-nine were obtained before any media coverage existed. Pre-publicity testimony is treated as the primary corpus; the remaining twelve are retained but weighted down for contamination risk.

Instrumented transits (n=64) were run at varied hours, speeds, and directions carrying four independent clock references: vehicle GPS, cellular network time, a disciplined oscillator, and a sealed mechanical chronometer. Toll-camera capture times were obtained under records request and compared against departure logs.

Structured cognitive interview, no leading prompts, conducted separately and blind to other accounts.

• W-1 (commercial driver, 19 yrs route experience): reports the sun's apparent position shifting 'a hand's width' across the corridor. Route familiarity argues against misjudged duration.
• W-7 (nurse, night shift): arrived at a timestamped badge-in 96 seconds before her phone recorded her leaving home. Badge system is on a separate network with independent time sync.
• W-22 (retired surveyor): the only witness to report no anomaly across eleven transits. His account is retained precisely because it is negative.

A uniform deficit of 112–124 s (mean 118 s) appears across all four clock references, including the sealed mechanical unit, which is not network-disciplined and cannot be spoofed by any signal-layer attack. Three drivers cleared a fixed toll camera before their logged departure. The corridor has been closed for 'resurfacing' four times in six years; resurfacing records do not correspond to any materials procurement.

• H1 · GNSS spoofing or jamming. Falsified: the mechanical chronometer is unaffected by RF and shows the same deficit.
• H2 · Collective time misestimation. Falsified: the toll camera is an external fixed reference and does not misestimate.
• H3 · Records error at a single administrative source. Weakened: the deficit reproduces across four unrelated systems under different custodians.
• H4 · Instrument fault in the survey vehicle. Not falsified for transits 1–12; excluded thereafter, when a second vehicle reproduced the result.

Recommend continuous instrumented occupancy for one full year and independent replication by a second institution. Under no circumstance should this file be cited as evidence of a specific mechanism.

The narrative was traced backward through staff turnover to a 1987 shift-log entry. That entry describes the same figure in the same posture — in a building not completed until 1991. The story therefore predates its own setting.

Archival regression: every retrievable telling was dated and mapped to its earliest source. Motif indexing was performed against regional folklore collections.

• W-3 (custodial supervisor): sincere, consistent, and demonstrably describing an event he places in a year he did not yet work there. Payroll records confirm the discrepancy. No indication of fabrication.
• W-11: account matches a published regional legend almost phrase for phrase, including a detail the legend gets wrong.

Three photographs offered as evidence. Two are the same image at different crops. The third resolves under level adjustment to a coat on a handrail.

• H1 · Genuine recurring apparition. Falsified by the pre-construction date of the earliest telling.
• H2 · Migratory folklore attaching to a new structure. Supported: motif matches a documented type with regional precedent.
• H3 · Deliberate hoax. Unsupported. No witness shows any sign of invention; they are transmitting, not inventing.

Retained as a methodological control. Any investigator who cannot correctly classify 077-K should not be trusted with 041-A.

Twenty-two complaints over four years, all localised to a single room and all reporting the same triad: dread, chest pressure, peripheral movement.

Because the reported symptoms match documented low-frequency acoustic exposure, the survey was designed to falsify that hypothesis first, not last. Instrumentation: calibrated infrasound array (0.1–20 Hz), thermal imaging, EMF survey, and a double-blind occupancy protocol in which subjects did not know whether the extraction fan was running.

• W-2 and W-9 independently described 'grey shapes at the edge of sight' — a hallmark of eyeball resonance under sustained infrasound near 19 Hz.
• W-14, who reported the strongest effect, is a musician; her sensitivity was the clue that redirected the survey.

A standing wave at 18.9 Hz, peak 92 dB, generated by an extraction fan whose mounting had failed and which was coupling to a duct run of near-exactly one half wavelength. The pressure node fell across the room's centre — precisely where occupants stood to report the sensation.

• H1 · Infrasonic resonance. Confirmed: under blind conditions, reports tracked fan state at a rate far beyond chance.
• H2 · Suggestion from prior reports. Contributory but insufficient: naive subjects reported the effect without prior exposure to the story.
• H3 · Anomalous presence. Falsified: remediating the mount eliminated the phenomenon entirely.

None. Filed as a demonstration that the mundane explanation, when it is the right one, explains everything — including why the witnesses were right about what they felt.

First reported by a documentary crew; subsequently reproduced by the investigator across three visits with independent equipment.

Parallel capture with four recorder classes: digital solid-state, analogue magnetic tape, optical film with a variable-density soundtrack, and a mechanical phonautograph tracing onto smoked glass. The last carries no electronics whatever and was included specifically to separate a recording-medium effect from an acoustic one.

• W-1 (field recordist, 30 yrs): describes the silence as 'arriving', not as an absence — a phenomenological detail the investigator considers significant and cannot yet operationalise.
• W-5: hearing-impaired, reports no subjective change. Her recorders were affected identically. This decouples the effect from perception.

Digital and analogue captures both show a noise floor collapse of roughly 40 dB across the affected zone. The phonautograph tracing is not yet analysed — the smoked-glass plates from the third visit await laboratory transfer. Until that reading exists, the central question is open: is this an effect on recording, or on the air?

• H1 · Broadband EM interference affecting electronics. Weakened by the analogue tape result but not excluded.
• H2 · Acoustic absorption by terrain and vegetation geometry. Plausible; requires the phonautograph plate to confirm or kill.
• H3 · Instrumentation artefact common to all four classes. Improbable, given how little they share.

Plate transfer scheduled. If H2 is confirmed, this file moves to RESOLVED and the site loses all anomalous status — an outcome the investigator regards as entirely acceptable.

Recovered from a university hydrophone array operated for cetacean research. The array had no interest in the object and no motive regarding it, which is the strongest provenance a record can have.

Multi-station time-difference-of-arrival across six hydrophones, with independent re-analysis commissioned from an acoustician given no context beyond the raw files.

• W-4 (array technician): reported the track as a fault and initiated a calibration cycle. Calibration returned nominal on all six stations.
• W-8 (blind re-analyst): derived the same track from raw data with no knowledge of the case, and independently flagged the absence of cavitation as the anomalous feature — not the speed.

Derived mean velocity approximately 310 kn. At that speed in seawater a conventional body would cavitate violently; dynamic pressure on any hull would exceed 29 MPa, and the acoustic signature should saturate the array. The record instead shows a broadband level barely above ambient. It is the quietness, not the speed, that resists explanation.

• H1 · Array timing fault. Falsified by post-event calibration and by blind re-analysis of raw files.
• H2 · Biological source. Excluded: no organism approaches this velocity by two orders of magnitude.
• H3 · Known military platform. Not excluded, but no disclosed platform reconciles the speed with the silence.
• H4 · Propagation artefact — a distant loud event mimicking a near quiet one. The most serious surviving objection. Testing requires array geometry data not yet released.

Records request outstanding for array geometry and maintenance logs.

Two hundred and six reports across eleven weeks. Geographic spread mapped against local newspaper circulation and, later, against social platform reach.

Epidemiological. Reports were treated as a transmissible unit and modelled for spread; the descriptive content was coded for drift across time and distance.

• W-6: gave a detailed account, then on re-interview at ten weeks reproduced it near-verbatim — including a phrase that first appeared in a news broadcast after her original statement.
• W-19 (child, interviewed with a specialist present): description diverged sharply from the canonical form. Notably, this is the account least contaminated by media.

Eleven photographs submitted; all resolve to reflections, condensation, or pareidolia on textured glass. The reproduction rate is zero — in two hundred and six sightings, across a season in which effectively every witness carried a camera, not one usable image exists.

• H1 · Real recurring intruder. Falsified: the spread front tracks media reach, not road networks or population density.
• H2 · Mass sociogenic transmission. Strongly supported: onset, peak, and decay fit a standard contagion curve, and descriptive drift decreases as coverage increases — the account converges because it is being copied.

W-19's divergent account is retained separately and remains open. It is the only report in the corpus that the contagion model does not explain.

Self-referred. The subject approached the institute directly and consented to a full protocol, including medical review, on the explicit condition that a negative finding would be published as readily as a positive one.

The investigator's standing rule applies: where a case concerns a living person's wellbeing, clinical assessment precedes and outranks anomalistic inquiry. Referral was made to an independent physician before any field instrumentation was deployed.

Details of the subject's clinical findings are withheld from this file. They are not the institute's to publish, and their absence here is deliberate rather than an omission. The referring physician identified a recognised, treatable cause consistent with the full symptom set, including its onset and its periodicity.

Symptoms resolved under treatment and had not recurred at eighteen months. The subject reviewed this file before archiving and approved its wording.

• H1 · Anomalous attachment. Not supported by any instrumental or documentary evidence.
• H2 · Identifiable medical cause. Confirmed independently by a treating clinician.

Where a report originates with a person in distress, the person takes precedence over the phenomenon. This is not a courtesy. It is the protocol.

Material held under a review agreement with a third party who supplied it on condition of embargo. The institute does not publish embargoed material, and does not pretend the embargo is more interesting than it is.

The file concerns instrumented data, not testimony. It has been independently duplicated and is held in two custodies. The embargo has a defined end date. No conclusion has been reached, and no conclusion is being concealed — there is not yet one to conceal.

The institute will not tease this file, will not trade on its existence, and will not permit its sealed status to imply a result. A sealed folder is an administrative fact, not evidence.

When the embargo lapses this entry will be replaced in full, whatever the outcome — including the outcome in which the data proves uninteresting.

Frames drawn from a continuous geological monitoring feed operated for eruption forecasting. The instrument had no anomalistic purpose, no operator present, and no editorial hand — the strongest class of provenance available.

Photogrammetric reconstruction using surveyed crater dimensions for scale, plus meteorological reanalysis for the interval, plus sensor-artefact testing against the preceding and following thirty days of feed.

• W-1 (volcanologist, feed custodian): reports that similar transits appear 'a few times a year' and are informally disregarded as insects or debris. His candour about the disregard is itself a finding — the base rate has never been measured.
• W-3 (imaging engineer): confirms no compression artefact class reproduces the observed trajectory.

Angular motion is inconsistent with a near-field insect at the frame rate, and inconsistent with ballistic ejecta on the return transit — ejecta do not come back. Thermal band shows no plume interaction at the crater rim. Scale remains the weak link: without a second camera the object's distance, and therefore its size, is unconstrained.

• H1 · Near-field insect or bird. Weakened by angular rate; not excluded without parallax.
• H2 · Ejecta. Excluded on the return transit.
• H3 · Aircraft or drone at altitude. Live and untested — no flight-track data has been obtained for the interval.
• H4 · Sensor artefact. Excluded by thirty-day comparison.

Priority action: install a second instrumented camera on a surveyed baseline. Parallax would settle this case in a single transit.

Six independent recordings from five surveyed positions — an unusually strong geometric dataset, and the reason this case could be closed rather than argued.

Triangulation from the five positions to fix altitude and spacing, followed by comparison against launch registries, satellite deployment catalogues, and local event permits for the date.

• W-2: described the objects as 'holding station against the wind'. Triangulation shows they were in fact drifting downwind at 4.1 m/s, matching the surface wind. The perception of stillness is a known parallax effect at that range.
• W-5: correctly estimated altitude to within 90 m — the only witness to do so, and a glider pilot.

Triangulated altitude 310–420 m, spacing 40–80 m, drift vector matching the recorded wind in both bearing and speed. A municipal permit records a memorial sky-lantern release 6.2 km upwind, forty minutes prior. Predicted drift for that release places the lanterns within the observed volume at the observed time.

• H1 · Sky lanterns. Confirmed on three independent grounds: drift, permit, and predicted position.
• H2 · Satellite train. Excluded — orbital objects do not drift with surface wind.
• H3 · Aircraft formation. Excluded by altitude and by silence.

None. Retained as the clearest demonstration in the archive that multiple independent witnesses can be simultaneously honest, mutually corroborating, and wrong.

Commercial sounder log, surrendered voluntarily. The crew reported it to the harbourmaster before contacting anyone else and sought no attention — a provenance pattern that correlates strongly with reliability.

Raw echo-sounder data requested from the manufacturer in native format rather than working from the screen capture already circulating. Marine biologists were consulted blind.

• W-1 (skipper, 31 yrs): 'I've seen every fish this boat can find. That wasn't a fish and it wasn't the bottom.' Experience is a genuine asset here, though the investigator notes it also produces confident misidentification.
• W-4 (deckhand): reports the vessel's compass swinging during the interval. No independent record exists. Uncorroborated.

Native-format data has been received but not yet processed — the manufacturer's decoder is under licence negotiation. A blind marine biologist consultation on the screen capture alone returned 'possible dense bait ball, possible artefact, insufficient to say', which is the correct answer to an insufficient dataset.

• H1 · Dense biological aggregation. Live, and the leading candidate on base rates alone.
• H2 · Sounder artefact or second-return echo. Live; testable once native data is decoded.
• H3 · Submerged vehicle. Live but unfavoured — no feature of the record requires it.

The compass report is the detail most likely to be quoted and least likely to be true. It is uncorroborated and should not be repeated as fact.

Traced to a single 1974 account, itself a near-verbatim reproduction of a short story published in a regional magazine eleven months earlier. The magazine issue was located in a county library holding.

Textual comparison of the earliest account against the published fiction, with independent literary analysis of shared distinctive phrasing.

• W-1 (originator, interviewed at 81): recalls the story vividly as personal experience. Presented with the magazine, he read it with visible distress and said he had subscribed. The investigator assesses this as genuine source-monitoring failure, not deception, and states so plainly because the distinction matters to the man.
• W-6, W-9: later accounts, both post-dating regional broadcast coverage.

Nine distinctive phrases appear in both texts, including one internally inconsistent detail that makes sense only as a printing error in the original fiction. That error propagates through every subsequent retelling — a textual fingerprint.

• H1 · Genuine event later fictionalised. Falsified by publication date.
• H2 · Fiction absorbed as memory, then transmitted. Confirmed by the propagating printing error.
• H3 · Deliberate fabrication. Rejected. Nothing supports it, and the investigator declines to imply it.

The originator is owed the record that he did not lie. This file states it.

Four affected properties, twenty-two animals over one season. Referred by a landowners' association that had already reached its own conclusion — a contamination risk the protocol had to control for.

Veterinary pathology on every recoverable carcass, timed decomposition study, invertebrate and scavenger survey, and camera traps at three sites. Pathologists were given no context.

• W-3 (landowner): 'Clean cuts. Surgical. No blood.' Reported in good faith and, in the investigator's judgement, describing exactly what he saw.
• W-8 (veterinarian, 24 yrs large-animal practice): identified the presentation immediately and without prompting.

All twenty-two presentations are consistent with post-mortem scavenging. Soft tissue is removed first by birds and small mammals, producing margins that appear cut because tissue tension retracts them; blood settles and pools ventrally, so the ground appears dry. Camera traps recorded the full sequence at two sites, start to finish.

• H1 · Scavenger predation post-mortem. Confirmed on camera at two independent sites.
• H2 · Human actor. Excluded — no tool marks under magnification, no access traces, no motive pattern.
• H3 · Anomalous cause. Falsified. Nothing in the pathology requires or suggests it.

Findings were presented to the affected landowners in person before publication. Three accepted them. One did not, and that is recorded here without comment.

Detected by an amateur optical tracking network across four continents, then independently confirmed by a fifth station. Amateur networks are habitually dismissed; this one publishes raw plate solutions, which makes its work more auditable than most institutional feeds.

Orbital elements derived from 340 observations across nine months, cross-referenced against public catalogues and against historical launch manifests back to 1961.

• W-2 (network coordinator): volunteered the object's most likely prosaic identity before being asked — an upper-stage fragment from an unlogged breakup. The investigator regards unprompted self-scepticism as a mark of a reliable source.
• W-7: reports a brightness periodicity of 4.1 s, consistent with an irregular tumbling body. Consistent, that is, with debris.

Orbit is stable, near-circular, moderately inclined, with a ballistic coefficient implying low area-to-mass — which argues against light fragmentary debris and toward something more compact. No catalogue entry matches. Two historical breakups could plausibly account for it, and neither has published fragment elements.

• H1 · Uncatalogued debris from a known breakup. The leading hypothesis and by far the most probable. Untestable without fragment data that has never been released.
• H2 · Undisclosed operational payload. Consistent with the ballistic coefficient; consistent also with a great many mundane objects.
• H3 · Anything more exotic. No feature of the record calls for it, and the investigator lists it only to record that it was considered and set aside.

Records request outstanding for fragment elements from both candidate breakups. The investigator fully expects this file to close as debris.