# Source-checking progress: pass 2 > Historical pass snapshot. Current status and methods: README.md; current measurement audit: measurement_audit_pass4.csv. Counts and pending items below describe this pass at the time, not the latest state. 23 October 2026. Working audit; final sample not frozen. ## Current files Read this update alongside README.md. The original pass-1 outputs remain as provenance. For current working values use measurement_audit_pass2.csv, crosswalk_pass2.csv and audit_summary_pass2.json. The 19 additional row-level decisions, their evidence links and limits are in source_decisions_pass2.csv. The queue is an inventory of 118 candidate measurements across 92 taxon labels, not 92 approved species. It includes rows already checked in part but still requiring eligibility or provenance work. ## Changes applied - Dog: source confirms domestic beagles. The corrected candidate joins to AnAge Canis familiaris, 27 years, acceptable, instead of Canis lupus, 20.6 years. Exact 84 Hz extraction and a defensible domestic-dog tree placement remain pending. No tree tips were silently renamed. - Cat: the compilation labels all four rows House cat. The appropriate candidate is AnAge Felis catus, 30 years, questionable, not wildcat Felis silvestris, 19 years, acceptable. Full-text source population remains to be confirmed. This removes the cat from the acceptable/high candidate tier, reducing that tier from 95 to 94. Neither number is a final eligible sample. - Cat measurements: 75 Hz is from a blood-flow study, 70 Hz from intracellular horizontal-cell recording, and 40 Hz from cortical recording. The last was incorrectly labelled Behavioural. These three endpoints are excluded from the behavioural and whole-retina ERG strata. The 47.5 Hz behavioural study remains separately traceable. - Painted turtle: Graf 1972 is behavioural psychophysics. Corrected Electrophysiology to Behavioural. Species identity is supported; the compiled 31 Hz needs full-text extraction. Replaced reliance on the unrelated iNaturalist link with the original article DOI in the evidence field. - Slider turtle: the original study names Pseudemys scripta and the taxonomic source supports Trachemys scripta. The corrected biological name is supported; the raw misspelled tree label is retained for traceability. The exact 52.6 Hz endpoint remains unverified. - Blue tit: confirmed Parus caeruleus/Cyanistes caeruleus identity and 130.3 Hz source mean. - Chicken (Lisney 2012) and tree shrew (Tigges 1967): both are ERG studies incorrectly labelled Behavioural. Corrected their audited method. Exact numeric extraction remains open. - Lionfish: 88 Hz is explicitly a median at maximum stimulus intensity, not a species mean. The inspected study reports Pterois spp.; a P. volitans-only longevity pairing remains withheld. - Honeybee: primary source identifies approximately 200 Hz movement fusion with rotating patterns. Kept separate from uniform-field flicker thresholds. Caste compatibility remains unresolved; no queen/worker longevity substitution was made. - Loggerhead and leatherback: the primary abstract describes frequency-following responses extending to about 15 Hz and not above 10 Hz respectively, unlike the compiled 40 and 15 Hz. Both rows are held. These response limits were not converted into invented exact CFF means. Their hatchling life stage and isolated-retina preparation are recorded. - Piglet: neonatal treatment study confirmed. The compiled 60 Hz is not validated as a fusion threshold by the accessible abstract. Held pending control-group and endpoint extraction. - Cane toad: rod-photoreceptor endpoint separated from whole-retina ERG. Historical synonym is supported but source locality/species-complex identity still requires care. - Green frog: nomenclatural connection supported. Original experiment is still needed; the compilation points to a review and its human-effects table. - Pigeon (Delius 2017): genuine behavioural threshold experiment confirmed. The symmetry-focused title is not grounds for exclusion; numerical table extraction and protocol selection remain pending. ## Consequences for methods The original broad method column cannot be used without source checks. Four explicit method corrections in this pass are painted turtle to Behavioural; cat Rager 1998 to cortical electrophysiology; chicken Lisney 2012 and tree shrew Tigges 1967 to retinal ERG. Other electrophysiological rows have been assigned more specific endpoint classes where supported. The original quality-rated overlap contained 40 rows labelled Behavioural across 30 taxa. This was a compiled-method inventory, not verified sample size. Exact uniform-field flicker, patterned movement fusion, intracellular responses, blood flow and cortical entrainment should remain distinct. A loss of sample size is preferable to silently merging incompatible endpoints. Retain reported medians as medians. The lionfish result highlights that a field called CFF can mix means, medians, individual extremes and stimulus limits. The working study-aggregation rule in README.md applies only after identifying compatible summaries; unverified statistic type must not be treated as a mean by default. Developmental studies require a stated life-stage scope, not automatic substitution of hatchling or intervention measurements for normative adult values. ## What is and is not resolved 19 additional measurement rows now have explicit source-based decisions or documented holds. These are not 19 fully validated records. Source identity, numeric extraction, endpoint comparability, longevity population and phylogenetic mapping are separate checks. All 280 source rows remain in the audit; nothing was silently deleted. No new CFF numeric values were changed in pass 2. The earlier two Ficedula corrections remain. The 94 acceptable/high candidates include taxonomically or methodologically unresolved records. The queue contains 92 taxa because current endpoint exclusions remove the sole available records for the cane toad and honeybee. It still includes ERG-only taxa, unresolved identities, partial checks and uncertain developmental scope. The questionnaire-like phrase final_analysis_eligible remains not_finalized throughout. Full-text extraction remains necessary for several older studies. Graf 1972 exposes a publisher access barrier; multiple indexed abstracts verify method but cannot support exact numeric claims. Many other rows have not yet received source review. The next work is to extract the remaining primary-source values and conditions, beginning with behavioural candidates and shared sources, then audit ERG candidates separately. Do not move to preregistration or final regression on the basis of this queue. Human independent review remains pending. No emails, registration, final model, manuscript revision, Zenodo replacement or website deployment occurred. ## Reproduction 1. Run `python apply_audit.py INPUT_DIRECTORY OUTPUT_DIRECTORY`. 2. Run `python continue_audit.py INPUT_DIRECTORY OUTPUT_DIRECTORY`. The scripts preserve raw inputs and assert row counts, the dog join, corrected methods and candidate counts. Evidence URLs are recorded per decision. Biological interpretation remains subject to human review; assertions test consistency, not scientific truth.