Extracts latitude/longitude for 155 of 164 Wuhan-resident December-2019 cases from maps in the WHO mission report, then tests their spatial distribution against a population-density-weighted null (Haversine distance to market vs. 1,000 age/density-matched random draws from worldpop.org data). Finds lineage-B cases (n=11, median distance 8.30 km, p=0.017) and the two earliest-known lineage-A cases (2.31 km, p=0.034; and a case adjacent to a hotel within ~500 m of the market) are significantly closer to the Huanan market than chance predicts, for both viral lineages independently. Separately documents live SARS-CoV-2-susceptible mammals (raccoon dogs among them) sold at the market with environmental-sample positives concentrated in the same stalls, and concludes the market was the pandemic’s early epicenter via the wildlife trade. Subject to a 2024 erratum (correction, not retraction).
relevance_note: the primary geospatial case-clustering analysis the “market epicenter” claim rests on; the direct target of the Stoyan/Chiu statistical critique and the Weissman ascertainment-bias critique in this same slice.
Results — case geography
O-30 - December-2019 Wuhan cases were significantly closer to the Huanan market than a population-weighted null
Case residences (155 of 164 Dec-2019 cases geolocated from WHO-report maps) were tested against 1,000 age/population-density-matched random draws from WorldPop, using Haversine distance to the market. The whole early case cloud centres on the market.
Link to original
O-31 - Both lineage A and lineage B early cases cluster near the Huanan market
The lineage-confirmed subsets are small (n=2 and n=11) and the p-values near threshold (0.034, 0.017); a 2024 erratum corrected the null-generation procedure. The claim rests on both independently-introduced lineages centring on the same market.
Link to original
A-43 - Both lineages independently centering on the market points to the market as the common emergence locus
Under a single point-source emergence at location X, the earliest cases of each introduced lineage should cluster around X. Observing that both lineage A (earliest cases) and lineage B (n=11) independently cluster near the same market is therefore expected if X = Huanan market, and requires a coincidence otherwise. A lab-origin scenario would instead predict the earliest cases to centre near the implicated institute (several km away across the river), not the market. Validity caveats (for scoring): the lineage-confirmed subsets are very small (n=2, n=11), the p-values are near threshold (0.034, 0.017), the case set is ascertainment-limited, and a 2024 erratum revised the null-model generation, all of which the Stoyan-Chiu and Weissman critiques press on.
Validity verdict
status: approved; reason_if_not_false: checked. Premises (granted, not tested here): two genuinely independent introductions (lineage A and lineage B) whose earliest cases each cluster near the market. Load-bearing step: under point-source emergence at X the earliest cases of each lineage cluster at X, so both lineages centering on the market is expected iff X = market and requires a coincidence otherwise, disfavouring a distant single origin including a lab. Probed the main undercutting defeater — that the market is a shared amplification venue rather than the emergence locus (the A-41/A-42 theme), so both lineages could center on it while emerging elsewhere. This defeater does not survive for the two-independent-introductions case: it requires a distant source to seed the market twice independently, which is itself improbable, whereas two independent zoonotic spillovers at an active wildlife market are natural; hence the step holds conditional on the independence premise. Note the listed caveats (small n, near-threshold p, ascertainment, erratum) bear on premise truth/strength and are priced at step 8, not here. The conclusion is already correctly hedged (“points to”/“disfavors”), so no correction needed. Approved, checked.
Link to original
O-32 - Market-unlinked December cases lived even closer to the market than market-linked cases
Surprising and load-bearing: the people with no known market connection were the ones living nearest the market, which the authors use to argue the clustering is not simply an artifact of investigators preferentially finding market-linked cases.
Link to original
A-44 - Market-unlinked cases clustering on the market argues the signal is not mere market-link ascertainment
If the market clustering were purely an artifact of case-finding that sought market connections, then cases lacking any market link (found by other routes) should not centre on the market; instead they centre on it even more tightly. The authors add that the spatial relationship persists after removing the two-thirds of unlinked cases residing nearest the market. This makes market-link ascertainment an implausible sole explanation for the clustering. Validity caveat (for scoring): critics (Stoyan-Chiu; Weissman’s proximity-ascertainment-bias argument) contend the residential locations themselves are ascertainment-biased - early case-finding concentrated on central-Wuhan hospitals near the market - so a residual location bias could still inflate the signal even among unlinked cases.
Validity verdict
status: approved; reason_if_not_false: checked. Premise: unlinked December cases lived even closer to the market than market-linked cases. Load-bearing step: if the clustering were purely an artifact of case-finding that sought market connections, cases lacking any market link (found by other routes) should not center on the market; they center on it even more tightly (and the relationship persists after dropping the two-thirds of unlinked cases nearest the market), so market-link ascertainment is an implausible sole explanation. Traced directly — this is a clean modus-tollens on the specific market-link mechanism, valid conditional on the premise. Probed the acknowledged Weissman/Stoyan-Chiu defeater (residential-proximity ascertainment: case-finding concentrated at central hospitals near the market): this is a distinct mechanism from market-link ascertainment and does not deny the stated conclusion (which is scoped to “market-link” ascertainment); it is a separate, upstream bias priced against the relevant observation at step 8. The conclusion is correctly scoped and hedged (“not mere”/“supporting”), so no correction. Approved, checked.
Link to original
Results — market environmental mapping
O-33 - Positive market environmental samples concentrated at a single live-mammal vendor stall
Worobey et al. re-mapped the China CDC environmental positives onto stall locations and found the earliest viral contamination physically co-located with a live-mammal-selling stall in the SW corner of the market.
Link to original
A-45 - Positive samples localizing to a live-mammal stall ties earliest contamination to the wildlife trade
If wildlife-trade spillover seeded the outbreak, viral contamination in the market should localize to stalls trading live susceptible mammals. Worobey et al. find the positives concentrate (relative-risk p<0.05) at a SW-corner live-mammal stall, matching that prediction and making a diffuse or non-wildlife source less likely. Validity caveat: environmental positivity marks where virus was shed, not the shedding species, so an infected human working that stall remains an alternative source of the same localized signal.
Validity verdict
status: approved; reason_if_not_false: checked. Premise: positive environmental samples concentrate (RR p<0.05) at a single SW-corner live-mammal vendor stall. Load-bearing step: contamination localizing to a live-mammal stall matches the wildlife-spillover prediction and raises the probability that emergence occurred via the live-wildlife trade. The conclusion is explicitly probabilistic, which is what lets it survive the acknowledged undercutting defeater (environmental positivity records where virus was shed, not by which species — an infected human working that specific stall could produce the same localized signal). That defeater weakens but does not break the step: it requires the human shedder to coincide with the one live-mammal stall out of the whole market, a coincidence, so localization at precisely the wildlife stall still shifts the likelihood ratio toward wildlife-trade emergence versus a diffuse or human-only source. Because the claim already stops at “raising the probability” (not “establishes”), no correction to a weaker form is needed. Approved, checked.
Link to original
Discussion
H-21 - The Huanan market was the early epicenter of the pandemic via the live-wildlife trade
The paper’s headline conclusion, resting jointly on the case-geography clustering (D-2) and the environmental positives at a live-mammal stall (D-1). It is the direct target of the Stoyan/Chiu and Weissman methods critiques in this slice.
Link to original