General inference: when case-finding preferentially searches for a particular exposure, the measured prevalence of that exposure among detected cases is inflated relative to the population of true cases, because cases lacking the exposure are less likely to be found or classified. Here the surveillance mechanism and field questionnaire explicitly targeted market/wet-market exposure, and the authors themselves caution that the recent-onset decline reflects under-ascertainment rather than a true incidence turn. Therefore both the high early market fraction and the all-market first-four-cases fact are partly artifacts of where investigators looked, weakening their use as unbiased evidence for a market origin. This does not show the market was not the source; it caps how much the raw fractions can bear.
Step 6 verdict — approved (checked)
Reconstruction. Premises: (i) case-finding preferentially searched each early case for market/wet-market exposure; (ii) a case lacking that exposure is therefore less likely to be detected/classified as a case; (iii) the early curve is under-ascertained (authors’ own caveat). Load-bearing step: differential ascertainment that favours the exposure inflates that exposure’s measured prevalence among detected cases above its true prevalence among all cases.
Evaluation. This is the standard selection-bias result and it is valid provided premise (ii) — not merely that investigators asked about the market, but that non-market cases were less findable. The body supplies exactly that (surveillance/case definition keyed to market contact), so the enthymeme is completed without force; this is also the documented mechanism (early case definitions incorporating market/Wuhan links). Conditional on the premises the upward bias on both O-7 and the all-market first-four (O-8) follows. The conclusion is already correctly capped (“does not show the market was not the source”), so no undercutting defeater pushes it to a still-weaker form. Traced myself; author-blind.