Weak structure and search sensitivity in the Big Ear Wow! signal archive

The weaker structures surrounding an isolated radio transient provide both an empirical control population and a test of
search sensitivity. We analyse an audited transcription of the complete 1977 August 15–16 Big Ear Wow! observing run:
5,974 records and 298,700 printed detector codes. A catalogue-free ranking defines 25 weak candidates. The strongest
unmarked feature, in channel 12, resembles a targeted next-night window at the same printed right ascension (r = 0.853),
whereas the same-clock comparison gives r = −0.458. The fixed-coordinate correlation remains at least 0.734 after
deletion of any five consecutive comparison points. A broad channel-20 structure also recurs, and 20 of 25 candidates
sample substantial external H I line brightness across nine tested pointing scenarios. Neither resemblance nor survey
association uniquely identifies a physical source. Separately, a search of selected windows on 12 surrounding dates,
comprising 3,836 records, yields no verified recurrence of Wow!. Thresholds fixed before transcription of five additional
dates produce no target trigger on those dates. Record-domain injection tests show why that null result has limited force:
conditional recovery at simulated amplitude 10 is 96 per cent in one channel but 37 per cent for an equal two-channel split
in the original seven-date experiment. A new paired width test, with unchanged thresholds, reduces recovery from 94
to 48 per cent for a shorter single-channel pulse and from 47 to 2 per cent for the corresponding split pulse. These are
model-dependent algorithm tests, not calibrated flux limits. The contribution is a reproducible weak-feature comparison
and an explicit account of archival search incompleteness; the origin and recurrence of Wow! remain unresolved.