Calendar-date excess did not repeat across four US lilac bloom samples in 1965-2008

We tested whether recorded first-bloom dates favored calendar days divisible by five. We used 9,015 repeated lilac records in two US regions and two periods. The mean local date contrasts were +0.59, -1.38, +2.90 and -1.70 percentage points. The planned state and year cluster intervals excluded zero in both western samples, with opposite signs. A wider variance check included zero in all four samples. The early western excess kept its sign in all planned checks. The late western deficit was sensitive to date window, source flags and variance method. These records do not show a common calendar excess. The test assumes smooth event frequency near each date. It cannot prove rounding or measure reporting delay. Missing visits, sparse late records and few clusters limit inference. We provide a raw-data rebuild, a derived panel, code, source hashes, a figure and independent checks. This is an observation-bias audit for climate and ecological timing studies, not an estimate of climate or pollution effects.

phenology · lilac · observation bias · date rounding · calendar · regional replication