Butterfly records had a lower weekend share than cherry records in southern Japan in 1988-2020

We tested whether butterfly first sightings had a lower weekend share than cherry first flowering at the same Japan stations. We used 4,446 paired station-years at 74 stations in 1953-2020. Four separate region and period tests gave each year equal weight. Butterfly minus cherry contrasts were +5.35, +1.64, +3.64 and -5.57 percentage points. Only the later southern test had a negative interval after a four-test correction: -9.16 to -1.99 points. Its corrected calendar-phase p value was 0.0438. Thus, a weekend deficit did not repeat across the four samples. Equal station weights kept the later southern estimate negative, but its corrected phase p value was 0.0874. In added setting checks, later on-site contrasts were -15.00 points in the North and -18.84 in the South. Nearby contrasts were uncertain. These selected setting samples have different support. The later southern deficit came mainly from Saturday dates. The results support an audit of observation schedules. They do not identify actual visits, biological delay, or a climate or pollution effect. Unknown effort, shared records, seasonal differences, and calendar assumptions limit inference. Code, source event tables, paired data, exact hashes, uncertainty checks, figures and an independent reconstruction permit offline reproduction.

phenology · observation bias · butterfly · cherry · Japan · weekend · calendar phase · species timing · paired observations