Song frequency had no clear link to tenure curvature in 283 North American bird species in 1966-2025

Question: does the reported late tenure loss of high-pitched birds have curvature that remains when observer cohort intercepts are free? We fitted new conditional Poisson models to 130,303 standard Breeding Bird Survey runs on 5,319 routes in 1966-2025. The model has free observer-route and year effects and a tenure spline. We test f(25) - 2 f(15) + f(5), which compares two ten-year log-count changes and cancels any linear drift. Of 446 species, 394 had stable primary fits. In 283 species outside the visual comparison group with frequency data, the slope of this contrast on frequency was +0.0084 log units per kHz (95% spatial-block interval -0.0063 to +0.0231). With family effects it was -0.0062 (-0.0271 to +0.0147). Eight species at 7 kHz or more had a mean contrast of -0.0129 (-0.1176 to +0.0919). Survey controls, region by year effects and log-count fits on a common set gave slope intervals that included zero. Some groups had curvature, but it did not decrease in order with frequency. Exact checks show that a linear tilt can change an endpoint loss while every fitted count stays fixed. The null frequency result cannot exclude linear hearing loss or identify its correction to climate trends. Tenure is not age, high-frequency species are few, unstable fits are excluded, and cohort-specific trajectories can violate the additive model. Code, source hashes, aggregate evidence, figures and numerical checks are included. Raw and record-level data stay outside the bundle.