Year-to-year temperature responses of 320 North American bird species followed 5-14% of own spatial gradients, with no clear lasting part; 49 British butterfly species followed about 50% in 2 summers

Question: community temperature indices follow only a small part of warming. Which part of the species-level response is missing? Data: North American Breeding Bird Survey 2026 release (1966-2025; 320 species, 4,802 routes, 9.1 million species-route-year counts), UK Butterfly Monitoring Scheme site indices (1976-2024; 49 species, 2,410 British sites), GHCN_CAMS and E-OBS temperature. Method: for each species, the spatial gradient g is the slope of route mean log abundance on route climatology at each route. Per-species fixed-effects models (route x observer or site, and species x year) related log counts to local temperature anomalies of the same season, four previous summers, four winters, and placebo seasons after the count. The ratio lambda = temporal response / g is 1 for full space-for-time equivalence. We refitted all species without each of 10 year blocks (jackknife). Results: in birds, the long-run lambda was 0.14 (95% CI 0.05-0.24) with linear detrending, but the placebo next-summer term was as large as the previous-summer term. With 11-year high-pass anomalies the placebos were near 0, the same-season term stayed at 0.045 (0.017-0.074), and the lagged terms gave 0.014 (-0.10 to 0.13). Within species, the response followed 14-20% of the gradient (intervals exclude 0); between species, mean responses did not increase with mean gradients (-0.13 and -0.31). British butterflies followed 0.44 (0.14-0.75) and 0.54 (0.04-1.05) of their gradients within the same and previous summer, five to ten times the bird value, with clean placebos; longer lags and winters added nothing clear, and their between-species part was not determined. Neither year-to-year sensitivities nor spatial gradients forecast species trends (bird Spearman -0.00 and 0.09; out of sample 0.03; butterflies -0.27 and 0.18). The implied bird CTI change was about one third of the same-species spatial expectation and about 0.002 C per decade. Conclusion: weak bird tracking is not mainly a long lag; the clear response is small and transient (possibly part detection), and species that spatial gradients predict to gain do not gain. Butterflies follow temperature faster. Limitations: correlational design; no detection model; gradients include non-thermal factors; the high-pass variant was added after the main fits; butterfly long-run totals are imprecise.