A Pre-Registered, Model-Independent Test of Sign Coherence Between Extreme Large-Scale Structure and CMB Temperature Residuals

Abstract We present a fully pre-registered observational test isolating a logically prior empirical question in studies of Cosmic Microwave Background (CMB) secondary anisotropies: whether extreme large-scale matter density contrasts necessarily imprint CMB temperature residuals of matching sign, independent of amplitude modeling or assumed physical mechanism. Using a fixed, blind target selection procedure applied to publicly available large-scale structure catalogs, we define a sample of ten extreme structures—five overdensities (superclusters) and five underdensities (supervoids) at redshift z < 0.15. We evaluate a single binary observable: the sign of a local CMB temperature contrast measured via locked aperture photometry. All analysis choices, including selection rules, apertures, masking, null simulations, and statistical decision criteria, were fixed prior to inspection of CMB data. All ten targets exhibit temperature residuals of the predicted sign. No realization among 10,000 full-sky Monte Carlo null simulations reproduces this outcome, yielding an empirical bound p < 10⁻⁴. We further demonstrate robustness under changes in CMB map realization and sky masking, and show via a scaled-sample expansion that sign coherence saturates for the most extreme structures and dilutes smoothly as less-extreme objects are included. The result does not falsify ΛCDM and does not require new physics. Instead, it establishes a logically prior constraint that any physical explanation of CMB–large-scale-structure correlations must satisfy. If incorrect, the result fails decisively under the pre-registered null tests defined herein.