Abstract Sea surface temperatures (SSTs) in the tropical Pacific vary as a result of the coupling between ocean and atmosphere driven largely by El Niño–Southern Oscillation (ENSO). ENSO amplitude is known to vary on long time scales, which makes it difficult to quantify its response to climate change and constrain the physical processes that drive it. To characterize the long-period variability in ocean–atmosphere coupling strengths, a linear regression of local SST changes is applied to the 4000-yr GFDL Climate Model, version 2.1 (CM2.1) and the 500-yr GFDL CM2 with Modular Ocean Model version 4p1 (MOM4p1) at coarse resolution (CM2Mc) preindustrial control runs, while also comparing to the observationally constrained Ensemble Coupled Data Assimilation (ECDA) dataset. The models produce regression coefficients that vary widely on multidecadal time scales. These variations are strongly reflected in the long-period modulation of ocean stratification and surface precipitation. During high variance periods, when there is stronger stratification and precipitation in the central equatorial Pacific, the ocean’s surface is less responsive to zonal wind stress perturbations, while the atmosphere is more responsive to SST perturbations. The mechanisms underlying this behavior are examined through an expansion of the linear regression equation to individual temperature tendency components. Long-term changes in ENSO amplitude are due to changes in both the oceanic response to the atmosphere, which is predominantly driven by regional changes in the advective and vertical diffusive heat tendencies, and the atmospheric response to the ocean.