Soil heterotrophic respiration (HR) is an important source of soil-to-atmosphere CO₂ flux, but its response to changes in soil water content (θ) is poorly understood. Earth system models commonly use empirical moisture functions to describe the HR–θ relationship, introducing significant uncertainty in predicting CO₂ flux from soils. Generalized, mechanistic models that address this uncertainty are thus urgently needed. Here we derive, test, and calibrate a novel moisture function, f_m, that encapsulates primary physicochemical and biological processes controlling soil HR. We validated f_m using simulation results and published experimental data, and established the quantitative relationships between parameters of f_m and measurable soil properties, which enables f_m to predict the HR–θ relationships for different soils across spatial scales. The f_m function predicted comparable HR–θ relationships with laboratory and field measurements, and may reduce the uncertainty in predicting the response of soil organic carbon stocks to climate change compared with the empirical moisture functions currently used in Earth system models.