Soil respiration (R_s) is the second-largest terrestrial carbon (C) flux. Although R_s has been extensively studied across a broad range of biomes, there is surprisingly little consensus on how the spatiotemporal patterns of R_s will be altered in a warming climate with changing precipitation regimes. Here, we present a global synthesis R_s data from studies that have manipulated precipitation in the field by collating studies from 113 increased precipitation treatments, 91 decreased precipitation treatments, and 14 prolonged drought treatments. Our meta-analysis indicated that when the increased precipitation treatments were normalized to 28% above the ambient level, the soil moisture, R_s, and the temperature sensitivity (Q₁₀) values increased by an average of 17%, 16%, and 6%, respectively, and the soil temperature decreased by −1.3%. The greatest increases in R_s and Q₁₀ were observed in arid areas, and the stimulation rates decreased with increases in climate humidity. When the decreased precipitation treatments were normalized to 28% below the ambient level, the soil moisture and R_s values decreased by an average of −14% and −17%, respectively, and the soil temperature and Q₁₀ values were not altered. The reductions in soil moisture tended to be greater in more humid areas. Prolonged drought without alterations in the amount of precipitation reduced the soil moisture and R_s by −12% and −6%, respectively, but did not alter Q₁₀. Overall, our synthesis suggests that soil moisture and R_s tend to be more sensitive to increased precipitation in more arid areas and more responsive to decreased precipitation in more humid areas. The responses of R_s and Q₁₀ were predominantly driven by precipitation-induced changes in the soil moisture, whereas changes in the soil temperature had limited impacts. Finally, our synthesis of prolonged drought experiments also emphasizes the importance of the timing and frequency of precipitation events on ecosystem C cycles. Given these findings, we urge future studies to focus on manipulating the frequency, intensity, and seasonality of precipitation with an aim to improving our ability to predict and model feedback between R_s and climate change.