This study examines the soil moisture-precipitation coupling during summer over sub-regions of the U.S. based on observations (CPC-VIC), reanalysis data (NARR and CFSR) and output from two GCMs and one RCM (CAM3-CLM3, CAM4-CLM4, and RegCM4-CLM3.5). The probability density function of conditioned correlation between soil moisture and subsequent precipitation or surface temperature over the years of large precipitation anomalies is used as a measure for the coupling strength. The lagged correlation is stronger in observational and reanalysis data than in numerical models, which contradicts previous suggestions that numerical models overestimate the land-atmosphere coupling strength. The relative comparison of lagged correlation among different numerical models are consistent with the relative comparison of coupling strength based on GLACE type of experiments using these models, which validates the use of lagged correlation as a representation of soil moisture-precipitation coupling strength. The role of scale in quantifying the land-atmosphere coupling strength will be also discussed.