Computer-based global climate models are important research tools for the study of global climate change, variability, and predictability. A major source of model dependence in simulation results, however, is the different ways that models simulate land-atmosphere interaction. The Global Land-Atmosphere Coupling Experiment (GLACE), a joint project of the GEWEX Global Land Atmosphere System Study (GLASS) and the CLIVAR Working Group on Seasonal-to-Interannual Prediction (WGSIP), is a multi-model inter-comparison experiment that aims to quantify the strength of land-atmosphere coupling in different global atmospheric models used for weather and climate studies. A pilot study with four atmospheric general circulation models (AGCMs) intriguingly suggested that coupling strength does vary widely across AGCMs. In GLACE, the degree of land-atmosphere coupling strength in a larger number of AGCMs is being examined. Importantly, GLACE also addresses the relative impact of fast and slow land surface states (e.g., surface temperature and deep soil moisture) on the evolution and predictability of weather.