This paper evaluates the simulation of precipitation in four successive versions of the UK Met. Office/Hadley Centre global climate model (GCM) that have been used to perform climate change experiments. The precipitation fields over land from the control integrations of these four experiments are compared with observed precipitation on a global scale and for selected regions. Globally, improvements in simulating mean monthly precipitation patterns occur with successive simulations. The poorest correlations and highest root mean square errors occur in the boreal summer months. A major improvement in the simulation of the South Asian summer monsoon by the most by the most recent versions of the model is noted. The pattern of temporal variability in the later versions over global land areas is generally well simulated. The observed pattern of temporal correlation between the Southern Oscillation Index and precipitation anomalies has improved in the latest model version. ¿ American Geophysical Union 1996