We study the responsiveness of rice production to climatic variation using a recent pseudo-panel data set from Sri Lanka. Output per farm is modelled as a non-linear function of temperature and rainfall (and other standard controls) using fixed effects regressions. We find that both climatic variables have concave, non-monotonic effects upon production, and that output is close to maximized at current climatic values. This implies that variations in growing conditions are likely to have negative effects upon production. Random simulations are used to model these impacts under various climate change scenarios, and we find that increasing temperatures will adversely affect rice production much more than varying rainfall, although the effects of a small ceteris paribus rise in temperature are positive. As rice production is a key component in economic output for Sri Lanka and other developing countries, our results have implications for food policy and poverty management in the future.