The characteristics of tree rings as natural archives of past climate are discussed. Special consideration is given to key issues affecting their robustness and reliability as sources of information on past climate. These issues include: the effects of sample design and in particular the importance of using networks of tree-ring records from many locations whenever possible; potentially complementary approaches to the identification of climate signal in tree rings, namely empirical-statistical and process-modeling approaches; statistical and mechanistic stability over time of the climate signal in tree rings; and the ongoing effort to isolate climate signal from noise, without introducing biases, in tree-ring based proxy records of climate.