Climate prediction occupies a regime intermediate to weather prediction, which is primarily an atmospheric initial value problem having a prediction horizon of a week or two, and future climate projection on time scales of decades to centuries, for which initial conditions are largely irrelevant. Climate prediction exploits the predictability in slowly varying components of the climate system, primarily the oceans, to forecast future climate evolution from observed initial states. Such forecasts have become routine at ranges of 1 to 12 months, and provide useful predictions of ENSO and its effects for example. The possibility that longer-term predictions spanning years to a decade might also be useful, e.g. through their ability to forecast changes in the Atlantic meridional overturning circulation, is suggested by climate predictability studies. However, attempts to produce such decadal forecasts have begun only within the past few years. This talk will describe the physical basis for decadal prediction, its origins and subsequent development, focusing on CCCma's contributions to this emerging field.