We propose a study of the mechanisms and predictability of interannual to interdecadal climate variability. Our general goals are to to understand the underlying mechanisms for climate variability on these timescales, to identify processes that might lead to predictability, and to understand what the intrinsic limits to climate predictability are. Our main tool is a novel hierarchy of climate models, developed using the Flexible Modeling System at GFDL. At the top of the hierarchy is the state-of-the-art, IPCC-class model at GFDL. Our models directly connect to that, but are simplified by using simpler and more economical physics packages, and/or by making simplifications in the geometry. Use of such models allows more experiments to be performed, including ensemble experiments, and mechanisms to be identified. We will focus on extra-tropical variability in the Atlantic sector, including the interannual and decadal variability of the NAO, although somes aspects of the proposed work are more general. The specific topics we propose to investigate include the timescales on which the atmosphere ocean system may be regarded as truly coupled, the timescales on which the atmosphere forces the ocean, the generation and persistence of sea-surface temperature anomalies and the effects of such anomalies on the atmosphere. As appropriate, we shall use still simpler theoretical tools and analyses to try to abstract the mechanisms to their esssentials. We shall also compare and validate our models against the full coupled climate model to ensure that we are in a realistic parameter regime that is relevant to reality.