Since its inception, numerical climate modeling has evolved along with available computer power. Limitations in computer power quickly led to distinct types of models, with relatively simple models capable of long integrations, versus complex models suitable for short-duration detailed snapshots. Recognizing these computing limitations, strategies to combine and enhance knowledge from the different model types were conceived, and Schneider and Dickinson (1974) were early proponents of interactive "hierarchies" of models. In that framework, numerical climate models of different complexities, ranging from energy balance models (EBMs) for longterm simulations, through zonal statistical-dynamic models and nowadays, Earth models of intermediate complexity (EMICs), to general circulation (or global climate) models (GCMs) for short-term weather details, are used in combination with each other. For instance, the GCM is used to determine key sensitivities (to orbital perturbations, for example), and then the EBM is tuned to have the same sensitivities. Knowledge and experience at each level of the hierarchy is applied interactively at other levels. Climate-model hierarchies have also been discussed by Henderson-Sellers and McGuffie (1987), Claussen et al. (2002) and Bartlein and Hostetler (2004), with Claussen et al. (2002) distinguishing between integration of components vs. detail of description, and proposing the term "spectrum" to avoid any suggestion that one hierarchical level is better than another (Fig. 1).This paper briefly surveys how these ideas have found form over the last several decades, in the area of coupled ice sheet-climate modeling. To some extent the original concept of hierarchies has been realized, but mostly it has been adapted in different ways than originally envisioned, driven by the need to address the very