The old adage that "a problem correctly formulated is half solved" is rarely challenged. Model-centered problem solving relies critically on a correct model. Complete, clearly stated assumptions, a precise statement of objectives, and an adequate representation of the model of the system under study are readily accepted as responsibilities of the modeler. But just how important are both the process and product of model representation? The modeler's responsibilities in producing a correct model can be either significantly aided or drastically inhibited by the representational mechanism - the language - being used. The comprehensibility of the model, subsequent extension, adaptation, or maintenance of the model, are highly dependent on the representational language. Language for model representation is therefore an indispensible tool. The pivotal role of this tool is described through its relationship to modeling methodology, model development environments, and automated model diagnosis. The need for an emphasis on the capabilities for conceptual expressiveness, in contrast with execution efficiency is discussed. While software engineering research clearly advocates this redefinition of emphasis, we observe that much of the research in the rising disciplines within parallel computation appears to ignore the lessons of this history.