This article takes a contrarian position: an "instructional design" or "teacher training" model, because of the sheer number of its interconnected parameters, is too complex to assess or to compare with other models. Models may not be the way to go just yet. This article recommends instead prior experimental research on limited numbers of individual parameters. Empirical, replicable results could then establish a dependable foundation for the development of models. Despite this off-the-reservation view about current models, conceding the merits of many of their rationales, the article offers several detailed recommendations for upgrading them