"Why models go wrong" BYTE, 1985
"A model can be a physical object: a scale-model of an ancient settlement inferred from archeological evidence, or a physical anthropologist's reconstruction of a skeleton from a few bones and fragments. Most social models, however, are mathematical isomorphisms that specify one-to-one relationships between elements of the model and observable processes or entities. This abstractness gives models great versatility, but it also opens doors to potential problems. You don't need to be a naval architect to see that a model battleship has no bottom, but flaws in an equally defective sociometric model that repeatedly factors a large covariance matrix might escape casual scrutiny." p.151.
My source:
Houston, T.R. 1985. Why models go wrong. BYTE 10(10) : 151-164.