It makes sense that the process of building new generalizations is grounded in individual cases. What is not so obvious is that the process of repairing them is similarly grounded. In order to repair generalizations, we need to construct explanations. And those explanations require data to support them. We cannot wonder if old fish makes for bad sushi unless we keep around those cases in which we have experienced sushi made from old fish. Likewise, if we later find out that one of our explanations is incorrect, we need access to the source cases for that explanation to try to construct an alternative. To construct a new explanation, we need to be able to access the details of our past cases and "see them in a new light."
Where am I in the content of the book?
