This study was designed to follow up earlier work on mapping instruction. The two main goals were to: (1) test the effectiveness of mapping instruction as a general cure for "buggy" subtraction algorithms, and (2) explore two alternative explanations of how this new form of instruction works. It was hypothesized that mapping cures bugs either by imparting understanding, or through feedback and practice. About 80 children from the fourth, fifth, and sixth grades of an urban parochial school were given a written subtraction test designed to diagnose buggy algorithms. From this, eighteen students diagnosed as having buggy algorithms were chosen and divided into two groups approximately matched on race, sex, grade, and teacher rating of ability. After pretesting, one group received mapping instruction, the other received prohibition instruction. After training, both groups took both immediate and delayed posttests. Results indicated that mapping instruction is not always successful in curing bugs, and that the relationship between bugs and understanding is not straightforward. Also, some questions have been raised about what is learned in mapping instruction, as data suggest children can miss the general-trade-borrow analogy that is a major factor in the training. (MP)