This study examined various misconceptions committed by junior high school students in fraction addition and subtraction problems. Almost 600 subjects were administered two tests, addition and subtraction, and their performances were analyzed by several computer programs, including two programs (FBUG and SPBUG) which are so flexible to any items generated by computer that they can be used for any teacher-made tests. A painstaking error analysis and construction of buggy programs were carried out and summary statistics are described. The analysis results indicate that individual differences in applying different strategies and procedural skills varied more among students than expected. Many erroneous rules were committed by students who used them sporadically. These rules are often observed only once per student and never used repeatedly by the same individuals. Various error types (sources of misconceptions) cover almost all the levels of tasks involved in solving fraction problems. A close examination of frequency distributions of erroneous rules revealed that some errors tend to appear among high-score students while others appeared only among low-score students. Systematic investigation of "bug-behaviors" will lead further understanding in human cognition and learning and thus it will bring about further improvement in American education. (JN)