Many selective regression testing strategies, which select a subset of the original tests for regression testing, have been proposed recently. Yet, the basic assumptions supporting selective regression testing have not been fully examined. The paper first introduces the notion of scope for change to represent the regression testing focus. A classification of faults and errors is then presented, together with the relationship between these faults and errors. The paper describes two basic assumptions that are implicitly made by most selective regression testing strategies, and identifies the types of fault that are undetectable due to these assumptions.