Object-oriented inspection in the face of delocalisation

Software inspection is now widely accepted as an effective technique for defect detection. This acceptance is largely based on studies using procedural program code. This paper presents empirical evidence that raises significant questions about the application of inspection to object-oriented code. A detailed analysis of the 'hard to find' defects during an inspection experiment shows that many of them can be characterised as 'delocalised' - the information needed to recognise the defect is distributed throughout the software. The paper shows that key features of object-oriented technology are likely to exaggerate delocalisation. As a result, it is argued that new methods of inspection for object-oriented code are required. These must address: partitioning code for inspection ('what to read'), reading strategies ('how to read'), and support for understanding what isn't read - 'localising the delocalisation'.