Protocol Analysis is a valuable tool for gaining qualitative data from obse
rvations of programmer behavior during software maintenance. However, there
are some major drawbacks with Protocol Analysis as it is currently practic
ed. Firstly, Protocol Analysis requires a daunting amount of effort at each
stage of analysis. Secondly, the results from one Protocol Analysis are of
ten difficult to compare with results from another. This paper describes a
coding scheme, AFECS, designed to reduce the effort required to perform Pro
tocol Analysis and to resolve the problem of noncomparable results. AFECS u
ses codes that consist of expandable and flexible segments. This allows AFE
CS to be tailored to the requirements of a variety of research studies, whi
le maintaining a degree of consistency. Explicit segmentation also makes AF
ECS easy to use. An example shows AFECS' use and ability to adapt to divers
e research questions.