The new technique to evaluate user interfaces by analysis of user session transcripts reported here involves the detection of patterns of repeated user actions. The hypothesis is that repetition of user actions indicates potential interface problems. To define a usable concept of repetition, the authors developed maximal repeating patterns (MRPs). MRPs are an important tool for discovering how users actually use the system. An advantage to the proposed process is that transcripts can be collected automatically and noninvasively over a long period of time, then processed to identify potential performance and usability problems. The proposed system does have some shortcomings, all of which are addressed in the paper. An O(n²) algorithm developed by Siochi uses position trees to detect all the MRPs in a transcript and to report the positions where they occur. Raw transcripts are first normalized; an MRP tool is then used to produce a list of MRPs, which may be screened and examined interactively by a human evaluator. This process was applied to a large image-processing system with a number of known problems, and the results were analyzed. Evidence presented indicates the value of the process. The paper is well organized and well written. It should interest both novices and experts in the field of user interface analysis and improvement. A good reference list is included.