This paper studies whether examinee characteristics are systematically related to mode effect across paper and computer versions of the same instrument, using data from the Graduate Record Examination (GRE) of the Educational Testing Service in its Computer-Based Testing Pilot Study of 1991. The following characteristics of 1,114 examinees were studied as contributors to mode effect: (1) demographic variables (gender, racial/ethnic background, and age); (2) computer use variables (variety and frequency of computer experience, frequency of mouse use, and test mode preference); and (3) test-taking strategy variables (strategy preference, and tendency to omit or review items). The typical method found in comparability literature was used, treating performance across paper and computer versions as a continuous dependent variable. Because a mode effect in a small subset of examinees could be masked, a method that isolated examinees most affected by test mode was used. For this method, mode effect was treated as a three-level, categorical, independent variable. Data demonstrate mode effect and support the conception of a small subset of examinees whose performance was more affected by mode than that of the total sample. The search for examinee characteristics that explain occurrence of mode effect, however, yielded inconsistent results, with only weak relationships to mode effect. Three figures and 21 tables present study findings. (Contains 31 references.) (SLD)