The availability of item response times made possible by computerized testing represents an entirely new type of information about test items. This study explores the issue of how to represent response-time information in item banks. Empirical response-time distribution functions can be fit with statistical distribution functions with known properties. For this study, data were obtained from a computerized adaptive test of arithmetic reasoning skills administered as part of a larger test battery (the Law School Admission Test) to 38,357 examinees. Four functions (the normal, lognormal, gamma, and Weibull) were fit to empirical distribution functions from a computer-administered test, and the various functions were evaluated to determine which described the empirical distributions the best and provided the most useful parameters for storing in an item bank. The lognormal distribution was found to best fit both exploratory and confirmatory samples. It provides meaningful and useful parameters that can be stored in an item bank. (Contains 3 tables, 8 figures, and 11 references.) (Author/SLD)