When items are selected during a computerized adaptive test (CAT) solely with regard to their measurement properties, it is commonly found that certain items are administered to nearly every examinee, and that a small number of the available items will account for a large proportion of the item administrations. This presents a clear security risk for testing programs that are available on more than a few scheduled testing dates throughout the year. Several approaches to this concern control item exposure rates through probabilistic mechanisms built into the selection process. While many of these exposure control procedures are quite effective in limiting rates of item use, they are also problematic to some extent. Several exposure control algorithms are described, including the rationale for their application and the nature of any inherent problems. An empirical comparison of the relative effectiveness of these methods is presented, based on simulated CATs. The unconditional Sympson-Hetter method (J. B. Sympson and R. Hetter, 1985) (USH), the conditional Sympson-Hetter method (M. Stocking and C. Lewis, 1995) (CSH), and the Davey-Parshall (T. Davey and C. Parshall, 1995) (DP) methods outperformed a no exposure control method, and the CSH and DP methods generally outperformed the USH method. The different targets and results of the CSH and DP methods are discussed. (Contains 1 table, 7 figures, and 17 references.) (SLD)