Item selection methods in computerized adaptive testing (CAT) can yield extremely skewed item exposure distribution in which items with high "a" values may be over-exposed while those with low "a" values may never be selected. H. Chang and Z. Ying (1999) proposed the a-stratified design (ASTR) that attempts to equalize item exposure distribution by uplifting the usage of low "a" items. The method has been demonstrated to be effective in improving the use of the entire pool, without sacrificing efficiency in ability estimation when it is used with certain types of item pools. Nevertheless, the ASTR may result in a number of items being over-exposed in some pools where the correlation between the "a-" and "b-parameters" is significant. To remedy this overexposure problem, H. Chang, J. Qian, and Z. Ying (1999) developed the a-stratified with b-blocking method (BASTR). These two stratified methods have not been tested under conditions where content specifications are imposed. To address the issue of content balancing, an adaptation of the general ideas of the constrained CAT to ASTR and BASTR was investigated in this study. In addition, the effects of incorporating Sympson-Hetter (SH) (J. Sympson and R. Hetter, 1985) exposure control into ASTR and BASTR were also examined. Findings from simulation indicate that ASTR and BASTR, with or without SH exposure control can meet the content specifications, make better use of item pools, and yield lower test-overlap rates. (Contains 2 tables, 2 figures, and 25 references.) (SLD)