An a-stratified design (H. Chang and Z. Ying, 1997) is a new concept proposed to address the issues of item security and pool utilization in testing. It has been demonstrated to be effective in lowering the test overlap rate and improving the use of the entire pool when content constraints are not main concerns. However, it cannot really solve the problem of high exposure rates when the test is moderately long. This paper proposes an enhanced stratified design that incorporates the Sympson-Hetter (SH) (J. Sympson and R. Hetter, 1985) procedure into the a-stratified design and compares the performance of the new procedure with the original a-stratified design and the Fisher-SH method through simulation studies. The results indicate that the enhanced stratified design yielded much well-balanced item exposure distribution, further reduced the test overlap rate and the numbers of overly exposed items, and maintained high reliability as well as low average bias, mean squared error, and scaled chi-square. (Contains 3 tables, 6 figures, and 17 references.) (Author/SLD)