STATEMENT OF PROBLEM: In contrast to gold crowns, in vitro determination of the retentive strength of all-ceramic crowns is more difficult because components allowing connection to testing apparatus are not as easily integrated into the all-ceramic material. Nevertheless, retentive strength data are crucial for obtaining information about the potential clinical performance of luting cements for all-ceramic restorations. Therefore, a new in vitro model was necessary to evaluate the retentive strength of all-ceramic crowns. PURPOSE: The purpose of this in vitro study was to determine the retentive strength of 4 resin-cement systems, a compomer, a glass-ionomer cement, a resin-modified glass-ionomer cement, and a self-adhesive resin for luting zirconium oxide ceramic crowns. MATERIAL AND METHODS: One-hundred-twenty extracted human teeth were randomly divided into 12 groups (n = 10) and prepared in a standardized manner (5-degree taper, 3-mm occlusogingival height). All-ceramic crowns (Lava) were fabricated in a standardized manner for each tooth. The following cements and corresponding bonding regimens were used to lute the crowns to the teeth according to manufacturers' recommendations: CO, Compolute/EBS Multi; CO/RT, Compolute/EBS Multi/Rocatec; CB, Superbond C and B; CB/RT, Superbond C and B/Rocatec; CB/PL, Superbond C&B/Porcelain Liner M; PA, Panavia F; DC, Dyract Cem Plus/Xeno III; CH/PL, Chemiace II/Porcelain Liner M; RL, RelyX Luting, K/C, Ketac Cem/Ketac Conditioner; K, Ketac Cem; and RU, RelyX Unicem. After thermal cycling (5000 cycles, 5 degrees C-55 degrees C), the outer surfaces of the cemented zirconium oxide ceramic crowns were treated (Rocatec) to improve bonding and then placed into a low-shrinkage epoxy resin block (Paladur). The block/crown and tooth components for each specimen were connected to opposing ends of a universal testing machine so that crown retention could be measured. Crowns were removed from teeth along their path of insertion. The retentive surface area (mm 2 ) was determined individually for each tooth. Statistical analyses were performed using the Wilcoxon exact test, (alpha =.05) and a Bonferroni correction (alpha =.001). RESULTS: The median (minimum/maximum) retentive strength values (MPa) were as follows: CO, 1.7 (0.6/4.3); CO/RT, 3.0 (1.3/5.4); CB, 4.8 (3.7/7.9); CB/RT, 8.1 (4.2/12.7); CB/PL, 5.3 (3.7/10.2); PA, 4.0 (3.3/5.1); DC, 3.3 (2.1/5.6); CH/PL, 4.0 (1.3/6.3); RL, 4.7 (2.8/6.6); K/C, 1.8 (0.6/2.3); K, 1.9 (0.2/4.5); and RU, 4.8 (2.5/6.7). Superbond C&B (+ Rocatec) specimens showed the highest median retentive strength, but were not significantly different from Superbond C&B without Rocatec pretreatment of the all-ceramic crown's inner surface. Compolute specimens also did not benefit significantly from the Rocatec pretreatment. Within the materials used without pretreatment of the ceramic, Superbond C&B, Panavia, Dyract Cem Plus, RelyX Luting, and RelyX Unicem showed the highest median retentive strength values and were not significantly different. CONCLUSION: Within the conditions of this study, the compomer-cement, the resin-modified glass-ionomer cement, and the self-adhesive resin luting agent had the same level of retentive quality as the resin luting agents, Superbond C&B, and Panavia. Rocatec pretreatment of the ceramic surface did not improve the retentive strengths of Compolute and Superbond C&B.