Objective. This study compared the antibacterial activities of two dentin bonding systems (ABF, Kuraray and Reactmer Bond, Shofu) by a conventional agar well technique and a newly designed in vitro test using tooth model. Methods. In the agar well technique, the test materials were filled in the wells of Muller Hinton agar plates inoculated with Streptococcus mutans NCTC10449, and the diameters of inhibition zones produced around the materials were measured after 24h of incubation. For the tooth model test, three cavities (diameter 1mm, depth 2mm) were prepared in the flat occlusal dentin of human extracted molar. After sterilization, the teeth were left in broth culture of 1.56×108CFU/ml of S. mutans at 37°C for 72h for allowing bacteria to invade the cavity. The dentin bonding systems were applied separately to each of the two infected cavities, and the third cavity was left unapplied for control. After sealing the occlusal surfaces, the teeth were kept in physiologic saline solution at 37°C for 72h. The standardized amounts of dentin chips (120±5mg) were obtained from the cavity walls and the number of bacteria recovered was determined. The results were analyzed by One Way ANOVA, Kruskal Wallis and Mann-Whitney's U tests. Results. The primer of ABF and Reactmer Bond produced inhibition zones with similar sizes (p>0.05), but the bonding resin of ABF did not produce any inhibition. When tested by the model cavity method, the application of ABF resulted in significantly less bacterial recovery than Reactmer Bond (p<0.05), demonstrating substantial antibacterial effects. Conclusions. The tooth model method used in this study was effective for evaluating the substantial antibacterial effects of dentin bonding agents, and the experimental dentin bonding system ABF was demonstrated to be able to inactivate the bacteria in the cavity effectively in comparison with little antibacterial activity shown by Reactmer Bond. © 2003 Elsevier Science Ltd. All rights reserved.