Objective: The purpose of this project was to evaluate the performance of a dentin adhesive system on carious and noncarious primary dentin in vivo. Method and materials: Forty-eight primary molars with carious lesions were randomly assigned to 2 different treatments: group 1 (control, n = 24) - All identifiable, irreversibly infected dentin was removed prior to the application of the bonding agent and restorative material; group 2 (experimental, n = 24) - Irreversibly infected dentin was partially removed prior to the application of the bonding agent and restorative material. The control and experimental teeth were clinically monitored every 3 months and evaluated 12 months after restoration. The teeth were extracted around the time of exfoliation and processed for scanning electron microscopy. Results: Retention rate, marginal integrity, and pulpal symptoms were identical in both groups. Radiographically, the radiolucent area associated with the experimental restorations did not increase with time in 75% of the cases. For the control group, the adhesive system formed a hybrid layer. In the experimental group, there was morphologic evidence of the formation of an acid-resistant 'altered hybrid layer.' An acid-resistant tissue, resulting from the interdiffusion of adhesive resin within the area of carious dentin, was observed adjacent to and under the altered hybrid layer. Conclusion: Application of an adhesive restorative system to irreversibly infected dentin did not affect the clinical performance of the restoration.