One hundred maxillary premolar teeth were randomly allocated to ten groups. Each group was restored with one of ten different restorative techniques. The teeth were stored in deionized water for 7d prior to longitudinal sectioning in a mesio-distal plane. Following sectioning, ten specimens from each group were chosen at random from the 20 available sections. The sectioned surfaces were polished using 600-grit SiC abrasive paper and etched for 10 s with 50% phosphoric acid to remove the smear layer produced by sectioning. Five tooth sections from the dentin bonding resin groups were allowed to dry at 20 degrees C for 24h. The glass ionomer-based groups were reimmersed in deionized water during this period. The remaining five sections from each group were replicated using an addition-cure vinyl polysiloxane impression material and an epoxy resin. A comparison was made of the sections and the replicas. All tooth specimens were sputter-coated with gold for 4 min and examined using a scanning electron microscope. Replicas were gold-coated for 3 min. Different tooth/restoration interfaces, associated with different materials, were observed. A marked difference between the replicas and tooth sections was observed for glass ionomer-based restorations but not for resin-based bonding systems. Representative samples of replicas and specimens are shown, and the significance of the observed differences is discussed.