Objectives: To evaluate the microleakage in Class II resin restorations at different margins and the polymerization shrinkage of the composites used. Method and materials: Four standardized Class II (3 3 5 3 2 mm) cavities were prepared in 32 teeth. The sample had the gingival margin either 1 mm below or above the cementoenamel junction. Teeth were divided and restored according to the following protocols: (1) Single Bond + P60; (2) Prime & Bond NT + Surefil; (3) Bond-1 + Alert; and (4) Prime & Bond 2.1 + TPH. After 7 days, the specimens were thermocycled (500 cycles between 5 to 55°C with a 15-second dwell time), immersed in 0.5% methylene blue for 24 hours, sectioned, and evaluated (both surfaces) at the gingival margin by two examiners, using a 0 to 3 marginal infiltration score system. The polymerization shrinkage of the composites (n = 6) was evaluated by the disk deflective method. Microleakage data was evaluated by nonparametric repeated measures ANOVA. The percentage of polymerization shrinkage was evaluated by ANOVA and Tukey's test. Results: Only protocol 3 showed a significant difference between enamel and cementum margin. No difference was detected among the protocols in the enamel margin. Only protocol 1 provided a good seal in the cementum margin. All packable composites showed lower polymerization shrinkage than the hybrid tested. Conclusion: All protocols are able to prevent dye penetration in enamel margins; however, protocol 1 is preferable to reduce the microleakage in the cementum margin. The packable composites showed lower polymerization shrinkage compared to the hybrid resin.