Purpose: To evaluate the marginal quality of composite resin restorations placed in extracted molars either in bulk (4 mm) or three increments. Materials and Methods: Sixteen extracted mandibular molars were selected and two two-surface cavities were prepared in each tooth (proximal depth 4 mm, occlusal width 5 mm). On one side of the tooth, Tetric EvoCeram Bulk Fill (Ivoclar Vivadent) was applied in a single increment; on the other side, Tetric EvoCeram (Ivoclar Vivadent) was applied in three increments: a horizontal gingival, an oblique buccal, and an oblique lingual increment. Each layer was light cured for 10 s with a Bluephase G2 curing light (1200 mW/cm2). Two adhesive systems were employed according to the instructions for use: the single-component etch-and-rinse system ExciTE F (Ivoclar Vivadent) and the self-etching two-component system AdheSE (Ivoclar Vivadent). The adhesive was light cured for 10 s with a Bluephase G2 curing light (1200 mW/cm2). Eight fillings were placed for each test group and all restoration margins were confined to the enamel. After 10,000 cycles of thermocycling (5°C/55°C), the quality of the proximal margins was semiquantitatively directly evaluated with a stereomicroscope at low magnification and a dental explorer using the SQUACE (semi-quantitative evaluation of restorations) method. In addition, replicas were made for SEM analysis, which was carried out four weeks later at high magnification (200X) by measuring the percentage of regular proximal margins in relation to the entire margin. Results: After thermocycling, statistically significantly higher percentages of regular margins were detected for those fillings placed with the etch-and-rinse system ExciTE F than for those placed with the self-etching system AdheSE - irrespective of the evaluation method (Mann-Whitney non-parametric test, p 0.05). There was no statistically significant difference between the resin restorations placed in bulk and those placed in three increments (Mann-Whitney, p > 0.05). The semi-quantitative evaluation by means of a light microscope yielded statistically significantly higher values for regular margin than did the SEM evaluation for all 4 test groups (p 0.05). Pearson's correlation coefficient for both evaluation groups was 0.87 (p 0.0001). Conclusions: The marginal quality of medium-sized Class II restorations of composite resins placed in one increment was similar to that of restorations placed in several increments. The semiquantitative evaluation of the marginal quality with an explorer at low magnification is an effective and rapid method to predict the clinical performance of direct restorations. Keywords: bulk-fill composite material, incremental technique, marginal quality, SEM, SQUACE