OBJECTIVES: To determine, in vitro, the influence of cavity dimensions on shrinkage stress development and microleakage of composite restorations. METHODS: Cylindrical cavities with 2, 3, 4 or 6 mm diameter and 1 or 2 mm depth were prepared in bovine incisors (with enamel margins) and reproduced in photoelastic resin. Cavities were restored in bulk (Single Bond+Filtek Z250). Stress fringes were analyzed 10 min after photoactivation. Restored teeth were stored for 24 h, coated with nail polish and immersed in 50% AgNO3 for 2 h, followed by 6 h in developing solution. Specimens were sectioned twice and microleakage was measured under 20x magnification. Photoelastic data were analyzed descriptively only, due to lack of variability in some groups. Microleakage was analyzed through (generalized linear) regression models. Pearson correlation coefficients were computed to evaluate the relationship between microleakage and 'C' factor or volume. RESULTS: Fringe order was higher at the internal angles than at the margins of the restorations. There was a trend for higher fringe orders with increasing diameter and depths. Regression analysis revealed a significant influence of restoration depth for diameters above 2mm. The effect of diameter on microleakage is significantly higher (p < 0.0001) for 2-mm deep restorations, compared to those 1-mm deep. Pearson correlation coefficients suggest some relationship between microleakage and volume (r = 0.724, p < 0.0001), but not between microleakage and 'C' factor (r = 0.048, p = 0.6120). CONCLUSIONS: Shrinkage stress and microleakage were higher in restorations with larger diameters and depths. Microleakage seemed to be related to a restoration's volume, but not to its 'C' factor.