Our objective was to compare the polymerization stress (sigma(pol)) of a series of composites obtained using poly(methyl methacrylate) (PMMA) or glass as bonding substrates, and to compare the results with those from in vitro microleakage of composite restorations. The tested hypothesis was that stress values obtained in a less rigid testing system (i.e. using PMMA) would show a better relationship with microleakage data. Five dental composites were tested: Filtek Z250 (FZ), Z100 (Z1), Concept (CO), Durafill (DU) and Heliomolar (HM). sigma(pol) was determined in 1mm high specimens inserted between two rods (Ø=5mm) of either PMMA or glass. The composite elastic modulus (E) was obtained by three-point bending. sigma(pol) and E data were submitted to a one-way analysis of variance/Tukey test (alpha=0.05). For the microleakage test (MI), bovine incisors received cylindrical cavities (Ø=5mm, h=2mm), which were restored in bulk. After storage for 24h in water, specimens were subjected to dye penetration using AgNO(3) as tracer. Specimens were sectioned twice, perpendicularly, and microleakage was measured (in millimeters) under 20x magnification. Data from MI were submitted to the Kruskal-Wallis test. Means (SD) of sigma(pol) (MPa) using glass/PMMA were FZ: 7.5(1.8)(A)/2.5(0.2)(bc); Z1: 7.3(0.5)(A)/2.8(0.3)(ab); CO: 6.8(1.1)(A)/3.2(0.5)(a); DU: 4.5(0.7)(B)/2.0(0.2)(bc); HM: 3.5(0.2)(B)/2.3(0.3)(c). sigma(pol) obtained using PMMA rods were 34-67% lower than with glass. Means (SD) for tooth average/tooth maximum microleakage were FZ: 0.92(0.19)(B)/1.53(0.30)(a); Z1: 1.19(0.21)(A)/1.75(0.20)(a); CO: 1.26(0.25)(A)/1.78(0.24)(a); DU: 0.83(0.30)(B)/1.68(0.46)(a); HM: 0.81(0.27)(B)/1.64(0.54)(a). The tested hypothesis was confirmed, as the composites showed the same ordering both in the polymerization stress test using PMMA rods and in the microleakage test.