Purpose: To compare contraction stresses in direct and indirect composite restorations using crack analysis. Materials and Methods: Glass disks with a central cylindrical hole were prepared. Initial cracks were made in the glass using a Vickers indenter at various distances from the edge of the hole. The indentation crack lengths were measured parallel to the tangent of the edge of the hole. Silanized holes were directly or indirectly filled with a resin composite. The composite was either self-curing and light curing or only self-curing. Various thicknesses of composite cement were applied to the indirect restorations. The crack lengths were re-measured 15 and 30 min after they were filled. The contraction stresses were calculated from the crack lengths and fracture toughness of the glass. Elastic moduli of light-cured and self-cured composite samples were measured using a nanoindentation method 15 and 30 min after curing. Results: The stress could not be calculated for the light-cured direct composite and light-cured indirect composite restorations with thicker cement, due to glass failure caused by the contraction stress. Glass failure did not occur in the indirect composite restoration with thinner cement or in the self-cured direct composite restoration. The stress in the glass was lower for greater distances or shorter times. Factors of indentation-hole distance and restorative procedure significantly affected the stress. The measuring time and the type of curing had significant influence on the elastic modulus. Conclusion: Light-cured indirect composite restorations with a cement thickness 200 µm generated less contraction stress than did light-cured direct composite restorations. The lowest contraction stress was developed in the self-cured direct composite restoration. Keywords: polymerization, shrinkage, resin composite