Objectives: To compare the tensile strength and failure modes of dentin and restored dentin using cast metal, resin composite and all-porcelain cemented with different resin cements, using mini-dumbbell test. Methods: Extracted human molars were vertically sectioned to prepare 2 mm dentin slabs which were divided into 1 control and 12 sample groups of 6 specimens each. Sample specimens were cross-sectioned using diamond burs and restored either with half 1) cast Co-Cr alloy 2) cured composite (Filtek Z 250, 3M) 3) porcelain (545 VMK 68, Vita), prepared in a standard half-dumbbell shape, and cemented either with 1) Panavia F 2) Superbond C&B 3) Single bond + RelyX or 4) Variolink II cements. All specimens were prepared into dumbbell shape, after soaked in water at 37ºC for 24 h, and tensile loaded using universal testing machine at a cross-head speed of 1 mm.min-1. Tensile strength in MPa were calculated and analyzed using 2-way ANOVA and Scheffe's test at p<0.05. SEM and light microscope were used to examine the mode of failure. Results: Significantly lower tensile strength of restored dentin than intact dentin was found. No significant difference in tensile strength was found among types of restorations. Superbond C&B provided significantly higher tensile strength with mixed failure of cohesive in resin and adhesive at resin/restoration interface. Single bond + RelyX specimens had significantly less tensile strength with adhesive failure at dentin/resin. No significant difference in tensile strength was found between Panavia F and Variolink II specimens. Conclusions: Reduction of dentin decreased the tensile strength. Complete hybridization of resin into dentin provided the higher tensile strength than that of resin/restoration interfaces.