Purpose: The objective of this study was to test the hypothesis that bulk fracture of glass-ceramic disks of variable thickness originates at the inner, resin-bonded surface and is dominant over Hertzian fracture at the lower range of thickness values. Materials and Methods: Eight groups of seven glass-ceramic disks (Dicor, Dentsply), 12 mm in diameter with thicknesses ranging from 0.4 to 2.4 mm, were cast, cerammed (to produce approximately 55 vol% of tetrasilicic fluormica cyrstals), air abraded, etched, and silane coated according to the manufacturer's instructions. The disks were bonded to an eposy die substrate (with an elastic modulus comparable to that of dentin) using a light-activated resin cement. The bonded samples were supported on a flat surface and loaded at the top center of each disk until crack initiation occurred. All disks exhibited an initial crack within the bonded surface. Three randomly selected samples for each thickness were loaded beyond the point of crack initiation until Hertzian failure occurred. Results: Although the crack-initiation force increased with increasing thickness, the failure stress approached a maximum level at a thickness of approximately 1.6 mm. These results suggest that the estimated maximum occlusal load for each patient should be used to select the minimum thickness of carmic crowns rather than using the arbitrary traditional selection of a 1.5 - mm thickness. Conclusions: The authors conclude that bulk fracture is initiated within the bonded surface of a glass-ceramic specimen (for samples 0.4 to 2.4 mm in thickness) when the glass-ceramic is supported by a substrate with an elastic modulus similar to that of dentin. Furthermore, a Hertzian failure mechanism is unlikely to cause bulk fracture for these conditions.