Abstract Results are presented of a Bureau of Mines study of the effect on the fracture initiation stress of loading a cylindrical specimen of brittle rock between steel end-plates with a diameter either equal to or slightly larger than the specimen diameter. The effect on fracture initiation stress of interposing a low-modulus insert material between the specimen and end-plates is discussed. In this paper, fracture initiation stress is defined to be the stress required to initiate microcracking (void growth) in the rock specimen. The three-dimensional axisymmetric finite-element method of solution is used to calculate the stress distribution in the specimen when a constant displacement is applied over the surface of the end-plates. These results are then used to calculate the fracture initiation stresses. Assuming that open and closed cracks are uniformly distributed throughout the rock structure, two classes of fracture initiation in brittle rock are considered: (1) Fracture initiation when the applied stresses are compressive occurs from closed cracks when the shear stress acting along the crack surface exceeds a critical value, i.e. when the modified Griffith criterion is operative; (2) fracture initiation when the applied stresses are tensile occurs from open cracks when the tensile stress acting normal to the crack surface exceeds a critical value. Interpretation of results indicates the following: (i) The intrinsic fracture initiation stress (i.e. the stress magnitude required to initiate failure in the specimen when no end-effects are present) can be measured with confidence in a specimen with an L/D (length-to-diameter) value of 2·00, provided that the measurements of volume dilatation are made in the central portions of the specimen; (ii) fracture initiation occurs first at the contact between the specimen and end-plates along the outer edge of the specimen; (iii) the geometry of the fracture surface depends on both the geometry and physical properties of the specimen and end-plates and, in particular, on whether a low-modulus insert material is placed between the specimen and end-plates.