It is often necessary to carry out tensile strength tests on fibre cement composite samples cut from a flat sheet in the as-received condition where further sample preparation is undesirable. Very often the sample coupons show defects such as bowing. These imperfections are most noticeable for materials which fail at low strains, such as some of the new asbestos cement board replacements. The effect of some sample defects and alignment errors on the measured tensile strength of glass fibrereinforced cement (grc) have been described by Green et al. [1] who found that in practice some misalignment errors had less effect than expected but that sample defects such as curvature produced large errors in measured strain. In the present work, data from tensile tests on grc have been further investigated to determine the significance of two possible test imperfections on the measured tensile strength. These factors are: (1) the eccentricity of loading due to nonuniform clamping of the sample. This occurs when there are thickness variations in the clamped areas of the sample and may be extreme when there is a uniform edge-to-edge thickness variation or an isolated thick area near the edge of the sample; (2) the bending strain due to straightening of an initially curved shape of the sample. This may give rise to strains during a tensile test even in the absence of any tensile load. A full discussion of the effects of combined bending and tension in curved or irregular samples is given by Green et al. [1]. Assuming linear elastic behaviour, loading through pin joints and the additivity of bending and tensile stresses, the maximum stress at the edge of a sample subjected to an off-centre tensile force, Fig. 1, is given by