In deducing capillary pressure curves by the centrifuge technique, the conventional Hassler-Brunner method assumes that the cylindrical sample is slender enough to neglect variation of capillary pressure across the sample width. In this work, the method for deducing capillary pressure curves for an arbitrarily shaped sample is evaluated. Capillary pressure curves are generated using model average saturation "data" and the error involved in neglecting the sample width by solving the Hassler-Brunner integral equation is determined. Neglect of width effects is seen to predict capillary pressure curves with lower fluid saturations. Errors associated with neglecting width effects are larger for larger diameter samples and for samples placed closer to the axis of rotation.