So-called “indirect” tensile tests have been adopted for various reasons, especially ease of execution and supposed freedom from several notable interferences. The extensive literature reveals much disparity in the conditions for such tests, and no consensus conclusion concerning the optimum protocol or even the interpretation of the results. Briefly, these, and other uniaxial tests in compression, are not well understood, despite long use and much investigation. Trust in centrally initiated tensile failure is misplaced. Experimentally, in a beam-split test, the effect of loading width is shown to be substantial; standard interpretations cannot explain the results. Failure initiated in shear is most probably the mechanism.