<p>The triplet test can be used to determine the cohesion and frictional coefficient that should be used in defining the Mohr–Coulomb relationship for masonry subject to shear. In the European standard, roller supports are used on the three-unit specimen to prevent bending and ensure shear action between the unit-mortar interface. However, in some recent studies and in the Chinese standard, flat supports are employed, so the test induces some bending in the specimen as well as shear along the bed joints between the centre and outer units. The question therefore arises as to whether using platen supports can yield similar results as using the roller supports, and to which normal stress within the test should be used when relating normal stress to shear strength in the Mohr–Coulomb relationship. Currently, it is typical to plot the average shear stress at failure of the specimen against the average normal stress at the beginning of the test. Here, the actual normal and shear stresses occurring at specimen failure are examined in conjunction with the failure mode. Relationships between these variables in hollow, partially grouted and fully grouted concrete masonry were formed, based on the results from tests on 53 triplet specimens from this and previous studies. The results from the triplet tests were also compared with those from shear walls. Data for 112 masonry shear walls built with similar material properties were examined, showing the prediction of wall capacity from the European standard was generally conservative for North American concrete masonry shear walls.</p>