Measurement and prediction of cohesive powder behaviour related to flowability, flooding or arching in silos is found to be very challenging. Previous round robin [52] attempts with ring shear testers did not furnish reliable data and have shown considerable degrees of scatter and uncertainty in key measurements. Thus studies to build a reliable experimental database using reference materials are needed in order to evaluate the repeatability and effectiveness of shear testers and the adopted procedures. In this paper, we study the effect of particle size on the yield locus for different grades of limestone (calcium carbonate). We use the nonlinear Warren Spring equation to obtain the values of cohesion C, tensile strength T, and the shear index n. We recover linear (n = 1) yield loci for d₅₀>70 μm with respectively small C and T, with consistent, finite macroscopic friction C/T = 0.7. With particle size decreasing below 70 μm the response becomes more and more cohesive and non-linear (1<n<2). Then we compare the values of the parameters C,T and n obtained from two different shear testers (Schulze and Brookfield PFT). Both testers run at positive confining stresses (slightly different ranges) and give identical results for large fractions (weakly cohesive). For strongly cohesive samples, the PFT results are very similar to the ring shear tester, with slightly smaller values for T, C, and n. Further experiments with a variety of cohesive powders are needed to confirm or rebut this systematic difference the two testers display for cohesive powders. Finally, we compare the (extrapolated) values of T with a direct, transverse measurement running at negative stresses, using the Ajax tensile tester, and found a very good agreement, which validates the Warren Spring equation for negative stresses.