The work reported was conducted to provide a basis for structural ceramic mechanical property standardization activities under way in the United States, Germany, Japan, and Sweden, All measurements reported here were conducted by 10 US. groups on GN-10 silicon nitride within an International Energy Agency program including the United States, Germany, Japan, and Sweden. This cooperative work included tensile strength studies of two geometries of button-head tensile specimens, The authors conducted some of the measurements and performed data analyses and interpretation. The tensile fracture behavior of GN-10 silicon nitride was studied at room temperature. A total of 150 strain-gaged button-head tensile specimens were measured, One hundred of a straight collet design and 50 of a tapered collet design were fractured. All specimens were highly strain gaged and the outputs for each were: measured during loading to fracture, Bending moments were calculated, Each participating laboratory group fractured 15 tensile specimens, 10 of the straight collet design and 5 of the tapered collet design under rigorously controlled testing conditions. Of 100 straight collet specimens 75 broke,in the gage section. Of 50 tapered collet specimens 34 broke within the gage section. Analysis of the Weibull m and a, estimators at upper and lower confidence bounds of 95% and 5% did not indicate a clear choice between the two designs. For specimens which fractured in the gage section, the unbiased maximum likelihood Weibull estimators for in and a, were 12.5 and 730 and 10.4 and 716, for the straight and tapered collet configurations, respectively. These are not statistically different at the 95% and 5% confidence levels. Strengths were also analyzed in terms of a three-parameter Weibull model. The straight collet specimen data fitted the three parameter model well with a threshold stress estimator gamma of 506 MPa, while the tapered collet specimens provided a poorer fit to the model and had a threshold stress estimator of 432 MPa, a difference of about 15%. Regression analysis indicated that the straight collet grip provided less bias of strength as a function of bending moment. The straight collet specimens showed essentially little dependence of tensile strength upon bending moment in the range of 0% to 6%, while the tapered collet specimens showed a decrease in strength as the bending moment increased from 0% to 4%, However, the regression parameter was low and no significant statistical conclusion could be made regarding the superiority of either of the grip designs.