Abstract
The fit of fracture strengths of glass to the normal, Weibull, and Type I extreme value distributions is compared. For abraded glass, the Type I distribution fits much better than the other two. Both the normal and Weibull distributions fit strengths of as‐received glass well. There seems to be no reason to use the maximum‐likelihood method for calculating Weibull parameters, although it is preferred by statisticians. Evidence for preferring the normal to the Weibull distribution for fracture strengths and log fatigue times is described. Extreme values of fracture strengths fit the normal, as well as the Weibull, distribution.