A testing procedure was developed to measure the effectiveness of tension proof testing, given that the gripped ends of the lumber do not fully experience the proof stress during a tension proof test. First the middle portion of each 14-foot 2X6 No 2 KDIS Southern Pine lumber specimen was tension proof tested, as if the pieces were only 10 feet long. Then the pips of the testing machine were moved out so that the entire middle 10-foot, section including the portions that had been gripped during the initial proof test would be fully stressed during a second tension test Some specimens, referred to as falldown, were broken in the second test at a load level below the proof load. This indicates that because of the effect of gripped ends the, initial proof test was ineffective in destroying some pieces of lumber with tension strength below the proof load level. The experimental design and analysis were aimed at validating the Showalter et al. (1987) tensile strength - length effect model as adapted to this specialized problem. Although the model predicted that falldown would occur, it was unable to predict the amounts of falldown subsequent to the amounts of proof-test breakage occurring for a sample of lumber. Through the use of end-grading rules, falldown amounts can be reduced to acceptable levels. A method was proposed for developing end-grading rules and deriving concomitant allowable tension stresses for proof-tested lumber. A significant finding was that the histogram for the tension strength of the surviving end-graded lumber had an abruptly-ending left tail, indicating that there were no severely reduced tension strength values for any of the falldown specimens remaining in the sample. / Master of Science