Two sizes (2x4 and 2x10) and two grades (2250f-1.9E and No. 2 KDl5) of Southern Pine lumber having three different test span lengths of 30, 90 and 120 inches were tested in tension parallel-to-grain. Results obtained from the tests indicated that the tensile strengths of the 30-inch test specimens were significantly higher than the tensile strengths of the 90- and 120-inch test specimens. A tensile strength-length effect model was developed for generating tensile strength values of lumber taking the length effect into consideration. The model generates tensile strength values for lumber longer than 30 inches in multiples of 30 inches, ie. 60-, 90- 120-inch lengths. The two sizes and two grades of Southern Pine lumber formed the data base for developing the model.

The tensile strength-length effect model utilized an MOE variability model which generated serially correlated MOE's along 30-inch segments for a piece of lumber using a second order Markov model. The segment MOE values were then used in a first-order Markov model to generate serially correlated tensile strength residuals for each 30-inch segment. The segment MOE values and the segment tensile strength residuals were then inputted into a weighted least squares regression to obtain the tensile strength parallel-to-grain for each 30-inch segment. The tensile strength of the generated piece of lumber was then determined using the weakest-link concept; the minimum segment tensile strength value was selected as the tensile strength of the generated piece of lumber.