This study utilizes state of the art information concerning roof truss behavior to simulate the effect of strength property correlations on roof truss reliability. Also, research efforts to define lumber property correlations are summarized. Computer models were constructed to simulate the behavior of five common roof truss patterns. The patterns studied were the Fink, double W, triple W, triple W farm, and flat Howe trusses. Monte Carlo simulation, based on 50,000 trials, was used to calculate probabilities of failure for each truss pattern, and level of strength property correlation. The simulated trusses were subjected to a 50 year maximum lifetime snow load distribution plus a dead load distribution. Bending, tensile, and compressive strength properties were modeled by a weighted least squares regression analysis. Modulus of elasticity was used as the independent strength predictor, and was modeled by a Weibull distribution. The residual strength correlations in the regression models were varied between simulations from 0.0 to 0.9. Member stresses were calculated using a pin joint analysis modified by load factors derived from the Purdue Plane Structures Analyzer. Truss failure was determined by a weakest link analysis which used failure criterion developed from current truss design equations. The influence of lumber property residual correlations on roof truss reliability was defined. The probability of failure increased as the levels of correlation increased, but the difference in magnitudes were considered insignificant. Also, the ratio of snow to dead load was found to be affect wood truss reliability. / M. S.