This paper uses a least squares regression analysis to verify the well-known PERT estimate of the mean activity time as a weighted average of the optimistic (a), most likely or modal (m) and pessimistic (b) activity times in the classical formula: ûx = ( a + am + b )/6, and the estimate of the standard deviation of activity duration as one-sixth of the range: σx = (b-a)/6. Based on properties of the beta distribution, the paper shows also that there are theoretical grounds for using these estimates for a specified range of modal activity times, because in this range, ûx closely approximates the actual relationship between the mean and the mode.