Recently Keefer and Verdini (Management Science, 39 (9), 1086-1091 (1993)) and Lau, Lau and Zhang (IIE Transactions, 28, 183–192 (1996)) showed that well-known PERT-type formulas for computing standard deviation from fractiles are very unreliable, and they presented alternative formulas that are more accurate. This note extends their results and present a formula that is simpler, more meaningful and yet more accurate.