In　the　paper　by　Li　(Fuzzy　Sets　and　Systems　106　(1999)　113),　a　minimum　average　weighted　deviation　method　was　investigated　to　determine　the　weights　of　attributes　using　incomplete　preference　information.　It　is　shown　in　this　paper　that　the　method　is　flawed　and　can　only　be　meaningful　if　it　is　revised.　Two　general　ways　are　presented　to　revise　it　and　are　justified　by　using　Lagrange　multiplier　method　and　Hessian　matrix.　Four　specific　revised　models　and　approaches　are　investigated　and　examined　through　a　numerical　example.　Different　membership　functions　are　suggested　and　utilized　to　normalize　decision　matrix　so　that　a　credible　ranking　order　can　be　achieved.　©　2004　Elsevier　B.V.　All　rights　reserved.