This　paper　investigates　the　issues　of　combination　and　normalization　of　interval-valued　belief　structures　within　the　framework　of　Dempster-Shafer　theory　(DST)　of　evidence.　Existing　approaches　are　reviewed,　examined　and　critically　analysed.　They　either　ignore　the　normalization　or　separate　it　from　the　combination　process,　leading　to　irrational　or　suboptimal　interval-valued　belief　structures.　A　new　logically　correct　optimality　approach　is　developed,　where　the　combination　and　the　normalization　are　optimised　together　rather　than　separately.　Numerical　examples　are　provided　throughout　the　paper.　©　2006　Elsevier　Inc.　All　rights　reserved.