A　two-stage　logarithmic　goal　programming　(TLGP)　method　is　proposed　to　generate　weights　from　interval　comparison　matrices,　which　can　be　either　consistent　or　inconsistent.　The　first　stage　is　devised　to　minimize　the　inconsistency　of　interval　comparison　matrices　and　the　second　stage　is　developed　to　generate　priorities　under　the　condition　of　minimal　inconsistency.　The　weights　are　assumed　to　be　multiplicative　rather　than　additive.　In　the　case　of　hierarchical　structures,　a　nonlinear　programming　method　is　used　to　aggregate　local　interval　weights　into　global　interval　weights.　A　simple　yet　practical　preference　ranking　method　is　investigated　to　compare　the　interval　weights　of　criteria　or　rank　alternatives　in　a　multiplicative　aggregation　process.　The　proposed　TLGP　is　also　applicable　to　fuzzy　comparison　matrices　when　they　are　transformed　into　interval　comparison　matrices　using　α-level　sets　and　the　extension　principle.　Six　numerical　examples　including　a　group　decision　analysis　problem　with　a　group　of　comparison　matrices,　a　hierarchical　decision　problem　and　a　fuzzy　decision　problem　using　fuzzy　comparison　matrix　are　examined　to　show　the　applications　of　the　proposed　methods.　Comparisons　with　other　existing　procedures　are　made　whenever　possible.　©　2004　Elsevier　B.V.　All　rights　reserved.