There　are　two　opposite　views　on　whether　working　memory　can　guide　visual　attention.　Some　researchers　have　reported　that　the　contents　of　working　memory　guide　visual　attention　for　capturing　target　efficiently.　However,　others　reported　that　they　could　not　find　any　evidence　of　attention　capture　by　working　memory.　In　this　study,　it　tried　to　find　evidence　for　the　first　view　with　computer　simulation.　Two　models　based　on　two　hypotheses　were　set　up　in　simulating　the　simplified　4x4　Sudoku　problem　solving　by　which　an　fMRI　(functional　Magnetic　Resonance　Imaging)　experiment　was　performed　at　the　same　time.　One　model　is　based　on　working　memory　guiding　visual　selective　attention　assumption　while　the　other　is　based　on　no　guiding　random　attention　assumption.　Both　of　the　models　predict　the　response　time　(RT)　and　blood　oxygenation　level-dependent　(BOLD)　response.　Cognitive　cost　analysis　on　the　predictions　shows　that　more　cost　was　occupied　on　no　guiding　model　resulting　in　more　differences　between　fMRI　real　data　and　predictions　while　the　other　can　reduce　the　cost　and　get　good　fitness.　This　study　confirms　the　first　view　and　shows　that　working　memory　guiding　visual　search　for　capturing　target　is　the　intelligence　of　human　brain　in　reducing　the　cognitive　cost.　©　2011　Springer-Verlag.