Graphical　variable　message　signs　(VMSs)　are　infrastructure-based　advanced　traveler　information　systems　widely　used　to　provide　drivers　with　real-time　traffic　condition　information　about　a　road　section　or　area.　However,　there　is　a　lack　of　research　on　the　suitable　volume　of　information　to　be　shown　on　graphical　VMSs.　In　practice,　an　overload　of　VMS　information　commonly　exists,　especially　on　China＇s　highways.　Building　on　our　earlier　findings　obtained　through　surveys　and　static　cognitive　experiments,　this　study　introduces　the　use　of　dynamic　simulation　experiments　to　assess　the　influence　of　the　volume　of　information　(i.e.,　number　of　roads　displayed)　on　graphical　VMSs　from　the　perspective　of　drivers＇　visual　perception　characteristics.　Thirty-two　drivers　participated　in　the　driving　simulation　experiment　and　questionnaires.　Five　indexes,　including　legibility　speed,　legibility　distance,　legibility　time,　comprehension　accuracy,　and　driver　subjective　scoring,　were　thoroughly　analyzed　to　evaluate　their　relationships　to　different　volumes　of　information　(i.e.,　four,　five,　and　six　roads　shown　on　a　VMS).　The　results　show　that　the　legibility　distance　notably　decreased　with　increasing　volumes　of　information.　The　comprehension　accuracy　decreased　significantly　when　the　number　of　roads　shown　increased　to　six.　The　legibility　speed,　legibility　time,　and　subjective　scoring　also　deteriorated　as　the　number　of　roads　displayed　on　the　VMS　increased.　The　index　scores　were　evaluated,　in　combination　with　the　data　of　the　drivers＇　subjective　scoring,　data-based　statistical　analyses,　and　comprehensive　evaluations　using　the　TOPSIS　method,　to　recommend　that　five　is　the　recommended　maximum　number　of　roads　to　be　shown　on　a　graphical　VMS.　The　results　of　this　study　support　the　goal　of　providing　understandable　and　effective　messages　for　drivers　by　addressing　issues　relating　to　how　much　information　should　be　displayed　on　a　VMS.　These　findings　provide　a　basis　for　policy　development　to　ensure　consistent　and　practical　designs　of　graphical　VMSs　on　highways.　(C)　2020　Elsevier　Ltd.　All　rights　reserved.