Previous　studies　analyzed　the　relationship　between　drivers＇　personal　characteristics　and　their　comprehension　of　traffic　signs　directly,　but　neglected　to　test　and　control　the　influence　of　small　class　of　traffic　signs　themselves.　This　error　might　cause　the　distortion　of　research　results.　So　a　comprehensive　survey　of　drivers＇　comprehension　of　all　warning　signs　was　conducted　in　this　paper.　Average　level　of　every　warning　sign　comprehension　was　assessed.　Warning　signs　were　divided　into　four　clusters　using　Hierarchical　Cluster　Analysis.　The　final　classification　scheme　was　determined　based　on　significant　difference　test　between　clusters　and　homogeneity　test　within　every　cluster.　Study　results　indicate　that　small　class　of　traffic　signs　has　certain　influence　on　the　level　of　traffic　signs　comprehension　and　it　may　inhibit　the　influence　of　drivers＇　personal　characteristics.　The　traffic　signs　classification　scheme　can　solve　this　problem　effectively.　This　study　will　provide　foundation　for　the　study　of　comprehension　and　design　of　traffic　signs.　©　2011　IEEE.