The problems about consonant clusters in old Chinese are hot issues in modern linguistics and traditional phonology in China. Joseph Edkins advanced some opinions about the consonant clusters of old Chinese for the first times. And Bernhard Karlgren also pointed out that there are some relationships between concord-characters in modern Chinese and consonant clusters in old Chinese. After that, Lin, Yu-tang et al. have studied on the problems about consonant clusters. So far, there are many studies about the forms of consonant clusters in the period of old Chinese and the rules of changes from old Chinese to middle Chinese, but there are shortcomings on the researches about internal constraints of changes. Therefore, this article will explain two problems below through optimality theory. First, what was the forms of consonant clusters in old Chinese? Second, what were the reasons for the loss of consonant clusters in middle Chinese? As a result, old Chinese had consonant clusters and their form was “C+L”, because Max-IO dominated *Complex and Sonority had higher hierarchy than Ident-IO(F). But in the period of middle Chinese, there was a reverse of hierarchy between MAX-IO and *Complex, and there were successive hierarchical raise from right to left among Ident-IO(F), *Onset-VO, Ph and Onset-Son, so that ‘voiceless obstruents(or nasal) + liquid’ and ‘voiced obstruents + liquid’ in old Chinese have changed into ‘voiceless obstruents(or nasal)’ and ‘liquid’ respectively in middle Chinese.