Used a 2*3*2 mixed design, we tested the difference of repeated priming effect between the same conditions and the relevant conditions, under different SOA conditions. The results showed that in a short 167ms SOA, the priming effect between repeat/same and repeat/related had no difference, and the relevant prime repetition did not lead to the reduction of semantic priming effect. In a long 1200ms SOA, there was a significant difference between repeat/same and repeat/related, and the relevant prime repetition do lead to the reduction of semantic priming effect, meanwhile the repeat/same did not lead to any reduction of semantic priming effect. After the offset of expectancy effect and the decay of semantic activation, we believe that the priming effect difference between repeat/same and repeat/related in 1200ms SOA is because of the obstruction of the initiation of spreading. Based on the results and the principle of least effort, we believe that in the long SOA conditions, there is no spreading occurred as scheduled. Therefor we concluded that prime repetition does affect the initiation of spreading in the long SOA, when P2 is not actually masked and can be clearly recognized.