This paper preliminarily researches the question making use of computational simulation. The results show that there exists an optimal interval among the different ratios of explicit and implicit knowledge base according to structural characteristics of small world networks. In this interval, the growth of network clustering coefficient is faster than that of the average shortest path length. From the evolutionary point of view, compared with network clustering coefficient, there exists a decline lag effect of the average shortest path length of networks, and when the ratio of explicit and implicit knowledge base falls within the optimal interval, the lag effect is more obvious.