本研究主要是以知識表徵系統的網路模式理論為主，以國小自然科學生學習成就為例，探訪國小學生的知識結構評量分析。 研究方法主要是量化的分析，以自然學科學能力為特定領域，利用路徑探測網路測量知識結構，探討知識結構與自然科學能力的關係，同時分析三個知識結構指數（PFC、PRX、GTD）對於受試者學習成就的解釋力。進一歩分析旨較不同學習成就、專家（老師）與生手（學生）在知識結構上的差異。 樣本採字簡單隨機抽樣，對台南市崇學國小五年級全體學生進行抽樣，抽出38位同學為研究樣本。分別對其施予研究者自編之國民小學學生知識概念表徵上之的差異。 本研究歸納提出下列結論： 一、知識結構三項相似性指數（PFC、PRX、GTD）與自然科學能力都有顯著的相關。同時，三個相似性指數對自然科學能力的解釋力，由高而低依序為PFC、GTD、PRX。 二、徑路探測網路分析在評量國小自然科目領域上的學生知識結構及能力是有其效度的。 三、徑路探測網路分析專家與生手，是有其區辨力的。 四、由專家與生手其概念結構圖中可以發現，其概念結構愈具有階層性，其概念結構與專家的概念結構有愈相似的趨勢存在。 五、本研究的概念選取偏少，因此無法獲得充分訊息，是為本研究的限制。

The purpose of this research is to evaluate and analyze the knowledge structure of elementary students in the method of network model theory of knowledge representation system, then further apply it to the learning achievements of natural science courses of elementary school students. This study will discuss the relationship between knowledge structure and learning capability of the natural science discipline by way of quantitative analysis. That is, in the scope of learning capability of natural science courses, this research probes the network as well as measures the knowledge structure. Meanwhile, three knowledge-structure indices, PFC, PRX and GTD, are evaluated to predict the learning achievements of students, then used further to analyze various learning achievements, and different learning structures between experts (teachers) and novices (students). Under the simple random sampling methods, a total of 38-student samples were selected out of all fifth-grade students of Chung-Share Elementary School in Tainan City as the population of this research. All fifth-grade students were firstly evaluated with the common basic concept scale of natural science discipline designed by the author. Then through an analytic software, which is called KNOT, the collected data are further analyzed to figure out the differences in knowledge concept representation among elementary school students. The following conclusions are reached from this research: 1.There is a significant relationship between the three similar indices (PFC, PRX, GTD) of knowledge structure and learning capability of natural science courses. And among these three indices, the predictability towards the learning capability of natural science courses rankings from high to low are PFC, GTD, then PRX. 2.The route probing network analysis methods proved effective in evaluating the knowledge structure and capability of students in natural science courses. 3.The route probing networking analysis method is able to differentiate the expert from the novice. 4.From the concept structure chart of the expert and notice, it is found that the more hierarchy that exists in the concept structure the closer it is to the expert’s concept structure.