像古典力學一類的科學理論如何被檢驗？這個問題有個長遠的歷史。在本文中，我將提議一個檢驗理論，建基在吉爾關於科學理論結構的研究上。但我的模型觀點與吉爾有些微不同。我將從理論性和歷史性兩個不同角度來展示：科學理論總是可以被理解為擁有一個分類的模型體系的理論，其中包括高層的模型與可落實模型；而且科學家總是透過可落實模型來檢驗理論。但是，透過可落實模型把檢驗的結果傳送到高層的模型或原理，乃是一件非常複雜的機制。因此，就算有一些可落實模型被印證或否證，整體理論完全被印證或否證實際上不太可能。最後我將以說明牛頓理論的模驗史為例，來佐證本文提議的檢驗理論。以拉卡托斯的話來說，這個檢驗理論在本質上也是個針對歷史的「合理重建」。

How is a scientific theory, especial a classical physical theory, tested? This problem has a long history. In this paper I’ll propose a theory of testing based on but differentiated from Giere’s studies on the structure of scientific theories (Giere 1988, 1994, 1999). I will show, from both theoretical and historical perspectives, that a scientific theory can always be understood as one contains a classified model population, including both higher-level models and realizable models, and that scientists test a theory through its realizable models. To transmit the consequences of testing realizable models to a higher-level model is a very complicated mechanism. Therefore, it is unlikely that a whole theory could ever be completely confirmed or falsified, even if some of its realizable models have been conclusively confirmed or falsified. Finally, I’ll illustrate such a theory of testing can give an adequate account of the testing history of a scientific theory, for example, the Newtonian theory. This theory of testing is a rational reconstruction, in Lakatosian sense, of the process of scientific testing.