多重实现是支持反还原主义的重要论据，但学界尚未对多重实现的界定标准达成一致。波格尔与夏皮罗提出了多重实现的严标准，该标准要求实现者与被实现必须是相关的。根据严标准，多重实现在自然界中十分罕见。相泽与吉列特提出了多重实现的宽标准，该标准认为事物的组合要素的差异足以构成多重实现。根据宽标准，多重实现在自然界中普遍存在。本文分别对严标准和宽标准提出批评。严标准的缺陷在于：（1）严标准是特设的，它缺乏清晰一致的标准判断哪些实现者对被实现者是相关的；（2）严标准对于多重实现的刻画是不必要的，它无法刻画因果-历史属性的多重可实现性。宽标准的缺陷在于：（1）宽标准会带来因果冗余问题，被实现者的因果力是冗余的；（2）宽标准对于多重实现的刻画是不充分的，它无法排除琐碎的多重实现。在此基础上，本文结合严标准和宽标准各自的内容，提出多重实现的适中标准。适中标准刻画了因果-历史属性的多重可实现性，因此比严标准更加宽松；适中标准排除了琐碎的多重实现案例，因此比宽标准更加严格。此外，多重实现的适中标准能够避免因果冗余论证的威胁。综上，相对于严标准和宽标准而言，适中标准更好地刻画了多重实现的范围，因此更适宜作为多重实现的界定标准。根据适中标准，多重实现在自然界中普遍存在；就此而言，本文更偏向反还原主义的立场。

The thesis of multiple realization serves as a significant argument for anti-reductionism, but philosophers have not reached a consensus on the criterion of multiple realization. Tom Polger and Lawrence Shapiro’s strict criterion of multiple realization requires that the realizers must be relevant to the realized properties, claiming that multiple realizations are quite rare in the natural world. On the other hand, Kenneth Aizawa and Carl Gillett’s broad criterion of multiple realization allows that mere compositional differences in realizers make for multiple realizations, claiming that multiple realizations are common in the natural world. In this thesis, I argue against both strict and broad criterion by revealing their defects. The strict criterion is deficient in the sense that (1) the strict criterion is ad hoc, since there is no clear way to figure out which realizers are relevant to the realized ones; (2) the strict criterion is not necessary to characterize the multiple realization, since it fails to capture the multiple realizability of etiological properties. The broad criterion is deficient in the sense that (1) the broad criterion brings about the redundancy problem according to which the realized properties might be causally redundant; (2) the broad criterion is not sufficient to characterize the multiple realization, since some trivial instances cannot be excluded. Based on strict and broad criteria, I propose to develop a moderate criterion of multiple realization. The moderate criterion captures the multiple realizability of etiological properties, and excludes the trivial instances of multiple realizations. In addition, the moderate criterion is able to avoid the threat of causal redundancy. Therefore, in contrary to strict and broad criteria, the moderate one is more plausible to characterize the concept of multiple realization. According to the moderate criterion, multiple realizations seem to be pretty common; in this case, I agree with anti-reductionism.