認知能力在工作層次的適應至今仍缺一個機械論，本文意在填補此缺口，尤其是要研究生物體如何可能在與複雜環境互動之下調適自我而完成工作。本文所採取的進路是認知科學的，並從事一個「主動知覺」的個案研究，根基於其演算法而建屯一個新概念──漸增式調適──用以描繪「認知適應」之機制。這個概念將取代傳統機械論的「時鐘」隱喻，並且成為複雜調適系統的一種理論原型。據此概念，本文尋求超越傳統知識論的主客對分，而將世異視為是認知活動本質的一環；並回應匯勒與克拉克（1999）提出的「基因式表徵」，提出更進一步的機制概念來解釋內外認知能力的關連。

　　　　　As yet, mechanicalism has not accounted for cognitive adaptation at the task level, although Darwinian evolutionary biology has made available significant understandings as to the adaptation of organism traits at the evolutionary level. The present paper, based on a case study of active perception, will fill this gap, by providing a mechanicalist account of task-level cognitive adaptation, through the notion of incremental adaptation. This notion is constructed on the basis of algorithms of three active perception systems: attentive control of saccades, Kalman filter implemented in navigation systems, and Dickmanns' car. Such a notion will replace the traditional "clock" metaphor of mechanicalism, and stand as a novel prototype of understanding for explaining Complex Adaptive Systems. The present account, beyond traditional epistemology, obviates a subjectobject dichotomy for understanding the world, and instead puts the world in the context of organism-environment interaction. This paper also provides a complement to Wheeler and Clark (1999), which is a recent endeavour at integrating internal and external cognition.