In the current literature on case study methodology, a ''causal mechanism'' tends to be defined as the process connecting an explanans with an explanandum. Accordingly, a causal explanation consists of a demonstration that the alleged cause brought about an outcome by reconstructing the intermediate steps between the two variables. ''Process-tracing'' functions thereby as the strategy for establishing the presence of such a process. This definition of mechanism (to be referred to as CM1) is not without alternatives, however. Particularly in the Life Sciences, a causal mechanism is understood as a ‘composite hierarchical system’ (Wright and Bechtel). Based on this conception (here referred to as CM2), a causal explanation consists of reconstructing an entity''s components, their relationships and the emergent causal powers of that entity. Even though CM2 does have some advocates in the Social Sciences, particularly among proponents of the philosophical school of Critical Realism, this conception is rarely adopted by Political Scientists. This paper also makes the case for a greater role of CM2 in case study research. Yet, rather than reiterating the philosophical arguments supporting a system conception of mechanisms, this paper takes a more constructive approach by clarifying the relationship between the different definitions of mechanisms and specifying the added value as well as the methodological implications of adopting CM2. I hold that rather than being mutually exclusive, the two definitions focus on different aspects of a causal relation. Whereas CM2 refers to the ability of the explanans to bring about the explanandum, CM1 refers to the transmission of the influence between the two entities. I then argue on the basis of research on causal cognition in Cognitive Science that the comprehension of causation also involves an account of an alleged cause’s ability to produce a change in its environment. Given this cognitive constraint on causal learning, explicitly including models of an explanans'' constitution into case studies would improve the capacity of Political Scientists to generate explanations that satisfy the cognitive requirements for understanding causal relations. Moreover, knowledge about an entity’s composition provides the kind of knowledge needed for intervening in causal processes. With regard to the methodological implications of adding constitutive analysis to the protocol of case study research, I suggest that process tracing (and other types of within-case analysis) are important strategies for determining the boundaries of a causal mechanism (CM2). Given that any entity is constituted by numerous elements, the key challenge for a causal explanation is to identify those parts and their relationships that are relevant for the generation of the outcome in question. In this respect, I find that the logic of inferential strategies used in the Life Sciences to establish constitutive relevance resonates with the principles behind process tracing in Political Science. Thus, although within-case methods would have to be modified to allow for interlevel inferences, the current methodical toolkit for case study research would still provide basic strategies for producing causal explanations that satisfy the demands of CM2.
