A recent trend within many scientific domains is the organization of information through the use of ‘ontologies.’ Ontologies allow scientific data and theoretical information to be expressed with the use of first-order logical systems, allowing for the creation of categories of objects and properties and the identification of relationships which hold between those categories. By organizing the information in this way, scientists are able to more effectively use the complex, overlapping web of information in wider scale projects, such as linking genetic information with relationships that hold between the levels of organization within a particular phenotype. This is undoubtedly helpful and a worthwhile project to be engaged in. Recently, there has been an increasing trend to try to get all ontologies to fit under the scope of a broader “upper level” ontology. These ontologies, such as the “Basic Fundamental Ontology” developed by Barry Smith and colleagues, attempt to identify the basic foundational categories and relationships of the world. The idea is then for the users of a wide range of domain-level ontologies to be able to nest each of the domain-level categories underneath one of the members of the upper level ontology, using one of the relationships of the upper ontology. The aim would be to ultimately have a large amount of scientific data and theoretical knowledge from the whole gamut of scientific domains to be nested under the same upper level ontology, allowing for the potential for interesting insights that might otherwise be missed. While I admire the pursuit of interdisciplinary thought that this attempt is at least partially founded upon, I think the use of upper level ontologies should be abandoned. Of primary concern is the problematic way in which the use of upper level ontologies with unknown philosophical commitments might create ways of understanding scientific information which precludes the full semantics of a theory (or a data model) to be fully expressed. The trouble could be identified in the attempt to reduce this information to a matter of first-order logic (even with the inclusion of temporalized logics, which have created their own problems). If we ignore this problem, there is another issue insofar as the basic relationships used by upper ontologies, although described in great detail in handbooks, are still vague enough that different scientists could use them in different ways. Even if these relationships were more solidly defined, there would still be trouble insofar as the relevant relationships of consideration in one domain (e.g. physics) might be non-starters in a different domain (e.g. biology). Another problem with the use of upper level ontologies is the way in which they are not created with the domain-specific knowledge in mind, and the way in which this could be problematic in the development of theories and the way in which questions will arise within the work of a field. Ultimately, the use of upper level ontologies requires a metaphysical commitment which has the potential to create problems within the practice of scientific investigation.
One of the most conspicuous developments in the philosophy of science over the past fifteen years has been an increasingly central concern with elucidating the role that mechanisms play in science, especially in the biological sciences. Much of the philosophical attention has focused on developing general accounts of mechanism that do justice to the way the term is used in scientific explanation. Although there is little agreement over how best to define this concept—Machamer et al. (2000), Glennan (2002), and Bechtel and Abrahamsen (2005) are the three most influential accounts—there is close to universal agreement regarding their metaphysical status. Whatever else they may be, one thing everyone appears to agree on is that mechanisms are “real systems in nature” (Bechtel 2006); that is, that they are “real and local”, as the title of a recent paper makes explicit (Illari and Williamson 2011). The reason for this consensus has to do with the way we tend to think about paradigmatic mechanisms of our everyday experience like a clock or a fridge. These are clearly “real and local,” and are of course “real systems in nature”. But does this realist understanding remain appropriate when ‘mechanism-talk’ is applied to biological phenomena? The history of the usage of the concept of mechanism in biology reveals that term has gradually come to be used to designate an extremely wide range of processes (such as natural selection, inheritance, or the immune response), and in doing so, it has lost its original machine connotations, becoming a dead metaphor. Unlike other scientific terms like microtubule, mitosis, or metabolism, ‘mechanism’ is not a technical concept; it does not appear in the glossaries of biology textbooks, nor is it listed in its indexes. Instead, it is a term that simply ‘comes up’ in scientific practice, and its meaning is inferred from the explanatory context in which it is invoked. Most philosophers have assumed that one thing that has remained attached to the mechanism metaphor as it has been imported into biology is that it still refers to ‘real systems in nature’ (like machines such as clocks and fridges). I challenge this conviction by taking seriously two implications that follow from the realist conception of mechanisms. If biological mechanisms are ‘real and local’, we should be able to answer two key questions: (a) how many mechanisms make up an organism? and (b) when is a description of a biological mechanism complete? By showing the impossibility of providing principled, unambiguous answers to these questions I will show that the best way to understand biological mechanisms is not as ontological building blocks of the living world, but as abstract and idealized spatiotemporal cross-sections of biological processes that heuristically pick out certain causal relations involved in the production of the phenomena that biologists are interested in explaining.