DECOMPOSING, RECOMPOSING, AND SITUATING CIRCADIAN MECHANISMS: THREE TASKS IN DEVELOPING MECHANISTIC EXPLANATIONS

Success in reductionistic research in cognitive science or biology is often portrayed as eliminating any need for independent explanations at higher levels. On the standard philosophical account, successful reduction of a higher level science means that its laws can be derived from those of a lower level science and hence perform no explanatory work of their own. But this misrepresents what successful reductionistic inquiry promises or can deliver. At least in the life sciences (including cognitive science), the usual focus of reductionistic inquiry is not the discovery of laws at a lower level than some law of initial interest. Instead, investigators begin with a phenomenon and general idea of the mechanism responsible for it and seek to discover its component parts and operations and how they work together. The focus of actual reductionistic inquiry is the decomposition of mechanisms, not the derivation of laws, and the desire to understand scientific inquiry in this way has led some of us to propose and develop a new mechanistic philosophy of science. Building this new approach has required a variety of case studies of scientific inquiry. Our own most recent case is research on the circadian rhythms exhibited in numerous behaviors and physiological functions. Researchers have had considerable success with the most basic reductionistic task in this field: identifying the parts within organisms that are important to the generation of the rhythms. In mammals, it has been found that many individual neurons in the suprachiasmatic nucleus function as clocks, and that key components include genes such as Period ( Per ) and Cryptochrome ( Cry ) and the proteins PER and CRY into which they are translated. Moreover, some key operations performed by these parts are known: PER and CRY form a compound (dimer) which is transported into the nucleus and inhibits Per and Cry , hence reducing the rate of production of further molecules of PER and CRY. Reductionistic research in the last 15 years has succeeded in identifying these and many other parts of the clockworks. Such inquiry, no matter how successful it is in finding the parts and characterizing the operations they perform, does not suffice to explain circadian phenomena. The operations performed by the parts in individual cells are organized and orchestrated such that the cell functions as a unit – one that displays complex temporal dynamics. Moreover, there are operations between SCN cells that synchronize their oscillations and between SCN cells and the receptors responsive to environmental cues that entrain the clock to the local time and between SCN cells and the many bodily organs that exhibit circadian behavior. Finally, there are operations connecting the organism to the environment, especially to sources of light and temperature. None of these operations at higher levels are discovered by focusing on the operations involving genes and proteins inside SCN cells—they require tools and techniques appropriate to the level at which the operations are occurring. An especially challenging part of inquiry in the life sciences involves relating parts and operations at different levels. Within the mechanistic framework, these are best handled not by invoking notions such as top-down or bottom-up causation, but by understanding the constitutive relation between a mechanism and its component parts and operations. When the mechanism is affected by operations impinging on it, so are some of its components. Conversely, when some of its components are changed by being operated on by other components, the mechanism as a whole and the operations in which it engages are changed. The ontological picture, as exemplified in modern biology, is one in which the capacity of mechanisms to operate in their environments is explained at lower levels but the mechanisms (as wholes) interact causally with other mechanisms at higher levels. This picture iterates as one goes to even lower levels by decomposing a part of the original mechanism into its parts or to higher levels as one treats the original mechanism as a part in a yet higherlevel mechanism. There are operations at multiple levels of organization and no level is eliminated by discovering the operations within it that enable a given mechanism to interact with others at its level.