Top-Down Causation in Biology and Neuroscience
Control Hierarchies William Bechtel
The notion of top-down causation has been fraught with controversy. Much of this turns on the notion of levels employed. What is it for one entity or causal process to be located at a higher level than another? In the context of biology and neuroscience, an important sense of level arises in the context of control—a controller is at a higher level than the system it controls, and if something is controlling the controller it is at a yet higher level.1 Thus, transcription factors are at a higher level of control than the genes whose expression they regulate, and neurons are at a higher level of control than muscles and other cells. The circadian clock is at a higher level than the transcription factors whose expression it regulates, and regions of cortex are at a higher level than sub-cortical areas they regulate. My goal is to unpack the notion of top-down causation required to understand the operation of control hierarchies that figure prominently in biology and neuroscience.
Control is exercised on a controlled system. A controlled system consists of a set of processes that causally interact and together bring about some effect. Human-made machines are exemplars of such controlled systems— an automobile consists of a number of parts that perform various different operations that together result in locomotion. In the context of biology and neuroscience, controlled systems (as well as controllers) are commonly referred to as mechanisms. In the recent literature on mechanistic explanation, mechanisms have been identified as entities or parts performing activities or operations organized so as to bring about a phenomenon (Machamer, Darden and Craver 2000; Bechtel and Abrahamsen 2005). For example, the heart circulates blood (the phenomenon) as a result of consisting of chambers in which muscles (parts) contract (operation) and valves (parts) limit flow to one direction (operation) in an organized and orchestrated manner.
Although not generally emphasized in philosophical accounts of mechanism, machines as well as mechanisms can be viewed as systems performing work by constraining the flow of Gibbs free energy2 (e.g., a pipe channels the free energy of water flowing downhill to move an another object). This work is often done in the service of a larger system of which the machine or mechanism is a part. For work to be performed in a manner that is useful to the larger system, control is needed. This requires that some of the constraints in the machine or mechanism be modifiable; control is exercised by altering these constraints, thereby redirecting the flow of free energy. In the case of a machine such as an automobile, the driver exercises control by, for example, pressing on the accelerator pedal. In traditional engines, there is a linkage from the accelerator pedal to the butterfly valve on the carburetor. The more the valve (the constraint) is pushed open, the more air, along with fuel, enters the combustion chambers of the engine. As a result of the increase of fuel and air, the combustion exerts more force, speeding up the engine’s operation. In a living cell, control is also exercised by altering constraints. An enzyme constrains a biochemical reaction and changing the concentration of the enzyme alters the rate of a reaction. The concentration is increased by an activator binding to the promoter site of a gene, allowing more transcription of that gene. Likewise, in a multi-cell organism, for control to be exercised there must be constraints that can be altered. To increase the flow of blood, the contraction of muscles in the various chambers must be increased. This is accomplished through the release of neurotransmitters that bind to receptors in the muscle cell, permitting the formation of cross-bridges between actin and myosin.
To provide a foundation for discussing control of mechanisms, I begin in section 2 by advancing a perspective that situates mechanisms as modules in networks whose endogenous function is modulated by activity elsewhere in the network. In section 3, I turn to human-made machines to introduce a basic mode of control realized by negative feedback. In section 4, I turn back to organisms and discuss why control is even more fundamental in understanding biological mechanisms than in the case of human-made machines and in section 5 consider cases in which feedback provides the needed control. Negative feedback not only is employed directly to control biological mechanisms but also, as I discuss in section 6, is a means of generating oscillations that facilitate controlling at what time a mechanism is operative. In section 7, I turn to neural control, emphasizing its importance in providing hierarchies of control in multicellular organisms whose component cells and mechanisms are endogenously active. I then conclude by emphasizing that top-down causation, as exhibited in the hierarchical control of biological mechanisms, is a fundamental feature of biological and neural systems. Such top-down control doesn’t pose any fundamental mysteries since the control mechanisms as well as the controlled mechanisms are all constructed by ordinary mechanisms within the organism.
Top-Down Causation in Biology and Neuroscience 205