Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
History of subatomic physics wikipedia , lookup
Introduction to gauge theory wikipedia , lookup
Modified Newtonian dynamics wikipedia , lookup
Standard Model wikipedia , lookup
Renormalization wikipedia , lookup
Time in physics wikipedia , lookup
Yang–Mills theory wikipedia , lookup
Condensed matter physics wikipedia , lookup
History of quantum field theory wikipedia , lookup
Electromagnetism wikipedia , lookup
STUART S. GLENNAN MECHANISMS AND THE NATURE OF CAUSATION In this paper I offer an analysis ABSTRACT. of causation based upon a theory a system's nisms "internal" parts interact to produce systems whose complex can be explained I argue that all but the fundamental laws of physics behavior. by an epistemologically to mechanisms. Mechanisms way provide unproblematic of mecha? "external" reference to explain This generalizations. laws from other the necessity which is often taken to distinguish account to which of necessity leads to a theory of causation according reasons related when that connects them. I present there is a mechanism account of fundamental physical causation does not undermine events why the mechanical are causally the lack of an account. 1. HUME'S PROBLEM [Experience only teaches us in the secret connexion, us, how which one event binds constantly them together follows another, and renders them without inseparable instructing (Hume, 1777, p. 63). to Hume, cannot tell us about the "secret con? according binds together events. When we attend to a supposed causal billiard ball colliding with a stationary interaction, for instance, a moving billiard ball, we can observe the motion of the first ball and then the motion of the second, but we can not observe a connection between the two. Fur? would allow us to observe thermore, no number of further observations Experience, nexion" which any connection. Hume's in nature, is that, although we can observe problem we can never see the "secret connexion" together. Itwould seem therefore one event to follow another. His define the notion regular conjunctions which binds them that we can never have know what causes to this problem is to no reference to that it makes solution" "skeptical in such a way to constant conjunction. of cause a connection, but only He defines a cause to be "an object, followed by another, and where all the objects similar to the first are followed by objects similar to the second" (Hume, 1777, p. 76). This definition is the germ of the regularity theory of causation which has of causation ever since.1 empiricist discussions Hume's can yield Although argument that no number of observations an impression of a connection in at view is, least, irrefutable, I do not my dominated Erkenntnis ? 44: 1996 Kluwer 49-71,1996. Academic Publishers. Printed in the Netherlands. STUARTS. GLENNAN 50 think that that argument requires us to adopt a regularity view of causation. In this paper I will try to suggest an alternative that I call a mechanical behind the The intuition of causation. theory is straightforward. theory I claim that some event causes another event, say that my turning When I have this simply because the key causes my car to start, I do not believe routinely I believe that turning the key is followed by the engine starting. I believe that connects that there is a mechanism observed this because to engine-starting. I believe that the key closes a switch which key-turning causes the battery to turn the starter motor and so forth. Furthermore, this I can look under the hood and see how the is not a "secret connexion". mechanism There works works.2 is an obvious to this sort of explanation. it Although objection the key starting the car, there is a very important for instance, to look for which it does not. It is not possible, for cases class of cases at the mechanism like which causes other. So far as we understand attract each to gravitationally there is no this interaction between bodies, a It "brute fact" about the it. is just explains two bodies which underlying mechanism we In this case we live. in which world cannot explain the interaction by form of regularity theory seems theory is not that it is an incorrect to any mechanism. Here some with the The regularity problem plausible. analysis of such cases, but that it fails to distinguish The vast majority there is a discernible mechanism. reference these from cases where of cases are of the latter sort. My aim in this paper is to offer an analysis of the concept of mechanism. can be put of the paper I indicate how this analysis the conclusion I emphasize the fact that this theory to work in a theory of causation. I suggest that there should in fundamental cannot explain causation physics. the case of between of causation in our understanding be a dichotomy At fundamental physics and that of other itself). Mechanical theories the Mechanist movement of causation sciences (including are not new. The much central of physics tenet of century was that all natural com? as the result of the action of mechanisms in the seventeenth phenomena were explicable To say that some event caused another event was posed of corpuscles.3 the two, where this linkage between just to say that there is a mechanical or a as bodies was of collection understood larger rigid corpuscles linkage to provide movement pushing on each other. The failure of the Mechanist I an enduringly stems, think, from two prob? adequate account of causation - one scientific and one The scientific problem has lems epistemological. of matter and views about the microstructure to do with the Mechanists' the forces which govern interactions between matter. Mechanists advocat MECHANISMSAND THENATUREOF CAUSATION 51 style theories in which the universe consists of a large (or of corpuscles which interact with each other only by infinite) collision of (or perhaps more extended pushes). The admissible principles interaction are what we would call "strictly mechanical i.e., principles", the sort of principles with which one would construct amechanical (rather ed Democritean collection non than say an electronic) device. It was believed that all apparently mechanical and magnetism, could ultimately be forces, such as gravitation as the actions of mechanisms of explained consisting pushing corpuscles on each other. This point of view has proved untenable in light of sub? scientific sequent developments. have resisted Physical forces such as gravitation and narrowly mechanical explanation. with Mechanism has problem seventeenth-century to do with the testability of mechanical theories. Mechanist explanations are often dominated of natural phenomena by baroque accounts of mech? anisms that are not of discernible (see, e.g., the consequence physical electromagnetism The epistemological discussion of gravity in Descartes, Descartes and other 1664). Although that all phenomena claimed could be explained in terms of of corpuscles, size, shape and motion very little was said about how these and their could be observed (or how we could make corpuscles properties Mechanists inferences about them on the basis of observables). that I will develop is largely inspired by I hope, however, that my anal? insights of the Mechanical philosophers. two will avoid these In the first must be mechanisms ysis pitfalls. place, in such a way that there are not a priori restrictions on the sorts conceived The account of allowable of mechanisms interactions which may take place analysis of causal connections when there are ways (even between a mechanism's in terms of mechanisms if indirect) of acquiring of their parts and the interactions between them. knowledge In the remainder of the paper I will present an account of mechanisms and show how it provides the foundation for a theory of causation. Section parts. Additionally, is only meaningful 2 of this paper presents my analysis of mechanisms. Section 3 applies to two simple systems. Section 4 discusses this analysis some relation? and laws. Section 5 shows how the analysis of ships between mechanisms can mechanisms be used to formulate a theory of causation. 2. are two senses There AN ANALYSISOFMECHANISMS in which the term 'mechanism' is commonly used. refers narrowly to the internal works of machines, as when of a clock mechanism. The second refers more generally to The first sense one speaks complex systems analogous to machines, as when one speaks of a human STUART S. GLENNAN 52 perceptual mechanism marized by a defmition or a market which mechanism. is meant My to capture analysis can be sum? this latter usage: a behavior is a complex system which underlying that behavior by of the interaction of a number of parts to direct causal laws. A mechanism (M) produces according a behavior' of a 'mechanism Notice that (M) is a definition underlying even a cannot rather than amechanism One mechanism simpliciter. identify without does. The boundaries of the saying what it is that the mechanism of it into parts, and the relevant modes of interaction these parts depend upon what the behavior we seek to explain. Furthermore, systems do many complex things at once. If one isolates a one can identify system by some kind of physical complex description, system, between the division ofthat 1970). A complex system (Kauffman, indefinitely many behaviors its different behaviors. has mechanisms many system underlying behavior of complex The polymorphous systems can be illustrated by the behaviors of the human body. Two of the many subsystems considering and respiratory body are the cardiovascular for has mechanisms systems doing certain things and with carbon-dioxide) inhaling oxygen and exhaling Each of of the human systems. these (pumping blood, regard to one of interact in such a way that the two systems blood) (oxygenating a as must These divide the considered be systems body composite. they it into the heart, system divides up in different ways. The cardiovascular etc. The respiratory system divides it into lungs, veins, arteries, capillaries, these etc. The physical of the systems extensions diaphragm, windpipe, mouth, there are, e.g., veins and arteries running through and their parts overlap; the various parts of the respiratory system. The choice of decomposition to be explained into parts depends upon the capacity or behavior (Wimsatt, 1972). Although the choice of decomposition are not merely is being upon what depends artifacts of the description. Veins bodies, even though they overlap. decompositions explained, and lungs are both really parts of human are good descriptions of mechanisms Descriptions what is "really" of dependence there. This point deserves decomposition systematic realism or relativism. Simple examples there is no such implication. insofar as they describe the context emphasis because to imply anti such as the one above show that is often taken was designed amechanism may be something two not of a behaviors be. Consider to do (or selected for), but it need of heat. of a drive shaft and the production combustion engine: the motion Either of these behaviors may be legitimately explained. mechanically The behavior in question MECHANISMS AND THE NATUREOF CAUSATION However, produced or systems is designed the engine a side-effect. is merely to move When the drive one considers 53 shaft, while designed the heat artifacts con? pressures, explanatory text often dictates analyzing the system in terms of its production of the or selected behavior, but one could choose any other behavior designed as In of these systems well. which fact, one can investigate mechanisms that have evolved under selection sense be said to have a purpose. One can con? in any interesting solar the the motions of sider, e.g., system as a mechanism underlying the earth (or the planets) even though one believes that this motion is not cannot purposive.4 In order for (M) to be sufficiently it is important that a very general of entities be of mechanisms. Parts may be simple may parts variety or complex in internal structure, they need not be spatially localizable, and in a purely physical In certain they need not be describable vocabulary. wide to consider genetic for instance, one might wish contexts, whose parts are genes or information processing mechanisms are software modules or data structures. There are, however, mechanisms whose parts certain kinds of entities which, to prevent to be parts of mechanisms. robustness (M) from being vacuous, should not be allowed must have a kind of The parts of mechanisms from their within that mechanism. It apart place and reality to take the part out of the mechanism should in principle be possible and consider its properties in another context. Care must be taken so that parts are neither merely properties of the system as a whole nor artifacts of the I shall summarize these restrictions descriptional vocabulary. by saying that parts must be objects.5 can be illustrated by considering The significance of these restrictions to give a mechanical whether or not it is possible of the behav? explanation ior of the electromagnetic field (as it is codified by Maxwell's I equations). in the only natural decomposition it is not, because, of the field into parts, the parts of the system are not objects in my sense. This case is sig? the electromagnetic nificant because field is an example of a law-governed believe entity whose behavior The electromagnetic is not subject to mechanical explanation. field is an important part of many mechanisms, from particle accelerators to TVs to the mechanism which produces the It is probably fair to say that electromagnetic Aurora Borealis. fields play a role in producing the great majority of physical phenomena. Furthermore, are fields electromagnetic objects of a kind. They have a variety of prop? erties for instance energy and momentum. The properties of a field can be completely described field B. magnetic by two vector fields, the electric field E and the 54 STUART Is there amechanism S. GLENNAN that explains the properties of the electromagnetic field? If there is, then it should be possible to decompose the field itself into parts. There is at least one sense to talk about the electric appropriate in which this can be done. It is quite in a region - for or magnetic field two capacitor plates, or the magnetic instance, the electric field between field of the earth. These fields are parts of the electromagnetic field, if as we conceive the entirety of space. What of this field jus? occupying is their relatively these parts as objects tifies characterizing high degree this kind of from the field in surrounding of separability However, space. to describe a mechanism that of parts would not be adequate articulation an electromagnetic field. produces articulation into parts consistent So far as I can see, the only possible field involves of the with the classical thinking theory electromagnetic as in Each of these of the parts of the field space. parts have the points fails field strength. Such an analysis of electric and magnetic properties in my sense. Such points are not because points in space are not objects to differentiate It is not possible in themselves. isolable or manipulable field that are suffi? between points in the electromagnetic experimentally in between points ciently close to one another. There are no boundaries the field between there are boundaries in which the way between, e.g., to alter fields.6 It is not possible the capacitor plates and the surrounding at a single point. The points, while part of the of the field as a whole, have no physical signif? account It is crucial to a mechanical icance apart from this description. that a system display some behavior which can be explained by reference But there are not, in this case, of its constituents. to underlying properties "Maxwell's with underlying constituents properties. As Hertz remarked, field the electromagnetic mathematical description equations". theory isMaxwell's be different The situation would if there were a detectable aether. If an underlying of for the transmission medium in any to then it might have been possible and magnetic electrical disturbances, of Itmight have been composed the properties of this medium. investigate these how forces have been there and particles explaining might particles interacted when disturbed by external sources (a light bulb or whatever). to our accepted physical But according theory there is no aether and there sense there were is no mechanical explanation in any physically meaningful between The interactions governed Goodman. of the electromagnetic sense go deeper.7 field. One cannot then to (M), are, according parts of mechanisms as same the use term in here 'law' I the laws. way essentially by which universal Laws are generalizations sup? (or propositions) port counterfactuals. Lawlike or nomic generalizations are distinguished MECHANISMSAND THE NATUREOF CAUSATION 55 from accidental because accidental offer generalizations generalizations no such support (Goodman, ch. 1947; Nagel, 1961, 4). Some may account of laws is not adequate argue that Goodman's because it does not provide a criterion for demarcating "deep" laws (like law of universal Newton's from less interesting counterfactual gravitation) whenever you leave bread on the counter (like supporting generalizations for two weeks, itmolds). Although it is true that Goodman's account does not provide such a criterion, for the purposes of my account of mechanisms, the lack of a clear distinction between these two sorts of counterfactual is a virtue. The laws used to describe the inter? supporting generalizations actions between parts of mechanisms and the laws which can be explained can be of the most profound or banal sort. In a by mechanisms explaining we must whose parts are interacting gravitationally invoke a rather deep law, while to explain mechanisms like lawn mowers we most? counterfactual such as ly invoke uninteresting supporting generalizations, mechanism laws about A more it appeals it seems the behavior of valves. tomy use of Goodman's serious objection to an unanalyzed notion of counterfactual analysis of laws is that support. Furthermore, that any analysis of counterfactuals that we could give will to causal notions. Given that I intend to use implicitly or explicitly likely appeal to explain the nature of causation, this appeal my analysis of mechanisms threatens to make my analysis of causation circular. In Section 4 I will can be appealed indicate how mechanisms to in a non-circular manner to explain the notion of counterfactual support. The final part of (M) that I wish to clarify is the stipulation that the laws governing interactions between parts of mechanisms laws. The stipulation that the laws be causal is meant be direct causal to exclude lawful which can be explained by common causes. For instance, generalizations it is a lawful generalization that night follows day, but certainly day does not cause night. Rather, the onset of day and of night are events which are both caused by the earth's rotation. Relations between parts must be otherwise the parts could not be said governed by causal laws because truly to interact.8 The further stipulation that causal laws be direct can be illustrated by the following a system consisting Consider of a series of three example: or more gears of various sizes. Given information about the number of teeth on the gears, one can state a law Li describing the rotation of the last does not gear as a function of the rotation of the first gear. Li, however, describe an direct interaction between the first and last gear. The interaction is mediated transmit the rotation from the first to the by other parts which last gear. By contrast, a law L2 describing the rotation of the second gear STUARTS. GLENNAN 56 as a function stipulation the behavior of the first is the sense is direct. What that a mechanism we are trying is a collection to capture by this of parts, in which stems from a series of local interactions the aggregate an In that immediate interaction is local, I am not parts. saying is that the interaction but supposing spatially local, only that there are no were a For if of instance, parts. intervening gravitation genuine example action at a distance, the gravitational interaction between the earth and the of between sun would be direct, because there would be no intervening that the notion of directness is relativized the consequence parts. This has to a particular 1992, ch. 3). (Glennan, decomposition The analysis that I have of mechanisms a number given resembles of decompositional for Kauffinan's artic? strategies explanation, notably ulation of parts explanation Cummins' and functional (Kauffman, 1970), and Haugeland's (Cummins, 1975,1980), system? analysis morphological several important atic analysis 1978). There are, however, (Haugeland, of these authors have differences. each the role of these First, emphasized a within for Cummins science strategies special (psychology explanatory and biology for Kauffman), whereas my account is intend? and Haugeland, I have fundamental ed to apply to all sciences Second, except physics. between tried to illustrate the connection decompositional strategies and most I have and role of mechanisms. the explanatory Finally, importantly, (and will at the end of this paper) argued and more generally, causal explanations, are that these sorts of explanations two events that a relation between is causal when and only when events) (other than fundamental physical in the appropriate way by a mechanism. these events are connected 3. can be analyzed two simple mechanisms in a to the The first (M). system regulate example, suggested by account. in a toilet tank, is clearly a mechanism by anyone's In this section the manner water TWO SIMPLEMECHANISMS level Iwill show how to which the parts interact according (excepting gravity) principles a voltage switch, relies on are "strictly mechanical". The second example, I will show that both of "non mechanical" different, operation. principles The of these systems may be analyzed are mechanisms inmy sense. A. A Float the lines of (M), and thus that they Valve a simple mechanism to regulate level in a the water a float valve. It should be familiar to anyone who has in accordance with the top of their toilet. Let us consider, 1 pictures Figure tank. It is called ever along opened MECHANISMSAND THE NATUREOF CAUSATION 57 Fill Point Water Fig. Source 1. A float valve. (M), what the behavior of this mechanism is, what its parts are, and what occur between causal interactions these parts to produce the behavior of the mechanism. The purpose of a float valve is to regulate the water level of a tank, so it is natural (though not required) to single out the maintenance of a certain water in the tank (the fill point) as the behavior of the mechanism. The operation of the mechanism is quite simple. A float is attached to a lever which opens and closes an intake valve. When the lever level is down the intake valve is open, allowing pressurized water to fill the tank. the lever is raised to a certain point, the intake valve closes, stopping the flow of water. The float is heavy enough that in the absence of water When it will pull the lever down, opening the intake valve. On the other hand, it is sufficiently that the buoyant rising water level will force the float up, water the intake valve when reaches the fill point. closing In describing number of parts: a the operation of this mechanism I have articulated the tank, the valve, water the pressurized source, the lever and the float. I have also specified the ways in which these parts are that is to say, I have described the causal interactions between connected; one one the parts. If could formulate precise laws describing these wished, interactions. One would do this by specifying of the parts, and properties to represent these laws. equations of the valve raises an important issue concerning the of mechanisms In the above description I have treat? generally. using mathematical The description description ed the valve as a kind of "black box" switch. It could be replaced by any into the tank only while the lever is in I have not specified how the valve itself works, only how certain positions. to the operation as a whole. it contributes of the float valve mechanism some more detail. According The accompanying to the diagram suggests a the valve is As the lever is the is low? raised, diagram, piston. piston ered by a mechanical Since under normal conditions linkage. operating device which allows water to flow STUART S. GLENNAN 58 SOURCE R2 'out in Ri Ground Fig. 2. A voltage switch. laws, it is easy to treat it as a black of its operation. The valve is itself box, abstracting a mechanism within the larger water-level regulator mechanism. Though I have treated the valve as a simple part, one could equally well remove the valve behaves according away from to simple the details to the piston, to the valve, and replace it with references the is the lever-piston chamber, linkage, etc. This kind of reductive analysis not limited to the valve. One could also, for instance, give an explanation the water and the float; One could, of the details of the interactions between so to speak, take the buoyancy mechanism out of its black box. In general, references are inex? the parts of a mechanism to it will be possible in terms of any deeper physical mechanism, plicable take apart the parts and look at how they themselves work. unless the laws of interaction between B. A Voltage Switch The float valve by pushing valve; water is amechanism each other. Water not understand ceases in the ordinary sense. The parts all interact pushes float; float pushes lever; lever closes century scientists would as we as well do today, the qualita? count as to them. The float valve would to push. Though the fluid mechanics were seventeenth well known in Figure 2 switch illustrated their lights. The voltage by an electric rather to This is different operates according quite principles.9 switch. There are, at least at the macrophysical than amechanical level, no tive principles a mechanism it is easy to provide a mechanical parts. Nevertheless, moving to the one provided for the float valve. this circuit analogous analysis of MECHANISMSAND THE NATUREOF CAUSATION V 59 SOURCE ?IIIIHII ?| Cutoff?? ^ || Region ?|| \ Saturation A* /.Region.. A A A A A A ..** A (vrF) CESSATlliilil?ll "?"a"?"?"a"a fi. A A A A A A VYOFFVvON Fig. Already 3. Input/output behavior for a voltage switching circuit. in describing the circuit in Figure 2 as a voltage switch, I have a characterization of its behavior. The switch has two terminals, suggested an input and an output. The behavior this circuit as a switch is the variation of importance of input voltage can be described in characterizing V{n with respect to output voltage Vout. This behavior quantitatively by = For circuits of this type, the graph of this giving a function Vout f(Vin). function has a distinctive shape illustrated in Figure 3.When V{n is below a certain voltage Voff, Vout is constant and equal to the voltage of the power source VsourceF?r values of V{n below T^ff the transistor is said to be in the cutoff region. Then, for a certain interval beyond T^off (the transient region), Vout decreases linearly as V{n increases. Finally, approximately as Vin increases beyond the voltage Vqn? the output voltage levels out at a value near to 0, the collector-emitter saturation voltage (T/ce)sat? and constant for values of If V?n. V{n < Voff, approximately larger the switch is off, indicated by Vout = ^sourceIf Vin > ^on? the switch is on, indicated by Vout = (Vce)sat ? 0. remains To analyze the mechanism for the switching behavior, we responsible now articulate the parts of the circuit. The central part is a junction transistor (hereafter, simply transistor). The transistor has three terminals: the base (left), the emitter (bottom) and the collector (top). The relevant of the transistor are given by the saturation voltage properties (Vce)sat and a parameter ? that determines the cutoff and saturation regions for the transistor. In addition there are two resistors, the bias resistor with must resistance and the load resistor with resistance ?2- There is also a a rail with and positive voltage (top) voltage Va ground (bottom). These might be terminals of a battery. Finally there are input and output as indicated in the circuit diagram. terminals. These parts are connected R\ source STUART S. GLENNAN 60 The circuit's behavior can be summarized by Define Voff to be 0 and Von to be (R\Vsource/?Ri)- two simple equations. Then the equations are: (1) Vin< V0?? =? (Vout= ^SOURCE) Switch off (2) Vin > V0N =* (Vout? 0) switch on a kind of current valve. When is functionally V?n < Voff is the base is the valve closed and no (i.e., voltage entering negative), > current passes from the emitter to the collector. When V{n Von (i.e., The circuit the base is above a small positive the valve is value), entering current to and from the emitter the collector. When passes open freely current < is < the valve restricted Vin Voff Von? part way open, allowing flow between the emitter and the collector. Properties of the load resistor voltage the voltage for the output terminal in the on (R2) and battery determine the valve is open, these properties also determine and off states. When the current the resistance of the bias resistor output Increasing Ic(= Va/R2). the valve opens. the voltage Von at which illustrates a number of important features of mechanisms in accordance with (M). Most and how they can be analyzed importantly, to mechanical this circuit is susceptible analysis even though it is not, in as in the water device. Additionally, the engineer's sense, a mechanical R\ increases This circuit are susceptible to themselves regulator valve, the parts of the mechanism to It is mechanical mechanical (in my give explanations analysis. possible of the resistors, transistor, battery and conductors sense) of the properties this circuit indicate why they have the properties they do. Moreover, itself be a part of larger mechanisms. Indeed, the chief interest of switching circuits is that they can be used as parts of larger electronic logic or control devices. Finally, this circuit indicates how what the mechanism which can I have given of the is depends upon how you look at it. The description Itwould also electronics. circuit has been largely in the terms of elementary to give a description of the system. in terms of the microphysics be possible the of the system in which lead to a decomposition This description would parts were be possible lattices, or other such entities. Itwould also electrons, molecular for other purposes. to consider this same circuit as amechanism that produces heat in the so that the input current always falls into bias resistor. Also, by arranging the transient region, the same circuit can be used as a linear amplifier. The has emphasized the potential logical given, though electronic, description to cases where of the circuit. Other descriptions, appropriate properties We could for instance input voltages typically linear behavior.10 consider lie within the mechanism the transient region, will emphasize its MECHANISMS AND THE NATUREOF CAUSATION 61 to illustrate my analysis are both them to illustrate the point that physical no a on nature restriction the of interactions (M) places priori physical between is in no way limited to parts. On the other hand, my analysis mechanisms that are physical to apply to in nature. It is meant equally The mechanisms mechanisms. that I have I have chosen selected and other higher level mechanisms. I biological, psychological this because the of the is emphasize point generality analysis key if it is to for a theory of causation.11 provide a foundation chemical, 4. FUNDAMENTALAND MECHANICALLYEXPLICABLELAWS of amechanism such as the voltage switch can be described in terms of one or more laws (i.e., counterfactual supporting generalizations) like those given in equations of the (1) and (2) above. A description internal structure of the mechanism this I behavior. call such laws explains The behavior mechanically explicable. There is an important class of laws that are not mechanically expli? cable laws. The essential fundamental feature of fundamental laws is that they are taken to represent facts about which no further explanation is possible.12 While it is difficult to define the notion of fundamental law and arguably impossible to devise an adequate test to determine whether or not a law is fundamental, it is not hard to come up with a small body are nearly unanimously Exam? regarded as fundamental. include the law of universal and ples from classical physics gravitation Maxwell's In this status is to accorded equations. contemporary physics, of laws which to space-time curvature or equation relating mass distribution to Schr?dinger for equations quantum mechanical systems. There are also are not fundamen? many laws which, while scientifically quite significant, tal: for example, the ideal gas law, Hooke's law, and laws of classical Einstein's genetics. The claim for which Iwill argue is that all laws are either mechanically or tertium non datur. I will refer to this thesis fundamental, explicable as the thesis of the mechanical of non-fundamental laws. explicability In arguing for this thesis, we must be careful to construe it in such a false nor trivial. A strong reading of it is way that it is neither obviously that every instance of a particular lawlike regularity is explained by the some of of mechanism. This seem would operation particular type reading to entail that laws describing the behavior of higher level mechanisms would be type-reducible to laws of a lower level theory. But a number of widely accepted arguments (Davidson, 1970; Fodor, 1974; Kitcher, 1984; STUARTS. GLENNAN 62 Putnam, 1973; Wimsatt, seldom possible. A weaker reading 1976) suggest that type reductions of this sort are of the thesis of the mechanical of explicability is that every instance of a non-fundamental law is explained of some mechanism; but it need not be the by the behavior case that the mechanisms which explain the various instances are all the laws non-fundamental same, or even of the same kind. For a higher level law to be mechanically itmust be realized by some lower-level mechanism, but itmay explicable, This reading would entail that laws describing be multiply-realized. the are token-reducible to lower-level of higher level mechanisms are not that but laws, they type-reducible.13 that there are higher level The weak reading allows for the possibility laws, every instance of which must be explained by a different mechanism, behavior perhaps even by a mechanism the laws in question would because these mechanisms, of a radically different kind. In such cases, not genuinely to be explained by reference can said about of be how the type nothing Such strongly irreducible laws is produced by mechanisms. mechanical but would, resist like fundamental laws, explanation, on lower-level mechanisms. fundamental laws, supervene of non the weaker reading, the thesis of the mechanical explicability lawful behavior would, unlike On laws is very plausible. The problem with the weaker reading is that it seems to entail little more than the claim that higher-level processes on physical insofar as I am committed However, processes. supervene about the thesis says something to a particular analysis of mechanisms, fundamental be in which the way occurs, and in this regard it might supervenience in discussed false. This point can be illustrated by the aether example continuous 7. If there were a genuinely 2 and footnote Section aether, we would of supervene on properties say that electromagnetic properties the aether, but we would not be able and magnetism the laws of electricity to say (without revising (M)) were mechanically explicable. that In a claim about the relation of higher is not primarily addition, but rather a claim about level laws to fundamental physical mechanisms, If laws to lower-level mechanisms the relation of higher-level generally. the thesis second that explains Mendel's one considers, for instance, the mechanism the natural in assortment of law of law (the genes gametes), independent second law holds (when it Mendel's is cytological. level of explanation because genes are often same on the chromosome), away created by choosing "randomly" holds) pair. And while one could on different chromosomes (or far and given how meiosis works, a gamete is one chromosome from each chromosome located look at the mechanisms that explain meiosis, MECHANISMSAND THENATUREOF CAUSATION that explain the mechanisms Mendel's second physical. A key feature of mechanically lematic way to understand a generalization about my explicable the counterfactuals 63 law are cytological, not laws is that there is an unprob which they sustain. Consider car starting when I turn the key. I am justified in to turn the key, the car would start" because I know that asserting I also which connects key-turning with car-starting. there is a mechanism "If Iwere know the sorts of circumstances in which the counterfactual would to be false, namely breakdown conditions for the mechanism it. I know for instance that my key-turning would not explains out turn which lead to is too cold, or if there is no gas in the gas tank, car-starting I understand the role of the battery and the gas in the ignition because can be understood in this way mechanism. Counterfactual generalizations if the weather without world, We appealing or the like. to unanalyzed notions of cause, propensity, possible are now to partially address the worry raised in section in a position use II about whether the of laws to describe the interactions of parts of mechanisms involves tion are mechanically character supporting) a circular appeal to causal notions. If the laws in ques? then their lawlike (i.e., counterfactual explicable, can be explicated to the further mech? by reference anisms which explain these laws. There remains the difficulty of under? source the of for not offer lawlikeness I fundamental laws. will standing a solution to this last problem here, but I will argue in the final section that the absence of an account of the lawlikeness of fundamental laws a mechanical does not undermine of laws and causal analysis higher-level relations. 5. I claimed towards a mechanical theory of causation at the outset of this paper that my theory of mechanisms could the foundation for a theory of causation. Although to spell out provide such a theory and defend it in any detail is beyond the scope of this paper, I can indicate briefly how such a theory would some go, and suggest of the theory. consequences Before outlining such a theory, we should consider very generally the are of a causation should solve. There stan? number of problems any theory dard ones: distinguishing real from spurious correlations; distinguishing lawlike from accidental real effects from generalizations; distinguishing artifacts. These are all instances of what we can, using Humean terminol? ogy, call the connection-conjunction problem. How does one distinguish connections from conjunctions? Humean seek to solve the approaches STUART S. GLENNAN 64 problem by giving ties. Anti-Humean necessity nections. (logical I think criteria how to some typically appeal approaches or natural) which distinguishes that both of these approaches not discuss but I will difficulties, to indicate true from accidental that distinguish a mechanical regulari? further notion of from con? conjunctions run into insurmountable I want here.14 Rather, the connection-conjunction these difficulties theory tackles problem. put, amechanical suggests that two events Roughly theory of causation con? are causally there is a mechanism connected when and only when a a case most in is the works clear where them. How such necting theory switch Take for instance the voltage by a conditional. 3. The behavior of the voltage switch can be three conditionals: by can be described behavior discussed summarized (1) (2) (4) in Section Vin < Vow => Vout = VsOURCE Vin > Von =*Vout = (Vcd)sat ~ 0 Voff < Vin < Von => (Vce)sat < Vout < ^source this behavior, we can say, which underlies there is a mechanism than Voff to greater than from less the that e.g., input voltage increasing to Vsource a causes Von change in the output voltage from (Vce)sat the connection-conjunction The chief virtue of the theory is that itmakes can one. one a formulate and confirm a theory that If scientific problem a two then one has produced mechanism events, connecting postulates Because evidence are causally from accidental that dis? The necessity is to be understood conjunctions that these events connected. connections tinguishes as deriving from a underlying mechanism, and nature where the existence The mechani? is open to empirical of such a mechanism investigation. cal account allows connection-conjunction to any metaphysical There is however us to escape with the the regularity theory's difficulties as Hume did, reference problem, while eschewing, notion of necessity. an obvious to the mechanical theory. Sooner of a system into parts must come to an laws. At this point we cannot point to how do we Since there is no mechanism, limitation or later the process of decomposition end. This is the level of fundamental any further or deeper mechanism. between explain the causal connection events at the level of fundamental and I will not try to theory offers physics? The mechanical to it is sufficient add anything about the problem here.15 For the moment, is in relations fundamental causal that whatever physics explains recognize at level. causal relations which that higher very different from explains us no answers, There presented. are two potentially The first of these serious alleges to the theory that I have objections of causation that that any analysis MECHANISMSAND THENATUREOF CAUSATION 65 on mechanisms of the concept since any explication is circular, of mechanism requires the use of causal concepts. Of course one may a system into parts and describing a mechanism identify by articulating the behavior of the various parts; one may formulate statements describing relies these parts and show how the behavior of the from the derives between interactions these (the effect) system mean to what it that Is does these interact? it not say parts. However, parts some to the mechanical that in essential the of theory changes properties the interactions between as a whole parts cause changes This circularity in the properties of other parts? is only apparent. In describing the mechanism that connects the two events I have explained how these events are causally connected. How the parts are connected is a different I can question. this second question another account of the try to answer by offering mechanisms which connect them, but I need not give such an account to between the events. Indeed, such an account would explain the connection the causally relevant features of the original explanation.16 only obscure to the apparent circularity The supposed circularity is analogous involved in the recursive definitions of sentences of predicate logic. A typical clause of such a definition 'if p and q are sentences, then (p & q) is a string of symbols also a sentence'. Whether is a sentence depends upon are sentences, whether certain other strings of symbols but we are not a circular because sentences the in used the definition definition, offering can themselves to be defined without reference the sentence in whose would be in giving account of how two they are being used. Similarly, are causally connected, I refer to a mechanism in turn refers which to causal relations, but these latter causal relations are different (and more one am than I relations to the which basic) explain. seeking definition events cannot so easily be defeated. Even granting that second objection can progressively at one level in terms of mecha? explain interactions nisms at the next, sooner or later, we are going to run out of levels and come to interactions laws. I grant without governed by fundamental argument The we that these fundamental theory. But interactions interactions cannot be explained by the mechanical theory gives us no account of such the fact that the mechanical combined with the fact that any mechanism depends ultimate? on there at causal connections the level of fundamental ly being physics that the mechanical us no has account might lead us to believe theory given of causation at all. The objection It is confessed, that the utmost of natural phenomena, effort to a greater ... But causes. into a few general their discovery-The attempt our ignorance a little longer.... most ismade of human eloquently reason is to reduce and to resolve simplicity, as to the causes of these perfect (Hume, 1777, philosophy pp. 30-31) by Hume: the principles, productive the many effects particular we in vain should causes, of a natural kind only staves off STUARTS. GLENNAN 66 Certainly is correct Hume that there must be some facts which we cannot to further more explain by reference must Our explanations this ultimate dependence theory and adopt a regularity theory of causation. why we are not forced to adopt a regularity theory, we up the mechanical To understand must (or mechanisms). general principles at somewhere. The issue is whether stop question on unexplained demands that we give regularities look at how it is that we in fact evaluate the truth of causal claims. to do this, I submit that the best way to there are many ways Although to evaluate such claims is find the mechanism for the supposed responsible causal connection. If for instance, we want to show that smoking causes the mechanism cancer, the best way to do so would be to discover by which tar, nicotine, etc., interact with might provide overwhelming between and cancer, smoking mechanism in question, we the body to produce cancerous cells. We to show the correlation statistical evidence but so long as we do not understand the can still wonder whether or not the correlation that smoking causes cancer.17 Not only are regularities insuffi? are as cient to establish causal connections, well. Once unnecessary they we need not acquire additional we have identified the mechanism, evi? indicates dence for the regularities it produces. Also, or operations of the parts of the mechanism further detail about the nature are not relevant. The best way to find out if it is a dead battery that is preventing my car from starting is to use a voltmeter to test the battery's charge. Once I have established that confirmed my hypothesis, and the battery has no charge, I have sufficiently evidence. Although the taking apart the battery will provide no additional two events may for connecting supervene upon responsible on mechanically and ultimately lower-level mechanisms, inexplica? it is not these laws which make the causal ble laws of fundamental physics, mechanism other true; rather, it is the structure of its parts. properties claim of the higher level mechanism and the treats causa? of a point of view which Iwould from level causation, higher physics differently to sketch how the point of view of the mechanical theory can To illustrate some of the benefits tion in fundamental like briefly in the quantum theory. It is often said shed light on an interpretive problem as a predictive successful while that the quantum extraordinarily theory, to cannot that it predicts. This be said instrument, explain the phenomena predicament correlations illustrated comes produced as follows. in the case of the "unexplainable" clearly EPR The problem may be type experiments.18 by to construct a device which shoots a It is possible out most to distant targets. These particles in opposite directions pair of particles can be prepared in such a way that, upon hitting these targets they will deflect in one of two directions, up or down. It is not possible using this MECHANISMS AND THE NATUREOF CAUSATION 67 to determine in advance which direction technique preparation they will mechanics and that if confirms go. However, quantum predicts experiment one particle deflects up then the other particle will deflect down and vice versa. seem There to be two explanations which might account for this corre? there might be some signal sent from one target to the other. lation. First, This possibility is ruled out (if relativity theory isright) by placing the tar? gets so far apart that any interactions than the speed of light. Alternatively, require signals to travel faster one might think that the preparation into a certain state which causes them would of the particle pair puts the particles or the other when to go in one direction they reach the target. Surpris? results (so called no-hidden-variable-theory ingly, however, mathematical results) indicate that there can be no such state. This result is generally stand. It indicates correlations an indirect considered with (consistently events where between to be very strange and hard to under? that there are evidence) experimental neither a direct causal connection nor via a common-cause can possibly account for that I think that the analysis of causation correlation. that I have offered shows is so puzzling about the cor? why we should not be so surprised. What relation in question is that there is no mechanism which could possibly connect the events occurring at the two targets and that there is no mech? connection anism which could possibly connect them each to some cause. However, if one believes that quantum mechanical the most fundamental then one does not believe physics, third common laws describe is a no mechanism if And there is rea? such what anyway. mechanism, deeper son is there to believe that distal events should or should not be correlated? Our uneasiness derives in part because we expect that the laws describing this quantum mechanical similar to those system should have properties of mechanically laws; but there is no reason to have such an explicable It is an artifact of our belief expectation. that there that there is something behind the regularities. The mechanical theory of causation rejects a wide-spread assumption about the nature of causation. I think that it is generally assumed that what? ever causal connections to do with the are, they ultimately have something most fundamental the more physics, further we from causal The closer we are to fundamental processes. physical our statements are about the true causes of things; the into the higher-level the more we move sciences, away stray statements however, assumption, causes this assumption damental physical and toward mere is what makes Hume's This generalizations. so skepticism devastating. On empirical are the ultimate metaphysical events together. Hume provides glue which a convincing holds fun? argument 68 STUART S. GLENNAN that we can have no knowledge of this glue, and that talk of such glue may to these difficulties be unintelligible. The solution is to reverse the initial assumption. Causal statements are typically statements about events even and mechanisms regulated by mechanisms, ties. Only when we talk about interactions does causal To what higher level enti? fundamental laws by governed talk become problematic. have we have solved Hume's extent problem? To what degree that binds together causally con? the level of fundamental still problem physics, Hume's have we uncovered nected are complex, the secret connexion At events? can observe but we cannot offer an expla? regularities, obtain. It is not good enough to say that in nation of why those regularities are there for there are still questions about which just regularities, physics are lawful and causal. the difficulties that remain, we regularities Despite remains. We certain one. In the case of is not a universal that Hume's problem we can between connections and conjunc? laws, distinguish higher-level can we understand the mechanisms which because tions, produce higher is not so secret after all. level regularities. Very often, the connexion have shown NOTES * I would like to thank Erich Reck, Mike Price, Bill Wimsatt, Ron McClamrock, Dan Howard Garber, Hume. him are unknowable? interpretation, Strawson (1989). purposes 2 I should the hood under 3 For there Whatever which experts a discussion regularities, Hume's interpretation taken as leading of Hume, of of interpretation lead skepticism that these powers or only to argue century empiricists For a defense is not decisive. the correct the for the latter it is sufficient former view, see for my or which is open and Corpuscularism to empirical investigation. in the seventeenth century see 2). or teleological it makes constraints, very few restrictions etiological to describe as a mechanism. For instance, it is quite possible my belt a mechanism such as this is not worthy bullets. Generally stopping as a mechanism for of but we investigation, death untimely understand of Mechanism ignores could count because not seem so silly. 5 here I do not mean can imagine contexts where a bullet off of my ricocheted it would be. If I was buckle then this mechanism saved from an would I do, however, of object. to give a definitive of the notion analysis to be it seems whatever term is plausible, because are, objects of contexts. that they can exist in a variety that my important the proper Does to the regularity is often theory of causation. argument to I not of what is under need have do that any knowledge personally emphasize a causal claim. that there is something that I believe It is sufficient in order to make Glennan(1992,ch. 4 4 Since (M) think evidence textual about question: tend that Hume's the hood on what twentieth on earlier and for comments discussions claims the following ignoring are no powers behind Although the to make introduction I am In particular, to believe that for fruitful and Ken Waters Stein drafts of this paper. 1 in this I do not wish use of the MECHANISMS AND THE NATUREOF CAUSATION 69 I would like entities which is that there are certain this stipulation with One problem sense properties. are in a natural beliefs and For instance, as objects which to consider as properties, to allow are often described but I want states generally) desires (or mental are parts. There that prevents is nothing, entities however, as objects under under some descriptions and as properties are properties under one description of our brains, while beliefs and desires others. Under and desires are themselves is right), beliefs another description (if some brand of cognitivism have various which computational properties. objects 6 of the requirement In the case of a physical like the electromagnetic field, "system" I have used the that the parts be discrete. is roughly the requirement isolability empirical such in which for mechanisms us from such viewing entities term because in cases where parts are not general of discreteness amounted. be clear to what a requirement more localizable spatially it would not 7 The possibility that the electromagnetic field could be explained in terms of properties of aether an explanation aether would about (M). A detectable provide question to say that aether it then be correct the electromagnetic field, but would waves? If (M) is correct, then it would that transmits electromagnetic a further raises of of properties is the mechanism if it were possible to decompose to call the aether a mechanism it into only be appropriate a turn out to that aether would be continuous But it is the discrete parts. genuinely possible so no such decomposition would that and if this were be available. medium, Supposing or not a detectable there would be a further empirical whether there were aether, question the aether was of discrete composed particles. of a genuinely continuous aether raises a question the possibility in such a way that all mechanisms (M). (M) defines mechanisms In short, erality of of discrete (in the sense "continuous have of decomposible) If we wished mechanism". to amend the definition parts, to count to allow (M) but a continuous the aether about the gen? are collections for "continuous" be a sort of aether would as a mechanism, mechanisms we would then a single having emendation. our best physical theories do not demand continuous such part. Currently, 8 one must an adequate to work, to distinguish If this account is ultimately have way to solving between these two kinds of laws. One of the most this promising approaches use as is of the statistical known off" (see e.g., Salmon, problem by relationship "screening For 1984). a discussion of is taken directly McCausland, 1978). See ?9.4 and 10 It is the higher level description chosen the context by considering what aspects of the lower-level point seeMcClamrock 11 Examples contains from a case of the limitations this approach together with alternatives, see 5). Glennan(1992,ch. 9 This example an undergraduate text electronics (Calvert and the circuit (e.g., whether in which the circuit occurs, behavior are significant or a switch), is an amplifier us to determine that allows as opposed to noise. For a similar (1995) Chapter 1. sciences study from ?14.2. of other than physics two models of vowel are discussed normalization in Glennan mechanisms 6 Chapter that have been (1992). developed by cognitive psychologists. 12 I am presuming that these laws are laws of physics. 13 It would to spell out the relationship take a fuller exposition between different tations of mechanical and various As framed types of reducibility. explicability interpre? by Fodor terms and are theses about the reducibility of theoretical type and token reducibility to theoretical terms and laws of another. While of one theory there is an analogous reduction relations between laws and the mechanisms which realize them, some work must to show how be done of laws and theoretical terms is related to type type-reducibility (1974), laws reducibility of mechanisms. STUARTS. GLENNAN 70 some 1 suggest an excellent reasons in Chapter 8 of Glennan See (1992). also Salmon for (1984) cannot theories solve the connection-conjunction regularity an account In 6 book and Salmon that of causation that 5, 9), proposes problem. (Chapters to address is meant this problem and that in certain ways parallels my own account. Space so I can only mention one problem does not permit me to discuss Salmon's theory in detail, discussion of why that his theory is incomplete. Salmon's with distinguish? suggests theory is concerned are the sorts what causal from he calls processes ing pseudo-processes". Pseudo-processes of things which non-causal but lawlike Salmon's criterion for distin? regularities. produce which can transmit "marks". is that causal processes these processes has Salmon not explained transmit marks where why causal processes pseudo-processes true difference can only be between causal processes and pseudoprocesses between guishing unfortunately do not. The in the mechanisms the differences them. view, by considering underlying to explicate in a way causal relations that dovetails with fundamental the account in Glennan (1995). in my explained, 151 discuss how mechanical 16 Kitcher (1984) 17 A similar point (1982). causal 18 For from theory of causation of correlations and artifactual a discussion of EPR thesis R3. a causal relationship between for attributing accounts of causation this that regularity sufficient conclude They properties). The mechanical of his of evolutionary by Sober and Lewontin biology events of actual or dispositional between regularities that the existence is not in his discussion observation in the case has been made argue They (or properties) a similar makes I have presented this sort. correlations see, e.g., (or are inadequate. to distinguish between one allows events those (1989). Shimony REFERENCES and M. J. M. Calvert, A. H. McCausland: R. F. Mandl, Series, Physics Chichester. 'Functional Cummins, Robert: 1975, Cummins, Robert: 1980, The Nature Davidson, Donald: 1970, 'Mental Abaris Fodor, Books, New Glennan, of Chicago. University Stuart S.: 1995, Glennan, Butler David: Sons, 72,741-765. of Philosophy MIT Explanation, and J.W. Swanson Press, Cambridge. (eds.), Experience Amherst. la Lumi?re, or The Disunity and Models 'A Two-Tiered 1777, edited and translated of Sciences Causation, of Causation', Theory by Michael as aWorking Ph.D. Hypo? the Dissertation, manuscript, unpublished Press, University Stuart: Kauffman, for Them', of Science, PSA reprinted 1975. Oxford, 1970, from the Human 1777 Understanding edition, 1970, Vol. Vm, and D. Reidel, Dordrecht. concerning 3rd ed., P. H. Nidditch, in Biology of Parts Explanation Studies and R. S. Cohen (eds.), Boston 'Articulation in R. C. Buck and Behavioral of Cognitivism', and Plausibility concerning Enquiries of Morals, Principles Search Press, de & University 'The Nature John: 1978, Haugeland, 215-226. Sciences Brain 1, Hume, of The Manchester John Wiley 1979. York, Sciences, 1974, 'Special 28, 97-115. Synthese S.: 1992, Mechanisms, Stuart (eds.), Journal Analysis', Jerry A: thesis', X Vol. J. Sandiford of Psychological in L. Foster Events', of Massachusetts and Theory, University ou Trait? Le Ren?: Monde, 1664, Descartes, S. Mahoney, Electronics 1978, and D. J. Ellison and the Oxford the Rational in the Philosophy MECHANISMSAND THE NATUREOF CAUSATION 71 Kitcher, '1953 1984, Philip: 335-74. 93, R.: and All A Tale That. 1995, Existential Cognition: Press, Chicago. sity of Chicago Ernest: 1961, The Structure Nagel, of Science, McClamrock, Putnam, Hilary: Harcourt, 1984, Scientific Explanation Wesley: Princeton Press, Princeton. University Abner: (ed.), The New Sober, Elliott, Philosophy Strawson, 1989, Brace Foundations 'Conceptual in the World, Jovanovich, of Psychology', and the Causal of Quantum William C: Univer? Inc., New Cognition Structure of York. 2,131-146. the World, in Paul Davies Mechanics', Press, Cambridge. Physics, Cambridge University and Richard C. Lewontin: Cause 1982, 'Artifact, of Science Galen: 1989, Review Philosophical Minds Computational Salmon, Shimony, Sciences', and the Nature 'Reduetionism 1973, of Two and Genie Selection', 49,157-80. The Secret Clarendon Connexion, Press, Oxford. the Logical of Function Structure Statements', 'Teleology inHistory Studies and Philosophy 3, 1-80. of Science William C: 1976, Levels of Organization, and the Mind-Body Wimsatt, 'Reductionism, Wimsatt, Problem', Brain: in G. G. Globus, A Scientific and 1972, G. Maxwell and Philosophical and Inquiry, Manuscript submitted October 10, 1994 Final version received July 13, 1995 Dept. of Philosophy and Religious Studies Butler 4600 University Sunset Ave. Indianapolis, IN 46220 U.S.A. I. Savodnik Plenum, (eds.), New Consciousness York. and the