EXCLUDING THOROUGHLY: THEORIES OF CAUSATION AND THEIR IMPACT ON THE EXCLUSION PROBLEM By ANDREAS FALKE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017
2017 Andreas Falke
To my mother and in memory of my father
4 ACKNOWLEDGMENTS Sometimes a dissertation is a manifestation of academic flourishing. People may contribute to such flourishing in different ways. They may di content, as my teachers, students, and peers have done; or they may be the backdrop against which the project has been possible in the first place by fostering intellectual curiosity and by providing emotional support, as some family members and friends have done. While it is impossible to mention every person who has affected this dissertation in some way I wish to thank at least those to whom I am most indebted. In the fall semester of 1999, I took my first Philosophy o f Mind course with Matthias Vogel at the Goethe University Frankfurt, Germany. It is hard to overestimate the impact that Philosophy of M ind I was exposed to most of the arguments, positions, and problems that I am subject and his striving of clarity infected me and thus set the course of my further academic development. I wish to thank him for that. It was also through him that I got invited to Wolfgang Detel colloquium. Populated by inspiring philosophers who possessed much greater knowledge and skill than I did at the time, it forced me to adapt intellectually, much to my own benefit. Though I am indebted to all the participants of the colloquium, I owe thanks particularly to Wolfgang Detel and Alexander Becker. Later, in 2006, they would constitute the committee for my Magister thesis in which I developed a criticism exclusion problem. In the meantime, I became a teaching and research assistant for Wolfgang. Having him as a teacher as well as a philosophical role model provided me with a solid foundation upon which I am stil l building. Without a doubt, I am most deeply indebted to him
5 among all the excellent teachers I had in Frankfurt. I wish to thank him for the many things I owe him for. There is not much of a point to building a foundation unless one plans to build someth ing on it. Since I was interested in an area and in a kind of philosophy that was and, I believe, still is best done in the United States, it seemed like a good idea to apply to graduate school there. I got an invitation from the philosophy department of t he University of Florida and readily accepted. I wish to thank the department as a whole for giving me the chance to study with philosophers who grew up in the analytic tradition and who have been an integral part of it. Particularly the courses I took wit h John Bi ro, David Copp, Michael Jubien, Chuang Liu, Kirk Ludwig, and Gene Witmer broadened my philosophical horizon and closed a good number of gaps. More importantly, they deepened my understanding of the arguments, positions, and problems and allowed me to see connections between concepts that I previously had kept separate in my mind. It is largely due to what I have learned from them that this dissertation focuses not on a problem within a particular area of philosophy but on a problem located at the i ntersection of the philosophy of mind, metaphysics, and the philosophy of science. Since I think this makes the dissertation more interesting and more valuable, I wish to thank them. Chuang, Gene, and John were also members of my dissertation committee. Fo r that, and for feedback along the way, I owe them as well as the external committee member Jonathan Edelmann thanks. Without a doubt I am most deeply indebted to Gene for being the superb advisor that he was and consequently for being the person I interacted most with philosophically. N umerous thoughts and insights on the following pages are the result of the many fruitful discussions I had with him H e made me see that many of the arguments and positions I d eemed clear during my Frankfurt years really are not clear after all. Thus he forced me to clarify where clarification was much
6 needed, thereby drastically improving the precision of my thinking and the clarity of my writing. I wish to thank him for all o f this and more. Regarding the backdrop, Andy Lcking has been most influential for my intellectual development. Had it not been for our nightlong discussions of Nietzsche and for our innocent marveling at Frege and Wittgenstein as teenagers, I likely wo uld not have become an academic, let alone a philosopher. It is telling that the owner of our local bookstore still remembers us as years later I found another fri end, Markus Rohling, with whom I could continue that tradition of excessively long conversations. I hope he knows how much I miss those tranquil evenings at the observatory. Without Andy and Markus I would not be who I am today. They deserve my gratitude more than anyone else. In Frankfurt, I became friends with Hana Grndler, Oliver Schtze, and Jan Erik Strasser. Fiercely smart and highly analytical, they have pushed my intellectual boundaries ever since. Letting our minds freely roam conceptual pastures has been among the most pleasurable and deeply satisfying activities in my life. I cannot imagine a life without it, thanks to them. I am confident that Hana, too, considers the loss of our beloved of ideas. In Gainesville, Ron Claypool and Aubrey Spivey have provided emotional support and social stability beyond what they may be aware of. I will miss our philosophizin g and cracking jokes during lunch breaks and I will miss our long bike rides on t he Hawthorne Trail. I hope they know how grateful I am for those times together. I also wish to thank Ray D. Fennell for having made me more aware of the beauty of simplicity; Michael Cullinan for not losing patience when I kept nagging him with questions about English grammar and punctuation ; Larry McKill for offering some culture when I was
7 starving for it; Roger Meyer for listening when I needed to blow off steam; Allison H. Roulston for going beyond words with me; Charles Stanley a.k.a. Carlo Pittore fo r showing me how to live a life of complete dedication; and William Neville for his encouragement and support when I do. These friends all have been a seemingly inexhaustible source of intellectual stimulation. That two of them are not able to witness the completion of my dissertation saddens me beyond words. Finall y, thanks are due to my brother, Christoph Falke and my parents, Herbert and Rita Falke for supporting me in numerous ways ever since I decided to study philosophy. I am sorry that my father too is among those who are not able to see the result of their support.
8 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 ABSTRACT ................................ ................................ ................................ ................................ ... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 11 1.1 Motivating the Project ................................ ................................ ................................ ...... 11 1.2 Structuring the Project ................................ ................................ ................................ ...... 19 2 THE EXCLUSION PROBLEM ................................ ................................ ............................. 23 2.1 Two Challenges ................................ ................................ ................................ ................ 23 2.1.1 The Lack of a Te rminological Convention ................................ ............................ 25 2.1.2 The Lack of a Canonical Version ................................ ................................ ........... 32 2.2 Formulating the Exclusion Problem ................................ ................................ ................. 37 2.3 Generalizing the Exclusion Problem ................................ ................................ ................ 50 2.4 Chapter Conclusion ................................ ................................ ................................ .......... 55 3 CAUSATION AND THE EXCLUSION PROBLEM ................................ ........................... 57 3.1 A Qualification ................................ ................................ ................................ ................. 57 3.2 Causal Explanations ................................ ................................ ................................ .......... 58 3.3 Theories of Causation ................................ ................................ ................................ ....... 62 3.3.1 Regularity Theories ................................ ................................ ................................ 64 3.3.2 Counterfactual Theories ................................ ................................ ......................... 72 3.3.3 INUS Conditions ................................ ................................ ................................ .... 79 3.3.4 Probabilistic Theories ................................ ................................ ............................. 83 3.3.5 Transference Theories ................................ ................................ ............................ 87 3.3.6 Single Difference Theories ................................ ................................ .................... 90 3.3.7 Causal Pluralism ................................ ................................ ................................ ..... 93 3.3.8 Causal Antirealism ................................ ................................ ................................ 99 3.4 Chapter Conclusion ................................ ................................ ................................ ........ 102 4 STANDARD RESPONSES TO THE EXCLUSION PROBLEM ................................ ....... 106 4.1 Organizing the Responses ................................ ................................ ............................... 106 4.2 Distinctness ................................ ................................ ................................ ..................... 109 4.3 Causal Impact ................................ ................................ ................................ ................. 116 4.4 Typicality ................................ ................................ ................................ ........................ 124 4.5 Completeness ................................ ................................ ................................ .................. 126 4.6 Overdetermination ................................ ................................ ................................ .......... 1 28 4.7 Exclusion ................................ ................................ ................................ ........................ 130
9 4.8 Chapter Conclusion ................................ ................................ ................................ ........ 134 5 REASSESSING THE EXCLUSION PROBLEM ................................ ................................ 139 5.1 Striking a Balance ................................ ................................ ................................ ........... 139 5.1.1 Strong Exclusion ................................ ................................ ................................ .. 140 5.1.2 Weak Exclusion ................................ ................................ ................................ .... 143 5.2 Co nclusion ................................ ................................ ................................ ...................... 145 APPENDIX: KANT EXCERPT ................................ ................................ ................................ .. 149 LIST OF REFERENCES ................................ ................................ ................................ ............. 150 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 158
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosop hy EXCLUDING THOROUGHLY: THEORIES OF CAUSATION AND THEIR IMPACT ON THE EXCLUSION PROBLEM By Andreas Falke December 2017 Chair: Gene Witmer Major: Philosophy The exclusion problem casts doubt on the idea that mental events can have a causal impact on other events in virtue of their mental properties While many responses to the problem focus on the relation ship between mental and physical properties some analyze o ther concepts involved, such as the completeness of the physical and overdetermination. However, surprisingly little effort has been made to investigate the exact role causation plays in the problem. An underlying premise of the debate thus seems to be tha t the problem remains the same regardless of what theory of causation one presupposes. To show that this premise is false I offer a version of the problem that is more carefully develop ed than the often sketchy versions discussed in the literature I then analyze dynamic an d its nature change depending on what theory of causat ion one presupposes. The results of this analysis can help to explain why and how some of the standard responses are not convincing at least as responses to my version of the exclusion problem In some respects that version is stronger and in other respects it is weaker than versions more commonly discussed. As a result, standard responses may not convince as response s to my version but succeed as responses t o other versions. Because of that, I review some of these alternative versions with respect to their virtues and vices and conclude the dissertation by defending my version as the one most suitable for future debate and research.
11 CHAPTER 1 INTRODUCTION 1. 1 Motivating the Project Metaphysics investigates what kinds of things there are in the world, what their nature is, and how they are related to one another. Philosophy of mind is commonly, though not necessarily, viewed as a branch of metaphysics. When viewed as such, it is that branch which and to the world in general. Since beliefs regarding the nature of the mind affect our theorizing about its relation to the world, and vice ver sa, careful inquiries in the philosophy of mind require keeping an eye on the broader metaphysical context. To some degree, such context is involved whenever the mental is contrasted with the physical. If, for example, mental events such as feeling pain a re essentially private, then they indiscernibility of identicals, according to which two things are identical only if they have all thei r properties in common, it se ems to follow that at least some mental events cannot b e identical with physical event s If, on the other hand, we endorse some kind of identity theory, we either have to deny that mental events are private or we have to explain how privacy can be accounte d for in physical terms. But from time to time we need to broaden the context even further exactly because the entanglement of the philosophy of mind and metaphysics goes deeper. For instance, when philosophers analyze the relationship between mental and physical properties, they often have recourse to other elements of their ontology such as events and laws of nature. Donald possible to reduce mental properties to physical properties exactly because there cannot be any
12 strict laws that would allow us to predict or to explain mental events, despite the fact that every mental event token is identical with a physical event token. 1 But while philosophers of mind reflec t on how metaphysical assumptions about events, properties, and laws of nature may affect their theories of mind, they remain remarkably silent about the concept of causation. This is remarkable for two reasons. First, in formulating and discussing theori es of mind many philosophers rely heavily on causal talk. If this talk remains unexplained, the resulting theories and discussions may remain as unclear as the causal talk they rely on. 2 Second, in contemporary metaphysics it is taken for granted that theo rizing about the nature of events and properties potentially has an impact on theories of causation, and vice versa. 3 If the philosophy of mind in fact is a branch of metaphysics, the silent assumption that theories of causation do not affect theories of t he mind then becomes rather puzzling unless some explicit justification for this assumption is provided. The relative silence regarding causation may be explained historically. With the exception of substance dualism and arguably idealism, all of the imp ortant theories of the mind property dualism, non reductive and reductive physicalism, functionalism, type and token identity theories, etc. are theories about how mental properties relate to physical properties. 4 1 See Davidson (1970/1980, 208) or Davidson ( 1993 3 ). As Evnine (1991 51 ) points out, Davidson later seems to admit the possible existence of psychophysical laws in the context of generalizations needed to spell out causal dispositi ons; see also Davidson (1987 45 ). It is an interesting question whether, or to what extent, this affects or even undermines his anomalous monism. 2 Stathis Psillos shares my worry. He argues that theories that rely heavily on the concept of causation, suc h as causal theories of knowledge, perception, or identity through time, appeal to little more than shaky prephilosophical intuitions, unless we do some serious groundwork to clarify what exactly we refer to when we speak of causation and e Psillos (2002, 2 3). 3 Shoemaker (1980 /1997 ), for instance, uses the notion of causation to define properties. According to him, properties are clusters of causal powers. 4 The fact that most, if not all, introductions to the philosophy of mind focus on this relation is indicative of this. In this respect, Kim (2006) is quite typical.
13 Most of the philosophical energy of the twe ntieth century philosophy of mind went into developing sophisticated versions of such theories. Some of them have obvious implications for entities such as events. If, for example, the identity of an event is determined by which properties it essentially b ears, a change in the nature of some of these properties might have implications for the identity conditions of events that bear them. But causation seems different. If for example, causation is a relation between entities that bear properties a change i n the nature of the causal relation may be less likely to affect the identity of the relata And this, in turn, may explain why philosophers did not primarily worry about causation. I t seems natural to reevaluate old problems like the problem of causal int eraction and to approach newer problems like the exclusion problem through the eyes of these theories of the mind first rather than throwing causation into the mix right away thereby potentially muddying the waters. The problem of causal interaction arise mental substances do not have any properties in common with physical substances. 5 In particular, it seems mysterious how something immaterial can causally interact with something material. 6 But even if we grant that mental substances have some physical properties, a mystery remains. For now we may be able to explain how mental substances can causally interact with physical substances, but it seems dubious that they do so in virtue of their mental properties Stripping this problem of its substance ontology, we almost automatically arrive at the problem 5 One exception is their both being temporal, i.e. (most of) their properties are instantiated at particular times. 6 Sometimes, the problem is not couched in terms of imma terial v ersus material but in terms of abstract v ersus concrete. Descartes famously thought that mental su bstances are unextended and non spatial whereas physical substances are spatially extended; see Descartes (1641 / 1996 II. 8 ). We might wonder, of cours e, how plausible this conception of the mind is. If and when I am having a thought while I am in my office on the UF campus, I might not be able to point to the exact location of my thought; but surely the thought takes place here rather than in India or o n Mars or nowhere at all.
14 of mental causation as it is discussed nowadays: how can mental events have a causal impact on physical events in virtue of their mental properties? The excl usion problem highlights how severe the problem of mental causation is by placing the assumption that some mental events have a causal impact on physical events in virtue of their mental properties into the context of further metaphysical assumptions that all seem quite plausible: mental properties are distinct from physical properties, the physical realm is causally closed, events are not systematically overdetermined, and no effect has more than one sufficient cause unless it is overdetermined. 7 The probl em is that, prima facie, not all of the assumptions can be true at the same time, so at least one of them must be false. Though there is disagreement about which assumption, if any, needs to be given up, most responses to the problem argue from the viewpoint of a particular theory of mind. Partly because of that, a good portion of the discussion focuses on th e relation between mental and physical properties while it either ignores or marginalizes the importance of the concept of causation. This ignorance or marginalization is not necessarily problematic. It might only be reasonable to utilize theories of causation if we have some evidence that in fact they matter. In the absence of such evidence, the most reasonable thing to do may be to pro ceed as if it were true that problems like the problem of causal interaction and the exclusion problem remain the same no matter what theory of causation we presuppose. The fact that so far this assumption has not been supported or justified in any syste matic manner then is not objectionable. After all, the burden of proof rests on the person who wants to reject the assumption because rejecting the 7 In the literature, the labels the problem of mental causation and the exclusion problem are not used consistently. Some philosophers use the problem of mental causation to refer to the exclusion problem; see, f or instance, Harbecke (2008, chapter 1). To avoid confusion, we will discuss these philosophers as if they adopted the distinction between the problem of mental causation and the exclusion problem
15 assumption implies positing a connection between the concept of causation and the above problems. As far as the burden of proof is concerned, this is correct. And as long as the historically dominating approach of spelling out the relation between mental and physical properties remains fruitful and provides new suggestions for how to answer the question how, if at all, mental events can cause physical events in virtue of their mental properties, there is no immediate need to change the focus of the debate. However, the historically dominating approach seems to have reached an impasse. Despite an ever growing bod y of literature on the exclusion problem, no satisfactory solution seems to have been offered so far. 8 Among the recent publications, the most promising and productive ones seem to be those that shift their attention from utilizing particular theories of t he mind to analyzing the concepts involved in the exclusion problem, such as completeness and overdetermination. 9 While neither these alternative approaches nor the traditional, partly property based responses to the exclusion problem have yielded convinci ng solutions so far, the recent debate has had some positive effects. One such effect is the realization that the exclusion problem has previously been presented and discussed in too sketchy a manner for it to be clear what the exact inferences are that ar e required for the problem to arise. More carefully developed versions of the exclusion problem are now available. 10 Through such fleshed out versions, the mechanics of the problem can be better understood, formerly overlooked defects can be 8 We will offer some support for this claim in Chapter 4. Of course, w e will not be able to discuss all proposed solutions. Instead, we will look at some representative strategies and illustrate why they are not fully satisfactory. 9 Haug (2009), for example, distinguishes two kinds of completeness and argues that taking t he se into account allows non reductive physicalists to solve the exclusion problem. Funkhouser (2002), to give another example, distinguishes three kinds of overdetermination and argues that not all of them are problematic. So the severity of the exclusion problem depends on which kind of overdetermination we presuppose. 10 A good example for a more carefully developed version especially with respect to the distinctness claim, can be found in Stoljar (2008).
16 addressed, and new responses emerge. In other words, the felt need for alternative approaches has brought some fresh air into the debate. It is in the context of a search for alternative approaches that the lack of justification for the assumption that the exclusion prob lem remains the same no matter what theory of causation we presuppose becomes particularly puzzling. Just like completeness and overdetermination, causation is a concept that plays a role in several of the assumptions. But unlike those other concepts, caus ation is, in addition, one of the key concepts in philosophy of mind in general. Since the entire discipline is saturated with causal talk, we should expect causation to be among the first targets of the shift, especially given that rather sophisticated me taphysical theories of causation are readily available to be utilized. But, surprisingly, relatively few philosophers entertain the possibility that theories of causation might contribute to solving the exclusion problem. Fortunately we have overdramatize d the situation to some extent. While indeed most philosophers endorse, implicitly or explicitly, the assumption that theories of causation will not help solving the exclusion problem, few philosophers claim that they have no impact whatsoever. Indeed, the re are counterexamples to the latter claim. Peter Menzies, for instance, utilizes the concept of contrastive causation to undermine the standard version of the exclusion assumption that no effect has more than one sufficient cause unless it is overdetermin ed. By showing how contrastive causation implies a modified, more plausible version of the exclusion assumption, Menzies illustrates how thinking about c ausation can indeed have an impact on the exclusion problem. 11 Of course, such an impact does not by its elf establish any previously 11 See Menzies (2008 ).
17 particular modification does not offer such wiggle room, not all possible ways in which thinking about causation might affect the exclusion pr oblem have been explored yet. Hence, it seems unreasonable to rule out the possibility that theories of causation might help solving the exclusion problem. And yet, some philosophers endorse such skepticism. Some such skeptics argue in too sketchy a manner to justify their confidence. Michael Esfeld, for example, devotes only fourteen pages to show ing that the exclusion problem is independent of theories of causation. 12 In a book of over four hund red pages, Jens Harbecke spends barely seven pages on arguing for this independence claim. 13 Most of the space on those pages is used to explain theories of causation It does not take a professional philosopher to suspect that, given the number of theories that need to be covered, the resulting justification is by no means systematic or satisfactory even if the justification provides interesting grounds for further debate and even if the conclusion drawn happens to be true. But there are laudable exceptions among the skeptics. Instead of arguing against the possibilit y that theories of causation in general might help solving the exclusion problem, they have developed criticisms against a particular theory of causation, usually the one that is most commonly utilized as a means to tackle the problem: the counterfactual t heory of causation. a caused event b to mean that, if event a had not occurred, event b would not have occurred. 14 Arguably, b does not counterfactually depend on the occurrenc e of a simpliciter but in virtue of a one or several properties. The general strategy then is to show that there are cases in which the 12 See Esfeld (2007). 13 See Harbecke (2008 85 92 ). 14 The version proposed by Lewis (1973 a ) will be discussed in section 3.3.2
18 better candidate for the property that b property r ather than a particular physical property. Terence Horgan and Lynne Rudder Baker are among the advocates of this strategy. 15 Some critics of this strategy, like Jaegwon Kim, remain skeptical of the counterfactual theory of causation in general and point out that the mere fact of a counterfactual dependence does not establish a causal relevance of the mental property in question. In the absence of an additional argument that establishes such relevance, the overall strategy to utilize the counterfactual theory of causation for solving the exclusion problem is not convincing. 16 But even if the critics win in this case, it remains an open question whether other theories of causation might or might not do the trick. 17 Where do these considerations leave us? There is evidence that some theories of causation affect the exclusion problem. At the very least, they help to clarify some of the assumptions as well as some of the inferences required for the problem to arise. Though it has not yet been convi ncingly shown that theories of causation can be used to solve the problem, it seems reasonable to explore this possibility by systematically analyzing the impact that different theories of causation might have on it. If the analysis reveals that in fact th eories of causation cannot, despite their impact, contribute to a solution, an important premise underlying the debate has been justified. If the analysis reveals that theories of causation can establish new ways to tackle the problem, an important contrib ution to the debate has been made A classic win win situation, or so it seems. 15 At least according to Zhong (2011, 130). Compare Horgan (1989), Horgan (1997), and Baker (1993). 16 S ee Kim (1998 67 72 ). 17 Lei Zhong, another critic of the counterfactual strategy, acknowledges this point; see Zhong (2011, 131).
19 1.2 Structuring the Project T he main goal of this dissertation is to demonstrate how theories of causation may affect the exclusion problem. But to get the project off the gro und, we will face t wo initial challenges in Chapter 2 First, s everal assumptions of the exclusion problem include causal claims However, there is no consensus on the natu re of causal relata. Are they aspects, events, facts, features, situations, states o f affairs, or tropes? Depending on how we answer that question, some of the assumptions would have to be phrased differently. To ensure coherence, we will treat events as the appropriate causal relata throughout the dissertation Second, many versions of t he exclusion problem are rather sketchy and differ in the number of assumptions, their strength, and their specificity Since there is no canonical version, we will develop our own one that avoids some of the shortcomings of more common versions Attention to detail will reveal the need for an assumption that is largely ignored in the literature namely the typicality assumption which will become relevant in later chapters Once all assumptions are established, we will generalize the exclusion pro blem Traditionally, it is understood as a threat to the causal impact of mental events in particular However, the problem may arise for other kinds of nonphysical events as well namely for all nonphysical event s that are claimed to have a causal impact on other events in virtue of their nonphysical properties. Many special sciences assert such causal impact The exclusion problem can thus be used to question the veracity of such claims thereby expanding the scope of the problem beyond the philosophy of mind into the philosophy of science In Chapter 3 we will pursue the main goal of the dissertation specified in Chapter 1 and Chapter 2 and build the foundation for assessing standard responses to the problem in the following chapter. Since too many the ories of causation are available, we will focus on a representative set Often, these theories have been developed in such detail that an exhaustive discussion even of only a representative set of theories would be unrealistic. Because of that, we
20 will present the main idea of each theory at a level of detail that suffices to achieve our goal, which is to show that and how they affect the exclusion problem or our responses to it even if it means that our discussion of each theory will remain sketchy. We will neither dis cuss the nuances nor argue for or against the plausibility of each theory unless doing so highlights something important about the exclusion problem. The resul ting analysis reveals that theories of causation sometimes affect the strength of particular assumpti ons of the problem and other times they affect the likelihood of rejecting particular assumptions. For example, regularity theories imply a particularly s trong interpretat ion of the typicality assumption; counterfactual theories may help to undermine the nonoverdetermination assumption and, by extension, the threat of exclusion for nonphysical events and properties; and ca usal pluralism reveals that the exc lusion problem might rest on the fallacy of equivocation. By showing how the dynamic and the nature of the problem change depending on what theory of causation we presuppose, the results of the analysis will help to explain why and how some of the standard responses to the problem are not satisfying Chapter 3 thus sets the stage for Chapter 4. Even though there are over one thousand articles on the exclusion problem, responses are sometimes divided into only three groups: autonomy solutions, inheritance s ol utions, and identity solutions. However, there are also responses that cut across this taxonomy as well as theories that could be but have not been yet, applied to the problem. To ensure that we cover a good amount of territory in spite of being forced to be highly selective, we will discuss at least one response from the three types of standard solutions, one that cuts across the taxonomy, and one that has not been discussed in the context of the exclusion problem yet. Ou r discussion will yield two results. First, the version of the problem that we have developed in Chapter 2 makes it indeed more difficult for some of the responses to succeed. Often, our version changes which of
21 the assumptions are likely to be given up in the light of a particul ar response or view. As an example, one traditional response is to distinguish between causal relevance and causal efficacy and argue that the exclusion problem m ay succeed in excluding mental events from being ca us ally efficacious but not in excluding the m from being causally relevant. Since our version specifie s the nature of the causal impact of such events in a way that many other versions do not that response amounts to little more than rejecting the very assumption that it was meant to support in a l ess fleshed out version. Unfortunately, the plausibility of such rejection will depend on how well the difference between physical and nonphysical events lines up with the distinction between relevance and efficacy, and w e can show that it does not line up well As a result, the response may fail Second, some of the responses work very well in combination with p articular theories of causation but not at all in combination with others. For example, one response claims that what motivates the completeness as sumption also motivates a further claim that is related to completeness but usually not discussed in the context of the exclusion problem, namely the claim that physics provides a complete inventory of the fundamental kinds and categories that are suffici ent for a realization basis for all of the events that have a causal impact. We can show that this assumption in combination with a transference theory of causation may imply that the impact assumption is false while it does not threaten impact if we presu ppose a regularity theory of causation. The conclusion of Chapter 4 will thus be that theories of causation have an impact not only on the exclusion problem itself but also on the plausibility or the impact of at least some of the responses to it. While s ome of the standard responses may not be convincing in response to our version of the problem, they may be promising as responses to other versions. For our discussion in Chapter 4 to have weight, we need to establish that our version is indeed a good one to endorse.
22 We will thus spend Chapter 5 on explaining why our version is to be preferred over some stronger and weaker versions sometimes discussed in the literature. This discussion will also serve as a justification for why o ur version is particularly well suited as a basis for future research. It strikes a balance between specificity and strength on the one hand and the avoidance of unnecessary ontologica l commitments on the other O ther versions rarely achieve such balance It makes it easier for philosophers who differ in their views on causation, completeness, events, overdetermination, and the mind to assess the problem and to respond to each other. In the final section, we will summarize the lessons learned and highlight some open ends that need to be addressed in future research.
23 CHAPTER 2 THE EXCLUSION PROBLEM 2.1 Two C hallenges In a paper published in 1968, Norman Malcolm asks whether it is conceivable that neurophysiological stat es and processes are sufficient causes of human behavior. His answer is negative for two reasons. If such states and processes were causally sufficient, the truth of such a view could not be asserted consistently because the act of asserting requires menta l states or processes to play a causal role in the production of the assertion. And if the assertion could not be produced by mental states or processes, it could not be made on rational grounds ; for that requires causally responding to reasons. 1 For these arguments to work, however, he needs to further assume that there being sufficient neurophysiological causes precludes other factors from playing a causal role in the production of the assertion. Malcolm thus arrives at an early formulation of the exclusi on problem. 2 In spite of occasional responses from prominent philosophers such as Daniel C. Dennett, Alvin Goldman, and G. H. von Wright in the 70s, it took almost 20 years for the problem to gain real momentum. 3 Partly due to the work of Jaegwon 1 See Malcolm (1968), especially pp. 45 46 and p. 71. 2 and Heil (2013 section 6.2 ) correctly point out, it was by no means the earliest. MacIntyre (1957) had developed similar arguments. And even a passage in Immanuel Critique of P ure R eason could be interpreted as fusing questions concerning the compatibility of free will and d eterminism with problems regarding causal exclusion; see Kant (1781/1997 A536/B564 ), cited in the appendix. If we interpret talk about freedom as talk about mental events that have certain features that are causally relevant, that passage could be viewed as a much earlier predecessor of the exclusion problem. Against this threat of the causal exclusion of the mental, is sometimes seen as the cont agency and physical determinism; see for instance, Kim (2003 129 ). Lately, such Kantian concerns have resurfaced in the form of exclusion arguments against free will. See, for e xample, List and Menzies (2017) for a criticism of such arguments. 3 See Dennett (1973), Goldman (1969), and von Wright (1971 118 31 papers from the 1980s and 1990s. As a starting point, see Kim (1989 /1993) and Kim (1993 ).
24 Kim, which helped structure the somewhat scattered responses up to that point, the exclusion problem has become one of the central problems in the philosophy of mind. The newly gained popularity also yielded an abundance of versions of the exclusion problem. T hese versions differ in the number of assumptions, their strength, and their level of specifi ci ty. This diversity has persisted. As of now, no canonical version of the exclusion problem has been developed. Because of that, our first order of business shoul d be to look at some of these versions, identify their similarities and differences as well as their strengths and weaknesses, and develop a version of the problem that seems suitable for our purposes. Our version needs to keep the assumptions informative enough for the problem to arise ; but it also needs to keep them general enough for the traditional responses to the problem to still make sense, although a fleshed out version might well illustrate how and why these responses are not satisfactory Finally, it needs to keep the assumptions flexible enough for them to be adjustable to different theories of causation. But for all that to be possible, we need to overcome a second obstacle first: it is unclear what the appropriate relata of causal relations are. Among the suggested candidates are aspects, events, facts, features, situations, states of affairs, and tropes. 4 In order to develop a suitable version of the problem, we thus need to decide which causal relata we wish to treat as the appropriate ones and how we want to phrase causal assertions, that is we need to establish a terminological convention regarding the ontological elements between which we posit causal relations. In the broader context of the dissertation, the purpose of such terminological c onvention is not to endorse a particular ontology. Rather, it is to allow for a coherent way of talking about the exclusion problem and to analyze and compare standard solutions to the problem more effectively. 4 Schaffer (2016, section 1) lists the key proponents and publications for each of these suggestions.
25 2.1.1 The Lack of a Terminological Conventio n In ordinary language, we do not worry much about causal relata. When a friend admires a hospit the falling down the stairs, as a cause of either a state, her ankle being broken, or an event, the breaking of her ankle. That is, we often cite different kinds of relata as causes and effects and seem quite happy mingling them. Doing so is not necessarily problematic, for it often is possible to translate a statement citing one type of relatum into a statement citing another, seemingly without loss of information After all, we could easily switch between talk about the state of our events and states may be different kinds of ontological entit y However, even if and when su ch translations seem successful, our metaphysical intuitions regarding the resulting assertions might vary. The worry here is reminiscent of an argument made by Tamar Szab Gendler in the context of personal identity. She suggests that our responses to tho ught experiments are sometimes based on how the scenarios are described rather than on what they describe. 5 If she is right and if her point generalizes to other types of descriptions, adopting a terminological convention that leads to particular descripti ons of the issues involved in the exclusion problem may not be as innocent as it initially seems. By the same token, some issues that seem particular to some causal relata may turn out to ustrate the idea, albeit in the 5 See Gendler (1998) and Williams (1970) Williams introduces many of the thought experiments Gendler discusses
26 context of the problem of free will. According to Chisholm, sometimes agents rather than events or states of affairs cause behavior. 6 Of course, an action such as intentionally lowering the right hand might involve many inte rmediate causes: there are neural events in the brain, muscle contractions, and so forth. However, when we trace the causal chain back to its origin, assuming that there is one, at some point we will find a primitive causal relation with no intermediate ca uses, and that causal relation has the agent and a particular brain event as its causal relata. At first, this view may seem puzzling at best and incomprehensible at worst. It is hard to resist the question how exactly an agent causes that brain event. But Chisholm responds that there is nothing to explain. The agent just does it. 7 According to him, if we are puzzled, we are puzzled about causation in general, not about agent causation in particular. To see what he has in mind, we need to ask ourselves how we would explain how an event that initiates a causal chain causes that first effect in the chain. If there are no intermediate causes and thus no causal mechanisms to refer to, it is hard to see what we could refer to beyond the fact that events of such type typically produce effects of that type. But Chisholm would make the same point about agents. Agents just are the sort of entities that For starters, we might ask what explains that the agent makes the decision to lower the hand now rather than later. To answer the question satisfactorily having certain desires, intentions, or thoughts now rather than later. But then t he agent causal view seems to collapse into a regular event causal view. 8 6 See Chisholm (1964 /2003 144). 7 Ibid., p. 150. 8 /2003 44 51 ) for a brief overview of the main criticisms launched against Chisholm.
27 some of our intuitions may be taken to be specific to certain kinds of causal relata a gents for instance when they really are about causation in general. And sometimes it takes unfamiliar ways of framing the issues to make us see how puzzling some of our conventions are even though they may appear very natural and clear to us. The danger then is that we may fail to pick up on some prob lems lingering in the background of the exclusion problem exactly because we stopped thinking about some types of relata, for they might highlight these problems in ways that our adopted terminological convention does not. 9 So adopting a terminological convention is risky for two reasons: our intuitions may be swayed in unintended ways and the convention may make us overlook issues that might hold the key to solving the exclusion problem. The good news is that every convention carries these risks, so as a starting point we should just opt for the most commonly used relata in the metaphysics of causation, namely events, 10 since this option will make it easiest for us to apply these theories of causation to the exclusion problem in C hapter 3. Two of the mos t prominent theories of the nature of events are those proposed by Donald Davidson and Jaegwon Kim. Davidson treats event s as datable and locatable, non repeatable particulars that are fundamental in the sense that they are not constituted or realized by an ything else. As such, they are unanalyzable. Happily we can refer to them by way of descriptions. As is typical of be described as her tripping on the th ird stair and landing, left foot first, on a particularly hard 9 Partly because of that, this dissertation should be thought of as a first step towards a larger project, namely a comprehensive primer to the exclusion problem. Such a primer would have to keep track of the possible impact of different terminological conv entions as well as of concepts other than causation, such as completeness, overdetermination, etc. 10 I am not alone in this assessment. Schaffer (2016, paragraph 1 ), too, claims that the standard view of causal relata e vent and that their number is two
28 slate floor. But while the latter description is more fine grained, it serves the same purpose: to ry seriously. Rather than privileging the latter description as the one that cites the features of the breaking, however described, and its being causally related to identity conditions. 11 He thus distinguishes between the way we describe and pick out events on the one hand and how events are ontologically individuated on the other hand. His theory might not suit our purposes well beca use the exclusion problem is based on the idea that causal descriptions and explanations involving the mental may sometimes compete or be inconsistent with those involving the physical. But if mental and physical descriptions just are different ways of pic king out causally related events, it is unclear how that inconsistency is supposed to arise. So, rather than utilizing a Davidsonian theory of events as the basis of our terminological convention, we should discuss it later, in C hapter 4, as a response to the exclusion problem. construes events in a manner that he deems most useful for our thinking about causation and explanation. Rather than being fundamental, events are compl ex, structured entities; they are exemplifications of a property by a substance at a time. And by substance he means any object or any kind of entity that can be the bearer of properties. 12 Thus construed, it is natural to cite some of the properties cons tituting a particular event as those in virtue of which it is causally related to another event. Arguably, if our friend screamed loudly while falling, the scream did 11 See Davidson (1969 /1980 various especially pp. 26 33. Compare 12 See, for instance, Kim (1976 /1993 34).
29 with her left foot first, perhaps in a certain angle with a certain force, did. While this sounds like a property exemplification constitutes a distinct ev ent and that every difference in properties implies a difference in identity. So may not be the same event as her tripping and screaming loudly, even though Kim can construe the latter as being part of the former. 13 For Davidson, these are merely different descriptions of the same cause. 14 Why does this matter? A Kimean account might sometimes force us to single out events and properties as causes in a way that a Davidsonian account would not. In doing so, it also might e xclude other contenders; that is, there may be cases where the exclusion problem arises in virtue of the theory of the nature of events endorsed. 15 Where does th e above leave us with respect to our desired terminological convention? On about events in that it is not overly fine grained. But there is a serious disadvantage to a dhering T he properties cited in causal statements mainly serve, on his theory, to pick out the events between which causal relations are posited. It remains unclear, though, what role they play ontologically, if any. Are an y of the properties cited in a description might simply reject such in virtue of talk altogether, if ontologically understood. Since most 13 See Kim (19 76 /1993 44 46). 14 See Evnine (1991, 30). 15 alleges that it lacks the resources to distinguish between events and facts. Hendrickson (2006) discusses this criticism and defends a modi fied Kimean account against it.
30 discussions of the exclus ion problem heavily rely on in virtue of talk, it thus seems unwise to compatible with such in virtue of talk. But then again, his theory seems overly fine gr ained. In addition, it would be, strictly speaking false to say that an event has a property because the property exemplification is partly what constitutes that event. 16 But it seems rather natural to speak that way of events and properties. In light o f these advantages and disadvantages, might there be a way of talking about events without committing to a particular theory of them, perhaps even a way that preserves some of the advantages? Suppose we simply asserted that event a has property P and that a causes event b in virtue of property P without specifying whether events are simple or structured or whether they bear properties or are partly constituted by them. Could we make sense of the assertions from the viewpoint of each theory of events? If we adopted a Davidsonian theory, the role of the in virtue of clause might be unclear, depending on whether we adopt his nominalist inclinations in addition to his theory of events. But if we rejected his nominalism, we might understand the statement as clai ming that a is the cause of b and P is a property of a that figures into a causal law that a and b fall under. 17 If we adopted a Kimean theory, we might understand the statement as claiming that a is an exemplification of P by an object at a time and it is in virtue of P that it is the cause of b So, while each theory will require some rephrasing of the assertions, we still can make sense of them. 16 This may be a simplification. At times, Kim seems to think that his account may be consistent with treating some properties as properties of events, namely when the properties are not essential, that is constitutive of it See Kim (1976 /1993, 42 49 ). 17 According to Davidson (1970 /1980, 207
31 Sometimes, clarity might require specifying the time at which an event has a certain a has property P at t 1 a at t 1 causes event b at t 2 nominalist version of the Davidsonian theory can easily accom m odate that distinction because events can persist independently of the properties they bear. But within a Kimean theory, that distinction may be more difficult to capture, for the time of the property exemplification is the time at which the e vent takes place. Within a Davidsonian theory, it is also informative to specify the times in the above way, for the identity of the event is determined by its causal relations to other events. They may hold even when an event in question occurs a bit earl exemplified property, and the time of the exemplification constitutes the event, to specify the ant or poorly phrased. So while each theory, if endorsed, would affect the way we phrase causal statements in general as well as the assumptions of the exclusion problem in particular, it does not seem necessary to commit to either one of them. It seems we understand the statements above without settling whether events are simple or structured and without specifying whether events are bearers of properties or partly constituted by them. Positing causal relations between events, considering the properties had by these events, and specifying the properties in virtue of which the causal relations hold seems to suffice as a basic terminological convention. We may consider the statement and the terminology ontologically neutral in the sense that it can be inte rpreted in a manner consistent with the Davidsonian and Kimean event theories while preserving some of their virtues. 18 Such neutrality seems to serve the main purpose of a terminological convention 18 the test of time. Other theories have been developed that might be more plausible as an ontological view. But since our goal was merely to
32 for our project, namely to increase coherence. Different authors cite different kinds of causal relata, utilize different theories of events, and so forth. A relatively neutral way of formulating the assumptions of the exclusion problem will then allow us to discuss these authors without switching back and forth between different ways of phrasing the assumptions and the arguments proposed. 2.1.2 The Lack of a Canonical Version Often the exclusion problem is discussed without being carefully developed. We are to understand the pull of the problem without the detai ls being clear. If we attempted to provide clarification, we would be bound to make interpretative decisions eventually yielding claims of varying strength. As a result, many versions of the exclusion problem exist some of which merely differ in their word ing without changing the content much while others vary in strength or even in the amount of assumptions included. Consider, for example, how Karen Bennett presents the problem: Distinctness: Mental properties (and perhaps eve nts) are distinct from physic al properties (or events). Completeness: Every physical occurrence has a sufficient physical cause. Efficacy: Mental events sometimes cause physical ones, and sometimes do so in virtue of their mental properties. Nonoverdetermination: The effects of mental causes are not systematically overdetermined; they are not on a par with the deaths of firing squad victims. Exclusion: [N]o effect has more than one sufficient cause unless it is overdetermined. 19 property exemplificatio n view on metho do logical grounds, see Zele k (2009). For a brief, general overview of theories of the nature of events beyond Davidson and Kim, see Simons (2003). 19 Bennett (2008, 281) ; bold in the original.
33 As suggested in C hapter 1, the exclusion problem consists in the prima facie inconsistency of these assumptions. One of them has to be false, given the truth of the others or so it seems. The big question is, of course, which of these assumptions should be rejected and for what reasons. But these reasons might be more or less compelling depending on how exactly we understand these assumptions and how exactly we phrase them. We might wonder, for instance, whether the even ts that have no cause? Some say quantum physics allows for such events. 20 If that is true, perhaps we need only to claim that every physical occurrence that has a cause at all has a sufficient physical cause. 21 This weaker version of the nonoverdetermination assumption seems to be all that is required for the problem. We then might be better off adopting such a weaker version. Some authors use the alleged truth of some of these assumptions as a reason for rejecting another. Stephen Yablo, for instance, introduces one such exclusion argument, an argument for epiphenom en alism: If an event x is causally sufficient for an event y then no event x distinct from x is causally relevant to y ( exclusio n ) For every physical event y some physical event x is causally sufficient for y ( physical determinism ) For every physical event x and mental event x *, x is distinct from x ( dualism ). So, for every physical event y no mental event x is causally relevant to y ( epiphenomenalism ) 22 20 See Healey (2009, 673). 21 See Kim (2005, 16 ). 22 Yablo (1992a, 426) and Yablo (1992b, 247 248) ; italics in the original.
34 Aside from the different labeling and wording of the assumptions and the fact that the rejection of one of thes e assumptions is turned into a conclusion, the seeming omission of one assumption is particularly striking. While it looks as if Yablo has dropped the nonoverdetermination assumption, he really seems to have merged it with the exclusion assumption. 23 As a r esult, the exclusion assumption seems stronger than the one suggested by Bennett. But this comes at a price: the exclusion assumption thus formulated may be implausible, for assuming the existence of some overdetermination cases might be perfectly reasonab le. What we should be worried about instead is systematic overdetermination, as Bennett correctly observes. Additional complications linger What exactly does it mean for an event to be causally relevant to another? What if the threat of exclusion stems no t from two distinct events but from two distinct properties of the same event? Can we not expect there to be a very similar threat of exclusion if the causation involved is probabilistic, even though in such cases the exclusion and determinism assumption a s formulated are false because in such cases the cause is not causally sufficient for the effect? Of course, Bennett and Yablo are perfectly aware of some of these issues and address them. Bennett, for example, spends significant time on explaining what the notion of overdetermination requires, namely the nonvacuous truth of certain counterfactuals that she specifies, thus illuminating the nonoverdetermination assumption. 24 And Yablo utilizes the rather 23 Some authors include more assumptions rather than fewer Jesper Kallestrup attributes the view to Jaegwon Kim that the causal exclusion problem involves an additional assumption namely a supervenience claim; see Kallestrup (2006 465 ). It is not clear, however, wh ether Kim in fact holds this. Kim (2005 21 ) distinguishes between the exclusion argument and the problem of mental causation, and it is only the latter broader problem that involves supervenience. Kim (1998) claims that the exclusion problem arises from the supervenience argument, but the exclusion problem itself, as described by him, does not include the supervenience claim. 24 See Bennett (2008, 288).
35 vague notion of causal relevance to explain how think ing about the relation between mental and physical events and properties as analogous to the relation between determinables and determinates can avoid the threat of overdetermination and exclusion. 25 Others, such as Matthew Haug and Eric Funkhouser careful ly distinguish between different kinds of completeness and overdetermination and investigate the impact each kind would have on the exclusion problem. Haug, for instance, begins his analysis with the following formulation of the completeness assumption: Ca usal sufficiency completeness (CSC) : For every physical event and every time t if that event has a sufficient cause at t then it has a physical sufficient cause at t In other words, every physical event is completely causally determined, insofar as it i s causally determined, by physical events. 26 and this before he even starts distinguishing di fferent kinds of completeness. The examples above may suffice to illustrate the range of differences regarding how the exclusion problem is presented and discussed in the literature. Unfortunately, there is no canonical formulation of the problem, just ma ny versions of it. Some versions are so sketchy that they merely appeal to our intuitive grasp of the problem. The inferences required for the problem to arise are far from clear. And if turned into an argument against the truth of one of the assumptions, the result is nowhere near a deductive argument. Others spell out the assumptions more, but do so in different ways resulting in claims of varying strengths. This makes it difficult to compare and assess responses to the problem, for they are responses to particular versions of 25 See Yablo (1992a, 432 434) and Yablo (1992b, 253ff.). 26 Haug (2009, 382).
36 it. What might seem like a promising approach to one version might not be promising at all regarding another version. More importantly, if we introduce a particular version of the problem, one that is not used and shared by those phi losophers whose approach we assess, it might be easy to get caught in a straw man fallacy of the following sort: Person X claims that theory T offers a solution to problem Ex 1 but theory T does not work as a solution to problem Ex 2 Therefore, X is mistaken. Of course, this is not what we are going to do. Since ultimately we want to analyze the impact that theories of causation might have on the exclusion problem, details will likely matter. We should thus aim for making the assumptions as clea r as possible, especially those that feature causal notions. Striving for such clarity by eliminating ambiguities and vagueness sometimes will result in a higher degree of specificity. Sometimes, such specificity is a virtue, but other times it may be coun terproductive. We need to keep a balance between specificity and general applicability. Our version should capture cases in which distinct types of events are causally excluded as well as cases in which a distinct feature or property of one and the same ev ent is excluded from being causally relevant. And it should be applicable to probabilis tic causation as well as to non probabilistic causation. In addition, we should strive for making the assumptions as strong as possible without undermining their plausib stronger in the sense that it excludes the possibility of overdetermination altogether whereas version is compatible with non systematic overdetermination. Given that examples of a unnecessarily strong, if not implausible in light of such examples. In addition, the plausibility of the assumptions w ill also depend on how strong a version of the exclusion problem they are part
37 of. An unnecessarily strong assumption embedded in a set of weaker assumptions that entail relatively little and that are more difficult to reject may sway our intuitions regarding which assumption should be rejected. By the same token, the weaker the assumptions, the less force our discussion will have; for it is possible that an approach avoids problems in weak versions while failing to solve problems in strong versions. A view that entails an occasional n on systematic overd If an approach can convince even as a response to the strong version, that is better evidence for its promise and quality than its managing to deal with weak versions. Applying standard responses to our strong version thus is to be considered a test for the quality of the response. It is not to be understood as a representation regarding what th e authors were arguing. It determinate distinction to deal with his version of the exclusion problem might be a promising a pproach while it might be a non starter in response to our version of it. In such situations, the conclusion will be that there is a version of the exclusion problem that such standard responses cannot deal with successfully. This is not yet to say that our version of the exclusion problem is the most plausible one available. In th e final chapter, we will assess the plausibility of a particularly strong version of the exclusion problem, compare it to weaker versions, and suggest that our version is a promising candidate for a canonical version that might prove most useful for future discussion. 2.2 Formulating the Exclusion Problem If mental properties were identical with physical properties, there may be no particular problem with mental causation. Mental events could have a causal impact on physical events in virtue of their mental properties simply because mental properties are just physical properties.
38 Hence, mental causation may be no more mysterious than physical causation. For the problem of mental causation and the exclusion problem to get off the ground, we need to assume tha t mental properties are not identical with physical properties. As we have seen earlier, this is more or less how the first assumption sometimes is phrased. However, it could be the case that nature is not that homogenous. Some mental properties may be ide ntical with physical properties, others may not be. 27 Hence, we should use a weaker version of the distinctness assumption: 1. Distinctness: There is an important kind k of mental property, such that, if a mental property is of that kind k it is distinct from any physical property. It is important to appreciate ju st how strong the last part of ( 1) is. If we thought of mental properties as complex, highly relational physical properties, for instance, we would not be thinking of mental properties of the kind k So (1) does not seem compatible with some versions of reductive physicalism, certainly not those that are effectively type identity theories. 28 Proponents of such theories will presumably remain unmoved by the exclusion problem to begin with. 29 The secon d assumption of the exclusion problem states that mental causation with re spect to bodily behavior exists; i.e. there are mental events that cause physical events in virtue of their mental properties. Most philosophers phrase this in terms of the causal e fficacy of mental events. Aga in, we have to be cautious causal efficacy is a term of art that acquires its content partly 27 Jaegwon Kim, for instance, claims that while most mental properties are fully functionalizable and physically reducible, qualia are not. See Kim (2005, ch apter 6). 28 This is not th e place to provide a comprehensive overview of the many variants of reductive physicalism. Suffice it to say that some versions are supposed to be reductive without assuming identity. Ernest Nagel, for example, argues that reduction merely requires that th ere are bridge laws connecting mental predicates M with neurological, or for our purposes physical predicates P such that x is M iff x is P That would seem compatible with (1). For a brief overview, see Stoljar (2015 section 7 ). 29 David Papinea u might agree, for he points out that it is the distinctness assumption that seems to get us into trouble in the first place by generating puzzles about how the brain generates mental states and how such states can ever be causally relevant; see Papineau ( 2002, 2f.).
39 through the terms it is contrasted with. Stephen Yablo, for instance, distinguishes causally efficacious events from causally r elevant events and explains the latter by analogy to the relation between determinables and determinates. 30 Fred Dretske, to give another example, calls causally efficacious events triggering causes and distinguishes them from structuring causes While trig gering causes explain why an event happens right now, structuring causes explain why a particular event rather than another event causes that event. 31 causally relevant events, the historical dimension of structuring causes is important. As a result, doubt, the most general formulation of the exclusion problem should be free of such theory dependencies. To highlight this lack of the ory dependence let us introduce another term of art: causal impact its functional role for the exclusion problem. In section 2.1.2, we saw that the exclusion problem is sometimes turned into an argument for epiphenomenalism. Roughly put, epiphenomenalism states that mental events play no causal role in the production of effects. 32 However, this can be interpreted in two different ways. For an event to play no ca usal role may mean that it is not strictly involved in the production of the effect; i.e. it is not part of the causal chain that leads to the effect. Alternatively, it may mean that it has no explanatory value ; i.e. it is not needed for a satisfactory c ausal explanation of why an event occurred. 33 Accordingly, a strong version of 30 See Yablo (1992b, 249ff.). 31 See Dretske (1993, 122). 32 See, for example, Robinson (2015). 33 Playing no causal role in this latter sense would rule out both causally relevant events la Yablo and structuring causes la Dretske.
40 epiphenomenalism would state that mental events are never strictly involved nor do they ever have any explanatory value. A weak version of epiphenomenalism would state that menta l events are never strictly involved but that they could have some explanatory value. 34 If an event having a causal impact merely implies that the event is strictly involved in the production of the effect, the exclusion problem could indeed be turned int o an argument for epiphenomenalism namely, if it concludes that no mental event can be strictly involved in the production of physical or mental events. This, however, would leave open which version of epiphenomenalism is true. If, on the other hand, an ev ent having a causal impact implies that the event is either strictly involved or that it at least has some explanatory power, the problem could be turned into a specific argument for strong epiphenomenalism namely, if it concludes that mental events can neither be strictly involved nor have any explanatory power. While the in a much more severe problem it might be too strong for what the exclusion problem is usually taken to show. Yablo (1992b), for i nstance, concludes that the exclusion problem succeeds in establishing that mental events cannot b e causally efficacious, but it does not establish that mental events are not causally relevant, i.e. he failure of the resulting arguments because they were too ambitious, it then seems reasonable to adopt the first interpretation of point. Hence, whenever we talk about an event having causal impact from now on, we will talk abou t an event that is strictly involved in the production of the effect, i.e. an event that is part of a causal chain that leads to the effect. We have to bear in mind, 34 It is an interesting question whether having explanatory value in this context amounts to anything ontologically. with commitment to epiphenomenalism des pite being a monist. For a discussion of this charge, see Davidson (1993) and McLaughlin (1993), for instance.
41 though, that this disqualifies the exclusion problem from being turned into an argument f or strong epiphenomenalism unless further assumptions are added. Let us now return to the second assumption of the exclusion problem. The claim that mental events have some causal impact on physical events in virtue of their mental properties is clearly to o strong because it is too general. It seems quite possible that while some types of mental events do have a causal impact in virtue of their mental properties, others do not. Thus, the claim needs to be weaker: 2. Causal Impact: Sometimes, mental events hav e a causal impact on physical events in virtue of their mental properties. Because we were so concerned about the generality of (1) and (2), we now run into another problem. The conjunction of (1) and (2) is consistent with two different stories about the causal impact of mental events. 2.1 Some mental events have a causal impact on physical events in virtue of mental properties that are identical with physical properties. 2.2 Some mental events have a causal impact on physical events in virtue of mental properties of the kind k that are distinct from any physical properties. As briefly mentioned above, the exclusion problem is sometimes used as an argument against non reductive physicalism and other times as an argument for epiphenomenalism. To be able to utilize the problem in this way, we need to ensure that the mental properties addressed in (2) are at least sometimes distinct from physical properties, i.e. at least sometimes they need to be of the important kind k as posited in ( 2.2 ) We should modif y ( 2) accordingly: 2 Causal Impact: Sometimes, mental events have a causal impact on physical events in virtue of their mental properties of the kind k There remains a problem. Its severity will become fully apparent only after the assumptions regarding completeness, non overdetermination, and exclusion have been introduced and
42 discussed. Nonetheless, this is the best time to address the problem because it will lead to a further assumption that best is seen as a qualifier of (2*). Many philosophers belie ve that a view according to which many events are or might be overdetermined is not plausible. This is why assumption (5) of the exclusion problem is a claim about the nonoverdetermination of physical events. In a strong version, assumption (5) then would claim that events are never overdetermined. 35 Thus interpreted, the claim would line up nicely with (1) and the yet to be discussed assumptions of the exclusion problem. If it is true that ( a) for every physical event there is another physical event that ha s caused it, that ( b) events are never overdetermined, and that ( c) no event has more than one sufficient cause unless it is overdetermined, it seems to follow that (2*) is false. However, we may be inclined to think that under certain circumstances partic ular events might very well be overdetermined. A popular example is the firing squad scenario. We are asked to imagine a firing squad executing a prisoner. Two of the bullets hit the prisoner in the same deadly manner, say in the head, at the same time. I t seems reasonable to consider both shots as causes of the death of the prisoner. Some judge such instances to be genuine cases of overdetermination. 36 Since it at least is possible that such cases exist, it seems wise to look for a weaker version of the nonoverdetermination assumption. Rather than not allowing for overdetermination to occur at 35 36 Of course, not everyone agrees. The general puzz le is this. A counterfactual analysis requires the following: if shot A is the cause of the death of the prisoner, then it must be true that if shot A had not occurred, the death would not have occurred (at the same time in the same fashion). Similarly, if shot B then had shot B not occurred, the death would not have occurred. But in the firing squad scenario, both counterfactuals are false. Thus, some argue, neither shot A nor shot B death. See Schaffer (2003a) for a defense of both shots being a cause as well as overdeterminers.
43 all, it seems to be safer to merely claim that there is no s ystematic overdetermination. And in fact, this is the version that most philosophers endorse. 37 But now (2*) cannot be rejected any longer based on the other assumptions If (2*) is true, then it merely follows from the other assumptions that any particula r mental event that causes a physical event in virtue of a mental property of the kind k must be an overdetermining event. But on its own, this does not imply systematic overdetermination because there might be something atypical about such cases. Sometime s, the instantiation of a mental property of the kind k might have a causal impact on physical events. Other times, it might not have such an impact. In other words, particular mental events that have a causal impact on physical events in virtue of mental properties of the kind k and thereby overdetermine them might not be representative for their mental types. And as long they are not representative, such cases of mental causation might not be particularly bothersome. 38 Consider the following analogy: suppo se it is true that in a Philosophy of Mind class only students with yellow socks got an A. This is not necessarily surprising for maybe only one student got an A. But suppose twelve out of twenty students got an A and all of them wore yellow socks and no other students wore yellow socks Most likely, this would make us wonder exactly because it would indicate some systematic correlation between the color of socks the students wear and the grades they got. Similarly, if every mental event that has a causal impact on physical events in virtue of its mental properties of the kind k is an overdetermining event, 37 We have seen that Bennett endorses such a version. See Stoljar (2010, 217) for a way to spell out the exclusion principle in a way that it is sensitive to a events e and e* if e directly causes e* and if there is a property F which is causally efficacious in e e *, then there is no property G such that (a) G is distinct from F and (b) G is causally efficacious in e e 38 I wish to thank Gene Witmer for pointing this problem out to me.
44 there would be a systematic correlation between mental causation of a certain kind and overdetermination. From there, it is not far from a clear violati on of the weak nonoverdetermination assumption. While the existence of non typical mental events that have a causal impact on physical events is conceptually possible, they not only le ad to an odd picture of reality but they also threaten to undermine the explanatory power of causal explanations that involve them. To illustrate this oddity and threat consider the following example. Suppose that my thought that it might rain today indeed had a causal impact on my behavior; it caused me to carry an umbrella. If the completeness assumption is true, then there was also a physical event in my brain, for instance that caused my behavior. To avoid systematic overdetermination, we now are committed to say ing that my thought that it might rain today is not represent ative of that kind of mental event, i.e. that there are many mental events in general and thoughts in particular that do not have any causal impact in virtue of being a thought or another mental property. But if thoughts, even of that particular kind, usu ally do not have such causal impact, then it seems that for whatever reason the thought had the causal impact it had, it was not because it was that thought, nor was it because it was a mental event. 39 And if this is true, the reference to my thought that i t might rain today was not an appropriate causal explanation for why I carry an umbrella today. 39 Interestingly, this seems to be more of a problem for theories of causation that presuppose some sort of regularity than for singularity theories of causation. In particular, reductionist theories of causation may not work if mental events turn out to be t oo unrepresentative for their kinds. According to Jonathan Schaffer, for example, causation reduces to history plus laws of nature; see Schaffer (2008 5 ). Typica lly, laws of nature express non accidental regularities or generalizations of some sort. It the n seems that if mental events are too atypical, there might not be any laws they can fall under. But if reductionism is right, that just means that they cannot have a causal impact. We will get back to this in chapter 4. A second interesting point may be l ingering here: if there are events that have a causal impact but that are not representative of their kinds in the way described above, there may be events that are strictly involved in the production of an event but that have no explanatory value. If such causes are possible, the notion of an adequate cause, which we will introduce below, needs rethinking and perhaps modification.
45 Given the above considerations and problems regarding completeness and typicality, we may be well advised to add an assumption not usually listed: 3. Typicality: Some mental events have a causal impact on physical events in virtue of mental properties of the kind k and this causal impact is not atypical of events that have mental properties of the kind k In other words, such mental events are representative of events that have mental properties of the kind k in that they tend to be strictly involved in the production of physical effects in virtue of properties of the kind k 40 The fourth assumption of the exclusion problem is motivated by our current scientific worldview. It is usually phrased in terms of causal sufficiency: every physical event has a sufficient physical cause, if it is caused at all. Bearing in mind that later we will discuss probabilistic causation and apply it to the exclusion problem, we cann ot simply adopt this formulation. Some events may not have a sufficient physical cause but rather one that made their occurren ce to some degree likely. Often this is put in terms of probability raising: if an event is a probabilistic cause of another eve nt then it contributed to raising the probability of the occurrence of that event 41 To say that an event has a probabilistic causal impact on another event then is to say that ( a) the event raised the probability of the occurrence of that type of event and that ( b) the event is part of a causal chain that leads to the effect We might wonder whether it is necessary to assume that a probabilistic cause of an event 40 It is worth noting that while the typicality assumption is usually not discussed in the context of the exclusion problem, it occasionally surfaces in the philosophy of science. For example, Harold Kincaid points out that at least cannot vary according to contex assumption (3). 41 For a good overview of various probabilistic theories of causation, see Williamson (2009).
46 qu alifies as having a causal impact, even if the cause lowers the probability. Consider, for example, a woman who gets pregnant in spite of having taken contraceptive pills. 42 On the one hand, it seems natural to say that having taken the pills has had a caus al impact on the likelihood of the pregnancy to occur. On the other hand, it seems wrong to say that taking the pills partly caused the pregnancy. We might be more inclined to say that such seemingly probability lowering event really contributed to the pre vention of an event, the pregnancy, albeit not sufficiently so ; that is, it increased the likelihood of the pregnancy not to occur. Happily w e need to take sides for we can simply state more neutrally that if an event has a probabilistic causa l impact on another event, it affected the pro bability of the effect to occur What is important, however, is that a complete probabilistic explanation does not require any reference to further events or other properties, even though probabilistic causes do not necessitate their effects. Since both kinds of causes, sufficient ones and probability affecting ones, can share this feature, we may use it to define another term of art: adequate cause An adequate cause is one that, if it were cited in a causal explanation, would be all that one needs to cite as a causal factor operative at that time in order to provide a satisfactory explanation of why another event occurs. 43 And a satisfactory explanation is one that does not require any reference 42 While Suppes (1970) defends the view of causes as probability rais ers, Rosen (1978, 605 606) argues that some causes may lower the probability of effects to occur. Rosen also discusses the pregnancy example in more detail. 43 adequate cause notice that our definition may differ subtly from previous uses. For example, D. M. Armstrong uses the term to refer to causal chains that appropriately connect cause and effect. Similar to our notion of causal impact, this guarantees that the cause is str ictly involved in the production of the effect. But unlike our notion, an Armstrongian adequate cause may not always yield a satisfactory explanation for why the effect occurs. It only may be one causal chain among many chains involved in the production of the effect; it may merely be a part of the overall cause. If so, it is that chain or that part that produces and explains the complexity of the effect. For example, human behavior often is very complex. An Armstrongian adequate cause needs to mirror this complexity in some way, or else it would not essarily to say that the adequate cause explains all features of the behavior, i.e., why it occurred in the exact way it did.
47 to further ev ents or properties. 44 Because the reference to an epistemic context is conditional, we can still consider the resulting concept a metaphysical one and use it to rephrase the fourth assumption in terms of adequate causes. 4. Completeness: If a physical event p 2 is caused at all, it has at least one adequate cause p 1 that is physical. In other words, if p 2 is caused at all, there is some event p 1 such that p 1 is a cause of p 2 and p 1 is an adequate cause of p 2 in virtue of its physical property A few remarks on the relation between causal impact and adequate causes are in order. Any cause at the very least involves the property or properties in virtue of which the event has causal impact. In some cases, citing these properties and that event might yield a satisfa ctory causal explanation of why the effect occurred. However, this is not necessarily all that an adequate cause involves. Suppose that a pigeon is trained to peck at red things. When confronted with a scarlet object, the pigeon pecks. Presumably, the wave length of the light that the object reflects causes the pigeon to peck. But this, in some sense, is not a satisfactory explanation ; for it is not because the objec t is scarlet that the pigeon pecks but because it is red. 45 While it is the wavelength corresp onding to scarlet in virtue of which the event has its causal impact, the fact that that particular wavelength is in a certain range is causally relevant and needed for a 44 explanations, albei t not in the context of probabilistic causation. For him, an adequate explanation is one that references causal conditions sufficient for an effect; see Malcolm (1968, 56). Kim, too, thinks of complete f causal conditions for the explanandum; see Kim (1989 /1993 251). We will need to address one problem at the beginning of Chapter 3, though. Which causal explanations provide satisfying answers to our causal questions may depend on the kind of question we asked as well as on the context in which the question is ask; that is, there may be pragmatic factors external to the situation that influence whether or not we consider a causal explanation satisfactory. But, as we will see, this is not an argument again st our notion of an adequate cause but rather indicates that it is not properly defined yet. It is worth noting that the causal questions we ask may also specify how far back on the causal chain leading to the effect we are supposed to go when identifying the cause, and that timespan may be part of what determines whether or not a cause is adequate. 45 The example is taken from Yablo (1992b, 257).
48 satisfactory causal explanation. Hence, adequate causes might include properties in a ddition to those in virtue of which the cause has its impact. We should interpret the completeness assumption as stating that, if there are physical events that require such additional properties, those properties need to be physical as well. 46 apitulate. So far, the exclusion problem assumes that sometimes mental events have some causal impact on physical events in virtue of their mental properties of the kind k This causal impact is not atypical of events that have mental properties of the kin d k But notice that the completeness assumption, be it on its own or in combination with the other assumptions, does not exclude the possibility that there are mental events that have a causal impact on physical events in virtue of their mental properties even if those mental properties are distinct from physical properties. It merely implies that whenever there is a mental event that has a causal impact on a physical event in virtue o f one of its mental properties there also is a physical event that is an adequate cause on its own. Regardless of whether the mental event is an adequate cause or not, as long as it has a causal impact in virtue of its mental properties, the physical effect is overdetermined. 47 As mentioned above, single overdetermination may not seem to be as outrageous as systematic overdetermination, and yet that is what the typicality assumption seems to commit us to. If there is a single case of overdetermination involving mental events, then that is a typical case of a cer tain type of me ntal events, mental events that have a causal impact on physical events in virtue of mental properties of the kind k typically overdetermine physical 46 This makes our version of the completeness assumption significantly stronger than those commonly discussed. We will address the question whe ther this might be too strong a claim later in C hapter 5. 47 This is only prima facie true. In the next chapter we will encounter one response to the exclusion problem that attempts to undermine this alleged implication.
49 events. 48 If we are inclined to think that nature most likely is not organized in such a way, we agree with the fifth assumption of the exclusion problem: 5. Nonoverdetermination: Events are not systematically overdetermined. Earlier we saw that an adequate cause is one that, if it were cited in a causal explanation, would be all that one needed to cite as a causal factor operative at that time in order to provide a satisfactory explanation of why another event occurs. And a satisfactory explanation is one that does not require any reference to further events or properties. This alone does not gu arantee that any explanation that involves more than one adequate cause is also a case of overdetermination. least one adequate cause and some other causal factor. This additional causal factor could be another adequate cause, but it also could be just another event that has some causal impact or another property that has some explanatory value. Such additional c ausal factors do not necessarily amount to a second adequate cause. But even if they did not amount to such they would be explanatorily redundant; for if they were not, the physical cause could not be an adequate one. Hence, we need a further assumption t hat ties the concept of an adequate cause to the concept of overdetermination and eliminates other ways for an adequate cause to coexist with any further causal impact. This is done by the so called exclusion assumption: 6. Exclusion: If an event has an adequate cause at time t then it does not have another cause at t unless it is overdetermined. We now have established all the assumptions of the exclusion problem. As mentioned earlier, the problem is that (1) (6) cannot all be true a t the same time, i.e. at least one of the 48 T he systematicity of overdetermination would threaten even more severely if the only events that genuinely overdetermine were events that have causal impact in virtue of properties of the kind k
50 assumptions has to be given up. Sometimes, the decision about which assumption to reject is based on which theory of mind an author favors. Other times, the decision is the result of weighing the plausibility of e ach assumption against the others and rejecting the one that seems the least promising or justified one. Depending on which assumption is rejected, the exclusion problem may in some cases turn into an argument for a particular theory of mind. 2.3 Generali zing the Exclusion Problem In section 2.1.2, we mentioned that the exclusion problem is sometimes turned into an argument for epiphenomenalism, the view that mental events never have a causal impact on physical events in virtue of their mental properties. That is, assumptions (1) and (3) to (6) are used as premises for the conclusion that the causal impact assumption (2) is false. Of course, (2) is crucial not only for the exclusion problem but for all arguments dealing with mental causation. Critics worry, for example, that (2) would violate conservation laws in physics 49 including those that the completeness assumption (4) may rely on 50 One type of response to these worries is to make them seem less pressing E ven if (2) implied such violation, so would any nonphysical event for which some causal impact is claimed in virtue of its nonphysical propert ies Presumably, many causal explanations in the special sciences invoke such impact. 51 In the claim ar kind of ch emical substance is cited as a cause of a particular kind of biological process or state. But while the properties being a toxin and being a cell membrane might relate to physical properties in important ways, they may not be identical with an y of them. Given that we cite such nonphysical events and properties 49 Fodor (1981 114ff. ) and Dennett (1991 35 ) use this point as a general argument against substance dualism as well as against anything nonphysical that allegedly causally impacts the physical. 50 See Crane (1995, 216ff). 51 See, for example, Burge (1993, 102), Baker (1993, 77), and Van Gulick (1992, 325).
51 in our causal explanations all the time, why then should we worry about mental causation in particular? Of course, th is strategy is a double edged sword, for we might as well just accept that the causal claims in the special sciences are problematic 52 But that just illustrates how relevant the problem is for the philosophy of science Prima face, this seems like a good rea son to generalize the exclusion problem by substituting references to mental events and properties with references to nonphysical events and properties While the threat of exclusion may apply to some events and properties cited in the special sciences, t he stakes are not necessarily the same in this context. If events never have a causal impact in virtue of their nonphysical properties, we may have to change our understanding of some of our epistemic institutions and practices including the special scien ces that invoke these properties in their causal explanations. Some of their explanations would be literally false, but they still might serve as abbreviated versions of more complex physical explanations. But if events never have a causal impact in virtue of mental properties such as having particular beliefs, desires, and intentions, we would have to revise the way we think about ourselves as human beings and, arguably, should strive to change some of our moral and social practices Neither our traditiona l concept of agency as behavior caused by intentions nor the view of ourselves as morally responsible agents might survive such drastic causal exclusion of mental events and properties. 53 Caution is in order, though. While there are good reasons for trying to save some kinds of mental events and properties from exclusion, excluding others might not have any effect at all on 52 Kim (1997, not a problem for us, but rather that geological and biological causation, too, is a problem. If anyone were to retort, 53 Robb and Heil (2013, section 1.1) agree that at least some types of moral responsibility require mental causation.
52 our understanding of human nature and practices. Jaegwon Kim, for example, believes tha t qualia, the phenomenal qualities of experience resist functionalization and face causal exclusion. But this mental residue, as he calls it, does not figure into our most promising physicalist accounts of intentions and other mental phenomena required for human agency. 54 A plausible and useful generali zed version of the exclusion problem thus should allow for the possibility that some kinds of nonphysical events and properties face e xclusion whereas others do not. Our version of the problem seems particularly well suited in that respect, for u nlike Be s our distinctness assumption (1) is not overly general. It does not claim that all mental events and properties are distinct from ph ysical ones but merely that there is a n important kind of mental property that is distinct from any physical property. (2) then assures that it is in virtue of mental properties of such kind that some mental events have a c a usal impact on physical events. Generalizing the assumptions to nonphysical events and properties would then not yield unduly strong assumptions, for they would merely state that some, but not necessarily all, nonphysical events have a causal impact in virtue of nonphysical properties of a n important kind. This may include some of the events and properties cited by the special sciences, but it does not have to. It could be the case that the only nonphysical events that have a causal impact are those that have an impact in virtue of a special kind of mental property. Using our version of the exclusion problem as the basis for the generalization thus has the virtue of potentially applying to the special sciences without committing us to t he claim that it does. Replacing all references to mental events and properties with references to nonphysical events and pr operties yields the following generalized version of the exclusion problem : 54 K im (2005), especially pp. 161 174. This is not a new idea. Armstrong (1968/1993, 234) already suggested that perceptions that are decoupled from beliefs may not have the same or even any causal impact on behavior. The beliefs do the causal work, but they c an be accounted for in terms of his central state materialism.
53 1. Distinctness: There is an important kind k of nonphysical property, such that, if a nonphysical property is of that kind k it is distinct from any physical property. 55 2. Causal Impac t: Sometimes, nonphysical events have a causal impact on physical events in virtue of their nonphysical properties of the kind k 3. Typicality: Some nonphysical events have a causal impact on physical events in virtue of nonphysical properties of the kind k and this causal impact is not atypical of events that have nonphysical properties of the kind k 4. Completeness: If a physical event p 2 is caused at all, it has at least one adequate cause p 1 that is physical. 5. Nonoverdetermination: Events are not systematic ally overdetermined. 6. Exclusion: If an event has an adequate cause at time t then it does not have another cause at t unless it is overdetermined. 56 Of course, this is not the only way to generalize the problem In fact, we may worry that contrasting the physical with the nonphysical may not be very informative for it is not clear what we are referring to properties currently cited in physics or to properties cited in a future complete d physics. 57 On the one hand, a future physics may qualify as being complete exactly because it includes all properties that in fact are physical assuming that being physical is a natural kind property Since do not know which properties are 55 The motivation for contrasting nonphysical with physical properties is explained further below. Thus contrasted, it is analytically true that nonphysical properties are not physical. But whet her or not a particular property is in fact nonphysical is an empirical question. When discussing particular properties, the nonphysical property in question is one that is treated by one or several of the special sciences as a serious contender for a nonp hysical property. 56 From now on, numbered assumptions in the text will refer to these generalized assumptions, unless otherwise noted. 57 Melnyk (199 7, 623) and Gillett and Witmer (2001, 302 303). See also Crane and Mellor (1990) for an argument against the suggested contrast. They argue that there is no good reason for excluding the mental, and perhaps by extension the nonphysical, from the physical r ealm. But then physicalism may be vacuous if understood as a theory of the mind, and perhaps by extension of the nonphysical in general.
54 threatened by exclusion. On the other hand, our current physics is likely subject to change and not complete in the abo ve sense But if it gets t cases the exclusion problem would threaten the exclusion of some events and properties that do not fa ce such a threat metaphysically Alternatively, we might consider a contrast between fundamental and higher level events and proper t i es 58 However, t he main problem with that approach is that it commits us to the existence of hierarchically stratified levels as well as to the assumption that there is a fundamental level in the first place. It might not be unreasonable to assume that reality is thus organized. 59 It might even be possible to provide criteria for what constitutes a particular level or how to individuate levels. But there are good reasons to remain critical of simply presupposing that a fundamental level exists. For example, Jonathan Schaffer claims that there is no scientific evidence that actually supports this assumption, even though some scientists endorse it as a presupposition or conjecture. 60 greater explanatory scope in that the workings of every single en tity is explicable in terms of the workings of its part s and infinite division yields a more elegant hypothesis in that the pattern of 61 Put more simply: not endorsing the assumption that there is a fundame ntal level might yield better theories. What seems most important in the context of the exclusion problem however, is that the notion of a fundamental 58 We could also contrast physical events with events that supervene on them. Humphreys (2009, 642), for example, sugges ts that the exclusion argument, and by extension the exclusion problem, can be applied to any kind of event that supervenes upon physical events. 59 John Tienson, for example, claims nonreductive physicalists must endorse such a level ontology. However, he of a general exclusion problem 60 See Schaffer (2003b, 502ff.). 61 Ibid., p. 500f.
55 level might carry the connotation of ontological or perhaps even causal priority. Or, p ut more colloquia lly, the fundamental level is where the real causal action takes place. Such suggestive priority might further bias us in unwanted ways regarding the possibility of a causal impact of nonphysical events, especially in the context of an already very strong completeness assumption (4) Thus, it seems best to avoid using the contrast between fundamental and higher level events and properties and opt for contrasting physical and nonphysical events and properties instead While that carries the risk of it remai ning somewhat unclear which particular events and properties are threatened by exclusion metaphysically, it may suffice to treat those events as serious contenders that according to the special sciences, are said to have a causal impact in virtue of their nonphysical properties Whether or not these properties are in fact nonphysical remains to be seen, but currently the special sciences offer the best contenders for such properties 2.4 Chapter Conclusion So far, we have accomplished three things. First, in order to develop a suitable version of the exclusion problem, we needed to adopt a convention regarding the ontological entities between which we posit relations. We have o pted for treating events as non repeatable particulars that have properties at a time. This should be understood as a way of talking about events, properties, and causation rather than as endorsing a particular theory of events. What matters is that we can specify the properties in virtue of which events have causal impact, not whether events in fact are the bearers of properties or partly constituted by them. Even though our We are thus bound to ignore alternative views according to which the relata are aspects, facts, features, situations, states of
56 affairs, or tropes. W hether we can modify our framework to incorporate such views goes beyond the scope of this dissertation, a s does the question whether these views yield responses to the exclusion problem that are distinct from those discussed here. Answering that question could be part of a larger project, namely a comprehensive primer on the exclusion problem. Second, we hav e developed a version of the exclusion problem that is more precise than most versions discussed in the literature. References to rather vague concepts such as being causally relevant have been replaced by references to more clearly defined concepts such a s having a causal impact and being an adequate cause While some versions of the problem boldly assume that mental properties are generally distinct from physical properties, our version is sensitive to and compatible with the possibility that some mental properties are identical with physical properties whereas others are not. And, as a result of trying to be more precise the need for an additional assumption became apparent, namely the typicality assumption, which helps illuminating some of the inference s needed for the problem to get off the ground. Third, we have generalized the exclusion problem. If the pr oblem indeed extends to some non physical nonmental events and properties, it threatens the status of some e xplanations in the special sciences. Not only does this highlight the gravity of the problem, it also allows us to contribute to arguments made in metaphysics and the philosophy of science regarding the proper scope of the problem. In return, we can utilize some of those arguments for bringing ou r concepts of causation and causal explanation into sharper focus. Generalizing the problem thus naturally sets the stage for merging debates in metaphysics, the philosophy of mind, and the philosophy of science.
57 CHAPTER 3 CAUSATION AND THE EXCLUSION PROBLEM 3 .1 A Qualification The goal of this chapter is twofold. First, we want to test whether the exclusion problem is affected by particular theories of causation. Perhaps we are more inclined to endorse or reject one or several of the assumptions when we presuppose a particular theory of causation ; o r perhaps the problem appears to be more or less seve re depending on which theory of causation we presuppose. Second, we want to set the stage for assessing some of the standard respo nses to the problem in Chapter 4 If theories of causation indeed affect the dynamic of the exclusion problem, they may also affect how successful some of the responses to the problem are both in virtue of affecting the exclusion problem itself and in vir tue of changing the meaning of claims involved in these responses. However, covering all theories of causation would be too ambitious a project. Over the past two and a half millennia, many different theories of causation and even more modifications and versions of these theories, have been developed By the same token, in section 1.1 we complained that the assumption that the exclusion problem remains the same no matter what theory of causation we presuppose has not been justified in any systematic mann er. We granted, though, that a potential impact of a few particular theories such as the counterfactual theory of causation has been explored. To decide whether or not th e above assumption can be justified, we thus need to look at more than just two or three theories, or else we may accidentally miss the theories that are most promising regarding a potential impact. So, while we need to cover enough theories for our inquiry to be systematic, we also need to exclude some theories and versions for the project to be feasible. An easy way to achieve this is to use standard anthologies such as Beebee, Hitchcock, and Menzies (2009) and Sosa and Tooley (1993 b ) to determine which
58 theories of causation are still relevant in contemporary metaphysics and which may be worth applying to the exclusion problem. But rather tha n just using the most prominent ones, we may also include a few lesser known theories because they might be interestingly d ifferent. The resulting set of theories may be considered representative enough for our project to be systematic, but it will not be comprehensive. While many theories of causation started out relatively simple, most of them have been developed into fairl y elaborate and complex theories. Naturally, we will not be able to do justice to that complexity, and, in a way, we do not need to. Since our goal is to show how such theories may affect the exclusion problem and the responses to it, we need to explain th em enough for that impact to surface, but not beyond. It may well be that a closer look would reveal further trouble f or the problem or the responses; but the goal of this dissertation is more limited in scope, namely to put pressure on the silent assumpt ion that theories of causation do not matter for the exclusion problem. If we can establish that they do matter without further elaboration we should discuss more theories of causation and assess more responses. But before we do so, it will be useful to b riefly look at the relation between causation and causal explanations. 3.2 Causal Explanations Except for the distinctness assumption (1), all assumptions of the exclusion problem include causal assertions. But the assertions involve differ ent types of causal explanation The nonoverdetermination assumption (5) and the exclusion assumption (6) reference an adequate cause that would satisfactorily explain the occurrence of the effect. 1 Whenever we expla in the occurrence of an event by referring to a caus e of it, let us call such explanations basic causal 1 At first glance, it may not seem that the nonoverdetermination assumption (5) is doing this. But when we established the assumption, we explained what it means for events not to be systematically overdetermined, namely, that they do not typically have mor e than one adequate cause, be it another adequate cause or just an additional causal factor. Thus understood, (5) uses the notion of an adequate cause.
59 explanations. But the causal impact assumption (2) the typicality assumption (3) and the completeness assumption (4) refer to properties that explain why the cause effect relation holds in first place namely in virtue of properties had by the cause Let us call explanations of this type causal grounding explanations. In the context of the exclusion problem as a whole, both types of explanations may also be understood contrastively. For example, we may take (2) to emphasize that sometimes nonphysical events have a causal impact on physical events in virtue of their nonphysical properties of the kind k rather than in virtue of some properties not of the kind k whether they are themselves physical or n onphysical. Let us call explanations that specify alternative or contending causes or properties It is important to distinguish these types of explanations because they have different explananda and can be used to answer dif ferent kinds of causal question 2 In everyday life, we tend not to clearly articulate and carefully distinguish between different kinds of causal question Often, it is the epistemic interests o r the conversational context that determine what we are really asking and what type of causal explanation we expect and would accept as satisfactory. If we wish to give an honest answer to a causal question or an explanation of a causal problem we need to be sensitive to that context. Interestingly, Paul and may serve to illustrate the point For example, the maxim of relation prescribes that our assertions be relevant to the purpose of the conversation 3 In causal contexts, that may limit how far back in the causal chain leading to the effect in question a cited cause can be If we were asked why the friend mentioned in section 2 Lipton (2009, 624) points out that not all causes are explanatory with respect to all explananda. For example, our love of Broadway shows may causally explain why we went out to see a show, but it would not explain why we went to see Jumpers rather than Candide Merely citing a cause in response to a contrastive question often does not yield a satisfactory explanation. 3 See Grice (1989, 27).
60 2.1.1 is in the hospital, responding with a reference to the B ig B ang would vi olate the maxim of relation, even if the B ig B ang indeed were an adequate physical hospital. But in an ordinary conversation, we would expect the cause cited to be either a disease or an accident such as her falling down the stairs and breaking her ankle. The maxim s of quantity, to give another example, a sk us to make our contributions as informative as is required but not more informative than is required by the context 4 If we responded to the question why our friend is in the hospital by giving a microphysical description of the event of her falling down the stairs this description may again cite an adequate cause, at least if the completeness assumption (4) is true ; b ut even though we avoid the mistake of going too fa r back in the causal chain, we now violat e the second maxim of quantity as well as the maxim of relation by providing too much detail Why is this important? First, it sometimes may seem as if the causal explanandum say a particular physical event in different domains of discourse is the same But attention to the conversational context might reveal that to be illusory If nonphysical events sometimes have a causal impact on physical events in virtue of their nonphysical properties of the kind k rat her than in virtue of some properties not of the kind k physical or nonphysical we need to tread carefully whenever the explanandum seems the same because properties of the kind k may be cited in explanations of different special sciences than properties not of the kind k That is, they may be cited in different domains of discourse This may be particularly important when thinking about the nonoverdetermination assumption (5) and the exclusion assumption (6), for both presuppose a common explanandum, the physical event p 2 4 Ibid., p. 26.
61 Second, if in fact there is reason for thinking that the explan an dum is never or rarely ever exactly the same when we cite physical causes and events that have a causal impact in virtue of their nonphysical properties we may wonder w hether physical causes really are adequate, as posited by the completeness assumption (4), at least for those aspect s of the physical event that the nonphysical cause is meant to explain. 5 Alternatively, we may wonder whether the nature of the causal relat ion involved in ea ch explanation is the same because we may take the lack of matching or overlapping causal explananda as a reason to endorse causal pluralism. Finally if the conversational context is as important as suggested, we may wonder that whether or not a causal explanation is satisfactory depends too much on our linguistic and by extension normative practices. 6 More specifically, we may worry that this might turn the concept of an adequate cause which is used by several assumption s into something epistemic. As a reminder, an adequate cause is an event that, if it were cited in a causal explanation, would be all that one n eeds to cite as a causal factor operative at the time in order to provide a satisfactory explanation of why ano ther event occurs. And a satisfactory explanation is one that does not require any reference to further events or properties. But if the conversational context r ather than the causal situation determines whether or not references to further events or prope rties are required, how then can an adequate cause be something ontologically robust, i.e. be a relatum in a mind independent causal relation? 5 Different theories of events may require us to phrase the issue here differently. As we have suggested in section 2.1.2, a fine grained Kimean account may construe every aspect of a physical event as a distinct event, for it construes events as exemplifications of a property by an object at a time. The question then is whether adequate physical causes would satisfactorily explain the occurrence of those events that nonphysical events are meant to expla in. 6 Menzies (2009, 342) suggests that this is a serious worry, albeit not specifically in the context of satisfactory causal are contras tive, normative, and context sensitive in character, and that these elements do not seat easily with the
62 The worry may be alleviated in the following way T he reason why we recommended paying attention to the context is that communication is fragile. are meant to increase the likelihood of a successful communication. Communication can fail, for example, when too many assumptions are not explicated or when assertions, arguments, and questions are not art iculated clearly or phrased properly But arguably physics and the special sciences attempt to articulate their background assumptions as explicitly as they can certainly much more so than we ordinarily do They also utilize definitions, formalisms, math ematical models, and such to maximize the clarity of their assertions. To the extent that they do this, they fix what exact causa l question we are asking or should be interested in and what types of causal explanation s would count as satisfactory. But once the different types of explanations and questions are properly distinguished and recognized there are objective facts in the world that determine what an adequate cause of a given event is 7 Hence, while we should be cautious regarding the types of causa l explanation given and the types of causal question asked, we should not worry too much about the concept of an adequate cause and explanations that reference it. 3. 3 Theories of Causation Theories of causation are theories about the nature of causal rela ta and relations 8 Since we have settled for events as the appropriate causal relata as a matter of terminological convention in Chapter 2 we will focus on the nature of causal relations in this chapter. To 7 Maslen, Horgan, and Daly (2009) may deny this. They argue that, within a counterfactual analysis, it always depends on the explanatory context which possible worlds will qualify as relevant and which do not. Section 3.3.2 explains what is meant by this. 8 The following sections will discuss various accounts of what the conditions are for a causal relatio n to hold between events. It is not clear in what sense exactly the mere listing of such conditions amounts to a theory of causation. The notion of a theory is thus not a very rich one. It may not mean anything more than there being an account or view of s omething. Yet, it is customary in the literature to talk of theories of causation. We will simply adopt this custom without discussing what constitutes a theory.
63 motivate the order in which we will discuss the theories of causation, it will be useful to briefly distinguish between causal realism and antirealism causal reductionism and antireductionism as well as causal monism and pluralism 9 Causal realism assumes that there are genuine, mind independent causal relations in nature S uch causal relations would hold between events even if no beings with a mind were around referencing those relations in their explanations 10 Causal antirealism in its boldest form eliminativism, denies that causation is a genuine, mind independent relation Eliminativists typically point out that our best epistemic practices and institutions, such as physics and the special sciences, do not rely on or need not make use of the concept of causation and take that as one reason to eliminate it from our ontology altogether Causal reductionism views causation as a complex phenomenon that can be analyzed in terms of its constituent parts. While reductionism does not entail antirealism, some reductionis ts causation explanations, for it can be replaced by whatever they think the concept and phenomenon of causation reduces to and is identical with But to explain causal relations by reference to other entitie s or relations in our ontology is not the same as deny ing the reality of causal relations Thus, causal reductionists are not eliminativists. Most r eductionist acknowledge the central role that causation plays in many of our explanations and thus would be reluctant to eliminate the term from our explanations even if they consider it successfully reduced Some types of n onreduct ion ism assert that the concept of causation is simple in the sense that it cannot be broken down into further concepts. Nonreductive realists 9 The distinctions introduced in this paragraph are loosely based on Humphrey (2009, 632 336) 10 See Psillos (2009, 133f.).
64 may see this as one reason to view causation as essential to our explanatory practices and epistemic interests 11 Causal monism asserts that there is only one type of causal relation, but causal pluralism denies this. U nlike regular c ausal realists, pluralists assert that there is more than one type of causal relation. Often, these types are domain specific. For example, mental causation may be one type of causation, and physical causation may be another. In theory, causal pluralist s c an mix and mingle all previous stances. Some type of causal relation may be reductive whereas others are no t. Some type of causal relation may be real whereas others may just have an epistemic or pragmatic value. Other types of causal relation may not hold up to scrutiny and should be abandoned altogether. The pluralist stance will partly depend on the number of causal relations posited and the function assigned to each of them. We will first discuss theories of causation that are usually considered instanc es of reductive realism T hese are regularity theories, counterfactual analyses, INUS conditions probabilistic theories, tra ns ference theories and single difference theories We will then look at causal pl urali sm which depending on how it is fleshed ou t, could be reductive; but it does not have to be. Finally, we will have a brief look at eliminativism as a kind of antirealism. 3. 3 .1 Regularity Theories According to Stathis Psillos, regularity theories are reductive in that they try to explain causation in terms of spatiotemporal contiguity, success ion and regularity: c causes e if and only if i. c is spatiotemporally contiguous to e ; ii. e succeeds c in time; and 11 There are many types of reduction, so this paragraph simplifies things considerably. It still might suffice to hint at some key differences between various theories of causation. Humphrey (2009, 632 336) briefly walks throu gh some of the standard views of reduction in the context of causation. The reader may recruit his text for further details.
65 iii. all events of type C i.e. events that are like c in some relevant respect, are regularly followed by or are constantly conjoined with events of type E i.e. events that are like e in some relevant respect. 12 A nave regularity theor y 13 would treat (i), (ii), and (iii) as individually necessary and j ointly sufficient conditions for causation. If they are necessary, notice that (i) may put pressure on the causal impact assumption (2), at least for some nonphysical properties. It is not cl ear, for example, whether mental events such as ha ving beliefs an d desires are spatially extended and locatable and in what sense they can be spatiotemporally, rather than merely temporally, contiguous with other events 14 In fact, some nonmental event s may face similar problems. For example, consider the closing down of a business. On the one hand, it is locatable in the sense that it is registered in a particular city, state, or country. One the other hand, some businesses might not be tied to a particular building, for the event mostly has to do with rev oking ownership. We might say that the closing down takes place wherever an official processes the legal documents ; or we might say that this just means that, while the event has a location, its boundaries are fuzzy. But either response seems somewhat arbi trary. 15 12 See Psillos (2009, 131). While Psillos phrases (i) and (ii) as listed, (iii) is slightly modified, so this is not a direct quote. On a side note, discussions of regularity theories usually start with Hume. Often, such theories are called preliminary definition is reductionist as it is. Sklar (2009) reads Hume specifically as an eliminativist. And according to Garrett (2009, 73 74), projectivists hold that, while a felt necessity is merely a feeling of the determination of the mind that then gets projected into the world, the neces sity is nonetheless an essential part of the concept of causation, which charges the concept of causation epistemically. But it is not clear whether Hume endorsed any of the above views. From the fact that we merely observe regularities but generate an imp ression of necessity in our mind, it does not follow that causal relations in nature do not possess features similar to necessity or sufficiency; see Garrett (2009, 89). To avoid a possibly misleading misnomer, we will avoid calling regularity 13 Armstrong (1983, 11) calls it that. 14 in section 1.1. However, the issue may arise even if one does not endorse a Cartesian substan ce dualism. 15 I thank Gene Witmer for this point and example.
66 Furthermore, a nave regularity theory would deny that the regularity expressed in (iii) needs further explanation, for example in the form of laws of nature that might explain why the regularit ies hold in the first place. 16 Of course, this denial does not suggest that there are no laws of nature. But such laws would merely express the regularities rather than ground and explain them. 17 Typically, the regularity referenced in (iii) is thought to be one without exception. If there is one event of type C that is not followed by an event of type E then c is not the cause of e This has interesting consequences for the typicality assumption (3) and the nonoverdetermination assumption (5 ) Recall that Stephen version of the exclusion argument in section 2.1.2 offer ed a very strong exclu sion assumption, one that merges exclusion with nonoverdetermination. His version d oes not allow for even a single case of overdetermination. 18 Many philosophers including Karen Bennett, think that that is too strong; i solated instances of overdetermination may be conceptually and perhaps metaphysically possible But in the context of the exclusion problem only systematically overdetermined effects should concern us which is wh at we tried to capture in (5). In part, this was what motivated the introduction of the typicality assumption (3). Nonphysical events that have a causal impact on physical events in virtue of the nonphysical properties would be threatened by exclusion qua (5) only if they are representative of their kind k i.e. only if they are typically strictly involved in the production of effects in virtue of properties of the kind k ; otherwise their causal impact would not imply systematic 16 See Beebee (2006, 510) and Psillos (2009, 132). 17 See Armstrong (1983, 11). 18 He is not alone. Bunzl (1979, 146) denies that some of the alleged standard cases are genuine overdetermination cases and doubts whether there could be some. See Kim (1989/1993, 250 254) for the claim that they are improbable. See Schaffer (2003a) for a criticism of Bunzl and Kim.
67 overdetermination. We thus restricted the scope o f the exclusion problem to nonphysical events that have such representative causal impact This restriction however, a llow s for there being atypical nonphysical events that only sometimes overdetermine effects in virtue of t heir nonphysical properties. After all, the distinctness assumption (1) merely states that, if a nonphysical property is of the kind k then it is distinct from any physical property. Since this is merely an implication and not a biconditional, (1) allows for nonphysical properties that are not of the kind k and yet distinct from physical properties. Some events may have a causal impact in virtue of such nonphysical properties of the non k kind say of the kind l without that impact being typical. But if t he nave regularity theory is correct, such atypical causes could not exist. T he assertion that c causes e in virtue of its nonphysical property of the kind l in cases where typically events of type C l are not followed by events of type E would imply that (iii) is false and consequently, that c is not the cause of e So, while (3) on its own is compatible with there being atypical nonphysical causes that overdetermine events that have adequate physical causes, (3) in combination with the nave regula rity theory captures all events that may have a causal impact in virtue of their nonphysical properties. In fact, we could just omit (3) for the nave regularity theory would guarantee typicality. By the same token, the falsity of any theory of causation t hat assumes or implies exceptionless regularities threatens the typicality assumption (3) of the exclusion problem. I f we accept the nave regularity theory the exclusion problem thus seems particularly severe, for all nonphysical events would be threaten ed by causal ex clusion. The problem could be turned into a deductive argument for instance by using the causal impact assu mption (2) to set up a reductio ad absurdum that results in the negation of (2). We may try to cast doubt on the validity by driving a wedge between the typicality of the causal impact and the
68 systematicity of the overdetermination. Perhaps instances of nonphysical events that have a causal impact are typical but rare enough not to amount to systematic overdetermination. But that seems confused Consider a standard overdetermination case such as two assassins that attempt to kill the same person. They pull the trigger of their respective rifles, both bullets hit the victim in a deadly manner, and as a result the person dies H owever, if the bullets hit the victim at different time s this may not qualify as genuine overdetermination for the earlier bullet may explain why the victim died an instant earlier. Similarly, if the bullets hit the victim at different places of his or her body, each shot may contri bute to the death in different ways which may not amount to overdetermination either. 19 For the assassina tion to be a genuine case of overdetermination, both bullets need to hit the victim at a time in such a way that their effect s contribute to the death of the victim in the exact same manner Such overdetermination cases may be very rare, but that is not wh at makes them lack systematicity. Rather, it is the fact that it is a mere coincidence that they happen to have the exact same effect, 20 for it is in virtue of the circumstances rather than in virtue of the properties of the causes that they overdetermine t he death of the victim. Increasing the pre valence of overdetermination by frequently recreating similar circumstances would not lead to systematic overdetermination. If, by contrast we found that whenever two assassins try to shoot a person the bullets hi t the victim at the same time in the same way, i.e. if there is little to no variability in the effects in spite of varying circumstances that would b e surprising exactly because it suggest s that there is something about the cause s that 19 grained overdetermination, effect occurs in precisely the same way regardless of which bullet hits the victim. Requiring such fine grained o verdetermination is much more natural in a Kimean account of events than in a Davidsonian one; see section 2.1.1 for details. 20 This point is inspired by Sider (2003, 722), even though he discusses coincidence and systematic overdetermination in a differen t context.
69 explains the overdetermination. That is what makes the overdetermination systematic and t hat is what nonphysical causation would look like according to the typicality assumption (3) and the completeness assumption (4). Every effect that has a nonphysical cause also h as an adequate physical cause. Consequently, all nonphysical causes that have a causal impact in virtue of their nonphysical properties are threatened by exclusion by way of the nonoverdetermination assumption (5) However, it seems difficult to support ( 5) if the nave regularity is true. So perhaps we rejecting (5). Armstrong and others have argued that the theory is bound to treat many regularities as laws of nat ure and as causal relations that intuitively are not For instance, there seems to be a correlation between per capita cheese consumption a nd the number of people who die by becoming tangled in their bedsheets 21 But it would be odd if there were an actual law these types of events fell under. Most likely, the correlation is s purious and purely coincidental. If Armstrong is right and the nave regularity theory is ontologically profligate it is difficult to see what reason s we could have for endorsing the n onoverde termination assumption (5). Systematic overdetermination by way of nonphysical events may yield an odd picture of reality, but it is not odder than an overabundance of laws of nature and causal relations. In fact, if nothing grounds regularities and if laws of nature supervene on those regularities, it is rather unclear what is odd about systematic overdetermination to begin with. Under the nave regularity theory, an event e being overdetermined implie s that there are at least two events, c and d that are spatiotemporally contiguous with e that precede e and 21 The example merely serves to illustrate the problem, for the correlation only covers 94.71 percent of the data entries. To satisfy (iii), we would need a perfect correlation. The example is drawn from the amusing collection of spurious correla tions offered by Vigen (2015).
70 for which there are laws they, together with e fall under. [L 1 ] such that events of the type C (where c is one of them) are fol lowed by events of type E (where e 22 and there is a law L 2 such that events of the type D (where d is one of them) are followed by events of type E (where e is one of them). But what exactly is odd about there always being two causal laws L 1 and L 2 ? One answer may be this: especially if one of these causes, c and by extension one of these causal laws, L 1 is ontologically privileged by way of the completeness assumption (4), d and L 2 seem epistemically redundant 23 for t not already explained by c and L 1 We may respond by apply ing that L 2 is a genuine causal law O r we may accept that it is a law that express es said regularity but deny it is als o a causal law T he regularity expressed by L 2 might supervene on the one expressed by L 1 If causes are events that bring about or produce effects, as we have suggested when we this solution may seem appealing at first But the nave regularity theo ry lacks the resources to draw the needed distinction in this case. While it can distinguish between laws that are causal and those that are not, it can do so only when the event s in question are not spatiotemporally contiguous or when e does not succeed d Since the relation between d and e satisfies (i), (ii), an d (iii) L 2 qualifies as a causal law We are left with overdetermination, or so it seems. However, the regularities that L 1 and L 2 express are not grounded in anything else In some sense, citing these laws does not explain the occurrence of e at least not in a way that causal production or necessitation would. But then it remains unclear why overdetermination is problematic, even when it is systematic. All it means is that two distinct events of certain types regularly precede effects of a certain type. 22 Psillos (2009, 146). 23 Sider (2003) discusses this epistemic argument against overdetermination in addition to a metaphysical objection and a coincidence objection.
71 To summarize: applying the nave regularity theory to the exclusion problem has two interesting results First, all nonphysical events that allegedly have a causal impact in virtue of their nonphysical properties are threatened by exclusion. I f overdetermination is a problem for nonphysical events, it always is Second, w hat exactly is probl ematic about overdetermination remains a mystery. When assessing whic h assumption of the exclusion problem should be rejected, assuming that (1) to (6) are in fact inconsistent, the nonoverdetermination assumption (5) may thus seem like t he best candidate To avoid an overly profligate nave regularity theory we can eithe r ground (iii) in something else or we can add further necessary condition s to (i) to (iii) Pursuing the first strategy, we may offer a thick notion of the laws of nature that might allow us to distinguish between accidental and nonaccidental regularities. 24 The Ramsey Lewis view on laws of nature may provide such a thick notion According to this view, regularities can be laws only if properly embedded as axioms in a deductive system i.e. a particular kind of scientific theory that balances simplicity and strength best The best candidates for nonaccidental regularities then are those that can be subsumed under a law that improves the quality of th e deductive system. 25 Within such systems, we may also have the resources to distinguish betwee n laws that are causal and those that are not, to privilege some laws over others, and to thus support the nonov erdetermination assumption (5). Still, it seems possible that regularities that lead to systematic overdetermination may be among those that imp rove the quality of the deductive 24 Strawson interacting in seemingly perfectly law g overned ways 25 See Psillos (2009, 142) for a brief sketch and Ramsey (1928/1978) and L ewis (1973b) for the fully developed view. Psillos also addresses the point that the Ramsey Lewis view charges the notion of laws of nature epistemically.
72 system. Since regularities and laws are grounded by the occurrence of events rather than vice versa, it seems rather unclear what principled reason we could have to reject that possibility. To the extent that we do not hav e such reason, we should not be particularly invested in the nonoverdetermination assumption (5). Alternatively, w e may declare laws of nature to be fundamental, not further analyzable items in our ontology that would allow us to distinguish between acci dental and nonaccidental or causal and noncausal types of regularity 26 While this may seem rather ad hoc for we could have postulated two distinct types of regularity instead, it would at least lend more plausibility to the completeness assumption (4) if the only laws that do not supervene on other laws happen to be physical laws. Pursuing the second strategy, we may add further necessary criteria to the causal biconditional by fusing the regularity theory with other theories of causation such as the counterfactual analysis of causation. 3. 3 2 Counterfactual Theories The best known counterfactual theory of causation is the one developed by David Lewis It shares two important features with the regularity theory discussed in section 3.3.1. First, it is reductive because it explains causation in terms of another relation, in this case counterfactual dependence rather than regularity. 27 Second, it is a realist theory, for the counterfactual dependence is based on facts about and relations between possible w orlds, which Lewis construes as entities that are as real as our actual world 28 Lewis himself presents his 26 See, for example, Maudlin (2007, 17 18). 27 Not all counterfactual theories are reductive, according to Paul (2009, 164). 28 See Lewis (1986) for a defense of this modal realism. However, the counterfactual analysis of causation does not rely on this type of realism. Plantinga (1976), for example, construes possible worlds as abstract states of affairs th at exist but do not obtain.
73 counterfactual analysis as an alternati ve to regularity theories of causation even to those that are more advanced tha n the nave regularity theory we discussed in the previous section 29 So he is not interested in grounding (iii) or in adding further necessary conditions to (i) (ii), and (iii) of the regularity theory. However, Helen Beebee has convincingly argued that his counterfactual analysis ultimately rests on an analysis of counterfactuals and laws of nature which in turn depend on regularities. It thus may be best understood as a sophisticated version of the regularity theory. 30 Let us now look at the analysis itself. Recall our friend fro m sections 2.1.1 and 3.2 who went to the hospital because she broke her ankle when she f ell down the stairs. The first thing to notice is that, for the breaking of her ankle to be a cause of her visiting the hospital, there needs to be a causal chain conne cting the two events. I f our friend b roke her ankle but treated it herself and went to the hospital to visit her brother the causal explanation would be false. If and when there is such a causal chain connecting the two events, it seems natural to say tha t her visiting the hospital causally depends on her falling down the stairs After all, had she not fallen down the stairs and broken her ankle she would not have visited the hospital unless some other accident had happened instead. Lewis thinks that we should take this ordina ry way of talking about causal dependence very seriously and offers his c ounterfactual analysis as a way of capturing it. 31 T o say that event c caused event e is best understood as to mean that, if event c had not occurred, event e would not have occurred. But this leaves it open whether causation explai ns the counterfactual dependence or vice versa. Lewis argues for the latter. What makes such counterfactual s true are not causal 29 30 See Beebee (2006, 511f.) for a defense of th is claim. 31 See Lewis (1973a, 557f.).
74 relations; rather, it is the degree of similarity bet ween relevant possible worlds where the degree of similarity is taken to be a primitive relation 32 P ossible worlds are ways our world might have been 33 To say that something was possible, even though it did not actually happen, is to say that there is at least one possible world in which it did happen To say that something happened necessarily is to say that it happened in all possible worlds. One problem is that f ew things if any, are metaphysically necessary in the strong sense that they hold in all p ossible worlds. Even the laws of nature are typically thought to be contingent, i.e. they could have been different. But that is just to say that there are possible wo rlds in which those laws do not hold. It then may seem difficult to analyze causation in terms of counterfactual dependence For example, there may be worlds in which broken ankles heal instantly. But in such worlds our friend probably would not have gone to the hospital after fa lling down the stairs. Similarly, there may be worlds in which bones break all the time so that people go to the hospital at least once a day. In such worlds, it would not be true that, had our friend not broken her ankle, she would not have gone to the ho spital. Since most counterfactuals will thus turn out to be false unless we specify the relevant worlds that are to be compared to our w orld, Lewis suggests a counterfactual is nonvacuously true iff it takes less of a departure from actuality to make the consequent true along with the antecedent than it does to make the 34 32 That is, one that is not further analyzable; see Lewis (1973a, 559). Interestingly, Lewis begins his paper with a brief discussion of problems with the regularity theory. But, as just noted, Beebee (2006) claims that counterfactual theory of causation can be understood as a sophisticated version of the regularity theory. 33 While Lewis (1973a) presents the basic counterfactual analysis, illuminating additional points can be found scattered throughout his work. Psillos (2002, 92ff.) and Horwich (1987/1993) both highlight some of these points own writing. 34 Lewis (1973a, 560).
75 T he above yields the following analysis of causation: i. c is a cause of e if and only if i.i e depends causally on c or 35 i.ii there is cau sal chain leading from c to e 36 ii e causally depends on c if and only if c and e occur red and e counterfactually depends on c iii e counterfactually depends on c if and only if iii .i if c occurred in the relevant worlds, e occur red as well, and iii .ii had c not occurred in the relevant worlds, e would not have occurred. iii .iii the relevant worlds in ( iii .i ) and ( iii .ii ) are those in which c and e occurred and that are closer to the actual world than worlds in which c occurred but not e 37 To compare possible worlds with respect to their relative closeness to the actual world, we need to weigh the following maxims against each other: (a) avoid widespread violations of laws of nature, (b) maximize t he spatiotemporal region with perfect ly match in g facts (c) avoid small, localized violations of laws of nature, and (d) it is not very important to secure an approximate similarity of particular facts 38 The earlier maxims are supposed to weigh more than the latter ones. Worlds in which the laws are ve ry different from ours tend to be more remote from ours than worlds in which only some local facts are different. A world in which bones heal instantly would be more remote than a world in which our friend ends up not going to the hospital merely because h er car is out of gas, for example. All key components counterfactual analysis are now on the table We can start applying it to the exclusion problem. 35 We may be tempted to think that causal dependence is necessary for causation, not merely sufficient. But, as Lewis (1973a, 563) points out, causation is usually considered transitive whereas causal dependence is not because counterfactual dependence is no t. Thus, causation does not imply causal dependence. 36 Ibid., p. 563. 37 This is a simplification, for Lewis phrases this in terms of propositions. See Lewis (1973a, 560f.) for a more careful way to spell this out. 38 See Lewis (1979, 472). Lewis (1973, 56 3) explains the notion of a causal chain by way of causal dependence: c d e d depends causally on c e on d and so on ch causal chains are transitive, not all causal dependence is.
76 One standard object ion to the counterfactual analysis of causation is that it has trouble handling some overdetermination cases. Consider the assassin case mentioned in section 3.3 1. independent causes c and d that satisfy ( iii .i) because they each lead to the death of the victim e If e counterfactually depends on c and d this implies that had c not occurred, e would not have occurred, and that, had d not occurred, e would not have occurred. But that seems false Had c not occurred, d still would have caused e ; and had d not occurred, c would have caused e Thus, e counterfactually depends neither on c nor d True there are two causal chains, one that connects c and e and one that connects d and e such that (i.ii) is satisfied and c and d still q ualify as causes. However, if we look at the events in each chain that occur right before the effect such ain, then the analysis might yield the conclusion that neither one of these events is a cause because neither (i.i) nor (i.ii) is satisfied. This may be a sign that the analysis is incomplete at best, but it may ultimately fail 39 While the counterfactual analysis may be able to handle some overdetermination cases and not others it remains uninformative regarding systematic overdetermination The nonoverdetermination assumption (5) should not be endorsed or rejected based on it. In that respect, it may differ at least from the nave regularity theory unless Beebee is corr ect when she states that the counterfactual analysis ultimately rests on a regularity theory of sorts, in which case our attitude towards (5) should not differ significantly. 39 problems with this simple account as well as of various ways the account can be developed further. See pp. 177 182 for a discussion of how counterfactual theories may deal with overdetermination cases. Much more could be said here, but the point here is merely to explain how the counterfactual theory may affect the exclusion problem. It may thus suffice to merely indicate that the analysis may have problems with some type of overdetermination cases,
77 Some have argued that the counterfactual analysis of causation can be used to re ject the completeness assumption (4) and, by extension, the exclusion assumption (6). Supposedly, there are cases in which the better candidate for the property that e occurrence counterfactually depends on is a nonphysical rather than a physical propert y. 40 Suppose that our friend was in severe pain after she fell down the stairs and br oke her ankle and we want to explain why she it hurt terribly Assuming that her pain is realized by the neuro nal activit y in her brain, we should also consider th e neuronal activity as a potential cause, especially with the completeness assumption (4) in mind If it is true that her being in pain caused her to go to the hospital, then, according to ( iii .ii) it is true that, had she not been in pain, she would not have gone to the hospital. 41 By the same token, if it is true that her neuronal activity caused her to go to the hospital assuming it is physical in nature to satisfy (4) then it is true that, had she not exhibited that neuronal activity, she would not have gone to the hospital. To avoid overdetermination, we need to decide which of these counterfactuals designate s an actual Either the physical cause is a dequate, in which case attributing actual causal impact to the pain implies overdetermination, or the physical cause cited in (4) is not adequate, i.e. (4) is false, in which case the pain might have causal impact. Thus, we need to determine which of the counterfactuals describes a world closer to ours. 40 Zhong (2011, 130) reads Horgan (1989), Horgan (1997), and Baker (1993) as defending this strategy. 41 Strictly speaking, this is too quick. Since c is a cause of e if and only if (i) or (ii), it could be the case that the pain is a cause of her going to the hospital merely by being part of a causal chain that leads to t hat. We cannot simply assume that her going to the hospital counterfactually depends on the pain. We will get back to this point in a bit.
78 While pain may be realized by neuronal activity in the brain, such pain can be expected to be multiply realizable. 42 The same kind of pain might have been realized by slightly different neurona l activity Since securing approximate similarity of particular facts is not very important by comparison to more extended spatiotemporally mismatching regions, a world in which our friend does not exhibit the particular neuronal activity and does not go t o the hospital seems more remote than a world in which our friend exhibits a slightly different neuronal activity and still goes to the hospital. But a world in which our friend does not feel pain and does not go to the hospital will arguably be closer to yet still goes to the hospital This indicates, first, that depends on her being in pain and perhaps, second, that her being in pain is a better candidate for being a cause than a particular neuronal activity. 43 While the ordering of possible worlds here involves conjecture and may be affected by our epistemic interests, this may at least indicate that a counterfactual theory of causation might make it easier to deny the completeness assumption (4). Let us generalize: there may be cases where the better candidate for a cause of a physical event is the nonphysical event that has a causal impact on the event in virtue of its nonphysi cal property. If that is so, the exclusion proximity to our actual world prevent s seemingly competing causes from actually being overdetermining quite simply because the completeness assumption (4) is false However, this strategy faces two problems. First, it may be question begging. Recall that we simply assumed 42 Putnam (1967) points out that organisms with quite diverse brain structures seem capable of experiencing pain. F urthermore, it is likely that even within one organism there will be slight differences in the brain areas that realize pain. Against that background, it seems reasonable to assume that pain is multiply realizable. 43 At least that is what Zhong (2011, 13 4) argues.
79 c is a cause of e if and only if (i) e depends causally on c or (ii) there is a causal chain leading from c to e this is too quick. For if being in pain were a cause merely in virtue of being part of a causal chain leading to the hospital visit, we would not have compared it to the adequate physical causes by way assessing the proximity of possible worlds. Second, the strategy may not appreciate what adequate physical causes entail. Arguably, the part that realizes the pain likely will be only a partial cause that contributes to the effect without being adequate itself. Adequate causes of p hysical events may often be extreme ly complex, for they must explain the occurrence of the effect in such a way that no reference to other events and properties is required. But the more complex they are, the more factors there are to keep in mind in figur ing out which world is close s t. So, unless we have a clearer idea of what being an adequate cause typically entails, we candidates for causes. 44 Still, w e may flag this and keep in mind that a counterfactual analysis may have an impact on how likely it is that we hold onto t he completeness assumption (4). 3.3.3 INUS Conditions According to r egularity theories and counterfactual theories, what counts as cause and effect depends on facts that go beyond the relation between these events themselves. Regularity theories construe causation as a relation that depends on how other events of the same type relate. C ounterfactual theories capture th e modality involved in causation by way of the similarity between and the proximity of possible worlds By contrast, INUS conditions locate the modality in proximity to the causal relation itself. 44 proximity of the possible worlds to the actual one as a feature of the world, they argue that the ordering of the worlds is always context ual. In some explanatory contexts, it may be in virtue of nonphysical properties that c causes e as Zhong claimed, whereas in others it may not be.
80 The basic idea of INUS conditions is best conveyed by way o f an example. Suppose a short circuit caused a fire in a house. On the one hand, the short circuit was not necessary for the fire H ad this particular short circuit not happened, another one might have happened shortly after. On the other hand, the short c ircuit on its own was not sufficient either, for without the presence of a complex set of conditions such as inflammable material, oxygen, and so forth, there would not have been a fire. So, the short circuit on its own was neither necessary nor sufficient for the fire. But together with these other conditions, the short circuit was sufficient for the fire. While the conditions and the short circuit were not necessary for the fire because something else might have caused it, the short circuit was a necessar y part of the conditions, for the other conditions on their own would not have produced the fire. Thus, the short 45 fire. We can express this i dea more schematically as follows : If c is a cause of e then c is an INUS condition of e i.e. c is an insufficient but necessary part of a condition which is itself unnecessary but exclusively sufficient for e 46 However, which conditions are relevant d epends on what exactly it is that the cause is supposed to explain. Often, the explanatory context of causal statements will illuminate this further. When we ask why the house caught fire, we may be interested in different things. As we have pointed out in section 3.2, such a question may be understood contrastively. Perhaps we know that other houses in the neighborhood have circuitry of the same type installed, and yet they have not caught fire. Or perhaps we know that the circuitry in this particular hous e was malfunctioning for some time already, so that we wonder why the house caught fire now rather than at an earlier 45 Mackie (1965, 245). The short circuit example is drawn from Mackie as well. 46 This is a slightly modified
81 time. In the former context, the conditions will include facts about houses with the same circuitry. In the latter context, the conditions will include historical facts about the cond ition of this particular circuitry. Mackie calls these regions in which causes and effects are assessed causal fields and includes them in his definition : If c is a cause of e relative to a particular causa l field F then, given the features that characterize F c is an INUS condition of e i.e. c is an insufficient but necessary part of a condition which is itself unnecessary but exclusively sufficient for e 47 The first thing to notice here is that this ac count, like the counterfactual theories of causation, has trouble handling overdetermination cases. Consider the assassination case again. conditions relevant for the causal field in question But when both pull the trigger, neither is a necessary part of the condition which is itself s death. The analysis th en implies that neither trigger 48 This is, of course, consistent with the nonoverdetermination assumption (5). However, if overdetermination is impossible in general it is not merely systematic overdetermination we have to d eny. As a consequence, atypical nonphysical events of the sort not captured by the causal impact assumption (2) could not have a causal impact. In that respect, INUS conditions seem to yield results similar to the nave regularity theory we discussed in se ction 3.3.1. However, the nave regu larity theory rejects atypical causes because they would violate the regularity requirement INUS conditions reject them because the concept of atypically overdetermining causes is as incoherent as any 47 See Mackie (1965, 249). As before, this is a slightly modified version to match our previous notation. 48 Only the disjunction of both trigger pulls is necessary
82 overdetermination. Counterfactual theories, on the other hand, have been able to handle overdetermination cases that involved intermediate events. Thus, they may allow for events that have a causal impact in virtue of their atypical nonphysical properties of the kind l that are not They face exclusion in cases only where overdetermination does not involve intermediate events. A second problem may be this. In cases where nonphysical events supposedly have a causal impact in virtue of their nonphysical properties of the kind k they supposedly overdetermine the physical effects that supposedly already have an adequate physical cause. But r of the alleged causes is in fact a cause While the nonphysical event may be excluded from having a causal impact in virtue of its nonphysical properties, i.e. while (2) and (3) may turn out to be false, some physical events too would be excluded, namel y those that causally explain the same features of the effect that the nonphysical properties would explain. But now this threatens the adequacy of physical causes in (4) and (6) and, by extension, the truth of those assumptions. However, proponents of an analysis of causation in terms of INUS conditions may avert such grave consequence by pointing out that, in all likelihood, the causal fields involved in explanations citing nonphysical causes and properties will differ from those that cite physical cause s and properties. If so, both physical and nonphysical events may qualify as a cause of the same event, but relative to different causal fields. As a result, there is no genuine overdetermination and the nonphysical cause is not threatened by exclusion. T his response is only convincing, though, if the adequacy of a cause must be relative to the causal fields as well. That requires some justification. Still, INUS conditions help to substantiate the worry we voiced
83 in section 3.2 that our epistemic and conve rsational context may determine what type of causal explanation we expect or would accept as satisfactory could be substantiate. 3. 3. 4 Probabilistic Theories P robabilistic theories of causation treat causes that necessitate their effects as special cases of a more general relation, namely as events that raise the probability of the effect occurrence to 1 In general, such theories merely state that c auses make a difference to or influence the probability of their effects. Patrick Suppes expresses that idea in the following way: An event c is a prima facie cause of an event e if and only if i. c occurs earlier than e and ii. the conditional probability of e occurring when c occurs is greater t han the unconditional probability of e occurring. 49 As mentioned in section 2.2., we may wonder whether a positive influence as required by (ii) is needed. Perhaps causes sometimes negatively affect the probability of an effect. 50 For example, we may assert that taking contraceptive pills lowe rs the likelihood of a pregnancy While this sounds like a natural thing to say, we may also worry. After all, if the pregnancy occurs, the contraceptive pills were not among its causes R ather, we think the pills had a causal impact on the prevention of the p regnancy, albeit not sufficiently so. For the purposes of this project, w e (ii) can easily be rephrased so that it can accommodate both the view that causes typically raise probabilities and the view that they sometimes lower probabilities : ii.* if an event c is the cause of another event e the probability of e occurring when c occurs is different from the unconditional probability of e occurring. This would capture that causes make a difference, regardless of whether they raise or lower the probability of an event to occur. 49 particular events. 50 Rosen (1978, 605 606) seems to think this is not unreasonable.
84 Even though this is barely even a sketch of a probabilistic theory, there are already interesting implications for the exclus ion problem, especially with respect to the nonoverdetermination assumption (5). Recall the overdetermination case of two assassins A and B, who pull the trigger s of their respective rifles so that the bullets hit the victim at the same time in exact ly th e same way Intuitively, it is correct to say that both assassins killed the victim. But consider how awkward the probabilities must be for this to qualify as a genuine case of overdetermination. Suppose that the probability of the victim from bein g hit by one of the bullets is 0.7 whereas the probability of the victim at the same time without being hit by any of the bullets, say by a heart attack instead, is 0.000 00001. If only assassin A took a shot, A qualify as a cause of the vict The trigger and the conditional 0.7, is different from the unconditional prob ability of the death, 0.00000001. (i) and (ii) of the causal biconditional wou ld be satisfied and the trigger pulling turns out to be a cause according the probabilistic analysis. In section 3.3.1 we pointed out that genuine overdetermination cases require that the effect brought about by each cause is the same in all its features 51 We thus should assume that the probability B pulls the trigger in the absence of A would be 0.7 as well. The analysis applies again and B pulling in the absence of A is a cause as well 51 ent occurs is always specified; see, for example, Suppes (1970, 12). Consequently, it would be impossible to have a genuine case of overdetermination where c and d would cause the event to appear at slightly different times in the absence of the other caus e, say, if the second shot killed the victim a split second later. So, the analysis requires temporal identity of the effects of c and d account of events.
85 However, overdetermination cases are supposedly different from regular joint causes. In the case of regular joint causes, even i f we assumed that P( e c ) = 0.7 and P( e d ) = 0.7, the joint probabili ty of e given c and d would be different than the probability given just c or d For example, we might get P( e c ,d ) = 0.9 52 Of course, we can not just add up the probabilities. I f the victim is already likely to bleed to death from one shot, the second shot might just lead to an increased blood loss which makes it more likely that the victim dies, but not necessarily drastically more likely. So what distinguishes a genuine overdetermination case from a regular case of joint causes? Answer: t he fact that the effect needs to be exactly the same. It looks like we have to say that the probability of e given c and d must equal the probability of e given either c or d 53 But that is problematic for now neither (ii) nor (ii*) is satisfied for c and d The conditional probability of e occurring when c occurs is no different from the probability when c does not occur: P( e c ,d ) = P( e d ) Therefore c is not a cause of e And the conditional probability of e occurring when d occurs is no different from the probability when d does not occur: P( e c ,d ) = P( e c ). Therefore d is not a cause of e Consequently, the simple p robabilistic theory above, at least on its own, has something in common with the simple counterfactual analysis in section 3.3.2: it has difficulties handling overdetermination cases While the probabilistic analysis like the INUS conditions discussed in section 3.3.3, can endorse (5) in that the ana lysis agrees with the claim that there is no systematic overdetermination, it does so because the alleged causes involved turn out not to be ca u ses at all. 52 Of course, this is very sketchy. See Pearl (2000) for one of the most detailed a ccounts of how to model probabilistic causation and how to calculate the impact of causes especially in Bayesian networks. 53 we should ask is why the presence of an additional, overdetermining cause would not change the probabilities of the
86 In that respect, it seems stronger than the counterfactual analysis discussed in section 3.3.2., whi ch could provide a convincing analysis of at least some overdete rmination cases and which seemed otherwise neutral with respect t o systematic overdetermination. 54 But like INUS conditions the probabilistic analysis has the further interesting consequence t hat it may cause trouble for the completeness assumption (4) As a reminder, i n cases where nonphysical events supposedly have a causal impact in virtue of their nonphysical properties of the kind k they supposedly overdetermine the physical events that supposedly already have an adequate physical cause. But in genuine overdetermination cases, the simple probabilistic analysis concludes that neither of the alleged causes are in fact causes. While the nonphysical event may be excluded from having a causa l impact in virtue of its nonphysical properties, i.e. while (2) and (3) may turn out to be false, some physical events would also be excluded, namely those that causally explain the same features of the effect that the nonphysical properties would explain. But now this threatens like INUS conditions the adequacy of physical causes in (4). As in the case of the counterfactual anal y sis of overdetermination by two events with no intermediate events it seems strange to think that the mere presence of another event that is a cause undermines the status of an event as a cause. It then seems natural to search for features intrinsic to th e causal relations that would explain why c and d are causes after all. 55 While process and transference theories do not have to be probabilistic, they can be and have been combined with the latter kind of theory and may be able to circumvent this issue, a s w e will see in the next section. If that were not an option, we might be well advised not to apply the simple 54 Unless Beebee is right in thinking that counterfactual theories ultimately rest on regularity theories of some sort. 55 We will get back to the point in the context of causal pluralism in section 3.3.7.
87 probabilistic theory of causation to the exclusion problem. If a theory of causation leads us to reject not only (2) and (3) but also (4) on an alytic grounds rather than on empirical ones, the best explanation may be that there is something wrong with the causal analysis as it stands Given that we merely provided a sketch of the basic idea underlying probabilistic theories, and given that most p robabilistic theories are rather sophisticated, this does not seem particularly surprising. Alternatively, we could just accept that in the context of the exclusion problem probabilistic theories imply that there are never nonphysical events that cause and overdetermine physical events, i.e., the threat of exclusion is very serious. 3.3. 5 Transference Theories Like process theories, transference theories are usually reductionist Some of the former theories explain causation in terms of processes and interactions; some of the latter theories explain it in terms of quantities such as momentum transferred from one object to another. 56 The basic idea may be construed as follows: An event c causes an event e if and only if i. e involves a change in an object, ii. at the time e occurs, the object the change in which constitutes the cause of e is in contact with the object that undergoes the change which constitutes e and iii. p rior to the time of the occurrence of e the body that makes contact with the effect object possesses a quantity which is transferred to the effect object when contact is made and manifested in e 57 56 (1984), for it connects well to our discussion of probabi listic theories and promises to overcome at least some of the discussion of his view and an application to the exclusion problem rather unnec essarily complicated and difficult. 57 This way of presenting the general idea of the transference theory is inspired by Aronson (1971, 421f.) but distinguishes between natural an d unnatural changes, where natural changes are internal changes in an object which he does not regard as instances of causation. Only external changes qualify.
88 We can easily make sense of (i) within a Kimean account of events, accord ing to which events are exemplifications of a property by an object at a time. 58 However, the account itself leaves it open whether the property exemplified needs to be distinct from properties previously exemplified by the object. (i) seems to require such distinctness ; at least it seems natural to understand change in that way (ii) is basically capturing the contiguity requirement we already encountered in the context of the nave regularity theories in section 3.3.1 59 And (iii) specifies that the quantity transfer happens between the objects constituting the events that are cause and effect, at least on a Kimean reading. Regardless of whether this kind of theory is ultimately defensible or not 60 it may appeal because i Paradigm cases of causation such as two billiard balls hitting each other usually involve physical contact. Furthermore, the way we describe such cases s uggests that we are often aware that phys ical quantities matter. For example, to explain why the red five had a certain impact on the black eight when hitting it, we may specify the speed of the red ball presumably because we are aware that the velocity of an object is related to its kinetic ene rgy which in turn is needed to explain the impact the red ball has. Thus, it should not surprise us that c ausation requires contact and involves change and the transfer of some quantity Plausible enough. W hen apply ing it to the exclusion problem, however, problems arise quickly 58 See section 2.1.1. 59 If causation consists in the transmission and transference of a quantity, contiguity seems particularly important. However, we may think that this is a disadvantage. In fact, Hall (2009, 168f.) considers it an advantage of counterfactual theories that they can handle situations where there is no process, no transfer of energy, or in general no contiguity. 60
89 The examples of transferred qu antities are usually physical : kinetic energy, momentum, velocity, and so forth. If they are, we can expect the physical cause in the completeness assumption (4) to account for all of them. It is hard to see how else it could be adequate. In that sense, the theory might even lend support to (4). But now consider the causal impact assumption (2). Supposedly, the n onphysical event has a causal impact on a physical event in virtue of its nonphysical properties of the kind k Should we assume that the properties that ground the transfer of such physical quantities ha ve to be physical as well ? If nonphysical properties can ground the transference of physical quantities, we need to explain h ow this can be 61 If we could, what would it mean for the adequate cause in (4)? If there is also a nonphysical property that grounds that transference, we would face systematic overdetermination. If there is not, (4) may be false depending on whether or no t we can preserve the adequacy of the cause in (4) even in the presence of transfer that is grounded by nonphysical properties only. I f only physical properties can ground such transference, then (2) turns out to be false. 62 Alternatively, we may consider t he existence of nonphysical quantities. But then it seems the impact a nonphysical event can have on a physical event cannot be of the same sort the adequate physical cause has, for one transfers a nonphysical quantity and the other a physical quantity. Un less nonphysical and physical quantities can convert into each other somehow 63 this would yield an implausible picture of the world in which nonphysical events can have a causal 61 This, of course, is a particular version of the problem of mental causation sketched in section 1.1. 62 Dennett (1991, 35), for example, takes e nergy conversation laws to be part of an argument against assumptions such as (2) exactly because no physical energy or mass is associated with minds. 63 ld argue that, like the gravitational field, the notion of energy simply cannot be defined for the mind, and hence one cannot even apply PEC to the mind/body interaction. The mind, like the gravitational field, could cause a real change in the energy of th used in physics.
90 impact on physical events but only by way of transferring nonphysical quantit ies that in turn cannot affect physical quantities. If however, nonphysical quantities can convert into physical quantities nonphysical events might indeed have a causal impact on physical events ; but that would violate the completeness assumption (4) 64 a nd the nonoverdetermination assumption (5). Considering how obscure such a view would be, it may be more reasonable to expect nonphysical events either not to have a causal impact at all or to have it in virtue of physical properties and qualities they possess. But in that case, the causal impact assumption (2) is false as we said earl ier That is not a particularly surprising result, of course. Given that the standard examples of transferred qualities are usually physical, we might suspect that propone nts of transference theories tend to generally be friendly towards a physicalist world view and skeptical of assumptions such as (2). 3.3.6 Single Difference Theor ies As mentioned in section 3.3. 4 regularity theories construe cause and effect as a relatio n that depends on how other events of the same type relate. Single difference theories, by contrast, deny that causation d epends on regularities. Whether two events relate as cause and effect is solely a matter of the nature of that relation and of the eve immediate environment. In that respect, single difference theories resemble INUS conditions except that the latter can construe causal fi elds as spatiotemporally widely extended Like some transference theories, single difference theories may conceiv e of events as changes in the property of object s If such an event is sufficient for another event to occur, it is to be considered the cause of that event. 65 Given such 64 of gravity, the electr 65 See Ducasse (1926/1993, 126)
91 understanding of events and causes, C. J. Ducasse claims that c was the cause of e if c was the only difference in e e occurred 66 Or, put more schematically: c is a cause of e if i. c occurred during a time and through a space terminating at the instant I at the surface S ii. e occurred during a time and through a space beginning at the instant I at the surface S and iii. n o change other than c occurred during the time and through the space of c and no change other than e during the time and through the space of e 67 caused her ankle to break. We can infer this without ever having seen other people breaking their ankles and without knowing anything about physical laws. In many respects, this is one of the most striking features of this kind of analysis It does n ot refer to any laws, regularities or external factors beyond the immediate environment of the event s Because of that, c can be the cause of e even if events of their type never occur again. Of course, this does not rule out regularities. But if there are regularities, they supervene on the causal relations, not vice versa. 68 Thi s has interesting consequences. Suppose c and e fall under event types C and E respectively. Even if another event of type C appears in a similar environment, there is nothing about the analysis that would require an event of type E to follow. That is, the analysis is perfectly compatible with atypical causes. 66 We relied on Psillos (2009, 146) for this concise formulation. Ducasse himself phrases this general idea of his K was such particular change C as alone occurred in the immediate environment of K /1993, 127). 67 See Ducasse (1926/1993, 127). We have adjusted the notion of events to match the one we are using in other sections. It remains somewhat unclear why Ducasse uses a conditional rather than a biconditional. He seems mainly interested in provi ding sufficient conditions for events to qualify as cause and effect. Still, this allows for there being causal relations due to factors other than those specified in (i) to (iii). Perhaps he merely thinks of the single difference account as one that may a ccompany other definitions and theories. 68 See Ducasse (1969, 21)
92 The theory is riddled with flaws. One, for example, is akin to a problem the nave regularity theory face s : the analysis seems to lack the resources to distinguish between events that follow each other accidentally and events that are causally related. 69 But let us put such wor ries aside for now. After all, we are not in the business of assessing the quality of theories of causation but in the business of applying them to the exclusion problem. So let us do just that. Whereas the nave regularity theory implie s that if there ar e nonphysical e vents that have a causal impact they always need to be typical, the single difference theory does not imply that at all. While the theory is compatible with the typicality assumption (3), especially given that (2) restricts the scope to a p articular kind of events, the theory might allow for atypical events to be widespread. So it may seem somewhat arbitrary that we restrict the exclusion problem to those kinds of events and properties. But if widespread atypical causal impact is possible, the completeness assumption (4) seems rather strong. Of course, a proponent of the single difference theory might still endorse it for reasons unrelated to the theory. But that is quite different from the transference theory which, depending on what kind o f quantity it allows for, might actually provide support for (4). Furthermore from the viewpoint of the single difference theory, it is not clear why we should worry about overdetermination, systematic or not After all, there seems nothing surprising a bout the fact that the two events c and d occur in close proximity right before e occurs Accordingly, it should not be surprising that they are both a cause of e provided th at nothing else happened nearby. So there does not seem t o be a threat of incons istency, as we suspected lingers in the case of counterfactual theories INUS conditions and probabilistic theories 69 See Psillos (2002, 68 71) for a brief discussion of problems with the theory.
93 Finally, it is unclear in what sense causes explain their effects, beyond the fact that nothing else happens in the immediate environment. We may worry what that lack of explanatory force does to our concept of an adequate cause. Suppose a physical and a nonphysical event occur in close proximity right before another physical event occurs. The completeness assumption suggests that the physic al event on its own should explain the occurrenc e of the effect satisfactorily that is, no reference to other events is necessary. But why should we privilege the physical event in such a way over the nonphysical, given that there are no laws that connect these events and that might have some hierarchical structure? The single difference theory might not be satisfying metaphysically speaking. However, at least it served to emphasize how drastic ally our responses to particular assumptions of the exclusion problem may change depending on what kind of theory of causation we presuppose, especially in contrast to the previous theories discussed. 3. 3 7 Causal Pluralism All of the previous theories of causation face serious challenges often in the form of counte rexamples that suggest that they are at best incomplete accounts While not all of these challenges are fatal, each theory faces at least one challenge that seems in surmountable 70 If this is true there are at least four ways to respond : (a) merge two or more of the above reductive theories into one complex theory of causation (b) give up the assumption that there is merely one type of causation as pluralists do, (c) give up causal reductionism, as some singularists do, or ( d ) give up the concept of causation altogether as eliminativists do We may expect that merging two or more of the above theories would have an impact on the exclusion problem that 70 Cartwright (2004) walks through some of these challenges, at least for the most promising theories of ca usation developed so far. She agrees that all suffer from fatal counterexamples and takes that to be an argument for causal pluralism. Psillos (2009, 154) also claims that there are counterexamples to all existing theories of causation.
94 is identical or similar to the combined impact of whichever theories are merged We thus will no t pursue (a) further. Instead, we will address (b) and (c) in this section and (d) in the following two sections. There are several kinds of causal pluralism. One kind claims that our ordinary notion of causation conflates several distinct concepts Once w e avoid that ambiguity we can identify a discrete set of criteria for each kind of causation It then can be shown that allegedly fatal challenges rest on mingling criteria from different sets. Since these criteria can come apart, the challenges lose their force as a response to each kind of causation. A second kind of causal pluralism agrees tha t there are several kinds of causation but denies that there are discrete sets of criteria that distinguish each kind Instead, these kinds resemble each other because their criteria overlap. Ned Hall pursues the former ambiguity strategy and Elizabeth A nscombe adopts the latter family resemblance strategy. 71 Hall distinguishes bet ween dependence and production To see the difference, consider a forest fire that breaks out in June rather than in May because of heavy rains in April and assume that, had it not rained in April, a fire would have broken out in May instead. The fire in June counterfactually depend s on the rain in April Had the rain not occurred in April the fire in June would not have occurred By the same token, it seems odd to claim that t he rain produced the fire. The counterfactual analysis fails to capture this difference between dependence and production and yields the claim that the rain in April was a cause of the fire in June. 72 According to Hall, t his failure is one in principle. Counterfactual theories are bound to fail because they 71 Of course, this i s not an exhaustive list of the kinds of causal pluralism that have been developed. Skyrms (1984), explanatory contexts. For a brief discussion of A nscombe, Hall, and Skyrms, see Godfrey Smith (2009). 72 See Hall (2004, 229 231). See also section 3.3.2 for a basic introduction to the counterfactual analysis.
95 cannot handle overdetermination and double prevention cases at the same time, even if combined with further assumptions. To demonstrate this, Hall introduces the following f ive theses : Transitivity : If event c is a cause of d and d is a cause of e then c is a cause of e Locality : Causes are connected to their effects via spatiotemporally continuous sequences of causal intermediates. Intrinsicness : The causal structure of a process is determined by i ts intrinsic, Dependence : Counterfactual dependence between wholly distinct events is sufficient for causation. Omissions : Omissions failures of events to occur can both cause and be caused. 73 T he proble m is that the counterfactual analysis fails for some types of overdetermination unless it adopts intrinsicness, locality, and transitivity. To illustrate Hall discusses a late preemption case of the following sort: Neurons a and c fire simultaneously, so that e fires at the same time as b ; the inhibitory signal from e therefore prevents d from firing. If c e still would have; for in that case d would not have been prevented from firing and so would have stimulated e t o fire. Likewise for every event in the causal chain leading from c to e : If that event had not occurred, e would nevertheless have fired. So the strategy of finding suitable intermediates breaks down; for it to succeed, e would have to depend on at least one event in the chain leading back to c and it does not. 74 But e lacks the right sort of counterfactual dependence only because of an accidental feature of its surro unding, namely the fact that there are two events, a and c and two causal chains each of which would lead to e in the absence of the other chain However, intuitively the relation 73 Ibid., p. 225; italics in the original. On page 339, Hall elaborates on the intrinsicness th esis in a useful way. It character as our given structure, also duplicates this aspect of its causal character that is, each duplicate of one of e e 74 Ibid., p. 235. Late preemption is a type of overdetermination, according to Hall.
96 between c and e seems causal, regardless of the lack of counterfactual dependence in virtue of the presence of another causal chain 75 The intrinsicness thesis tries to make sense of this But if we adopted the intrinsicness thesis to get the right results for late preemption cases such as the above we would fail in our analysis of double prevention cases such as following : Neurons a b and c all fire simultan eously. The firing of c prevents e from firing; if e had fired, it would have caused f to fire, which in turn would have prevented g from firing. Thus, if c had not fired, g would not have. So c is a cause of g 76 For c to qualify as a cause of g we need to accept that omissions can be causes After all, c prevents e and f from occurring. There is no causal chain connecting c and g T his violates the intrinsicness thesis because nonoccur r ing events cannot enter nomological relations Hall then demonstrates that similar inconsistencies arise if we try pairing the counterfactual dependence with locality and transitivity. The problem can be solved by assuming that we are really dealing with two distinct kinds of causal relation The depe ndence relation capture s the dependence and omission these s whereas the product ion relation captures the intrinsicness, locality, and transitivity theses 77 How does this distinction between dependence and production affect the exclusion problem? Recall that a ssumptions (2) to (6) use the notion of causal impact. In section 2.2, we said that when events have a causal impact they are strictly involved in the p roduction of the effect, i.e. they are part of a causal chain that leads to the effect. It then looks as if the causal imp act 75 Interestingly, this point echoes one that we made in section 3.3.1. We suggested that overdetermination cases such as the assassin case lack systematicity in virtue of the circumstances rather than in virtue of the properties of the causes that overdetermine the death of the victim. 76 Hall (2004, 241f.). 77 merely trying to illustrate why Hall distinguishes between two rather than three or more kinds of causation and to explain why he adopts depen dency and production rather than other kinds of relation.
97 locality and transitivity. It is, howeve r, not clear whether causal explanations invoking nonphysical properties of the kind k always fit the bill. Suppose a colleague is surprised that we are still in the office at nine kids to t Having forgotten a promise made earlier seems like a perfectly natural, arguably causal explanation of why we are still in the office But if it is, it rests on an omission, which according to Hall is captured by counterfactual dependence rather than by production. But changing that definition to allow for both counterfact ual dependence and production is risky business, for it may lead to ambiguous assumptions To avoid an equivocation, we need to make sure that the kind of causal relation involved in the assumptions (2) to (6) remains the same. Of course, the easiest w ay to avoid complications is simply to ignore mere counterfactual dependence and omission cases and stick to production. But this easy opt out may not always be an option Different kinds of causal pluralism distinguish between diff erent kinds of causal relat ion Some m ay introduce causal relations that are all compatible with our concept of causal impact resulting in different kinds of causal impact. E xcluding one kind of causal impact may seem ad hoc and unjustified so we may have to allow for several kind s. To avoid ambiguity, any application of the exclusion problem to specific events and properties would then have to specify what kind of causal impact is used in (2) to (6) However, Elizabeth Anscombe s kind of causal pluralism may undermine this strateg y 78 78 tions. As will become clear loosely connected by way of family resemblance. But it also is singularist in spirit in that it is antireductioni st. Causal relations could hold even if there were no laws that connected the types of events in question. That is why Moore (2009, 4) views Anscombe as a singularist.
98 Anscombe line up in the way that Hall suggests. When we speak of causal relations in ordinary language, we often refer to v ery specific causal concepts such scrape push wet carry eat burn knock over keep off squash make (e.g. noises, paper boats), hurt 79 And sometimes we posit specific relations between various kinds of relata such as compress the air in the carburetor chamber, the sun attracts the planets, the loss of skill among long term unemployed workers discourages firms from opening new jobs 80 In English, w e can refer to the relata in these effect and to the rela ; but when we do so we are merely grouping them based on their family resemblance. Doing so requires mastery of the concepts in question. Because of that, such concepts seem semantically pri or to the concepts of cause, effect and causation However, it is not as if ever y causal relation shared a single s et of criteria that all others share. For any kind of causal relation there may be another kind that shares some of its criteria, but it is not guaranteed that any two relations share any such criteria. 81 While our notion of causal impact may exclude some such causal relations, it likely will group together several specific causal relations, especially given how different the vocabularies of various special sciences can be. Suppose we wanted to explain how a cell developed the shape it has. In biology, we might state that the cell membrane anchors the cytoskeleton and attaches 79 Anscombe (1971/1993, 93); italics in the original. 80 Cartwright (2004, 814); italics that sentences with a specific causal verb can be replaced by sentenc loss of information. He does acknowledge exceptions, though. 81 Famously, Wittgenstein (1953, paragraphs 66 67) introduces the concept of family resemblance. Cartwright suggests we are abstracting from more specific content of the description and phenomena they describe. We thus are bound to lose information. For her, that is why these concepts have semantic priority.
99 to the extracellular matrix and other cells to hold them together to form tissues. 82 Suppose then that anchor ing attach ing and hol d ing together are specific kinds of causal impact that events have in virtue of whichever properties of the cell membrane explain said effects. If we treated these as contenders for nonphysical events and properties, the exclusion problem then would get of f the ground only under two conditions. Either the kinds of causal impact in (4), (5), and (6) match the kinds of causal impact in (2) and (3) or we explain how the latter kinds relate to the former kinds. The less they resemble each other the less obvio us it may be in what way such nonphysical events and properties are excluded by an adequate physical cause. We may be able to explain what the exact r elation between these different kinds of causal impact is in specific instances, but without investing ext ra effort we may have no way of telling whether the exclusion problem is valid when we apply it to specific events and kinds of causal relations. 3. 3 8 Causal Antirealism All of the previous theories of causation were realist in spirit; they assumed that there are genuine, mind independent causal relations in nature Most of these theories were also reductive : they attempted to explain the nature of causal relations by reference to other items in our ontology for example laws of nature. But even when they were reductive, they did not eliminate the concept of causation altogether. After all, p roviding an analysis of the concept by breaking it do wn into its constitutive parts is not necessarily to deny that the analyzed relation is real. Elimi nativists such as Bertrand Russell want to do just that deny its reality, albeit not necessarily for reasons related to reduction 83 82 83 The best anthology for contemporary views that at least in part rely on Russell is Price and Corry (2007b), although the views included do not always amount to eliminativism. In Price and Corry (2007a), the authors defend perhaps indispensable, in our dealings with the world, it is a category provided neither by God nor by physics, but rather independence required for
100 Russell offers two main arguments for eliminating the concept of causation from our ontology and explanatory practices. Fi rst, the concept itself is confused; and second, our best science, physics, does not employ it but relies instead on laws of nature. 84 To develop the first argument, Russell presupposes a necessitarian theory of causation, one that endorses what he calls th e law of causality: Given any event c there is an event e and a time interval T such that, whenever c occurs, e follows after an interval T 85 Since c and e are event s, they are temporally extended; that is they must endure for some time. 86 But if they do, we face a problem. For c to be a cause of e c needs to be spatiotemporally contiguous with e as we saw in section 3.3.1. But then only the latest part of c can have a causal impact on e for the earlier parts are not contiguous with e 87 Since they are not contiguous, there is a time interval between them and the earliest part of e Because things could happen during this interval that prevent the occurrence of e we must ensure that they do not. But then c is not sufficient for e Even wo rse, varying environments will require varying ways to ensure no interference. But the more of the environment we need to include, the less likely it is that e occurs whenever c occurs. 88 While Russell seems to think that this ultimately undermines the necessitarian theory of causation, he mainly uses the point about varying environments to establish his second argument Scientific realism as we have d efined it in 3.3 or they propose some kind of epistemic pragmatism but ultimately endorse an ontological eliminativism. 84 In fact, Russell (1912 13) offers a third reason related to determinism and teleology, but will ignore it here. 85 Ibid., p. 4. We have adjusted the way Russell refers to particular events to match the way we have done throughout this dissertation. 86 sil ent on the issue. 87 Ibid., p. 5. 88 Ibid., pp. 7f.
101 explanations rely on laws of nature rather than on the law of causality. 89 The antecedent s in actual nomological relations are likely so complex that the effect never will occur in exactly the same way. Consequently the regularities suggested by the law of causality, that whenever c occurs e follows, have no equivalent in science. 90 Regardless of whether of science is accurate 91 let us assume that he is right in thinking that the concept of causation is expendable and should be eliminated from our ontology and explanatory practices. How would this affect the exclusion problem? As it stands, all but the distinctness assumption (1) include causal assertions. If causation is not a real relation, (2) to (5) all turn out to be false and the threat of exclusion is averted. Well, not really. I t seems possible to construct a nomolog ical versi on of the exclusion problem instead. To illustrate what such version might look like, we may replace the causal impact assumption (2) with a claim involving laws of nature : Nomological S ubsumption : Sometimes, nonphysical events are subsumed under a law in virtue of their nonphysical properties of the kind k and the law connects these nonphysical events and physical events. Assuming that laws are exceptionless, we might omi t the typicality assumption (3) and rephrase the completeness assumption (4) as follows: 89 Not all eliminativists replace causation with laws of nature. Niels Bohr, for example, tried to replace causation specifically with complementarity on the atomic domain; see Stltzner (2009, 112). 90 See Russell (1912 13, 9). Later, on page 12, he also points out that the laws in physics are probabilistic and thus developed, as explained in section 3.3.4, this point seems rather moot nowadays. 91 See Lange (2009, 249ff.) for a critical discussion. Schaffer (2016, section 2.1) summarizes the main worry as he main objection to eliminativism is that causation is too central to elimina te. Causation, according to various contemporary philosophers, is required for the analysis of metaphysical concepts such as persistence, scientific concepts such as explanation and disposition, epistemic concepts such as perception and warrant, ethical co ncepts such as action and responsibility, mental concepts such as functional role and conceptual content, and
102 Completeness: If a physical event p 2 is subsumed under any law at all, there is at least one ph ysical law it is subsumed under; i.e. there is at least one law that connects p 2 and another phys ical event p 1 Likely, we will need to specify the nomological subsumption assumption further. The problem is that there are many laws that supervene on physical laws without competing with them in any way. But without some kind of competition, we w ould n ot get the t h reat of overdetermination and exclusion. One solution might be to assume that the laws referenced in the nomological subsumption assumption do not themselves supervene on physical laws. But this assumption would require a proper defense. So i t may be trickier to rephrase the nonoverdetermination assumption and the exclusion assumption in terms of laws. Still, doing so seems possible in principle. I f it is, causal antirealism, especially in the form of eliminativism, does not have much of an advantage when dealing with the exclusion problem at least not more of an advantage than those reductive theories of causation that rely on laws of nature And ct on the problem is concerned, it will probably depend on what kin d of relation causation is replaced with. If laws of nature take its place, the impact may be identical to the one that the nave version of the regularity theory or any theory of causation that grounds causal relations in the laws of nature has If someth ing like complementarity takes its place as suggested by Niels Bohr, the impact may be different. 3. 4 Chapter Conclusion After restricting the scope of our inquiry to a representative rather than a comprehensive set of theories of causation, we cautione d against mingling domains of di scourse too easily. Doing so may obscure what the causal explananda in a given explanatory context are We then explained key features of various theories of causation and illustrated how they may affect the exclusion proble m.
103 The exclusion problem allowed for, but did not apply to events that have atypical causal impact on physical events in virtue of their properties of the kind l rather than k This partly motivated the introduction of the typicality assumption to allow for nonsystematic cases of overdetermination. However, if we presuppose a nave regularity theory, such atypical causes are not possible. That makes the way we phrased the causal impact assumption (2) unnecessarily co mplicated and renders the typicality assumption (3) superfluous. While the theory is compatible with the nonoverdetermination assumption (5), it is not clear why we would be particularly invested in it if we accepted the nave regularity theory. If we were forced to give up one assumption to avoid exclusion, (5) might be a particularly promising candidate. For some kinds of overdetermination, the basic counterfactual theory suggests that the allegedly overdetermining events are not causes at all, for the ef fect neither depends on them counterfactually nor is connected to them by way of a causal chain. Perhaps that provides particularly strong reasons to hold onto (5). Unlike the nave regularity theory, the basic counterfactual theory may provide reasons for rejecting the completeness assumption (4) as well There may be cases where the better candidate for a cause of a physical event is the nonphysical event that has a causal impact in virtue of its nonphysical property of the kind k While the basic counter factual theory sees trouble only for specific kinds of overdetermination cases at first glance INUS conditions must consider all genuine cases as impossible. The implications would be much more severe than in the case of the nave regularity theory or the basic counterfactual theory. Not only would we have to reject the causal impact assumption (2) and the typicality assumption (3) because all alleged cases of nonphysical causal impact would lead to overdetermination, even the adequacy of physical causes i n ( 4) would be threatened by overdetermination. Proponents of INUS conditions thus would have even stronger
104 reasons to hold onto (5) when deliberating which assumption to give up. To avoid these consequences, they could postulate that the causal fields inv olved in explanations citing nonphysical causes and properties are generally different from those citing physical causes and properties. If they are, there is the danger of an equivocation lingering unless we specify the causal fields involved in the caus al claims made by each assumption This last option is not one available t o the basic probabilistic theory I t denies that the alleged causes in overdetermin ation cases are in fact causes and sees that as a threat to the adequacy of physical causes in (4) and (6). As the previous theories, it can either be particularly invested in holding onto (5), or it can merge with other theories of causation such as process theories which may provide the resources needed to deliver a more convincing analysis of overdet ermination cases. The basic transference theory seems to affect the exclusion problem in a manner t hat is quite different from the previous theories If such a theory understands causation to be the transfer of physical quantities in particular, it can pr ovide support for the completeness assumption (4) a nd will not give it up lightly. Furthermore, the causal impact assumption (2) and the typicality assumption (3) may turn out to be false, for these require the causal impact of the event in virtue of its n onphysical properties. If such properties can ground the transfer of physical quantities this consequence may be avoidable, though. Unlike the nave regularity theory, the single difference theory is compatible with there being many events that have an atypical causal impact. While it can endorse (2) and (3), it is unclear whether it would be particularly invested in the typicality of the causal impact of nonphysical events. In addition, if widespread atypical causal impact is possib le, we might not be
105 particularly invested in the completeness assumption (4) when we are asked to give up one or several of the assumptions to avoid exclusion. Depending on what kind of causal pluralism we endorse, if at least two of the causal relations distinguished from each other are compatible with our notion of causal impact, we may run the risk an equivocation again unless we qualify which causal relation is involved in each because the notions of causal impact and adequate cause line up nicely with his notion of and we have an abundance of different causal relations loosely related by way of family resemblance, it may be rather difficult to get the exclusion problem off the ground. Finally, causal eliminativism does not seem to undermine the exclusion problem bec ause it seems possible to construct a nomological rather than a causal exclusion problem. How such a nomological version may be affected by different theories of laws of nature may be a topic for another project.
106 CHAPTER 4 STANDARD RESPONSES TO T HE EXCLUSION PROBLEM 4.1 Or ganizing the Responses We started our project wondering why many philosophers seem to assume, implicitly or explicitly, that differences among theories of causation do not change the nature or the dynamic of the exclusion problem Prima facie, it seems reasonable to expect that they at least might have an impact on how we respond to particular assumptions of the problem. I f the problem can be resolved only by rejecting one of the assumptions, it would matter if different theories incline us to endorse or reject different assumptions. However, Chapter 3 has shown that, while theories of causation indeed affect the degree to which we might be invested in particular assumptions, they affect the exclusion problem far beyond that. Some theories allow for or even suggest the existence of events that have an atypical impact, but the exclusion problem as we have developed it does not capture such events. From the viewpoint of such theories, the exclusion problem seems incomplete. It migh t succeed in excluding some no nphysical events and properties; but depending on how widespread we think atypical causes are, this exclusion might not be much of a threat. Other theories of causation imply that such atypical causes cannot exist. From the vi ewpoint of those theories, the threat of exclusion is much more severe. The typicality assumption (3) also would be redundant and could be omitted. Some theories would also allow for or imply the existence of different kinds of causal relation. But our version of the exclusion problem seems to utilize only one, based on the notion of causal impact. From the viewpoint of such theories, the problem then might either not be particularly serious or involve an equivocation. It would not be particularly serious if it only applie d to cases where nonphysical events enter the same kind of causal relation that physical
107 events do But if there are other kinds of causal relation we might not be particularly invested in the idea that nonphysical ev ents enter the same kind of causal relation to begin with, for they could be said to be causes in another sense. The exclusion problem would involve an equivocation if there were different kinds of causal relation that might qualify as different kinds of c ausal impact. If we distinguished between these kinds, completeness or overdetermination worries, and by extension the threat of exclusion, might d issolve. We have thus accomplished what we set out to do, namely to show that theories of causation indeed h ave an impact on the exclusion problem. We still might wonder, though, whether this impact in turn affects arguments made in the literature regarding the exclusion problem. This is the task of Chapter 4. There are two parts to this task. First, in many res pects, our version of the exclusion problem is more carefully developed than many versions discussed in the literature. If used as an argument for exclusion, where the assumptions (1) and (4) to (6) serve as premises to reject (2) and perhaps (3) our vers ion seems valid whereas most other versions are not. Consequently, we need to test whether some of the responses to the other versions work as responses to our version as well Second, given the potential impact of theories of causation on the exclusion pr oblem, we should investigate whether they also affect the plausibility of the responses to the problem. In section 3.1, we established that we could look only at a representative set of theories of causation, for there are too many theories and modificati ons of theories out there to discuss them all. This is even more of a n issue for responses to the exclusion problem. P hilpapers.org lists over one thousand publications that include the term exclusion problem in their abstract. Consequently, we will need to be very selective. To avoid picking responses at random, we can use taxonomies offered in the literature to make sure we cover at least a reasonable amount of
108 territory. For example, David Robb and John Heil suggest that many responses fall in one of three categories: identity solutions, autonomy solutions, and inheritance solutions. 1 Put briefly identity solutions reject the distinctness assumption (2) to prevent the exclusion problem to get off the ground. Autonomy solutions reject the exclusion assumption (6) by arguing that causal explanations involving the physical and the nonphysical tend to be domain specific and independent of each other in such a way that physical and nonphysical causes do not compete Inheritance solutions avoid such competition by pointing out that physical and nonphysical properties are intimately related in such a way that nonphysical properties inherit the causal impact, or rather their causal powers, from physical properties. We will make sure to include at least one alleged solution from each of these categories. However, some responses do not fit into this seemingly neat taxonomy either because they do not fall into any of these categories or because they cut across them. If so me of the alleged solution s respond to particular assumptions of the problem, it may be interesting to see whether such alternative responses explore options that the standard responses do not explore. Finally, some philosophers hold views that could be us ed as a response to the exclusion problem but that have not yet been applied to the problem Applying such views may offer insights that have not been addressed in the literature yet. Given the variety of responses to be covered the easiest way to structu re this ch apter is by running through the assumptions of the exclusion problem one by one and discuss ing particular responses whenever they have something interesting to say about an assumption Sometimes they m a y provide reasons for rejecting one ; other times they may provide reasons to hold onto one ; and yet other times they may just introduce important distinctions that help us to 1 1 Robb and Heil (2013, section 6).
109 get clearer on what the claims are and how the concepts involved relate to other important concepts. 4. 2 Distinctnes s I n some ways the easiest response to the exclusion problem is to reject that there are nonphysical properties of the kind k that are distinct from any physical propert ies Eliminative materialism claims that standard examples of nonphysical properties s uch as beliefs, desires, and intentions do not exist. If they do not exist, then there are no events to exclude. Some kinds of r ealization physicalism claim that seemingly nonphysical properties are in fact identical with their physical realizers. If they are identical, there are no events to exclude either, for any case of alleged nonphysical causal impact really is a case of physical impact. Eliminative materialists typically focus on mental events and properties in particular Paul Churchland, for examp le, argues that we ordinarily explain human behavior by reference to beliefs, desires, intentions, perceptions, and so forth. Such explanations presuppose regularities among mental events and properties cited in the explanations and the behavior to be expl ained. 2 Sometimes s uch common sense laws are used to explain behavior, other times they are used to predict behavior, and yet other times they are modified when they do not work. Given that they are used in this way, they should be considered to be part of a genuine theory, usually called folk psychology. 3 Unfortunately, this theory is radically incomplete and likely to be wrong It is incomplete because it offers no explanation of a wide range of important phenomena. F or example, the theory is una ble to explain the function of sleep or the nature, dynamics, and 2 If the explanatory force of folk psychology indeed rests on regularities, the falsity of any theory of causation that assumes or implies exceptionless regularities would threaten folk psychology. We have made a similar point before, in section 3.3.1, with res pect to the typicality assumption (3). 3 See Churchland (1981, 68f.).
110 origins of mental illness. 4 In spite of this radical incompleteness, folk psychology has been more or less stagnant for several thousands of years. We are not much better than the ancient Gr eeks were at explaining human behavior when we exclusively rely on folk psychology 5 It is likely to be wrong because we have an admittedly incomplete but increasingly powerful set of theories neuroscience, evolutionary biology, organic chemistry, physics that connect reasonably well with each other and that provide increasingly nuanced explanations of human behavior in particular and the world in general But folk psychology does not integrate into this larger corpus of theories and knowledge. 6 The best ex planation for this failure of folk psychology may be that the kinds of events and properties it posits do not actually exist, especially if our best theories do not cite them Therefore, we should eliminate them 7 If we can eliminate them, we should reject the distinctness assumption (1) of the exclusion problem. As a result, the problem does not even get off the ground, as we stated in section 2.2 when establishing (1). However, to the extent that eliminative mate rialism is a defensible philosophical position, it may serve as a response to versions of the exclusion problem that threaten to exclude mental events and properties in particular. But our version is generalized. It threatens to causally exclude all events that allegedly have a causal impact in virtue of their nonphysical properties of the kind k These properties include those cited by the special sciences that Churchland claims are part of 4 Ibid., p. 73. 5 Ibid., p. 74. 6 Ibid., p. 75. 7 Note that this move with respect to the mental events and properties is similar to the one that causal eliminativists make regarding th e concept of causation (compare section 3.3.9). Note also that the vocabulary of folk psychology may not be sophisticated and nuanced enough to square well with some of the theories of causation we have discussed. That may be an additional weakness. For ex that processes remain uniform in the absence of interactions with other processes. But most of the causal explanations employed by folk psychology are much too crude to include such level of detai l.
111 8 Unlike folk psychology, these special sciences have not been stagnant o n the contrary. Consequently, Churchland could not use his main argument against the existence of mental events and properties as a reason to reject the distinct ness assumption (1 ) Doing so might undermine his reason for rejecting folk psychology in the first place There are two obvious ways for Churchland to respond. He could consider the special sciences to be useful heuristics that cite properties that ultimately can be elimin ated as well, once physics will have reached a sufficient degree of complexity. Or he could argue that there is an important difference between folk psychology and the special sciences that warrants treating them differently. One such difference we hinted at earl ier namely that the special sciences connect well with physics whereas folk psychology does not. At least for some properties in the special sciences bridge laws that connect them with physical properties and laws have been established. The existe nce of such bridge laws is often thought to be an important condition for a successful reduction of one theory to another. 9 While reduction may suggest identity, it does not require elimination. And w hile at this point in history we have not yet succeeded at reduc ing all special sciences to physics, we may have good reasons to expect a future science to exhibit such reductive unity 10 However, proponents of scientific pluralism reject this picture of science progressing towards a unified theory John Dupr for example, claims that the reduction of one theory to 8 Churchland (1981, 75). 9 See Horgan (1978) for a brief explanation what a reduction by way of bridge laws may look like. However, Horgan suggests that successful reduction does not necessarily commit us to property identity. The classical model of i ntertheoretical reduction by way of bridge laws is spelled out in Nagel (1961), especially chapter 11. 10 The unity of science view is popular, but it can be understood in various ways. See Cat (2013) for a general overview. See sections 2, 3.1, and 4.1 fo r a discussion of reductive unity in particular.
112 another theory T relies on the identification of structurally complex objects cited by to structures of simpler objects cited by T For this mapping to work, we need to assume that the objects h ave essential properties that are both necessary and sufficient for something to belong to a kind that subsumes the objects in question. Within each theory, these essential properties are the ones that fall under laws spe cific to the theory that then expla in phenomena in the respective domain. Dupr argues that many if not all, special sciences do not assume the existence of s uch domain specific essential properties. To explain ecological systems, for example, ecology levels multicellular organisms, single cells, and molecules (as nutrients in the environment) but the understanding of such systems typically involves such factors as climate and geology, factors that would have to be assigned to parallel 11 S uch mixing and mingling of objects and properties from different levels of organization to explain events amounts to categorizing and dividing up the world in radically different ways, so much so that it undermines any promise of mapping the properties and laws in the way required for reduction. While different disciplines may categorize and divide up the world in different ways, the causal and nom ological relations they posit between events involving different kinds of object are perfectly real. But since they do not map onto those posited by other disciplines, there is no threat of exclusion even if (1) to (3) are true 12 While scientific pluralism might allow for the completeness assumption (4) and the overdetermination assumption (5) to be true within physics without undermining the causal 11 Dupr (1993, 103). 12 Of course, this is a very sketchy summary of a position that is based on the analysis of various special sciences. Specifically, our summary does not explain how we should understand the claim that the world has a pluralistic structure. We also may be sym pathetic to a pluralism in the sciences without assuming some kind of ontological pluralism, however that is to be understood. Horst (2007) offers such an epistemic kind of pluralism, which he dubs
113 impact of nonphysical events in other disciplines, it is not clear why we should th ink that the kinds of events objects, and properties which figure into the explanations of various disciplines have to be so different that they never map onto each other Why think, for example, that biologists and ecologist cannot both explain features of the very same organism? If such is consistent with his scientific pluralism Dupr has two options. He may admit that this can happen. If these sciences attempt to explain not just the same events and objects but the same properties of these events and objects as well (4) and (5) might be false. Or he may admit that it can happen but deny that these sciences explain the same properties of the events and objects in question. The latter strategy resembles the dual explanandum strategy pursued by Terry Hor gan which we will get back to in section 4.6. 13 The general idea is that causation is a four place relation which always includes the specific properties of the effect that are necessitated by a properties of an effect might sometimes be explained by different causal chains. Dupr could then add that these different causal chains might figure into the ex planation s of different sciences, which is why they can all include true explanations without competing with each other. If that is a possibility, scientific pluralism may indeed be compatible with (4) and (5). 14 Regardless of whether we with pluralism, it is important to recognize that these views relate differently to the theories of causation di scussed in Chapter 3 ems compatible with all the theorie s ce various sciences assume probabilistic laws, both the nave regularity theory and INUS conditions seem like an 13 Horgan (1989, 50). 14 Dupr himself merely states that his view is not committed to completeness; see Dupr (1993, 216).
114 unlikely match. And the transference theory might be incompatible depending on what kind of quantities it assumes there are Dupr himself endo rses probabilistic causation as a promising candidate in the context of scientific pluralism. 15 However, that might not work the way he envisions. Recall that we argued in section 3.3.4 that at least basic probabilistic theories have good reasons to endorse the nonoverdetermination assumption (5). But if scientific pluralism implies widespread overdetermination, assuming that various sciences at least sometimes attempt to explain the same events and properties, probabilistic theories would be at odds with su ch pluralism. As pointed out at the end of section 3.3.4, we could combine it with other kinds of theory such as process theories, in which case Dupr might be able to offer a consistent view. Still, in the context of the exclusion problem, the incompatibi lity with several theories of for it will only be a convincing response if it is plausible to begin with and if the theory of causation it relies on is likely to be true On the other hand, if w e deem eliminativism too high a price to pay we may consider type identity theories. Unlike eliminativist materialism, realization physicalism does not deny that there are mental properties. It is just that every property instance is either an instance of a physical 16 Functional properties are specified by way of the causal role they play in a given physical system. For example, to be in pain is to be a in state that is typically produced by some kind of injury, that produces the 15 See Dupr (1993, chapter 9). 16 Melnyk (2003, 26).
115 desire to be out of that state, that often leads to wincing and so forth 17 Usually nonphysical properties such as biological, chemical, or economical ones are considered to be functionalizable a s well. But the way we generalized the exclusion problem in section 2.3 may yield an awkward sounding analysis because w e opted to contrast physical with nonphysical properties. Applying the functional analysis to the distinctness assumption (1), we then would have to say that instances of nonphysical properties are really just physically realized instances of some functional property. That makes it sound as if we first treat it as an analytic truth that nonphysical properties are distinct from physical pr operties just to then assert that they are not distinct at all in spite of the distinctness being analy tic That makes little sense. R ecall that we specified in section 2.3 that by nonphysical events and properties we meant those events and properties t reated by the special sciences as serious contenders for being nonphysical events and properties 18 Whether they are indeed nonphysical remains an empirical question. Against that background, the analysis would simply state that allegedly nonphysical proper ties really just are physical properties or physically realized functional properties How strong a response to the exclusion problem this so called identity solution 19 is ultimately depends on how defensible realization physicalism is That discussion would lead us very far astray. Since the exclusion problem as well as the entire project of this dissertation is premised on the assumption that we can make sense of the distinctness claim and should take it 17 The example is taken fro m Levin (2013, section 1). By the way, it would be perfectly fine to refer to other mental states such as beliefs or desires in such functional analyses, for they in turn are physically realized instances of functional properties. 18 Melnyk (2003, 32) calls scientific or honorary physicalism does not need to deny the existence of such types, but it does require to treat them as identical with physical types. 19 Robb and Heil (2013, section 6) c onsider identity solutions as one of three standard responses to the exclusion problem, the other two being autonomy solutions and inheritance solutions, to which we will return to in later sections.
116 seriously, it seems most usefu l to look at standard responses to the other assumptions of the problem. 4. 3 Causal Impact e versions discussed in the literature are. Since there is no canonical version of the pro blem, we developed our own. But apart from dropping a few hints here and there, we have not yet discuss ed own response to the exclusion problem. To do so, let us remind ourselves of how exactly he sets up the problem as an argument for epiphenomenalism: If an event x is causally sufficient for an event y then no event x distinct from x is causally relevant to y ( exclusion ) For every physical event y some ph ysical event x is causally sufficient for y ( physical determinism ) For every physical event x and mental event x *, x is distinct from x ( dualism ) So, for every physical event y no mental event x is causally relevant to y ( epiphenomenalism ) 20 exclusion premise seems to be a blend of our nonoverdetermination assumption (5) and the exclusion assumption (6), except that he uses the notions of causal sufficiency and causal relevance to state it He then argues that this exclusion premise is false. From the fact that there is a sufficient cause of an event, it does not follow that no other event or property is causally relevant to it. The mistake according to Yablo, is to think that physical and mental causes compete. I f the first place of a causal relation is already occupied by an event, in this case by a physical event, then no other event can also occupy that place. This view in turn relies on the assumption that the competitors for the place are independent of each other in important 20 Yablo (1992a, 426) and Yablo (1992b, 247 248) ; italics in the original.
117 ways. We m ay be swayed by the dualism premise, for it asserts that the events in question are distinct. Alas, distinctness does not imply independence. Consider the properties being scarlet and being red The former is a determinate of the latter, and the latter is a determinable of the former. Being thus related implies that, if anything has the property being scarlet it necessarily has the property being red as well, whereas the reverse does not hold. 21 Yablo then suggests that there are cases where an event is causally sufficient for another ev ent in virtue of its determinate while it is also true that, had it not been the determinate of said determinable, it would not have been the property in virtue of which the causal relation holds. Consider a pigeon that h as been trained to peck at all and only red shapes Since scarlet is a shade of red, the pigeon also pecks at scarlet shapes. When it does seeing scarlet shapes is causally sufficient for its pecking, and yet it seems the redness of the shape s is what the pigeon is responding to and what causally explains its pecking in light of its past training But then it seems the redness was causally relevant after all 22 Yablo then suggests that mental and physical properties may stand in a relation to e ach other that is similar though not necessarily identical, to the d eterminable determinate relation Consequently, mental properties, which are like determinables of physical properties, may remain causally relevant even if physical properties are the on es in virtue of which physical events are causally sufficient for other events. Epiphenomenalism thus does not follow from his premises of exclusion, determinism, and dualism. as a response to our version of the e xclusion problem. In some respects adequate causes play a role in our version that is similar to the one 21 Yablo (1992a, 429). 22 Ibid., p. 432.
118 that To be an adequate cause is to be an event such that, if it were cited in a causal explanation, no ref erence to further events or properties would be needed for the resulting explanation of the event to be satisfactory. We may be tempted to conclude that if an adequate physical cause occupies the first place of a causal relation, it exclude s all other events from having a causal impact as well But this works only if we presuppose particular types of theor ies of causation. If a nonphysical competitor is thought to have a causal impact in virtue of its nonphysical properties of the kind k the n the typicality of such impact as specified in (3) would indeed imply systematic overdetermination and consequently the event would face exclusion But remember that in section 2.2 and then later in section 3.3.1 on regularity theori es of causation and in sect ion 3.3. 6 on single difference theories we pointed out that there may be nonphysical causes that have a causal impact in virtue of an atypical kind l Such atypical causes would not be threatened by exclusion because ic overdetermination. So, from the fact that there is an adequate physical cause it does not follow that no other events can have a causal impact. It merely follows that nonphysical events of the kind referred to in the causal impact assumption (2) are th reatened by exclusion. In that sense, the threat unless we adopted a type of theory of causation that would not allow for an atypical causal imp act of some nonphysical events The nave regularity theory seems to b e such a theory whereas single difference theories may not be But while causal sufficiency might resemble causal impact to some degree, there is no counterpart to causal relevance in our version of the exclusion problem. A ssumption (2) does not merely st ate that nonphysic al events are causally relevant; it claims that they sometimes have a causal impact on physical events They may not ever be adequate causes like physical events
119 sometimes are, but physical and nonphysical events both can have a causal impact ; that is they can both be relata in the same kind of causal relation Whereas properties are threatened not e ven to be causally relevant, which is a more sev ere threat, nonphysical events and properties in our version are merely threate ned not to have a causal impact, i.e. to not be part of a causal chain that leads to the effect. This may be compatible with the ir being causally relevant in some way, but such relevance must be of a sort that does not undermine the adequacy of the physical causes referred to in the completeness assumption (4). In this context, it may be important how exactly the notion of underst oo d and what determi nables and by analogy mental and some nonphysical properties are supposed to explain. D eterminable properties are relevant in that they specify the degree to which the effect lacks sensitivity to the highly specific physical properties of the cause. 23 But what does that mean? Ya blo himself is not very clear in that respect partly because he never At times, Yablo seems to affirm that determinate properties explain the causal sufficiency of the cause albeit in too much detail That sounds like a causal grounding claim: an event is causally sufficient for another event in virtue of its determinate properties. But how then are we to understand the claim that determinable properties explain as well why an effect oc curred specifically that they are relevant for the effect ? We could take Yablo to advocate a dual explanandum strategy of the sort we will discuss in more detail in section 4.6. While determinate properties explain the determinate properties of the effect determinable properties explain determinable properties of the effect But if he indeed adopts a dual explanandum strategy, he could easily do without his determinable determinate distinction as we will see in section 4.6. In fact, even the distinction between causal relevance 23 Yablo (1992b, 278).
120 and causal sufficiency may not be needed. Be that as it may, the dual explanandum strategy may lead to a rejection of the completeness assumption (4) because there are properties of the effect that are not explained by the physical properties of the cause. If causes are adequate only if they explain all the properties of an effect, 24 then (4) would turn out to be false. Alternatively, we might take Yablo to advocate two distinct kinds of grounding relations. Though he himself does no t state it that way, this view may be consistent with most if not all of his claims. While determinate physical properties ground causal relations, i.e. they are the ones in virtue of which a cause event is causally sufficient for an effect event, deter minable nonphysical properties are causally relevant because they ground that grounding relation. It is in virtue of the properties being determinates of determinables that the cause event is causally sufficient in virtue of its determinate property. But if determinables play this role in causal explanations our version of t he exclusion problem is guilty of an equivocation because it fails to distinguish between two types of in virtue of relation one that grounds causal relations and one that g rounds that grounding relation. 25 For our version of the exclusion problem to get off the ground, it needs to interpret the in virtue of relation in the assumptions (2) to (6) in the same 24 This is a simplification. As explained earlier, an adequate cause is one that, if it were cited in a causal explanation, would be all that we need to cite as a causal factor operative at that time in order to provide a satisfactory explanation for why ano ther event occurs. Such satisfactory explanation would have to locate the cause in time; that is, it needs to identify the event in the causal chain leading to the event the citing of which would yield a satisfactory in the hospital may be rather exhausted, but that may be the result of a party the night before and have nothing to do with the tripping. An adequate cause thus only needs to explain those properties of an effect that makes it an effect of that cause but not necessarily all properties. 25 While the charge is r eminiscent of the one addressed in section 3.3.7 on causal pluralism, it locates the equivocation elsewhere. Curiously, Robb and Heil (2013, section 6.4) list Yablo as a proponent of an inheritance solution. According to such a solution, some nonphysical p roperties inherit their causal powers from the property they are intimately related to. The determinable determinate relation may sound like a good candidate for such intimate relation, but if the suggested analysis is accurate, determinables do not inheri t causal powers from seem to presuppose that the causal power theory is correct, at least not as it was discussed in section 3.3.6.
121 way. Because of that, Yablo would have to reject the causal impact assumptio n (2) as well as the typicality assumption (3). However, whether or not we should reject (2) and (3) depends on how plausible it is to treat mental properties, and by extension at least some nonphysical properties, as if they were related to physical prop erties as determinables are related to determinates. There are at least two problems with this. First, a slight modification of the pigeon example demonstrates that the causal relevance of nonphysical properties does not depend on th eir being like determin ables of physical properties. Second, the description of the pigeon example includes a historical use of the determinable determinate distinction. Suppose th e color red corresponds to a wavelength range of 620 to 750 nanometers. Now imagine that we trained the pigeon not to peck at all red shapes but only at shapes in two color ranges, for example those that correspond to 620 to 660 and 700 to 750 nanometers. Scarlet falls within the lower of these two ranges. If our pigeon now pecks at a scarlet shape, it would be false to say that it did so because scarlet is a shade of red. Consequently, the determinable being red was not causally relevant. E ither we posit that there are determinables that correspond to whatever wavelength range we pick or sometimes determinables are not causally relevant in the way Yablo describes. 26 The first option seems to suggest that the pigeon was trained to peck at two distinct determinables rather than at one One feature of determinables is that they are less specific, i.e. more general, than their determinates. 27 It is hard to see how they could then include gaps of the sort needed in our example. But this is a risky response for we could train the pigeon to peck at shapes of specific colors scattered all over the spectrum so that there is no 26 Falke (2006, 41ff.) makes this point. 27 See Wilson (2017, section 2.1).
122 determinable that corresponds to this scattered range and that could be causally relevant However, the causal explanation would remain the same, namely that the pigeon pe cks at shapes of these particular colors because we trained it to. It is the training, not the determinable determinate relation that grounds the fact that the pigeon pecks at a scarlet shape. It thus may be good to offer a causal analysis that captures t his historical dimension of the original causal explanation of why the pigeon pecks at scarlet shapes. In section 2.2 a historical dimension. For Dretske, there is a clear differe nce between causally explaining why the pigeon pecks at a scarlet shape at a particular time and why what it did was pecking at a scarlet shape rather than doing something else. 28 Triggering causes, like event s that have a ca usal impact, explain the former; structuring causes explain the latter. To clarify the distinction, consider the following example: general gets in the car and turns the key to start the engine. The b omb is detonated (triggered by turning the key in the ignition) and the general is killed. Who kill ed him? The terrorist, of course. How? By planting a bomb in his car. Although the he did a week ago, that will certainly be singled out, in both legal and moral inquiries, as the cause of the explosion that resulted in death. 29 While s tructuring causes provide the conditions for triggering causes to do their work, they are not sufficient for the effect ; they depend on later events such as turning on the engine. Since they depend on later events to conspire in the right sort of way for the effect to occur, they y triggering causes do. 30 Furthermore, structuring causes can 28 See Dretske (1988, 42 43). 29 Dretske (1993, 122 123). 30 3.3.3. Recall that one of the reasons for establishing INUS conditions was that the cause was itself not sufficient for
123 persist through time in ways triggering causes do not. 31 Each peck at a scarlet shape is explained by a different triggering cause, th e perception of each individual color and shape. But the structuring cause explains all of these instances, which the training of the pigeon does as well. It useful seem to be the same. If the causal impact assumption (2) and the typicality assumption (3) cite structuring causes rather than triggering causes, our version of the exclusion argument involves an equivocation. S tructuring causes explain why certain physical ev ents cause their physical effects in virtue of their physical properties. Since they are relata in a different kind of in virtue of relation than the one that figures into assumptions (4) to (6), we should reject (2) and (3) as they stand. But this strateg y seems promising only under two conditions. First, the way nonphysical events and properties figure into special science explanations need to match the way structuring Consider our friend with the broken ankle again. She ma y have a background belief that whenev er she experiences serious pain she should go to the hospital. Such belief may serve as a structuring cause, for it would explain every instance of her going to t he hospital when she is in severe pain. But at least in some cases the pain itself seems to be a triggering cause, even though there are the effect. Only together with the background conditions is it so. We may thus wonder what the exact difference between structuring causes and mere b ackground conditions is. The terrorist example may suggest that triggering causes must be more strictly involved in the production of the effect than most background conditions are. They are do not have to be. An additional worry we may have is whether a counterfactual analysis would be able to capture the distinction between structuring causes and triggering causes. 31 and triggers. Both structuring causes and enabling conditions depend on triggering causes to set off the causal chain leading to the effect; both have a narily would call background conditions; see Lombard (1990, 203ff.). However, Sartorio (2009, 582f.) points out that one could view enablers as causes. If that is right, the difference between structuring causes and enabling conditions might blur further.
124 adequate physical triggering causes, according to the completeness assumption (4). Of course, this is not yet a reason nonphysical events cited by the special sciences are usually treated as structuring causes or not. Second, even though structuring causes figure into a different in virtue of relation, this may undermine the degree to which physical causes are adequate. If a satisfactory explanation of a physical event requires not only reference to the physical cause but also a grounding explanation of the sort that structuring causes are supposed to provid e, then we need to reject (4) if we want to hold onto (2) and (3) 4. 4 Typicality Since the typicality assumption (3) is not usually part of the exclusion problem as discussed in the literature there are no standard responses to it. As a starting point for a discussion, let us briefly remind ourselves of the contexts in which typicality came up in earlier sections. We introduced (3) because we did not want to make the overdetermination assumption (5) too strong. In its strongest form, it would claim that there is no overdetermination. In the light of possible overdetermination examples such as the double assassination case, we opted to merely reject systematic overdetermination. This had us worried that nonphysical events may have an a typical impact such that, if and when they have it, they would overdetermine but not in a systematic way. The typicality assumption was a way to ensure that the causal impact of the nonphysical event would not be atypical. Instead, it would be representati ve of events that have a causal impact in virtue of a nonphysical property of kind k Any such impact would then yield systematic overdetermination, or so it seems. Before we introduced the assumption in section 2.2, we drew an analogy that is worth getti ng back to at this point. To explain why atypical causes are not worrisome or why typical ones are we considered a philosophy class in which only students with yellow socks got an A.
125 This may not be surprising, especially if only one student happened to wear yellow socks all the time. It seems like a mere coincidence. Similarly, if some nonphysical events sometimes have a causal impact on a physical event which already has an adequate physical cause, the overdetermination might simply be due to an acciden tal arrangement of external circumstances, as we pointed out in section 3.3.1. But if twelve out of twenty students got an A and all of them wore yellow socks and no other students wore yellow socks, we would wonder whether there is something about the gra ding procedure that connects the color of socks student wear and their assessment by the teacher. Similarly, if nonphysical events of a certain kind and only events of that kind have an overdetermining causal impact on physical events, it may suggest that the overdetermination is not just due to the accidental arrangement of external circumstances. That is, we will begin to suspect that the overdetermination has something to do with the nature of the nonphysical events or perhaps the nature of the causal re lation they enter. Any theory of causation that implies regularities will exacerbate the oddity of this picture. Single difference theories, however, could simply reject (3). Perhaps there are some kinds of nonphysical event that do have a typical causal impact, but those may indeed be threatened by exclusion qua (4) and (5). But since (5) allows for nonsystematic overdetermination, a proponent of a single difference theory might claim that nonphysical events that have a causal impact tend to be atypical. For example, there may not be laws that events of their types fall under. While this is an option, we now may wonder what exactly single difference causes explain. In section 2.2, we considered the example of our carrying an umbrella The thought that it m ight rain today caused our carrying an umbrella. If the completeness assumption (4) is true, then there was also a physical event in our brain, for example that caused our behavior. Fortunately we are not facing systematic overdetermination because the th ought is supposed to be atypical. But that
126 amounts to say ing that normally, or in other circumstances, it might not have led us to carry an umbrella. But then the reference to the thought loses all explanatory force. So, while the single difference theory may avoid the threat of systematic overdetermination and thereby avoid causal exclusion, this comes at a high price. The causal explanations involving such atypical causes lose all explanatory force, or so it seems. 4. 5 Completeness In section 2.1.2, we p ointed out that some versions of the completeness assumption are too strong. Karen Bennett, for example, phrased it as the claim that every physical occurrence has a sufficient physical cause. 32 T hus stated, it may be false according to our best physical theories, for there may be physical events that have no cause. To make sure that the assumption cannot be shot down easily, we conditionalized the assumption. That is not the only way to do it, of course. Barbara Mo 33 She also cautions us not to interpret the assumption of completeness or closure, as she calls it as stating that physical effects have only physical causes. In the context of the exclusion problem this would not only make the nonoverdetermination assumption redundant but also would imply a stronger claim, namely that overdetermination, even of the nonsystematic kin d, does not exist. Our version of the assumption would have such strong implications only in combination with theories of causation such as the nave regularity theory and so me kinds of probability theory. Finally, she warns not to confuse the completeness assumption with the claim that there is no causal 32 Bennett (2008, 281). 33 Montero (2003, 174).
127 interaction between the physical and the nonphysical. 34 After all, completeness is compatible with physical events causing nonphysical events as long as such causation does not violate some condition for c losure, say conservation laws or the like 35 While our version of the completeness assumption (4) avoids all the pitfalls that Montero cautions against, it does not include what Matthew C. Haug calls the fundamental category ics provides a complete inventory of the fundamental kinds and categories that are sufficient for a realization basis for all of the events that causally impinge on 36 According to Haug, this is a distinct completeness claim that phys icalists are committed to. While this assumption is not needed for the exclusion problem, we may ask what impact it would have on it. If we are willing to endorse the completeness assumption (4), for example because we have physicalist inclinations, then w e should at least be sympathetic towards the fundamental category t assumption (7). If (7) is true and we wanted to endorse the transference theory, we would be bound to restrict the quantities to be transferre d to physical quantities. In that case, however, it looks like the impact assumption (2) is threatened, though not necessarily rejected. Remember that we stated in section 3.3.5 that the quantities transferred are not necessarily the properties in virtue o f which the transfe r takes place. So we could maintain the causal impact of nonphysical 34 Ibid., p. 175. 35 Montero hints at this when she discusses the argument for the causal closure from the success of science. Roughly, if the completeness assumption (4) were false, physics would necessarily b e incomplete, but since it is not necessarily so, (4) is true. This would be a stronger claim than even physicalists would need to endorse. See Montero (2003, 177f.). 36 tegories that characterizes the
128 properties, but only by way of a transfer of physical quantities. Essentially, this is a form of inheritance solution. Typically, such inheritance solutions argue by w ay of a particularly intimate relation between nonphysical events and properties and their physical realizers. But notice that we did not need to specify this relation beyond the fundamental category completeness assumption. The inheritance flowed from a p articular theory of causation in combination with that assumption. But if we presupposed mere counterfactual dependence, for example, we would need to spell out the relation between nonphysical events and their realizers more so as to arrive at that view. When weighing the plausibility of the assumptions of the exclusion problem against each other, we may find that the motivations underlying some of the assumptions come with additional commitments which in turn may be more or less compatible with particula r theories of causation. 4. 6 O verdetermination Horgan claims that causation is really a four place rather than a two p lace relation. Rather than being a relation between events only, it is a relation between events and particular properties they have: For any two events c and e and any two properties F and G c qua F causes e qua G iff: (i) c causes e ; (ii) c instantiates F ; (iii) e instantiates G ; and (iv) the fact that c instantiates F is explanatorily relevant to the fact that e occurs and instantiates G 37 In our version of the exclusion problem, we specified only the property of the cause that grounds the causal relation between c and e 37 Horgan (1989, 50).
129 Such a property does not ground the causal relation simpliciter but only the causal relation with respect to a particular property of the effect. If the completeness assumption (4) is correct, there is a physical event that grounds the relation between c and e in virtue of a physical property, say F. And if the causal impact assumption (2) is correct, there is an event that grounds the relation in virtue of a nonphysical property, say N But notice that neither claim specifies the propert ies of the effect th at are supposed to be explained. Assuming that the nonphysical event in (2) is distinct from c say d 38 it could be the case that c F causes e G whereas d N causes e H But that is an instance neither of overdetermination nor of exclusion. For this strategy to be a good response to our version of the exclusion problem we need to assume that being an adequate cause does not entail that c satisfactorily explains all properties of e for then we get both o verdetermination and exclusion. That may be a reasonable assumption. Physical explanations usually come with a much hig her degree of specificity than explanations in the social sciences, for example. That is, many explanations involve a certain degree of abstraction. Physics might consider a causal explanation of an event satisfactory to the degree that it explains all hig hly specific physical properties of that event. And if the physical cause referred to in the completeness assumption (4) does that, it should qualify as an adequate cause. Notice, however, that this does put constraints on the kind of propert y of effects t hat nonphysical events can causally explain in virtue of their nonphysical properties : either abstract physical properties not captured by the adequate cause or nonphysical properties. 38 It could also be identical with c so that c e H rather than of e G As mentioned in section 2.1.2, this distinction might partly depend on which account of events we endorse. A Kimean account might force us to deny the identity of c and d whereas a D avidsonian account could easily capture this.
130 Abstract physical properties may seem odd from the viewpoint of probability theories. The adequate physical cause would necessitate the probability of the effect occur ring with all its physical properties. Abstract physical properties are intimately related to the particular properties perhaps they stand in a determinable determinate relation to each other. But if they do, why would we think that they would have any impact on the probability of the effect? For the nonphysical event in (2) to be causally related to the abstract physical effect, it needs to be make a differen ce to the probability of the effect But if it indeed does, it is questionable that we were right in considering the physical cause as adequate. If it is not, the completeness assumption (4) is threatened. Alternativ ely, consider transference theories. If the quantities transferred must be physical, the physical cause in (4) must satisfactory explain all the quantities transferred. But then there is no e H to relate to and to explain d N d, he would have to assume that there are nonphysical quantities that transfer to e If they can convert into physical quantities, this will probably undermine (4). If they cannot convert into physical quantities, nonphysical events could explain only nonp hysical properties of events. T o avoid some strange parallel between the physical and the nonphysical, it may be most reasonable to assume a different kind of causal relation at work here counterfactual dependence for example. That is, we might move towar ds a Hall type causal pluralism. 4. 7 Exclusion The way the exclusion assumption (6) is phrased is somewhat suggestive. It makes it sound as if any nonphysical event will have to compete with an adequate physical cause. If causation is a two place relation, and if the first place is already taken by an adequate physical cause, the nonphysical event has no place to occupy. However, w e might question this picture and advocate that physical and nonphysical descriptions are relatively autonomous. For example,
131 n e ither psychology nor physics may describe the world as it is Both describe it relative to a languag e and its terms and categories. To put it crudely, the language of psychology carves up the world differently than the language of physics and it does so in such a way that no systematic mapping of the categories provided by each language is possible. 39 For Donald Davidson, that is not a problem. As we saw in section 2.1.1, he thinks that we use different descriptions of eve nts as a way to refer to them Even two radically different descriptions that refer to two distinct sets of properties still may refer to the same event. But reference to the se sets of properties merel y serves to pick out the event in question, they are not part of the identity condit ions of an event as they are in a Kimean account. Confusion arises if we then cite some of the properties to ground causal relations as we did in previous sections. For Davidson, an event is not a cause of another event in virtue of any of its properties Rather, the conditions. 40 What makes events stand in cause effect relations to each other is the fact that they fall under strict laws. 41 However, laws for David son are linguistic in nature. As such, events relate causally only under a specific description. 42 Since some descriptions differ radically events may relate causally under one des cription but not under another. So even if a causal statement such as the c ompleteness assumption (4) refers to physical cause of an even t, n othing 39 Davidson (1974 /1980, 230f. ) argues that there cannot be any bridge laws connecting the mental and the physical partly because of this the exclusion problem. Nonetheless, his anomalous monism is often used to establish the autonomy of psychological causal explanations from physical ones. Jackson (1996 380) presents Davidson as the main proponent of one of three important kinds of autonom y solu 40 See Davidson (1969/1980, 179). 41 See Davidson (1970/1980, 207). 42 Ibid., p. 215.
132 seems to follow for possible causes of the same event under a different description, such as a psychological one. In that sense, causal explanations involving the physical and the non physical are autonomous. Consequently Davidson would simply reject the exclusion assumption (6). Note, however, that this response is semantic and epistemological in nature. 43 It shifts the focus from what happens in the world to the way we talk about the world and how we explain what is going on. 44 It remains unclear what such a theory implies metaphysically. We could, for instance, argue that the very concept of causation is as relative to language as the other categories are and maintain that there is no such thing as causation on a metaphysical l evel i.e. we could be causal eliminativists. On the other hand, if we are already monists of some sort, we could also argue that although we distinguish between physical and mental events the events are really of one kind neither physical nor mental i n nature. But if they are, the problem of exclusion seems to vanish just as it vanished for the identity theorists in section 4.2. In sections 1.1 and 3.2, we hinted at a more promising strategy to put pressure on the exclusion assumption (6) We mentioned that it is often the conversational context that determines what kind of causal explanation we are after Menzies takes this very seriously and points out that the context often determines which property values are taken to be default values. These defaul t values may then provide the background against which we specify which properties of the effect we refer to in our causal explanations. Menzies gives the following example: patient 100 mg of penicillin cu red him the con trastive focus serves to hi g hl ight the fact that only certain dosages 43 44 This is a point frequently made against Davidson, or at least, if not intended as criticism, it is emphasized that Davidson 12, section 5.3 ).
133 within a range of possible dosages were causally effective. It suggests that we should take the cause variable to be a quantitative variable, which can take various values such as 0, 50 mg, 100 mg, 200 mg, and so on. And that the causal statement recovery rather than non recovery. 45 Since such assumptions often remain in the background of the conversational context, the contrastive structure of causal statements often is obscured. Rather than merely stating that c causes d we should state that c rather than d causes e rather than f Rather than stating that c caused d in virtue of property P c being P rather than P* causes event d Q rather than Q* Explicating the contrastive structure of causal statements can avoid confus ions and problems Consider the pigeon from section 4.3 which was trained to peck at red shapes According pecking. Interestingly, (a) is true whereas (b) is false. Had the shape not been red, the pigeon would not have pecked f or it. But had it not been scarlet but crimson instead, the pigeon might still have pecked at it, as long as it was a shade of red. 46 Without having to employ the distinction between causal relevance and causal sufficiency, as Yablo did, and merely relying on the notion of making a causal difference, we can see that determinates and determinables are not in causal competition. We thus may have reasons to doubt the exclusion assumption (6). 45 Menzies (2008, 208). Menzies account introduces several elements foreign to our terminological convention. For example, he talks about values rat idealized manipulation that sets the value of X by a causal process that is independent of all other possible causes of es causing each other, which is not in line with our convention. We will simplify his account by adjusting it to our terminology to make it applicable to our discussion more easily. 46 Ibid., p. 209.
134 While Menzies relies on the notion of mak ing a causal difference, his strategy may also t assumption (2), the following contrastive statement may be true: (a) a particular nonphysical k rather than properties of the non k kind (or physical properties. Similarly, according to the completeness assumption (4), the following contrastive statement ma nonphysical properties are those in virtue of which that event has a causal impact on a physical roperties. However, to apply Menzies contrastive ac cou nt properly we would also have to specify the properties of the effect as well as the contrast properties. The issues here seem explanandum strategy. It seems we must assume that t he properties of the effect in (a) must not be the same as the properties of the effect in (b), or else we the nonphysical cause in (a) faces exclusion. But if the properties to be explained must be nonphysical rather than physical, (2) and (3) might turn out to be false, at least if we understand the reference to the physical event as to causally. For that, (b) seems the better candidate. By the same token, if we spe cify that the properties of the effect in (2) are nonphysical, then this will restrict the properties that adequate physical causes can explain causally 4. 8 Chapter Conclusion The goal of this chapter was twofold. First, we wanted to see how some of the s tandard responses to the exclusion problem fare as a response t o the particular version we developed in Chapter 2 S econd, we wanted to check whether theories of causation also affect these responses in some way, especially given that Chapter 3 established that theories of causation indeed affect
135 the exclusion problem in various ways Because thousands of articles have been written about the exclusion problem, we were forced to be selective, but we tried to cover some of the most commonly discussed as well as some lesser known responses to cover a reasonable amount of territory Th ese were the main lessons : Eliminativist may deny the distinctness assumption (1) but at least one argument against distinctness does not work as it stands against our particu lar version of it If the main reasons for eliminating mental properties are that the theories citing them have not improved much over time and that they do not integrate well into the set of our most successful scientific theories, then many nonphysical properties that play a role in the special sciences remain untouched, for the special sciences have progressed and they integrate better. If those properties are to be eliminated for some other reason, we would have to reth ink the status of a large body of theories. Alternatively, the eliminativist can embrace a type identity theory for nonphysical properties. Adopting such a theory may be based on the hope that all the kinds of events and properties cited by the special sci ences will eventually be successfully reduced to physics. The prospect of such unity of science, however, may be dim, especially if there are reasons to think that scientific pluralism is true. Depending on how such pluralism is construed, it may imply tha t the completeness assumption (4) and the nonoverdetermination assumption (5) are false. While eliminativism and type identity theories are compatible with all theories of causation discussed in Chapter 3, scientific pluralism may not be compatible with so me regularity theories and INUS conditions Even worse, there may be a tension with probabilistic theories even though proponents of scientific pluralism sometimes explicitly endorse them Some philosophers read the causal impact assumption (2) in a weaker way They respond to the exclusion problem by distinguishing between causal relevance and causal sufficiency and
136 then claim that nonphysical events are causally relevant to but not causally sufficient for their effects. Unlike causal pluralism, these are not two distinct causal relations. Rather, the former is a grounding relation of the latter. Our version of the exclusion problem however, specifies that some nonphysical events have a causal impact, which is a production relation of sorts. Because of that, the strategy seems to amount to little more than a rejection of (2) and (3). Since these are the key assumptions of the problem, the strategy is only as good as the reasons for endorsing the suggested di stinction between causal relevance and causal sufficiency. It turns out that at least s for it are not convincing The distinction between structuring causes and triggering causes on the other hand, is more plausible in some respects but it may lead us to reject the completeness assumption (4). Furthermore, it may be unclear whether counterfactual theories of causation can capture the distinction. If the typicality assumption (3) is correct, then many nonphysical events that have a causal i mpact in virtue of their nonphysical properties are overdetermining causes. Any theory of causation that implies regularities will exacerbate the oddity of such a picture. Single difference theories may point out that assumptions (1) to (3) are compatible with there being atypical nonphysical causes that are not threatened by exclusion qua the nonoverdetermination assumption (5). But the price may be that citing such atypical causes loses all its explanatory force. The reasons for endorsing the completeness assumption (4) may be the same as those motivating another assumption, which is related to completeness but distinct from (4). That further assumption is not needed for the exclusion problem to arise, but it may have an impact on which of its assumptions we are likely to endorse or reject. In combination with a transference theory of causation, for example, it may imply that the causal impact assumption (2) is false. In
137 combination with a regularity theory of causation, to consider another example, it may force us to spell out the relation between nonphysical and physical events more to decide whether or not they face exclusion. Even if the o verdetermination assumption (5) is true, the threat of overdetermination might be averted by a dual explanandum strategy One such strategy construes causal relations as four place relations. Once we realize that causes rarely explain all properties of their effects, and once we concede that even adequate causes might not do so without ceasing to be adequate it is easy to see how some nonphysical events avoid overdetermination and exclusion. They explain different features of effects than adequate physical causes do. Depending on what we think these features are, the resulting view may be incompatible with some theo ries of causation such as the probability theory. And in combination with the transference theory, the strategy may imply rather implausible views such as parallelism. The exclusion assumption (6) suggests a picture according to which nonphysical events c ompete with physical events for their place in causal relations. Autonomy solutions question this picture. They may do it in the form of a dual explanandum strategy or in the form of scie ntific pluralism, but they may also do it based on views about langua ge or the theory dependence of descriptions. If the categories and distinctions that causal descriptions and explanations rely on are relative to frameworks, either of a semantic or of a theoretical sort, then the fact that there is a description according to which a physical event is an adequate cause of an event does not imply that descriptions according to which nonphysical events are causes of the same event are false, for the tension only arises within one framework or theory. However, it is hard to se e to what degree this response is ontological in nature, for language and theory dependence seem mostly epistemic issues. The most promising autonomy solutions may thus be
138 the dual explanandum strategy and scie ntific pluralism, for they are both ontologica l in spirit. Menzies contrastive account of causation, however, seems to have more interesting consequences. If we specify the properties of the effect to be explained by the physical and nonphysical event in the right way, we may avoid the causal exclusi on of nonphysical events and properties. However, the view does not seem helpful if we insist that nonphysical events sometimes explain some of the same properties of an effect that physical adequate causes explain.
139 CHAPTER 5 REASSESSING THE EX CLUSION PROBLEM 5.1 Striking a Balance Chapter 3 and Chapter 4 demonstrated that theories of causation affect both the exclusion problem and some of the standard responses to the problem Sometimes they affect the str ength of particular assumptions; at other times they affect the likelihood of rejecting particular assumptions. Since our version is importantly different from versions often discussed in the literature as illustrated in sections 2.1.2 and 2.2, we may worry that perhaps it is so strong tha t it includes a number of unreasonable assumptions. While our analysis serves to show that there is one version of the exclusion problem that is affected by theories of causation and to which some standard responses do not work well, it may not be a versio n we should endorse But if we should not, rejecting otherwise promising responses based on it would be misleading as well as unfair to the responses. To address these worries we will review a stronger and a weaker version of the exclusion problem. These versions involve assumptions that we have addressed in previous chapters We can then explain why it is reasonable to hold onto our version Doing so will provide a background against which to honestly assess the strengths and weaknesses of our version. Ho wever, w e will see that, all things consider ed our version seems to strike a good balance between specificity and strength on the one hand and the avoidance of unnecessary ontological commitments on the other hand The latter is especially important to al low for and to ease the discussion of the exclusion problem among philosophers who differ in their views on c ausation, completeness, events overdetermination, and the mind A version that accomplishes such balance thus seems particularly well suited for future research We will end this chapter and dissertation with a brief summary of the main findings and highlight some of the loose ends that may require attention in later projects
140 5.1.1 Strong Exclusion In previous chapters, we encountered various versions of the assumptions of the exclusion problem. They differed in the way they were phrased, in their level of specificity, in their strength, and in their ontological commi tments. Let us look at a particularly strong version that we may construct based on some of the assumptions we have looked at. A. Distinctness: Nonphysical properties are distinct from physical properties B. Causal Sufficiency : Sometimes nonphysical events ar e causally sufficient for physical events C. Completeness: Every physical event has a sufficient physical cause D. Nonoverdetermination: Events are not overdetermined E. Exclusion: If an event has a sufficient caus e, it has no further cause T his version seems m uch simpler than our version. It is easy to see how it may be used as an argument against (B). If every physical event has a sufficient physical cause, and if events do not have more than one sufficient cause because there is no overdetermination, then nonphysical events cannot be causally sufficient for physical events The argument also appears to be valid. While b oth simplicity and validity are virtues, these virtues do n ot have much weight unless we have good reasons for endorsing the premises. Unfortunately, there are very good reasons for rejecting several of them. Assumption (A), a version of which we discussed in section 2.2, is extremely general and because of that d oes not allow for any nuance. For example, we may endorse a type identity realization physicalism of the sort discussed in 4.2 for many nonphysical properties cited by various special sciences, but we may want to reject it for particular kinds of mental pr operties (A) would force us to treat them in the same way But why should we accept that? By contrast, our distinctness assumption (1) allows f or such nuanced discussion of particular kinds of
141 nonphysical properties, which seems particularly important in the context of a generalized problem. In section 4.5, we discussed a version of (B) The first thing to notice is that sufficiency is such a modal notion. Not all theories of causation posit such modal relation between cause and effect. In fact, theories such as the nave regularity theory in section 3.3.1 were specifically designed to avoid modality. Thus understood, (B) rules out the application of some theories of causation to the exclusion problem. At the very least, (B) makes strong er claims about how cause and effect are related than some kinds of regularity theories, probabilistic theories, and single difference theories. But if the problem rests on positing a relation that is stronger than these theories posit, what reason would a proponent of such a theory have to endorse it? If he or she has to give up one assumption to ensure consistenc y among the others, (B) may be a promising candidate. Since probabilistic theories of causation are often considered to be those that are consist ent with current science, we should make sure to phrase the exclusion problem in a way that is consistent with the application of such theories. Assuming that a sufficient cause is one that raises the probability of another event to 1.0, i.e. assuming that sufficiency is a special case of probabilistic causation, the problem is not particularly severe from the viewpoint of such a theory, for it does not capture most instances of causation Our causal impact assumption (2) managed to do that wit hout undermining the validity of the argument and the threat of exclusion But it required, especially in the light of a weaker nonoverdetermination assumption the introduction a term of art, causal impact, and the typicality assumption (3) which has n o equivalent in the strong version. So we h ad to trade some simplicity for a higher degree of compatibility with theories of causation.
142 In sections 2.1.2 and 4.5 we mentioned a particularly strong version of the completeness assumption along the lines of (C). The problem is that even a physicalist is not committed to that. After all, our best physical theories posit that quantum fluctuations are uncaused. Since these are not events in the ordinary sense of the word, they may not count as proper counterevi dence. But cosmological theories are compatible with the claim that the big bang was not caused, for example. 1 By contrast, our completeness assumption (4) accommodates this nicely. In section 2.1.2 we addressed a nonoverdetermination assumption along the lines of (D) (D) may be true. As we saw in several sections of Chapter 3, several theories of causation struggle with providing a coherent and plausible analysis of genuine overdetermination cases. However, some theories of causation are at least compati ble with there being nonsystematic cases. The question then is whether we should exclude isolated ins tances by turning (5) into (D). The answer should not be based on what our stance on overdetermination is but on the role the assumption plays in the pro blem and in resulting arguments. As a general rule for gument to work, but no stronger are count erexamples. Minimizing our ontological commitments is a good strategy for the very same reason. Finally, (E) is implausibly strong. Even proponents of INUS conditions point out that causes on their own rarely, if ever, suffice. An event having more than two causes is probably the rule rather than the exception. Of course, that does not negate (E), since it is phrased as a conditional. But if we follow the earlier advice that (B) should be modified so that it allows for 1 Montero (2003, 174) uses the big bang as an example as well.
143 probabilistic caus ation and we should go with a weaker nonoverdetermination assumption (5) rather than (D), (E) needs to be adjusted accordingly. It seems, then, that we should keep the assumptio ns strong enough for them to serve as premises in a valid deductive exclusion a rgument. But they should not be so strong that they face counterexamples, eit her conceptual or empirical The assumptions (1) to (6) in our version are weaker than the assumptions (A) to (E) in the strong version. However, we may ask whether perhaps we sho uld endorse yet weaker assumptions. 5.1.2 Weak Exclusion While adopting weaker assumptions may be beneficial in that they may require fewer ontological commitment s they may also undermine the potential validity of the problem when it is used to construct an exclusion argument If, for example, we weakened all the sufficiency claims by replacing them with causal relevan cy, the problem seems to disappear : A. Distinctness: Nonphysical properties are distinct from physical properties B. Causal Sufficiency : Sometime s nonphysical events are causally relevant to physical events C. Completeness: For every physical event there is a causally relevant physical event. D. Nonoverdetermination: Events are not overdetermined E. Exclusion: If there already is a causally relevant event to an other event there is no further causally relevant event Typically, to say that an event is causally relevant is taken to be a weaker claim than one about its causal impact or causal sufficiency. For example even if we think that the presence of oxygen is a mere background condition rather than a partial cause of an explosion, we still may say it was causally relevant. Without the oxygen, the explosion might not have happened, after all. But if that is right the exclusion assumption (E) seems clearly false, for it implies that the cause could not have been relevant in the presence of the background conditions. That makes
144 little sense. But if that is right, then there seems to be little reason for thinking th at nonphysical events cannot be causally relevant whenever there is a physical event that is also relevant. That is, the problem ceases to be one. One significant difference between both the weak and the strong versions, on the one hand, and ours, on the other hand seems to be that our version is much more specific. It relies heavily on in virtue of talk to specify the properties that ground the causal relations. Notice that in neither the strong nor the weak version is it clear what feature of an event e xplains that it is a cause of the effect. It may be suggested that it is the physical features of physical events and the nonphysical features of nonphysical events, but that does not follow from anything actually stated. Inheritance solutions sometimes ex ploit such vagueness in pointing out that nonphysical events may inherit their causal powers from their physical realization base. That may be so, but notice how, once we specify the properties in virtue of which events cause their effects, it becomes much clearer what that strategy amounts to. For nonphysical events and properties to inherit their causal powers from physical events and properties, there has to be a rather tight relation between the nonphysical and physical properties in question. If the strategy is merely to postulate that nonphysical events have a causal impact in virtue of their physical properties, it could not serve as a response to our version of the problem. It seems then, that trading further simplicity for the sake of explanatory power is reasonable. Doing so will clarify the connections between the assumptions and it will allow us to explain better how responses to the exclusion problem are supposed to work. Of course, none of this establishes that our version is the only one av ailable or even the one that best achieves a good balance between specificity and strength on the one hand and the
145 avoidance of unnecessary ontological commitments on the other hand. But it does seem like our version is at least one of them. 5.2 Conclusion We began this project because we were puzzled by how few philosophers analyzed and utilized theories of causation in the context of the exclusion problem, even though most of its assumptions include causal claims. It seemed that there is an implicit, and sometimes even an explicit, assumption that theories of causation do not affect the problem at all While a few philosophers have tried to use the counterfactual theory in particular to develop a response or a solution to the problem, to date there is no s ystematic survey that would justify the assumption that theories of causation generally do not matter This dissertation had the rather hum ble goal to investigate th is assumption. Had it turned out that in fact theories of causation have no impact on eithe r the exclusion problem itself or on how we respond to it, we would have provided support for an assumption that seems to underlie the debate. But, fortunately our investigation revealed quite the opposite. There are many ways in which theories of causati on affect the problem. To the extent that we do not have any obvious reason s to reject a particular assumption of the exclusion problem we have to weigh their plausibility against each other to decide which assumptions should be rejected in order to resolve the tension s or inconsistenc ies among the m As it turns out, theories of causation often affect how invested we should be in h old ing onto particular assumptions. If, for example, we presuppose a regularity theory of causation, we cannot allow for a typical causes. So, the typicality assumption, under such theory, cannot be given up since it is implied by the theory. Single difference theories, on the other hand, may accept the assumption, but they will not be as invested in it, for they can handle at ypical causes quite well.
146 At o ther times, a theory may require stronger assumptions than others do For example, theories that run into inconsistencies when analyzing overdetermination cases, as some kinds of probabilistic theories do might not only endo rse the nonoverdetermination assumption (5) but also may even imply that it is not just systematic overdetermination that is a problem. No overdetermination exists. That in turn has consequences for how to think about the severity of the problem as a whole As we have set up the problem, it allows for atypical nonphysical causes that nonsystematically overdetermine physical events. That is, the scope of the problem is somewhat limited for it does not capture a certain set of events But from the viewpoint of a theory of causation that does not allow for any overdetermination, the exclusion problem covers all cases of nonphysical causation a significant difference. As the basis for the discussion of the possible impact of theories of causation on the exclusion problem, we first carefully developed our own version of it, since the versions discussed in the literature were often either too strong or too weak. Once our version was on the table and once it w as clear that and how theories of causation affect the exclusion problem, the question whether any of this would impact standard responses to the problem naturally arose. Again, the answer was affirmative. The vers ion of the problem that we developed in Ch apter 2 indeed makes it more difficult for some of the responses to succeed. Yablo, for example, endorsed a much weaker version of the causal impact assumption (2). Partly because of that, his response could not convince as a response to our version. Other responses turned out to work particularly well in combination with p articular theories of causation but not so well in combination with others Horgan place relation, for example, seems to be particularly compatible wit h counterfactual theories or causal pluralism but it might not work well with probabilistic theories or transference theories.
147 In spite of these positive results, there is much we did not accomplish On the one hand, we tried to discuss a representative s et of theories of causation, partly by analyzing the impact of some of the most popular theories of causation as well as the impact of some more obscure theories. The fact that all of them had some kind of impact on the exclusion problem may justify the co njecture that perhaps all theories of causation do On the other hand, we had to neglect some of the most promising theories of causation available. For example, we pointed out that some of the problems that a simple probabilistic theory of the sort we dis cussed in section 3.3.4 faces might be dealt with by combining it with another theory or by adding further assumptions to build a more complex theory that can handle problematic cases of overdetermination better. Of course, that is exactly what proponents of probabilistic theories have often done. One reason why we did not pursue this further is that in many cases these theories diverge very far from theory, for example, treats proc esses as more fundamental than events. But that complicates an application of it to the exclusion problem as we have construed it. One project could thus be to look at such promising theories of causation and analyze their impact on the exclusion problem a s well. Investigating theories of causation not yet covered may have a further benefit. As just pointed out, s ome of these theories consider causal relata other than events T his may give us the opportunity to test a suspicion we voiced in section 2.1.1, namely that different causal relata may have an impact on the problem that events do not have It may seem as if translations of statements involving one type of causal relatum into statements involving another kind are easy to come by. But, as we sugges ted such translations may be deceptive Perhaps other types of causal relata woul d generate new kinds of problems In this context, it also is worth reminding ourselves that one shortcoming of our investigation is that, in order to cover a good number of
148 theories of causation and responses to the problem, we were bound to be superficial in our analysis. Looking at particular theories in more detail might thus reveal further unexpected ways in which different types of causal relata as well as particular the ories of causation affect the problem. In section 3.3.8, we sketched the beginning of a nomological version of the exclusion problem. Since many reductive theories of causation rely on the concept of a law of nature, developing such a nomological exclusio n problem in detail should prove fruitful not just for a deeper analysis of the impact of reductive theor ies of causation on the problem but also for exploring might co nnect better to the problem of free will. Simply put, the problem is whether it is possible for free will to exist in a deterministic world. Determinism is usually introduced as the thesis that, given any state of the universe at a given time and the laws of nature, any subsequent state follows necessarily. To articulate t his problem, the notion of causation is not needed. A nomological version of the exclusion problem will thus make it easier to discuss interesting and important connections between the problem of free will and the exclusion problem. While completeness, law s of nature, overdetermination, and properties played an important role in our analysis, we did not systematically explore how different theories of them may affect the problem. Ultimately, a thorough analysis of the exclusion problem should do so Thus, t his dissertation is best understood as the first step towards a comprehensive primer on the exclusion problem. As of now, no such primer exists but it should. Regardless of which of these projects we may pursue first, having developed a version of the pro blem that avoids some of the pitfalls and unnecessary ontological commitments of other versions will serve as a good starting point for future research.
149 APPENDIX KANT EXCERPT The following passage Critique of p ure r eason could be inter preted as fusing questions concerning the compatibility of free will and determinism with problems regarding causal exclusion. [T]hus the difficulty we encounter in the question about nature and freedom is only whether freedom is possible anywhere at all, and if it is, whether it can exist together with the universality of the natural law of causality, hence whether it is a correct disjunctive proposition that every effect in the world must arise either from nature of freedom, or whether instead both each in a different relation, might be able to take place simultaneously in one and the same occur r ence. The correctness of the principle of the thoroughgoing connection of all occurrences in the world of sense according to invariable natural laws is already co nfirmed as a principle of the transcendental analytic and will [not] suffer violation. Thus the only question is whether, despite this, in regard to the very same effect that is determined by nature, freedom might not also take place, or is this entirely e xcluded through that inviolable rule? And here the common but deceptive presupposition of the absolute reality of appearance immediately shows its disadvantageous influence for confusing reason. For if appearances are things in themselves, then freedom can not be saved. Then nature is the completely determining cause, sufficient in itself, of every occurrence, and the condition for an occurrence is always contained only in the series of appearances that, along with their effect, are necessary under the law o f nature. 2 2 Kant (1781/1997 A536/B564 ) which we added in brackets. The German version is clear on this.
150 LIST OF REFERENCES Anscombe, G. E. M. (1971/19 9 3). Causality and determination. In E. Sosa & M. Tooley (Eds.), Causation Oxford: Oxford University Press. Armstrong, D. M. (1968/1993). A materialist theory of the mind (Rev. ed.). Lond on: Routledge. Armstrong, D. M. (1983). What is a law of nature? Cambridge: Cambridge University Press. Synthese 22 (3 4), 414 30. Baker, L. R. (1993). Metaphysics and mental causation. In J. Heil & A. Mele (Eds.), Mental causation Oxford: Clarendon Press. Beebee, H. (2006). Does anything hold the universe together? Synthese, 149 (3) 509 533. Beebee, H., Hitchcock, C., & Menzies P. (Eds.) (2009). The Oxford handbook of causation Oxford : Oxford University Press. Bennett, J. (1988). Events and their names Oxford: Clarendon Press. Bennett, K. (2008). Exclusion a gain. In J. Hohwy & J Kallestrup ( Eds.), Being r educed: New e ssays on r eduction, e xplanation, and c ausation Oxford: Oxford University Press. Bunzl, M. (1979). Causal overdetermination. The Journal of Philosophy, 76 (3), 134 150. Burge, T. (1993). Mind body causation and explanatory practice. In J. Heil & A. Mele (Eds.), Mental causation Oxford: Clarendon Press. Cartwright, N. (2004). Causation: One word, many things. Philosophy of Science, 71 (5), 805 820. Cat, J. (2013). The unity of science. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy ( Spring 201y Edn.) http://plato.stanford.edu/ entries/scientific unity/ Cell membrane. ( 2017 ) In Wikipedia https://en.wikipedia.org/wiki/Cell_membra ne / Chisholm, R. (1964 /2003 ). Human freedom and the self. In G. Watson (Ed.), Free will ( 2 nd ed ) Oxford: Oxford University Press. Churchland, P. M. (1981). Eliminative materialism and the propositional attitudes. The Journal of Philosophy, 78 (2), 67 90. Collins, R. (2008). Modern physics and the energy conservation objection to mind body dualism. American Philosophical Quarterly, 45 (1), 31 42. Crane, T. (1995). The mental causation debate. Proceedings of the Aristotelian Society 69 211 236.
151 Crane, T. & Mellor, D. H. (1990). There is no question of physicalism. Mind, 99 (394), 185 206. Davidson, D. (1969 /1980 ). The i ndivid uation of events. Essays on actions and events Oxford: Clarendon Press. Davidson, D. (1970 /1980) Mental events. Essays on actions and events Oxford: Clarendon Press. Davidson, D. (1974/1980). Psychology as philosophy. Essays on a ctions and e vents Oxford: Clarendon Press. Davidson, D. (1987). Problems in the explanation of action. In P. Pettit, R. Sylvan, & J. Norman (Eds.), Metaphy sics and morality: Essays in honour of J. J. C. Smart Oxford: Basil Blackwell. Davidson, D. (1993). Thinking causes. In J. Heil & A. Mele (Eds.), Mental causation Oxford: Clarendon Press. Dennett, D. C. (1973). Mechanism and responsibility. In T. Honderi ch (Ed.), Essays on freedom and action London: Routledge & Kegan Paul. Dennett, D. C. (1991). Consciousness explained Boston: Little, Brown, & Company. Descartes, R. ( 1641 / 1996). Meditations on f irst p hilosophy (J. Cottingham (Ed. & Trans.) ). Cambridge: Cambridge University Press. Dretske, F. (1988). Explaining behavior: Reasons in a world of causes Cambridge: MIT Press. Dretske, F. (1993). Mental events as structuring causes of behavior. In J. Heil & A. Mele (Eds.), Mental causation Oxfo rd: Clarendon Press. Ducasse, C. J. (1926/1993). On the nature and the observability of the causal relation. In E. Sosa & M. Tooley (Eds.), Causation Oxford: Oxford University Press. Ducasse, C. J. (1969). Causation and types of necess ity. New York: Dover. Dupr, J. (1993). The disorder of things: Metaphysical foundations of the disunity of science Cambridge: Harvard University Press. Earman, J. (1976). Causation: A matter of life and death. Journal of Philosophy, 73 (1), 5 25. Esfeld, M. (2007). Mental causation and the metaphysics of causation. Erkenntnis, 67 (2) 207 2 20. Evnine, S. (1991). Donald Davidson Stanford: Stanford University Press. Falke, A. ( 2006 ). Zur kausalen Relevanz hherstufiger Eigenschaften: Eine essentialistische Theorie der Verursachung als Strategie gegen das Argument der Exklusion (Unpublished m agister thesis). Goethe University Frankfurt, Germany.
152 Fodor, J. A. (1981). The mind body problem. Scientific American, 244, 114 123. Funkhouser, E. ( 2002). Three varieties of causal overdetermination. Pacific Philosophical Quarterly, 83 (4) 335 3 51. Garrett, D. (2009). Hume. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Gendler, T. S. (1998). Exceptional persons: On the limits of imaginary cases. Journal of Consciousness Studies, 5 (5 6), 592 610. Analysis, 61 (4), 302 309. Godfrey Smith, P. (2009). Causal pluralism. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Goldman, A. (1969). The compatibility of mechanism and purpose. Philosophical Review, 78 468 482. Grice, P. (1989). Studies in the way of words Cambridge: Harvard University Press. Hall, N. (2004). Two concepts of causation. In J. Collins, N. Hall, & L. Paul (Eds.), Causation and counterfactuals Cambridge: MIT Press. Harbecke, J. (2008). Mental causa Frankfurt: Ontos. Haug, M. C. (2009). Two kinds of completeness and the uses (and abuses) of exclusion principles. Southern Journal of Philosophy, 47 (4), 379 401. Healey, R. (2009). Causation in quantum mechanics. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Hendrickson, N. (2006). Towards a more plausible exemplification theory of events. Philos ophical Studies, 129 (2) 349 375. Horgan, T. (1978). Supervenient bridge laws. Philosophy of Science 45(2), 227 249. Horgan, T. (1989). Mental quausation. Philosophical Perspectives, 3 47 76. Horgan, T. (1997). Kim on mental causation and causal exclusion. Philosophical Perspectives, 11 (s11), 165 84. Horst, S. (2007). Beyond reduction: Philosophy of mind and post reductionist philosophy of science Oxford: Oxford University Press. Horwich, P. ( Causation Oxford: Oxford University Press.
153 Humphreys, P. (2009). Causation and reduction. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford Unive rsity Press. Jackson, F. (1996). Mental Causation. Mind, 105 (419), 377 413. Kallestrup, J. (2006). The causal exclusion argument. Philosophical Studies, 131 (2), 459 4 85. Kane, R. (2005). A contemporary introduction to free will Oxford: Oxford University Press. Kant, I. ( 1781 / 1997). Critique of pure reason (P. Guyer & A. W. Wood (Eds. & Trans. ). Cambridge: Cambridge University Press. Kim, J. (1976 /1993 ). Events as property exemplifications. Supervenience and mind: Selected philosophical essays Cambridge: Cambridge University Press. Kim, J. (1989 /1993 ). Mechanism, purpose, and explanatory exclusion. Supervenience and mind: Selected philosophical essays Cambridge: Cambridge University Press. Kim, J. (1993). The non troubles with mental causation. Supervenience and mind: Selected philosophical essays Cambridge: Cambridge University Press. Kim, J. (1997). Does the problem of mental causation g eneralize? Proceedings of the Aristotelian Society, 97 (3) 281 97. Kim, J. (1998). Mind in a p hysical w orld Cambridge: MIT Press. Kim, J. (2003). Philosophy of mind and psychology. In K. Ludwig (Ed.), Donald Davidson Cambridge: Cambridge University Press. Kim, J. (2005). Physicalism, or something near enough Princeton : Princeton University Press. Kim, J. (2006). Philosophy of mind 2nd E d. Boulder: Westview. Kincaid, H. (2009). Causation in the social sciences. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Lange, M. (2009). Causation in classical mechanics. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Levin, J. (2013). Functionalism. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy ( Winter 2016 Edn.) http://plato.stanford.edu/ entries/functionalism / Lewis, D. (1973a). Causation. Journal of Philosophy, 70 (17), 556 567. Lewis. D. (1973 b ). Counterfactuals Cambridge: Harvard University Press. Nos, 13 (4), 455 476. Lewis, D. (1986). On the plurality of worlds Oxford: Basil Blackwell.
154 Lipton, P. (2009). Causation and explanation. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. List, C. & Menzies, P. (2017). My brain made me do it: The exclusion argument against free will, and what's wrong with it. In H. Beebee, C. Hitchcock, & H. Price (Eds.) Making a difference: Essays on the p hilosophy of c ausation Oxford: Oxford University Press. Lombard, L. B. (1990). Causes, enablers, and the counterfactual analysis. Philosophical Studies, 59 (2), 195 211. MacIntyre, A. C. (1957). Determinism. Mind 66 (261) 28 41. Mackie, J. (1965). Causes and conditions. American Philosophical Quarterly, 2 (4), 245 264 Malcolm, N. (1968). The conceivability of mechanism. Philosophical Review, 77 45 72. Maslen, C. Horgan, T., & Daly, H. (2009). Mental c ausation. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Ox ford University Press. Maudlin, T. (2007). The metaphysics within physics New York: Oxford University Press. In J. Heil & A. Mele (Eds.), Mental causation Oxford: Clarendon Press Journal of Philosophy, 94 (12) 622 637. Melnyk, A. (2003). A physicalist manifesto: Thoroughly modern materialism Cambridge: Cambridge University Press. Menzies, P. (2008). The exclusion problem, the determination relation, and c on trastive Being reduced: New essays on reduction, explanation, and causation Oxford: Oxford University Press. Menzies, P. (2009). Platitudes and counterexamples. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Montero, B. (2003). Varieties of causal closure. In S. Walter & H. D. Heckmann (Eds.), Physicalism and mental causation: The metaphysics of mind and action Exeter: Imprint Academic. Moore, M. (2009). Introduction: The nature of singularist theories of causation. The Monist, 92 (1), 3 22. Nagel, E. (1961). The structure of science: Problems in the logic of explanation New York: Harcourt, Brace & World. /2003 ). Agent causation. In G. Watson (Ed.), Free will 2 nd edition. Oxford: Oxford University Press.
155 Papineau, D. (1996). A universe of zombies? The problem of consciousnes s and the temptations of dualism. Times Literary Supplement 21, 3 4. Papineau, D. (2002). Thinking about consciousness Oxford: Clarendon Press. Paul. L. A. (2009). Counterfactual theories. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Pearl, J. (2000). Causality: Models, reasoning, and inference Cambridge: Cambridge University Press. Plantinga, A. (1976). Actualism and possible worlds. Theoria, 42 (1 3 ), 139 160. Price, H. & Corry, R. (2007a). A case for causal republicanism? In H. Price & R. Corry (Eds.), Oxford: Oxford University Press. Price, H. & Corry, R. (Eds.). (2007 b ). Causat ion, physics, and the constitution of reality: Oxford: Oxford University Press. Psillos, S. (2002). Causation and explanation Montreal: McGill Psillos, S. (2009). Regularity theories. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Pu tnam, H. (1967). Psychological p redicates. In W. H. Capitan & D. D. Merrill (Eds.), Art, Mind, and Religion Pittsburgh: University of Pittsburgh Press. Ramsey, F. P. (1928/1978). Universals of law and of fact. In D. H. Mellor (Ed.), Foundations: Essays in philosophy, logic, mathematics and economics London: Routledge & Kegan Paul. Robb, D. & Heil, J. (2013). Mental causation. In E. N. Zalta (Ed.), The S tanford encyclopedia of philosophy (Spring 2014 Edn.) http://plato.stanford.edu/entries/mental causation/ Robinson, W. (2015). Epiphenomenalism. In E. N. Zalta (Ed.), The Stanford encycl opedia of philosophy (Fall 2015 Edn.) http://plato.stanford.edu/entries/epiphenomenalism/ Rosen, D. (1978). In defense of a probabilistic theory of causality. Philosophy of Science, 45 (4 ) 604 613. Russell, B. (1912 13). On the notion of cause. Proceedings of the Aristotelian Society, 13 1 26. Salmon, W. (1984). Scientific explanation and the causal structure of the world Princeton: Princeton University Press. Sartorio C. (2009): Causation and ethics. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press.
156 Schaffer, J. (2003a). Overdetermining causes. Philosophical Studies, 114 (1 2), 23 45. Schaffer, J. (2003b). Is there a fundamental level? Nos, 37 (3), 498 517. Schaffer, J. (2008). Causation and laws of nature: R eductionism. In T. Sider, J. Hawthorne, & D. W. Zimmerman (Eds.), Contemporary debates in metaphysics Malden: Blackwell. Schaffer, J. (2016). The metaphysics of causation. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Fall 2016 Edn.) http://plato.stanford.edu/entries/causation metaphys ics/ Shoemaker, S. (1980 /1997 ). Causality and properties. In D. H. Mellor & A. Oliver (E ds.) Properties Oxford: Oxford University Press. Philosophy and Phenomenological Research, 67 (3) 719 726. Simons, P. (2003). Events. In M. J. Loux & D. W. Zimmerman (Eds.), The Oxford handbook of metaphysics Oxford: Oxford University Press. Sklar, L. (2009). Causation in statistical mechanics. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbo ok of causation Oxford: Oxford University Press. Skyrms, B. (1984). EPR: Lessons for metaphysics. Midwest Studies in Philosophy, 9 (1), 245 255. Sosa, E. & Tooley, M. (1993a). Introduction. In E. Sosa & M. Tooley (Eds.), Causation Oxford: Oxford Universit y Press. Sosa, E. & Tooley, M. (Eds.) (1993 b ). Causation Oxford: Oxford University Press. Stoljar, D. (2008). Distinctions in distinction. In J. Hohwy & J. Kallestrup (Eds.), Being reduced: New essays on reduction, explanation, and causation Oxford: Oxford University Press. Stoljar, D. (2010). Physicalism London: Routledge. Stoljar, D. (2015). Physicalism. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Spring 2016 Edn.) h ttp://plato.stanford.edu/entries/physicalism/ Stltzner, M. (2009). The logical empiricists. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Strawson, G. (1987). Realism and causation. Ph ilosophical Quarterly, 37 (148), 253 277. Suppes, P. (1970). A probabilistic theory of causality New York: Basic Books. Suppes, P. (1984). Conflicting intuitions about causality. Midwest Studies in Philosophy, 9 (1), 151 168. Tienson, J. (2002). Higher order causation. Grazer Philosophische Studien, 63 (1) 89 101. Yablo, S. (1992a). Cause and essence. Synthese, 93 (3), 403 49. Yablo, S. (1992b). Mental causation. Philosophical Review, 101 (2), 245 80. Van Gulick, R. (1992). Three b ad a rguments for i ntentional p roperty e piphenomenalism. Erkenntnis 36 ( 3 ), 311 331.
157 Vigen, T. (2015). Spurious correlations New York: Hachette Books. von Wright, G. H. (1971). Explanation and understanding Ithaca: Cornell University Press. Williams, B. (1970). The self and the future. Philosophical Review, 79 (2), 161 180. Williamson, J. (2009). Probabilistic theories. In H. Beebee, C. Hitchcock, & P. Menzies (Eds.), The Oxford handbook of causation Oxford: Oxford University Press. Wilson, J (2017). Determinables and determinates. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Spring 2017 Edn.) http://plato.stanford.edu/entries/determinate determinables/ Witmer, G. (2003). Dupr essentialist objection to reductionism. The Philosophical Quarterly, 53 (211), 181 200. Wittgenstein, L. (1953). The philosophical investigations (G. E. M. Anscombe (Trans.)). Oxford: Basil Blackwell. Yalowitz S. (2012). Anomalous monism. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy ( Winter 2014 Edn.) http://plato.stanford.edu/entries/anomalous monism/ Zele k, E. (2009). T wo approaches to event ontology. Organon F, 16 (3), 283 303. Zhong, L. (2011). Can counterfactuals solve the exclusion problem? Philosophy and Phenomenological Research, 83 (1), 129 47.
158 BIOGRAPHICAL SKETCH A ndreas Falke was first exposed to the exclusion problem as an undergraduate at the Goethe U niversity in Frankfurt, Germany, where he majored in philosophy and minored in historical ethnology and religious studies He became so intrigued by the problem that h e ended up writing his thesis on it. In 2006, he graduated with a Magister degree. Andreas then attended the philosophy graduate progr am at the University of Florida degree in 2010, just to get back to studying the exclusion problem in more detail. In his dissertation Excluding thoroughl y: Theories of causation and their impact on the exclusion problem Andreas presented the results of his inquiry. He received a docto rate in philosophy in Fall 2017.