Introduction
We regularly talk about omissions, that is, what people did not do or failed to do. In some cases, we are indifferent between describing what someone did as an omission or as an action. For example, ‘Ali's staying still’ and ‘Ali's not moving’ seem to pick out the same event. It is thus tempting to hold that by saying ‘Ali's not moving,’ we are referring not to an absence or negative action, but simply to Ali's staying still. In other cases, such as ‘Ali's failure to take her medication’ or ‘Ali's abstaining from alcohol,’ we are reluctant to equate an omission with an action, or some positive event. According to what I call the identity view, we should surmount this reluctance and identify all omissions with actual events. More carefully, on this view, a nominal such as ‘Ali's non-jogging’ denotes an event (rather than the absence of an event) under a negative description: it denotes whatever Ali is doing at the time she is said not to be jogging—say, her walking (Davidson Reference Davidson, Vermazen and Hintikka1985; Payton Reference Payton2018; Schaffer Reference Schaffer and Hitchcock2004, Reference Schaffer2012; Varzi Reference Varzi, Bottani and Davies2006, Reference Varzi, Castellani and Quitterer2007; Vermazen Reference Vermazen, Vermazen and Hintikka1985).
The identity view has a number of virtues, which I describe in the next section while explaining the view further. However, my goal in this essay is not to argue for the identity view, but to defend it against one major objection. Opponents of the view claim that omissions cannot be actions, since the two do not have the same causal relations. I call this the causal problem. (For other objections against the identity view, see Bernstein [Reference Bernstein2014: 4–5]; Clarke [2012: 135–36]; Moore [Reference Moore2009: 438–39].) In section 2, I explain this problem, illustrating it with an example. As I show in section 3, any view according to which events have many names faces a similar causal problem. In that section, I propose a general solution to the causal problem. According to this solution, the relata of a causal relation are events. However, the same event could be described in different ways: different descriptions pick out different properties of the event. When the property picked out by the predicate F is not causally relevant to the effect, the claim ‘The F causes the G’ is odd. I explain how we can account pragmatically for the claim's oddness while maintaining its truth. In section 4, I show how this solution can be applied to omissions, and in section 5, I address a related puzzle.
1. The Identity View
I first explain briefly what motivates the identity view and present some of its advantages. Although each of my points deserves further scrutiny, my aim here is merely to clarify the identity view and show that it has strong prima facie appeal.
First, as noted in the introduction, we often speak of agents’ omissions as if they actually happened (Davidson Reference Davidson, Vermazen and Hintikka1985; Vermazen Reference Vermazen, Vermazen and Hintikka1985). For example, we say, ‘Those in attendance witnessed his not standing for the national anthem.’ In this case, it is natural to construe ‘his not standing’ as picking out an actual event, namely his kneeling, under a negative description. Speakers often use negative and positive descriptions interchangeably. We could describe her staying as her ‘not leaving’ and her leaving as her ‘not staying’ (Varzi Reference Varzi, Bottani and Davies2006: 133). Granted, it would be odd to describe his kneeling as, say, ‘his not dancing.’ But identity theorists can respond that the oddness is due to the fact that he was not supposed or expected to dance. By contrast, we have no problem describing his kneeling as a ‘non-standing,’ for he was supposed (or expected) to stand.
Second, events are particulars: they are unrepeatable entities located in space and time. So are individual things and persons. In appropriate contexts, different definite descriptions would designate the same person: ‘Tad's best friend,’ ‘the winner of the race’ and ‘the woman drinking a martini’ would all pick out Ali. And in the right context, one could even designate Ali by means of a negative description such as ‘the non-smoker.’ Note that since the above definite descriptions are all incomplete, context (including, especially, the speaker's intention) is needed to fix what they designate. Accounting for the specific mechanisms by which an incomplete definite description successfully picks out an individual is a matter of debate; however, there is no reason to think that the account will be substantially different when the incomplete definite description is negative. The same is true of event nominals, according to the identity view: ‘Ali's walk,’ ‘Ali's brisk walk’ and ‘Ali's non-jogging’ could all pick out the same action under different descriptions. The identity view thus allows us to offer a uniform treatment of definite descriptions, whether they pick out events or things.
Third, the identity view offers a simple account of omissions that avoids any commitment to a problematic ontology of absences, nonexisting things, or possibilities. On this view, omissions are actual events. Hence, once we have provided a metaphysical account of actions and events, no special addendum is needed to deal with omissions. (See Bernstein [Reference Bernstein2014] for a defense of the view that omissions are de re possibilities of actual events; see Silver [Reference Silver2018] for the view that omissions are fined-grained events exemplifying omissive properties.)
It is natural to take identity theorists to be committed to the metaphysical view that actions and events are coarse-grained. Most identity theorists, including me, embrace this commitment: Ali's walk, Ali's brisk walk and Ali's non-jogging are one and the same event. Some of the advantages I just mentioned could certainly be invoked in support of the coarse-grained view. For example, it is natural not only to construe ‘his not standing’ as picking out the same event as ‘his kneeling,’ but also to hold that his not standing is one and the same thing as his kneeling. However, it is worth noting that the identity view is primarily a semantic view about what nominals designate rather than a metaphysical thesis concerning the nature and individuation of actions and events. Once again, the analogy with definite descriptions will help. Earlier, I wrote that the same person, say, Ali, could be designated by different definite descriptions, such as ‘Tad's best friend,’ ‘the woman drinking a martini’ and ‘the non-smoker.’ But this point about designation is independent of metaphysical issues about the nature of persons and personal identity. Similarly, we should not regard the identity view as being strictly committed to a particular position about the nature of events.
The identity view is best construed as an extension of a general view according to which the same event can be designated by different nominals. Call this the many names view. According to the many names view, the nominals ‘Ali's run,’ ‘Ali's exercise session’ and ‘Ali's favorite moment of the day’ would all pick out the same event. Jonathan Bennett (Reference Bennett1988, Reference Bennett and Gale2002) cautions about drawing metaphysical conclusions from semantic considerations about event names. He holds that events are property exemplifications, or tropes. This metaphysical position, he points out, is perfectly compatible with the many names view: ‘Although each event is a trope, I contend, an event name (‘the sparrow's fall,’ ‘his assault on her’) need not wear in its face every detail of the trope that it names’ (Reference Bennett and Gale2002: 47). Bennett distinguishes between a partial describer of an event (‘his assault on her’) and a complete describer the same event, which would fully describe the property that the event is an exemplification of. As Bennett puts it (Reference Bennett and Gale2002: 47), the complete describer would be extremely detailed and start like this: assaulting by kicking hard with the right foot . . . etc. (I hasten to add that his endorsement of the many names view does not commit Bennett to the identity view.) The identity view and the many names view are compatible with alternative metaphysical positions, such as W. V. Quine's (Reference Quine1960: 171; Reference Quine, Lepore and McLaughlin1985) construal of events as the material contents of (potentially discontinuous) space-time regions. In what follows, I steer clear of metaphysical debates about the nature of events, and will thus construe the identity view as neutral between Bennett's and Quine's views.
(The identity view and the many names view would not sit well with authors such as Jaegwon Kim [Reference Kim1993]. Kim's metaphysics of events is intertwined with a semantics that is incompatible with the many names view. Like Bennett, Kim holds that events are property exemplifications; however, Kim contends that an event name captures its nature. For example, according to him, ‘the sparrow's fall’ is—using Bennett's terminology—a complete describer of the event that consists in the exemplification by the sparrow of the property of falling at a certain time. Kim thus rejects the many names view, and holds that ‘the sparrow's fall’ and ‘the sparrow's irregular fall’ name different events.)
One last remark before I present the causal problem. Suppose Ali has not run for fifteen years. It is far from clear what action or event ‘Ali's non-running for fifteen years’ picks out (Clarke Reference Clarke2012: 136). Call this the identification problem. This problem raises several complex issues that I do not have the space to address here. I thus assume that the identification problem does not occur in the cases I discuss. (See Payton [2018]; Schaffer [2012: 421–22] for useful remarks concerning this problem.)
2. The Causal Problem
To illustrate the causal problem, I use the following story, inspired by Carolina Sartorio (Reference Sartorio and Hazlett2010). (Note that Sartorio takes her case to pose a problem for counterfactual theories of causation, and does not specifically discuss the identity view.)
The Queen's Gardener. The queen's gardener has promised to water the queen's plant while she is away. The plant needs to be watered every afternoon; otherwise it dies. Instead of watering the plant, the gardener spends the whole afternoon touring the local pubs. When he returns home, the gardener is boozy. The plant is lifeless.
Recall that by assumption, the identification problem does not arise in this case. The story makes clear what the gardener was doing during the crucial period of time during which he was supposed to water the plant: he was touring the local pubs. Because my focus is on events, I make use of perfect nominals such as ‘his touring of the pubs’ instead of imperfect nominals such as ‘his touring the pubs.’ As Zeno Vendler (Reference Vendler1967) and Jonathan Bennett (Reference Bennett1988, Reference Bennett and Gale2002) argue, the latter arguably denotes a fact rather than an event. This is because in ‘his touring the pubs,’ the verb ‘touring’ can be modified by tenses or modalities, can have a direct object, and should be modified by an adverb rather than an adjective. The identity view thus holds not that ‘the gardener's touring local pubs’ and ‘the gardener's not watering the plant’ designate the same thing, but that ‘the gardener's touring of the local pubs’ and ‘the gardener's non-watering of the plant’ denote the same event.
Consider the following causal claims:
1. The gardener's non-watering of the plant caused the plant's death.
2. The gardener's touring of the local pubs caused the plant's death.
3. The gardener's touring of the local pubs caused the gardener's drunkenness.
4. The gardener's non-watering of the plant caused the gardener's drunkenness.
Claims 1 and 3 are very plausibly true. But if ‘the gardener's non-watering of the plant’ and ‘the gardener's touring of the local pubs’ designate the same event, then claims 2 and 4 should also be true. However, these claims seem odd, if not false.
My solution to the causal problem is inspired by Varzi's (Reference Varzi, Bottani and Davies2006: 141–42) brief remarks about a similar case. Applied to the Queen's Gardener, Varzi's suggestion is to explain the oddness of claims 2 and 4 pragmatically. Varzi does not specify the pragmatic mechanism at work. I try to do so in the following section.
3. A Similar Problem and Its Solution
3.1 The Causal Problem and the Many Names View
As a first step toward a solution to the causal problem, I examine a similar problem that affects the many names view of events. Consider the following case:
The Gunshot: A killer shoots someone with a gun. The loud shot kills the victim.
Ernest Sosa writes, ‘The loud noise is the shot. Thus, if the victim is killed by the shot, it's the loud noise that kills the victim . . . Yes, in a certain sense the victim is killed by the loud noise; not by the loud noise as a loud noise, however, but only by the loud noise as a shot, or the like . . . Besides, the loudness of the shot has no causal relevance to the death of the victim: had the gun been equipped with a silencer, the shot would have killed the victim just the same’ (Reference Sosa1984: 277–78).
As Sosa acknowledges, while we would readily accept one of the two claims below, claim 5, the other, claim 6, is odd:
5. The gunshot caused the victim's death.
6. The loud noise caused the victim's death.
Claim 5 is true. But which properties of the cause are relevant to the effect? The mass and velocity of the bullet plausibly are. However, the noise produced by the shot and the color of the bullet are irrelevant. Causal relevance is typically construed as a relation between a cause's property and an effect's property. Clearly, as Sosa points out, the property of being loud is not causally relevant to the victim's death, for it does not pass the counterfactual test he suggests: the victim would have died even if the shot had been faint or silent. If we assume that ‘the loud noise’ and ‘the gunshot’ pick out the same event, we ought to deem claim 6 true. According to Sosa, we find claim 6 odd because the loudness of the shot has no causal relevance to the victim's death. Below, I make this proposal more specific. First, I argue that in making simple causal reports such as claims 5 and 6, a speaker would (typically) pragmatically presuppose a proposition about causal relevance. I then provide counterfactual accounts of causal relevance and causation.
Bennett (Reference Bennett1988: 108–12) imagines a case similar to The Gunshot, which he uses as an objection against Quine's metaphysics of events. Suppose Lexi swims for a period of time during which she catches a cold. If we assume that the space-time region occupied by Lexi's swim matches the one occupied by the onset of her cold, then, on Quine's view, we have to hold that Lexi's swim is the onset of Lexi's cold. Bennett contends that this is implausible. Since Lexi's swim was healthful, we could truly say, ‘Lexi's swim caused the improvement in Lexi's health.’ However, ‘The onset of Lexi's cold caused the improvement in Lexi's health’ sounds odd. In response, one could adopt Sosa's suggestion and explain the oddness of the claim by holding that although true, it implies something false, namely that Lexi's cold is causally relevant to the improvement in her health. Bennett briefly considers this suggestion, but writes that he ‘cannot see how to combine Quinean theory with a counterfactual analysis’ (Reference Bennett1988: 111). In the rest of this section, I explain how this can be done.
3.2 Pragmatic Presupposition
Suppose Sylvia says,
7. Ahmed doesn't know that the monitor is defective.
In making claim 7, Sylvia takes for granted that the monitor is defective. Her main point, or proffered content, is that Ahmed does not know that. A pragmatic presupposition is, roughly, a proposition that a speaker assumes (or acts as if she assumes) in order for her utterance to be appropriate in the current context (Potts Reference Potts, Lappin and Fox2015; Simons Reference Simons, Sbisà and Turner2013; Stalnaker Reference Stalnaker1973). (Some authors hold that utterances are the primary bearers of presuppositions, while others insist that only speakers could presuppose propositions. For my purposes, it will not be necessary to settle this dispute.) In making claim 7, Sylvia pragmatically presupposes that the monitor is defective. In some cases, the truth of a pragmatic presupposition is not a condition on the truth of the utterance. Claim 7, for example, would be true if the monitor were not defective.
A pragmatic presupposition differs from a conversational implicature. The latter is part of what a speaker means—that is, what she intends to communicate (Grice Reference Grice1989). Consider the following:
8. Wei: Are you going to the picnic?
Fatima: It's raining.
In uttering ‘It's raining,’ Fatima implicates that there will be no picnic. This implicature is part of the main point of her utterance; it is a proposition that she intends to convey to Wei. But Fatima's utterance also involves a presupposition. In uttering ‘It's raining,’ Fatima presupposes that one does not picnic when it rains. While Fatima is informing Wei that there will be no picnic, she would surely not take herself to inform Wei that picnics do not occur when it is raining. She takes for granted that Wei already knows that.
To return to causal reports. In my view, in uttering ‘The gunshot caused the victim's death,’ a speaker pragmatically presupposes that the property of being a gunshot is causally relevant to the victim's death. Intuitively, the speaker would not intend to convey that a gunshot is a kind of thing that can kill; she would take for granted that her interlocutor is aware of that fact. This is why the proposition about causal relevance is best construed as a pragmatic presupposition rather than a conversational implicature. The same considerations would apply to causal reports such as ‘The gardener's beer drinking caused his drunkenness,’ ‘The rock's collision with the bottle caused the breaking of the bottle,’ and ‘The dropping of a lit cigarette caused the bushfire.’ In these cases, too, because the propositions about causal relevance are commonly known, they are best construed as pragmatically presupposed. Hence, in uttering a sentence of the form ‘The F caused the G,’ a speaker (typically) pragmatically presupposes the proposition that the cause's being F is causally relevant to the effect's being G.
It is worth noting that a true utterance associated with a false pragmatic presupposition can seem very odd. As I mentioned earlier, utterances of ‘S does not know that P’ typically presuppose that P. Consider:
9. Einstein didn't know that the earth is about 6,000 years old.
Although the claim is true, it sounds terrible. It sounds terrible, because it has a false presupposition, namely that the earth is about 6,000 years old. The oddness of claim 6, above, ‘The loud noise caused the victim's death’, can be explained similarly. Although the claim is true, an utterance of this sentence is odd because it presupposes something false, namely that that the loudness of the noise is causally relevant to the victim's death.
3.3 Causal Relevance
How should we account for causal relevance? Ernest Lepore and Barry Loewer (Reference Lepore and Loewer1987) offer a promising counterfactual condition: c's being F is causally relevant to e's being G only if e would not have been G if c had not been F. Applied to claim 6, we have: the loudness of the shot is not causally relevant to the victim's death, since the victim would have died even if the shot had been faint or silent. (As shown above, this counterfactual condition is suggested by Sosa. See also Horgan [Reference Horgan1989]. See McKitrick [Reference McKitrick2005] for a useful overview of various accounts of causal relevance.)
Some clarifications and refinements are needed. First, we should give the counterfactual conditional an elimination rather than a replacement reading. Suppose that the killer gave herself two options: if she had not used her gun to kill the victim, she would have stabbed him. Hence, in a nearby world in which the cause lacks the property of being a gunshot, it has the property of being a stabbing, and the victim still dies. This is not how the counterfactual test should be understood here. The relevant nearby world is one in which the property of being a gunshot is simply eliminated rather than replaced by another property such as being a stabbing. (In their discussion of a similar case, Payton [2018], Sartorio [2010], and Weslake [Reference Weslake2013] all rely on a replacement reading.)
Now, the counterfactual test raises a delicate issue about the cross-world identification of events. In the actual world, the cause, c, has the property of being a gunshot. Consider an event c' in a nearby world that would result from simply eliminating that property from c. It is not clear that c' is the same event as c. That depends on how modally fragile events are. The same question could be asked about the effect, e, the victim's death. The counterfactual test assumes that the very same event e could occur without the victim's dying. Again, this possibility rests on a contentious assumption about the fragility of events. This is a thorny metaphysical question that I wish to avoid.
Fortunately, David Lewis's notion of alteration provides a way out:
Let an alteration of event E be either a very fragile version of E or else a very fragile alternative event that is similar to E, but numerically different from E. If you think E is itself very fragile, you will think that all its alterations (except for the actual alteration) are alternatives. If you think E is not at all fragile, you will think that all its alterations are versions. You might think that some are alternatives and others are versions. Or you might refuse to have any opinion one way or the other, and that is the policy I favor. (Reference Lewis2000: 188)
Different alterations of the cause in The Gunshot would be obtained by modifying the velocity of the bullet or its direction, or by making the gun not shoot at all. These all count as alterations of the cause, but we do not have to settle the question whether they are versions of the same event or entirely different events. Hence, talk of alterations enables us to dodge controversies about the cross-world identification of events. It is worth noting that an alteration may involve the addition of a property. Consider a variant on The Gunshot in which the gun actually jams. Some alterations of this event would involve the gun shooting a bullet. (This point becomes relevant below, in section 4.)
The counterfactual test could thus be restated as follows: c's being F is causally relevant to e's being G only if an alteration of e would not have been G if an alteration of c had not been F. Note, once again, that the counterfactual should be given an elimination reading: the test asks us to consider a specific alteration of c in which the property designated by ‘F’ is not replaced but simply eliminated. Moreover, only this property should be eliminated: the other properties of the cause (and surrounding events) should be preserved, if possible. It is, for example, possible to eliminate the property of being loud without eliminating the property of being a gunshot. Hence, being a loud noise does not pass the test for causal relevance: had an alteration of the cause not been loud, the victim would still have died.
(More would need to be said about causal relevance. First, to accommodate probabilistic causal processes, we may want to replace talk of counterfactual dependence by talk of probability raising. Roughly, a property is causally relevant to an effect only if it increases the probability of the effect; see Hitchcock [Reference Hitchcock1993]. This alternative account would not affect the points I made about The Gunshot: while the cause's being a shot raises the probability of the victim's death, the cause's being loud does not. Second, the condition would have to be amended to accommodate cases of overdetermination, preemption and the like. Third, there are no doubt additional necessary conditions on causal relevance. Lepore and Loewer [1987] hold that c's being F and e's being G should be metaphysically independent. Fortunately, given our purposes, the rough account presented here will suffice.)
3.4 Causation
I have explained why claim 6 is odd: although it is true, it pragmatically presupposes the false proposition that the loudness of the noise is causally relevant to the victim's death. I have explained how propositions about causal relevance could be evaluated through a counterfactual test. But how are we to account for the truth of the claim? To answer this question, we need to provide the truth conditions of simple causal reports of the form ‘c causes e.’ This is a notoriously challenging endeavor. However, for our purposes, a rough account will suffice.
A simple counterfactual analysis of claim 6, such as ‘Had the loud noise not occurred, the victim would not have died’ will not do. To determine what it means for the loud noise not to occur, we need to know how modally fragile this event is. As I remarked earlier, this is a controversy I want to avoid. Once again, a promising approach relies on alterations: ‘c causes e’ is true just in case some alterations of c are such that if they had occurred, substantial alterations of e would have occurred. (Lewis [Reference Lewis2000] accounts for influence in terms of alterations, and then defines causation in terms of patterns of influence. This complication, which is meant to address cases of preemption, does not matter to my purposes. I thus stick with my simpler account. See also Woodward [Reference Woodward2003] for a sophisticated account of causation in terms of manipulation that is similar in spirit to the one proposed here.)
Again, a few clarifications are in order. First, as was mentioned before regarding causal relevance, when considering alterations of c, only the properties of c should be modified: properties of other events should all be preserved, if possible. Second, the truth of ‘c causes e’ requires the counterfactual alterations of e to be substantial in order to avoid what Lewis (Reference Lewis2000) calls spurious causes. By the law of gravitation, a distant planet affects very slightly the bullet's trajectory, thereby making a very small difference in the victim's dying. To avoid counting the planet's gravitational field as a cause of the victim's death, we should require that some alterations of a cause counterfactually entail substantial alterations of the effect. Third, not every alteration of a cause ought to be accompanied by an alteration of its effect: slight changes in the mass or velocity of the bullet would not counterfactually entail substantial alterations of the victim's death. However, some alterations would. If a blank cartridge had been used, or if the gun had not been shot, then a substantial alteration of the effect would have occurred: the victim would not have died. The proposed truth-conditional account clearly entails the truth of claim 5. And since, according to the identity view, ‘the loud noise’ designates the same event as ‘the gunshot,’ claim 6 is also true.
4. A Solution to the Causal Problem
To return to the original problematic statements, claims 2 and 4:
2. The gardener's touring of the local pubs caused the plant's death.
4. The gardener's non-watering of the plant caused the gardener's drunkenness.
On the proposed account, claim 2 is odd because someone who makes it would pragmatically presuppose the proposition that the gardener's touring the pub is causally relevant to the plant's dying. In the previous section, I held that c's being F is causally relevant to e's being G only if an alteration of e would not have been G if an alteration of c had not been F. The gardener's touring the pub is not causally relevant to the plant's dying, since if the gardener had not been touring the pubs, then (keeping everything else the same, including his not watering the plant) the plant would still have died.
Claim 4 is a little different from the causal claims I have considered so far: in it the cause is described negatively. In my view, it pragmatically presupposes the proposition that the gardener's not watering the plant is causally relevant to his drunkenness. Plausibly, the counterfactual test for this proposition is as follows: if the gardener had watered the plant, he would still have been drunk. However, this test asks us to consider an alteration that involves the addition of a property, namely the watering of the plant. Previously, when testing for the causal relevance of a property, we were simply asked to remove counterfactually that property. Why think that it is acceptable to add a property when testing for causal relevance? My answer is that if we are asking whether the absence of a property is causally relevant, we should consider what would happen if this property were added.
This answer can be motivated further by considering the following case. A soldier's staying still enables him to avoid being detected and then killed by the enemy. We thus have:
10. The soldier's staying still caused his survival.
Intuitively, staying still is causally relevant to survival. The counterfactual test yields the right result: if the soldier had not stayed still, he would not have survived. The alteration the test considers is one in which the soldier is not staying still. This is, obviously, an alteration in which the soldier is moving. Hence, in this case, the counterfactual test for causal relevance considers an alteration that involves the addition of a property. When testing for causal relevance, it is thus acceptable to move to a nearby world in which a property is added to the cause.
The argument of the previous paragraph does not assume any particular view about omissions. In claim 10, the cause is not described negatively. However, for identity theorists, this claim has the same truth-conditions as
11. The soldier's non-moving caused his survival.
This claim sounds just as fine as claim 10. In my view, this is because in addition to being true, claim 11 pragmatically presupposes the true proposition that not moving is causally relevant to survival. Now, the counterfactual test for the causal relevance of not moving is very plausibly similar to the test for the causal relevance of staying still: if the soldier had moved, he would not have survived. Once again, the counterfactual test considers an alteration involving the addition of a property.
The foregoing considerations suggest the following generalization: a speaker who utters ‘The non-F caused the G’ would (typically) pragmatically presuppose that the cause's not being F is causally relevant to the effect's being G. Moreover, c's not being F is causally relevant to e's being G only if an alteration of e would not have been G if an alteration of c had been F. To apply this account to claim 4: someone who makes claim 4 would pragmatically presuppose the proposition that the gardener's not watering the plant is causally relevant to his drunkenness. This proposition would not pass the counterfactual test: if the gardener had watered the plant (in addition to touring the pubs), he would still have been drunk.
To explain why claims 2 and 4 are true: According to the account presented in the previous section, ‘c causes e’ is true just in case some alterations of c are such that if they had occurred, substantial alterations of e would have occurred. Recall that an alteration results from modifying or eliminating an event's features. And as I showed above, an alteration could also result from adding a feature the event lacks. Claim 11 is true. Why? Because if the soldier had moved, he would have been killed. This means that to produce an alteration that counterfactually entails the soldier's death, we need to add the property of moving to the soldier. We should apply the same reasoning to claims 2 and 4. Recall that according to the identity view, ‘the gardener's non-watering of the plant’ and ‘the gardener's touring of the local pubs’ designate the same event. Consider an alteration of that event that results from adding the property of watering the plant to the gardener. Since that alteration counterfactually entails the plant's survival, claim 2 is true. As to claim 4, a relevant alteration here would be one in which the gardener lacks the property of touring the pubs. Since this alteration counterfactually entails the gardener's sobriety, claim 4 is true.
One might worry that if alterations could involve the addition of properties an event lacks, this will lead to a proliferation of implausible causes. Consider:
12. The emperor of Japan's non-watering of the plant caused the plant's death.
Clearly, if we added the property of watering the plant to the emperor, then the plant would not have died. Nevertheless, claim 12 is intuitively false, or at least very odd.
Identity theorists have two options: they could explain the awkwardness of claim 12 semantically or pragmatically. I start with the semantic account, which I favor. Identity theorists could insist that only alterations that could obtain according to some norm or other are admissible when evaluating a causal claim. These norms would include not only moral norms (‘His failure to keep his promise caused her disappointment’), but also prudential norms (‘His not exercising caused his weight gain’), social norms (‘Her not making eye contact caused his discomfort’), norms of proper functioning (‘Her alarm clock's not ringing caused her late arrival’) and statistical norms (‘The absence of summer rain caused the drying of the grass’) (Alicke, Rose, and Bloom Reference Alicke, Rose and Bloom2011; Halpern and Hitchcock Reference Halpern and Hitchcock2015; Hitchcock and Knobe Reference Hitchcock and Knobe2009; McGrath Reference McGrath2005). On the current proposal, ‘c causes e’ is true just in case some admissible alterations of c are such that if they had occurred, substantial alterations of e would have occurred. And what makes an alteration admissible is the fact that its occurrence would satisfy some norm or other.
Compare claim 1 (‘The gardener's non-watering of the plant caused the plant's death’) with claim 12. Since it is the gardener's duty to water the plant, his watering of the plant is an admissible alteration. For this reason, claim 1 is true. (The proposal here is not that the gardener's non-watering of the plant caused its death because he is responsible for its death. He is responsible for the plant's death in part because his failure to water it caused the plant's death. His failure caused the plant's death in part because it was his duty to water the plant. And it was duty to water the plant because he promised he would do so while the queen was away.) By contrast, it is not the emperor's role to water the queen's plant. Suppose his duty is to attend a reception with foreign ambassadors, which he does. Hence, the emperor's watering of the plant is not an admissible alteration of his activity that afternoon. Claim 12 is false because no admissible alteration of the emperor's activity is such that if it had occurred, the plant would not have died.
A second option is to insist that despite appearances, claim 12 is true, but that its oddness can be explained pragmatically. This is the option that Lewis favors:
At this very moment, we are being kept alive by an absence of nerve gas in the air we are breathing. The foe of causation by absences owes us an explanation of why we sometimes do say that an absence caused something. The friend of causation by absences owes us an explanation of why we sometimes refuse to say that an absence caused something, even when we have just the right pattern of dependence. I think the friend is much better able to pay his debt than the foe is to pay his. There are ever so many reasons why it might be inappropriate to say something true. It might be irrelevant to the conversation, it might convey a false hint, it might be known already to all concerned. (Reference Lewis2000: 196)
Lewis's suggestion is underdeveloped: his brief remarks about inappropriate but true utterances do not explain why, specifically, claim 12 appears to be false. However, as a first step, Lewis could get inspiration from the semantic account I just described. He could insist that claim 12 is true, since if we were to add the property of watering the plant to the emperor, the plant would not die. But it is awkward, because the emperor's watering of the plant is not an alteration that would obtain in virtue of some norm or other. More would need to be said about the specific pragmatic mechanism at work here: Is it implicature, presupposition, or some other type of pragmatically conveyed proposition? For reasons of space, I will not explore this account further.
5. One Last Puzzle
There is more to explain. Why do denials of claims 2, 4, and 6 seem correct? The following claims seem fine:
13. The loud noise didn't cause the victim's death; the gunshot did.
14. The gardener's touring of the pubs didn't cause the plant's death; his non-watering of it did.
15. The gardener's non-watering of the plant didn't cause his drunkenness; his touring of the pubs did.
In my view, the claims seem true because the negation is read metalinguistically (Horn Reference Horn1989: 362–444). Consider:
16. I don't like chocolate; I love it.
It would be a mistake to understand claim 14 as equivalent to
17. I dislike chocolate; I love it.
Clearly, a person who makes claim 16 does not deny that she likes chocolate. She does not object to the truth-conditional content of ‘I like chocolate’; instead, she objects to a certain way of describing her attitude about chocolate. Here are some other examples:
18. He's not my father. He's my male progenitor.
19. They are not mongeese. They are mongooses.
The word ‘not’ is sometimes used to deny the way a proposition is conveyed rather than the proposition itself. There is a good deal of controversy about how metalinguistic negation should be understood. Laurence Horn (Reference Horn1989: 362–444), for example, holds that the duality of use of ‘not’ is a case of pragmatic, rather than semantic ambiguity. But for my purposes, it is not necessary to figure out the specific mechanism of metalinguistic negation (Carston Reference Carston1996; Geurts Reference Geurts1998; Pitts Reference Pitts2011). The crucial point is that in claim 16, ‘not’ is not used to reject the proposition expressed by ‘I like chocolate.’ A similar point can be made about claims 18 and 19.
Consider claim 15 again. In making this claim, I argue, one does not deny the proposition that the gardener's non-watering of the plant was a cause of his drunkenness. Instead, one objects to a particular description of the cause, because this description suggests that the gardener's not watering the plant is causally relevant to his drunkenness. The apparent truth of claims 13 and 14 can be similarly explained.
(Why not say that claims 13 through 15 are false, but seem true, because each pragmatically presupposes a true proposition? In making claim 13, for example, one would pragmatically presuppose the true proposition that the loudness of the noise is not causally relevant to the victim's death. The problem with this proposal is that a false utterance still seems false, even when it pragmatically presupposes a true proposition. Consider ‘Einstein didn't know that the earth revolves around the sun.’ This false knowledge denial does seem false, even though it has a true presupposition, namely that the earth revolves around the sun.)
I grant that the fact that ‘didn't’ is used metalinguistically in claims 13–15 is not obvious. But non-obvious uses of metalinguistic negation are not uncommon. Consider:
20. Clark Kent didn't come out of the phone booth; Superman did.
It is not obvious that ‘didn't’ should be interpreted metalinguistically in claim 20. But plausibly, someone who knows that Clark Kent is Superman and makes this claim denies not that Clark Kent comes out of the phone booth, but that he is still in Clark Kent mode when he does. In other words, the speaker uses ‘didn't’ metalinguistically: she objects to naming Superman ‘Clark Kent’ in this context. (See Barber [Reference Barber2000] and Saul [Reference Saul1997, Reference Saul1999] for pragmatic treatments of claim 20. Although neither invokes metalinguistic negation, my proposed construal is compatible with these treatments.) More would need to be said in defense of this account of claim 20. But my point here is simply that a metalinguistic use of negation can sometimes be non-obvious.
6. Conclusion
Identifying omissions with actual events is problematic, some argue, because the two have different causal relations. I have proposed a pragmatic account of this apparent difference. It is crucial to distinguish between the truth of a causal claim, which concerns a relation between events, and the truth of what someone who makes that claim pragmatically presupposes, namely a relation of causal relevance. A true causal claim of the form ‘The F causes the G’ would be odd, if it is not the case that the cause's being F is causally relevant to the effect's being G. I have shown how counterfactual accounts of causation and causal relevance can be used to support this position. This approach, I have argued, provides a plausible solution not only to the causal problem about omissions, but also to a similar causal problem that concerns events in general.
