Concept
The But-For Test
Actual causation (but-for cause)
The But-For Test
Section 2.03(1)(a) of the Model Penal Code states the first condition of causation in a single clause: conduct is a cause of a result when it is an antecedent but for which the result in question would not have occurred.1 The clause can be restated as a question to put to any fact pattern. If the defendant's conduct is deleted from what happened and the rest of the story is left as it was, does the result, as and when it actually occurred, still happen? When it does, the conduct was not a cause in the but-for sense, and the prosecution has failed on the first step of causation. When it does not, the conduct was an actual cause, and the inquiry can move on to the further requirements taken up later in this chapter.
This is the but-for test, and the law gives it several names. But-for cause, actual cause, cause in fact, and factual cause all describe the same relationship, the Latin sine qua non refers to the same idea of a necessary condition, and criminal-law writing sometimes calls the test the had-not test.2 Courts applying the common-law version reach the same place, treating conduct as a cause in fact of a prohibited result when the result would not have happened in the absence of that conduct.3 Legislatures have also written the formula into their own codes. The Delaware Supreme Court applied Delaware's version, which uses the Code's wording, in Oxendine v. State.4
The plain case shows how little the test demands. A, intending to kill B, aims at the heart of B, a person in good health, and pulls the trigger. No one else is trying to kill B at that moment, A's aim is good, and B dies within moments. Delete the shot and leave everything else as it was, and B simply goes on living, so A's conduct was a but-for cause of B's death, and the death that occurred is exactly the one A intended.5 Where one person shoots another and the other dies of the wound, this is the ordinary result, which is why the but-for relationship explains almost every case in which conduct is held to have caused a result.6 The prosecution's task at this step is accordingly light. It must prove beyond a reasonable doubt that the result would not have occurred when it did and in the way that it did without the defendant's act, but in the usual case that proposition is not seriously contested.7
The test is a thought experiment. No one can rerun events without the defendant's act, so the factfinder pictures the world as it would have been if everything had been the same except for the allegedly causal event, and asks whether the result would still have come about.8 In Bostock v. Clayton County, which construed Title VII's ban on employment decisions made because of sex, the Supreme Court described the same procedure as changing one thing at a time and seeing whether the outcome changes; if it does, a but-for cause has been found.9 Two features of the experiment decide close cases. The first is that only the defendant's conduct is removed, so what other people did and the condition the victim was in stay as they were. The second is that the result being tested is the result as it actually occurred, with its actual timing and route.10 That second feature matters because everyone dies eventually. If the question were whether the victim would have died at some point without the defendant, no killer would ever be a but-for cause, so the test asks about this death, at this time, in this way. The refinements below show how that wording works when a victim was already dying.
The third step carries most of the weight. The experiment changes one fact only, so a student who removes the defendant's act and also quietly improves the victim's health, or erases some other person's act, has answered a different question from the one clause (a) poses. Once the other facts really are held constant, the answer is usually obvious, and the harder cases are those in which the answer turns on what the evidence can establish about the world without the act.
The experiment returns a negative answer whenever the result would have come about anyway. The missed shot from the previous section is the clearest case. A fires at B intending to kill and misses, and at that instant B drops dead from a cause wholly unconnected with the shot. Delete the shot, keep B's unrelated condition in place, and B still dies at that moment, so the shot is not a but-for cause even though the intent, the act, and the death all coincide.11 A standard tort illustration shows that the wrongfulness of the conduct is beside the point. Phillips negligently removes the blade guard from a band saw, and Mancuso, cutting plastic on that saw, has a severe allergic reaction to the plastic fumes. The fumes would have reached Mancuso with the guard in place, so the removal was careless but was not a but-for cause of the reaction.12
Two reasons make the negative answer useful rather than technical. The first is that the counterfactual turns a question about blame into a question of fact that evidence can address: a pathologist can say what killed the victim and whether the victim would have died at that moment anyway, and the prosecution fails on the element if the proof leaves reasonable doubt on the point.13 The second is that a bare result invites a temptation the test checks. A jury that sees a dead victim and a defendant who acted wrongfully is inclined to link the two, and the counterfactual question replaces that inclination with something specific to verify, namely whether the death would have happened without the act.
Clause (a) asks only whether the conduct was necessary to the result. It does not ask whether the defendant was the sole or principal cause. Every result has many necessary conditions, and removing any one of them would have prevented it, so a shooting death depends on the victim's being within range and the gun's being loaded as much as on the pull of the trigger. Because one event may have many causes, pointing to a second cause does not eliminate the defendant's.14 A narrower test would force the factfinder to rank causes, which is a judgment about responsibility, and leaving that judgment to the later inquiry into legal cause keeps this step answerable by evidence. Three refinements follow from the wording and are developed next: a result that would have occurred anyway defeats the test, a defendant who shortens the life of a dying victim still satisfies it, and a single result may have several but-for causes.
Results That Would Occur Anyway
The but-for test gives a negative answer whenever the result, as and when it actually occurred, would have come about without the defendant's conduct. Section 2.03(1)(a) treats conduct as a cause only if it is an antecedent but for which the result would not have occurred, so conduct that made no difference to the result fails the clause however culpable it was.16 The evidence can lead to that conclusion in two ways. It may affirmatively show that the same result would have followed anyway, or it may leave a reasonable doubt about the matter. The second route repays attention, because the burden of proof decides it.
The burden matters for a structural reason. The counterfactual asks the factfinder to picture a world that no one observed, so some uncertainty attends every application of the test, and the only question is who bears it. Causation is an element of a result crime, and the prosecution must prove every element beyond a reasonable doubt, which places the risk of that uncertainty on the state. If a reasonable possibility remains that the death would have occurred regardless of the defendant's act, the element is unproved. A rule that allowed conviction whenever the act could have caused the result would punish defendants for results their conduct may not have altered, and the result is the very feature that raises the grade of the offense. The uncertainty route is easiest to see in a pair of medical facts that differ in one respect.
The two versions share the same duty, the same failure to act, and the same death. The only difference is how strongly the evidence says the operation would have prevented it. That difference decides the element because clause (a) asks what the proof establishes about the world in which the surgeon appeared, and the prosecution loses when the proof leaves that world uncertain. The surgeon in the second version has proved nothing and is not required to, since the gap in the state's case is enough. The other route to a negative answer arises when the evidence affirmatively shows that the same result would have occurred, which is clearest when an independent second force stands ready to produce it.
Set beside the plain case from the previous section, in which a well-aimed shot at a healthy victim was followed by death within moments, the graze shows which fact carries the analysis. In the plain case, deleting the shot leaves the victim alive. Here, deleting the shot leaves X's act untouched, so V dies at the same moment and in the same way, and D's shot is not a but-for cause even though D intended the death and the death occurred. D's intent, D's act, and V's death are all present, and the missing link is the one the test is built to detect. What D loses is a conviction for causing the death. The shot may still support liability that does not depend on that result, a point taken up with preempted causes below.
Uncertainty about the counterfactual arose in a different form in the Dyos prosecution. The defendant admitted inflicting a forehead injury on the victim, who also bore an injury to the ear of unknown origin, and the pathologist regarded it as a reasonable possibility that the ear injury, perhaps from a brick, was what caused the death. The defense argued that the prosecution (the Crown) therefore had to exclude the possibility that the other injury caused the death before the murder count could go to the jury: there was another injury, it may reasonably have been the cause of death, and it could not be shown not to have been.17 The defendant's admitted act could have caused the death, but nothing proved that it did, and the beyond-a-reasonable-doubt standard resolves that gap in the defendant's favor. After the judge's ruling, the defendant pleaded guilty to affray and to grievous bodily harm, charges that do not depend on having caused the death.18
The reasoning has a limit. The defense put the point as saying the defendant's act cannot be the cause of death if the death would or could have occurred without it, but the word could overstates the rule, because the doubt that defeats the prosecution must be a reasonable one, supported by evidence such as the pathologist's testimony.19 The inquiry looks to what did happen, rather than to what might have happened by some other route.20 For that reason a bare imaginable alternative does not create reasonable doubt. A defendant who points to evidence that another injury, illness, or actor may reasonably have produced the death stands where the defendant in Dyos stood. One who says only that anything is possible does not.
Civil decisions run the same counterfactual on ordinary facts and show the burden working against whichever party carries it. In a civil suit over a collision at a railroad crossing, the plaintiff showed that the train exceeded the concededly safe speed, but the engineer's testimony and the train's braking distances showed that it could not have stopped in time even at the safe speed, so the excess speed was not a cause in fact of the collision.21 The plaintiff's fallback was that the car's driver would conceivably have had a little more time to escape, and the court rejected that theory as conjecture with no evidentiary support.22 The same discipline binds a prosecutor who can only speculate that the patient would have survived had the surgeon appeared.
A standard torts problem adds a lesson about the kind of evidence that settles the counterfactual. A landlord negligently fails to install a required fire escape, and a tenant is found dead in his bed from smoke inhalation, apparently without ever trying to leave. A dozen fire escapes would have made no difference, so the landlord's omission is not a factual cause of the death, while the other tenant who fell asleep while smoking is a cause, because without that act there would have been no fire.23 If the body is instead found in the middle of the hall, the facts can support an argument that the tenant reached for an exit that was not there and turned back, and the omission may then be shown to have made a difference.24 In each version the outcome depends on how the death actually unfolded, and that is the evidence a factfinder uses to decide whether removing the conduct would have changed the result. A civil plaintiff need only show that but-for cause is more likely than not.25 A criminal prosecution must clear the higher bar of proof beyond a reasonable doubt, so it fails on this element more readily when the evidence about the counterfactual is thin.
Hastening a Death
The but-for test asks about the result as it actually occurred, and that wording decides what happens when the victim was already dying. Section 2.03(1)(a) treats conduct as a cause when the result in question would not have occurred without it, and the result in question is a particular death at a particular time and in a particular manner.27 A death that arrives an hour earlier than it otherwise would have is therefore a different result from the one that would have followed had the defendant stayed out of the story. The standard illustration involves a victim at death's door. A shoots and instantly kills B, who from disease or from a prior mortal wound inflicted independently by X had only an hour left to live. Delete the shot and leave the disease or the earlier wound in place, and B would have lived another hour, so the death that actually occurred would not have happened when it did. A has caused B's death, and the general proposition follows that one who hastens a victim's death is a cause of it.28
The rule takes this shape because every person dies eventually, and a rule that required the victim to have otherwise lived indefinitely would remove the killing of the sick, the injured, and the elderly from causation altogether. Treating the remaining span of a life as something the law protects avoids that result, however short the span is. The rule also removes an incentive problem. A defendant who learns that the victim was already doomed would otherwise have a ready answer to every homicide charge, namely that the death was coming regardless, and the victims least able to defend themselves would receive the least protection. Because the result is the death as it actually happened, that answer fails whenever the defendant's act moved the death earlier. The motive for the act does not alter this analysis: a caregiver who disconnects life support from a patient with a single day left to live has caused the death, whatever society may think of the mercy behind it, because the patient's actual death would have come later.29
This rule can look as though it conflicts with the previous section, where a result that would have occurred anyway defeated the test. The two fit together once the variable that separates them is identified. In the graze example, an independent killer's act stood ready to produce the same death at the same moment, so deleting the defendant's shot left the death unchanged. Here, deleting the act leaves the victim alive for a further stretch, so the death as it occurred has a different time. The question asks whether the death that happened would still have happened when and as it did, and a victim's general doom does not answer it.
Two wounds can each shorten the same life. When A shoots B and leaves her to die, and C then stabs B and hastens her death, C is guilty of murder for intentionally shortening another's life, and most courts also hold A liable, because at the time of death both wounds were contributing to the life-shortening process.30 The hastening rule is what makes C a cause even though B was already dying from A's shot, and the shared contribution of the two wounds is what keeps A in the picture. Courts find it harder to hold A responsible if C's act is more extreme, as when C decapitates B, but that question belongs to legal cause, which this chapter reaches at the end.31 For present purposes, the point is that each act that moves a death earlier or contributes to bringing it about can satisfy clause (a).
The hardest part of a hastening case is usually proof. In Oxendine v. State, a child died after two beatings, and the State had to establish causation as Delaware's statute defined it, as the antecedent but for which the result in question would not have occurred.32 The medical examiners identified two injuries, one inflicted more than twenty-four hours before death and one less than twenty-four hours before. One pathologist could not separate the effects of the two or say which hemorrhage caused the death, and another regarded the earlier injury as the underlying cause but, when asked directly whether the second hemorrhage accelerated the death, answered that he did not know.33 A third pathologist, called by a codefendant, said only in general terms that another blow to a weakened child would tend to shorten life, and that answer responded to a hypothetical assuming two blows when he had himself found one injury.34 The prosecutor's question about acceleration was the right one, because proof that the second beating moved the death earlier would have made it a but-for cause even if the first injury was the underlying cause. The case shows the evidentiary side of the rule: a defendant is not liable for the death unless there is evidence that the act caused or accelerated it.35
The hastening rule has a boundary that the same reasoning draws. Hastening requires a life to shorten, so the rule reaches only a person who is still alive when the defendant acts. If the victim was already dead, the death element was complete before the conduct, and deleting the conduct leaves that death exactly as it was. Students lose this issue when they treat a doomed victim as a dead one, because a person with one hour to live is still alive, and the hour belongs to the victim.
This pair differs in one variable: whether the victim was alive when the defendant acted. In the poison version, the shot ended a life that still had hours in it, and that remaining time was the thing the defendant cut short. In the heart-attack version, nothing remained to cut short. The conclusion concerns the causation element only, and whether D incurs any other liability for the shot is a separate question that does not depend on having caused a death.
Several But-For Causes
The but-for test asks whether the defendant's conduct was a necessary condition of the result, and it does not ask whether that conduct was the only one. A death, a fire, or a collision rests on many necessary conditions at once: the victim had to be in a certain place, an injury had to go unremedied or be mishandled, and the defendant had to act. Deleting any one of those conditions changes what happens, so each of them satisfies clause (a) of § 2.03(1).37 The consequence is that naming a second cause says nothing about the first, because an event has many causes and identifying one does not eliminate the others.38 Students lose this point when they treat causes as competitors, reasoning that proof of what else contributed to a death disproves the defendant's role in it.
The test cannot work as a competition because of how it is built. It removes the defendant's conduct, holds everything else in the story fixed, and asks whether the result changes. Another person's act, the victim's frailty, or a doctor's mistake is part of the fixed background in both versions of the story, so its presence tells the factfinder nothing about whether the defendant's act made a difference. A rule demanding that the defendant be the sole cause would also hand every contributor the same defense, since each could point to the others, and a result that several people together brought about would be attributable to none of them. The law deals with the fairness of reaching a remote or minor contributor in a different place. Because but-for causation is only half the inquiry, a defendant who was a but-for cause is still not responsible unless the conduct was also the imputable, or legal, cause.39
The Supreme Court's explanation of but-for causation in Burrage v. United States shows the principle at work. When a defendant's act combines with other factors to produce a result, the act is still a but-for cause so long as those other factors alone would not have produced it, the straw that broke the camel's back. Poison given to a man weakened by several diseases is therefore a but-for cause of his death even though the diseases played a part, provided he would have lived without the poison's added effect.40 The same structure appears when the second cause is another person. If A inflicts a wound that would not have been fatal with proper treatment, and the attending doctor then treats it so negligently that the victim dies, A's wound is a but-for cause, because without it there would have been nothing to treat and no death.41 The doctor's negligence is a but-for cause as well, and the two facts sit side by side. Civil law reasons the same way. In Bostock v. Clayton County, a Title VII decision, the Court observed that a collision may occur both because one driver ran a red light and because the other failed to signal a turn, so that each is a but-for cause, and a defendant cannot avoid liability by citing some other factor that also contributed.42
The treatment case and the threat case look alike on the surface, because in each a second person also played a role in the death, yet the second person is not what separates them. In the treatment case, deleting A's wound removes the whole sequence and the death does not occur. In the threat case, deleting A's threat leaves the sequence intact. The variable that moved the answer is whether the result survives deletion of the defendant's conduct, and another person's involvement does not move it in either direction. That is also where this section meets the previous one: a defendant escapes only when the result would have happened anyway.
Proving that a defendant is one of several but-for causes settles only the factual link. Whether A should answer for a death that a negligent surgeon also helped produce, or that a victim's prior weakness made possible, is the legal-cause question taken up at the end of this chapter, and that inquiry is where the fairness of holding a contributor responsible is decided.43