Concept
Divisible Harm
Apportionment of liability and contribution among tortfeasors
Divisible Harm
Full liability for an indivisible harm rests on one fact: each defendant's wrong was a but-for cause of the entire injury. Fermi, a passenger in Joule's car, breaks a collar bone when Joule and Edison, driving negligently, collide. The two drivers did not act together and may not have known of each other, yet the negligence of each is a but-for cause of the single injury.1 Courts therefore refused to divide Fermi's damages in half, because each driver had caused all of the harm, not half of it.2
When the harm comes apart, that premise survives for only one piece of it. Restatement (Second) of Torts § 881 states the consequence: if two or more persons, acting independently, tortiously cause distinct harms, or a single harm for which there is a reasonable basis for division according to the contribution of each, each is subject to liability only for the portion of the total harm that he has himself caused.3 A defendant whose conduct did not cause a separate injury, or a measurable part of a single injury, has no causal link to that part, and the ordinary rule that liability follows causation stops the defendant's exposure at the edge of what the defendant did.
The Restatement's general apportionment rule draws the same line from the other side. Damages are apportioned among two or more causes where there are distinct harms, or where there is a reasonable basis for determining the contribution of each cause to a single harm.4 Every other harm is one that cannot be apportioned, and § 879 makes each tortfeasor whose conduct is a legal cause of such a harm liable for all of it, whether the conduct was concurring or consecutive.5 A federal court applying CERCLA described the structure in just these terms: independent actors who cause a single harm with a reasonable basis of apportionment each answer for the portion they caused, while those who cause a single and indivisible harm answer jointly and severally for the entire harm.6 Apportionment is therefore the exception, and joint and several liability is the default that governs whenever the exception cannot be shown.
Three limits frame what follows. Section 881 speaks only of persons acting “independently,” so the discussion assumes defendants whose acts were not joined by concert of action. Divisibility also asks what each defendant's conduct did to the plaintiff, not how much blame each defendant deserves: a jury that rates one defendant 15 percent at fault and another 85 percent is saying that the first was considerably less negligent, even though each was a but-for cause of all of the plaintiff's damages.7 Those percentages belong to the later sections on comparative fault, and they do not divide a single injury. Finally, many jurisdictions have altered joint and several liability in recent years, and a few have abolished it, so the lawyer must look to the statutory and other law of the jurisdiction at hand.8
Three questions organize the analysis. Are the harms distinct? If the harm is single, does the evidence supply a reasonable basis for dividing it? And who must prove the division?
Distinct Harms
Section 881's first alternative applies when independent tortfeasors cause harms that can be told apart by what was damaged, where, or when. Each is then liable only for the portion of the harm that the defendant's own conduct caused.9 The reasoning inverts the case for full liability: a defendant whose conduct was not a but-for cause of a separate injury has no causal connection to it, and charging that defendant would make him pay for another person's wrong.
Two hunters in separate hunting parties show how the rule works. Each negligently fires at the same moment, and one wounds the plaintiff's arm while the other wounds his leg. Absent concert of action, each hunter is liable only for the harm traceable to him, because the case is essentially a problem of proof of causation, and the wounds themselves show who caused what.10 Fermi's collar bone was different: one injury had two but-for causes and no seam. Here the seam runs between the arm and the leg.
Had the fires merged and destroyed the whole parcel, with nothing to show which fire burned which acres, no seam would exist. The harm could not be apportioned, and § 879 would make each farmer liable for all of it.11
Distinct harms can also be separated in time. In the Restatement's illustration, A negligently drives into B, fractures his skull, and leaves him lying unconscious in the highway. Shortly afterward C negligently runs over B and breaks his leg. A is liable for the damages from both fractures, but C only for the broken leg.12 The two results fit together. Section 881 caps C at what C's conduct added, while A answers for the second injury because it followed from the position in which A's own wrong had left B, as the earlier discussion of consecutive tortfeasors explained.
The two accidents are separate, but the harm is no longer plainly distinct, because the second accident aggravated an injury already present in the same joint. Whether an aggravation is the same injury or a new one is the question on which cases like Specialty Hospital-Gainesville v. Barth turn. The plaintiff there, treated at the defendant hospital, developed paralysis and also a bed sore, and the bed sore got worse at a later facility that was not sued. A Florida appellate panel held that negligence by a subsequent tortfeasor does not diminish the initial tortfeasor's liability where each caused different injuries, and that fault is apportioned to the later actor only where that actor caused the same injury, not a different injury or an aggravation of an existing one.13 Distinctness thus depends on the injury and not on the number of accidents. Overlapping harm may still be divisible if the evidence permits, which the next section takes up.
The enhanced-injury cases show that a single event can also produce distinct harms.
The truck driver's liability is a separate question; the manufacturer's is narrower. Under Restatement (Third) of Torts: Products Liability § 16(b), when proof supports a determination of the harm that would have resulted from other causes without the defect, the seller's liability is limited to the increased harm attributable solely to the defect.14 The roof did not break the arm, so that injury stays outside the manufacturer's responsibility even though the same crash produced it. A manufacturer answers for the increase at all because crashes are foreseeable: it cannot say that it does not intend its automobile to be involved in an accident when it knows that the probability over the product's life is high.15 If the proof cannot support a determination of what harm would have occurred without the defect, § 16(c) treats the harm as indivisible, and the seller is liable for all of the harm attributable to the defect and the other causes. Section 16(d) then leaves it to the applicable rules of joint and several liability whether that seller is jointly and severally liable or severally liable with the other responsible parties.16
Reasonable Basis for Division
Section 881 contains a second alternative: when independent tortfeasors cause a single harm for which there is a reasonable basis for division according to the contribution of each, each is liable only for the portion that tortfeasor caused.18 The general apportionment rule agrees, directing apportionment wherever there is a reasonable basis for determining the contribution of each cause to a single harm.19 This alternative is harder to spot than distinct harms, because the plaintiff has one injury or one loss and nothing on its face says who caused what. The practical question is whether the evidence supplies a measuring stick, such as animals, acres, years, or gallons, that converts each defendant's conduct into a share of the damage. Where it does, the harm is divisible although single.
The standard serves two interests. Charging every defendant with the whole when the evidence can divide the harm makes defendants pay for damage their conduct never caused. Excusing a defendant on a guessed division has its own cost, and the plaintiff bears it: if Fermi's damages were split so that each driver were severally liable for half, Fermi would collect half from Joule and nothing from an insolvent Edison, whereas under joint and several liability Joule takes the risk of Edison's insolvency and Fermi has twice the chance of full compensation.20 Courts therefore ask for less than precision and more than speculation. The division must rest on evidence.
If a sheep is a sheep when it comes to appetite, the head count supports an inference that the larger flock ate about three-quarters of the grass and the smaller flock the rest. That inference is a reasonable basis for division, and damages are then apportioned on it rather than charged to each owner in full.21 The head count works as a proxy for two reasons: evidence can establish it, and the assumption of equal appetite links each animal to a roughly equal quantity of grass. Weaken the second link, as when one flock is sheep and the other is cattle, and the count alone would need support about how much each kind eats. The result approximates what each flock did to the pasture, and the rule accepts an approximation.
Environmental cleanup litigation applies the same standard. In Burlington Northern & Santa Fe Railway Co. v. United States, chemicals spilled while being transferred at an agricultural distributor's facility, part of which sat on land leased from two railroads, and the EPA spent $8 million on the cleanup. The Supreme Court held that a contamination event may cause a single harm and still be divisible if a defendant shows a rational basis for apportionment, relying on Restatement § 433A(1)(b). The district court had assigned the railroads 9 percent of the costs, and the Court found the geographic, chronological, and volumetric evidence sufficient to uphold that figure. The district court multiplied three factors, 19 percent for the parcel's share of the facility's area, 45 percent for its share of the relevant time, and 66 percent for the share of the injurious spills attributable to it, and then added a 50 percent margin of error.
A different cleanup case reached the opposite result under the same standard. In a CERCLA cost-recovery action over the Bluff Road waste site, the generators who had shipped waste there conceded that the contamination was a single harm, and each could point to a potentially identifiable volume of waste it had sent. They argued that liability should be divided according to those volumes, and the court, in light of the conditions at Bluff Road, refused to accept volume as the basis for apportionment.22 Volume shows how much a defendant sent. It divides the harm only if conditions at the site let a unit of volume be translated into a unit of damage, and the court held that the conditions at Bluff Road did not allow that. The railroads in Burlington Northern offered area, time, and volume together, and the courts accepted the combination.
The standard asks for evidence and not exactness. Environmental harm is estimated even before it is divided: natural resource damages, for example, are usually measured by shadow-pricing methods such as contingent valuation, because no commercial market prices threatened species or the nutrient filtering of an ecosystem.23 In McAllister v. Pennsylvania R. Co., the plaintiff was injured in separate accidents with some combined effect, and the court accepted a division supported by some evidence even though the proof fell short of the exactness needed for a precise dividing line. A rule demanding exact measurement would make almost no harm divisible, and the exception would exist only on paper. A bare assertion that a defendant surely caused less than the whole, on the other hand, gives the trier of fact nothing to apply. Where the trier of fact can determine on some reasonable basis which aspects of the plaintiff's loss are attributable to which defendant, damages ordinarily must be apportioned on that basis.24
The commonest false basis for division is a jury's percentages of fault. When a jury finds one defendant 15 percent at fault and another 85 percent, it is saying that the first was considerably less negligent than the second, even though each was a but-for cause of all of the plaintiff's resulting damages.25 The jurors are rating how faulty each defendant was in causing a single indivisible harm, and they are making no finding that either caused that fraction of the damage. The sheep differ because a count of animals measures what each flock did to the pasture, whereas a percentage of fault measures how blameworthy each defendant was. Only the first kind of figure divides a harm.
Each driver's negligence was a but-for cause of the entire injury, so the harm stays indivisible unless other evidence, of the kind the sheep and the spill sites supplied, shows a way to divide it. The later section on comparative fault takes up what jurisdictions do with such percentages. The point here is narrower: fault percentages alone never establish a reasonable basis for dividing the harm.
Many harms offer no measuring stick at all. When two negligent drivers collide and one car bounces onto a sidewalk and injures a pedestrian, the pedestrian's injury is indivisible and each driver is liable for all of it. If two separate accidents combine to kill the plaintiff, the death is likewise an indivisible result for which each defendant is liable.26 Where no fair and practicable method of apportioning the loss is available, the common-law default is that each defendant whose breach of duty was a substantial factor in producing the damage is jointly and severally liable for the total loss.27 That default is § 879's rule for harm that cannot be apportioned, and the statutes that alter it belong to the later sections on alternatives to joint liability.
Burden of Proving Division
Section 881 limits each independent defendant to the portion of the harm that defendant caused when the harms are distinct or a single harm has a reasonable basis for division, but it does not say who must prove the division.29 The omission matters because evidence of division is often thin, and the side that carries the burden loses when the evidence runs out. Restatement § 433B(2) supplies the answer: where the tortious conduct of two or more actors has combined to bring about harm to the plaintiff, and one or more of them seeks to limit liability on the ground that the harm is capable of apportionment among them, the burden of proof as to the apportionment is upon each such actor.30 Division is therefore an exit from full liability that the defendant must earn.
A defendant that offers nothing stays liable for the whole, and the whole can be large. Under pure joint and several liability, each tortfeasor is liable for the entire harm, and the plaintiff may sue any one of them or all, obtain judgments for the full amount against as many as possible, and collect from one or any combination, though only one satisfaction is allowed. With three liable defendants and $100,000 in damages, the plaintiff receives a $100,000 judgment against each.31 The burden rule decides whose silence produces that result.
Carter v. Wallace & Gale Asbestos Settlement Trust shows the sequence in a jurisdiction that retains joint liability. The Maryland court reasoned that a tortfeasor whose conduct was a substantial factor in an indivisible injury can be sued for the full amount of the damages even though others contributed. Only when the harm is reasonably or theoretically divisible does apportionment become a question of fact for the jury, and the burden of proof then shifts to the defendant to show that apportioning the damages is appropriate. The court borrowed a Montana formulation of the two steps: the plaintiff makes a prima facie showing that the defendant's act or omission was at least a contributing proximate result of the harm, and the defendant must then either deny all liability or prove that the harm can be divided. The same opinion notes that nondivisibility can result either because the harm cannot theoretically be divided or because the plaintiff cannot practically divide it among the wrongdoers, with death or the destruction of a building as examples of the first.32 The burden rule does its main work in the second group, where a division may exist in principle but only the defendants hold the information needed to show it.
Summers v. Tice supplies the reasons for placing the burden on the defendants. Two hunters negligently fired toward the plaintiff, and one shot struck his eye. The California court observed that the defendants, both wrongdoers, had put the injured party in the unfair position of having to say which of them caused the harm, and that each should therefore be required to absolve himself if he can. The court carried the reasoning into damages: where apportionment is incapable of proof, the innocent plaintiff should not be deprived of redress, and the wrongdoers should be left to work out any apportionment among themselves.33 Access to evidence points the same way. The owner of a flock knows its size, and a polluter keeps discharge records, while the injured plaintiff has none of that information.
When no one proves a division, the result follows from § 879. A harm that no one can divide is a harm that cannot be apportioned, and each tortfeasor whose conduct was a legal cause of it is liable for the entire harm.34 Uncertainty about the division therefore costs the defendants and leaves the plaintiff whole. A rule that put the burden on the plaintiff would reverse that result: a plaintiff who failed to prove the division would recover only the portion that could be proved, and the unprovable remainder would fall on the injured party.
A defendant that does the work reaches a different result. In Burlington Northern the Supreme Court held that CERCLA defendants who want to avoid joint and several liability bear the burden of proving a rational basis for apportionment. The railroads carried it with geographic, chronological, and volumetric evidence about their parcel, its period of use, and the spills on it, and the Court upheld an allocation of 9 percent of the $8 million cleanup. The government did not have to disprove the railroads' share, because the railroads had to prove it. They stood in the same structural position as the two plants above, and the only difference was that the railroads produced the evidence.
The enhanced-injury cases show how the burden depends on the jurisdiction's liability regime. When a defective design makes a crash injury worse, the manufacturer answers only for the increase, so someone must prove how much the defect added. One view requires the plaintiff to prove it more probable than not that the defect aggravated the injuries from the initial collision. The Eighth Circuit rejected that view in Mitchell v. Volkswagenwerk A.G., reasoning that it would force the plaintiff to prove that the defect was the sole cause of the enhanced injury and that he would not otherwise have been injured, a negative fact that leaves the victim in an almost hopeless position. Restatement (Third) § 16(c) takes a similar course: if the proof does not support a determination of the harm that would have resulted without the defect, the seller is liable for all of the harm attributable to the defect and the other causes.35 That is unallocability charged to the defendant, as § 879 charges it.
Utah resolved the question the other way. In Egbert v. Nissan Motor Co., the Utah Supreme Court rejected the burden-shifting approach and relied on a Utah statute that bars liability for any amount beyond the proportion of fault attributed to the defendant. Because every injury can and must be apportioned under that statute, the court held, no burden shifts, formally or informally, to the defendant seller. The plaintiff keeps the burden in an enhanced-injury case, and a product seller cannot become liable for the entire injury merely by being a codefendant.36
Both results follow from one piece of logic applied to two liability regimes. A shifted burden settles who pays when the evidence runs out. Where joint liability survives, running out of evidence means each defendant pays everything, so the defendants have every reason to produce the division and the plaintiff has no means of producing it. Where a statute caps each defendant at its own proportion of fault, running out of evidence cannot make a minor defendant answer for the whole, and the reason for tilting the burden disappears. The plaintiff then keeps the ordinary burden of proving damages, and the factfinder allocates shares on the evidence it hears, which is the pattern the later section on alternatives to joint liability develops. The governing statute therefore decides the burden, and because the statutes vary significantly, the lawyer must read the one in force and confirm that it still stands.37 Even reform legislation can fail state constitutional review: Alabama's court struck down a malpractice cap on noneconomic damages, and Arkansas's court struck down a statutory cap on punitive damages.38