Proof and Causation: When Courts and Scientists Disagree

[This column was written with the asstance of Meredith James.]

One of the challenges in environmental work is the inconsistent and erratic relationship between law and science. To be effective, environmental policy needs to be based on good science, which is why current government cuts to key environmental research are so harmful, in both the short and the long run. But even when good science exists, the law struggles to properly incorporate it, and sometimes science can’t (yet?) answer the questions the courts are interested in. What should happen then?

Criminal cases

The courts sometimes seem to have a poor grasp of the difference between scientific uncertainty and bad science. (Actually, they’re not the only ones. Ben Goldacre writes a wonderful, funny but horrifying blog called Bad Science, which illustrates some of the abuses of bad science, especially in government policy.)

In criminal and regulatory cases, it is a breach of fundamental justice to penalize someone without reliable evidence that they committed an offence. In Mitchell v. Minister of National Revenue, the SCC restated the fundamental principles of the law of evidence:

The rules of evidence should facilitate justice, not stand in its way. Underlying the diverse rules on the admissibility of evidence are three simple ideas. First, the evidence must be useful in the sense of tending to prove a fact relevant to the issues in the case. Second, the evidence must be reasonably reliable; unreliable evidence may hinder the search for the truth more than help it. …

Scientists spend immense effort establishing objective data quality standards precisely because poor quality data can lead to unreliable and incorrect results. Alas, this hasn’t stopped our regulators from laying charges, and courts from occasionally entering convictions, on evidence that doesn’t meet those standards, even when no scientist knows whether the alleged result is either reliable or correct. This is grossly unfair, and it can lead to a miscarriage of justice.

Are judges misled because test results look like science? They admit that juries are, as the SCC noted in R. v. Mohan:

There is a danger that expert evidence will be misused and will distort the fact-finding process. Dressed up in scientific language which the jury does not easily understand and submitted through a witness of impressive antecedents, this evidence is apt to be accepted by the jury as being virtually infallible and as having more weight than it deserves.

Civil cases

In civil cases, there is an additional problem: the fundamental difference in the standard of proof. Civil law requires proof on the balance of probabilities (i.e. 51% likelihood) while the scientific community does not accept causation unless the results are statistically significant. The most common level of scientific statistical significance requires the results to have a 95% chance of being true, and even then the results often turn out to be wrong.

Civil courts, by definition, are content to accept evidence that is 49% likely to be wrong, almost ten times as unreliable as the most junior scientist’s first paper. And even this standard is occasionally considered too strict. The courts often want to give a remedy that seems “fair”, even if (current) science cannot prove that the defendant is at fault. This is especially true when the defendant has deep pockets and the injured plaintiff is in great need.

Consider, for example, proof of causation, i.e. a plaintiff’s claim that a particular defendant is actually responsible for her injuries. Scientific evidence is often called upon to establish the causal link between the defendant’s conduct and the plaintiff’s harm, i.e., “but for” the act or omission of the defendant, there is a 51% chance that the injury would not have occurred. In some circumstances, the civil courts will accept an even lower standard of causation – the “material contribution” test. Two recent Supreme Court of Canada cases have explored this “material contribution” test, and the troubled relationship between legal and scientific proof.

Causation: “Material contribution” or “But for”?

In Resurfice Corp. v. Hanke, an ice resurfacing machine operator was horribly burned after he put water into its gasoline tank. He sued the manufacturer and distributor of the machine, alleging that the gasoline and water tanks were similar in appearance and placed confusingly close together. The trial judge dismissed Hanke’s claim, because he found that Hanke’s carelessness, and not the alleged design defects, was responsible for his injuries. The Court of Appeal ordered a new trial, holding that the trial judge should have considered whether the design defects made any “material contribution” to the accident. The Supreme Court restored the trial judgement, ruling that causation should normally be decided on the “but for” test. This test ensures that a defendant will not be held liable for the plaintiff’s injuries where there is less than a 51% chance that the defendant was actually at fault.

However, the SCC’s analysis encouraged lawyers to bring more lawsuits that do not meet the “but for” test, and broadened options to use the “material contribution” test as an easier way to prove causation:

First, it must be impossible for the plaintiff to prove that the defendant’s negligence caused the plaintiff’s injury using the “but for” test. The impossibility must be due to factors that are outside of the plaintiff’s control; for example, current limits of scientific knowledge. Second, it must be clear that the defendant breached a duty of care owed to the plaintiff, thereby exposing the plaintiff to an unreasonable risk of injury, and the plaintiff must have suffered that form of injury. In other words, the plaintiff’s injury must fall within the ambit of the risk created by the defendant’s breach. In those exceptional cases where these two requirements are satisfied, liability may be imposed, even though the “but for” test is not satisfied, because it would offend basic notions of fairness and justice to deny liability by applying a “but for” approach. (paragraph 25)

Clements v. Clements was the first case to go back to the SCC using this analysis. Ms. Clements suffered a severe brain injury after riding on the back of her husband’s overloaded, speeding motorcycle. Unknown to them, a nail had punctured one of the tires. When the husband tried to pass another vehicle, the tire deflated and the motorcycle crashed. Ms. Clements sued her husband’s insurer for damages. If the husband’s negligence caused the crash, the insurer had to pay. If the accident would have happened anyway, because of the nail, the insurer did not have to pay.

At trial, the judge held that the wife was unable to meet the “but for” test due to the limitations of accident reconstruction science, and applied the material contribution test to find the husband liable. The SCC ruled that this was in error, and ordered a new trial.

The Chief Justice took the opportunity to narrow the comments she made in Resurfice. She reminded readers that the civil law of negligence does not require scientific (i.e. 95%) proof of causation:

In many cases of causal uncertainty, it is conceivable that with better scientific evidence, causation could be clarified. Scientific uncertainty was referred to in Resurfice in the course of explaining the difficulties that have arisen in the cases. However, this should not be read as ousting the “but for” test for causation in negligence actions. The law of negligence has never required scientific proof of causation; to repeat yet again, common sense inferences from the facts may suffice. (paragraph 38)

Nevertheless, she said, the courts will normally insist on the “but for” test. As most earlier cases showed, the material contribution test is only available to prove causation where:

1. the plaintiff has established that her loss would not have occurred “but for” the negligence of two or more tortfeasors, each possibly in fact responsible for the loss; and
2. the plaintiff, through no fault of her own, is unable to show that any one of the possible tortfeasors in fact was the necessary or “but for” cause of her injury, because each can point to one another as the possible “but for” cause of the injury, defeating a finding of causation on a balance of probabilities against anyone. (paragraph 46)

Toxic torts?

What about pollution? “Material contribution” was partly developed in a pollution case where many were exposed and one got sick, and it may still have a place there. The Chief Justice in Clements left the door open to consider the material contribution test in toxic tort class actions in which probabilistic evidence of harm is all there is:

This is not to say that new situations will not raise new considerations. I leave for another day, for example, the scenario that might arise in mass toxic tort litigation with multiple plaintiffs, where it is established statistically that the defendant’s acts induced an injury on some members of the group, but it is impossible to know which ones. (paragraph 44)

This makes sense, and is consistent with the precautionary principle:

Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation.[2]

But courts need to understand, and to act, on the difference between bad science and scientific uncertainty:

• Scientific uncertainty shouldn’t stop prevention: Polluters who create a plausible risk of harm, by dumping their filth into our shared air and water, should have to take cost-effective measures to prevent that harm. And uncertainty, as described by the Chief Justice, may even have a place in compensation.
• But bad science should be thrown out of court. When alleged scientific data fails to meet relevant, objective quality standards specifically developed for that kind of data, no one knows whether the claimed result is either reliable or correct. It is fundamentally unfair to punish anyone based on such data.