Burden of proof

At dinner, a friend says the restaurant’s “house-made” pasta is actually bought frozen. You ask how he knows. He says he saw a delivery of boxes from a frozen-food supplier at the back door last week.

He made the claim, so he owed a reason, and he gave one. Now the burden has moved, at least a little: if you still think the pasta is made in-house, you need something to say about the boxes (perhaps they were for desserts). If he had said instead, “Well, can you prove it isn’t frozen?”, he would have been shifting the burden. And if he had no reason at all, the right response would be to not believe him yet, not to conclude that the pasta must be house-made. That last step would be an Argument from ignorance.

• Whoever asserts, and asks for agreement, carries it. C. L. Hamblin, in his influential study of fallacies, cites the principle from the law of evidence: “He who asserts must prove.” In pragma-dialectics, an approach that treats good argument as following the rules of a good-faith discussion, the corresponding rule is that you must defend your standpoint if asked to do so.
• It isn’t about whether a claim sounds positive or negative. “There’s no mouse in the kitchen” is a claim like any other, and someone who asserts it and wants it believed needs a reason. The philosopher Steven Hales points out that any claim can be rephrased as a negative (“P” is equivalent to “not not-P”). If wording alone could excuse a claim from needing support, any claim could be excused.
• It moves during a discussion. Once one side gives a reasonable argument, the other side has to respond to it, by criticizing it or offering a counterargument, or the argument stands. Walton describes this for law as a global burden that stays where it started and a local burden of answering the latest argument that goes back and forth; the same pattern shows up in ordinary argument.
• Its size depends on the claim and the stakes. Deciding what to cook needs less support than deciding whether to sell a house. What counts as enough is usually a matter of judgment outside law, and a matter of explicit rules inside it.

Law makes the principle explicit, and much of the vocabulary comes from there. Walton describes law as distinguishing two burdens:

• The burden of persuasion, fixed before the trial begins. It doesn’t change during the argument, and it decides who wins at the end if the evidence isn’t strong enough.
• The burden of production (also called the evidential burden, or the burden of going forward with evidence), which shifts back and forth as each side presents evidence the other must answer.

The standard of proof says how strong the evidence has to be to meet the burden of persuasion. In plain terms, as described in the Stanford Encyclopedia of Philosophy‘s entry on legal probabilism:

Legal scholars who analyze these in terms of probability sometimes read them as thresholds (above 0.5 for the civil standard, for example), but how to define the standards, or whether to define them at all, is contested, and courts have warned that trying to define “reasonable doubt” can make it less clear.

Why have different standards? A stricter standard produces fewer false convictions at the cost of more false acquittals, and a looser one does the reverse. Criminal cases use the strictest standard because convicting an innocent person is treated as much worse than acquitting a guilty one. A burden of proof also lets a court reach a decision at all when the facts can’t be fully known: witnesses may be missing, and evidence lost. Walton makes this point about why the device is necessary.

Meeting or failing a burden settles the case, not the facts. A “not guilty” verdict means the prosecution didn’t meet its burden, not that the court found the defendant innocent.

A presumption is a rule for where the burden starts. It lets something be taken as true until someone gives a reason to think otherwise.

• In law. Walton quotes a common minimal account: a presumption is a legal mechanism which, unless enough evidence is introduced against it, “deems one fact to be true when the truth of another fact has been established”. A properly addressed and stamped letter that was mailed is presumed to have been received; a person not heard from for a long period set by law can be presumed dead.
• The presumption of innocence. The 19th-century logician Richard Whately, as quoted by Hamblin, stressed that it doesn’t mean we are to take for granted that the defendant is innocent, nor that innocence is “antecedently more likely than not”. It “means only that the ‘burden of proof’ lies with the accusers”.
• In argument generally. Whately carried the idea into rhetoric. On Hamblin’s reading, a presumption in Whately’s sense doesn’t carry weight as an argument: it “merely decides which party, if there is to be an argument, must lead the attack.” Presumptions vary in strength, and one can be rebutted with another, which shifts the burden to the other side. Whately also held that calling a person or book an “authority” grants it a presumption (compare appeal to authority).
• Everyday defaults. Treating every wire as live until it’s tested is a presumption chosen because one mistake is far worse than the other.

This slogan says the burden grows with how much a claim conflicts with what’s already known. Its history, as traced by David Deming of the University of Oklahoma (2016):

• Carl Sagan popularized the wording in his book Broca’s Brain (1979), discussing claims about the paranormal. He did not define “extraordinary”.
• Marcello Truzzi had earlier written that “an extraordinary claim requires extraordinary proof”, in a 1975 letter to Parapsychology Review and a 1978 article in Zetetic Scholar.
• Pierre-Simon Laplace wrote in 1812 that “the more extraordinary the event, the greater the need of its being supported by strong proofs.”
• David Hume‘s essay on miracles (1748) argued that the weight of testimony for a reported fact diminishes “in proportion as the fact is more or less unusual.”

Read carefully, it’s a point about weighing evidence: a claim contradicted by a large body of existing evidence starts far behind, so it takes a lot of new evidence to overcome that. Deming argues that on Hume’s account, “extraordinary evidence” isn’t a special kind of evidence, just a great deal of it, and that a claim that is merely new or unpopular isn’t extraordinary in this sense.

• Burdens can be contested. Outside a courtroom, nobody assigns them in advance, and each side can argue that the other owes the proof. Walton devotes part of his book to these disputes, which he models as a metadialogue: a discussion about how the discussion should go. He also takes up the objection that legal ideas of burden and presumption have been carried into public debate, where argument isn’t organized the same way, and argues that they can be adapted to it, though not in every respect. Whately’s claim of a presumption in favor of existing institutions, which puts the burden on anyone proposing a change, has itself been controversial. Hamblin reports the controversy, though he considers the main charge against Whately ill-founded.
• The courtroom model may not fit ordinary argument. Ulrike Hahn and Mike Oaksford argue that in law the goal is a practical decision, while in critical discussion the goal is often just to raise or lower how strongly a claim is believed. There, whether a claim crosses a threshold may not matter, and they explore what that means for theories that explain fallacies in terms of the burden of proof.
• “You can’t prove a negative” isn’t a rule of logic. Hales argues that negatives can be proved as well as anything else. A valid Modus tollens argument can conclude that something doesn’t exist (“if unicorns had existed, there would be evidence in the fossil record; there isn’t; so they didn’t”), and a thorough search can support a negative inductively. What no inductive argument gives is certainty, and that’s true of positive claims too. The real problem is narrower: a demand for conclusive disproof, or for disproof of a claim built so that nothing could count against it (see Unfalsifiability), sets a burden no one can meet.
• “Extraordinary” can be stretched. Because Sagan’s version never defines the word, Deming argues it has been misused to dismiss unconventional research and even well-supported mainstream hypotheses, by labeling any unwelcome claim “extraordinary” and then demanding more evidence than anyone could supply.
• An unmet burden doesn’t make a claim false. It means the claim hasn’t been shown. Treating the failure to meet a burden as proof of the opposite is the argument from ignorance.