September 08, 2007

Types of truth

Last week, I commented on Rahm Emanuel's clever remark that "Alberto Gonzales is the first attorney general who thought the truth, the whole truth and nothing but the truth were three different things". I observed that the formal semantics and pragmatics of the traditional oath are subtle things, and asked for a logical translation of the joke ("Political semantics quiz", 8/28/2007).

The answers were surprisingly diverse -- summarizing crudely, some people think that the three sub-phrases are just lawyerish repetition for emphasis, while others think that they reference three different relations between statements and the world, or perhaps three different relations among speakers, statements and the world.

Mr. Emanuel's quip seems to rely on the redundancy theory being true, or at least being the view of all previous holders of Mr. Gonzales' office. I thought that this implication was probably wrong, although the joke is still funny.

Email has piled up faster than I can post it or answer it, as Barbara Partee has recruited several other interesting people into a lively discussion. I've posted a couple of the more interesting messages below.

Jerry Hobbs wrote:

Construct the following matrix:

said: not said:

To tell the truth: A > 0
To tell the whole truth: B = 0
To tell nothing but the truth: C = 0

(The first line depends on how you interpret "the" in "the truth". A better way of saying "A > 0" would be "Make sure there is some truth in what you are saying." Actually, I think "Tell the truth." is just a summary statement for what is elaborated in the next two statements; so it would really mean "B = 0 & C = 0".)

Since A/(A+B) is recall, promising to tell the whole truth is promising 100% recall. Since A/(A+C) is precision, promising to tell nothing but the truth is promising 100% precision.

I've always thought a good answer to that standard TV lawyer question, "Please answer the question yes or no." would be "I'm under oath to tell the whole truth. The whole truth is more complicated than just a yes or a no." But fortunately I've never been in a position to use this argument.

In fact, I think the Supreme Court has ruled that a witness is not required to tell the whole truth if he or she is not asked the right question.

I think that Jerry's initial picture of the semantics is similar to what I implied by suggesting that truth is like pie: ("Do you swear to eat the pie, the whole pie and nothing but the pie?"). But he concludes that the first clause is really "just a summary statement for what is elaborated in the next two" -- while Tim Finin thought that clause #1 and clause #3 are synonymous, while clause #2 means something different. And I bet that Jerry meant truth to be constrained by relevance, since his analysis would otherwise require rather lengthy testimony in answer to any question (the number-theory segment alone would empty the courtroom and exhaust the witness...)

Larry Solan commented on the relevant SCOTUS decision, and more generally on the subtle philosophy of language involved in legal discussions of perjury:

The perjury cases are very interesting. The lead case is called Bronston v. United States, decided by the Supreme Court in 1973. Bronston had filed for bankruptcy. He was questioned under oath as follows:

Q. Do you have any bank accounts in Swiss banks, Mr. Bronston?

A. No, sir.

Q. Have you ever?

A. The company had an account there for about six months, in Zurich.

Q. Have you any nominees who have bank accounts in Swiss banks?

A. No, sir.

Q. Have you ever?

A. No, sir.

It turns out that he also had had a Swiss bank account in the past. He was convicted of perjury, but the Supreme Court reversed 9 - 0, on the theory that he didn't say anything literally false and that it was up to the questioning lawyer to pursue the truth. So far, implicature loses.

But there is a footnote in the opinion. It gives the following hypothetical situation, which it agrees does constitute perjury:

"The District Court gave the following example as an illustration only: If it is material to ascertain how many times a person has entered a store on a given day and that person responds to such a question by saying five times when in fact he knows that he entered the store 50 times that day, that person may be guilty of perjury even though it is technically true that he entered the store five times."

The Court comments: "it is very doubtful that an answer which, in response to a specific quantitative inquiry, baldly understates a numerical fact can be described as even "technically true." Whether an answer is true must be determined with reference to the question it purports to answer, not in isolation. An unresponsive answer is unique in this respect because its unresponsiveness by definition prevents its truthfulness from being tested in the context of the question -- unless there is to be speculation as to what the unresponsive answer "implies."

Furthermore, there is a more recent case by a court of appeals which held someone guilty of perjury for testifying truthfully when the questioner had misstated the question. The witness answered truthfully to the question as put (it contained a wrong date), but the witness clearly knew what the questioner had meant and purposely took advantage of the mistake to attempt to create a false impression. So the law seems to care about implicature just when further inquiry is not likely to undo the perlocutionary effect of deceit.

Peter Tiersma and I write about this in our book, Speaking of Crime, focusing largely on the Clinton scandal. I think it shows that even though lawyers don't know the vocabulary, they have a pretty good intuitive sense of the relevant linguistic concepts. As for the morality of cases like Bronston, that's a different matter.

I take all this to confirm my suspicion that Alberto Gonzales was by no means the first Attorney General to think that the truth, the whole truth and nothing but the truth are three different things. But he may have been uniquely ineffective at exemplifying the distinction in testimony to congress.

[Update -- Bob Ray writes:

I was comforted by following statement in your post:

"In fact, I think the Supreme Court has ruled that a witness is not required to tell the whole truth if he or she is not asked the right question."

Many years ago, I was an expert witness on spelling errors in a highly publicized kidnapping case. The word "approuch" appeared in the ransom note and in a letter written by one of the defendants years earlier. This spelling error was one of three key pieces of evidence against the two defendants.

I told the defense lawyers when they asked me to testify that I could think of a number of reasons why "approuch" is a reasonable way to misspell the word "approach," but that if I was asked if it was a "common" misspelling of the word, I would have to say I didn't think so. This was ante-Google and all I had to go on were compiled lists of common spelling errors.

When I was on the stand, the defense attorney asked me if it was a reasonable misspelling and I gave a number of reasons why I thought it was. The prosecutor asked me nothing at all. The defendants were acquitted.

I think I can rest easy about having been guilty of perjury, but I still wonder if I violated my oath to tell "the whole truth." In all these years, I've never seen the word spelled that way in the wild but it's also a comfort that Google shows over 29,000 hits for it (and 366 million for "approach").


Posted by Mark Liberman at September 8, 2007 09:37 AM