LITTLE OF WHAT WE KNOW is based on our own perceptions. This claim may seem surprising, but further reflection reveals its truth. Yes, I know it is snowing outside because I can look out the window and see snow. I know this wine is bad because I taste the mustiness that wine experts call “corking.” And I know the orchestra is playing Beethoven’s Fifth Symphony because I, personally, recognize its characteristic theme.
But these examples are unrepresentative. I know the date of my birth because my mother told me and because it is written on a document prepared by a public official in Newark, New Jersey. Of course, I did not see, hear, taste, smell, or touch my own birth. I know when I was born not because of direct perception, but because of what others have told me.
This is unremarkable. Much of our knowledge comes from the assertions of others; from testimony, in the broad and not court-centric sense of that word. The philosopher R. F. Atkinson is surely right when he observes that “so much of what we believe, not just about the past but about everything, is based on testimony that it is scarcely credible that we could get along without it.”1 Testimony, in this broad sense, consists of the assertions of others, especially the assertions of the others we are asked to credit just because someone has asserted something. So understood, testimony can be written as well as oral, the image of the witness in court giving testimony notwithstanding. I believe it was cold in Montana yesterday because a friend who lives there told me so in a telephone conversation. I believe that the rate of Covid-19 infection in my city is decreasing because I heard it on the evening news. I believe there are bears in the neighborhood, although I have never seen one, because multiple neighbors say that they have seen one. I believe that Indonesia is the fourth largest country in the world in population because I read it on the US Census Bureau’s website. I confidently pour a yellow liquid into my mouth because the label on the bottle testifies that the bottle contains lemonade. And on and on. The questions to be addressed here are about the extent to which testimony is evidence for the proposition that the testifier asserts. Or that a testifier asserts to another testifier who asserts it to still another testifier, creating a long chain of testimony whose end link is the person relying on an original assertion passed along through multiple others. The questions about testimony not only are about the status of testimony as evidence, but are also about the extent to which, if at all, that testimonial evidence is different from other types of evidence, about how testimonial evidence compares to personal observation or perception, and about the special evidentiary problems that testimony does or does not create. The focus of this chapter is secondhand knowledge. It is knowledge that comes, not from firsthand observation or experience, but from others’ knowledge, typically transmitted by language.2 The aim of this chapter and Chapters 6–8 is to expose and explore the complexities arising from treating secondhand knowledge as information and as evidence.3
Most Americans know of George M. Cohan, if they know of him at all, as the composer and entertainer who gave us “Give My Regards to Broadway,” “You’re a Grand Old Flag,” “Over There,” “Yankee Doodle Boy,” and many other popular songs of the first half of the twentieth century. But tax lawyers know of Cohan for different reasons. For them Cohan is the central character in what is now known as the “Cohan rule,” and it is an important part of tax law and tax practice.4
The Cohan rule emerged out of a dispute between Cohan and the Internal Revenue Service regarding his 1921 and 1922 tax returns. Cohan, whose penchant for living large matched his fame, was accustomed to picking up the tab at restaurants and nightclubs for business associates, fellow entertainers, friends, acquaintances, and assorted hangers-on. And when tax time came, Cohan estimated what he had spent on these acts of professional largesse and claimed that amount as a deductible business expense. An auditor for the IRS asked for proof of these expenditures, but Cohan could supply nothing in writing. After all, famous big spenders don’t ask for receipts. So Cohan offered as evidence only his own oral representations of what he had spent. The IRS, not surprisingly, refused to allow the deduction. Litigation ensued.
Cohan’s dispute with the IRS eventually reached the US Court of Appeals for the Second Circuit, whose jurisdiction encompasses New York, Connecticut, and Vermont. And although Cohan lost on some of his arguments, he prevailed on the question whether his oral representations—his testimony—could count as evidence for his deductible expenses. In a decision written by Judge Learned Hand, perhaps the most distinguished of those American judges who never became Supreme Court justices, the court sided with Cohan.5 Judge Hand seemed skeptical of Cohan’s representations, but he was just as skeptical of the Internal Revenue Service’s blanket policy of disallowing any claimed deduction for which there was nothing in writing. What Cohan had said, Judge Hand concluded, was evidence of what Cohan had actually done, and the court instructed the IRS to assess Cohan’s oral representations more carefully, and without the benefit of a preemptive dismissal of anything that could not be documented on a piece of paper.6
The Cohan rule is still with us, although Congress amended the Internal Revenue Code in 1962 to require written substantiation for business travel and various other commonly exaggerated business expenses.7 And even though it is still not a good idea to claim deductions for which there is no written documentation, Internal Revenue Service agents are nevertheless instructed to recognize that taxpayer oral representations and recollections are evidence and cannot be dismissed out if hand. Written evidence is still better than oral evidence, at least for tax purposes, and the oral evidence needs to be “credible.” But oral evidence—oral testimony—remains a permissible, even if risky, basis for a tax deduction.8 Despite its subsequent narrowing, the tax rule that bears Cohan’s name reminds us, and reminds the Internal Revenue Service, that what people say—testimony—can be evidence of the truth of what they are saying.
Treating testimony as evidence requires trusting what the testifier has said.9 Most obviously, this trust may come from a belief—one itself based on evidence—that the testifier is someone who can be trusted. Perhaps our trust in the truth of what is said—our epistemic trust—comes from our knowledge that what the testifier has said on previous occasions has turned out to be true. Perhaps it comes from knowing that the testifier is the kind of person who does not say things without good reason for believing them true. And perhaps the testifier’s position or profession or other source of expertise creates a reason for trust. There may be other reasons as well, but the point is that what someone says, as the Internal Revenue Service still acknowledges, even if grudgingly, is evidence for the truth of what they are saying, at least when they are asserting the kind of thing that there can be evidence for.10
The previous paragraph passed quickly over the possibility that the testifier is lying, a topic that will occupy all of Chapter 7, and the possibility that the testifier is honestly mistaken, which we address in Chapter 8. But even more generally, the extent to which testimony can be evidence for a proposition asserted by someone presupposes testifiers who are sincere and competent—that they know what they are talking about. As we will see, these constraints of honesty (or sincerity) and competence (or accuracy, or reliability) are substantial, creating many of the problems with treating testimony as evidence. But as an introduction to the idea of testimony as evidence, we can recognize that testimony is everywhere, that it can be evidence, and that it is perhaps the most predominant form of evidence that exists. Even what little remains of George M. Cohan’s dispute with the Internal Revenue Service is a reminder that what people say, even without anything in writing, and even without anything more tangible by way of corroboration, can be evidence for what they are saying.
In the midst of the trial of the impeachment of former President Donald Trump in February 2021, the House impeachment managers proposed to call Representative Jaime Herrera Beutler of the state of Washington as a witness. And although Representative Herrera Beutler’s proposed oral testimony was eventually replaced with a written statement, both the proposed oral testimony and the actual written testimony raise an important issue. Representative Herrera Beutler’s testimony was not about something she had seen or otherwise experienced firsthand. She testified that House Republican leader Kevin McCarthy had told her and others about an angry and expletive-laden phone call between McCarthy and Trump in which Trump had made it clear that he was not going to call off the rioters who at the time were storming the US Capitol.11
The House impeachment managers originally sought to use Herrera Beutler’s oral testimony as evidence for what McCarthy had told to her, and to use what Herrera Beutler said that McCarthy had said as evidence of Trump’s unwillingness to take actions to quell the riot. But if this was to be evidence of Trump’s behavior, it required that the members of the Senate believe what Herrera Beutler had said that McCarthy had said, and if the senators believed Herrera Beutler’s testimony about what McCarthy had said, they had to believe, further, McCarthy’s account of what Trump had said. In other words, this was hearsay. Twice.
Consider also a June 14, 2020, story in the New York Times reporting on a recently released book about first lady Melania Trump.12 The book, by Washington Post reporter Mary Jordan, was described in the Times article as follows:
And she [Melania Trump] has also constructed an image of herself that is not always supported by fact. For instance, Ms. Jordan raises questions about Mrs. Trump’s claims of being fluent in a number of languages.
“Photographers and others who have worked with her over the years—including native speakers of Italian, French, and German—told me that they never heard her use more than a few words of those language,” Ms. Jordan writes.
Now suppose we are interested in the question whether Melania Trump speaks French. And then think about how we might consider the evidence in the Times article. If the proposition is that Melania Trump does not speak French, that proposition is supported by the article, which offers itself as evidence for the conclusion that Melania Trump does not speak French. But the reporter who wrote the Times article, Maggie Haberman, does not have firsthand knowledge of whether Ms. Trump speaks French. The article merely reports on what photographers and others told Mary Jordan, whose book is described in the Times article. And Ms. Jordan also does not have firsthand knowledge of whether Melania Trump speaks French. Ms. Jordan relies on what others have told her. The photographers may know from their own interactions with Ms. Trump that she does not speak French, but the readers of the article must rely on an article by a Times reporter, who relies on a book by Mary Jordan, who relies on what the photographers and others have told her. So we have a chain that goes from the photographers to the book author to the Times reporter to the reader who might be interested in whether Melania Trump speaks French. This is triple hearsay. And it is risky business.
The law considers hearsay risky for reasons aptly illustrated by the chain that started with the photographers and ended up with the reader of the Times article. At every step in the chain—photographer to book author to newspaper reporter to reader—there is a possibility of mistake. And the reader, the one who wants to know whether Melania Trump speaks French, has no way of determining whether there has been a mistake at any of the links, except to rely on the possibly mistaken next link. The reader cannot test what the photographers say except to trust the book author’s trusting of the photographers, and the reader has no way of testing the book author except to trust the Times reporter’s trusting of the book author. And the reader cannot even test the Times reporter.
This is why the law has traditionally distrusted hearsay. When a witness in court testifies about what someone else said, and when the jury is asked to accept the truth of what that someone who is not in court said, the normal safeguards of the oath, cross-examination, and observation of the demeanor of the witness are absent. Why, in the extended example above, should we trust the photographers? Why should we trust what the book author said the photographers said? Why should we trust what the reporter said the book author said? Similarly, the law distrusts the statements of those who are not in court, under oath, observed, and subject to cross-examination.
In recent years American law has attached so many exceptions to the rule against hearsay that there is not much of it left.13 Depending on how one counts, the Federal Rules of Evidence contain at least thirty-one exceptions to the prohibition on hearsay evidence, and then the Rules add a catchall exception just in case there are forms of reliable hearsay that are not covered by the thirty-one listed exceptions.14 Going even further, common-law jurisdictions such as Canada and the United Kingdom have pretty much eliminated the exclusion of hearsay, as do many American judges, informally, when there is no jury.15
The continuing weakening of the bar on hearsay evidence is explained by the fact that it would be hard to imagine a world without hearsay, or knowledge without hearsay. Perhaps few people care whether Melania Trump speaks French.16 But many more people care what Representative Herrera Beutler said Representative McCarthy had said former President Trump had said. So much of what we know is based on hearsay, starting with the dates of our birth, that it is impossible to imagine a world in which we did not use hearsay as evidence. Even the quickest scan of a daily newspaper reveals that most of what we learn from so-called news is actually news about what other people have said, and most of why we care about what the people in the news have said is because those people are making factual assertions that we are implicitly being asked to credit. To return to the mob invasion of the Capitol building on January 6, 2021, much of what was reported in the news about that event consisted of various allegedly first-person accounts by those who were there, even though those who provided the accounts typically provided them to journalists, who in turn provided them to the public. Even if we trust the journalists to report accurately on what the participants and observers told them, as we generally do, we are still expected to form our beliefs about what happened on the basis of accounts offered by the people who were there, people whose testimony we cannot observe or cross-examine, and whose accounts we are asked to simply assume are both honest and accurate. We routinely do this, and we would not know much about the world and its goings-on without just this kind of reliance on hearsay. The law’s traditional skepticism about hearsay evidence is good reason to think twice about each step in the chain when we evaluate a hearsay statement, but it is important to remember that hearsay evidence is still evidence. And it is usually pretty good evidence at that.
One final cautionary note on hearsay, and on testimony generally: The law says that a statement is hearsay if the out-of-court statement is offered for the truth of that statement. That is an important qualification, as the classic example from the legal literature, here modified slightly, makes clear.17 If my colleague says to me that the dean just told him that she was Joan of Arc, my colleague is not asking me to take the dean’s statement as true. Neither of us think that she is Joan of Arc, nor that there is even the remotest possibility that she is. Rather, my colleague is reporting the dean’s statement as evidence that she has become mentally unhinged, or as evidence that she was joking. But not as evidence that she is actually Joan of Arc. Only when one person says that some other person said something, and only when we are asked to take what the other person said as evidence for the truth of the proposition asserted, is the problem of hearsay presented.
This understanding of hearsay is relevant to the entire topic of testimony. People say all sort of things, and for all sorts of purposes. When a child comes out from behind a bush on Halloween and shouts “Boo!,” she is not testifying, and she is not asking me to believe anything. So too when the train conductor says “All aboard,” when the boss says “You’re fired,” when my friend says “I’m sorry,” or, to use another example of so-called performatives from J. L. Austin and speech act theory, some notable personage breaks a bottle of champagne on the bow of a ship and says “I christen this ship H.M.S. Queen Elizabeth.”18 Only when someone asserts something that can be true or false do we have testimony, and only then do we confront the myriad questions surrounding the issue of whether we can take that assertion—the testimony—as evidence for what it is that the asserter was asserting.
The problem of hearsay is exacerbated, but also perhaps alleviated, when we are considering whether gossip, rumor, reputation, and other forms collective and unattributed testimony might be considered as evidence. From one perspective, the problems surrounding hearsay seem even worse when the original assertion is not from an individual and identifiable source. When we say that someone has a bad reputation for honesty, or a good reputation for reliability, we are relying on an aggregation of hearsay statements, although we rarely recognize it as such. But hearsay it is, usually thorough multiple iterations of hearsay, and without knowing the source of the original assertion. Whatever the risks of hearsay, they are compounded when we not only have no opportunity to examine the source of the original statement—the declarant, in lawspeak—but when we do not even know who the source or sources are. When we are asked to rely on—to treat as evidence—what some unidentified accumulation of people happen to believe, the justification for believing rumor, gossip, and reputation seems especially thin. And even if we do not go as far as insisting that such sources of knowledge should not count as evidence at all, it seems sensible at the very least to discount heavily the evidence that comes from such unattributed and attenuated sources.
But this is too quick. Perhaps rumor, gossip, and the social mechanisms that produce someone’s or something’s reputation should be understood not as heightening but as lessening the problems with hearsay. One reason for thinking that rumor, gossip, and reputation might alleviate the hearsay problem, therefore, is that rumor, gossip, and reputation are the collective products of a mechanism that might—and only might—incline toward reliability. Some years ago James Surowiecki, a writer for The New Yorker, wrote a popular book entitled The Wisdom of Crowds.19 When groups are diverse and decentralized and when the members are independent of each other, Surowiecki argued, using a number of engaging examples, the groups tend to make better decisions than individuals, better even than highly trained experts making decisions in isolation.
As several critics pointed out, other research casts doubt on Surowiecki’s conclusions.20 Moreover, he may have built so much into his conditions of diversity, decentralization, and independence, especially the last, that the conclusions he reached about wise crowds may have been highly unrepresentative of the behavior of real crowds making real decisions. What the psychologist Irving Janis influentially labeled “groupthink” back in the 1970s is often a large part of how people come to their factual conclusions. And just as collective wisdom may sometimes incline toward truth, perhaps the phenomenon of following the crowd, as people are often prone to do, inclines in just the opposite direction.
The possible wisdom of crowds has in recent years become the subject of more serious academic research, often under the label “collective intelligence”—which is the title of a new academic journal, although its first issue has yet to appear as of this writing. And the Massachusetts Institute of Technology now has the MIT Center for Collective Intelligence. But whether in its popularized or more seriously academic versions, the same issues present themselves.21 Modern information technology and related advances in artificial intelligence and so-called big data undoubtedly increase the potential for collective intelligence, but the same technology can also be the instrument of collective stupidity. The key question is the extent to which good ideas or true assertions about facts will be more likely to emerge from this or that collective process, or whether instead the same processes that increase the potential for the collective production of good ideas and accurate factual assertions will also increase the potential for the collective production and distribution of bad ideas and erroneous factual assertions..
This debate between the wisdom of crowds crowd, on one side, and the groupthink group, on the other, raises the question whether collective wisdom can be evidence, and whether it is ordinarily better than the noncollective wisdom that is the source of much of our testimonial evidence. If and when the collective wisdom has substantial evidentiary value, it is because of the ability of people, at times, to learn from each other, and, at times, to correct the mistakes that others make. But although some “crowds” are actually diverse, and although many crowds are decentralized, Surowiecki’s independence condition limits the value of his conclusions. Most forms of collective judgment are produced by the nonindependent groups that we might, less flatteringly but more accurately, call “mobs.” And somewhat less pejorative than “mob” is simply the idea of conventional wisdom, which at times devolves into mere rumor. But whatever label we use, there is a serious question about whether the kinds of things that “everyone knows” can count as evidence for whatever it is that everyone knows, or thinks they know.
It is important to distinguish evidence that “everyone knows” something from the fact of everyone knowing something being evidence for the something that everyone knows. Consider Donald Trump’s frequent assertion, made most notoriously to the crowd of people assembled in Washington, DC, on January 6, 2021—significant numbers of whom became the mob that invaded the Capitol building shortly thereafter. Addressing the crowd on the issue of the integrity of the election, the president said that “everyone knows” that he won the election by a landslide and that “everyone knows” that this obvious outcome was not made official because the election was “stolen.”22 This statement contained (at least) two falsehoods, which must be distinguished. One was the question whether President Trump in fact won the election by a landslide. He did not, so his claim that he had was false. But the other is whether this is something that “everyone knows.” That was false too. Sometimes, however, it is true that everyone thinks they know something that is in fact false. Even acknowledging the empirical claim of “everyone” as hyperbole, it is still true that everyone in 1491 knew, or at least thought they knew, that the earth was flat. And to a first approximation, everyone knows now that Marie Antoinette said “Let them eat cake,” that Napoleon was short, and that eating shortly before swimming produces muscle cramps. As claims about what it is that everyone knows, or thinks they know, these claims are, roughly, true. As claims about the truth of what everyone knows, or thinks they know, the claims are false.
Once we disaggregate claims that everyone knows something from claims about the truth of what everyone knows, we can focus more carefully on the latter—on the claim that everyone knowing something—or, more plausibly, the consensus knowing something—is evidence for the truth of what it is that everyone knows. The “wisdom of crowds” claim, or the collective-intelligence claim, oversimplified, is that this is so. But the various social, psychological, and political forces leading people to want to align themselves with others—independent of the basis for the alignment—gives pause. In high school some of us wanted to do what the cool kids were doing just because we wanted to be part of that crowd, even if what the crowd was doing was often pretty stupid. And to the extent that such a dynamic exists in other social domains, the ability to rely on the collective wisdom of that domain as evidence for the truth of what the domain believes diminishes commensurately.
The “wisdom of crowds” claim bears an interesting affinity with the “marketplace of ideas” slogan that has been a central part of free speech discourse, policy, and law at least since John Milton asked, rhetorically, “Who ever knew Truth put to the worse in a free and open encounter?”23 Milton’s basic idea, echoed in John Stuart Mill’s On Liberty, in Oliver Wendell Holmes’s claim that the “best test of truth is the power of the thought to get itself accepted in the competition of the market,” and in civil libertarian rhetoric for at least the last hundred years, is that the process of collective discussion will incline toward truth—that collectivity produces collective intelligence more than it produces collective folly.24 Clothed in modern terminology, the claim, as the late Anthony Lewis engagingly put it, is that the collective discourse fostered by a free speech regime is a “search engine for truth.”25
As a question about evidence, the “marketplace of ideas” slogan raises two different evidentiary claims. One is whether the truth of some proposition is evidence of the likelihood that people will accept it. Obviously, people believe many true propositions. And they believe many false ones as well. And we know, and Madison Avenue knows, that the truth of some proposition is hardly the only determinant of the likelihood that some population will accept it, or even that some significant percentage of some population that will accept it. After all, roughly 30 percent of the population accepts the truth of the (false) predictive and (false) behavioral claims of astrology.26 Even assuming that truth matters to acceptance, as we hope it does, the identity, authority, and charisma of the person who is asserting something also matter. The same holds for the frequency with which something is said, the technological way in which it is said, the extent to which what is said fits with the prior beliefs of the audience, the way in which what is said satisfies some emotional or psychological needs of the hearer, and much more. Indeed, the long-standing and seemingly increasing acceptance of so-called conspiracy theories provides strong support for the conclusion that people’s acceptance of some proposition depends on far more than that proposition’s truth.27 If the question is whether the truth of a proposition is evidence of its likely acceptance—perhaps of less importance in this book—then there is much reason to believe, the faith of Milton and his successors notwithstanding, that although truth matters as evidence, it may matter less than we hope and less than we often suppose. The truth of a proposition, which we know as a result of evidence, is itself not very good evidence, or at least not as good evidence as has long been assumed, of whether it will be accepted by those to whom it is communicated.28
More important to us here, however, is the reverse—not the evidentiary value of truth in determining acceptance, but the evidentiary value of acceptance in determining truth. This is the “everyone knows” or “wisdom of crowds” or “collective intelligence” claim—that the fact of acceptance is at least some evidence of the correctness of what it is that is accepted.29
On this question, the “wisdom of crowds” claim is one side of the story. But the other side of the story is as skeptical of collective wisdom as evidence as the marketplace of ideas is celebratory. This other side of the story, traceable to the worry about the spread of false rumors during the Second World War, and of a piece with the game of “telephone” that many of us played as children, views collective discussion as the hotbed of misinformation, disinformation, and the basket of human psychological needs that are often in conflict with truth. As falsehoods that are spread widely and quickly by the internet and social media make this hotbed ever hotter, the value of widespread belief as a reliable indicator of likely truth becomes ever smaller.
All the same, evidence that might not be very good evidence is often, although not necessarily, evidence with some value, depending on what we want to with it, or what decisions we want to make as a result of it. The law has long believed, for example, that a person’s reputation for honesty (or lack thereof) might be relevant to assessing their credibility and thus the reliability of their formal testimony.30 The law may be wrong in this, but that long-standing belief suggests that reputation—what everyone knows—may in some contexts have some evidentiary value. And however skeptical we might be about some uses of reputation, including the law’s, most of us are comfortable in using reputation evidence when we select our contractors, our restaurants, and perhaps even our babysitters and investment advisors. To the extent that we are relying on collectively produced reputation in this way, we are relying on testimony, even though the original sources of that testimony remain hidden from us. When we rely on reputation, which we often more pejoratively call rumor and gossip, we are relying on also-anonymous assessors of the reliability of what other have told them. This is all the more reason to be especially skeptical of what we can and should understand as indirect testimony. But it might not be reason to discard it entirely.